exposure to be extended from pluripotency
through an intermediate stage of corticogenesis
(day 50), a stage that features a particularly
relevant representation of progenitor cells,
early neurons, and developmental trajecto-
ries that match, at the transcriptomic level,
the gestation time when SELMA women were
profiled (fig. S2A).
Differential expression analysis showed, as
expected, a much broader impact than that
in the acute setting and included three genes
(DHRS3,LGI4, andNEFL) that were dysre-
gulated upon both acute and chronic exposure
(fig. S2B). The effect of MIX N chronic expo-
sure on gene expression highlighted a high
degree of concordance between the two doses
for both experimental systems (Fig. 2, A and B,
and fig. S2C) and revealed a major impact of
MIX N exposure (132 DEGs for fetal progen-
itors and 599 DEGs for organoids; FDR < 0.05,
logFC > 0.5, logCPM > 0). In both fetal pro-
genitors and organoids, genes down-regulated
upon MIX N exposure were enriched for gene
ontology (GO) terms related to extracellular
compartment regulation, such as extracellular
matrix organization in fetal progenitors and
collagen fibril organization in organoids, whereas
up-regulated genes were enriched for categories
related to cellular differentiation, such as post-
embryonic animal organ development in fetal
progenitors and neuronal fate specification
in organoids (fig. S2D). Moreover, gene set
enrichment analysis (GSEA) revealed significant
(adjustedP< 0.1 and normalized enrichment
score > 2) enrichment for disease-relevant genes,
such as genes associated with neurodegenerative
diseases ( 32 ), cell proliferation, and several types
of cancer ( 33 ) (Fig. 2C).
We next complemented the analysis with a
more comprehensive approach by surveying,
in both experimental systems, the transcrip-
tional changes of genes modulated in at least
one of the examined conditions and applying
a clustering approach to identify the genes that
were concordantly dysregulated, as well as the
ones that have an opposite response in organo-
ids and fetal progenitors (Fig. 3A). This showed
that concordantly regulated genes are enriched
for GO categories related to the plasma mem-
brane, cell polarity, and extracellular matrix
organization, whereas genes with opposite
regulation are enriched for categories related
to the regulation of translation and RNA
metabolism (Fig. 3B).
Finally, given the founding epidemiological
evidence associating MIX N exposure to
language delay, we tested whether the genes
significantly (FDR < 0.05) altered by MIX N in
fetal progenitors or organoids were enriched
for genes underlying neurodevelopmental
disability, harnessing the best-established
databases of causative gene annotations for
neurodevelopmental disorders (NDDs), includ-
ing ASD ( 19 , 34 – 42 ). As shown in Fig. 3, C and
D, and fig. S2E, we found not only that MIX
N DEGs had significant (P< 0.05) overlaps
with the NDD genetic burden but also that,
of particular note, for some cases in which the
genetic cause is known to be a copy number
gain or loss, the direction of MIX N induced
dysregulation was concordant with the genetic
lesion (table S4). This is exemplified by the
down-regulation by MIX N of the following
bona fide ASD-causing genes, annotated with
the highest score genes in the Simons Foun-
dation Autism Research Initiative (SFARI)
database: (i) in organoids, the subunit alpha 2
of sodium voltage-gated channelSCN2A, which
is essential for the generation and propagation
of action potentials and whose disruptive muta-
tions impair channel function and cause ASD;
and (ii) in fetal progenitors, the transmembrane
glycoproteinCD38, for which genetic mutations
and knock-out mouse models have shown a
causative role for ASD ( 21 , 43 – 47 )(Fig.3,Cand
D). These results indicate that EDC exposure
converges on the interference of the same
gene targets and biological pathways of ASD
( 48 , 49 ).
To validate mixture-specific effects, both
with regard to the current regulatory prac-
tice (single compound at a time) and in light of
the complexity of molecular events triggered
by endocrine disruption, our experimental de-
sign included a panel of additional exposures
performed on the same human in vitro systems.
