Epigenomics of Retinal Development in Mice and Humans

Neuron

Previews daughter lineages, but the work from Hippenmeyer and colleagues provides a solid foundation on which to build. REFERENCES Ables, J.L., Breunig, J.J., Eisch, A.J., and Rakic, P. (2011). Nat. Rev. Neurosci. 12, 269–283. Beattie, R., Postiglione, M.P., Burnett, L.E., Laukoter, S., Streicher, C., Pauler, F.M., Xiao, G., Klezovitch, O., Vasioukhin, V., Ghashghaei, T.H., et al. (2017). Neuron 94, this issue, 517–533.

Breunig, J.J., Haydar, T.F., and Rakic, P. (2011). Neuron 70, 614–625.

sheva, E., Insolera, R., Chugh, K., et al. (2014). Cell 159, 775–788.

Breunig, J.J., Levy, R., Antonuk, C.D., Molina, J., Dutra-Clarke, M., Park, H., Akhtar, A.A., Kim, G.B., Hu, X., Bannykh, S.I., et al. (2015). Cell Rep. 12, 258–271.

Klezovitch, O., Fernandez, T.E., Tapscott, S.J., and Vasioukhin, V. (2004). Genes Dev. 18, 559–571.

Bridges, C.B., and Brehme, K.S. (1944). The Mutants of Drosophila Melanogaster (Carnegie Inst. Wash. Publ. 552). Gao, P., Postiglione, M.P., Krieger, T.G., Hernandez, L., Wang, C., Han, Z., Streicher, C., Papu-

Li, X., Newbern, J.M., Wu, Y., Morgan-Smith, M., Zhong, J., Charron, J., and Snider, W.D. (2012). Neuron 75, 1035–1050. Rasin, M.R., Gazula, V.R., Breunig, J.J., Kwan, K.Y., Johnson, M.B., Liu-Chen, S., Li, H.S., Jan, L.Y., Jan, Y.N., Rakic, P., and Sestan, N. (2007). Nat. Neurosci. 10, 819–827.

Epigenomics of Retinal Development in Mice and Humans Nicolas Lonfat1 and Connie Cepko1,2,* 1Departments

of Genetics and Ophthalmology Hughes Medical Institute Harvard Medical School, Boston, MA 02115, USA *Correspondence: [email protected] http://dx.doi.org/10.1016/j.neuron.2017.04.029 2Howard

In this issue of Neuron, Aldiri et al. (2017) present an analysis of epigenetic changes during retinal development, and use these data to probe reprogramming of retinal iPSC cells, as well as the origin of retinoblastoma cells. The development of the retina into a highly organized and conserved structure, composed of six types of neurons as well as Mueller glia, is a process tightly regulated in time and space. These different cell types are born in a stereotypical order, evolutionarily conserved across species. To achieve such diversity, lineage analyses have shown that there are multipotent proliferative cells, termed retinal progenitor cells or RPCs, which change over time. These analyses have also shown that some of these RPCs are specified to make only a single type of neuron, or a specific pair of very different types of neurons, in terminal divisions (Cepko, 2014). In addition to lineage studies, much attention has been focused on the roles of transcription factors in the determination of the fate of retinal cell types, and in cataloguing gene expression, using all manner of RNA profiling methods, including singlecell RNA-seq. These previous studies have set the stage for analyses of the role of chromatin in regulating gene expression in retinal development and disease.

To decipher developmental gene regulatory programs, extensive efforts have been made to map cis-regulatory elements, particularly enhancers, that control spatio-temporal gene expression. Advances in genome-wide methods that map DNA and histone modifications, transcription factor binding, chromatin accessibility, and genome organization, along with advances in computational methods, have allowed an appreciation of the features that characterize hundreds of thousands of enhancers. The dynamic nuclear organization of the genome has an important role in transcription regulation during development, with topologically associated domains (TADs) emerging as fundamental structural units confining interactions between dispersed enhancers and their target genes (Dixon et al., 2015; Spitz 2016). Since the chromatin landscape in which transcription factors operate is crucial to their ability to perform their regulatory roles, epigenetic modifications acquired during differentiation can lead to restrictions in the ability of

420 Neuron 94, May 3, 2017 ª 2017 Elsevier Inc.

transcription factors to affect cell fate, or other events. Aldiri et al. (2017) comprehensively characterized the dynamics of epigenome and transcriptome across multiple stages of normal retinal development in the mouse and in the human (Figure 1). DNA methylation was previously shown to be inversely correlated with transcriptional activity in the brain (Mo et al., 2015), and Aldiri et al. (2017) profiled and analyzed these changes with respect to RNA levels in the retina. While changes in DNA methylation significantly accompanied changes in expression during murine development, further analyses on validated retinal gene sets showed that only a small subset (3%–38%) of those developmentally regulated genes had an inverse correlation. As expected, genes associated with retinal differentiation were upregulated during development, while progenitor and cell-cycle genes were silenced, and DNA methylation was more often associated with the former. To characterize other epigenetic changes,

