Ontology extension is based on the reuse of the classes existing in other ontologies to properly characterize concepts and reuse existing frameworks and axioms. The reuse of ontology terms creates links between data, making the ontology more valuable. 28
For example, ChEBI contains more than 3 million axioms. Thus, only relevant subsets of ontologies were reused. After searching for classes within the working ontology using the owlready2 Python package, 29 missing chemical compounds and possible reaction types are searched for in other ontologies as listed in Table 1 . To accomplish this, a nested dictionary is used, containing all IRIs of terms alongside their corresponding labels, prefLabels, altLabels, and the names used within the ontology. Applying a class information extraction process that incorporates functions from the owlready2 package, the dictionary was generated for 22 ontologies relevant to the domain of catalysis research. 24 Once the dictionary is loaded into the Python environment, it is searched for classes missing in the working ontology. If one of the labels, prefLabels, altLabels, or names matches the searched entity, the corresponding IRI is added to the dictionary along with the matching value. The IRIs of found terms that are still missing in the working ontology are stored in automatically created text files. The names of the text files include the acronym of the source ontology for further reference.
This ensures that the module retains all the same logical entailments in the full ontology, providing consistency in the ontology subset. The chosen SMLE approach is the BOTTOM module, which contains the terms in the seed, their corresponding superclasses, and the interrelations between them. As the name implies, the class hierarchy is built from the bottom up, gathering the superclasses of the selected class. Thus, for each ontology, a separate subset of relevant classes is created in rdf/xml (owl) format.
The second task for which ROBOT is used in this work is to merge the created subsets of classes and the main ontology into a single ontology with a single .owl-file. Thus, the merging process is used to update the working ontology within existing terms of other ontologies.
Because some of the merged ontologies are aligned with different top-level ontologies, terms that theoretically share the same definition are located at different positions within the class hierarchy. For example, the OBO is the top-level ontology of ChEBI, while the AFO is aligned with the BFO. Both ontologies have, for example, the term “atom”, but at different positions in the class hierarchy. Another factor why the same terms are represented differently is related to the granularity problem of ontologies. This issue arises because ontologies often adopt different levels of details when representing identical knowledge to support different applications. 30
Since all of the utilized ontologies are connected to the domain of catalysis 24 and chemistry, 31 terms with identical designation are assumed equivalent. The equivalence of classes indicates that respective classes share all their instances, and the descriptions of both classes are interlinked. However, the use of the equivalence relation does not imply class equality. Both relations are defined differently in OWL. Equality is denoted by “owl:sameAs”, while the equivalence is represented by owl:equivalentClass. Class equality can only be defined by the description language OWL-Full, and owlready2 supports only equivalence. 29
To identify terms with the same designation that originate from different ontologies and consequently have different IRIs, the mappings created in previous work 24 are used in the processing. These mappings represent all terms shared by two ontologies according to the same IRIs or the same set of labels, prefLabels, names, or altLabels. After merging ontologies to reuse existing terms, the process of creating new classes and subsequently populating the working ontology with new instances is initialized. First, a new instance of a publication is generated as an instance of the “publication” superclass. The DOI and title of the publication retrieved at the beginning of the process are added to the publication instance as datatypes using the “has doi” and “has title” datatype properties, respectively. Extracted chemical compounds that do not exist in the working ontology after merging are then created as new classes within the working ontology, utilizing the context information of the new compounds. Chemical compounds that can be further broken down into compounds and atoms, such as “Al 2 O 3 ” or “titanium dioxide”, or those that are recognized as compounds using pubchempy are created as subclasses of the “molecule” class.
Support material entities, which represent a combination of two or more carrier compounds, like “TiO 2 –SiO 2 ”, or materials such as “MCM-41” are created as subclasses of the “support material” class. Each newly created class and instance are automatically assigned a generated name linked to the number of the processed publications in the working ontology.
Entities from the “Reactant”, “Product” and “Catalyst” categories that represent specific types of chemical entities, such as “light olefin” and “vapour phase propene”, are created as instances of the corresponding chemical compound. Extracted and preprocessed catalyst entities are created as instances of the “chemical substance” class.
Chemical entities, which represent catalysts in the form of
“ <catalytic compound>/<support compound> ” or
“ <catalytic compound>@<support compound> ”
are labelled in the ontology
“ <catalytic compound> supported on <support compound> ”
and linked with their chemical compounds based on their roles in the entity using the “catalytic component of” and “support component of” object properties. A schematic example of interconnections within the ontology is shown in Fig. 3 .
Examples of created entities and assigned relations. Entities within dashed boxes represent instances, while continuous bordered boxes represent classes. |
Table 4 in the appendix lists the object properties and their inverse properties that need to be defined within the working ontology in order to assign the relations between the newly created entities.
The creation of the classes corresponding to the catalyst types is based on the creation of subclasses of the term “catalyst role”, which already exists in the AFO ontology. Roles in ontologies are used to reduce the amount of object properties and thus to speed up reasoning. The corresponding roles of terms are provided as classes in the ontology and terms are linked to them via the “has role” object property. The hierarchical structuring of the catalyst roles is based on the content of the classes extracted from the entities. For instance, within the text corpus, the extracted class of an entity after preprocessing might be a “dispersed catalyst role”, while the catalyst type corresponding to another entity is an “atomically dispersed catalyst role”. Since the second class is identical to the first but with an additional word, it is considered a subclass of the first class. In case no entity from the text corpus has a “dispersed catalyst role” as an extracted class, then an “atomically dispersed catalyst role” is created as a subclass of the “catalyst role”.
Chemical reactions are created as subclasses of the previously extracted reaction “heads”. If there is no corresponding reaction found in other ontologies, a new class is created as a subclass of “chemical reaction (molecular)”, which is also a class within the AFO. For each class created after the merging that corresponds to an extracted entity or a chemical compound, an instance is created with an automatically generated name.
The label of the instance is the same as the label of its corresponding class. The same procedure is applied to the newly created classes. All classes and instances, once created, can be reused for ontology extension with the respective publications. The newly created classes of chemical compounds are linked to their corresponding components via “has part” relations at the instance level.
Created instances are linked to their roles according to the categories and the context using the “has role” relation. The used roles include the “support role”, “reactant role”, “product role”, and “catalyst role”, and all are created as subclasses of the “catalyst role”.
Finally, all created and used instances that are mentioned in the processed publication are linked to the instance of the publication through the “mentioned in” object property. Entities labelled “Characterization” and “Treatment” are added to annotations of publications as comment.
The following competency questions were implemented as SPARQL queries and can thus be easily retrieved from the knowledge graph resulting from the extension of the ontology. The corresponding SPARQL queries are numbered and exemplary input and output of the queries are listed in Table 5 in the appendix and an exemplary SPARQL query is listed in Table 6 in the appendix:
• Give me a list of reactions (1), reactants, support materials, catalysts, and products mentioned in one specific publication, which is a part of the knowledge graph, in one list (2) or separately,
• Retrieve the abstracts from publications in the ontology (3),
• Give me a list of DOIs of publications from the working ontology, which mention the same reactions (4) or the specific reaction (5) or catalyst (6),
• Give me a list of reactions, reactants, support materials, catalysts, and products mentioned in all publications of the knowledge graph (7),
• I need a list of all possible synonyms for the extracted reactants (8), support materials (9), catalysts (10), and products (11) in the form of chemical entities,
• I need possible catalysts where the support material from this paper can be used (12).
The retrieved entities can be used to query Scopus for new publications with similar context. Using the pybliometrics Python package, the search is performed, leading to a query, which has the same structure as a query that works in the Scopus advanced search. With the chosen query type ‘TITLE-ABS-KEY()’ (as depicted exemplarily in Fig. 4 ), the search is performed within the titles, abstracts, and keywords of the publication.
Two types of query formulations for the advanced search for further publications in Scopus, executed by the Python API. |
Since there are multiple ways to name a specific chemical compound, to avoid a large number of possible queries and at the same time allow diversity in the naming of chemical compounds, a trade name or common name and a formula listed in the class annotations of a chemical compound are used for queries' formulation. Moreover to exclude mismatches, the publication will be skipped if during text mining no reaction was found in the text. After a query is executed, its results are downloaded and cached to speed up the subsequent analysis.
After the results are concatenated into one table, duplicates are removed from it. As Scopus contains records of articles published since 1970, an option to filter the results by publication date is integrated into the process, to allow for the inclusion of primarily newer publications into the knowledge graph. Utilization of the pandas Python module 33 allows the resulting DataFrames to be stored as sheets in an Excel file.
Moreover, a set of 28 publications on methanation processes (dataset 2) was used to evaluate how well the created tool works on the different types of catalysed reactions. Hereby the focus was laid on the heterogeneously catalysed conversion of carbon monoxide and carbon dioxide to methane via hydrogenation, which is important for the production of synthetic natural gas. In particular, the valorization of CO 2 together with renewable hydrogen might be considered an integral sustainable path towards the production of renewable gaseous fuels. 36 For that, an extension of an alternate ontology setup similar to the first dataset was performed.
The dataset for training of the model was complemented with 151 sentences manually labelled in label-studio 37 from 18 abstracts of papers to the topic of hydroformylation in the liquid and gas phase. Checkpoints from the model trained by the authors of CatalysisIE and the model trained on the complemented dataset were compared with each other.
To evaluate the difference in the prediction of the checkpoints, ten manually labelled abstracts from papers to the same topic were compared to predictions of both models. Since it is important to gain as many correct distinct predictions from the text as possible to be able to describe the content of the publication using extracted entities, the recall R of the model was evaluated with the number of true positives TP and false negatives FN using eqn (1) . To obtain the true positives and false negatives, the amount of distinct entities was counted and compared with the number of distinct entities from the prediction after qualitative manual labelling of the texts. This comparison for each extracted abstract from dataset 1 is shown in Table 9 in the appendix.
Besides recall, the precision Pr was selected for evaluation of multi-label classification. Because class imbalances are present in the dataset, the precision was calculated using eqn (2) with the number of positives P i instead of true positives TP and the number of used labels N . Furthermore, the standard deviation σ of the precision was selected as a metric and calculated using eqn (3) . The sum of true positives corresponds to the number of correctly predicted instances. Precision and its standard deviation were calculated for the six categories for each of the abstracts.
(1) |
(2) |
(3) |
Extraction of sequences was treated as a binary classification problem, where the sum of TP is correctly extracted from distinct entities and is independent from the assigned label. The sum of true positives and false negatives is the total number of distinct manually labelled entities in the text. The metrics were calculated for the ten manually labelled abstracts from papers to the same topic and compared to predictions of both checkpoints I and II. Here, checkpoint I addresses the complemented model, while checkpoint II addresses the pre-trained checkpoint, provided by the developers of CatalysisIE. The resulting metrics are listed in Table 2 . Deviations in the metrics of the fourth publication may be due to formatting errors in the retrieved abstract, causing extracted tokens to end with citation numbers ( e.g. , “catalysts9”), thus not being counted as found entities.
TP + FN | CP | TP | R | Pr | σ | |
---|---|---|---|---|---|---|
1 | 12 | I | 11 | 91.7 | 83.3 | 40.8 |
II | 10 | 83.3 | 66.7 | 51.6 | ||
2 | 15 | I | 14 | 93.3 | 93.3 | 14.9 |
II | 12 | 80.0 | 80.0 | 44.7 | ||
3 | 15 | I | 13 | 86.7 | 75.6 | 43.3 |
II | 12 | 70.0 | 73.3 | 43.5 | ||
4 | 16 | I | 6 | 37.5 | 35.0 | 23.8 |
II | 6 | 37.5 | 38.7 | 30.7 | ||
5 | 15 | I | 10 | 66.7 | 64.6 | 9.9 |
II | 7 | 66.0 | 61.8 | 26.7 | ||
6 | 15 | I | 13 | 86.7 | 58.3 | 49.2 |
II | 14 | 93.3 | 66.7 | 51.6 | ||
7 | 37 | I | 33 | 89.2 | 88.2 | 21.7 |
II | 24 | 64.9 | 62.7 | 16.3 | ||
8 | 29 | I | 25 | 86.2 | 68.2 | 10.5 |
II | 24 | 82.8 | 64.7 | 11.4 | ||
9 | 25 | I | 12 | 48.0 | 63.6 | 26.0 |
II | 13 | 52.0 | 68.6 | 25.1 | ||
10 | 6 | I | 6 | 100.0 | 100.0 | 0.0 |
II | 6 | 100.0 | 100.0 | 0.0 |
The entities labelled “Characterization” were predicted least accurately. Additionally, there were no “Treatment” labels in the evaluation dataset. Overall, the model trained on the expanded dataset (CP I) was better at predicting entities labelled “Catalyst”. The average recall of the newly trained model for ten abstracts is equal to 86.67% with a standard deviation of 20.85% and shows a high average precision of 71.90%. In comparison, the recall of the old model (CP II) is 80.00% with a standard deviation of 19.37% and an average precision of 66.67%. In both cases only in one text, precision and recall fall under 50%. Furthermore, for the ten publications shown in Table 2 , in the cases where CP I achieved higher precision, σ was lower. This indicates that the dispersion across the different classes in relation to Pr has decreased and therefore the model makes more stable predictions across the classes.
To investigate the performance of the extended model further, ten abstracts from dataset 2 are labelled manually and classified with CP I and CP II to evaluate the metrics as in Table 2 . The resulting metrics are presented in more detail in Table 7 . An average recall of 82.81% with a standard deviation of 22.39% and an average precision of 71.46% was achieved for CP I. Furthermore, an average recall of 79.47% with a standard deviation of 19.92% and an average precision of 73.20% was achieved for CP II. Thus, the extended model can also be applied on dataset 2.
Title recognition by 19 out of 23 processed PDFs from dataset 1 was successful and 26 from 28 publications from dataset 2 could be recognized correctly. Publications of “Royal Society of Chemistry” could not be correctly recognized because the layout of the publications is not integrated in the workflow of the used pdfdataextractor package.
The AFO was chosen as the initial ontology, because of its linkage to the chemical domain and well-defined structure in the class hierarchy. Table 8 lists the terms and textual definitions assigned as equivalent in ontologies for both datasets, which exist in the AFO and are merged into the working ontology from ChEBI.
Chemicals which could not be found in PubChem or in ChEBI are created as instances of the class “chemical substance”. For dataset 1, the ontology is extended with 53 instances of “chemical substance”. Dataset 2 results in 55 instances of “chemical substance” that were also created automatically. Each of the generated instances representing extracted entities and their chemical components is provided with a connection to the publication in which it is mentioned and linked to the corresponding roles as shown in an excerpt of the resulting ontology in Fig. 5 . The reactions that are mentioned within the publication are listed, including the respective participants of the reactions within the knowledge base (upper area of the figure). The individual “cobalt atom”, for example, is connected with the individual “Co-containing catalyst” via the object property “catalytic component of” (right area of the figure), thus indicating the suitable catalytic component of the concept extracted from text. Furthermore, the role of a “bimetallic catalyst role” is asserted to the three individuals on the bottom right of Fig. 5 . The class “bimetallic catalyst role” is created as a subclass of the “catalyst role”, which in turn also has an individual that is connected to other substances via the “has role” object property (bottom left of the figure).
Excerpt from the created ontology for dataset 1 created with Protégé. Boxes marked with yellow circles represent classes and those with purple rhombi are instances. Arrows denote the relationships between them, color-coded as listed in the legend on the right. Small boxes with a plus (+) inside indicate that not all relations of the entity are shown in the figure. |
The knowledge graph with publications from dataset 1 was extended by 48 classes from the other ontologies, including their superclasses and interrelations. In total, 331 new classes, 9 new object properties, 2 new data properties, and 155 new individuals were added to the working ontology. From the new classes, 288 were merged from other ontologies, while none of the new individuals were merged from other ontologies, as expected. The new object and data properties were merged from other ontologies.
In the knowledge graph with dataset 2, 39 classes from other ontologies were imported from other ontologies together with their respective superclasses and interrelations. With this, 222 new classes, 4 new object properties, 2 new data properties, and 130 new individuals were added to the working ontology. Here, 198 from the 222 new classes were merged from other ontologies, while also none of the new individuals were merged from other ontologies. The new object and data properties were merged from the other ontologies listed in Table 1 and counted without the ones already presented in Table 4 in the appendix. The explained ontology metrics are listed in Table 3 .
Metric | Initial ontology | Extended ontology dataset 1 | Extended ontology dataset 2 |
---|---|---|---|
Classes | 3116 | 3447 | 3338 |
Instances | 47 | 203 | 178 |
Logical axioms | 5755 | 6936 | 6596 |
SubClassOf | 4823 | 5372 | 5174 |
Equivalent classes | 178 | 188 | 185 |
Fig. 6 shows the individual “0.5% Co–0.5% Rh supported on Al 2 O 3 ” in an excerpt of Protégé after reasoning with HermiT. 38 The implicit knowledge is highlighted in yellow, showing an increased semantic expressiveness for the individual describing the catalyst complex. Thus, the individual now also can be found when searching the knowledge graph, e.g. , for catalysts that contain cobalt.
Excerpt from Protégé with inferred relations after reasoning for the individual “0.5% Co–0.5% Rh supported on Al O ”. Knowledge inferred by the reasoner is highlighted in yellow, showing increased semantic expressiveness of the individual. |
Most of the terms in both knowledge graphs originate from the ChEBI ontology and identify chemical compounds and atoms. But also, the classes for such terms as “hydrogenation”, “hydroformylation”, and “acylation” are reused from the RXNO ontology. In the current process, entities representing some chemical groups, such as “phenolic substances” or “phenolic species”, can be recognized with the text mining module, but the extension of the ontology with them is not implemented. This includes, for example, entities such as “phenolic substances”, “phenolic species”, and “alkyl species” which are usually classified as products or reactants in the text. Such entities cannot be queried in PubChem, and in ChEBI, the presumed superclasses are placed in different positions. All queries are formulated within the functions in the module “queries” provided in the GitHub repository of this work 39 and can be executed by the Jupiter notebook “user_queries.ipynb” contained in the repository. It contains descriptions of code cells, which execute specific queries, which can answer competency questions formulated in Methods. In addition, some examples of executed functions are also provided in the notebook.
Listings of the resulting publications are found in the provided GitHub repository of this work 39 in the “output” directory.
To further rate the quality of the query result, a random sample in size of 50 publications from the resulting filtered list of 731 publications similar to those of dataset 1 was selected for the evaluation of the queried content. The list of chosen publications for evaluation is provided in the appendix in Table 9 .
These publications are rated as similar to the publications in the knowledge graph (of dataset 1) if the following requirements are fulfilled, based on the evaluation of their abstracts and titles:
• Heterogeneous or homogeneous catalysis or catalysts are mentioned.
• Hydroformylation or hydrogenation is mentioned.
• Rh-, Co-, and Ni-based catalysts with silica, zeolite or aluminium oxide as a support material are mentioned.
According to these restrictions, 34 out of 50 publications of the sample provided are rated as similar to the content of the publications within the knowledge graphs, which is equal to 68% accuracy.
The quality of matched articles to the content of the knowledge graphs was evaluated with dataset 1, dealing with hydroformylation reactions. A random sample of 50 publications was investigated in more detail and abstracts, keywords, and titles were screened for the mention of catalysts, hydrogenation or hydroformylation, and whether Rh-, Co-, and Ni-based catalysts with silica, zeolite or aluminium oxide as a support material were mentioned. With these quite strict criteria and with the lack of case sensitivity in the Scopus API, 68% accuracy was achieved for the random sample. This allows for more structured searches of relevant scientific literature in the domain of catalysis research, which is highly important, especially in this domain, as research is quite heterogeneous and the number of relevant publications in the field is quite high. However, a more thorough post-processing of these found publications needs to be conducted, e.g. by a post-processing that conducts a case-sensitive automated search of the respective keywords in the extracted abstracts from Scopus, to improve the accuracy of the output related publications.
Object property | Reverse object property | ||
---|---|---|---|
Name | Rdfs:Label | Name | Rdfs:Label |
supported_on | Supported on | support_material_of | Support material of |
support_component_of | Support component of | has_support_component | Has support component |
catalytic_component_of | Catalytic component of | has_catalytic_component | Has catalytic component |
mentioned_in | Mentioned in | Mentions | Mentions |
Query no. | Input parameters | Output |
---|---|---|
1 | Doi = r‘10.1021/acsami.0c21749.s001’ | [‘hydroformylation’] |
2 | list_type = ‘all’ | [‘hydroformylation’, ‘olefin’, ‘rhodium atom’, ‘Rh-based atomically dispersed catalyst’, ‘Rh supported on ZnO modified with Pi’, ‘zinc oxide’, ‘phosphate ion’, ‘aldehyde’, ‘linear aldehyde’] |
Doi = r‘10.1021/acsami.0c21749.s001’ | ||
3 | Doi = r‘10.1021/acsami.0c21749.s001’ | Abstract: in the study of heterogeneity of homogeneous processes, effective control of the microenvironment of active sites… |
4 | Doi = r‘10.1021/acsami.0c21749.s001’ | [[‘10.1021/acscatal.1c02014.s001’], [‘10.1021/acscatal.0c04684.s001’], [‘10.1021/acscatal.1c00705.s002’], [‘10.1021/acscatal.1c04359’],…] |
5 | Reac = “hydrogenation”, Doi = none | [[‘10.1021/acscatal.0c04684.s001’], [‘10.1021/acscatal.1c00705.s002’],…] |
6 | Cat = “RhCo”, Doi = none | [[‘10.1021/acscatal.0c04684.s001’], [‘10.1021/acscatal.1c00705.s002’], |
7 | list_type = ‘all’ | [‘styrene’, ‘cobalt atom’, ‘0.5% Co–0.5% Rh supported on Al O ’, ‘Co-containing catalyst’, ‘aluminum oxide’, ‘hydroformylation’,…] |
Doi = none | ||
8 | list_type = ‘reactant’ | [‘olefin’] |
Doi = r‘10.1021/acsami.0c21749.s001’ | ||
9 | list_type = ‘product’ | [‘Aldehyd’, ‘RCHO’, ‘aldehidos’, ‘aldehydes’, ‘Aldehyde’, ‘aldehydum’, ‘an aldehyde’, ‘RC( |
Doi = r‘10.1021/acsami.0c21749.s001’ | ||
10 | Doi = r‘10.1021/acsami.0c21749.s001’ | [[‘45Rh’, ‘rhodium’, ‘Rh’, ‘rodio’, ‘Rh(111)’, ‘rhodium atom’]], [[‘Rh on ZnO modified with Pi’, ‘Rh supported on ZnO modified with Pi’]] |
11 | Doi = r‘10.1021/acsami.0c21749.s001’ | [[‘zinc oxide’, ‘oxyde de zinc’, ‘Zinkoxid’, ‘oxido de cinc’, ‘ZnO’], [‘phosphate ions’, ‘Pi’, ‘phosphate’, ‘phosphate ion’]] |
12 | Doi = r‘10.1021/acscatal.1c02014.s001’ | [[‘silicon dioxide’, ‘Rh supported on SiO catalyst’], [‘silicon dioxide’, ‘Rh P nanoparticle supported on SiO support material’], [‘silicon dioxide’, ‘Rh7Co1P4 supported on SiO ’],…] |
only_doi = false |
TP + FN | CP | TP | R | Pr | σ | |
---|---|---|---|---|---|---|
1 | 15 | I | 13 | 86.7 | 68.3 | 32.5 |
II | 12 | 80.0 | 61.7 | 36.1 | ||
2 | 19 | I | 15 | 78.9 | 82.7 | 28.9 |
II | 15 | 78.9 | 82.7 | 28.9 | ||
3 | 9 | I | 9 | 100.0 | 93.3 | 14.9 |
II | 9 | 100.0 | 93.3 | 14.9 | ||
4 | 11 | I | 6 | 54.5 | 66.7 | 28.9 |
II | 7 | 63.6 | 70.8 | 26.0 | ||
5 | 6 | I | 6 | 100.0 | 100.0 | 0.0 |
II | 6 | 100.0 | 100.0 | 0.0 | ||
6 | 14 | I | 8 | 57.1 | 43.8 | 51.5 |
II | 8 | 57.1 | 43.8 | 51.5 | ||
7 | 12 | I | 12 | 100.0 | 73.6 | 29.7 |
II | 11 | 91.7 | 72.6 | 42.6 | ||
8 | 12 | I | 7 | 58.3 | 69.3 | 41.3 |
II | 6 | 50.0 | 49.3 | 46.6 | ||
9 | 9 | I | 4 | 44.4 | 69.0 | 27.0 |
II | 5 | 55.6 | 73.8 | 25.1 | ||
10 | 15 | I | 15 | 100.0 | 92.6 | 13.5 |
II | 15 | 100.0 | 92.6 | 13.5 |
Term label | IRIs + Definitions | |
---|---|---|
In the AFO | In ChEBI | |
‘Chemical substance’ | http://purl.allotrope.org/ontologies/material#AFM_0001097 | http://purl.obolibrary.org/obo/CHEBI_33250 |
A chemical substance is a portion of material that is matter of constant composition best characterized by the entities (molecules, formula units, atoms) it is composed of [IUPAC] | A chemical entity constituting the smallest component of an element having the chemical properties of the element | |
‘Anion’ | http://purl.allotrope.org/ontologies/material#AFM_0000161 | http://purl.obolibrary.org/obo/CHEBI_22563 |
An anion (−) is an ion with more electrons than protons, giving it a net negative charge (since electrons are negatively charged and protons are positively charged) | A monoatomic or polyatomic species having one or more elementary charges of the electron | |
‘Ion’ | http://purl.allotrope.org/ontologies/material#AFM_0000077 | http://purl.obolibrary.org/obo/CHEBI_24870 |
An ion is an atom or molecule in which the total number of electrons is not equal to the total number of protons, giving the atom or molecule a net positive or negative electrical charge | A molecular entity having a net electric charge | |
‘Role’ | http://purl.obolibrary.org/obo/BFO_0000023 | http://purl.obolibrary.org/obo/CHEBI_50906 |
B is a role means: b is a realizable entity and b exists because there is some single bearer that is in some special physical, social, or institutional set of circumstances in which this bearer does not have to be and b is not such that, if it ceases to exist, then the physical make-up of the bearer is thereby changed [BFO] | A role is particular behavior which a material entity may exhibit | |
‘Cation’ | http://purl.allotrope.org/ontologies/material#AFM_0000189 | http://purl.obolibrary.org/obo/CHEBI_36916 |
A cation (+) is an ion with fewer electrons than protons, giving it a positive charge | A monoatomic or polyatomic species having one or more elementary charges of the proton | |
‘Group’ | http://purl.obolibrary.org/obo/BFO_0000023 | http://purl.obolibrary.org/obo/CHEBI_24433 |
A group is an aggregate of people | A defined linked collection of atoms or a single atom within a molecular entity | |
‘Atom’ | http://purl.allotrope.org/ontologies/material#AFM_0001028 | http://purl.obolibrary.org/obo/CHEBI_33250 |
An atom is a smallest particle still characterizing a chemical element. It consists of a nucleus of a positive charge carrying almost all its mass (more than 99.9%) and Z electrons determining its size | A chemical entity constituting the smallest component of an element having the chemical properties of the element | |
‘Chemical substance’ | http://purl.allotrope.org/ontologies/material#AFM_0001097 | http://purl.obolibrary.org/obo/CHEBI_59999 |
A chemical substance is a portion of material that is matter of constant composition best characterized by the entities (molecules, formula units, atoms) it is composed of | A chemical substance is a portion of matter of constant composition, composed of molecular entities of the same type or of different types |
Author contributions, conflicts of interest, acknowledgements.
