for transforming growth factor–b (TGFb) and
fibroblast growth factor (FGF) pathway members
(binomial test, P = 2.42 × 10−3). Notably, both
module 10 and module 3 include gain-associated
genes belonging to the TGFb and FGF pathways
(Fig. 3C). The association of gains with biologically
related genes across multiple enriched modules
suggests that there may be regulatory coordination and potential transcription factor (TF) cross-talk among these modules.
Consistent with this hypothesis, gain-enriched
modules exhibited significantly higher gene expression correlations with each other than with
other modules in the network (Wilcoxon rank
sum test, P < 1 × 10−15) (Fig. 4A). Moreover, gain-associated genes in enriched modules converge
on related biological functions (Fig. 4B). To identify regulatory signatures underlying the correlation of these modules, we predicted transcription
factor binding sites in all active promoters and
enhancers in our data set, including human
lineage gains. We then identified enriched TF
motifs in enhancers or promoters assigned to
each module. Many motifs were enriched in
promoters and enhancers assigned to the same
module as the transcription factor itself. Surprisingly, we also identified TF motifs enriched
across multiple modules. For example, SMAD
binding motifs were enriched in active promoters
in module 10, although SMAD transcription factors are not included in this module (BH permutation test, P = 7.92 × 10–3) (table S5). The observed
transcription factor binding site enrichment patterns suggest regulatory cross-talk among gain-enriched modules that may contribute to their
highly correlated expression.
Our results reveal a marked convergence of
human lineage epigenetic gains on common biological processes and regulatory pathways in
corticogenesis. Epigenetic gains are enriched in
modules important for neuronal proliferation,
cortical patterning, and the ECM. Moreover, gain-associated genes in each module are enriched for
similar conserved biological functions as all genes
in the entire module (table S4). These findings
suggest that many human lineage regulatory
changes operate within, and have potentially
modified, older regulatory mechanisms and developmental processes essential for building the
The epigenetic changes associated with these
conserved biological pathways also predominantly occur at sequences with ancestral regulatory
activity. The majority of human lineage gains
involve potential modification of promoters or
enhancers marked by H3K27ac in rhesus or mouse
cortex (fig. S11) (10). A smaller proportion of gains
may arise from co-option of ancestral regulatory
sequences active in noncortical tissues. Human
gains not marked in any of the 2 rhesus or 20
mouse tissues we examined may include de novo
regulatory functions arising on the human lineage. We note that epigenetic gains may be due
to genetic changes in humans that directly altered
regulatory functions, or they may reflect coordinated changes in cellular composition in the human
cortex compared with rhesus and mouse cortex.
Distinguishing between these two modalities of
evolutionary change will require functional analysis of the sequences underlying epigenetic gains
using mouse transgenic assays and humanized
mouse models. Such studies would also provide
insight into the biological relevance of the molecular changes described here.
The convergence of human regulatory innovations on developmentally related functions is
also consistent with the biological complexity of
the cortex. Neocortical development requires the
orchestration of spatially and temporally distinct
but biologically interconnected mechanisms. In
the context of this interdependency, it has been
postulated that human cortical evolution involved
coordinated changes in multiple processes during corticogenesis (3). For example, changes in
progenitor proliferation probably required concomitant changes in patterning and connectivity
to generate novel cortical functions (1). The inventory of human lineage regulatory changes
that we identified provides the means to evaluate
this hypothesis and dissect the genetic mechanisms underlying the evolution of the human
REFERENCES AND NOTES
1. D. H. Geschwind, P. Rakic, Neuron 80, 633–647 (2013).
2. M. L. Schwartz, P. Rakic, P. S. Goldman-Rakic, Proc. Natl. Acad.
Sci. U.S.A. 88, 1354–1358 (1991).
3. P. Rakic, Nat. Rev. Neurosci. 10, 724–735 (2009).
4. P. Rakic, A. E. Ayoub, J. J. Breunig, M. H. Dominguez, Trends
Neurosci. 32, 291–301 (2009).
5. S. Prabhakar, J. P. Noonan, S. Pääbo, E. M. Rubin, Science 314,
6. J. A. Capra, G. D. Erwin, G. McKinsey, J. L. R. Rubenstein,
K. S. Pollard, Philos. Trans. R. Soc. London Ser. B 368,
7. G. Konopka et al., Neuron 75, 601–617 (2012).
8. J. Cotney et al., Genome Res. 22, 1069–1080 (2012).
9. J. Cotney et al., Cell 154, 185–196 (2013).
10. A. Rada-Iglesias et al., Nature 470, 279–283 (2011).
11. J. Ernst et al., Nature 473, 43–49 (2011).
12. T. S. Mikkelsen et al., Cell 143, 156–169 (2010).
13. P. Rakic, Science 183, 425–427 (1974).
14. P. Rakic, Science 241, 170–176 (1988).
15. P. Rakic, R. L. Sidman, J. Neuropathol. Exp. Neurol. 27,
16. T. Takahashi, T. Goto, S. Miyama, R. S. Nowakowski,
V. S. Caviness Jr., J. Neurosci. 19, 10357–10371 (1999).
17. See supplementary materials and methods on Science Online.
18. C. Y. McLean et al., Nat. Biotechnol. 28, 495–501 (2010).
19. A. Visel, S. Minovitsky, I. Dubchak, L. A. Pennacchio, Nucleic
Acids Res. 35 (suppl. 1), D88–D92 (2007).
20. A. Visel et al., Cell 152, 895–908 (2013).
21. J. R. Dixon et al., Nature 485, 376–380 (2012).
22. P. Langfelder, S. Horvath, BMC Bioinformatics 9, 559
23. BrainSpan Atlas of the Developing Human Brain;
24. J. C. Quinn et al., Dev. Biol. 302, 50–65 (2007).
25. F. M. Vaccarino et al., Nat. Neurosci. 2, 246–253 (1999).
26. A. L. Pearlman, A. M. Sheppard, Prog. Brain Res. 108, 119–134
This work was supported by NIH grants GM094780 (to J.P.N.),
DA023999 (to P.R.), NS014841 (to P.R), and F32 GM106628 (to
D.E.); a Brown Coxe Fellowship in the Medical Sciences (to J. Y.);
and an NSF Graduate Research Fellowship (to S.K.R.). Human
tissue was provided by the Joint Medical Research Council
(UK)/Wellcome Trust (grant 099175/Z/12/Z) Human
Developmental Biology Resource (HDBR) ( http://hdbr.org). The
human tissues used in this study are covered by a material
transfer agreement regarding their transfer, but tissues may be
requested directly from the HDBR. We thank S. Mane, K. Bilguvar,
S. Umlauf, and A. Lopez at the Yale Center for Genome Analysis for
sequencing data; the members of the BrainSpan consortium for
providing human brain transcriptome data to the research
community; N. Carriero and R. Bjornson at the Yale University
Biomedical Performance Computing Center for computing support;
T. Nottoli and C. Pease at the Yale Animal Genomics Service for
generating transgenic mice; and S. Wilson and M. Horn for
veterinary care of nonhuman primates. All ChIP-seq data are
available through the Gene Expression Omnibus under accession
Materials and Methods
Figs. S1 to S12
Tables S1 to S5
8 September 2014; accepted 4 February 2015