Neuron

Previews

Figure 1. Dynamics of Retinal Epigenomes and Transcriptomes during Development and Disease Aldiri et al., (2017) characterized the transcriptional and epigenetic changes that occur during development of the retina in mice and humans. They also profiled the epigenomes of iPSCs derived from the retina and those from retinoblastoma cells. Chromatin architecture was analyzed at different stages of development, at different scales, focusing on the organization of histone marks and DNA methylation, interactions among genes in the context of TADs, and analysis of the overall nuclear structure within rods.

Aldiri et al. (2017) used ChIP-seq to map several histone marks, which have been associated with active or repressive chromatin. The temporal changes of epigenetic marks at the promoters and within gene bodies correlated with changes in gene expression. Aldiri et al. (2017) also described several bivalent or poised promoters in genes that became active during late stages, and they found that repressive marks and DNA methylation were associated with the silencing of certain genes during development, but not all, suggesting additional repression mechanism(s). To integrate these multiple epigenetic datasets, they used the algorithm chromHMM that models combinations of chromatin marks (Ernst and Kellis 2012). They identified 11 chromatin states, such as actively transcribed, enhancer, bivalent promoter, polycomb-repressed, etc., and used them to annotate the

genome, enabling the analyses and comparisons of the epigenetic changes during retinal development and across species. Interestingly, they identified genes that were upregulated, even though they showed a repressive state. By examining these genes in isolated rods, they found that they are epigenetically repressed in rods, while they are expressed in other, non-rod cell types. Using chromatin state modeling and changes in DNA methylation, they found that 40% of upregulated genes and 36% of downregulated genes had at least one type of correlated epigenetic change, more prevalent in differentiation genes than in progenitor genes. Aldiri et al. (2017) overlapped histone marks with regions of open chromatin, as defined by ATAC-seq, to identify a few hundred poised and active enhancers with high dynamic profiles between embryonic and adult stages. Since long-

range interactions between enhancers and their target promoters tend to occur within TADs that are generally stable across cell types, Aldiri et al. (2017) used TAD boundaries from the mouse cortex to map putative enhancers to their target genes. They identified clusters of enhancers, or super-enhancers, transitioning between chromatin states during development and that were associated with rod genes, G2/M or retinal progenitor cell genes. Correlated changes in gene expression were more frequent for genes within 100 kb of a super-enhancer. Using the chromHMM data, they also classified the TADs as either active, polycomb repressed, null chromatin, or constitutive chromatin. TADs that had no marks, or were in constitutive heterochromatin at early stages, remained stable at later stages, but some TADs transitioned between active and polycomb or Neuron 94, May 3, 2017 421

Neuron

Previews heterochromatin states (10% become active and 22% become repressed). Rod nuclei show areas of dense heterochromatin within the center of the nucleus, with two concentric rings of facultative heterochromatin and euchromatin surrounding the dense center. This unique inverted arrangement is seen among nocturnal species and is thought to allow light to channel more effectively through the layer of rod nuclei toward the lightsensing outer segments (Solovei et al., 2009). Aldiri et al. (2017) used DNA FISH to explore the relationship of genes with different chromatin states to the nuclear structure of rods. Interestingly, the celltype specificity or chromatin state did not predict the nuclear location, e.g., genes expressed in the inner ear were localized to the euchromatin. Some progenitor genes with H3K27me3 repressive marks were found in the facultative heterochromatin and other retinal genes that are not expressed in rods and that are also H3K27me3 repressed were found in the euchromatin, along with rod genes previously shown to be localized there. Although the subnuclear localization could not be correlated to the epigenetic states, they did see that the localization was consistent across nuclei, suggesting that the 3D organization is precisely regulated. The data from Aldiri et al. (2017) provide a major contribution to the field that will enable future studies on the role of chromatin in a variety of developmental and other processes. As a first application of these datasets, Aldiri et al. (2017) analyzed the epigenetic signatures of retinoblastoma (Rb) tumors, as well as the signatures of iPS cells derived from the retina and a non-retinal cell type (Figure 1). Retinoblastomas are pediatric tumors of the retina that occur in individuals who inherit a loss of function allele in the Rb1 gene. The window of tumorigenesis is restricted to early childhood, suggesting an event that takes place early in retinal development, e.g., perhaps within the mitotic population. Several mouse models of this disease have been created using inactivation of Rb1 and related genes. The cell type(s) of origin for both human and mouse Rb tumors has been an area of interest. The approach to this question has been to analyze both morphology and gene expression in 422 Neuron 94, May 3, 2017

