Abstract
Maintenance of cellular identity is essential for development and homeostasis and problems in such control can have dire consequences such as organ malformation or cancer. At the molecular level, cell identity is determined by the coordinated activation and repression of defined sets of genes. The tumor suppressor L(3)mbt was shown to secure cellular identity in Drosophila larval brains by repressing germline specific genes, although the mechanisms remain elusive. Here we show that L(3)mbt safeguards the integrity of the somatic and germline tissues forming the Drosophila ovary. L(3)mbt mutation causes multiple developmental defects and simultaneous expression of original identity markers and new tissue-specific misexpression signatures. We demonstrate that the ovarian developmental defects are largely due to the ectopic expression of nanos, a key regulator of germline fate in the somatic cells of the ovary, rather than to the loss of cellular identity. Importantly, using genome-wide analysis we found that L(3)mbt exerts boundary activity between differentially regulated genes by binding to high occupancy insulator sites enriched at the borders of topological domains. Using nanos as an example, we propose that L(3)mbt maintains cell fate and specific gene expression programs by locally insulating repressed genes from closely apposed transcribed neighbors.
Introduction
Development requires tight control of gene expression as differentiating cells must express lineage-specific genes while repressing genes promoting other fates. Mechanisms ensuring the maintenance of cellular identity must be robust, as changes in cell fate are fairly uncommon in wild-type conditions, with only one documented case of regulated transdifferentiation being described in C. elegans (Jarriault et al. 2008). Rare cases of transdifferentiation have been observed in mutants affecting chromatin complexes, suggesting a role for chromatin structure in the maintenance of cellular identity (Tursun et al. 2011; Petrella et al. 2011). A notable example is given by mutations affecting the Drosophila lethal (3) malignant brain tumor (l(3)mbt) gene, which lead to malignant brain tumors that ectopically express germline-specific genes and have been proposed to cause soma-to-germline transformations (Gateff et al. 1993; Janic et al. 2010). Gene expression profiling of l(3)mbt brain tumors and L(3)mbt-depleted cultured somatic cells identified a group of upregulated genes known as the Malignant Brain Tumor Signature (MBTS) that is enriched for factors specifically expressed in germ cells (Georlette et al. 2007; Janic et al. 2010; Meier et al. 2012; Sumiyoshi et al. 2016). Mutations of germline-specific genes, including those impairing the piRNA factors piwi, aub, and vasa, as well as the translational repressor nanos, were found to suppress the neural overgrowth induced by loss of L(3)mbt (Janic et al. 2010). However, another study suggested that the Hippo growth control pathway is critical for tumor formation in l(3)mbt mutant brains, and questioned tissue transformation as the cause of tumorigenesis (Richter et al. 2011). Furthermore, our lab showed that strong l(3)mbt mutations cause a maternal, germline autonomous effect phenotype that precludes normal embryonic development including primordial germ cell formation (Yohn et al. 2003), suggesting a broader role for L(3)mbt in regulation of tissue identity.
L(3)mbt encodes a 1477 amino-acid protein that is ubiquitously expressed in Drosophila and is conserved from worms to humans. L(3)mbt harbors three MBT repeats that were shown to bind to mono- and possibly di-methylated lysine 20 of Histone 4 in vitro (Bonasio et al. 2010). L(3)mbt also contains a C2/HC Zinc-Finger domain and a carboxy-terminal region thought to mediate protein-protein interactions (Bonasio et al. 2010). Notably, L(3)mbt was purified in two non-enzymatic repressive chromatin complexes: the Drosophila RBF, E2F2, and Myb-interacting proteins (dREAM complex, also called Myb-Muv B) as well as the L(3)mbt-Interacting complex (LINT) (Lewis et al. 2004; Meier et al. 2012). DREAM is a large multi-subunit complex that controls gene expression throughout the cell cycle but also represses developmental genes. L(3)mbt associates at sub-stoichiometric levels with dREAM and is strictly found in its repressive forms (Georlette et al. 2007; Lewis et al. 2004). The LINT complex is composed of L(3)mbt, the novel transcriptional repressor Lint1, as well as the corepressor CoREST, and was shown to silence developmental genes in cultured cells (Meier et al. 2012). Interestingly, dREAM and LINT complexes repress overlapping sets of genes in somatic cells, including genes normally expressed in the germline.