Specifically, in addition to 0.1% dimethyl sulfox-
ide (DMSO) as a negative control, we also tested
5 nM triiodothyronine (T3) exposure to define
the baseline of the effect of thyroid hormone
(TH) modulation, given its essential role in
brain development ( 50 , 51 ); indeed, previous
epidemiological ( 52 , 53 ) and experimental
( 54 ) evidence had identified the main chemical
classes of the tested mixture as TH disruptors
( 55 ). We thus exposed the human neurodeve-
lopmental models to T3 and identified, as
expected, a major transcriptional impact (391
DEGs for fetal progenitors and 2272 DEGs for
organoids; FDR < 0.05, logFC > 0.5, logCPM >
0) and observed different patterns of dysre-
gulation relative to MIX N (fig. S3, A to G).
Moreover, in organoids, T3 and MIX N ex-
posures showed an opposite effect for a sizable
portion of the genes modulated by both treat-
ments; in particular, cell proliferation-related
genes (CDC20Band histone-related genes) as
well asNEUROG1, which was shown to act
as a negative regulator of neocortical neuro-
genesis ( 56 ), were significantly (FDR < 0.05)
up-regulated by MIX N and down-regulated
by T3 (fig. S3G).
Furthermore, because BPA has been reported
to affect human brain development and behav-
ior by epidemiological, in vitro, and in vivo
evidence ( 57 – 60 ), we also probed its impact as
a single compound at the same concentration
at which it is present in 1X MIX N. For bothfetal progenitors and organoids (fig. S3, H
and I), the effect of MIX N, although show-
ing an expected partial overlap, extended
well beyond that of BPA alone (which deter-
mined 49 DEGs for fetal progenitors and
871 DEGs for organoids; FDR < 0.05, logFC >
0.5, logCPM > 0).
In addition to the molecular dissection of
the MIX N effects, the same organoids that
were profiled at the transcriptomic level after
chronic exposure were also subjected to im-
munofluorescence analysis to score the cellular
phenotypes and to investigate the MIX N impact
on the developmental processes of neuronal
differentiation. In particular, we stained for
KI67, which marks proliferating progenitors,
and DCX, a protein involved in neuronal mi-
gration that marks early neurons. The results
showed a significant (P< 0.05) increase of
KI67-positive cells for MIX N–exposed organo-
ids coupled with a decrease of DCX-expressing
cells (P< 0.05) for 1000X MIX N exposure,
thus suggesting an effect favoring neural
progenitor proliferation while hindering neu-
ronal differentiation (fig. S4, A to D), which is
in line with recent observations that hormo-
nal exposure affects the same developmental
processes that regulate neuronal progenitor
proliferation and neuronal maturation of
genetic mutations, increasing ASD vulner-
ability ( 61 ). This result was corroborated at
the level of gene expression, which showed that
a subset of genes related to neural differentia-
tion (DCX,SYP,MAP2, andRBFOX3) were
down-regulated (P< 0.05 forSYPandRBFOX3),
whereas cell proliferation genes (MKI67,CCNB1,
CDC20, andHMGB2) were up-regulated by
MIX N (fig. S4, E and F;P< 0.05 forMKI67).
This result is also in line with the observa-
tion that gene sets related to cell proliferation
were among the top positively enriched sets in
GSEA (“mitotic spindle,”“G2M transition”;
Fig. 2C). When the cellular phenotypic analysis
was applied to T3-treated samples, it displayed
the opposite effect of MIX N, both for pro-
liferation and maturation, whereas BPA par-
tially recapitulated MIX N effects only for
progenitor proliferation. For T3 and BPA, the
transcriptomics data were also in agreement
with the immunostaining results (fig. S4).MIX N alters hormonal pathways
Next,wesetouttoidentifythekeyregulators
that are responsible for the transcriptomic
phenotypes as an inroad to the molecular
pathways leading to the cellular and clinical
adverse effects. To this end, we performed a
master regulator analysis starting from the
dysregulation observed in the chronic MIX
N exposure setting, using recently released
data from the PsychENCODE consortium ( 62 )
integrated into a human brain–specific gene
regulatory network across genomic, transcrip-
tomic, and epigenomic layers. We leveragedCaporaleet al.,Science 375 , eabe8244 (2022) 18 February 2022 6 of 15
RESEARCH | RESEARCH ARTICLE