this network to infer the transcription factors
(TFs) whose altered activity (inferred from
their downstream output and hence aggre-
gating both their transcriptional or post-
transcriptional regulation) was most likely
mediating the impact of MIX N. The 92 reg-
ulators pinpointed by this analysis include up-
regulation ofSOX9, which is known to control
neurogenesis and astrogenesis during devel-
opment ( 63 – 66 ); down-regulation ofMEF2C,
which regulates multiple neurodevelopmen-
tal processes and whose haploinsufficiency is
causatively linked to ASD ( 67 – 69 ), and down-
regulation of the TH–dependent factorKLF9,
which plays a key role in neurogenesis ( 70 , 71 )
(fig. S5, A and B).
To further characterize how different endo-
crine pathways were affected by MIX N, we
intersected both the DEGs (FDR < 0.05) and
the master regulator TFs with genes related
to specific hormonal pathways, selected on the
basis of their relevance in the context of
neurodevelopment and endocrine disruption
( 33 , 72 – 75 ). Among the positive enrichment
results, estrogen and thyroid-related genes
were the most enriched at the level of DEGs
and TFs, respectively (fig. S5, C and D). More-
over, plotting the expression of all the endo-
crine-relevant DEGs and TFs showed MIX
N–dependent patterns of modulation for both
(fig.S5,EandF).Finally,withinthislandscape
of endocrine-related transcriptional dysregu-
lation, the Genomatix Pathway System (GePS)
program was used to identify critical hubs of
disruption (Fig. 4 and table S5). From the
MIX N targets (both DEGs and upstream TFs),
65 were found to interact with each other
using the expert level within GePS as the
minimum evidence level, with most DEGs
within the network found to be related to TH
( 43 ), estrogen ( 40 ), and peroxisome proliferator-
activated receptor (PPAR) ( 36 ), in line with the
PsychENCODE network–based analysis showing
the thyroid and estrogen pathways as the most
enriched ones.MIX N disrupts TH signaling in developmental
in vivo models
Having determined the impact of MIX N in
pathophysiologically relevant human models
in vitro, we extended our analysis in vivo to
define the impact of the mixture in regulatory
toxicology-relevant whole-organism models
(Fig. 5, A to H). We focused on the TH axis as
an example of convergent EDC-mediated
dysregulation, given our in vitro findings, its
essential roles in brain development, and previ-
ous epidemiological and experimental evidence
identifying the main chemical classes present
in the mixture as TH disruptors ( 76 ).The TH-disrupting capacity of the mixture
was investigated using theXenopuseleuther-
oembryonic thyroid assay (XETA), an assay
that probes TH disruption at multiple levels
(synthesis, transport, and metabolism) on
the basis of TH sensitivity and metabolic
competence of early-stageXenopus laevis
embryos. One-week-old transgenic tadpoles,
harboring a green fluorescent protein (GFP)
thyroid-responsive element reporter, were
exposed for 72 hours to T3 (5 nM) and MIX N
(Fig. 5A and fig. S6A). MIX N altered TH
activity in the head, resulting in a significant
(10X:P= 0.0210; 100X:P< 0.0001; 1000X:P<
0.0001) reduction in fluorescence at concen-
trations 10X, 100X, and 1000X (Fig. 5, B and
C). Next, we investigated neural gene expres-
sion in tadpoles exposed to MIX N for 3 days
by reverse transcription quantitative poly-
merase chain reaction (RT-qPCR) and found
that, in line with in vitro results, MIX N had
already decreased expression of the TH-
dependent TFklf9at 1X concentration, whereas
it decreased expression of the TH transporter
oatp1c1at 10X and increased the expression
of the TH transportermct8at 1000X. In the
high-dose range (100X and 1000X), mRNA
concentrations of brain-derived neurotrophic
factor (bdnf), a factor vital for synaptogenesis,
were also significantly up-regulated by MIX NCaporaleet al.,Science 375 , eabe8244 (2022) 18 February 2022 7 of 15
Thyroid
Estrogen
Androgen
Corticoid
PPAR
Progesterone
RetinoicFig. 4. DEGs and TFs identified in HFPNSCs and COs after MIX N exposure are linked to multiple endocrine pathways.Regulatory pathways were generated
by the Genomatix GePS program that connects differentially expressed genes. Box sizes indicate absolute fold change, with red lettering showing up-regulation
and blue showing down-regulation in organoids. Within this network, of the multiple endocrine pathways investigated, we found the following order of hormonal targets
in the network: thyroid > estrogen > PPAR > progesterone > androgen > corticoid > retinoic, reflecting the overall number of DEGs identified.
RESEARCH | RESEARCH ARTICLE