Despite extensive biochemical studies, we still know little about which chromatin complex mediates L(3)mbt’s role in tissue identity. L(3)mbt binding sites were characterized in larval brains and embryonic cells by Chromatin Immuno-purification followed by sequencing (ChIP-seq, (Richter et al. 2011; Van Bortle et al. 2012)). These studies revealed that L(3)mbt binds approximately 3500 regions and is enriched at the promoters of repressed genes, suggesting a direct role in transcriptional repression. On the other hand, a large number of L(3)mbt binding sites overlaps with regions occupied by the insulator proteins CP190, BEAF-32 and to a lesser extent CTCF as well as Su(Hw), which led to an alternative model in which L(3)mbt works as an insulator accessory factor (Richter et al. 2011; Van Bortle et al. 2012; 2014).
To further understand how L(3)mbt secures tissue identity, we combined genetic and genomic approaches to characterize the functions of L(3)mbt in Drosophila melanogaster ovarian development. We find that L(3)mbt affects gene expression in a tissue-specific manner, in somatic cells of the ovary it represses germline genes and in the female germline it targets genes normally expressed in the testis and the nervous system. Mutant somatic ovarian tissues continue to express signatures of the tissue of origin, indicating that loss of L(3)mbt does not induce transdifferentiation. Consistent with a primary role of repressing the “germline program” in the soma, ectopic expression of a single gene, the translational repressor and key regulator of germline fate nanos, is largely responsible for aberrant soma development. Analysis of L(3)mbt associated proteins reveals a primary function for the LINT complex in the ovary. We observed that L(3)mbt bound regions required for regulation are high-occupancy insulator binding sites found at the boundaries of chromosomal domains. Finally, we demonstrate that L(3)mbt depletion results in loss of boundary between differentially-regulated head-to-head genes, allowing expression of genes incompatible with normal ovarian development.
Results
Sterility in l(3)mbt mutant females is associated with aberrant ovarian development
L(3)mbt was previously shown to be required for the development of the nervous system, and l(3)mbt mutant flies grown at restrictive temperatures (29°C) develop malignant brain tumors and die at larval stages (Gateff et al. 1993; Janic et al. 2010; Richter et al. 2011). We observed that, when grown at lower temperature (25°C), null l(3)mbt mutants were viable but fully sterile, indicating that L(3)mbt is critically required for germline development. In the wild type, each ovary is composed of 16 to 20 egg assembly chains called ovarioles (Fig. 1A). At the tip of each ovariole in a region called the germarium, germline stem cells (GSCs) divide asymmetrically to generate a new GSC and a differentiating daughter cell. The differentiating GSC daughter undergoes four rounds of mitosis with incomplete cytokinesis to form a sixteen-cell germline cyst. Sibling germ cells remain interconnected through cytoplasmic bridges called ring canals. Only one of the cyst germ cells develops into an oocyte, while the other 15 cells become supportive nurse cells. The somatic cells of the ovary play important roles in supporting oogenesis: they compose the GSC niche, promote cyst divisions and differentiation, and the follicle cells enclose and individualize egg chambers, contributing to oocyte-nurse cell differentiation. At the macroscopic level, l(3)mbt mutant ovaries were atrophied and adult females did not lay eggs. To characterize the ovarian phenotype in detail, we used antibodies against the germline marker Vasa and α-Spectrin (α-Spe), which labels the membranes of somatic cells and distinguishes between undifferentiated germ cells and connected cysts. Similar to the wild-type, l(3)mbt mutant germaria contained GSCs adjacent to the somatic niche (Supplemental Fig. S1A-B). In contrast to the wild type, mutant ovarioles contained fewer individualized egg chambers (1.35 egg chambers/ovariole in average vs. 6 in WT, p<10-4), which were highly abnormal and displayed several defects (Fig. 1C, quantified in J). These included extra-numerous germ / nurse cells (Fig. 1C), somatic follicle cells that failed to fully enclose germ cells (Fig. 1E, arrowhead), and accumulation of undifferentiated germ cells marked by branched fusomes, a structure stained by α-Spe that connects sibling germ-cells within early cysts (Fig. 1G, arrow). The majority of mutant egg chambers contained multiple oocytes, as indicated by oo18 RNA-binding protein (Orb, (Christerson and McKearin 1994)) staining and ring canal numbers, suggesting that l(3)mbt mutant ovarioles undergo abnormal egg chamber separation or fusion (Fig. 1H-I, arrowheads; Supplemental Fig. S1C-D). In rare cases (2%), we observed multiple germaria connected to the same aberrant egg chamber (Supplemental Fig. S1E-F), suggesting that ovariole fusion can also occur. Taken together, our results indicate that in addition to its previously reported requirement for brain and embryonic development, L(3)mbt is critical for proper ovary development.