freshly explanted human and murine tumors, with the hope that a close resemblance of an Rb tumor and a particular retinal cell type will reveal the cell type of origin (Dyer 2016). The gene expression profiles have revealed a mix of markers specific to different retinal cell types, including progenitor cells, photoreceptors, and interneurons, even within a single cell (McEvoy et al., 2011). Such a mixture of gene expression patterns also occurs in normal development. Regarding morphology, one of the most striking observations was that some murine Rb tumors have extensive neural processes and synapses, which strongly resemble those of mature horizontal cells, a type of interneuron. Nonetheless, these cells divide. The current study sought to add to these previous characterizations by analyzing the epigenetic state of human and murine Rb tumors, while also performing additional RNA expression analyses. Both mouse and human Rb tumors again showed a mix of gene expression patterns, with the human tumors showing more photoreceptor markers and the murine tumors showing more markers of interneurons, in keeping with the previous studies. The epigenomes of the tumors were found to match the developmental stages likely corresponding to the presumed times of origin, namely the period of photoreceptor genesis for humans and that of inner retinal neuron genesis for mouse. One approach for the treatment of retinal degeneration is to replace lost photoreceptors by engrafting rod and/or cone cells derived from iPSCs. During reprogramming of somatic cells into iPSCs, the epigenetic memory acquired during development often persists and needs to be erased and remodelled to confer pluripotency (Nashun et al. 2015). Cells reprogrammed from rod photoreceptors (r-iPSCs) are more efficient at producing differentiated retinal cells than are fibroblast iPSCs (f-iPSCs) or ESCs (Hiler et al., 2015), raising a question regarding differences in epigenetic marks in iPSCs originating from these two cell types. Aldiri et al. (2017) found that repressive marks, H3K27me3 and H3K9me3, were the most differential epigenetic marks between r-iPSCs and f-iPSCs, suggesting that histone modifications have a greater role than DNA methylation in

epigenetic memory, perhaps as a barrier to reprogramming. They also show that, contrary to their expectation, few rod genes retained epigenetic memory in r-iPSCs. As mentioned above, rods from nocturnal species have a peculiar heterochromatin structure thought to be optimized for light penetration. Aldiri et al. (2017) suggest that the sequestration of developmental genes in the facultative chromatin could serve as a silencing mechanism. Two recent studies also addressed the chromatin structure of photoreceptors, focusing on the DNA methylation and chromatin accessibility of mature rods, as well as of cones, the photoreceptor type used in the brighter light conditions of the day (Hughes et al., 2017; Mo et al., 2016). They found that rods have a very closed chromatin architecture compared to cones and compared to other cell types, which might be related to the unique nuclear architecture of rods (Hughes et al., 2017). Another striking observation was the enrichment of hypomethylated DNA specifically in rods. These areas were in closed chromatin, where they may be protected from methyltransferases (Mo et al., 2016). The hypomethylated regions have been proposed to mark vestigial enhancers active earlier in development (Mo et al., 2016; Hughes et al., 2017). As these previous studies were confined to mature cell types, the data provided by Aldiri et al. (2017) can now be used to investigate this hypothesis. In contrast to nocturnal animals, diurnal mammals harbor a conventional chromatin architecture in rods (Solovei et al., 2009). Perhaps in keeping with this, Aldiri et al. (2017) found a strong conservation of epigenetic features between mice and humans, except in DNA methylation, which was the least conserved. Since Aldiri et al. (2017) found that H3K9me3, enriched in heterochromatin, contributed more than 10-fold to the epigenetic memory of r-iPSCs, compared to other epigenetic marks, it would also be interesting to analyze the epigenetic memory of human r-iPSCs, to compare to those of murine r-iPSCs. Differences may reflect the differences between the epigenetics of human and murine rods. The genome is organized in a highly hierarchical structure and chromatin

Neuron

Previews architecture undergoes extensive reorganization during development in association with gene expression (Dixon et al., 2015). At birth, rod nuclei exhibit a conventional chromatin architecture in mice, with the inverted organization developing over the next several weeks or months (Solovei et al., 2009). Further study of the changes in the three-dimensional organization of the rod genome during different stages of differentiation using FISH and comprehensive mapping of TADs, via chromosome conformation capture methodologies such as Hi-C (Lieberman-Aiden et al., 2009; Dixon et al., 2015), would be of interest. This would allow further investigation of the role of nuclear organization in transcriptional regulation, and potentially, vice versa. Previous studies have shown that the genome is organized into A and B compartments, which correspond to relatively active and inactive regions respectively, with TADs forming the basis of these higher-order chromatin structures (Lieberman-Aiden et al., 2009; Dixon et al., 2015). A combination of retina-specific nuclear organization datasets and the epigenome data from Aldiri et al. (2017) would enable a study of the dynamics of compartment A/B patterns during retinal development. In addition, such an anal-