† Electronic supplementary information (ESI) available: https://github.com/AleSteB/CatalysisIE_Knowledge_Graph_Generator. See DOI: |
Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.
Nature Communications volume 15 , Article number: 6367 ( 2024 ) Cite this article
10k Accesses
321 Altmetric
Metrics details
Male sex, early life chemical exposure and the brain aromatase enzyme have been implicated in autism spectrum disorder (ASD). In the Barwon Infant Study birth cohort ( n = 1074), higher prenatal maternal bisphenol A (BPA) levels are associated with higher ASD symptoms at age 2 and diagnosis at age 9 only in males with low aromatase genetic pathway activity scores. Higher prenatal BPA levels are predictive of higher cord blood methylation across the CYP19A1 brain promoter I.f region ( P = 0.009) and aromatase gene methylation mediates ( P = 0.01) the link between higher prenatal BPA and brain-derived neurotrophic factor methylation, with independent cohort replication. BPA suppressed aromatase expression in vitro and in vivo. Male mice exposed to mid-gestation BPA or with aromatase knockout have ASD-like behaviors with structural and functional brain changes. 10-hydroxy-2-decenoic acid (10HDA), an estrogenic fatty acid alleviated these features and reversed detrimental neurodevelopmental gene expression. Here we demonstrate that prenatal BPA exposure is associated with impaired brain aromatase function and ASD-related behaviors and brain abnormalities in males that may be reversible through postnatal 10HDA intervention.
Introduction.
Autism spectrum disorder (ASD or autism) is a clinically diagnosed neurodevelopmental condition in which an individual has impaired social communication and interaction, as well as restricted, repetitive behavior patterns 1 . The estimated prevalence of ASD is approximately 1–2% in Western countries 2 , with evidence that the incidence of ASD is increasing over time 3 . While increased incidence is partly attributable to greater awareness of ASD 4 , other factors including early life environment, genes and their interplay are important 5 . Strikingly, up to 80% of individuals diagnosed with ASD are male, suggesting sex-specific neurodevelopment underlies this condition 5 .
Brain aromatase, encoded by CYP19A1 and regulated via brain promoter I.f 6 , 7 , 8 converts neural androgens to neural estrogens 9 . During fetal development, aromatase expression within the brain is high in males 10 in the amygdala 11 , 12 . Notably, androgen disruption is implicated in the extreme male brain theory for ASD 13 , and postmortem analysis of male ASD adults show markedly reduced aromatase activity compared to age-matched controls. Furthermore, CYP19A1 aromatase expression was reduced by 38% in the postmortem male ASD prefrontal cortex 14 , as well as by 52% in neuronal cell lines derived from males with ASD 15 . Environmental factors, including exposure to endocrine-disrupting chemicals such as bisphenols, can disrupt brain aromatase function 16 , 17 , 18 .
Early life exposure to endocrine-disrupting chemicals, including bisphenols, has separately been proposed to contribute to the temporal increase in ASD prevalence 19 . Exposure to these manufactured chemicals is now widespread through their presence in plastics and epoxy linings in food and drink containers and other packaging products 20 . Although bisphenol A (BPA) has since been replaced by other bisphenols such as bisphenol S in BPA-free plastics, all bisphenols are endocrine-disrupting chemicals that can alter steroid signaling and metabolism 21 . Elevated maternal prenatal BPA levels are associated with child neurobehavioral issues 20 including ASD-related symptoms 22 , 23 , with many of these studies reporting sex-specific effects 20 , 22 , 23 , 24 . Furthermore, studies in rodents have found that prenatal BPA exposure is associated with gene dysregulation in the male hippocampus accompanied by neuronal and cognitive abnormalities in male but not female animals 20 , 23 , 24 . One potential explanation is that epigenetic programming by bisphenols increases aromatase gene methylation, leading to its reduced cellular expression 16 and a deficiency in aromatase-dependent estrogen signaling. If such is the case, it is possible that estrogen supplementation, such as with 10-hydroxy-2-decenoic acid (10HDA), a major lipid component of the royal jelly of honeybees, may be relevant as a nutritional intervention for ASD. Indeed, 10HDA is known to influence homeostasis through its intracellular effects on estrogen responsive elements that regulate downstream gene expression 25 , 26 , as well as its capacity to influence neurogenesis in vitro 27 .
Here, we have investigated whether higher prenatal BPA exposure leads to an elevated risk of ASD in males and explore aromatase as a potential underlying mechanism. We demonstrate in a preclinical (mouse) model that postnatal administration of 10HDA, an estrogenic fatty acid, can ameliorate ASD-like phenotypes in young mice prenatally exposed to BPA.
We examined the interplay between prenatal BPA, aromatase function and sex in relation to human ASD symptoms and diagnosis in the Barwon Infant Study (BIS) birth cohort 28 . By the BIS cohort health review at 7-11 years (mean = 9.05, SD = 0.74; hereafter referred to as occurring at 9 years), 43 children had a pediatrician- or psychiatrist- confirmed diagnosis of ASD against the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) criteria, as of the 30th of June 2023. ASD diagnosis was over-represented in boys with a 2.1:1 ratio at 9 years (29 boys and 14 girls; Supplementary Table 1 ). In BIS, the DSM-5 oriented autism spectrum problems (ASP) scale of the Child Behavior Checklist (CBCL) at age 2 years 29 predicted diagnosed autism strongly at age 4 and moderately at age 9 in receiver operating characteristic (ROC) curve analyses; area under the curve (AuC) of 0.92 (95% CI 0.82, 1.00) 30 and 0.70 (95% CI 0.60, 0.80), respectively. The median CBCL ASP score in ASD cases and non-cases at 9 years was 51 (IQR = 50, 58) and 50 (IQR = 50, 51), respectively. Only ASD cases with a pediatrician-confirmed diagnosis of ASD against the DSM-5, as verified by the 30th of June 2023, were included in this report. We thus examined both outcomes (ASP scale and ASD diagnosis) as indicators of ASD over the life course from ages 2 to 9 years (Supplementary Table 1 ). Quality control information for the measurement of BPA is presented in Supplementary Table 2 .
Of the 676 infants with CBCL data in the cohort sample, 249 (36.8%) had an ASP score above the median based on CBCL normative data (Supplementary Table 1 ). From a whole genome SNP array (Supplementary Methods), a CYP19A1 genetic score for aromatase enzyme activity was developed based on five single nucleotide polymorphisms (SNPs; rs12148604, rs4441215, rs11632903, rs752760, rs2445768) associated with lower estrogen levels 31 . Among 595 children with prenatal BPA and CBCL data, those in the top quartile of the genetic predisposition score, that is, children with three or more variants associated with lower levels of estrogens were classified as ‘low aromatase activity’ with the remaining classified as ‘high aromatase activity’ (Fig. 1 ). Regression analyses stratified by this genetic score and child’s sex were performed and an association between high prenatal BPA exposure (top quartile (>2.18 μg/L) and greater ASP scores was only seen in males with low aromatase activity, with a matched OR of 3.56 (95% CI 1.13, 11.22); P = 0.03 (Supplementary Table 4 ). These findings were minimally altered following adjustment for additional potential confounders. Among males with low aromatase activity, the fraction with higher than median ASP scores attributable to high BPA exposure (the population attributable fraction) was 11.9% (95% CI 4.3%, 19.0%). These results indicate a link between low aromatase function and elevated ASP scores. A sensitivity analysis using an independent weighted CYP19A1 genetic score confirmed these findings. For the additional score, the Genotype-Tissue Expression (GTEx) portal was first used to identify the top five expression quantitative trait loci (eQTLs; rs7169770, rs1065778, rs28757202, rs12917091, rs3784307) for CYP19A1 in any tissue type that showed a consistent effect direction in brain tissue. A functional genetic score was then computed for each BIS participant by summing the number of aromatase-promoting alleles they carry across the five eQTLs, weighted by their normalized effect size (NES) in amygdala tissue. This score captures genetic contribution to cross-tissue aromatase activity with a weighting towards the amygdala, a focus in our animal studies. The score was then reversed so that higher values indicate lower aromatase activity and children in the top quartile were classified as ‘low aromatase activity’ with the remaining classified as ‘high aromatase activity’. Again, a positive association between prenatal BPA exposure and ASP scores was only seen in males with low aromatase activity, with a matched OR of 3.74 (95% CI 1.12, 12.50); P = 0.03. Additional adjustment for individual potential confounders provided matched ORs between 3.13 to 3.85 (Supplementary Table 5 ).
Conditional logistic regression models were run where participants were matched on ancestry and time of day of urine collection and, for ASD diagnosis at 9 years, each case within these matched groups was individually matched to eight controls based on nearest date of and age at year 9 interview. BPA was classified in quartiles with the top quartile above 2.18 μg/L as high BPA exposure vs the other three quartiles. ‘Low aromatase enzyme activity’ means being in the top quartile and ‘high aromatase enzyme activity’ means being in the lower three quartiles of an unweighted sum of the following genotypes associated with lower estrogen levels 31 (participant given 1 if genotype is present, 0 if not): CC of rs12148604, GG of rs4441215, CC of rs11632903, CC of rs752760, AA of rs2445768. ‘Greater ASD symptoms’ represents a T-score above 50 (that is, above median based on normative data) on the DSM-5-oriented autism spectrum problems scale of the Child Behavior Checklist for Ages 1.5-5 (CBCL). Data are OR ± 95% CI. Source data are provided as a Source Data file. * Since there were only two ASD cases at age 9 in the girls with low aromatase enzyme activity group, the regression model was not run.
In subgroup analyses where we stratified by child’s sex and unweighted CYP19A1 genetic score, the results were consistent with those found at 2 years. A positive association between high prenatal BPA exposure and ASD diagnosis was only seen in males with low aromatase activity, with a matched OR of 6.24 (95% CI 1.02, 38.26); P = 0.05 (Supplementary Table 4 ). In this subgroup, the fraction of ASD cases attributable to high BPA exposure (the population attributable fraction) was 12.6% (95% CI 5.8%, 19.0%). In a sensitivity analysis where the weighted CYP19A1 genetic score was used, a similar effect size was observed in this subgroup; matched OR = 6.06 (95% CI 0.93, 39.43), P = 0.06 (Supplementary Table 4 ).
We investigated the link between BPA and aromatase further by evaluating epigenetic regulation of the aromatase gene at birth in the same BIS cohort. CYP19A1 (in humans; Cyp19a1 in the mouse) has eleven tissue-specific untranslated first exons under the regulation of tissue-specific promoters. The brain-specific promoters are PI.f 6 , 7 , 8 and PII 17 . For a window positioned directly over the primary brain promoter PI.f, higher BPA was positively associated with average methylation, mean increase = 0.05% (95% CI 0.01%, 0.09%); P = 0.009 (Fig. 2 ). Higher BPA levels predicted methylation across both PI.f and PII as a composite, mean increase per log 2 increase = 0.06% (95% CI 0.01%, 0.10%); P = 0.009. Methylation of a control window, comprising the remaining upstream region of the CYP19A1 promoter and excluding both PI.f and PII brain promoters, did not associate with BPA, P = 0.12. These findings persisted after adjustment for the CYP19A1 genetic score for aromatase enzyme activity. Thus, higher prenatal BPA exposure was associated with increased methylation of brain-specific promoters in CYP19A1 . Sex-specific differences were not observed. While these effects were identified in cord blood, methylation of CYP19A1 shows striking concordance between blood and brain tissue (Spearman’s rank correlation across the whole gene: ρ = 0.74 (95% CI 0.59, 0.84); over promoter PI.f window: ρ = 0.94 (95% CI 0.54, 0.99) 32 . Thus, prenatal BPA exposure significantly associates with disruption of the CYP19A1 brain promoter and hence likely the level of its protein product, aromatase.
Visualized using the coMET R package. A Association of individual CpGs along the region of interest with BPA exposure, overlaid with three methylation windows: a 2 CpG window positioned directly on promoter PII, and 7 and 15 CpG windows overlapping PI.f. The red shading reflects each CpG’s level of methylation (beta value). B The CYP19A1 gene, running right to left along chromosome 15, and the positions of both brain promoters. Orange boxes indicate exons. C A correlation matrix for all CpGs in this region. Highlighted in tan are the two CpGs located within the PII promoter sequence and the single CpG located within PI.f. For the 7 CpG window over promoter PI.f, higher BPA associated positively with methylation, mean increase = 0.05% (95% CI 0.01%, 0.09%); P = 0.009, after adjustment for relevant covariates including cell composition. The BPA-associated higher methylation of the brain promoter PI.f region remained evident when the window was expanded to 15 CpGs (mean increase = 0.06%, 95% CI [0.01%, 0.11%], P = 0.04). For PII, the BPA-associated mean methylation increase was 0.07%, 95% CI [-0.02%, 0.16%], P = 0.11). BPA also associated positively with methylation across both PI.f and PII as a composite, mean increase = 0.06% (95% CI 0.01%, 0.10%); P = 0.009. For the remainder of CYP19A1 , excluding both PI.f and PII brain promoters, there was no significant association, P = 0.12. Higher CYP19A1 brain promoter methylation leads to reduced transcription 17 . All statistical tests are two sided. Source data are provided as a Source Data file.
Previously, the Columbia Centre for Children’s Health Study-Mothers and Newborns (CCCEH-MN) cohort (Supplementary Table 3 ) found BPA increased methylation of the BDNF CREB-binding region of promoter IV both in rodent blood and brain tissue at P28 and in infant cord blood in the CCCEH-MN cohort 33 . In rodents, BDNF hypermethylation occurred concomitantly with reduced BDNF expression in the brain 33 . Re-examining the CCCEH-MN cohort, BPA level was also associated with hypermethylation of the aromatase brain promoter P1.f (adjusted mean increase 0.0040, P = 0.0089), replicating the BIS cohort finding.
In BIS, we aimed to reproduce these BDNF findings and extend them to investigate aromatase methylation as a potential mediator of the BPA- BDNF relationship. A link between aromatase and methylation of the BDNF CREB-binding region is plausible given that estrogen (produced by aromatase) is known to elevate brain expression of CREB 34 , 35 . In BIS, male infants exposed to BPA (categorized as greater than 4 µg/L vs. rest, following the CCCEH-MN study) had greater methylation of the BDNF CREB-binding site (adjusted mean increase = 0.0027, P = 0.02). This was also evident overall (adjusted mean increase = 0.0023, P = 0.006), but not for females alone (adjusted mean increase = 0.0019, P = 0.13). We then assessed whether methylation of aromatase promoter P1.f mediates this association. In both cohorts, aromatase methylation was positively associated with BDNF CREB-binding-site methylation in males (BIS, adjusted mean increase = 0.07, P = 0.0008; CCCEH-MN, adjusted mean increase = 0.91, P = 0.0016). In the two overall cohorts, there was evidence that the effect of increased BPA on BDNF hypermethylation was mediated partly through higher aromatase methylation (BIS, indirect effect, P = 0.012; CCCEH-MN, indirect effect, P = 0.012).
Bpa reduces aromatase expression in human neuroblastoma sh-sy5y cell cultures.
To validate the findings of our human observational studies on BPA and aromatase expression, we began by studying the effects of BPA exposure on aromatase expression in the human neuroblastoma cell line SH-SY5Y (Fig. 3A ). Indeed, the aromatase protein levels more than halved in the presence of BPA 50 μg/L ( P = 0.01; Fig. 3B ) by Western Blot analysis.
A Western Blot (1 representative blot) demonstrates that increasing BPA concentrations reduced immunoblotted aromatase protein signals (green fluorescence, 55 kDa) in lysates from human-derived neuroblastoma SH-SY5Y cells. Each sample was normalized to its internal house keeping protein β-Actin (red fluorescence, 42 kDa). B Aromatase immunoblotted signals in SH-SY5Y cells treated with vehicle or BPA ( n = 3 independent experiments/group). Five-day BPA treatment of SH-SY5Y cells leads to a significant reduction in aromatase following 50 mg/L(MD = 89, t(6) = 4.0, P = 0.01) and 100 mg/L (MD = 85, t(6) = 3.9, P = 0.01) BPA treatment, compared to vehicle. C BPA treatment (50 µg/kg/day) of Cyp19 -EGFP mice at E10.5-E14.5 results in fewer EGFP+ neurons in the medial amygdala (MD = -5334, t(4) = 5.9, P = 0.004) compared to vehicle mice, n = 3 mice per treatment. Independent t -tests were used and where there were more than two experimental groups ( B ), P -values were corrected for multiple comparisons using Holm-Sidak. All statistical tests were two-sided. Plots show mean ± SEM. Source data are in a Source Data file. Note: UT = untreated.
There is a prominent expression of aromatase within cells of the male medial amygdala (MeA) 11 . To visualize aromatase-expressing cells, we studied genetically modified, Cyp19 -EGFP transgenic mice harboring a single copy of a bacterial artificial chromosome (BAC) encoding the coding sequence for enhanced green fluorescent protein (EGFP) inserted upstream of the ATG start codon for aromatase ( Cyp19a1 ) 11 (see Methods). As shown, EGFP (EGFP+) expression in male mice was detected as early as embryonic day (E) 11.5 (Supplementary Fig. 1 ), indicating that aromatase gene expression is detectable in early CNS development.
To study the effects of prenatal BPA exposure on brain development, pregnant dams were subject to BPA at a dose of 50 μg/kg/day via subcutaneous injection, or a vehicle injection during a mid-gestation window of E10.5 to E14.5, which coincides with amygdala development. This dose matches current USA recommendations 36 , 37 as well as the Tolerable Daily Intake (TDI) set by the European Food Safety Authority (EFSA) at the time that the mothers in our human cohort were pregnant 28 , 38 . In these experiments, we observed that prenatal BPA exposure led to a 37% reduction ( P = 0.004) in EGFP+ neurons in the MeA of male EGFP+ mice compared to control mice (Fig. 3C ). These results are consistent with our findings in SH-SY5Y cells that indicate that BPA exposure leads to a marked reduction in the cellular expression of aromatase.
Next, we evaluated post-weaning social approach behavior (postnatal (P) days P21-P24) using a modified three-chamber social interaction test 39 (Fig. 4C ). As shown, male mice with prenatal exposure to BPA were found to spend less time investigating sex- and age-matched stranger mice, when compared with vehicle-treated males (with a mean time ± SEM of 101.2 sec ± 11.47 vs. 177.3 s ± 26.97, P = 0.0004; Fig. 4A ). Such differences were not observed for female mice prenatally exposed to BPA (Fig. 4A ). As a control for these studies, we found that the presence of the EGFP BAC transgene is not relevant to behavioral effects in the test (Supplementary Fig. 2 ), and the proportions of EGFP transgenic mice were not significantly different across BPA-exposed and vehicle-exposed cohorts.
Sociability is the higher proportion of time spent in the stranger interaction zone compared to the empty interaction zone. In the three chamber social interaction test ( A ) BPA-exposed mice ( n = 30, MD = 75 s, t(96) = 3.7 P = 0.0004) spent less time investigating the stranger mouse as compared with male control ( n = 21) mice. B Male ArKO ( n = 8, MD = 43 s, t(30) = 2.3, P = 0.03) mice also spent less time with the stranger compared to male WT littermates ( n = 9). C A schematic of the 3-Chamber Sociability Trial. Created with BioRender.com. D Male BPA-exposed mice ( n = 12, MD = 8.2, U = 11, P = 0.048.) spent more time grooming compared to control ( n = 5) mice. There were no differences between female BPA-exposed ( n = 9) and female control ( n = 6) mice. Independent t -tests were used P -values were corrected for multiple comparisons using Holm-Sidak. For ( C ), a Mann–Whitney U test was used. All statistical tests were two-sided. Plots show mean ± SEM. Source data are in a Source Data file. Note: Veh = vehicle.
To determine if the effects of prenatal BPA exposure were developmentally restricted, we delivered subcutaneous injections (50 µg/kg/day) of BPA to pregnant dams at early (E0.5–E9.5), mid (E10.5–E14.5), and late (E15.5–E20.5) stages of gestation. From these experiments, we found that while male pups exposed to BPA in mid-gestation developed a social approach deficit, such behavioral impairments were not observed for early or late gestation BPA exposure (Supplementary Fig. 3 ). In addition, we performed experiments in which BPA was available to dams by voluntary, oral administration (50 µg/kg/day) during mid-gestation. As shown, a social approach deficit was again observed in male mice (Supplementary Fig. 4 ), consistent with results from prenatal (mid-gestation) BPA exposure by subcutaneous injections. Thus, we find that prenatal BPA exposure at mid-gestation (E10.5-E14.5) in mice leads to reduced social approach behavior in male, but not female offspring. Notably, the amygdala of embryonic mice undergoes significant development during mid-gestation 40 .
Having demonstrated that prenatal BPA exposure reduces aromatase expression in SH-SY5Y cells and affects the postnatal behavior of mice, we next asked if the aromatase gene ( Cyp19a1 ) is central to these phenotypes. To address this, we performed social approach behavioral testing (Supplementary Fig. 5 ) on aromatase knockout (ArKO) mice 41 which have undetectable aromatase expression. The social preference towards the stranger interaction zone compared to the empty zone was only evident for the wildtype ( P = 0.003 Fig. 4B ) but not the ArKO ( P = 0.45 Fig. 4B ). This male-specific social interaction deficit is similar to the BPA exposed pups. Further, postnatal estrogen replacement could reverse the ArKO reduction in sociability seen in males ( P = 0.03 Supplementary Fig. 5 ) resulting in a similar stranger-to-empty preference in the E2-treated ArKO as observed for wildtype. The female ArKO pups did not have a sociability deficit (Supplementary Fig. 5 ).
Further, we did not observe any behavioral differences between ArKO vs WT (or BPA exposed vs unexposed) mice of both sexes in Y-maze test. All groups were able to distinguish the novel arm from the familiar arm. All groups spent significantly more time in the novel arm compared to the familiar arm (Supplementary Fig. 6 ), excluding major short-term memory, motor and sensory intergroup difference contributions.