L(3)mbt functions in ovarian somatic cells to safeguard ovary development
Previous experiments had shown that loss of l(3)mbt in germ cells caused developmental defects in the resulting embryos, but allowed oocytes to mature (Yohn et al. 2003). Thus, we wondered whether the gross abnormalities of mutant ovaries were indicative of a role for L(3)mbt in somatic cells of the ovary. We induced homozygous l(3)mbtGM76 mutant clones in the somatic ovarian cells by using the FRT-FLP system (Harrison and Perrimon 1993) under the transcriptional control of the Ptc-Gal4 or c587-Gal4 drivers, which drive expression in the somatic cells of the germarium (Hinz et al. 1994; Zhu and Xie 2003). L(3)mbt mutation in a subset of somatic cells was sufficient to perturb germline development (Fig. 2A) and led to egg chambers containing multiple oocytes (Fig. 2B). To further test a requirement for L(3)mbt in somatic cells, we generated an inducible UAS-l(3)mbt::myc transgene. Its expression, specifically in somatic ovarian cells using the tj-Gal4 driver, was sufficient to rescue the aberrant morphology of mutant ovaries (Fig. 2C-D and Supplemental Fig. S2A-C). Embryos laid by mutant females somatically expressing l(3)mbt::myc failed to hatch, in agreement with the described maternal requirement for L(3)mbt (Yohn et al. 2003). While most egg chambers developed normally in somatically-complemented l(3)mbt mutant ovaries, we noticed that 69% of ovarioles contained apoptotic egg chambers, identifying a germline autonomous role for L(3)mbt in egg chamber development (Supplemental Fig. S2C-D). Our results demonstrate that in addition to its germ cell autonomous role, L(3)mbt is required in the somatic tissues of the ovary to support normal oogenesis.
L(3)mbt mutant somatic cells simultaneously express somatic gonad and germline-specific genes
Previous studies suggested that loss of L(3)mbt resulted in a soma to germline transformation in the larval brain and cultured somatic cell (Janic et al. 2010; Georlette et al. 2007; Meier et al. 2012; Sumiyoshi et al. 2016). Therefore, we performed RNA-sequencing (RNA-seq) analysis to gain a genome-wide view of gene expression changes induced by loss of L(3)mbt in ovarian somatic cells in vivo. To distinguish between germline and somatic tissues of the ovary, we took advantage of tud maternal mutations (tudM), which give rise to progeny that lack germ cells and develop into adults devoid of germline ((Arkov et al. 2006; Smendziuk et al. 2015), see Supplemental File S1 and Materials and Methods). Comparisons between tudM; l(3)mbtGM76/l(3)mbtDf and tudM; l(3)mbtGM76/+ control ovaries identified 600 genes differentially expressed (p-value<0.05) in somatic cells of the ovary. Of these 459 were upregulated and 141 downregulated in mutant somatic ovarian tissues (Supplemental Table S1). 44 upregulated genes were shared with the 101 MBTS genes (Janic et al. 2010), and 115 out of the 681 genes found upregulated in ΔΔl(3)mbt Ovarian Somatic Cells (OSC; Sumiyoshi et al., 2016). These derepressed genes include piRNA pathway components and germline-specific genes such as nos, Pxt, vas, aub, tej, krimp, AGO3, and CG9925 (Fig. 3A, Supplemental Fig. S3A). The effect of l(3)mbt mutation on gene expression was more pronounced for repressed genes, while genes normally expressed in the soma showed only low fold-changes in the mutant (124/141 had a Log2 fold-change between −0.3 and −1). This suggested that multiple cell identities may coexist in l(3)mbt mutant tissues. Indeed, key somatic ovarian transcription factors such as Traffic Jam (TJ) and Zfh1, which are essential for gonad development and