RESEARCH ARTICLE
◥ENVIRONMENTAL TOXINS
From cohorts to molecules: Adverse impacts
of endocrine disrupting mixtures
Nicolò Caporale1,2,3†, Michelle Leemans^4 †‡, Lina Birgersson^5 †, Pierre-Luc Germain^1 †§,
Cristina Cheroni1,2,3†, Gábor Borbély^6 ¶, Elin Engdahl6,7¶, Christian Lindh^8 , Raul Bardini Bressan^9 ,
Francesca Cavallo^1 , Nadav Even Chorev^1 , Giuseppe Alessandro DÕAgostino^1 #, Steven M. Pollard^9 ,
Marco Tullio Rigoli1,2, Erika Tenderini^1 , Alejandro Lopez Tobon^1 , Sebastiano Trattaro1,2, Flavia Troglio^1 ,
Matteo Zanella^1 *, Åke Bergman6,10,11¶, Pauliina Damdimopoulou6,12¶, Maria Jönsson^7 ,
Wieland Kiess^13 , Efthymia Kitraki^14 , Hannu Kiviranta^15 , Eewa Nånberg^16 , Mattias Öberg6,17¶,
Panu Rantakokko^15 , Christina Rudén^10 , Olle Söder^18 , Carl-Gustaf Bornehag19,20††,
Barbara Demeneix^4 ††, Jean-Baptiste Fini^4 ††, Chris Gennings^20 ††, Joëlle Rüegg6,7¶††,
Joachim Sturve^5 ††, Giuseppe Testa1,2,3*††
Convergent evidence associates exposure to endocrine disrupting chemicals (EDCs) with major
human diseases, even at regulation-compliant concentrations. This might be because humans are
exposed to EDC mixtures, whereas chemical regulation is based on a risk assessment of individual
compounds. Here, we developed a mixture-centered risk assessment strategy that integrates
epidemiological and experimental evidence. We identified that exposure to an EDC mixture in early
pregnancy is associated with language delay in offspring. At human-relevant concentrations, this
mixture disrupted hormone-regulated and disease-relevant regulatory networks in human brain
organoids and in the model organismsXenopus leavisandDanio rerio, as well as behavioral
responses. Reinterrogating epidemiological data, we found that up to 54% of the children had
prenatal exposures above experimentally derived levels of concern, reaching, for the upper decile
compared with the lowest decile of exposure, a 3.3 times higher risk of language delay.
H
uman populations are exposed to a
large number of chemicals with known
or suspected endocrine-disrupting prop-
erties [endocrine-disrupting chemicals
(EDCs)] ( 1 ). Whereas real life entails
simultaneous exposure to multiple EDCs in
mixtures, with additive effects manifesting at
lower doses than experimental effect thresh-
olds for single compounds ( 2 ), regulation has
so far operated on a single-chemical basis.
Because exposure to single EDCs has been
repeatedly associated with major diseases and
impaired development ( 3 ), including neuro-
developmental disorders ( 4 ), assessing the
health impact of real-life mixtures is a crucial
unmet challenge. However, experimental evi-
dence with mixtures is, at present, limited to
combinations within the same chemical class
or to observational measurements for more
complex mixtures and thus lacks the mecha-
nistic weight to link actual population-based
exposures with adverse health outcomes in
humans.
In the context of chemical regulatory affairs,
most authorities, including the Organisation
for Economic Co-operation and Development(OECD), recommend integrated approaches
for testing and assessment that incorporate
results from multiple methodologies. Empha-
sis is placed on molecular initiating events and
key events that lead to physiologically mea-
surable adverse outcomes. The regulatory
framework is, however, still entirely predicated
on the assessment of individual chemicals,
leaving the impact of simultaneous exposure
to multiple compounds, even at modest con-
centrations, unexamined. Here, leveraging
the Swedish Environmental Longitudinal,
Mother and child, Asthma and allergy (SELMA)
epidemiological study, we focused on language
delay as a test case for a prototypical adverse
outcome in the neurodevelopmental domain
and tested whether a mixture-based approach
was epidemiologically and experimentally viable
in orienting a more realistic risk assessment.