Using the water squirt test, we have previously reported that male ArKO, but not female ArKO mice displayed excessive grooming, a form of repetitive behavior, compared to WT mice 42 . Thus, we conducted the water squirt test on BPA-exposed mice to find that male but not female mice exhibited excessive grooming behavior ( P = 0.048; Fig. 4D). Thus, male prenatal BPA-exposed mice and ArKO mice, but not females, exhibited such repetitive behaviors compared to control mice.
The development and function of the amygdala are highly relevant to human brain development and ASD 43 , 44 . Notably, the medial amygdala (MeA) is central to emotional processing 45 , and this tissue is a significant source of aromatase-expressing neurons. Given that aromatase function in the amygdala is significant for human cognition 46 and behavior 12 , 47 , and that aromatase is highly expressed in the mammalian MeA, as particularly observed in male mice 11 , we investigated changes to the structure and function of this brain region. We performed stereology analyses on cresyl violet (Nissl)-stained sections of male MeA, we observed a 13.5% reduction in neuron (defined by morphology, size, and presence of nucleolus) number. Compared to the vehicle-exposed males, BPA-exposed males had significantly reduced total neuron number (mean count of 91,017 ± SEM of 2728 neurons vs 78,750 ± SEM of 3322 neurons, P = 0.046; Supplementary Fig. 7 ).
We further examined the characteristics of cells within this amygdala structure in detail using Golgi staining (Fig. 5A, B ). We found that the apical and basal dendrites in the MeA were significantly shorter in male BPA-exposed mice vs. vehicle-treated mice (apical: 29.6% reduction, P < 0.0001; basal, P < 0.0001). This phenotype was also observed for male ArKO vs. WT mouse brains (apical 35.0% reduction, P < 0.0001; basal 31.9% reduction, P < 0.0001; Fig. 5A ). Dendritic spine densities of apical and basal dendrites of male ArKO mice, as well as male mice exposed to BPA, were also significantly reduced (KO vs WT apical, P = 0.01; KO vs WT basal, P < 0.0001; BPA treated vs vehicle treated apical P < 0.0001; BPA treated vs vehicle treated basal, P = 0.004; Fig. 5B ). The dendritic lengths (Fig. 5A ) and spine densities (Fig. 5B ) for apical and basal neurites within the MeA of female ArKO mice or BPA-exposed mice were not significantly different compared to control. Thus, in the context of aromatase suppression by prenatal BPA-exposure, or in ArKO mice lacking aromatase, we find that the apical and basal dendrite features within the MeA are affected in a sexually dimorphic manner.
A Golgi staining showed shorter apical and basal dendrites in male BPA-exposed (apical: n = 27β = −136 μ m, 95% CI [−189, −83], P = 6.0 × 10 −7 ; basal: n = 27 neurons β =-106, 95% CI [−147, −64], P = 5.1 × 10 −7 ) and ArKO mice (apical: n = 27 β = −194, 95% CI [−258, −130], P = 2.9 × 10 −9 ; basal: n = 27, β = −121, 95% CI [−143, −100], P = 1.2 × 10 −29 ) compared to male vehicle (apical n = 27, basal n = 27) or WT (apical n = 27, basal n = 29). Female BPA-exposed mice had longer basal dendrites vs. vehicle ( n = 26 neurons/group, β = 133, 95% CI [76, 191], P = 5.5 × 10 −6 ), while female ArKO mice had shorter basal dendrites vs. WT ( n = 22 neurons/group, β = −45, 95% CI [−91, −0.2], P = 0.049). Significant sex-by-BPA-treatment interaction effects were observed for apical ( P = 0.0002) and basal ( P = 3.0 × 10 −11 ) dendritic lengths, and a sex-by-genotype interaction for basal length ( P = 0.003) but not apical length ( P = 0.19). B Golgi staining showed male BPA-exposed (apical: n = 39, β = −7.0, 95% CI [−10.1, −4.0], P = 5.5 × 10 −6 ; basal: n = 90, β = −3.4, 95% CI [−5.7, −1.1], P = 0.004) and ArKO (apical: n = 51, β = −6.8, 95% CI [−12.2, −1.3], P = 0.01; basal: n = 97, β = −3.8, 95% CI [−5.4, −2.2], P = 5.2 × 10 −6 ) mice had lower spine densities on apical and basal dendrites vs. vehicle (apical n = 46, basal n = 106) or WT (apical n = 53, basal n = 109) mice. Female mice exhibited no spine density differences for BPA exposure (apical n = 74, basal n = 106) vs. vehicle (apical n = 61, basal n = 103) and ArKO (apical n = 94, basal n = 86) vs. WT (apical n = 88, basal n = 83). There was a significant sex-by-BPA-treatment interaction for apical spine density ( P = 0.0005) but not basal ( P = 0.99), and no significant sex-by-genotype interactions (apical: P = 0.08; basal: P = 0.19). For golgi staining experiments, 3 mice/group with 6–9 neuron measures/mouse. Spine count datapoints represent the number of spines on a single 10μm concentric circle. C c-Fos fluorescent immunostaining in adult male PD-MeA revealed fewer c-Fos+ve cells in BPA-exposed ( n = 3) vs. vehicle mice ( n = 3; mean difference MD = 3687, t (4) = 16.12, P < 0.0001) and ArKO ( n = 4) vs. WT mice ( n = 4; MD = −10237; t (4) = 6.48, P = 0.0002). Early postnatal estradiol restored ArKO c-Fos to WT levels ( n = 4; MD = −3112; t(4) = 1.97, P = 0.08). D Microelectrode array electrophysiology showed a lower rate of change in EPSP over 1-4 volts in male BPA-exposed mice ( n = 5 mice, n = 11 slices) vs. vehicle ( n = 7 mice, n = 12 slices; P = 0.02). Generalized estimating equations were used clustering by mouse ( A , B ) or voltage input ( D ) and assuming an exchangeable correlation structure. For ( C ), independent t -tests were used and where there were more than two experimental groups (ArKO analysis), P -values were corrected for multiple comparisons using Holm–Sidak. All statistical tests were two-sided. Plots show mean ± SEM. Source data are in a Source Data file.
The amygdala, a social processing brain region, is hyporesponsive in ASD (see review ref. 48 ). A post-mortem stereology study reported that adolescents and adults diagnosed with ASD feature an ~15% decrease in the numbers of neurons within the amygdala 43 . Also, functional MRI studies report amygdala hypoactivation in participants with ASD compared to controls 49 . Given that the amygdala is a significant source of aromatase-expressing neurons, we next conducted a series of studies to explore how aromatase deficiency influences the male mouse amygdala, using a combination of c-Fos immunohistochemistry, Golgi staining of brain sections, as well as electrophysiological analyses.
To investigate amygdala activation responses after interacting with a stranger mouse, we performed c-Fos immunohistochemistry (a marker for neuronal activation 50 ; Supplementary Fig. 8 ). As shown, prenatal BPA-exposed mice featured 58% fewer c-Fos positive neurons than in the amygdala of vehicle-exposed mice brains ( P < 0.0001; Fig. 5C ). Similarly, we found that the MeA of ArKO mice had a marked deficit of 67% c-Fos-positive neurons when compared with WT ( P = 0.0002) mice, which was ameliorated by early postnatal estradiol replacement (Fig. 5C ). Therefore, prenatal BPA exposure or loss of aromatase expression in ArKO mice leads to amygdala hypoactivation.
Next, we measured the synaptic excitability (I/O curve) of the MeA using multiple electrode analysis, with excitatory postsynaptic potential (EPSP) output indicative of electrical firing by local neurons. As shown, compared to corresponding controls, we find that MeA excitability (I/O curve) is significantly reduced in male mice prenatally exposed to BPA as well as in male ArKO mice (Figs. 5 D and 9D ). As shown, at 4-volt input, BPA treatment resulted in a 22.8% lower ( P = 0.02) excitatory EPSP output than the vehicle treatment, while a 21% reduction ( P = 0.03) in signal was observed for male ArKO mice compared to male WT mice. Thus, prenatal BPA exposure leads to functional hypoactivation of the amygdala of male mice, and this pattern is also evident in male ArKO mice.
It has been reported that individuals with ASD show distinct anatomical changes within the somatosensory cortex, including in neurons of cortical layer V 51 . We previously reported that layer V within the somatosensory cortex is disrupted in ArKO mice 52 . Thus, we performed Golgi staining to study the apical and basal dendrites of neurons within layer V of the somatosensory cortex following prenatal BPA exposure, as well as in ArKO mice. As shown, we found that apical and basal dendrite lengths of layer V cortical neurons were significantly decreased in male mice prenatally exposed to BPA, compared with vehicle-treated mice (apical P = 0.04; basal P < 0.0001, Fig. 6A ). Such reductions in dendrites were also reported in male ArKO vs. WT mice (apical P < 0.0001; basal P = 0.02; Fig. 6A ). Furthermore, we found that dendritic spine densities on apical dendrites were also reduced (BPA-exposed mice vs. vehicle, P = 0.04; ArKO vs. WT mice, P = 0.01 (Fig. 6B ).
A Golgi staining showed shorter apical and basal dendrites in male BPA-exposed (apical: n = 36, β =-350μm, 95% CI [−679, −20], P = 0.04; basal: n = 36, β = −217, 95% CI [−315, −119], P = 1.4 × 10 −5 ) and ArKO mice (apical: n = 35, β = −541.9, 95% CI [−666, −417], P = 1.3 × 10 −17 ; basal: n = 35, β = −163, 95% CI [−308, −17], P = 0.02) compared to male vehicle (apical n = 35, basal n = 36) or WT(apical n = 36, basal n = 36). B Golgi staining showed male BPA-exposed (apical: n = 186, β = −4.7, 95% CI [−9.2, −0.2], P = 0.04; basal: n = 40 β = −6.7, 95% CI [−16, 2.8], P = 0.17) and ArKO (apical: n = 148 β = −4.4, 95% CI [−7.7, −1.0], P = 0.01; basal: n = 51 β = −5.2, 95% CI [−14.4, 4.1], P = 0.27) mice had lower spine densities on apical but not on basal dendrites vs. vehicle (apical n = 189, basal n = 56) or WT mice (apical n = 185, basal n = 55). For golgi staining experiments, 3 mice/group with 9–12 neuron measures/mouse. Spine count datapoints represents the number of spines on a single 10 μm concentric circle. C Representative photomicrographs of golgi stained cortical neurons, scale bar is 100μm. D Electrocorticograms (ECoG) revealed an increased in the average spectral power at 8 Hz in BPA-exposed ( n = 4; * a 8 Hz MD = −0.5; t(325) = 3.4 P = 0.01) mice and (E) 4–6 Hz in ArKO mice ( n = 4; * b 4 Hz MD = −0.2; t(120) = 4.3, P = 0.0006; * c 5 Hz MD = −0.2, t (120) = 6.1, P < 0.0001; * d 6 Hz MD = −0.2, t (120) = 5.2 P < 0.0001) vs. vehicle ( n = 7) or WT ( n = 4)mice. Generalized estimating equations were used clustering by mouse (Panels A, B) and assuming an exchangeable correlation structure. For ( D ) and(E) Independent t test were used used, P -values were corrected for multiple comparisons using Holm-Sidak. All statistical tests were two-sided. Plots show mean ± SEM. Source data are in a Source Data file.
To explore the effects of reduced aromatase on cortical activity, we performed electrocorticography (ECoG) recordings from mice in both experimental models (Fig. 6C ). As shown, spectral analysis revealed an increased power in the range of 4–6 Hz for ArKO vs. WT mice (4 Hz P = 0.0006, 5 Hz P < 0.0001, 6 Hz P < 0.0001; Fig. 6D ) and at 8 Hz for BPA-exposed vs. vehicle mice ( P = 0.01; Fig. 6C ). These data indicate that BPA-exposure or loss of aromatase in ArKO mice affects cortical activity, a result which is reminiscent of cortical dysfunction evidenced by EEG recordings on human participants diagnosed with ASD 53 .
It has been reported that BPA interferes with estrogen signaling through its competitive interaction and binding with estrogen receptors α (ERα) and β (ERβ) 54 . To explore this in the context of our findings, we used high-resolution in silico 3D molecular docking simulations to model the binding affinity of the natural ligand 17β-estradiol, the putative ligand BPA, as well as a putative therapeutic ligand of interest 10HDA with ERα (Protein Data Bank (PDB) ID: 5KRI) and ERβ (PDB ID: 1YYE). As shown, our spatial analysis indicated that all three ligands have robust binding affinity (Fig. 7 ; Supplementary Movie 1 ). However, while docking alignment revealed that the predicted fit for 10HDA is strikingly similar to that of 17β-estradiol 25 , 55 , BPA showed a greater mismatch (Fig. 7D ), consistent with previous reports that BPA is 1000-fold less estrogenic than the native ligand 56 . Thus, at least for ERα and Erβ, we find that 10HDA may be effective as a competitive ligand that could counteract the effects of BPA on estrogen signaling within cells.
In silico molecular docking analysis of estrogen receptor β (ERβ, Protein Data Bank (PDB) ID: 1YYE; encoded by the ES gene) using the DockThor platform, showing binding predictions for ( A ) the native ligand 17β-estradiol (E2), ( B ) bisphenol A (BPA), ( C ) Trans-10-hydroxy-2-decenoic acid (10HDA), and ( D ) E2 and BPA (left) and E2 and 10HDA (right) superimposed for spatial alignment comparison. While the molecular affinities of BPA and 10HDA for Erβ were comparable (−9.2 vs. −7.9, respectively), 10HDA aligns better with the binding conformation of the endogenous ligand E2, which activates the receptor. BPA is previously reported as sub-optimally estrogenic 106 — >1000-fold less compared to natural estradiol 54 , 56 —whereas 10HDA has an estrogenic role in nature 25 , 55 . Thus, 10HDA may compensate for E2 deficiency caused by a reduction in aromatase enzyme, and in competition with binding by BPA. Please see Supplementary Movie 1 for a video of the above molecular docking of ERβ with E2 superimposed with BPA and Supplementary Movie 2 for the above molecular docking of ERβ with E2 superimposed with 10HDA.
Examining male fetal primary cortical culture, BPA alone shortened neurite lengths (Fig. 8A , BPA; quantified in Fig. 8B-C ) and decreased the spine density. BPA treatment reduced both neurite length ( P = 0.0004) and spine densities ( P < 0.0001; Fig. 8A , BPA; quantified in Fig. 8B–C ). Co-administration with 10HDA ameliorated these adverse effects of BPA (Fig. 8A , 10HDA + BPA; quantified in Fig. 8B, C ).
A Representative photomicrographs of primary cultures of embryonic (ED15.5) mouse cortical neurons, red staining is βIII tubulin and green is aromatase. Scale bar is 100μm. B Compared to the BPA group, the vehicle group (β = 79.9, 95% CI [36, 124], P = 0.0004) and BPA + 10HDA group (β = 174, 95% CI [102, 247], P = 2.4 × 10 −7 ) have significantly longer neurites. The BPA + 10HDA group (β = 94, 95% CI [8, 180], P = 0.03) has longer neurites compared to the vehicle group, and there is no difference between the vehicle and 10HDA groups. C Compared to the BPA group, the vehicle group(β = 16, 95% CI [11, 21], P = 1.7 × 10 −8 ) and BPA + 10HDA group (β = 30, 95% CI [21, 40], P = 8.4 × 10 −10 ) have significantly higher spine densities. The BPA + 10HDA group (β = 14, 95% CI [4, 24], P = 0.006) has a higher spine density compared to the vehicle group, and there is no difference between the vehicle and 10HDA groups. n = 10 neurons/group. Primary cortical cell culture was obtained from 12 male mouse embryoes. Spine count datapoints represent the number of spines on a single 10μm concentric circle. Generalized estimating equations were used clustering by mouse and assuming an exchangeable correlation structure. All statistical tests were two-sided. Plots show mean ± SEM. Source data are in a Source Data file.
Guided by our findings in cultured neurons, we next investigated the effects of postnatal 10HDA administration on mice prenatally exposed to BPA at mid-gestation, as follows. After weaning, pups (six litters, 3 weeks of age) were administered daily injections of 10HDA (0 and 500 μg/kg/day; dissolved in saline, i.p.) for 3 weeks, following which pups were assessed for behavioral phenotypes. Strikingly, 10HDA treatment significantly improved social interaction (Fig. 9A ). To determine whether the effect of 10HDA administration is permanent, all treatments were withdrawn for 3 months, and mouse behaviors were subsequently re-tested. Withdrawal of 10HDA treatment in BPA-exposed male mice resulted in a deficit in social interaction (Fig. 9B ), and this deficit was once again ameliorated by a subsequent 10HDA treatment (Fig. 9C ) at 5 months of age, in adulthood. Taken together, these data demonstrate that continuous, postnatal 10HDA administration is effective for ameliorating social interaction deficits in male mice following prenatal BPA exposure.
A 10HDA treatments increased social approach in males ( n = 10/group, MD = 41.14, U = 11, P = 0.03,) but not females ( n = 8/group), compared to saline controls. After 3 months of treatment withdrawal ( B ), male mice (saline n = 8, 10HDA n = 7) no longer spent more time interacting with strangers, compared to vehicle treatment. C When male mice ( n = 8/group) were subsequently treated with a second round of 10HDA, social approach behavior was once again significantly elevated (MD = 39.2, U = 5, P = 0.003), indicative of a rescue of this behavioral effect. D Compared to the WT Saline ( n = 10) group (β = 57.1, 95% CI [47.4, 66.8]), EPSP increases at a 21% lower rate with increasing input in the ArKO Saline ( n = 14) group (β = 45.1 μV, 95% CI [40.1, 50.1], P = 0.03). No differences in slope were detected when comparing the WT Saline group with each of the other two treatment (WT n = 18, KO n = 12) groups. Mann–Whitney U tests were used and for ( D ), Generalized estimating equations were used clustering by voltage input ( D ) and assuming an exchangeable correlation structure. All statistical tests were two-sided. Plots show mean ± SEM. Source data are in a Source Data file. Note: Sal = saline, w/d = withdrawal.
Next, we wanted to determine if hypoactivity arising from the absence of aromatase in the amygdala may be influenced by 10HDA. To address this question, we studied ArKO mice using multiple-electrode analyses, following 3 weeks of treatment with 10HDA (500 μg/kg/day, i.p.). As shown in Fig, 9D , the electrical activity of the male ArKO amygdala treated with 10HDA was similar to male WT activity levels, whereas saline-treated male ArKO amygdala showed significantly lower activity ( P = 0.03) when stimulated by an input/output paradigm, suggesting that 10HDA treatment was effective to compensate the absence of aromatase. Therefore, we interpret these results to suggest that 10HDA restores signaling deficits arising from aromatase deficiency. Given that prenatal BPA exposure suppresses aromatase, 10HDA supplementation may be relevant to aromatase-dependent signaling in that context as well.
MiSeq Next-Gen Sequencing was performed on the transcriptome libraries generated from the brain cortex of the E16.5 fetuses after maternal mid-gestation BPA or vehicle exposure. The action of 10HDA was analyzed by RNAseq of transcriptome libraries from total RNA extracted from primary mouse fetal cortical cultures treated with vehicle or 10HDA. Firstly, pathway analysis of the RNAseq data was performed for Gene Ontology (GO) categories using the clusterProfileR R package. No individual pathways in the BPA analysis survived correction for multiple comparisons using an agnostic (non-candidate) approach. Further candidate investigation using the binomial test showed a significant inverse effect of BPA and 10HDA on pathways previously linked to autism 57 , with 10HDA treatment counteracting the effects of BPA on these pathways (Supplementary Fig. 9 ). Based on our Golgi staining experimental findings relating to altered dendrite morphology, we further assessed the category “dendrite extension” as a candidate pathway. Genes in this pathway were downregulated by BPA (Supplementary Figs. 10 A, 9A) and upregulated by 10HDA (Supplementary Figs. 9B , 10B ). More broadly, Fisher’s exact test showed a significant BPA-associated down-regulation ( P = 0.01), and 10HDA-associated up-regulation ( P = 0.0001), of pathways with the terms “axon” and “dendrite”. Notably, the majority (82%; 9 of the top 11 available) mid gestational biological processes whose activity is overrepresented in induced pluripotent stem cells of autism cases vs. controls 57 were impacted by BPA, and in the opposite direction to 10HDA ( P = 0.03; Supplementary Fig. 9 ).
Next, we performed pathway enrichment analysis, also using a candidate pathway approach, of the RNAseq data using Ingenuity (Fig. 10 ). Strikingly, the effects of BPA and 10HDA on gene expression were diametrically opposed across many functional domains (Fig. 10 ). For example, the canonical pathways “Synaptogenesis Signaling pathway” and “CREB signaling” were downregulated by BPA but upregulated by 10HDA. Similarly, key brain functions, e.g., growth of neurites and neural development were down regulated by BPA and reciprocally upregulated by 10HDA (Fig. 10 ). Taken together, prenatal BPA exposure is detrimental to gene expression through a mechanism that may be ameliorated by postnatal 10HDA administration. The full list of differentially expressed genes can be found in Supplementary Dataset 1 .
The Canonical pathways and Disease and Function—Brain pathway databases were selected in Ingenuity. Several key signaling pathways and brain functions were downregulated ( Z -score less than zero) by BPA and also upregulated ( Z -score greater than zero) by 10HDA. 10HDA downregulated four brain disorder related pathways—hyperactive behavior, seizures, seizure disorder and behavioral deficit. Colored boxes indicate significant ( P < 0.05, Fisher’s Exact Test with Benjamani-hochberg) changes in z -score. Boxes shaded in gray indicate non-significant gene expression changes ( P = 0.05 or greater). Source data are provided as a Source Data file.
Here, we report that prenatal BPA exposure leads to ASD endophenotypes in males, and that this involves the actions of the aromatase gene, as well as its functions in brain cells. Our multimodal approach, incorporating both human observational studies and preclinical studies with two mouse models, offer significant insight into how prenatal programming by BPA disrupts aromatase signaling to cause anatomical, neurological, as well as behavioral changes reminiscent of ASD in males.
In the Barwon Infant Study (BIS) human birth cohort the adverse effect of high prenatal BPA exposure on ASD symptoms (ASP score) at age 2, and clinical ASD diagnosis at age 9, was particularly evident among males with a low aromatase enzyme activity genetic score. Further, studying cord blood gene methylation as an outcome we have demonstrated that BPA exposure specifically methylated the offspring CYP19A1 brain promoter in the BIS cohort, replicated in the CCCEH-MN cohort. Previously, in a meta-analysis of two epigenome-wide association studies (EWAS), two CpGs in the region around the brain promoters PI.f and PII were significantly associated ( P < 0.05) with ASD in both EWAS 58 . Past work 16 , and our findings of BPA reduced aromatase expression in a neuronal cell culture and reduced aromatase-eGFP expression in mouse brain, are consistent with the brain-specific suppression of aromatase expression. We also demonstrated that BPA exposure led to a reduction in steady-state levels of aromatase in a neuronal cell line. Further, we replicated past work that higher prenatal BPA levels are associated with BDNF hypermethylation 33 , previously demonstrated to be associated with lower BDNF expression in males 33 .
We find that prenatal BPA exposure at mid-gestation in mice induces ASD-like behaviors in male but not female offspring, concomitant with cellular, anatomical, functional, and behavioral changes (Supplementary Fig. 11 and Supplementary Fig. 12 ). We found that these features were also observed in male ArKO mice, and this is important because aromatase expression is disrupted in BPA-treated male mice. In the Y-maze test, both the ArKO and BPA offspring did not show any differences with the respective control indicating that there are no major memory, sensory or motor issues in these animals. Given the distinct parallels between the effects of BPA that suppresses aromatase, as well as ArKO mice that lack aromatase, we surmise from our studies that BPA disrupts aromatase function to influence the male mouse brain which manifests as: (i) reduced excitatory postsynaptic potentials in the amygdala, (ii) reduced neuron numbers as well as dendritic lengths and spine densities for neurons within the MeA, (iii) altered cortical activity as recorded by ECoG concomitant with decreased dendritic length and spine density in layer IV/V somatosensory cortical neurons, as well as (iv) enhanced repetitive behaviors and reduced social approach to a stranger. These results in mice are consistent with studies with human participants that report abnormal neuronal structure in these comparative regions within the brains of individuals with ASD 51 . Furthermore, we investigated our gene expression dataset and found that the majority of the top biological processes over-represented in cells derived from ASD cases compared to non-cases in a human pluripotent stem cell analysis with a focus on mid gestational brain development 57 were impacted in opposite directions by BPA compared to 10HDA in our gene expression studies on the male mouse brain (Supplementary Fig. 9 ). Of note, the sexually dimorphic effect we report is consistent with work of others demonstrating that prenatal BPA exposure of rodents led to dysregulation of ASD-related genes with neuronal abnormalities, and learning and memory problems only in males 59 .