exclusively expressed in somatic cells (Leatherman and Dinardo 2008; Li et al. 2003), were still expressed in l(3)mbt mutants (Fig. 3B-C and Supplemental Fig. S3B-C). Despite the gross morphological abnormalities, we observed TJ expressing cells with epithelial character, which resembled wild-type follicle cells, surrounding germ cells in mutant ovaries. However, individual TJ-positive cells occasionally expressed the germline marker Vasa (Fig. 3C, arrows). As TJ and Vasa expressions are mutually exclusive in the wild type, our results indicate that l(3)mbt mutant somatic cells retain somatic and epithelial/follicular characteristics while ectopically expressing hallmark genes of the germline fate.
Expression of Nos is necessary and sufficient to cause ovarian defects
Aberrant growth of l(3)mbt brain tumors was shown to rely on the ectopic expression of nos, aub, and vasa (Janic et al. 2010). We therefore asked whether misexpression of any of these genes contributed to the ovarian defects observed in l(3)mbt mutants. We found that aub, l(3)mbt double mutant ovaries were phenotypically similar to l(3)mbt single mutant, containing many apoptotic cells and aberrant egg chambers with more than 16 germ cells (Supplemental Fig. S4A) (we were unable to test vas, l(3)mbt double mutants because of lethality). In contrast, nosBN/nosL7 mutations dramatically suppressed the l(3)mbt ovarian defects: nos, l(3)mbt double mutant ovaries contained late stages egg chambers, their ovarioles were phenotypically indistinguishable from wild type and 85% of double mutant egg chambers contained 16 germ cells and only one oocyte (Fig. 4C-E, n=88). In agreement with this, depletion of Nanos’ co-factor Pumilio in l(3)mbt mutant ovaries also suppressed the mutant phenotypes, with 81% of pum680, l(3)mbt double mutant ovarioles resembling wild-type morphology (Fig. 4F-H, n=181). To test if nos misexpression in somatic ovarian cells was sufficient to cause defects similar to those observed in l(3)mbt mutants, we expressed nos under tj-Gal4-Gal80ts control (McGuire et al. 2004) to restrict expression in the soma to adult stages. Nos somatic expression did not fully recapitulate the l(3)mbt phenotype but perturbed ovarian morphology resulting in poorly individualized egg chambers (Supplemental Fig. S4B). In contrast, ectopic expression of aub or vas in somatic ovarian cells did not yield a morphologically significant phenotype (Supplemental Fig. S4C, D). Together, these results suggest that ectopic expression of nos, but not aub or vas, is necessary and sufficient to perturb somatic ovarian development.
L(3)mbt functions through the LINT complex to secure ovary development
L(3)mbt has been associated with two chromatin complexes that repress developmental and germline genes: the dREAM and LINT complexes (Fig. 5A-B; (Georlette et al. 2007; Meier et al. 2012)). To determine the function of these complexes in ovarian development, we depleted somatic cells from E2f2 and Mip120, two core repressors of the dREAM complex. Mutant somatic clones for E2f2 (Fig. 5C) or ovaries deficient for mip120 (Fig. 5E) did not result in phenotypic aberrations reminiscent of those observed in l(3)mbt mutants. We also asked whether mip120 mutation affected L(3)mbt nuclear localization in the ovary, as previously described in salivary glands (Blanchard et al. 2014). Using the TdTomato::L(3)mbt fusion protein we observed L(3)mbt nuclear localization comparable to wild type in mip12067.9A.9 mutants (Supplemental Fig. S5B). Together, these results suggest that L(3)mbt’s critical function in somatic ovarian cells is independent of the dREAM complex.
