Science - USA (2022-02-18)

this selection was based on the following: (i)
established evidence, within each group of com-
pounds, for both their endocrine interfering
properties and their association to neurodevel-
opment [from single-compound studies, as in

the case of bisphenol A (BPA) ( 11 )]; and (ii) our

previous demonstration, with a single-compound

approach, that prenatal exposure for three of

the chemicals—diethyl phthalate (DEP), dibutyl

phthalate (DBP), and benzyl butyl phthalate

(BBzP)—was associated with language delay

in children at 30 months of age ( 9 ).

Based on these choices of end point and

compounds, we determined a chemical mix-

ture associated with language delay in children

Caporaleet al.,Science 375 , eabe8244 (2022) 18 February 2022 2 of 15

Fig. 1. Overview of the study.
(A) Identification of an EDC mixture
associated with adverse health outcomes
in neurodevelopment. In the SELMA
pregnancy study, 20 analytes from 15 EDCs
were measured in the urine or serum of
women around pregnancy week 10.
Associations between these exposures
and language delay of the children at
2.5 years of age were established using
WQS regression. This resulted in the
identification, within the mixture, of sEDCs
that contributed to the association with the
adverse health outcome (language delay).
Based on the ratios found in SELMA
mothers, the identified sEDCs were mixed
to compose MIX N for subsequent use in
the experimental systems in concentra-
tions corresponding to 0.01X, 0.1X, 1X, 10X,
100X, and 1000X the serum concentrations
in SELMA mothers. MEP, mono-ethyl
phthalate; MBP, mono-n-butyl phthalate;
MBzP, monobenzyl phthalate; MiNP, Mono-
isononyl phthalate; PFHxS, perfluorohexane
sulfonate; PFNA, perfluorononanoic acid;
PFOS, perfluorooctane sulfonate.
(B) Identification of gene regulatory
networks and cellular responses dysregu-
lated by MIX N, along with their dose-
response relationships. Transcriptome
analyses were carried out in HFPNSCs and
COs upon treatment with 0.1X to 1000X
MIX N at several model-relevant time
points. Examples of dose-response
patterns are shown across models.
Significant (FDR < 0.05) transcriptional
changes were already detected at 1X
concentrations (red dotted line). Down-
regulation by MIX N of the bona fide ASD-
causing geneSCN2Ain COs. The horizontal
black bars represent the mean expression
value for each exposure condition.
(C) Validation of key pathways affected by
MIX N and their physiological impact in in vivo
models. Effects on key genes and pathways
identified in the cellular models were
assessed inX. laevisandD. reriolarvae upon
short-term treatment (48 to 72 hours),
exemplified by brain-derived neurotrophic
factor (bdnf) expression inXenopus(left) and
thyroid hormone receptora(thra) expression
in zebrafish (right) upon MIX N exposure,
quantified using RT-qPCR. In the box plots,
boxes and whiskers represent the interquartile range and minimum and maximum values, respectively. (D) Referring experimental in vivo evidence back to the SELMA cohort for risk
assessment by SMACH. In the first step of SMACH, sufficient similarity was tested between prenatal exposures and the reference MIX N mixture. In the second step, a SMRI was
calculated, which is based on the BMDL established in (C). Of the SELMA women, 54% had a SMRI higher than 1, indicating that their children are at risk of adverse neurodevelopmental
outcomes. In the SMRI graph, the red dotted line indicates the median, N indicates the number of women in the analysis, and P10 to P99 indicate the different percentiles.

Identication of EDCs of Concern

EPIDEMIOLOGY

MIX N

Estimation of daily intake

association bet

ween language delay

and the WQS index

BIOSTATISTICS

Formulation of EDC concentrations

in mix and synthesis

CHEMISTRY

CELL AND DEVELOPMEN

TA

L

BIOLOGY

Weighted Quantile Sum regression

Exposure to synthesized mixtures

Dose-dependent response

relative concentration

Gene expression, Reporter assays, Developmental phenotypes

10 weeks

gestation

7 years old

children

EDC levels in urine, blood + clinical data

30 months-old children

animal - in vivo models

Transcriptomics, Gene Network Analysis, Cellular phenotype

A

B

C

Phthalates

PFAS

bisphenol-A BPA

MEP

MBP

MBzP

MiNP

PFHxS

PFOS

PFNA

0.25 0.20 0.15 0.10 0.05 0

FF FF FF FF

FF FF FF

R

MIX N

R

R

Cortical

organoids DMSO 1X1000X

SCN2A

Corrected expression

Fetal neural

progenitors DM

SO0. 1 X1X 10 X100X 1000 X

Log Foldchange

SMAC

H

D Similar mixture approach (SMACH)

Risk Assessment

MIX N log10(conc+1)^1

human - in vitro models

RESEARCH | RESEARCH ARTICLE