Speech and language development was chosen
as an epidemiological end point because
language delay is an early marker for intellec-
tual functioning deficits in children and is
routinely used for further assessment that might
lead to habilitation and support. Several
studies have reported that language develop-
mental milestones in early childhood have
predictive validity for intellectual functioning
later in childhood ( 5 , 6 ) and even in adulthood
( 7 ). Indeed, in the SELMA study, language
delay at 30 months was strongly associated
with cognitive function at 7 years of age ( 8 ).
Finally, in terms of the general applicability
of this measure, its prevalence (10% in the
SELMA cohort) was also similar across con-
texts; we found a language delay prevalence
of 10% in the US The Infant Development
and the Environment Study (TIDES) ( 9 ), and
another Swedish study found a prevalence
of 6% ( 10 )
In terms of composition, to define the
feasibility of a mixture-centered approach
for EDC risk assessment, we focused on 15
chemicals that belong to the three chemical
groups of phthalates, alkyl phenols, and per-
fluoroalkyl substances (PFAS). The rationale forRESEARCH
Caporaleet al.,Science 375 , eabe8244 (2022) 18 February 2022 1of15
(^1) High Definition Disease Modelling Lab, Stem Cell and Organoid Epigenetics, IEO, European Institute of Oncology, IRCCS, 20141 Milan, Italy. (^2) Department of Oncology and Hemato-oncology,
University of Milan, 20122 Milan, Italy.^3 Human Technopole, V.le Rita Levi-Montalcini, 1, 20157 Milan, Italy.^4 UMR 7221, Phyma, CNRS–Muséum National d’Histoire Naturelle, Sorbonne Université,
75005 Paris, France.^5 Department of Biological and Environmental Sciences, University of Gothenburg, 41463 Gothenburg, Sweden.^6 Swedish Toxicology Sciences Research Center (SWETOX),
Södertälje, Sweden.^7 Department of Organismal Biology, Environmental Toxicology, Uppsala University, SE-752 36 Uppsala, Sweden.^8 Division of Occupational and Environmental Medicine,
Department of Laboratory Medicine, Lund University, SE-221 85 Lund, Sweden.^9 Medical Research Council Centre for Regenerative Medicine and Edinburgh Cancer Research UK Centre,
University of Edinburgh, Edinburgh, UK.^10 Department of Environmental Science, Stockholm University, SE-10691 Stockholm, Sweden.^11 School of Science and Technology, Örebro University, SE-
70182 Örebro, Sweden.^12 Division of Obstetrics and Gynecology, Department of Clinical Science, Intervention and Technology, Karolinska Institutet and Karolinska University Hospital, 141 86
Stockholm, Sweden.^13 Hospital for Children and Adolescents, Department of Women and Child Health, University Hospital, University of Leipzig, 04103 Leipzig, Germany.^14 Lab of Basic Sciences,
Faculty of Dentistry, National and Kapodistrian University of Athens, 152 72 Athens, Greece.^15 Department of Health Security, Finnish Institute for Health and Welfare (THL), Kuopio 70210,
Finland.^16 School of Health Sciences, Örebro University, SE-70182 Örebro, Sweden.^17 Institute of Environmental Medicine, Karolinska Institutet, SE-17177 Stockholm, Sweden.^18 Department of
Women’s and Children’s Health, Pediatric Endocrinology Division, Karolinska Institutet and University Hospital, SE-17176 Stockholm, Sweden.^19 Faculty of Health, Science and Technology,
Department of Health Sciences, Karlstad University, SE- 651 88 Karlstad, Sweden.^20 Icahn School of Medicine at Mount Sinai, New York, NY 10029, USA.
*Corresponding author. Email: [email protected] (C.-G.B.); [email protected] (B.D.); [email protected] (J.R.); [email protected] (G.T.)
†These authors contributed equally to this work.
‡Present address: Université Paris -Est Créteil, INSERM, IMRB, Team CEpiA, F-94010 Créteil, France.
§Present address: D-HEST Institute for Neurosciences, ETH Zürich and Lab of Statistical Bioinformatics, University of Zürich, Zürich, Switzerland.
¶SWETOX was closed at the end of 2018.
#Present address: Lee Kong Chian School of Medicine, Nanyang Technological University, Singapore, Singapore.
**Present address: Evotec SE, Hamburg, Germany.
††These authors contributed equally to this work.