Next we examined the role of the LINT complex as a mediator of L(3)mbt function. Since mutations in Lint1 had not been identified, we generated CRISPR-induced Lint1 alleles (Gratz et al. 2013; 2014). Lint11 deletes two cytosines (350 and 351), creating a premature stop codon at position 66/540 (Lint1-C) or 128/601 (Lint1-A, Fig. 5F). Homozygous mutant flies were viable and fertile at 25°C, and their ovaries developed normally although 2% of egg chambers contained misplaced or extra-numerous oocytes (Fig. 5G). However, when grown at 29°C, lint11 females were fully sterile, laying eggs that failed to hatch, and 15% of their ovarioles contained aberrant egg chambers reminiscent of those observed in l(3)mbt mutant (Fig. 5H). We conclude that L(3)mbt may function via the LINT complex for ovary development.
In the female germline, L(3)mbt is required to represses neuronal- and testis-specific genes
Our data suggest a role for L(3)mbt in suppressing germline specific genes in somatic tissues of the ovary. However, L(3)mbt is also autonomously required in the germline for embryo development (Yohn et al. 2003) and egg chamber survival (Supplemental Fig. S2D). To gain a genome-wide view of changes in gene expression induced by the germline loss of L(3)mbt, we performed RNA-seq analysis on embryos laid by l(3)mbt mutant or heterozygous mothers expressing the L(3)mbt::myc fusion in the somatic ovary. We collected early embryos prior to activation of the zygotic genome as the early embryonic RNA pool is exclusively composed of maternally provided transcripts and reflects the germline expression profile during oogenesis (Edgar and Schubiger 1986). Our analysis identified 878 differentially expressed genes (p<0.05), of which 342 were upregulated and 536 downregulated in embryos laid by mutant females (Supplemental Table S2). Interestingly, we observed only a limited overlap between the genes derepressed in l(3)mbt mutant somatic and germline ovarian tissues, with two thirds of genes being specifically upregulated in one of the two tissues (Fig. 6A). To better characterize the group of genes upregulated in the l(3)mbt mutant germline, we performed two-way hierarchical clustering to identify any tissue specific expression signatures. Two major groups were readily identified: one composed of genes highly expressed in neuronal tissues (brain, thoracicoabdominal ganglion (tag), larval CNS, eye, and head; Fig. 6B) and another comprising testis-specific genes. As in mutant somatic ovaries, two-thirds (365/536) of the downregulated genes had their expression level reduced by less than two-fold compared to the heterozygous controls (Supplemental Fig. S6). These results indicate that L(3)mbt’s function is not restricted to repressing the germline program, but that L(3)mbt regulates distinct sets of genes, in a tissue-specific manner.
L(3)mbt-bound regions are high-occupancy insulator sites
L(3)mbt has long been thought to act as a transcriptional repressor (Georlette et al. 2007; Meier et al. 2012). However, L(3)mbt-binding sites were recently found to overlap with insulator elements, suggesting that L(3)mbt may harbor insulating function (Richter et al. 2011). These models are not mutually exclusive, so we asked whether L(3)mbt bound regions show characteristics of insulator elements. We observed that in cultured Ovarian Somatic Cells (OSC), L(3)mbt binding sites coincide with open and transposon-accessible chromatin and are devoid of nucleosomes (Supplemental Fig. S7A-B). Using published datasets (Li et al. 2015; Van Bortle et al. 2014), we further determined that L(3)mbt distribution resembles the binding profiles of the canonical insulators CP190 and BEAF32, being enriched upstream of Transcription Start Sites (TSS) (Supplemental Fig. S7C) and associated with 5’ Nucleosome Free Regions that are flanked by nucleosomes containing the histone variant H2Av (Supplemental Fig. S7D) (Raisner et al. 2005; Soboleva et al. 2014). However, gene expression changes after CP190 (Ali et al. 2016) or Beaf32 depletion (Lhoumaud et al. 2014) do not parallel those observed after l(3)mbt knockdown (Meier et al. 2012) in Kc167 cells (Supplemental Fig. S7E-F). Further, our ATAC-seq data showed that nucleosome depletion at regions normally bound by L(3)mbt was not affected in ΔΔl(3)mbt OSCs (Supplemental Fig. 7B, (Sumiyoshi et al. 2016; Schep et al. 2015)). Therefore, the effect of L(3)mbt on gene repression is not solely mediated by the canonical insulators CP190 and BEAF32 and other factors must mediate local chromatin remodeling at L(3)mbt-binding sites independently of L(3)mbt.