In our investigations of BPA, we recognized that 10HDA may be a suitable compound as a ligand in the context of brain ER signaling 27 because of its positive effect on gene expression through stimulation of estrogen responsive DNA elements 25 and its role in neurogenesis 27 —characteristics that altogether may compensate for a relative lack of aromatase-generated neural estrogens. Administration of 10HDA alongside BPA protected neuronal cells in culture from the adverse sequelae observed for BPA alone at the same dose. Three weeks of daily postnatal 10HDA treatment significantly enhanced the sociability of the male BPA-exposed mice and dendrite morphology in primary cell culture. The adverse decrease in dendrite lengths and spine densities of the BPA-exposed mice was also corrected by 10HDA administration (Fig. 8 ). Furthermore, postnatal 10HDA treatment restored amygdala electrical activity in the ArKO mice, indicating that 10HDA likely acts downstream of, rather than directly upon, the aromatase enzyme, given that ArKO mice lack functional aromatase. Transcriptomic analyses revealed that 10HDA upregulated, whereas BPA downregulated, gene expression for fetal programming such as for synaptogenesis and growth of neurites. Some of these pathways could be activated by factors downstream of aromatase, such as 17β-estradiol (Supplementary Fig. 13 ). In this study, the ArKO model was useful because it provided an estrogen deficient comparison 41 . We were able to demonstrate that early postnatal E2 administration restored both MeA neural activation and social preference behavior in the ArKO males.
The molecular docking simulations indicate that ERα and ERβ both comprise docking sites for 10HDA and BPA, however, 10HDA is strongly estrogenic 25 , 55 while BPA is greater than 1000-fold less potent than natural estrogen 54 . Such differences in binding are likely relevant to the diverse transcriptomic effects observed in the cells we analyzed by RNAseq.
Strengths of this study include the multimodal approach to test the hypothesis of the interplay of BPA, male sex, and aromatase suppression. In our human epidemiological studies, extensive information was available to allow confounding to be accounted for using matched analyses for the BPA-ASD cohort finding, and findings persisted after adjustment for further individual confounders. Using a modern causal inference technique, molecular mediation 60 we demonstrate in both birth cohorts that aromatase gene promoter I.f methylation underlies the known effect of higher prenatal BPA on BDNF hypermethylation. Other key features that support an underlying causal relationship include: the consistency of the findings across studies in this program (Supplementary Fig. 11 ); and the consistency with which our experimental laboratory work maps to prior studies of people with ASD (summarized in Supplementary Fig. 12 ) in relation to neuronal and structural abnormality in the amygdala 43 and abnormality in amygdala connectivity 44 , and resting-state cortical EEG 53 . Our findings are also consistent with past work indicating reduced prefrontal aromatase levels in individuals with ASD at postmortem 14 , 15 . The finding that BPA-associated gene methylation patterns in the BIS cohort were not sex-specific but that BPA-associated ASD symptoms and clinical diagnosis were more evident in males with a low genetic aromatase score would be consistent with the male vulnerability to BPA reflecting not differential epigenetic programming, but a greater vulnerability to reduced aromatase function in the developing male brain. This is reinforced by the ArKO model which resulted in an ASD-like phenotype in males not females. We have provided experimental evidence not only on the adverse neurodevelopment effects of BPA, but also experimental evidence of the alleviation of the behavioral, neurophysiological, and neuroanatomical defects following postnatal treatment with 10HDA. A human randomized controlled prevention trial that achieved bisphenol A elimination during pregnancy, with a resultant reduction in ASD among male offspring, would be a useful next step to provide further causal evidence of BPA risks but the feasibility and ethics of such an undertaking would be considerable. We demonstrate that postnatal administration of 10HDA may be a potential therapeutic agent that counteracts the detrimental impacts on distinct gene expression signatures directly impacted by prenatal BPA exposure. Furthermore, 10HDA may ameliorate deficits in ArKO mice which further suggests its utility as a replacement therapy for aromatase deficiency.
Two limitations of our human study were that BPA exposure was measured in only one maternal urine sample at 36 weeks, and that the assay may have low sensitivity 61 . We partially redressed the latter by focusing on categorical BPA values, as recommended 61 , and undertook a matched ASD analysis where determinants for BPA variation, such as the urine collection time of day, were matched to reduce misclassification. Also, functional gene expression studies were unavailable for human samples in our study, but whilst the misclassification introduced by a reliance on a SNP based score would likely lead to an underestimation, an effect was still found among males with a low genetic aromatase score. It would be useful in further studies to consider altered aromatase function with a combined epigenetic-genetic score to reflect environment-by-epigenetic and genetic determinants of low aromatase function. Direct brain EWAS measures were not available, but for the key brain promoter PI.f region of CYP19A1 , the brain-blood correlation is very high: Spearman’s rho= 0.94, 95% CI [0.80, 0.98] 32 . Although ASD symptoms (ASP score) at 2 years were based on parent report, we have previously reported that a higher ASP score was predictive of later ASD diagnosis by age 4 30 . ASD diagnosis at age 9 was verified to meet DSM-5 criteria by pediatrician audit of medical records, thereby reducing diagnostic misclassification.
In our preclinical studies, we performed the 10HDA studies and some of the BPA mechanistic studies only on male animals because our extensive laboratory and human studies, with more than 25 analyses, demonstrated that BPA exposure had significantly more adverse effects in males than females. In addition, we performed the RNASeq on the cortex of fetal mice exposed to BPA in vivo, whereas 10HDA was performed in vitro on cortical primary cell culture. This would likely increase the variability between the BPA RNASeq and the 10HDA RNASeq, yet many of the same pathways were impacted but in opposing directions. Furthermore, the changes we observed in our RNAseq data could be due to changes in the cell type or cell state. This could be clarified by future single cell RNAseq experiments now that this specific issue has been identified.
The BPA exposure of mouse dams under our experimental conditions (50 µg/kg bodyweight) matches the current Oral Reference Dose set by the United States Environmental Protection Agencyc, the current safe level set by the U.S. Food and Drug Administration (FDA) 37 , as well as the Tolerable Daily Intake set by the European Food Safety Authority 38 at the time that the mothers in our human cohort were pregnant 28 . The EFSA set a new temporary TDI of 4 µg/kg bodyweight in 2015 62 and, in December 2021, recommended further reducing this by five orders of magnitude to 0.04 ng/kg 63 ; although this was subsequently revised to 0.2 ng/kg in EFSA's scientific opinion published in 2023 64 . Therefore, the timing of this new evidence is particularly pertinent and provides direct human data to support the reduced TDI.
Consistent with typical human exposure in other settings 20 , BPA exposure in our birth cohort was substantially lower than the above, and yet we see adverse effects. Assuming fractional excretion of 1 65 and average daily urine output of 1.6 L 65 , the median urinary bisphenol concentration of 0.68 µg/L—for which we see increased odds of ASD diagnosis—equates to a total daily intake of just 13 ng/kg, given a mean maternal bodyweight of 80.1 kg at time of urine collection. Notably, while we find an adverse association at 13 ng/kg, we do not have sufficient participants with lower exposure to evaluate a safe lower limit of exposure below this. Our findings in cell culture, with concentration 5 µg/L, parallels these human findings in terms of dose response. Although there are limitations in translating concentrations across body compartments without a stronger understanding of pharmacokinetics of BPA, 5 µg/L corresponds to the 90th and 95th percentile of BPA in urine in our human cohort, and allowing for a standard factor of 10 for variability in human sensitivity used when setting TDIs 38 indicates relevance down to at least 0.5 µg/L, below the median urine concentration. Our findings in laboratory animal studies, with exposure of 50 µg/kg bodyweight, are a little higher, as they were designed to correspond to the then current recommendations 36 , 37 , 38 , but implications nevertheless have relevance within the range of exposure in our cohort. Allowing for standard factors of 10 for interspecies variability 38 and variability in human sensitivity 38 , our animal study findings support a TDI at 500 ng/kg or below, which corresponds to the upper 0.5% of our human cohort. The findings of the human study, also allowing for a factor of 10 for variability in human sensitivity 38 , therefore, support a TDI at or below 1.3 ng/kg.
Despite bans on its use in all infant products by the European Union in 2011 and the U.S. FDA in 2012, BPA remains widespread in the environment 66 . The main source of human exposure to BPA is dietary contamination 68 . Bisphenols are used in the production of common food contact materials, and migrate from those materials during use 69 , including polycarbonate food and beverage containers and the epoxy linings of metal food cans, jar lids, and residential drinking water storage tanks and supply systems 64 . Additional sources of exposure include BPA-based dental composites and sealant epoxies, as well as thermal receipts 64 . BPA levels in pregnant women have previously been reported to be higher for young mothers, smokers, lower education, and lower income 70 . A substantial proportion of ASD cases might be prevented at the population level if these findings were causal and prenatal maternal BPA exposure were reduced. Here, exposures in the top quartile of BPA (>2.18 ug/L) correspond to a population attributable fraction (PAF) for males with low aromatase of 12.6% (95% CI 5.8%, 19.0%) although this estimate is imprecise as it is based on low case numbers. The only other available study with data on BPA exposure (>50 ug/L) and ASD provides an estimate in all children of 10.4% 67 . These studies have misclassification issues (e.g., a single urine measure for BPA and, in the Stein et al. study, an ASD diagnosis derived from health care sources 67 ) but these misclassifications are likely non-differential and thus would bias findings towards the null. Additionally, we need to consider that the above findings of RfD/TDI and PAF are based on BPA alone. Factoring in that most exposures occur as part of a chemical mixture adds additional concern 71 . For example prenatal valproic acid exposed mice (an established ASD mouse model) also have a lower brain aromatase expression 72 .
In summary, this multimodal program of work has shown an adverse effect of higher maternal prenatal BPA on the risk of male offspring ASD by a molecular pathway of reduced aromatase function, which plays a key role in sex-specific early brain development. Overall, these findings add to the growing evidence base of adverse neurodevelopmental effects from bisphenol and other manufactured chemical exposure during pregnancy. The case is compelling and supports broader evidence on the need to further reduce BPA exposure, especially in pregnancy. We also envision that our findings will contribute to new interventions for the prevention and/or amelioration of ASD targeting this specific pathophysiological pathway and we have identified one possible neuroprotective agent—10HDA—that has strong laboratory support. This agent now warrants further study, including human safety and efficacy evaluation.
The human Barwon Infant Study cohort study was approved by the Barwon Health Human Research Ethics Committee, and families provided written informed consent. Parents or guardians provided written informed consent at prenatal recruitment and again when the child was 2 years of age. The human Columbia Center for Children’s Environmental Health Mothers and Newborn cohort study was approved by the Institutional Review Boards of Columbia University and the Centers for Disease Control and Prevention, and all participants in the study provided informed consent. All procedures involving mice were approved by the Florey Institute of Neuroscience and Mental Health animal ethics committee and conformed to the Australian National Health and Medical Research Council code of practice for the care and use of animals for scientific purposes and All experiments were designed to minimize the number of animals used, as well as pain and discomfort. This work adheres to the ARRIVE essential 10 guidelines.
Participants.
From June 2010 to June 2013, a birth cohort of 1074 mother–infant pairs (10 sets of twins) were recruited using an unselected antenatal sampling frame in the Barwon region of Victoria, Australia 28 . Eligibility criteria, population characteristics, and measurement details have been provided previously 28 ; 847 children had prenatal bisphenol A measures available (Supplementary Table 1 ).
We used a direct injection liquid chromatography tandem mass spectrometry (LC-MS/MS) method, as previously described in detail 73 . In summary, a 50 µL aliquot of urine was diluted in milli-Q water and combined with isotopically-labeled standards and b-glucuronidase (from E. Coli -K12). Samples were incubated for 90 min at 37 °C to allow for enzymatic hydrolysis of bisphenol conjugates before quenching the reaction with 0.5% formic acid. Samples were centrifuged before analysis, which was performed using a Sciex 6500 + QTRAP in negative electrospray ionization mode. The BPA distribution and quality control attributes for the application of this method to the Barwon Infant Study (BIS) cohort are shown in Supplementary Table 2 .
Between the ages of seven and ten, a health screen phone call was conducted to gather information on autism spectrum disorder (ASD) diagnoses and symptomology. Out of the 868 individuals who responded to the health screen, 80 had an ASD diagnosis reported by their parents/guardians or were identified as potentially having ASD. The parent-reported diagnoses were confirmed by pediatric audit of the medical documentation to verify an ASD diagnosis as per DSM-5 guidelines. Participants that had a parent-reported diagnosis and then a verified pediatrician diagnosis by 30 June 2023 and whose diagnoses occurred before the date of their 9-year health screen were included as ASD cases in this study’s analyses ( n = 43). Participants were excluded if (i) their parent/guardian responded with ‘Yes’ or ‘Under Investigation’ to the question of an ASD diagnosis on the year-9 health screen but their diagnosis was not verified by 30 June 2023 ( n = 26), or (ii) they had a verified diagnosis of ASD by 30 June 2023 but their date of diagnosis did not precede the date of their year-9 health screen ( n = 15). The DSM-5-oriented autism spectrum problems (ASP) scale of the Child Behavior Checklist for Ages 1.5-5 (CBCL) administered at 2-3 years was also used as an indicator of autism spectrum disorder.
Blood from the umbilical cord was gathered at birth and then transferred into serum coagulation tubes (BD Vacutainer). Following this, the serum was separated using centrifugation as described elsewhere 74 . Genomic DNA was extracted from whole cord blood using the QIAamp DNA QIAcube HT Kit (QIAGEN, Hilden, Germany), following manufacturer’s instructions. Genotypes were measured by Erasmus MC University Medical Center using the Infinium Global Screening Array-24 v1.0 BeadChip (Illumina, San Diego, CA, USA). The Sanger Imputation Service (Wellcome Sanger Institute, Hinxton, UK) was used for imputing SNPs not captured in the initial genotyping using the EAGLE2 + PBWT phasing and imputation pipeline with the Haplotype Reference Consortium reference panel 75 . Detailed methods are provided elsewhere 76 .
Genome-wide DNA methylation arrays and analysis methods can be found in the Supplementary Methods.
The study participants consisted of mothers and their children who were part of the prospective cohort at the Columbia Center for Children’s Environmental Health Mothers and Newborn (CCCEH-MN) in New York City (NYC). They were enrolled between the years 1998 and 2003, during which they were pregnant. The age range for these women was between 18 and 35, and they had no prior history of diabetes, hypertension, or HIV. Furthermore, they had not used tobacco or illicit drugs and had initiated prenatal care by the 20th week of their pregnancy. Every participant gave informed consent, and the research received approval from the Institutional Review Boards at Columbia University as well as the Centers for Disease Control and Prevention (CDC) 33 .
Epigenetic methods have been previously described 77 . Briefly, DNA methylation was measured in 432 cord blood samples from the CCCEH-MN cohort using the 450 K array (485,577 CpG sites) and in 264 MN cord blood samples using the EPIC array (866,895 CpG sites) (Illumina, Inc., San Diego, CA, USA).
BPA measures in the CCCEH were based on spot urine samples collected from the mother during pregnancy (range, 24–40 weeks of gestation; mean, 34.0 weeks) 33 , 78 .
Maternal urinary BPA concentrations were corrected for specific gravity to control for differences in urine dilution. Given a high proportion of the sample (46%) had BPA concentrations that were not detected or below the limit of detection (LOD), a dichotomous BPA exposure variable was formed using the 75th percentile as the cut-point. Dichotomizing the measurements in this way is also likely to give similar results regardless of whether indirect or direct analytical methods were used 79 . This is desirable since indirect methods might be flawed and underestimate human exposure to BPA 61 .
To evaluate whether autism spectrum problems at 2 years could be used as a proxy for later ASD diagnosis, receiver operating characteristic (ROC) curve analyses were used. CBCL ASP at age 2 years predicted diagnosed autism strongly at age 4 and moderately at age 9 with an area under the curve of 0.92 (95% CI 0.82, 1.00) and 0.70 (95% CI 0.60, 0.80), respectively.
According to the normative data of the CBCL, T-scores greater than 50 are above the median. Due to a skewed distribution, ASP measurements were dichotomized using this cut point, which has respective positive and negative likelihood ratios of 2.68 and 0.00 in the prediction of verified ASD diagnosis at 4 years and 1.99 and 0.49 in the prediction of verified ASD diagnosis at 9 years.
A CYP19A1 genetic score for aromatase enzyme activity was developed based on five genotypes of single nucleotide polymorphisms (CC of rs12148604, GG of rs4441215, CC of rs11632903, CC of rs752760, AA of rs2445768) that are associated with sex hormone levels 31 . Participants were classified as ‘low activity’ if they were in the top quartile, that is, they had three or more genotypes associated with lower levels of estrogen and as ‘high activity’ otherwise. Conditional logistic regression model analyses investigating the association between prenatal BPA levels and (i) early childhood ASP scores and (ii) verified ASD diagnosis at 9 years were conducted in the full sample, repeated after stratification by child’s sex (assigned at birth based on visible external anatomy), and repeated again after further stratifying by the CYP19A1 genetic score. Matching variables included child’s sex (in the full sample analysis only), ancestry (all four grandparents are Caucasian vs not) and time of day of maternal urine collection (after 2 pm vs before). Within these matched groups, we additionally matched age-9 ASD cases and non-cases based on the date of the health screen and child’s age at the health screen using the following procedure. Each case was matched to a single non-case based on nearest date of and age at health screen. Once all cases had one matched non-case, a second matched non-case was allocated to each case, and so on until all cases had 8 matched non-cases (8 was the most possible in the boys with high aromatase activity sub-sample and so this number was used across all sub-samples). The order by which cases were matched was randomly determined at the start of each cycle.
The guidelines for credible subgroup investigations were followed 80 . Only two categorical subgroup analyses were conducted, and these were informed a priori by previous literature and by initial mouse study findings. The adverse BPA effects in males with low aromatase enzyme activity (as inferred from the CYP19A1 genetic score) were expected to be of higher magnitude, based on the prior probabilities from the laboratory work. A systematic approach was used to evaluate non-causal explanations and build evidence for causal inference, considering pertinent issues such as laboratory artefacts that are common in biomarker and molecular studies 81 .
A second CYP19A1 genetic score was developed for use in sensitivity analyses. The Genotype-Tissue Expression (GTEx) portal was used to identify the top five expression quantitative trait loci (eQTLs) for aromatase in any tissue type that showed a consistent effect direction in brain tissue. A functional genetic score was then computed for each BIS participant by summing the number of aromatase-promoting alleles they carry across the five eQTLs (AA of rs7169770, CC of rs1065778, AA of rs28757202, CC of rs12917091, AA of rs3784307), weighted by their normalized effect size (NES) in amygdala tissue. The score was then reversed so that higher values indicate lower aromatase activity. The score thus captures genetic contribution to reduced cross-tissue aromatase activity with a weighting towards the amygdala, a focus in our animal studies. The variable was dichotomized using the 75th percentile as the cut-point and the above stratified analyses were repeated with this new weighted score replacing the original, unweighted score.
For the human epigenetic investigations, we used multiple linear regression and mediation 60 approaches. As in past work 33 , BPA was classified as greater than 4 μg/L vs less than 1 μg/L in the CCCEH-MN cohort. A comparable classification was used for the BIS cohort, with greater than 4 μg/L vs the rest. In both cohorts the regression and mediation analyses were also adjusted for sex, gestational age, self-reported ethnicity, and cord blood cell proportions. In the BIS cohort, ethnicity was defined as all four grandparents are Caucasian vs not (see Table S3 ). For the CCCEH-MN cohort, ethnicity was defined as Dominican vs African American 33 . We used statistical software packages R v3.6.3 82 and Stata 15.1 83 .
Shsy-5y cell culture study, bpa treatment on aromatase expression in cell culture.
Human neuroblastoma SHSY-5Y cells were chosen because they were known to express aromatase and SH-SY5Y have been used in ASD research 84 . SHSY-5Y cells (CRL-2266, American Type Culture Collection, Virginia, USA) were maintained in Dulbecco’s Modified Eagle Medium (DMEM) (10313-021, Gibco-life technologies, New York (NY), USA) supplemented with 10% heat-inactivated Fetal Bovine Serum (FBS) (12003C-500 mL, SAFC Biosciences, Kansas, USA), 1% penicillin streptomycin (pen/strep) (15140-122, Gibco-life technologies, NY, USA) and 1% L-Glutamine (Q) (25030-081, Gibco-life technologies, NY, USA) at 37 °C in a humidified atmosphere of 95% air and 5% CO 2 . SHSY-5Y cells were grown in 175 cm 2 cell culture flasks (T-175) (353112, BD Falcon, Pennsylvania, USA). Cells were passaged when the seeding density of the T-175 flasks was reached (roughly 80-90% confluence). Cells were passaged by aspirating media from flasks and flasks were then washed once with 10 mL of DPBS (14190-136 Gibo-life technologies, NY, USA) to remove the FBS (inhibits the actions of trypsin). Next, cells were incubated with trypsin (2 mL/T-175 flask) at 37 °C for 5 min to detach cells from the flask wall. To prevent further action of trypsin, media (8 mL/T-175 flask) was added, and contents were pipetted up and down to disperse cell clumps. The cell suspension was then transferred to a 15 mL centrifuge tube (430791, Corning CentriStar, Massachusetts, USA) and centrifuge (CT15RT, Techcomp, Shanghai, China) for 5 min at 1000 RPM at room temperature (RT). The media was then aspirated from these tubes and the cell pellet resuspended in 1 mL of media. Cell viability counts were performed using a hemocytometer (Hausser Scientific, Pennsylvania, USA) to determine the number of live versus dead cells in solution. Two μL of cell suspension was diluted with media (98 μL) and then trypan blue (100 μL) (T8154, Sigma-Aldrich Co., St. Louis, MO, USA) (which labeled dead cells) in a sterile microcentrifuge tube (MCT-175-C-S, Axygen, California, USA). Ten μL of this solution was loaded into the hemocytometer and imaged using a light microscope (DMIL LED, Leica, Germany). Dead cells appeared blue under the microscope because these cells take up the dye whereas live cells were clear (i.e., unstained). Cells were counted in the outer four squares located in each chamber (two chambers, eight squares), with their dimensions known. The average of the eight counts was multiplied by the dilution factor and by 104, yielding the concentration of cells/mL solution. Average cell counts were plotted against treatment groups using GraphPad Prism 9.0 (GraphPad Software, Inc., San Diego, CA).
Bisphenol A (BPA) (239658-50 G, Sigma-Aldrich Co., St. Louis, MO, USA) was used for cell treatment. Prior to treatment, stock solutions of each drug were prepared as stated below. BPA was dissolved in pure ethanol (EA043-2.5 L, Chem supply, South Australia, Australia) and the final concentration of the stock solutions was 0.0435 g/mL. Cells in T-175 flasks were randomly assigned to receive treatment with BPA at a dosage of 100 μg/L, 50 μg/L, 25 μg/L or 0 μg/L (vehicle). There was also a no treatment (no vehicle added) flask.
Cell treatment, protein assay, SDS-Page, and western blotting methods can be found in the Supplementary Methods.
Two colonies of mice, maintained at the Florey Institute, were used in this study. The Aromatase knockout (ArKO) mouse model and the Aromatase-enhanced fluorescent green protein (Cyp19-EGFP) transgenic mouse model. Animals were monitored daily except for weekends. If animals showed general clinical signs, an animal technician or a vet was consulted for advice and euthanasia performed as required.
Mice were maintained under specific pathogen-free (SPF) conditions on a 12 h day/night cycle, with ad libitum water and soybean-free food (catalog number SF06-053, Glen Forrest Stockfeeders, Glen Forrest, Western Australia, Australia). Facial tissues were provided for nesting material, and no other environmental enrichment was provided. The room temperature ranged from 18 °C-23 °C and the humidity ranged from 45%-55%.
The sex of mice was determined by SRY genotyping if fetal, otherwise sex was determined by examining the anogenital region around PND9 and again at weaning. Sex was confirmed by inspecting the gonads during dissection. Cyp19 -EGFP mice were toe and tail clipped for identification and genotyping at PND9, ArKO mice were ear notched and tailed clipped at two weeks of age. The oligonucleotide sequences (custom oligos, Geneworks, Australia) for ArKO, GFP and SRY genotyping can be found in Supplementary Data File 2 .