ysis would allow one to ask about changes in the inter- and intra-TAD interactions between enhancer and promoters during development. The multiple datasets generated by Aldiri et al. (2017) provide a valuable resource for deciphering the gene regulatory networks controlling retinal development, e.g., the many putative enhancers suggested by their work will enable a more efficient route toward discovery of the cis-regulatory elements and cognate transcription factors for specific genes. These data may improve our use of 3D retinal organoids and/or iPSCs for deriving cell-based therapies for retinal degenerations. In addition, their data provide a starting point for an analysis of chromatin changes in specific types of retinal progenitor cells and their progeny, as well as potentially provide a window into some of the higher-order changes in the genome that may drive the temporal progression of retinogenesis.

Cepko, C. (2014). Nat. Rev. Neurosci. 15, 615–627. Dixon, J.R., Jung, I., Selvaraj, S., Shen, Y., Antosiewicz-Bourget, J.E., Lee, A.Y., Ye, Z., Kim, A., Rajagopal, N., Xie, W., et al. (2015). Nature 518, 331–336. Dyer, M.A. (2016). Trends Mol. Med. 22, 863–876. Ernst, J., and Kellis, M. (2012). Nat. Methods 9, 215–216. Hughes, A.E.O., Enright, J.M., Myers, C.A., Shen, S.Q., and Corbo, J.C. (2017). Sci. Rep. 7, 43184. Lieberman-Aiden, E., van Berkum, N.L., Williams, L., Imakaev, M., Ragoczy, T., Telling, A., Amit, I., Lajoie, B.R., Sabo, P.J., Dorschner, M.O., et al. (2009). Science 326, 289–293. McEvoy, J., Flores-Otero, J., Zhang, J., Nemeth, K., Brennan, R., Bradley, C., Krafcik, F., Rodriguez-Galindo, C., Wilson, M., Xiong, S., et al. (2011). Cancer Cell 20, 260–275. Mo, A., Mukamel, E.A., Davis, F.P., Luo, C., Henry, G.L., Picard, S., Urich, M.A., Nery, J.R., Sejnowski, T.J., Lister, R., et al. (2015). Neuron 86, 1369–1384. Mo, A., Luo, C., Davis, F.P., Mukamel, E.A., Henry, G.L., Nery, J.R., Urich, M.A., Picard, S., Lister, R., Eddy, S.R., et al. (2016). eLife 5, e11613. Nashun, B., Hill, P.W., and Hajkova, P. (2015). EMBO J. 34, 1296–1308.

REFERENCES Aldiri, I., Xu, B., Wang, L., Chen, X., Hiler, D., Griffiths, L., Valentine, M., Shirinifard, A., Thiagarajan, S., Sablauer, A., et al. (2017). Neuron 94, this issue, 550–568.

Solovei, I., Kreysing, M., Lanctoˆt, C., Ko¨sem, S., Peichl, L., Cremer, T., Guck, J., and Joffe, B. (2009). Cell 137, 356–368. Spitz, F. (2016). Semin. Cell Dev. Biol. 57, 57–67.

What Do Sensory Organs Tell the Brain? Avner Wallach,1 Satomi Ebara,2 and Ehud Ahissar3,* 1Department of Cellular and Molecular Medicine, Faculty of Medical Science and Department of Physics, Faculty of Science, University of Ottawa, Ottawa, ON K1N 6N5, Canada 2Department of Anatomy, Meiji University of Integrative Medicine, Kyoto 629-0392, Japan 3Department of Neurobiology, the Weizmann Institute of Science, Rehovot 76100, Israel *Correspondence: [email protected] http://dx.doi.org/10.1016/j.neuron.2017.04.031

Understanding how perception emerges depends on the understanding of sensory acquisition by sensory organs. In this issue of Neuron, Severson et al. (2017) present a brilliant leap towards understanding active sensory coding by mechanoreceptors. Evolution hides its secrets. Comparative studies trying to trace the birth of a new function in the brain usually fail. Instead, neuronal precursors of almost every studied function is found in earlier species. On this background, sensory organs

stand out. Dissecting a sensory organ seems like opening a treasure box for a comparative scientist. The distinction between different receptor types is indubitable, and the differences between one type of sensory organ to the other are

overwhelming. Such dramatic differences must reflect significant functional differences—different evolutionary solutions to different environmental challenges. What stands behind these differences? What kinds of information do the sensory

Neuron 94, May 3, 2017 ª 2017 Elsevier Inc. 423