cortex compared with mouse cortex (fig. S4, A
To identify promoters and enhancers showing
quantitative epigenetic gains in the human genome
versus both the rhesus and mouse genomes, we
compared the level of H3K27ac or H3K4me2 signal
in replicating human peaks to the signals at corresponding orthologous sites in the other two species (9, 17) (fig. S5). Human gains were called on
the basis of an increase in H3K27ac or H3K4me2
signal compared with all rhesus and mouse data
sets for each mark (17). It is possible that we may
be overestimating gains at 7 p.c.w., due to the
lack of an early developmental stage in rhesus.
However, this concern is mitigated by our inclusion of a comparable mouse time point and our
requirement that each human site exhibit an
epigenetic gain compared with all mouse and
rhesus time points and tissues. In total, 8996
nonoverlapping enhancers and 2855 promoters
show epigenetic gains in humans (Fig. 1B). To
assess the robustness of these gains, we compared epigenetic signals at the orthologous human and mouse genomic locations for 77 human
gains by chromatin immunoprecipitation (ChIP)–
quantitative polymerase chain reaction using additional biological replicates (fig. S6, A and B).
Sixty-seven of these sites (87%) showed a gain in
humans, supporting the reproducibility of the
epigenetic gain calls from our genome-wide analysis (17). We then explored this high-confidence
set of gains to obtain insight into their origins
and relevance to human cortical evolution.
We first considered whether epigenetic gains
could be attributed to human-specific sequence
changes. Forty-eight highly conserved noncoding
regions displaying accelerated evolution in humans
exhibit increased H3K27ac or H3K4me2 in the
human cortex (table S1) (5, 6). However, gains in
general do not show increased rates of human-specific sequence change, suggesting that the
majority of our gains cannot be identified by
sequence acceleration alone (table S1).
In light of this result, we examined epigenetic
gains at known human enhancers active in the
embryonic forebrain to determine whether gains
reveal changes in regulatory function (19) (table S1).
In a proof-of-principle experiment, we compared
the activities of a human forebrain enhancer
exhibiting a gain and its rhesus ortholog using
a mouse embryonic transgenic enhancer assay
(20). The human enhancer drove reproducible
reporter gene expression in two telencephalon
domains: a wide caudal-dorsal domain and a
caudal-ventral stripe (Fig. 1C). The rhesus ortho-
log drove qualitatively weaker reporter gene ex-
pression in a similar caudal-dorsal domain but
did not drive reproducible activity in the human
caudal-ventral domain. Upon sectioning, we de-
termined that the dorsal domain was restricted
to the neocortex, whereas the human ventral do-
main corresponded to the caudal ganglionic
eminence (fig. S7C).
We also searched for genomic regions with a
high density of enhancers or promoters exhibiting gains. We used previously defined maps of
long-range genomic interactions to demarcate
putative regulatory domains maintained across
tissues and species (17, 21). This analysis revealed
genes within topologically delimited domains
that are hotspots of epigenetic gains (fig. S8, A to
D, and table S2). We identified 301 genes within
a gain-enriched hotspot that included at least
one gene with a promoter gain, notably TGFb R3,
COL13A1, EPHA2, and LMX1B.
To obtain global insights into biological pathways associated with human lineage epigenetic
gains, we integrated gains with gene coexpression network analyses (22). We generated a coexpression network using public RNA sequencing
data from multiple neocortical areas spanning
8 to 15 p.c.w., which includes the periods of corticogenesis in which we mapped H3K27ac and
H3K4me2 signatures (Fig. 2A, fig. S9A, and table
S3) (23). This network consists of 96 modules,
each of which is a set of genes showing highly
correlated expression across multiple neocortical
regions and developmental stages. Genes in each
module may be co-regulated and may participate
in related biological processes. Hub genes are defined as genes with connectivity values in the top
5% for each module, suggesting that they include
important regulators that drive correlated gene
expression. Epigenetic gains at promoters were
directly assigned to their target genes, whereas
gains at enhancers were assigned on the basis of
their proximity to annotated genes (17, 18).
We used permutation analysis to identify modules significantly enriched in human lineage gains
at enhancers or promoters (fig. S9, B and C) (17).
Seventeen modules are enriched for H3K27ac or
H3K4me2 gains in at least one human developmental stage. Overall, gains are consistently
enriched in modules containing genes associated with biological processes crucial for cortical
development (table S4). For example, module 3
(Fig. 2B) is enriched for human lineage H3K27ac
enhancer gains that are associated with genes
implicated in neuronal progenitor proliferation.
Gene ontology categories showing significant
enrichment include neuronal differentiation (
binomial test, P = 2.13 × 10–4) and neuron fate commitment (binomial test, P = 3.67 × 10−4) (Fig. 2D).
Epigenetic gains in this module are associated
with genes critical for cortical development, including PAX6, GLI3, and FGFR1. Each of these is a
hub gene, consistent with their known contributions to fundamental processes in corticogenesis.
Notably, PAX6 controls cortical cell number by
regulating cell cycle exit of neural progenitor cells,
and Pax6-null mice have a depleted progenitor
Fig. 2. Identifying modules of coexpressed genes enriched for epigenetic gains in human corticogenesis. (A) Schematic illustrating integration of epigenetic gains into coexpression networks. (B) Coexpression module enriched for H3K27ac enhancer gains. Genes associated with gains are highlighted,
and genes representative of the biological enrichments associated with the module are labeled. The
module was rendered using multidimensional scaling (17). (C) Fold enrichment of H3K27ac enhancer
gains at each human time point in this module. *P < 0.01 (BH-corrected permutation). (D) Gene
ontology enrichments for genes associated with gains in this module. P values were calculated using a
binomial test in DAVID (the Database for Annotation, Visualization and Integrated Discovery) (17).