The ArKO mouse is a transgenic model having a disruption of the Cyp19a1 gene. Exon IX of the Cyp19a1 gene was replaced with a neomycin-resistant cassette 41 . Homozygous Knockout (KO) and wild-type (WT) offspring were bred by mating heterozygous (het) ArKO parents and then PCR genotyped. ArKO mice were backcrossed onto a C57BL/6 J background strain, >10 generations (obtained from Animal Resources Centre, Western Australia) and the colony maintained at the Florey institute.
The Cyp19-EGFP mouse model (backcrossed onto the FVBN background strain >10 generations, obtained from Animal Resources Centre, Western Australia) is a transgenic model having a bacterial artificial chromosome containing the full length of the Cyp19a1 gene with an Enhanced Green Fluorescent Protein (EGFP) gene inserted upstream of the ATG start codon 11 . Thus, EGFP expression is an endogenous marker for Cyp19a1 expression. This allows for the visualization and subsequent localization of EGFP as the marker for aromatase without the use of potentially nonspecific aromatase antibodies 11 . We have previously characterized this transgenic model and its brain expression of EGFP 11 . Based on our characterization studies, this transgenic model does not have phenotypes that are significantly different to wildtype mice.
Mice were allocated into three groups: (1) WT mice receiving a sham implantation; (2) ArKO mice receiving a sham implantation; and (3) ArKO mice undergoing implantation with a 17β-estradiol pellet (sourced from Innovative Research America). This estradiol pellet was designed to release 0.2 mg of 17β-estradiol steadily over a period of 6 weeks. A corresponding sham pellet, identical in size but devoid of E2, was implanted in the control groups.
The implantation procedure was carried out on postnatal day 5. For anesthesia, mice were exposed to 2% isoflurane (IsoFlo, Abbott Laboratories, VIC, Australia) within an induction chamber. The efficacy of anesthesia was confirmed by the lack of response to foot-pinch stimuli. During the surgical procedure, mice were maintained on a heated pad to regulate body temperature. A small, 5 mm incision was made in the dorsal region for the subcutaneous insertion of the pellet, preceded by an injection of Bupivacane in the same area. Following the implantation, the incision was carefully sutured. Post-surgery, mice were placed in a thermal cage (Therma-cage, Manchester, UK) for recovery and monitoring until they regained consciousness and could be returned to their respective litters. Any mice exhibiting complications such as opened stitches were excluded from the study.
Plugged FVBN dams were randomly assigned, blocking by weight gain at E9.5 and litter/cage where applicable, to receive daily scruff subcutaneous injections (24 G x 1”, Terumo, Somerset, New Jersey, USA) of BPA (239658-50 G, Sigma-Aldrich Co., St Louis, MO, USA) in ethanol and peanut oil (Coles, Victoria, Australia), either between E0.5-E9.5, E10.5-E14.5 or E15.5-birth at a dosage of 50 μg/kg (deemed as the safe consumption level by the Food and Drug Administration, FDA) 37 or 0 μg/kg (vehicle) of maternal body weight. The injection volume was 1.68 μL/g bodyweight. Mice were weighed directly before each injection. BPA and vehicle exposed litters did not differ in litter size (Supplementary Fig. 14 ).
Cyp19 -EGFP or ArKO mice were randomly assigned by blocking on sex and litter to receive daily intraperitoneal injections (31 G x 1”, Terumo, Somerset, New Jersey, USA) of 500 μg/kg 10HDA (Matreya, USA) in saline or vehicle saline for 21 consecutive days. The injection volume was 2.1 μL/g bodyweight. Mice were weighed directly before each injection.
Plugged FVBN dams were exposed to jelly at E9.5. The jelly contained 7.5% Cottee’s Raspberry Cordial (Coles, Victoria, Australia) and 1% bacteriological agar (Oxoid, Australia) in milli-Q water. The pH was increased to between 6.5-7.5 with a pallet of NaOH to allow the jelly to set. Dams were then randomly assigned, blocking by weight gain at E9.5 and litter/cage where applicable, to receive a daily dose of jelly, which contained either ethanol or BPA dissolved in ethanol, at a dosage of 50 μg/kg or 0 μg/kg (vehicle) of maternal body weight. Dams received doses between E10.5-E14.5, and only dams that were observed to have consumed all the jelly each day were included in the study.
Three-chamber social interaction test.
The three-chamber social interaction test is extensively used to investigate juvenile and adult social interaction deficits, including in sociability 85 , 86 . BPA exposed pups were habituated in the experimental room on P21, directly after weaning. Following a two-to-three-day habituation, testing was conducted from P24 to P27, as only a maximum of ten mice could be tested during the light phase per day. ArKO mice treated with estrogen or sham pallet began habituation at PND28-29, with testing at PND31-33. Both male and female mice were tested. Testing was performed in a dedicated room for mouse behavior studies; no other animals were present in the room at the time of acclimatization and testing. The temperature of the room was maintained at approximately 21 °C.
The test apparatus, a three-chambered clear plexiglass, measuring 42 cm x 39 cm x 11 cm, had two partitions creating a left, right (blue zones), and center chamber (green zone) in which mice could freely roam via two 4 cm x 5 cm openings in the partitions (Supplementary Fig. 15 ). The two side chambers contained two empty wire cages. A 1 cm wide zone in front of each wire cage was defined as the interaction zone (yellow zones). The chamber was set on a black table for white mice, and on a white covering for black mice to aide tracking.
Each test consisted of two consecutive 10-min trials, a habituation trial (T1) and a sociability trial (T2). T1 allowed the test mouse to habituate, and any bias for either empty interaction zone was noted. For T2, a C57BL/6 J novel stranger mouse matched with the test mouse for age and gender was introduced into the cage on the opposite side to which the test mouse demonstrated an interaction zone bias. Thus, any evidence of sociability is bolstered as interaction zone bias would have to be overcome.
For each trial, the test mouse began in the center chamber, and its activity, both body center point, and nose point was tracked and quantified by TopScan Lite (Clever Sys Inc., Reston VA, USA). In this study, the key measure extracted was the average duration of the nose point in each interaction zone.
Social approach and sociability were analyzed. We define social approach as the time the test mouse’s nose point was tracked in the stranger cage interaction zone. Sociability is the higher proportion of time the test mouse to spends with the nose point in the stranger cage interaction zone compared to the empty cage interaction zone.
Details on the Y-maze and grooming methods can be found in the Supplementary Methods.
Mice had not undergone any behavioral testing. For Golgi staining and analysis, Wild Type (WT) and Knockout (ArKO) and Cyp19 -EGFP littermate males (aged P65-P70); one mouse from n = 3 litters for each genotype) were deeply anesthetized with isopentane rapidly decapitated and fresh whole brain tissues were collected. Brains were first washed with milli-Q water to remove excess blood and then directly placed in the solution obtained from the FD Rapid GolgiStain TM Kit (FD Neuro-Technologies, Inc., MD, USA). Brains were stored at room temperature in the dark and the solutions were replaced after 24 hours, and the tissues were kept in the solution for two weeks. After two weeks, tissues were transferred into solution C for a minimum of 48 hours at room temperature. For sectioning, brains were frozen rapidly by dipping into isopentane pre-cooled with dry ice, and 100 µm thick coronal sections were cut at -22 °C and mounted on 1% gelatin-coated slides. The sections were then air dried in the dark at room temperature. When sections were completely dry, slides were further processed and rinsed with distilled water and placed in the solution provided in the kit for 10 min and washed again with distilled water followed by dehydration for 5 min each in 50%, 75%, 95%, and 100 % ethanol. Sections were further processed in xylene and mounted with Permount.
Neuron tracing was conducted on the amygdala and somatosensory cortex of BPA-exposed mice (exposed ED10.5-14.4) and untreated ArKO mice. Neuron tracing in the amygdala was conducted in both male and female mice, and in the somatosensory cortex, only in male mice. Stained slides were coded to ensure that morphological analysis was conducted by an observer who was blind to the animals’ treatment. Morphological analysis followed a previously described protocol 87 with the following modifications: layer V pyramidal cells of the somatosensory cortex, which were fully impregnated and free of neighboring cells or cellular debris, were randomly selected for analysis (Supplementary Fig. 16 ). Golgi-stained coronal sections containing medial amygdala and somatosensory cortical area were visualized under Olympus BX51 microscope. Neuronal tracing was carried out with the help of Neurolucida and Neuroexplorer software (MicroBright Field Inc., Williston, USA). Up to three pyramidal cells in the MeA and four pyramidal cells in the somatosensory cortex per section over 3 sections (9 (MeA) and 12 (cortex) cells per animal respectively) were sampled 88 , 89 . For Sholl analysis 90 , concentric circles were placed at 10 µm intervals starting from the center of the cell body and the parameters i) total dendritic length (sums of the length of individual branches) of apical and basal dendrites of pyramidal cells and ii) number of spines (protrusions in direct contact with the primary dendrite) and their density (number of spines per 10 µm) were recorded.
Neuron selection criteria: Neurons were selected based on the following criteria. They had to be fully stained, and the cell body had to be in the middle third of the section thickness. The dendrites of the neuron had to be unobscured by the other nearby neuron. Also, neurons had to possess tapering of the majority of the dendrites towards their ends. Representative images of neurons from vehicle and BPA-exposed adult mice can be found in Supplementary Fig. 17 .
Stranger exposure paradigm procedure.
Cyp19-EGFP mice of both sexes as well as male ArKO mice, together with male WT littermates were utilized in this study. Mice had not undergone any other behavior testing. All test mice were acclimatized to the testing room in individual cages for three nights prior to testing. All mice were age P24 on the day of testing, which was performed between 10 am-2 pm. Testing was performed in a dedicated room for mouse behavior studies and no other animals were present in the room at the time of acclimatization and testing. The temperature of the room was maintained at approximately 21 °C.
On the day of testing, each mouse cage containing the isolated test mouse was placed on a stage (a trolley). The lid containing food and water was removed and immediately following, a sex-/age-matched C57Bl/6 J stranger mouse or a novel object (new 1 mL syringe) was placed into the cage and a clean, empty lid was placed on the top. New gloves were used to handle each syringe to avoid transferring another mouse’s olfactory signature to it. The 10 min trial began as soon as the cage lid was shut. After 10 min had elapsed, the stranger or the novel object was removed, the test mouse with home cage was returned to its original location with the original cage lid with food and water for 2 hours prior to perfusion. Once it was established that there was a difference in c-fos expression between stranger exposure and novel object exposure in the medial amygdala, BPA and vehicle exposed (ED10.5-14.5) Cyp19-EGFP mice as well as estrogen and sham pallet treated ArKO and WT mice were exposed to an age and sex matched stranger as described above. C-fos expression was quantified in male mice only.
Histology and stereological analysis methods can be found in the Supplementary Methods.
Neuron count brain collection, staining, brain region delineation and stereology can be found in the Supplementary Methods.
Microelectrode array electrophysiology.
Male mice aged 8 weeks weighing between 15 and 20 g were used for this study. They had not undergone any behavioral testing prior to electrophysiology. We studied synaptic activity parameters such as the Input/Output (I/O) curve. Stimulation of the glutamatergic synapses terminate in the basolateral amygdala (BLA) and the basomedial amygdala (BMA), which were integrated with multiple inputs that compute to produce an output (field excitatory postsynaptic potential, fEPSP). I/O curve serves as an index of synaptic excitability of large neuronal populations. Mice were anesthetized with isoflurane (IsoFlo TM , Abbott Laboratories, Victoria, Australia) and decapitated. The whole brains were quickly removed and placed in ice-cold, oxygenated (95% O 2 , 5% CO 2 ) cutting solution (composition in mmol/L: 206 sucrose, 3 KCl, 0.5 CaCl 2 , 6 MgCl 2 -H 2 O, 1.25 NaH 2 PO 4 , 25 NaHCO 3, and 10.6 D-glucose). Coronal brain amygdala slices (300 µm) were prepared with a VT 1200 S tissue slicer (Leica) and quickly transferred to 34 o C carbogen bubbled artificial CSF (aCSF) (composition in mmol/L: 126 NaCl, 2.5 KCl, 2.4 CaCl 2 , 1.36 MgCl 2 -H 2 O, 1.25 NaH 2 PO 4 , 25 NaHCO 3, and 10 D-glucose) for 30 min. After further recovery of 1 h equilibrium in oxygenated aCSF at room temperature, the slices were transferred to a submission recording chamber, an MEA chip with 60 electrodes spaced 200 μm apart (60 MEA 200/30 iR-Ti: MCS GnbH, Reutlingen, Germany). The slice was immobilized with a harp grid (ALA Scientific Instruments, New York, USA) and was continuously perfused with carbogenated aCSF (3 mL/min at 32 °C). fEPSPs produced in BLA and BMA were by stimulation of a randomly chosen electrode surrounding the target area with a biphasic voltage waveform (100 μs) at intermediated voltage intensity. The electrode could only be chosen if it produced a fair number of fEPSPs in the surrounding recording electrodes. The width of the EPSP wave ranged from 20 to 30 ms was selected. We chose slices where BLA and BMA were greatly represented according to Allen Mouse Brain Atlas 91 . Care was taken to choose the stimulating electrode in the same region from one slice to the other. The peak-to-peak amplitude of fEPSP in BLA and BMA was recorded by a program of LTP-Director and analyzed using LTP-Analyzer (MCS GnbH, Reutlingen, Germany).
Electrocorticogram recordings.
Male mice aged 8 weeks were used for this study. Mice had not undergone any behavioral testing prior to ECoG recording. For ECoG, surgeries were performed as previously described 92 . Mice were anesthetized with 1–3% isoflurane and two epidural silver ‘ball’ electrodes implanted on each hemisphere of the skull. Electrodes were placed 3 mm lateral of the midline and 0.5 mm, caudal from bregma. A ground electrode was placed 2.5 mm rostral from bregma and 0.5 mm lateral from the midline. Mice were allowed to recover for at least 48 hours after surgery. ECoGs were continuously recorded in freely moving mice for a 4–6-hour period during daylight hours following a standard 30-min habituation period. Signals were band-pass filtered at 0.1 to 40 Hz and sampled at 1 kHz using the Pinnacle EEG/EMG tethered recording system (Pinnacle Technology Inc, KS). Power spectrums were calculated using Hann window with a resolution of 1 Hz using Sirenia Pro analysis software (Pinnacle Technology Inc) on stable 30-min periods of ECoG recordings.
Neuroprotective effect of 10hda against injury induced by bpa on embryonic mouse cortical neurons.
Primary cortical neurons were obtained from male Cyp19 -EGFP mouse embryos at gestational day 15.5. Embryos were genotyped for SRY to determine sex, and only male embryos were used. Cells were seeded in 24-well plates containing 12 mm glass coverslips, coated with 100 µg/mL poly D-lysine to a density of ~0.45 x 10 6 cells/well and incubated in a humidified CO 2 incubator (5% CO 2 , 37 °C). Cells were pre-treated with vehicle (DMSO), 1 mM 10HDA (Matreya, PA,USA), 25 nM BPA and 1 mM 10HDA with 25 nM BPA. For each group, 10 neurons were measured, and the experiments were duplicated. Each replicate was from a separate culture.
Cells were fixed in 4% paraformaldehyde and stained with mouse anti-βIII tubulin monoclonal primary antibody (1:1000; cat #ab41489, Abcam, United Kingdom) and goat anti-mouse secondary antibody, Alexa Fluor 488, (1:2000; cat#A11017; Invitrogen, USA) to label neuronal cells. Aromatase was stained using Rabbit anti-aromatse Antibody (1:2000 cat# A7981; Sigma Aldrich, St. Louis, MO, USA) and donkey anti-rabbit Alexa594 (1:2000; cat# A-21207; Invitrogen, USA). Cell nucleus was stained with Hoechst 33258 solution (Sigma 94403 (2 µg/mL)). Images were captured using an Olympus IX51 microscope (X40 objective). Neurites were quantified using Neurolucida and Neuroexplorer software (MicroBright Field Inc, Williston, USA) as described in the neuron tracing section.
Total RNA was extracted using PARIS kit (cat#: AM1921, Invitrogen™PARIS™ Kit) according to the protocol supplied by the manufacturer. cDNA libraries were generated using the SureSelect.
Strand-Specific RNA Library Prep for Illumina Multiplexed Sequencing kit (Agilent Technologies, CA, USA), according to manufacturer’s instructions.
Pregnant Cyp19-EGFP dams were injected subcutaneously with BPA or vehicle ED10.5-14.5 as described in previous section, and culled on ED15.5 by isoflurane overdosed. Fetuses were harvested and placed in chilled PBS. Embryo brain cortical tissue was dissected from fetuses, snap frozen in liquid nitrogen and stored in −80°C until RNA extraction. The sex of fetuses was determined by visual assessment of the gonads and Sry (a male-specific gene) genotyping. Each RNA seq run, 6 cDNA libraries (derived from total RNA samples with 3 biological samples per group), were analyzed by MidSeq Nano run, 50 bp, Single end read on the Illumina platform. Because of undetectable levels of Cyp19a1 RNA in the fetal brain, Cyp19a1 RNA levels were not included. This is consistent in that Aromatase+ cells represent <0.05% of neurons in the adult mouse brain 93 . Subsequent in vivo transcriptomic analyses were completed in males only. Read quality was then assessed with FastQC. The sequence reads were then aligned against the Mus musculus genome (Build version GRCm38). The Tophat aligner (v2.0.14) was used to map reads to the genomic sequences. Sequencing data were then summarized into reads per transcript using Feature counts 94 . The transcripts were assembled with the StringTie tool v1.2.4 using the reads alignment with Mus_musculus.GRCm38 and reference annotation based assembly option (RABT) using the Gencode gene models for the mouse GRCm38/mm10 genome build. Normalisation and statistical analysis on the count data were executed using EdgeR (version edgeR_3.14.2 in R studio, R version 3.14.2). The data were scaled using trimmed mean of M-values (TMM) 95 and differentially expressed genes between all treatment group (Benjamini–Hochberg false discovery rate >0.1). Differentially expressed genes (DEGs) were identified by comparing mice exposed to 50 μg/kg/day BPA with those exposed to the vehicle.
Primary brain cortical neurons were obtained from C57BL/6 mouse embryos at GD 15.5. Neuronal cell cultures were treated with vehicle (DMSO) or 1 mM 10HDA (Matreya, PA,USA) as described above. The libraries were sequenced with 50 bp single end reads using an Illumina Hiseq and read quality assessed using FastQC. Untrimmed reads were aligned to mouse mm10 genome using Subjunc aligner (version 1.4.4) within the Subread package 96 . Sequencing data were then summarized into reads per transcript using Feature counts 94 and the Gencode gene models for the mouse GRCm38/mm10 genome build (August 2014 freeze) 97 . Normalisation and statistical analysis on the count data were executed using EdgeR (version edgeR_3.4.2 in R studio, R version 3.0.2) 98 after removing features with less than 10 counts per million for at least 3 of the samples. The data were scaled using trimmed mean of M-values (TMM) 95 and differentially expressed genes between all treatment group (Benjamini–Hochberg false discovery rate >0.1). Annotation was added using the ensemble mouse gene annotation added using bioMart package 99 . Differentially expressed genes were identified by comparing cells exposed to 10HDA with those exposed to the vehicle.
The BPA and 10HDA differential expression data for enriched pathways were analyzed using Ingenuity (QIAGEN) and tested against the Canonical Pathway Library, Brain Diseases and Functions Library and the Brain Disorders pathway libraries. We included the top 8 Canonical pathways. Then we included only pathways which were p < 0.05 in both the BPA and the 10HDA data for the Brain Diseases and Functions pathway libraries, and included all P -values for the Brain Disorders pathway library ( p > 0.05 are in gray).
An additional analysis of the gene expression data was performed using the c lusterProfileR R package 100 , which provides a range of statistical tests to detect pathways from a query gene set. The test used here was Gene Set Enrichment Analysis (GSEA) 101 , and the genes were tested against the Gene Ontology pathway database (specifically, GO: Biological Process) 102 .
The DockThor molecular docking platform 103 was used to assess binding affinities between estrogen receptor beta (encoded by ESR2 gene) and the ligands 17-beta estradiol (E2; the native ligand), BPA, and 10HDA.DockThor takes as input 3D molecular structures for a putative receptor-ligand pair and employs a genetic-algorithm-based optimization strategy to identify optimal binding position within a specified search region. The crystal structures of estrogen receptor alpha (Erα) and beta (Erβ) were sourced from the Protein Data Bank (PDB) with respective PDB IDs: Erα - 5KRI and Erβ - 1YYE. For each ligand, the search grid was restricted to the known estrogen receptor beta ligand-binding domain, centered at x = 30, y = 35, z = 40, with total grid size of x = 25, y = 28 and z = 22. Default settings were used for the optimization procedure.
Researchers were blind to treatment during the conduct of the experiment and the outcome assessment but not during statistical analysis.
Mean, standard deviation (SD), and standard error of the mean (SEM) were calculated with GraphPad Prism version 9.4 (GraphPad Software).
Data were tested for equal variances and normality using the Shapiro Wilk test. As electrophysiology, Golgi staining and primary cell culture experiments utilized several data points per animal, observations were not independent, and this non-independence was accounted for in our analyses. We used generalized estimating equations (GEEs) in R version 4.1.2. in a marginal modeling approach that estimates population-averaged effects while treating the covariance structure as a nuisance. We specified the covariance structure as exchangeable (that is, assumed equal correlation between pairs of measurements on the same animal). Given the small number of clusters (i.e. animals), bootstrapped standard errors were estimated using 200 repeats to maintain a conservative type 1 error rate 104 . An interaction term was added to the amygdala Golgi staining study to assess a sex * genotype or sex * BPA exposure interaction.
Where data were normally distributed, parametric tests were conducted. For more than two groups, a one-way ANOVA was conducted with Holm-Sidak post hoc FDR correction, with alpha set to 0.05. Otherwise, unpaired two-tailed Student’s t -tests were used to compare two variables. In cases where normality was not assumed, a Mann-Whitney (comparing two groups) or Kruskal-Wallis with Dunn’s post hoc (comparing three or more groups) was used. In the case of the three-chamber data, a two-way mixed ANOVA was used to assess group x cage side interaction (stranger cage interaction zone vs empty cage interaction zone) with post hoc testing adjusted by the Holm-Sidak method.
Comparisons made are indicated on the Figure legends, and p -values < 0.05 were considered significant. All tests were two-sided (two-tailed) where applicable.
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
The BIS data including all data used in this paper are available under restricted access for participant privacy. Access can be obtained by request through the BIS Steering Committee by contacting Anne-Louise Ponsonby, The Florey institute of Neuroscience and Mental Health, [email protected]. Requests to access cohort data will be responded to within two weeks. Requests are then considered on scientific and ethical grounds and, if approved, provided under collaborative research agreements. Deidentified cohort data can be provided in Stata or CSV format. Additional project information, including cohort data description and access procedure, is available at the project’s website https://www.barwoninfantstudy.org.au . Source data underlying Figs. 1 – 6 , 8 – 10 and Supplementary Figs. 2 , 3 – 7 , 14 have been provided as a Source Data file with this paper. The RNAseq data discussed in this publication have been deposited in NCBI’s Gene Expression Omnibus 105 and are accessible through GEO Series accession numbers; fetal brain expression with and without prenatal BPA exposure, GSE266401 and primary cortical culture treated with and without 10HDA, GSE266400 Source data are provided with this paper.
American Psychiatric Association aAPAD-TF. Diagnostic and Statistical Manual of Mental Disorders: DSM-5 (American Psychiatric Association, 2013).
Maenner, M. et al. Prevalence of autism spectrum disorder among children aged 8 years—autism and developmental disabilities monitoring network, 11 sites, United States, 2016. MMWR Surveill. Summ. 69 , 1–12 (2020).
Article PubMed PubMed Central Google Scholar
Li, Q. et al. Prevalence of autism spectrum disorder among children and adolescents in the United States From 2019 to 2020. JAMA Pediatrics 176 , 943–945 (2022).
Atladottir, H. O. et al. The increasing prevalence of reported diagnoses of childhood psychiatric disorders: a descriptive multinational comparison. Eur. Child Adolesc. Psychiatry 24 , 173–183 (2015).
Article PubMed Google Scholar
Lai, M.-C. & Lombardo, M. V. Baron-Cohen S. autism. Lancet 383 , 896–910 (2014).
Boon, W. C., Chow, J. D. & Simpson, E. R. The multiple roles of estrogens and the enzyme aromatase. Prog. Brain Res. 181 , 209–232 (2010).
Article CAS PubMed Google Scholar
Harada, N. & Honda, S.-i Analysis of spatiotemporal regulation of aromatase in the brain using transgenic mice. J. Steroid Biochem Mol. Biol. 95 , 49–55 (2005).
Tan, W., Zhu, Z., Ye, L. & Leung, L. K. Methylation dictates PI.f-specific CYP19 transcription in human glial cells. Mol. Cell Endocrinol. 452 , 131–137 (2017).
Roselli, C. E., Liu, M. & Hurn, P. D. Brain aromatization: classic roles and new perspectives. Semin Reprod. Med. 27 , 207–217 (2009).
Article CAS PubMed PubMed Central Google Scholar
Ruiz-Palmero, I. et al. Oestradiol synthesized by female neurons generates sex differences in neuritogenesis. Sci. Rep. 6 , 31891 (2016).
Article ADS CAS PubMed PubMed Central Google Scholar
Stanić, D. et al. Characterization of aromatase expression in the adult male and female mouse brain. I. Coexistence with oestrogen receptors α and β, and androgen receptors. PLoS One 9 , e90451 (2014).
Article ADS PubMed PubMed Central Google Scholar
Takahashi, K. et al. Association between aromatase in human brains and personality traits. Sci. Rep. 8 , 16841 (2018).
Baron-Cohen, S. The extreme male brain theory of autism. Trends Cogn. Sci. 6 , 248–254 (2002).
Crider, A., Thakkar, R., Ahmed, A. O. & Pillai, A. Dysregulation of estrogen receptor beta (ERβ), aromatase (CYP19A1), and ER co-activators in the middle frontal gyrus of autism spectrum disorder subjects. Mol. Autism 5 , 46–46 (2014).
Sarachana, T. & Hu, V. W. Genome-wide identification of transcriptional targets of RORA reveals direct regulation of multiple genes associated with autism spectrum disorder. Mol. Autism 4 , 14 (2013).
Santangeli, S. et al. Effects of BPA on female reproductive function: the involvement of epigenetic mechanism. Gen. Comp. Endocrinol. 245 , 122–126 (2017).
Ali, A. A. et al. Developmental vitamin D deficiency increases foetal exposure to testosterone. Mol. Autism 11 , 96 (2020).
Andrade, A. J., Grande, S. W., Talsness, C. E., Grote, K. & Chahoud, I. A dose-response study following in utero and lactational exposure to di-(2-ethylhexyl)-phthalate (DEHP): non-monotonic dose-response and low dose effects on rat brain aromatase activity. Toxicology 227 , 185–192 (2006).
Moosa, A., Shu, H., Sarachana, T. & Hu, V. W. Are endocrine disrupting compounds environmental risk factors for autism spectrum disorder? Hormones Behav. 101 , 13–21 (2018).
Article CAS Google Scholar
Ejaredar, M., Lee, Y., Roberts, D. J., Sauve, R. & Dewey, D. Bisphenol A exposure and children’s behavior: a systematic review. J. Expo. Sci. Environ. Epidemiol. 27 , 175–183 (2017).
Chen, D. et al. Bisphenol analogues other than BPA: environmental occurrence, human exposure, and toxicity—a review. Environ. Sci. Technol. 50 , 5438–5453 (2016).
Article ADS CAS PubMed Google Scholar
Hansen, J. B. et al. Prenatal exposure to bisphenol A and autistic- and ADHD-related symptoms in children aged 2 and 5 years from the Odense Child Cohort. Environ. Health 20 , 24 (2021).
Lim, Y. H. et al. Prenatal and postnatal bisphenol A exposure and social impairment in 4-year-old children. Environ. Health 16 , 79 (2017).
Ferri, S. L., Abel, T. & Brodkin, E. S. Sex differences in autism spectrum disorder: a review. Curr. Psychiatry Rep. 20 , 9 (2018).
Suzuki, K. M. et al. Estrogenic activities of fatty acids and a sterol isolated from royal jelly. Evid. Based Complement. Alternat. Med. 5 , 295–302 (2008).
Pirgon, O., Atar, M., Çiriş, M. & Sever, M. Effects of royal jelly supplementation on growth plate zones and longitudinal growth in young rats. Mellifera 19 , 1–13 (2019).
Google Scholar
Hattori, N., Nomoto, H., Fukumitsu, H., Mishima, S. & Furukawa, S. Royal jelly and its unique fatty acid, 10-hydroxy-trans-2-decenoic acid, promote neurogenesis by neural stem/progenitor cells in vitro. Biomed. Res. 28 , 261–266 (2007).
Vuillermin, P. et al. Cohort profile:the Barwon Infant Study. Int J. Epidemiol. 44 , 1148–1160 (2015).
Achenbach T. M. & Rescorla L. A. Manual for the ASEBA Preschool Forms & Profiles (University of Vermont, Research Center for Children, Youth, & Families, 2000).
Pham, C. et al. Early life environmental factors associated with autism spectrum disorder symptoms in children at age 2 years: a birth cohort study. Autism.: Int. J. Res. Pract. 26 , 1864–1881 (2022).
Article Google Scholar
Kidokoro, K. et al. Association between CYP19A1 polymorphisms and sex hormones in postmenopausal Japanese women. J. Hum. Genet. 54 , 78–85 (2009).
Braun, P. R. et al. Genome-wide DNA methylation comparison between live human brain and peripheral tissues within individuals. Transl. Psychiatry 9 , 47 (2019).
Kundakovic, M. et al. DNA methylation of BDNF as a biomarker of early-life adversity. Proc. Natl Acad. Sci. USA 112 , 6807–6813 (2015).
Panickar, K. S., Guan, G., King, M. A., Rajakumar, G. & Simpkins, J. W. 17beta-estradiol attenuates CREB decline in the rat hippocampus following seizure. J. Neurobiol. 33 , 961–967 (1997).
Solum, D. T. & Handa, R. J. Estrogen regulates the development of brain-derived neurotrophic factor mRNA and protein in the rat hippocampus. J. Neurosci. 22 , 2650–2659 (2002).
EPA USEPAU. Bisphenol A; CASRN 80-05-7. Integrated Risk Information System (IRIS) Chemical Assessment Summary (National Center for Environmental Assessment, 1988).
U.S. Food & Drug Administration. Bisphenol A (BPA): Use in Food Contact Application (U.S. Food & Drug Administration, 2023). [cited 2024 Jul]. Available from: https://www.fda.gov/food/food-additives-petitions/bisphenol-bpa-use-food-contact-application .
Authority, E. F. S. Opinion of the Scientific Panel on food additives, flavourings, processing aids and materials in contact with food (AFC) related to 2, 2‐BIS (4‐HYDROXYPHENYL) PROPANE. EFSA J. 5 , 428 (2007).
Semple, B. D., Dixit, S., Shultz, S. R., Boon, W. C. & O’Brien, T. J. Sex-dependent changes in neuronal morphology and psychosocial behaviors after pediatric brain injury. Behav. Brain Res. 319 , 48–62 (2017).
Soma, M. et al. Development of the mouse amygdala as revealed by enhanced green fluorescent protein gene transfer by means of in utero electroporation. J. Comp. Neurol. 513 , 113–128 (2009).
Fisher, C. R., Graves, K. H., Parlow, A. F. & Simpson, E. R. Characterization of mice deficient in aromatase (ArKO) because of targeted disruption of the cyp19 gene. Proc. Natl Acad. Sci. USA 95 , 6965–6970 (1998).
Hill, R. A. et al. Estrogen deficient male mice develop compulsive behavior. Biol. Psychiatry 61 , 359–366 (2007).
Avino, T. A. et al. Neuron numbers increase in the human amygdala from birth to adulthood, but not in autism. Proc. Natl Acad. Sci. USA 115 , 3710–3715 (2018).
Cai, S., Wang, X., Yang, F., Chen, D. & Huang, L. Differences in brain structural covariance network characteristics in children and adults with autism spectrum disorder. Autism Res. 14 , 265–275 (2021).
Keshavarzi, S., Sullivan, R. K. P., Ianno, D. J. & Sah, P. Functional properties and projections of neurons in the medial amygdala. J. Neurosci. 34 , 8699–8715 (2014).
Alia-Klein, N. et al. Human cognitive ability is modulated by aromatase availability in the brain in a sex-specific manner. Front Neurosci. 14 , 565668 (2020).
Biegon, A. et al. Relationship of estrogen synthesis capacity in the brain with obesity and self-control in men and women. Proc. Natl Acad. Sci. USA 117 , 22962–22966 (2020).
Sato, W. & Uono, S. The atypical social brain network in autism: advances in structural and functional MRI studies. Curr. Opin. Neurol. 32 , 617–621 (2019).
Herrington, J. D., Miller, J. S., Pandey, J. & Schultz, R. T. Anxiety and social deficits have distinct relationships with amygdala function in autism spectrum disorder. Soc. Cogn. Affect Neurosci. 11 , 907–914 (2016).
Krukoff, T. L. C-fos expression as a marker of functional activity in the brain. in Cell Neurobiology Techniques. Neuromethods . Vol. 33 213–230 (Humana Press, 1999).
Stoner, R. et al. Patches of disorganization in the neocortex of children with autism. N. Engl. J. Med. 370 , 1209–1219 (2014).
Anthoni, H. et al. The aromatase gene CYP19A1: several genetic and functional lines of evidence supporting a role in reading, speech and language. Behav. Genet. 42 , 509–527 (2012).
Wang, J. et al. Resting state EEG abnormalities in autism spectrum disorders. J. Neurodev. Disord. 5 , 24 (2013).
Matuszczak, E., Komarowska, M. D., Debek, W. & Hermanowicz, A. The impact of Bisphenol A on fertility, reproductive system, and development: a review of the literature. Int. J. Endocrinol. 2019 , 4068717 (2019).
Moutsatsou, P. et al. Fatty acids derived from royal jelly are modulators of estrogen receptor functions. PLoS ONE 5 , e15594 (2010).
Rubin B. S. Bisphenol A: An endocrine disruptor with widespread exposure and multiple effects. J. Steroid. Biochem. Mol. Biol . 127 , 27–34 (2011).
DeRosa, B. A. et al. Convergent pathways in idiopathic autism revealed by time course transcriptomic analysis of patient-derived neurons. Sci. Rep. 8 , 8423 (2018).
Andrews, S. V. et al. Case-control meta-analysis of blood DNA methylation and autism spectrum disorder. Mol. Autism 9 , 40 (2018).
Thongkorn, S. et al. Sex differences in the effects of prenatal bisphenol A exposure on autism-related genes and their relationships with the hippocampus functions. Sci. Rep. 11 , 1241 (2021).
VanderWeele, T. J. Explanation in causal inference: developments in mediation and interaction. Int J. Epidemiol. 45 , 1904–1908 (2016).
PubMed PubMed Central Google Scholar
Gerona, R., vom Saal, F. S. & Hunt, P. A. BPA: have flawed analytical techniques compromised risk assessments? Lancet Diabetes Endocrinol. 8 , 11–13 (2020).
Bolognesi, C. et al. Scientific Opinion on the risks to public health related to the presence of bisphenol A (BPA) in foodstuffs: Executive summary. efsa J. (2015).
Authority E. F. S. Re-evaluation of the Risks to Public Health Related to the Presence of Bisphenol A (BPA) in foodstuffs (European Food Safety Authority, 2021).
EFSA Panel on Food Contact Materials E, et al. Re-evaluation of the risks to public health related to the presence of bisphenol A (BPA) in foodstuffs. EFSA J. 21 , e06857 (2023).
Gao, H. et al. Cumulative risk assessment of phthalates associated with birth outcomes in pregnant Chinese women: a prospective cohort study. Environ. Pollut. 222 , 549–556 (2017).
Welch, C., Mulligan, K. & Does Bisphenol, A. Confer risk of neurodevelopmental disorders? what we have learned from developmental neurotoxicity studies in animal models. Int J. Mol. Sci. 23 , 2894 (2022).
Stein, T. P., Schluter, M. D., Steer, R. A., Guo, L. & Ming, X. Bisphenol A exposure in children with autism spectrum disorders. Autism Res. 8 , 272–283 (2015).
Rudel, R. A. et al. Food Packaging and Bisphenol A and Bis(2-Ethyhexyl) Phthalate Exposure: Findings from a Dietary Intervention. Environ. Health Perspect. 119 , 914–920 (2011).
Geueke, B. et al. Systematic evidence on migrating and extractable food contact chemicals: Most chemicals detected in food contact materials are not listed for use. Criti. Rev. Food Sci. Nutrition 63, 9425–9435 (2022).
Arbuckle, T. E. et al. Exposure to free and conjugated forms of bisphenol a and triclosan among pregnant women in the MIREC cohort. Environ. Health Perspect. 123 , 277–284 (2015).
Caporale, N. et al. From cohorts to molecules: adverse impacts of endocrine disrupting mixtures. Science 375 , eabe8244 (2022).
Hameed, R. A., Ahmed, E. K., Mahmoud, A. A. & Atef, A. A. G protein-coupled estrogen receptor (GPER) selective agonist G1 attenuates the neurobehavioral, molecular and biochemical alterations induced in a valproic acid rat model of autism. Life Sci. 328 , 121860 (2023).
Heffernan, A. L. et al. Harmonizing analytical chemistry and clinical epidemiology for human biomonitoring studies. A case-study of plastic product chemicals in urine. Chemosphere 238 , 124631 (2020).
Burugupalli, S. et al. Ontogeny of circulating lipid metabolism in pregnancy and early childhood—a longitudinal population study. Elife 11 , e72779 (2022).
McCarthy, S. et al. A reference panel of 64,976 haplotypes for genotype imputation. Nat. Genet. 48 , 1279–1283 (2016).
Ponsonby, A. L. et al. Prenatal phthalate exposure, oxidative stress-related genetic vulnerability and early life neurodevelopment: a birth cohort study. Neurotoxicology 80 , 20–28 (2020).
Wang, Y. et al. A methodological pipeline to generate an epigenetic marker of prenatal exposure to air pollution indicators. Epigenetics 17 , 32–40 (2022).
Perera, F. et al. Prenatal bisphenol a exposure and child behavior in an inner-city cohort. Environ. Health Perspect. 120 , 1190–1194 (2012).
Hunt, P. A., Vom Saal, F. S., Stahlhut, R. & Gerona, R. BPA and risk assessment—Authors’ reply. Lancet Diabetes Endocrinol. 8 , 271–272 (2020).
Burke James, F., Sussman Jeremy, B., Kent David, M. & Hayward Rodney, A. Three simple rules to ensure reasonably credible subgroup analyses. BMJ 351 , h5651 (2015).
Ponsonby, A.-L. Reflection on modern methods: building causal evidence within high-dimensional molecular epidemiological studies of moderate size. Int J. Epidemiol. 50 , 1016–1029 (2021).
Team RC. R: A Language and Environment for Statistical Computing (R Foundation for Statistical Computing, 2013).
StataCorp. Stata Statistical Software: Release 14 (StataCorp LP, 2015).
Thongkorn, S. et al. Investigation of autism-related transcription factors underlying sex differences in the effects of bisphenol A on transcriptome profiles and synaptogenesis in the offspring hippocampus. Biol. Sex. Differ. 14 , 8 (2023).
Nadler, J. J. et al. Automated apparatus for quantitation of social approach behaviors in mice. Genes, brain, Behav. 3 , 303–314 (2004).
Moy, S. S. et al. Sociability and preference for social novelty in five inbred strains: an approach to assess autistic-like behavior in mice. Genes, Brain, Behav. 3 , 287–302 (2004).
Champagne, D. L. et al. Maternal care and hippocampal plasticity: evidence for experience-dependent structural plasticity, altered synaptic functioning, and differential responsiveness to glucocorticoids and stress. J. Neurosci. 28 , 6037–6045 (2008).
Wang, I. T., Reyes, A. R. & Zhou, Z. Neuronal morphology in MeCP2 mouse models is intrinsically variable and depends on age, cell type, and Mecp2 mutation. Neurobiol. Dis. 58 , 3–12 (2013).
Kelly, E. A., Opanashuk, L. A. & Majewska, A. K. The effects of postnatal exposure to low-dose bisphenol-A on activity-dependent plasticity in the mouse sensory cortex. Front. Neuroanat. 8 , 117 (2014).
Sholl, D. A. Dendritic organization in the neurons of the visual and motor cortices of the cat. J. Anat. 87 , 387–406 (1953).
ADS CAS PubMed PubMed Central Google Scholar
Allen Institute for Brain Science AMBA. https://mouse.brain-map.org/static/atlas (2004).
Tan, H. O. et al. Reduced cortical inhibition in a mouse model of familial childhood absence epilepsy. Proc. Natl Acad. Sci. USA 104 , 17536–17541 (2007).
Unger, E. K. et al. Medial amygdalar aromatase neurons regulate aggression in both sexes. Cell Rep. 10 , 453–462 (2015).
Liao, Y., Smyth, G. K. & Shi, W. featureCounts: an efficient general purpose program for assigning sequence reads to genomic features. Bioinformatics 30 , 923–930 (2013).
Robinson, M. D. & Oshlack, A. A scaling normalization method for differential expression analysis of RNA-seq data. Genome Biol. 11 , R25 (2010).
Liao, Y., Smyth, G. K. & Shi, W. The Subread aligner: fast, accurate and scalable read mapping by seed-and-vote. Nucleic Acids Res. 41 , e108 (2013).
Harrow, J. et al. GENCODE: producing a reference annotation for ENCODE. Genome Biol. 7 , S4.1–9 (2006).
Robinson, M. D., McCarthy, D. J. & Smyth, G. K. edgeR: a Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics 26 , 139–140 (2010).
Durinck, S., Spellman, P. T., Birney, E. & Huber, W. Mapping identifiers for the integration of genomic datasets with the R/Bioconductor package biomaRt. Nat. Protoc. 4 , 1184–1191 (2009).
Wu, T. et al. clusterProfiler 4.0: a universal enrichment tool for interpreting omics data. Innovation 2 , 100141 (2021).
CAS PubMed PubMed Central Google Scholar
Subramanian, A. et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc. Natl Acad. Sci. USA 102 , 15545–15550 (2005).
The Gene Ontology Consortium. The Gene Ontology Resource: 20 years and still GOing strong. Nucleic Acids Res. 47 , D330–d338 (2019).
Santos, K. B., Guedes, I. A., Karl, A. L. M. & Dardenne, L. E. Highly flexible ligand docking: benchmarking of the DockThor Program on the LEADS-PEP protein–peptide data set. J. Chem. Inf. Model 60 , 667–683 (2020).
Leyrat, C., Morgan, K. E., Leurent, B. & Kahan, B. C. Cluster randomized trials with a small number of clusters: which analyses should be used? Int J. Epidemiol. 47 , 321–331 (2018).
Edgar, R., Domrachev, M. & Lash, A. E. Gene Expression Omnibus: NCBI gene expression and hybridization array data repository. Nucleic Acids Res. 30 , 207–210 (2002).
Xiao, X. et al. Bisphenol AP is anti-estrogenic and may cause adverse effects at low doses relevant to human exposure. Environ. Pollut. 242 , 1625–1632 (2018).
Download references
The authors thank the BIS participants for their generous contribution to this project. The authors also thank current and past cohort staff. The establishment work and infrastructure for the BIS was provided by the Murdoch Children’s Research Institute, Deakin University, and Barwon Health, supported by the Victorian Government’s Operational Infrastructure Program. We thank all the children and families participating in the study, and the BIS fieldwork team. We acknowledge Barwon Health, Murdoch Children’s Research Institute, and Deakin University for their support in the development of this research. We thank Dr Shanie Landen for statistical advice, and Alex Eisner for independent statistical review of the analyses in the manuscript. We thank Soumini Vijayay and Kristie Thompson for human BPA lab measurement and Dr Steve Cheung for assistance preparing the primary cortical culture. We thank Chitra Chandran, Georgia Cotter, Stephanie Glynn, Oliver Wood and Janxian Ng for manuscript preparation. Manuscript editor Julian Heng (Remotely Consulting, Australia) provided professional editing of this article. This multimodal project was supported by funding from the Minderoo Foundation. Funding was also provided by the National Health and Medical Research Council of Australia (NHMRC), the NHMRC-EU partnership grant for the ENDpoiNT consortium, the Australian Research Council, the Jack Brockhoff Foundation, the Shane O’Brien Memorial Asthma Foundation, the Our Women’s Our Children’s Fund Raising Committee Barwon Health, The Shepherd Foundation, the Rotary Club of Geelong, the Ilhan Food Allergy Foundation, GMHBA Limited, Vanguard Investments Australia Ltd, and the Percy Baxter Charitable Trust, Perpetual Trustees, Fred P Archer Fellowship; the Scobie Trust; Philip Bushell Foundation; Pierce Armstrong Foundation; The Canadian Institutes of Health Research; BioAutism, William and Vera Ellen Houston Memorial Trust Fund, Homer Hack Research Small Grants Scheme and the Medical Research Commercialisation Fund. This work was also supported by Ms. Loh Kia Hui. This project received funding from a NHMRC-EU partner grant with the European Union’s Horizon 2020 Research and Innovation Programme, under Grant Agreement number: 825759 (ENDpoiNTs project). This work was also supported by NHMRC Investigator Fellowships (GTN1175744 to D.B., APP1197234 to A.-L.P., and GRT1193840 to P.S.). The study sponsors were not involved in the collection, analysis, and interpretation of data; writing of the report; or the decision to submit the report for publication.
Nhi Thao Tran
Present address: The Ritchie Centre, Department of Obstetrics and Gynaecology, School of Clinical Sciences, Monash University, Clayton, Australia
These authors contributed equally: Christos Symeonides, Kristina Vacy.
These authors jointly supervised this work: Anne-Louise Ponsonby, Wah Chin Boon.
Minderoo Foundation, Perth, Australia
Murdoch Children’s Research Institute, Parkville, Australia
Christos Symeonides, Toby Mansell, Martin O’Hely, Boris Novakovic, David Burgner, Mimi L. K. Tang, Richard Saffery, Peter Vuillermin, Fiona Collier, Anne-Louise Ponsonby, Sarath Ranganathan, Lawrence Gray & Anne-Louise Ponsonby
Centre for Community Child Health, Royal Children’s Hospital, Parkville, Australia
Christos Symeonides, Sarath Ranganathan & Anne-Louise Ponsonby
The Florey Institute of Neuroscience and Mental Health, Parkville, Australia
Kristina Vacy, Sarah Thomson, Sam Tanner, Hui Kheng Chua, Shilpi Dixit, Jessalynn Chia, Nhi Thao Tran, Sang Eun Hwang, Feng Chen, Tae Hwan Kim, Christopher A. Reid, Anthony El-Bitar, Gabriel B. Bernasochi, Anne-Louise Ponsonby, Anne-Louise Ponsonby & Wah Chin Boon
School of Population and Global Health, The University of Melbourne, Parkville, Australia
Kristina Vacy
The Hudson Institute of Medical Research, Clayton, Australia
Hui Kheng Chua & Yann W. Yap
Department of Pediatrics, The University of Melbourne, Parkville, Australia
Toby Mansell & David Burgner
School of Medicine, Deakin University, Geelong, Australia
Martin O’Hely, Boris Novakovic, Chloe J. Love, Peter D. Sly, Peter Vuillermin, Fiona Collier & Lawrence Gray
Columbia Center for Children’s Environmental Health, Columbia University, New York, NY, USA
Julie B. Herbstman, Shuang Wang & Jia Guo
Department of Environmental Health Sciences, Columbia University, New York, NY, USA
Julie B. Herbstman
Department of Biostatistics, Columbia University, New York, NY, USA
Shuang Wang & Jia Guo
Department of Anatomy and Developmental Biology, Monash University, Clayton, Australia
Kara Britt & Vincent R. Harley
Breast Cancer Risk and Prevention Laboratory, Peter MacCallum Cancer Centre, Melbourne, Australia
Sir Peter MacCallum Department of Oncology, The University of Melbourne, Melbourne, Australia
Faculty Medicine, Dentistry & Health Sciences, University of Melbourne, Parkville, Australia
Gabriel B. Bernasochi, Lea M. Durham Delbridge, Mimi L. K. Tang, Leonard C. Harrison & Sarath Ranganathan
Sex Development Laboratory, Hudson Institute of Medical Research, Clayton, Australia
Vincent R. Harley & Yann W. Yap
Departments of Paediatrics and Community Health Sciences, The University of Calgary, Calgary, Canada
Deborah Dewey
Barwon Health, Geelong, Australia
Chloe J. Love, Peter Vuillermin, Fiona Collier & Lawrence Gray
Department of General Medicine, Royal Children’s Hospital, Parkville, Australia
David Burgner
Department of Pediatrics, Monash University, Clayton, Australia
Child Health Research Centre, The University of Queensland, Brisbane, Australia
Peter D. Sly
WHO Collaborating Centre for Children’s Health and Environment, Brisbane, Australia
Queensland Alliance for Environmental Health Sciences, The University of Queensland, Brisbane, Australia
Jochen F. Mueller
Monash Krongold Clinic, Faculty of Education, Monash University, Clayton, Australia
Nicole Rinehart
Centre for Developmental Psychiatry and Psychology, Monash University, Clayton, Australia
Bruce Tonge
School of BioSciences, Faculty of Science, The University of Melbourne, Parkville, Australia
Wah Chin Boon
Walter and Eliza Hall Institute, Parkville, Australia
Leonard C. Harrison
You can also search for this author in PubMed Google Scholar
Conceptualization—laboratory experiments: W.C.B., N.R., B.T., L.M.D.D. Laboratory experiments and analysis: W.C.B., K.V., S.D., H.K.C., J.C., F.C., CR, T.K., G.B.B., A.E.-B., S.E.H., N.T.T., K.B. Supervision of lab data collection: W.C.B., K.V., S.D., C.R. Laboratory statistical analysis: W.C.B., K.V., F.C., C.R., S.Th., V.H., Y.W.Y. Design and conduct of the Barwon Infant Study: C.S., A.-L.P., P.V., D.B., P.S., C.L., M.L.K.T., BIS Investigator Group. Design, conduct and analysis of the CCCEH-MN study: J.B.H., S.W., J.G. Design and conduct of BPA study measures in BIS: J.M., C.S., A.-.L.P. Design, conduct, and analysis of gene methylation studies: S.Ta., B.N., T.M., R.S., D.D., A.-L.P. Human studies statistical analysis: C.S., S.Th., A.-.L.P., S.Ta., K.V., M.O.H. Writing—reports and original draft: C.S., K.V., S.Th., S.Ta., A.-.L.P., W.C.B. Writing—editing: all authors. Results interpretation: all authors. Kara Britt did the laboratory experiment—estrogen pellet implantation.
Correspondence to Wah Chin Boon .
Competing interests.
W.C.B. is a co-inventor on ‘Methods of treating neurodevelopmental diseases and disorders’, USA Patent No. US9925163B2, Australian Patent No. 2015271652. This has been licensed to Meizon Innovation Holdings. A.-L.P. is a scientific advisor and W.C.B. is a board member of the Meizon Innovation Holdings. The remaining authors declare no competing interests.
Peer review information.
Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information, peer review file, description of additional supplementary files, supplementary movie 1, supplementary movie 2, supplementary dataset 1, supplementary dataset 2, reporting summary, source data, source data, rights and permissions.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .
Reprints and permissions
Cite this article.
Symeonides, C., Vacy, K., Thomson, S. et al. Male autism spectrum disorder is linked to brain aromatase disruption by prenatal BPA in multimodal investigations and 10HDA ameliorates the related mouse phenotype. Nat Commun 15 , 6367 (2024). https://doi.org/10.1038/s41467-024-48897-8
Download citation
Received : 01 December 2022
Accepted : 16 May 2024
Published : 07 August 2024
DOI : https://doi.org/10.1038/s41467-024-48897-8
Anyone you share the following link with will be able to read this content:
Sorry, a shareable link is not currently available for this article.
Provided by the Springer Nature SharedIt content-sharing initiative
By submitting a comment you agree to abide by our Terms and Community Guidelines . If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.
Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.
BMC Public Health volume 24 , Article number: 2212 ( 2024 ) Cite this article
119 Accesses
Metrics details
Increased consumption of ultra-processed foods (UPFs) which have additives such as artificial colours, flavours and are usually high in salt, sugar, fats and specific preservatives, are associated with diet-related non-communicable diseases (NCDs). In India, there are no standard criteria for identifying UPFs using a classification system based on extent and purpose of industrial processing. Scientific literature on dietary intake of foods among Indian consumers classifies foods as unhealthy based on presence of excessive amounts of specific nutrients which makes it difficult to distinguish UPFs from other commercially available processed foods.
A literature review followed by an online grocery retailer scan for food label reading was conducted to map the types of UPFs in Indian food market and scrutinize their ingredient list for the presence of ultra-processed ingredients. All UPFs identified were randomly listed and then grouped into categories, followed by saliency analysis to understand preferred UPFs by consumers. Indian UPF categories were then finalized to inform a UPF screener.
A lack of application of a uniform definition for UPFs in India was observed; hence descriptors such as junk-foods , fast-foods , ready-to-eat foods , instant-foods , processed-foods , packaged-foods , high-fat-sugar-and-salt foods were used for denoting UPFs. After initial scanning of such foods reported in literature based on standard definition of UPFs, an online grocery retailer scan of food labels for 375 brands (atleast 3 brands for each food item) confirmed 81 food items as UPFs. A range of packaged traditional recipes were also found to have UPF ingredients. Twenty three categories of UPFs were then developed and subjected to saliency analysis. Breads, chips and sugar-sweetened beverages (e.g. sodas and cold-drinks) were the most preferred UPFs while frozen ready-to-eat/cook foods (e.g. chicken nuggets and frozen kebabs) were least preferred.
India needs to systematically apply a food classification system and define Indian food categories based on the level of industrial processing. Mapping of UPFs is the first step towards development of a quick screener that would generate UPF consumption data to inform clear policy guidelines and regulations around UPFs and address their impact on NCDs.
Peer Review reports
Non-communicable diseases (NCDs) are one of the leading causes of premature morbidity and mortality resulting in over 7 out of 10 deaths worldwide [ 1 ]. Mortality due to NCDs has been on the rise in India, increasing from 37.9% of all deaths in 1990 to 61.8% in 2016 [ 2 , 3 ]. Overweight/obesity have been identified as a contributing factor [ 4 ]. The recent national-level data shows an increase of 25% in the prevalence of overweight and obesity among Indian men and women over 14–15 years and 3% among children under five years [ 5 , 6 ]. Due to their thin fat phenotype, Indian infants and children, who comprise almost one quarter of the total population, are predisposed to obesity [ 7 , 8 ]. These risk factors are further amplified by changing food environments and behavioural variables such as tobacco, alcohol, drug use and low physical activity [ 9 ]. Exposure to unhealthy food environments in genetically predisposed children, along with other behavioural risk factors, increases their risk of developing obesity and diet-related non-communicable diseases (DR-NCDs) in the long term [ 10 ].
The rapidly changing food environment is characterized by diets transitioning from minimally-processed staple foods (such as pulses and whole cereals) high in vitamins, minerals and fibre to refined, processed and ultra-processed foods (UPFs) [ 11 ]. The Indian population is exposed to a wide variety of UPFs which are hyper-palatable, packaged, convenient, affordable and have a long shelf life, such as sugar-sweetened beverages, chips, biscuits and bread, and ready-to-eat/ ready-to-cook (RTC) meals [ 12 ]. The sales data of UPFs in India demonstrates an exponential increase, from USD 0.9 billion in 2006 to USD 37.9 billion in 2019 [ 13 ]. This growth indicates a notable expansion of these food products in the market, coupled with widespread advertising efforts that specifically target vulnerable populations, including children and youth [ 14 , 15 , 16 , 17 , 18 ]. Consumer demand for UPFs has increased due to higher disposable incomes, nuclear families, single-member households, and less availability of time for housework [ 19 , 20 ]. UPFs have penetrated the rural boundaries of the country and are likely making their way into households of diverse geographic and socio-economic attributes [ 21 , 22 ].
The Nova food classification system categorizes foods based on the purpose and the level of processing and includes four categories: (i) unprocessed/ minimally processed foods, (ii) processed culinary ingredients, (iii) processed foods, and (iv) ultra-processed foods [ 23 , 24 , 25 , 26 ]. UPFs are a category of food that undergo a series of industrial processes like extrusion and moulding, and have presence of classes of additives whose function is to make the final product palatable or more appealing, such as flavours, flavour enhancers, colours, emulsifiers, thickeners, sweeteners, etc. Although not unique to UPF, they also include additives that prolong the product duration and protect original properties or prevent proliferation of microorganisms [ 23 , 24 , 25 , 26 ]. In addition to this, several of these products are high in saturated fats or trans-fats, added sugars, and salt and low in dietary fibre, various micronutrients and other bioactive compounds [ 27 , 28 , 29 , 30 , 31 , 32 , 33 , 34 , 35 ].
Overconsumption of UPFs has been associated with higher body mass index (BMI), obesity, type-2 diabetes, hypertension, cardiovascular diseases, and certain types of cancers [ 36 , 37 , 38 ]. Given the diversity in UPFs, there is a need to systematically map the range of UPFs accessed by the Indian population. This is an important first step to understanding their potential role in contributing to the NCD burden in India and in developing strategies to encourage the substitution of the most frequently consumed UPFs to healthier alternatives. Identifying specific UPF categories could also help inform the development of dietary assessment instruments like food frequency questionnaires (FFQ) and screeners.
The present study aimed at: (i) mapping the specific categories of UPFs accessed and consumed in India, (ii) assessing the ingredient composition of these products, (iii) ranking the UPFs by consumer preference, and (iv) developing a list of categories of UPFs commonly consumed in India. For this, a secondary review of available literature complemented by an online grocery retailer scan and a saliency analysis were conducted between April 2021 and February 2022 (Fig. 1 ).
Flow chart briefly explaining the 3 steps of methodology
Step 1. The literature review was conducted to map and identify the various types of UPFs accessed, consumed, preferred and/ or purchased (as reported behaviours) in India. This review included published cross-sectional and observational research studies that used surveys, focus group discussions and interviews to elicit reported behaviours across different population groups and regions in India. International and national survey reports on UPF food intake and purchase among Indian population were also included in the review. Articles for review were identified from two electronic databases (NLM NCBI and Google Scholar). To ensure the search captured the diversity of UPFs, search terms included proxy descriptors identified in Indian policy documents [ 39 , 40 , 41 , 42 , 43 , 44 ], including: junk food* , fast food* , modern food* , westernized food* , ultra-processed food* , UPF* , convenience food* , ready-to-eat food* , ready-to-eat snack* , ready-to-cook food* , instant food* , frozen food* , canned food* , tinned food* , processed food* , packaged food* , high fat , sugar and salt food* and HFSS*. The literature search and data extraction was conducted by two authors (MS and GK).
To be eligible for the review, studies needed to: (i) include UPFs or their proxy descriptors, with examples of products, (ii) be conducted in either rural and/or urban areas of India, (iii) be published in the English language, between January 2012 and December 2022. This time frame was chosen to capture the high growth in UPFs sales during this decade [ 45 , 46 ]. Review articles, and publications that did not define the food category studied or did not cite any examples of foods were excluded.
Data from eligible articles were extracted in MS-Excel to record key variables on UPFs or their proxy descriptors with examples, location of the study (national/specific state), geographical area (urban/rural), sample size, sampling method, study participants’ age (years) and dietary data collection tools such as 24 h dietary recalls, FFQ, interviews and structured questionnaires (Additional file 1 ). A free list of UPF foods and beverages identified from the reviewed studies, was developed.
Step 2. An online grocery retailer scan for extracting detailed information on the UPFs identified in Step 1, was also conducted. The objective was to review and scrutinize the ingredient list provided in the food labels and to confirm that the food item qualified as UPFs. For this online scan, three researchers (GK, IKB, MS) reviewed the online grocery websites of the largest grocery retailers in India - Big Basket, Grofers, and Amazon [ 47 ]. Individual foods and beverages from Step 1, were checked for their ingredient composition and the presence of additives. This activity was guided by the FAO document ‘Ultra-processed foods, diet quality, and health using the Nova classification system’ [ 23 , 24 , 25 , 26 ] and the expertise of the co-authors (NK and FHML). Food items were specifically scrutinized for the use of additives (flavours, flavour enhancers, colours, emulsifiers, emulsifying salts, artificial sweeteners, thickeners, foaming, anti-foaming, bulking, carbonating, gelling and glazing agents), specific ingredients such as industrially derived sugars (fructose, invert sugar, maltodextrin, dextrose, lactose, high fructose corn syrup, fruit juice concentrates), modified oils (hydrogenated fats, interesterified fats), extracted proteins (hydrolysed proteins, soy protein isolate, gluten, casein, whey protein, mechanically separated meat) [ 23 , 24 , 25 , 26 , 48 ]. All food items were assessed for at least three different brands and if a majority of the items (2 out of 3) qualified as UPFs, the product category was confirmed as UPF. The free-list of UPFs identified from the literature and confirmed through label reading using the online grocery retailer scan were then categorized on the basis of the primary ingredient of their composition and/or functionality of the product. A 23 category UPF list was developed at the end of Step 2.
Step 3. The confirmed UPF categories (23 categories), were then subjected to a saliency analysis, conducted by two authors (GK, MS). Saliency is a statistical accounting of items for rank and frequency of mention, across all respondents within a given domain. For example, the colour chosen most often from a free list of ten colours by a study population is referred to as the most salient [ 49 ]. The saliency test indicators included the commonly accessed, consumed, preferred and/or purchased (collectively referred to as ‘preferred’ in this paper to identify the common UPF categories accessed by the Indian consumers). These categories were limited to the food items that were confirmed as UPFs in Step 2. For example, if a study used “junk food” as a descriptor of UPFs and included freshly prepared savouries like “samosa/kachori” along with chips and soft drinks, we included data for only chips and soft drinks for the purpose of saliency analysis. The UPFs were then sorted from the most to the least preferred UPFs (Additional file 2 ). The steps and formulas [ 49 ] used to calculate composite salience scores for each UPF category have been illustrated in Fig. 2 . The UPF categories were classified per consumer preference, to the composite salience score cut-offs, defined after dividing the distribution of the composite salience scores obtained into tertiles as follows: (i) ≥ 0.61 as frequently preferred, (ii) 0.61 − 0.51 as infrequently preferred, and (iii) < 0.51 as rarely preferred UPFs.
Saliency analysis method for free-listed UPF categories
The literature search and study selection process of Step 1 is illustrated in Fig. 3 . A total of 23 research articles that matched the inclusion criteria were included in the final review. An overview of the extracted variables is provided in Additional File 1 . These studies were conducted in both rural (5 out of 23) and urban areas (17 out of 23), in different regions of the country, among a diverse population aged between 9 and 69 years (Table 1 ). Table 2 provides the outcome of the literature review with proxy descriptors along with the food items listed under them. These foods were verified as UPFs and non-UPFs.
Flow diagram reporting the screening and selection of studies reporting consumption and availability of UPFs in India
The online grocery retailer scan, label readings of 375 packaged foods were completed (atleast 3 brands per product) and 81 of those food products qualified as UPFs. Several of the packaged Indian traditional foods and snacks such as bottled and packaged pickles, namkeens (cereal and pulse-based extruded snacks), papads , frozen non-vegetarian meals and snacks, and frozen RTE meals (like rajmah curry and rice, biryani , dal makhni , etc.) had UPF ingredients and additives in their formulation that qualified them as UPFs (Table 3 ). Food products such as RTE breakfast cereals (e.g. poha , upma , etc.), RTE Indian curries (e.g. paneer makhani , butter chicken, etc.), Indian RTE bread (e.g. thepla , paratha , etc.), RTC mixes (e.g. idli mix, dal vada mix, etc.) also qualified as UPFs. However, some RTE traditional meals such as RTE biryani , RTE rajmah curry with rice, RTE kadhi pakoda with rice were not categorized as UPFs as these did not include UPF ingredients.
Consumer preferences for the confirmed UPF food categories identified above, were assessed using saliency scores. Table 4 lists these categories and shows the order of preference based on the saliency scores. The last column in the table indicates ‘frequently’, ‘infrequently’ or ‘rarely’ preferred UPFs by consumers in India. The frequently preferred UPFs were breads, chips and other extruded snacks (such as potato chips, cheese balls, puff corns, etc.) and sugar-sweetened beverages (such as cold drinks, diet coke, sodas, and energy drinks. The three rarely preferred UPFs were margarine and frozen/ packaged vegetarian and non-vegetarian snacks and meals (such as stuffed/plain parantha , naan , palak paneer , rajma , cutlets, fish/seafood snacks, salami, and sausages).
The present study aimed to identify the specific categories of UPFs accessed in India and rank them by consumer preference using a literature review, an online grocery retailer scan and saliency analysis. We found 23 categories of UPFs accessed by Indian consumers. After analysing the ingredient list of UPFs, we found that product formulation of several traditional Indian foods has transitioned from being processed to ultra-processed category with the use of industrially processed ingredients and presence of additives such as artificial colours, flavour enhancers, anti-caking agents. These ultra-processed versions of traditional foods even though have similar nutrient composition to home-prepared meals, are increasingly consumed, displacing home-cooked meals, and substituting staples. While the health effects of this displacement from minimally processed food ingredients to UPFs is an area of on-going research, we have growing evidence that UPF dietary patterns are linked to poor health outcomes [ 23 , 69 ]. It is crucial to track reformulation of traditional recipes to ultra-processed convenience foods especially since traditional meals are thought to be healthier [ 70 ]. The increasing market of ultra-processed traditional Indian recipes with poor nutritional profile needs more scrutiny and research.
Saliency analysis identified the preferred UPFs among the Indian population with breads, chips and sugar-sweetened beverages being the most preferred UPFs and frozen non-vegetarian snacks being the least preferred. This finding is consistent with the sales trends reported by Euromonitor International in 2020, which has also highlighted a substantial contribution of similar categories of packaged foods, such as bakery items, biscuits, packaged dairy products, savory snacks, and sauces and condiments [ 46 ]. Further, saliency analysis also indicates the preference of Indian consumers towards UPFs such as fruit-based preserves, cookies and biscuits, Indian sweet mixes, sauces and pickles, instant noodles/soups/ pasta and savoury puff rolls. Studies from other low and middle-income countries (LMICs) demonstrates similar trends in preference (consumption of UPFs and contribution to percentage of total calories) towards packaged confectioneries, savoury snacks, deep-fried foods, biscuits, candy/ chocolate, savoury snacks, canned red and luncheon meats, pre-fried French fries, mayonnaise, ketchup, fast-food such as sandwiches and pizzas, chips and salty snacks (including tortillas and pretzels), sweets and sweetened beverages and sausages (including canned) [ 71 , 72 ].
Our results also suggest a benefit of utilizing a classification system based on processing. Currently several UPFs are being captured by proxy descriptors like junk foods, fast foods, convenience foods, instant foods, packaged foods, etc. This limits comparability with other studies, monitoring the preference for and consumption of these products by the population, developing targeted interventions, tracking product reformulation and other regulatory measures to control exposure of these foods to vulnerable age groups through food advertising, etc. [ 73 ]. Using UPFs more consistently in studies reporting unhealthy food consumption pattern in India will help with global comparisons and in also elucidating the health effects of these foods. Additionally, as per the packaged food sales data from 2015-19, the Indian UPF market is slowly expanding with increasing sales of RTE meals, savoury snacks, processed fruits, vegetables, meats and other packaged foods [ 46 ]. The Nova food classification system can serve this purpose and may be explored as an option for categorization of foods by regulatory authorities. This classification system is used to assess dietary patterns in several high and middle-income countries [ 23 , 70 ]. Food based dietary guidelines of several countries such as Brazil, Uruguay, Ecuador, Peru and Israel have utilized Nova classification system to inform their dietary recommendations [ 74 , 75 , 76 , 77 , 78 ].
The present paper identified only a limited number of Indian studies which were primarily reported from 2 geographical regions. More such surveys on the consumption of UPFs are desirable to identify common regional UPFs. In the Indian context, several UPFs are indigenously produced by local retailers apart from the huge market share of nationally known branded UPFs [ 79 ]. These locally accessible UPFs have greater penetration into the local markets.
The categories of UPFs in India developed in the present study after due validation can be developed into a UPF consumption screener. This tool can be used for monitoring the UPF consumption in India and can address critical gap in scientific literature. This information on quantitative estimate of UPF consumption among Indian population can be useful for assessing impact of UPF consumption on increasing burden of NCDs in India.
This study is one of the first attempts to explore the types of UPFs in the Indian food market, identify the types of packaged traditional recipes that have been converted to UPFs, and map their saliency.
Studies reviewed were majorly from South India and largely represented the urban population, hence the results cannot be extended to the rural population. The study could only conduct saliency mapping of preferred foods without quantity of intake of UPFs and their contribution to total day’s energy intake. We could not explore traditional variants of UPFs that may be sold in the local unregulated markets.
India needs to develop a food classification system while systematically defining food categories based on level of processing. This should be followed by an assessment of the extent of UPFs consumption in India. The mapping of the UPFs in India reported in this paper provides the first step in developing a quick screener that systematically lists all the UPF categories. The data generated on consumption of UPFs using the screener is likely to inform policies on regulating the Indian UPFs market, undertake consumer education initiatives and create nutrition literacy around UPFs and thus contain their indiscriminate consumption. This may address the impact of UPF consumption on increasing burden of NCDs in India. There is an urgent need for strengthening the food regulatory environment to check the infiltration of several unhealthy UPFs in the Indian food market.
No new data was created or analyzed under the literature review part of the study. The datasets used as part of a particular component is available from the corresponding author on reasonable request.
Non-communicable diseases
Diet-related non-communicable diseases
Food frequency questionnaires
High fat sugar salt
Low and middle-income countries
Ready-to-eat
Ready-to-cook
World Health Organization. Non-Communicable Diseases Country Profiles 2018. https://apps.who.int/iris/handle/10665/274512 . [Cited 18 January 2023].
World Health Organization. Non-Communicable Diseases Progress Monitor 2022. https://apps.who.int/iris/handle/10665/353048 . [Cited 18 January 2023].
Indian Council of Medical Research (ICMR), Public Health Foundation of India (PHFI) and Institute for Health Metrics and Evaluation (IHME). India: Health of the Nation’s States-The Indian State-level Disease Burden Initiative. New Delhi: ICMR, PHFI, and IHME; 2017). https://phfi.org/downloads/171110_India_Health_of_Nation_states_Report_2017.pdf [Cited 20 January 2023].
Alwan A. Global status report on non-communicable diseases 2010. World Health Organization, 2011. https://apps.who.int/iris/bitstream/handle/10665/44579/9789240686458_eng.pdf . [Cited 8 May 2023].
International Institute for Population Sciences (IIPS) and ICF. National Family Health Survey (NFHS-5), India, 2019-21: Mizoram. Mumbai: IIPS. 2021. https://pdf.usaid.gov/pdf_docs/PA00ZBN4.pdf . [Cited 18 January 2023].
Sethi V, Lahiri A, Bhanot A, Kumar A, Chopra M, Mishra R, Alambusha R, Agrawal P, Johnston R, de Wagt A. Adolescents, diets and Nutrition: growing well in a changing World, the Comprehensive National Nutrition Survey. Thematic Rep, (1), 2019.
D’Angelo S, Yajnik CS, Kumaran K, Joglekar C, Lubree H, Crozier SR et al. Body size and body composition: a comparison of children in India and the UK through infancy and early childhood. J Epidemiol Community Health 2015 July 16;69(12):1147.
India - Sample Registration System (SRS)-Statistical Report. 2020. https://censusindia.gov.in/nada/index.php/catalog/44376 . [Cited 18 January 2023].
Dandona L, Dandona R, Kumar GA, Shukla DK, Paul VK, Balakrishnan K, et al. Nations within a nation: variations in epidemiological transition across the States of India, 1990–2016 in the global burden of Disease Study. Lancet. 2017;390(10111):2437–60.
Article Google Scholar
Tak M, Law C, Green R, Shankar B, Cornelsen L. Processed foods purchase profiles in urban India in 2013 and 2016: a cluster and multivariate analysis. BMJ open. 2022;12(10):e062254.
Article PubMed PubMed Central Google Scholar
Baker P, Friel S. Processed foods and the nutrition transition: evidence from Asia. Obes Rev. 2014;15(7):564–77.
Article CAS PubMed Google Scholar
Dunford EK, Farrand C, Huffman MD, Raj TS, Shahid M, Ni Mhurchu C, Neal B, Johnson C. Availability, healthiness, and price of packaged and unpackaged foods in India: a cross-sectional study. Nutr Health. 2022;28(4):571–9.
Article PubMed Google Scholar
The growth of ultra-processed foods in India: an analysis of trends, issues and policy recommendations. New Delhi: World Health Organization, Country Office for India; 2023. https://www.who.int/publications/i/item/9789290210672
Bassi S, Bahl D, Arora M, Tullu FT, Dudeja S, Gupta R. Food environment in and around schools and colleges of Delhi and National Capital Region (NCR) in India. BMC Public Health. 2021;21(1):1–3.
Harris J, Heard A, Schwartz M. Older but still vulnerable: all children need protection from unhealthy food marketing. Yale Rudd Center for Food Policy & Obesity. 2014; www.uconnruddcenter.org/files/Pdfs/Protecting_Older_Children_3_14.pdf . [Cited 24 December 2022].
Euromonitor International. (2019). Euromonitor International; London. Passport Global Market Information Database. https://www.euromonitor.com/
Pandav C, Smith Taillie L, Miles DR, Hollingsworth BA, Popkin BM. The WHO South-East Asia region nutrient profile model is quite appropriate for India: an exploration of 31,516 food products. Nutrients. 2021;13(8):2799.
The growth of ultra-processed foods in India: an analysis of trends, issues and policy recommendations. New Delhi: World Health Organization, Country Office for India; 2023.
Monteiro CA, Moubarac JC, Cannon G, Ng SW, Popkin B. Ultra-processed products are becoming dominant in the global food system. Obes Rev. 2013;14:21–8.
Magalhães V, Severo M, Correia D, Torres D, de Miranda RC, Rauber F, Levy R, Rodrigues S, Lopes C. Associated factors to the consumption of ultra-processed foods and its relation with dietary sources in Portugal. J Nutritional Sci. 2021;10.
Prakash J. Consumption trends of processed foods among rural population selected from South India. Int J Food Nutr Sci. 2015;2(6):1–6.
Google Scholar
Gupta A, Kapil U, Singh G. Consumption of junk foods by school-aged children in rural Himachal Pradesh, India. Indian J Public Health. 2018;62(1):65.
Monteiro CA, Cannon G, Lawrence M, da Costa Louzada ML, Pereira Machado P. Ultra-processed foods, diet quality, and health using the NOVA classification system. Rome: FAO; 2019. p. 48.
Monteiro CA, Cannon G, Levy RB, Moubarac JC, Louzada ML, Rauber F, Khandpur N, Cediel G, Neri D, Martinez-Steele E, Baraldi LG. Ultra-processed foods: what they are and how to identify them. Public Health Nutr. 2019;22(5):936–41.
Monteiro CA, Cannon G, Levy R, Moubarac J-C, Jaime P, Martins AP, et al. NOVA: the star shines bright. World Nutr. 2016;7(1–3):28–38.
Monteiro CA, Levy RB, Claro RM, Castro IRR, Cannon G. A new classification of foods based on the extent and purpose of their processing. Cad Saude Publica. 2010;26:2039–49. https://doi.org/10.1590/S0102-311X2010001100005 .
Moubarac JC, Batal M, Louzada ML, Steele EM, Monteiro CA. Consumption of ultra-processed foods predicts diet quality in Canada. Appetite. 2017;108:512–20.
Rauber F, Louzada ML, Steele EM, Millett C, Monteiro CA, Levy RB. Ultra-processed food consumption and chronic non-communicable diseases-related dietary nutrient profile in the UK (2008–2014). Nutrients. 2018;10(5):587.
da Costa Louzada ML, Ricardo CZ, Steele EM, Levy RB, Cannon G, Monteiro CA. The share of ultra-processed foods determines the overall nutritional quality of diets in Brazil. Public Health Nutr. 2018;21(1):94–102.
Louzada ML, Martins AP, Canella DS, Baraldi LG, Levy RB, Claro RM, Moubarac JC, Cannon G, Monteiro CA. Impact of ultra-processed foods on micronutrient content in the Brazilian diet. Rev Saude Publica. 2015;49.
Louzada ML, Martins AP, Canella DS, Baraldi LG, Levy RB, Claro RM, Moubarac JC, Cannon G, Monteiro CA. Ultra-processed foods and the nutritional dietary profile in Brazil. Rev Saude Publica. 2015;49.
Monteiro CA, Levy RB, Claro RM, de Castro IR, Cannon G. Increasing consumption of ultra-processed foods and likely impact on human health: evidence from Brazil. Public Health Nutr. 2011;14(1):5–13.
Moubarac JC, Martins AP, Claro RM, Levy RB, Cannon G, Monteiro CA. Consumption of ultra-processed foods and likely impact on human health. Evidence from Canada. Public Health Nutr. 2013;16(12):2240–8.
Poti JM, Mendez MA, Ng SW, Popkin BM. Is the degree of food processing and convenience linked with the nutritional quality of foods purchased by US households? Am J Clin Nutr. 2015;101(6):1251–62.
Article CAS PubMed PubMed Central Google Scholar
Luiten CM, Steenhuis IH, Eyles H, Mhurchu CN, Waterlander WE. Ultra-processed foods have the worst nutrient profile, yet they are the most available packaged products in a sample of New Zealand supermarkets. Public Health Nutr. 2016;19(3):530–8.
Beslay M, Srour B, Méjean C, Allès B, Fiolet T, Debras C, Chazelas E, Deschasaux M, Wendeu-Foyet MG, Hercberg S, Galan P. Ultra-processed food intake in association with BMI change and risk of overweight and obesity: a prospective analysis of the French NutriNet-Santé cohort. PLoS Med. 2020;17(8):e1003256.
Levy RB, Rauber F, Chang K, Louzada ML, Monteiro CA, Millett C, Vamos EP. Ultra-processed food consumption and type 2 diabetes incidence: a prospective cohort study. Clin Nutr. 2021;40(5):3608–14.
de Miranda RC, Rauber F, de Moraes MM, Afonso C, Santos C, Rodrigues S, Levy RB. Consumption of ultra-processed foods and non-communicable disease-related nutrient profile in Portuguese adults and elderly (2015–2016): the UPPER project. Br J Nutr. 2021;125(10):1177–87.
Food Safety and Standards. (Safe food and balanced diets for School Children) Regulations 2020. https://www.fssai.gov.in/upload/uploadfiles/files/Gazette_Notification_Safe_Food_Children_07_09_2020.pdf . [Cited 12 May, 2023].
Report of Working Group on Addressing Consumption of Foods High in Fat. Salt and Sugar (HFSS) and Promotion of Healthy Snacks in Schools of India. Working Group Constituted by Ministry of Women and Child Development Government of India 2015. https://www.nipccd.nic.in/file/reports/hfss.pdf . [Cited 12 May, 2023].
What India Eats ICMR-NIN. 2020. https://www.millets.res.in/pdf/what_india_eats.pdf . [Cited 12 May, 2023].
National Programme for Prevention and Control of Cancer, Diabetes, Cardiovascular Diseases and Stroke (NPCDCS). Directorate General of Health Services, Ministry of Health and Family Welfare, Government of India. Handbook for Counsellors Reducing Risk Factors for Non-Communicable Disease. 2017 Sep. https://main.mohfw.gov.in/sites/default/files/Handbook%20for%20Counselors%20-%20Reducing%20Risk%20Factors%20for%20NCDs_1.pdf . [Cited 12 May,2023].
Recommended Dietary Allowances and Estimated Average Requirements Nutrient Requirements for Indians. A Report of the Expert Group - Indian Council of Medical Research National Institute of Nutrition 2020. https://www.nin.res.in/RDA_Full_Report_2020.html . [Cited 12 May, 2023].
National Multi-sectoral Action Plan for Prevention and Control of Common. NCDs 2017–2022. https://main.mohfw.gov.in/sites/default/files/National%20Multisectoral%20Action%20Plan%20%28NMAP%29%20for%20Prevention%20and%20Control%20of%20Common%20NCDs%20%282017-22%29_1.pdf ,. [Cited 12 May, 2023].
Baker P, Machado P, Santos T, Sievert K, Backholer K, Hadjikakou M, Russell C, Huse O, Bell C, Scrinis G, Worsley A. Ultra-processed foods and the nutrition transition: global, regional and national trends, food systems transformations and political economy drivers. Obes Rev. 2020;21(12):e13126.
Packaged Food in India. Euromonitor Int 2020. https://www.euromonitor.com/
India Online Grocery Market Size. Report 2021–2028. Grand View Research. https://www.grandviewresearch.com/industry-analysis/india-online-grocery-market . [Cited 12 May, 2023].
Food Safety and Standards (Food Products Standards and Food Additives) Regulations. 2011. https://www.fssai.gov.in/cms/food-safety-and-standards-regulations.php . [Cited 8 May, 2023].
Quinlan M. Considerations for collecting free-lists in the field: examples from ethobotany. Field Methods. 2005;17(3):219–34.
Jain A, Mathur P. Intake of ultra-processed foods among adolescents from low-and middle-income families in Delhi. Indian Pediatr. 2020;57(8):712–4.
National Non-communicable Disease Monitoring Survey (NNMS 2017-18), Bengaluru ICMR-NCDIR. India. 2021. Available from: chrome-extension://efaidnbmnnnibpcajpcglclefindmkaj/viewer.html?Pdfurl = https%3A%2F%2Fwww.ncdirindia.org%2Fnnms%2Fresources%2fchapter_4_2_3.pdf&clen = 181451&chunk = true. [Cited 18 Jan, 2023]
Law C, Green R, Kadiyala S, Shankar B, Knai C, Brown KA, Dangour AD, Cornelsen L. Purchase trends of processed foods and beverages in urban India. Global Food Secur. 2019;23:191.
Goel S, Kaur T, Gupta M. Increasing proclivity for junk food among overweight adolescent girls in district Kurukshetra, India. Int Res J Biol Sci. 2013;16:17.
Bhavani V, Prabhavathy Devi N. Junk and Sink: A Comparative Study on Junk Food Intake among Students of India’. Shanlax Int J Arts Sci Humanit. 2020:13–8.
Naveenkumar D, Parameshwari S. A study on junk food consumption and its Effect on adolescents in Thiruvannamalai District, Tamil Nadu, India. Adalya J August. 2019;8(8):781–93. [ISSN 1301–2746].
Hiregoudar V, Suresh CM, Singode C, Raghavendra B. Proportion, patterns, and determinants of junk food consumption among adolescent students. Annals Community Health. 2021;9(2):46.
Kumari R, Kumari M. Quality aspects of fast foods and their consumption pattern among teenagers of rural-urban region of Sabour block in Bhagalpur district of India. The Pharma Innovation Journal. 2020. 2020;9(4):96–102.
Amin T, Choudhary N, Jabeen AN. Study of fast food consumption pattern in India in Children aged 16–20 years. Intl J Food Ferment Technol. 2017;7(1):1–8.
Singh M, Mishra S. Fast food consumption pattern and obesity among school going (9–13 year) in Lucknow District. Int J Sci Res. 2014;3(6):1672–4.
Kotecha PV, Patel SV, Baxi RK, Mazumdar VS, Shobha M, Mehta KG, Mansi D, Ekta M. Dietary pattern of school going adolescents in urban Baroda, India. J Health Popul Nutr. 2013;31(4):490.
CAS PubMed PubMed Central Google Scholar
Joseph N, Nelliyanil M, Rai S, YP RB, Kotian SM, Ghosh T, Singh M. Fast food consumption pattern and its association with overweight among high school boys in Mangalore city of southern India. J Clin Diagn Research: JCDR. 2015;9(5):LC13.
Prabhu NB. Examining fast-food consumption behaviour of students in Manipal University, India. Afr J Hospitality Tourism Leisure. 2015;4(2):621–30.
Khongrangjem T, Dsouza SM, Prabhu P, Dhange VB, Pari V, Ahirwar SK, Sumit K. A study to assess the knowledge and practice of fast food consumption among pre-university students in Udupi Taluk, Karnataka, India. Clin Epidemiol Global Health. 2018;6(4):172–5.
Landge JA, Khadkikar GD. Lifestyle and nutritional status of late adolescent in an urban area of Western Maharashtra: cross sectional study. Int J Community Med Public Health. 2020;7(8):3027.
Iyer ML, Kumar NP. A study of Breakfast habits of Urban Indian consumers. Research Journals of Economics and Business Study. Res J Econ Bus Stud. 2014;3(8):107–16. [ISSN: 2251 – 1555].
Rathi N, Riddell L, Worsley A. Food consumption patterns of adolescents aged 14–16 years in Kolkata, India. Nutr J. 2017;16(1):1–2.
Vasan M. Consumers’ preference and consumption towards instant Food products. Think India J. 2019;22(14):8333–7.
Harrell M, Medina J, Greene-Cramer B, Sharma SV, Arora M. Understanding eating behaviors of New Delhi’s Youth. J Appl Res Child. 2015;6(2):8.
Mediratta S, Ghosh S, Mathur P. Intake of ultra-processed food, dietary diversity and the risk of nutritional inadequacy among adults in India. Public Health Nutr. 2023;26(12):2849–58.
Monteiro CA, Cannon G, Moubarac JC, Levy RB, Louzada MLC, Jaime PC. The UN Decade of Nutrition, the NOVA food classification and the trouble with ultra-processing. Public Health Nutr. 2017;21(1):5–17.
Islam MR, Rahman SM, Rahman MM, Pervin J, Rahman A, Ekström EC. Gender and socio-economic stratification of ultra-processed and deep-fried food consumption among rural adolescents: a cross-sectional study from Bangladesh. PLoS ONE. 2022;17(7):e0272275.
Pries AM, Filteau S, Ferguson EL. Snack food and beverage consumption and young child nutrition in low-and middle‐income countries: a systematic review. Matern Child Nutr. 2019;15:e12729.
Costa CD, Faria FR, Gabe KT, Sattamini IF, Khandpur N, Leite FH, Steele EM, Louzada ML, Levy RB, Monteiro CA. Nova score for the consumption of ultra-processed foods: description and performance evaluation in Brazil. Rev Saúde Pública. 2021;55:13.
Monteiro CA, Cannon G, Moubarac JC, Martins AP, Martins CA, Garzillo J, Canella DS, Baraldi LG, Barciotte M, Louzada ML, Levy RB, Claro RM, Jaime PC. Dietary guidelines to nourish humanity and the planet in the twenty-first century. A blueprint from Brazil. Public Health Nutr. 2015;18(13):2311–22. Epub 2015 Jul 24. PMID: 26205679; PMCID: PMC10271430.
Ministerio, de Salud. Guía Alimentaria para la Población Uruguay. 2016. https://www.gub.uy/ministerio-salud-publica/sites/ministerio-salud-publica/files/documentos/campanas/MSP_GUIA_ALIMENTARIA_POBLACION.pdf
Ministerio de Salud Pública del Ecuador y la Organización de las Naciones Unidas para la Alimentación y la Agricultura. Documento Técnico De las Guías Alimentarias Basadas en Alimentos (GABA) Del Ecuador. Quito. 2020. https://doi.org/10.4060/ca9928es .
Ministerio de Salud Instituto Nacional de Salud y Centro Nacional de Alimentación y Nutrición. Guías Alimentarias para la Población Peruana. 2020. https://repositorio.ins.gob.pe/bitstream/handle/20.500.14196/1247/Gu%c3%ada-alimen21.pdf?sequence=1&isAllowed=y
Nutritional Recommendations: The Israeli Ministry of Health. 2019. https://www.health.gov.il/PublicationsFiles/dietary%20guidelines%20EN.pdf
Downs SM, Thow AM, Ghosh-Jerath S, Leeder SR. Developing interventions to reduce consumption of unhealthy fat in the food retail environment: a case study of India. J Hunger Environ Nutr. 2014;9(2):210–29.
Download references
We would like to appreciate the contribution of Dr. Shukrani Shinde for supporting the study team during the literature search and online grocery retailer scan.
This work is funded through the Innovative Methods and Metrics for Agriculture and Nutrition Action (IMMANA) programme (Grant IMMANA 3.06), led by the London School of Hygiene & Tropical Medicine (LSHTM). IMMANA is co-funded with UK Aid from the UK government and by the Bill & Melinda Gates Foundation. This work was supported, in part, by the Bill & Melinda Gates Foundation [INV-002962 / OPP1211308]. Under the grant conditions of the Foundation, a Creative Commons Attribution 4.0 Generic License has already been assigned to the Author Accepted Manuscript version that might arise from this submission.
Suparna Ghosh-Jerath and Neha Khandpur contributed equally to this work.
The George Institute for Global Health, 308, Third Floor, Elegance Tower, Plot No. 8, Jasola District Centre, New Delhi, Delhi, 110025, India
Suparna Ghosh-Jerath & Sahiba Kohli
Wageningen University, Wageningen, The Netherlands
Neha Khandpur
Public Health Foundation of India, New Delhi, India
Gaurika Kumar, Meenu Singh, Inderdeep Kaur Bhamra & K Srinath Reddy
Center for Epidemiological Research in Nutrition and Health, Faculty of Public Health, University of Sao Paulo, Sao Paulo, Brazil
Fernanda H Marrocos-Leite
You can also search for this author in PubMed Google Scholar
The study was conceptualized by KSR, SGJ, NK and FHML. The literature review and online grocery retailer scan were conducted by MS, GK, IKB. The first draft of the manuscript was prepared by MS, GK, IKB, SK and SGJ. The manuscript was critiqued and edited by SGJ, NK, FHML and KSR. SGJ had primary responsibility for final content; and all authors read and approved the final manuscript.
Correspondence to Suparna Ghosh-Jerath .
Ethics approval and consent to participate.
This study was conducted according to guidelines laid down in the Declaration of Helsinki, and all procedures involving humans were approved by the Institutional Ethics Committee at the Public Health Foundation of India, and the ethics committee of University of Sao Paulo. The current manuscript, however, reports findings from an exhaustive literature review and online grocery retailer scan for which informed consent process is not applicable.
Not applicable.
The authors declare no competing interests.
Publisher’s note.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Below is the link to the electronic supplementary material.
Supplementary material 2, rights and permissions.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .
Reprints and permissions
Cite this article.
Ghosh-Jerath, S., Khandpur, N., Kumar, G. et al. Mapping ultra-processed foods (UPFs) in India: a formative research study. BMC Public Health 24 , 2212 (2024). https://doi.org/10.1186/s12889-024-19624-1
Download citation
Received : 04 July 2023
Accepted : 29 July 2024
Published : 14 August 2024
DOI : https://doi.org/10.1186/s12889-024-19624-1
Anyone you share the following link with will be able to read this content:
Sorry, a shareable link is not currently available for this article.
Provided by the Springer Nature SharedIt content-sharing initiative
ISSN: 1471-2458
COMMENTS
Tips on how to write a review of related literature in research. Given that you will probably need to produce a number of these at some point, here are a few general tips on how to write an effective review of related literature 2. Define your topic, audience, and purpose: You will be spending a lot of time with this review, so choose a topic ...
A literature review is a survey of scholarly sources (such as books, journal articles, and theses) related to a specific topic or research question. It is often written as part of a thesis, dissertation , or research paper , in order to situate your work in relation to existing knowledge.
A literature review is meant to analyze the scholarly literature, make connections across writings and identify strengths, weaknesses, trends, and missing conversations. A literature review should address different aspects of a topic as it relates to your research question. A literature review goes beyond a description or summary of the ...
A literature review may consist of simply a summary of key sources, but in the social sciences, a literature review usually has an organizational pattern and combines both summary and synthesis, often within specific conceptual categories.A summary is a recap of the important information of the source, but a synthesis is a re-organization, or a reshuffling, of that information in a way that ...
A literature review is an integrated analysis-- not just a summary-- of scholarly writings and other relevant evidence related directly to your research question.That is, it represents a synthesis of the evidence that provides background information on your topic and shows a association between the evidence and your research question.
Literature Review is a comprehensive survey of the works published in a particular field of study or line of research, usually over a specific period of time, in the form of an in-depth, critical bibliographic essay or annotated list in which attention is drawn to the most significant works.. Also, we can define a literature review as the collected body of scholarly works related to a topic:
Closely related to the semi-structured review approach is the integrative or critical review approach. ... review, an integrative review usually has a different purpose, with the aim to assess, critique, and synthesize the literature on a research topic in a way that enables new theoretical frameworks and perspectives to emerge (Torraco, 2005 ...
What kinds of literature reviews are written? Narrative review: The purpose of this type of review is to describe the current state of the research on a specific topic/research and to offer a critical analysis of the literature reviewed. Studies are grouped by research/theoretical categories, and themes and trends, strengths and weakness, and gaps are identified.
The broader scholarly literature should be related to the chosen main topic for the LR (how to develop the literature review section). The LR can cover the literature from one or more disciplines, depending on its scope, but it should always offer a new perspective. ... Ideas and theories in the field are related to research methodologies .
A literature review is a written work that: Compiles significant research published on a topic by accredited scholars and researchers; Surveys scholarly articles, books, dissertations, conference proceedings, and other sources; Examines contrasting perspectives, theoretical approaches, methodologies, findings, results, conclusions.
A literature or narrative review is a comprehensive review and analysis of the published literature on a specific topic or research question. The literature that is reviewed contains: books, articles, academic articles, conference proceedings, association papers, and dissertations. It contains the most pertinent studies and points to important ...
2. MOTIVATE YOUR RESEARCH in addition to providing useful information about your topic, your literature review must tell a story about how your project relates to existing literature. popular literature review narratives include: ¡ plugging a gap / filling a hole within an incomplete literature ¡ building a bridge between two "siloed" literatures, putting literatures "in conversation"
A literature review is a comprehensive summary of previous research on a topic. The literature review surveys scholarly articles, books, and other sources relevant to a particular area of research. The review should enumerate, describe, summarize, objectively evaluate and clarify this previous research. It should give a theoretical base for the ...
A literature review is a critical analysis and synthesis of existing research on a particular topic. It provides an overview of the current state of knowledge, identifies gaps, and highlights key findings in the literature. 1 The purpose of a literature review is to situate your own research within the context of existing scholarship ...
A review of related literature (RRL) is a part of the research report that examines significant studies, theories, and concepts published in scholarly sources on a particular topic. An RRL includes 3 main components: A short overview and critique of the previous research.
Literature reviews are in great demand in most scientific fields. Their need stems from the ever-increasing output of scientific publications .For example, compared to 1991, in 2008 three, eight, and forty times more papers were indexed in Web of Science on malaria, obesity, and biodiversity, respectively .Given such mountains of papers, scientists cannot be expected to examine in detail every ...
Types of Literature Review are as follows: Narrative literature review: This type of review involves a comprehensive summary and critical analysis of the available literature on a particular topic or research question. It is often used as an introductory section of a research paper. Systematic literature review: This is a rigorous and ...
A review of literature is a classification and evaluation of what accredited scholars and. researchers have written on a topic, organized according to a guiding concept such as a research ...
The literature review provides a way for the novice researcher to convince the proposal the reviewers that she is knowledgeable about the related research and the "intellectual traditions" that support the proposed study. The literature review provides the researcher with an opportunity to identify any gaps that may exist in the body of ...
A comparison/contrast of the choices different authors or characters make in a work. A reading of a work based on an outside philosophical perspective (Ex. how would a Freudian read Hamlet ?) A study of the sources or historical events that occasioned a particular work (Ex. comparing G.B. Shaw's Pygmalion with the original Greek myth of ...
Join the 250,000+ researchers, students, and professionals using Litmaps to accelerate their literature review. Find the right papers faster. Get started for free! ... Using Litmaps for my research papers has significantly improved my workflow. Typically, I start with a single paper related to my topic. ... University of Applied Sciences ...
Writing the Literature Review Barry Mauer and John Venecek. We discuss the following topics on this page: The Literature Review; ... The aim of a scoping review is to provide a comprehensive overview or map of the published research or evidence related to a research question. This might be considered a prelude to a systematic review that would ...
Finding "The Literature" Organizing/Writing; APA Style This link opens in a new window; Chicago: Notes Bibliography This link opens in a new window; MLA Style This link opens in a new window; Sample Literature Reviews. Sample Lit Reviews from Communication Arts; Have an exemplary literature review? Get Help!
R.I.E., Bhopal. [email protected]. Introduction. "Review of related Literature" is mostly known as the second chapter in almost every. empirical thesis. But it is the very first step to ...
A review of related literature (RRL) puts your research in the context of what is currently known about the topic. Accordingly, it is a good way to highlight the knowledge gaps and what your study offers in terms of novelty and significance. Your question on outlining the subtopics is a pertinent one. However, outlining subtopics would heavily ...
As already indicated by others, a literature review should relate to your research question, so have a look what the relation is between your approach and that of others, where it differs, adds ...
Looking in the reference section of a literature or systematic review can be a good place to find original research studies. Below is a screenshot of the abstract of the article The Effect of Dietary Glycaemic Index on Glycaemia in Patients with Type 2 Diabetes: A Systematic Review and Meta-Analysis of Randomized Controlled Trials , with words ...
Generating knowledge graphs through text mining of catalysis research related literature†. Alexander S. Behr * a, Diana Chernenko a, Dominik Koßmann b, Arjun Neyyathala c, Schirin Hanf c, Stephan A. Schunk d and Norbert Kockmann a a Department of Biochemical and Chemical Engineering, Laboratory of Equipment Design, TU Dortmund University, Dortmund, Germany.
Male sex, early life chemical exposure and the brain aromatase enzyme have been implicated in autism spectrum disorder (ASD). In the Barwon Infant Study birth cohort (n = 1074), higher prenatal ...
Increased consumption of ultra-processed foods (UPFs) which have additives such as artificial colours, flavours and are usually high in salt, sugar, fats and specific preservatives, are associated with diet-related non-communicable diseases (NCDs). In India, there are no standard criteria for identifying UPFs using a classification system based on extent and purpose of industrial processing.