In higher eukaryotes, chromosome three-dimensional structure is organized into topologically associated domains (TADs, (Sexton et al. 2012; Hou et al. 2012; Van Bortle et al. 2014)). Recent studies suggest that co-operative binding of insulators and additional architectural factors such as condensin and cohesin proteins defines insulator strength. Specifically, high-occupancy sites (bound by 7 or more factors) often cluster at TAD boundaries and exhibit robust insulator activities, while low-occupancy sites tend to regulate gene expression more locally (Van Bortle et al. 2014). To investigate binding of insulators and associated factors at L(3)mbt sites upstream regulated genes, we compared the previously generated matrix of architectural proteins with the genes derepressed in L(3)mbt depleted Kc167 cells (Van Bortle et al. 2014; Meier et al. 2012). On average, these functionally relevant L(3)mbt sites were occupied by 6.8 factors, notably including the insulators CP190 (88% of overlap), Chromator (83% overlap), and BEAF32 (69% overlap; Fig. 7A). Of note, Lint1 also occupies 60% of these high occupancy sites (Supplemental Table S3). Next, we investigated the distribution of derepressed genes in l(3)mbt mutant somatic ovarian cells in relation to TADs identified in OSC cells (Ulianov et al. 2016). We observed a striking positional enrichment for L(3)mbt-regulated genes at the borders of TADs (Fig. 7B). Notably, nos is located at the boundary of a TAD in OSC cells (Fig. 7C). We conclude that L(3)mbt bound regions required for proper gene regulation are preferentially high-occupancy insulator sites enriched at TAD boundaries.
L(3)mbt exerts boundary activity between differentially regulated head-to-head genes
The genome of Drosophila melanogaster is compact and around 32% of its genes are divergently-paired, i.e. in head-to-head orientation with less than 1 kb separating the adjacent TSS (Yang and Yu 2009). Insulator proteins are often enriched between divergent pairs of genes where they act as boundaries between differentially regulating genes (Nègre et al. 2010; Yang et al. 2012). To investigate an insulating role for L(3)mbt, we examined the orientation of repressed genes relative to their adjacent neighbors. In l(3)mbt mutant somatic ovarian cells, 66 out of 97 (69%) upregulated genes (Log2 fold change >1.5) were in head-to-head (HtH) orientation with the upstream adjacent gene (Fig. 7D, “opposite orientation”), which significantly differs from an expected random distribution (p<5x10-4). In contrast, genes directly downstream of derepressed genes were randomly oriented, with 50.6% of the pairs in a tail-to-tail conformation and 49.6% tail-to-head (p=0.99). Interestingly, we observed a comparable orientation bias for genes upregulated upon L(3)mbt depletion in other tissues (germline: 73% HtH, p<10-4; brain: 72% HtH, p<10-4 (Janic et al. 2010); and Kc167 cells: 68% HtH, p<10-4 (Meier et al. 2012)). Next, we asked whether L(3)mbt binds between head-to-head targets. Indeed, 186 out of 235 (79%) regulated head-to-head pairs affected in the somatic ovary contain L(3)mbt binding sites in their Upstream Intergenic Region (UIR) and/or 5’UTRs (Richter et al. 2011). The proportion of pairs containing L(3)mbt sites was of 37% (60/164) for other conformations, indicating an enrichment for L(3)mbt binding in between head-to-head targets.
Head-to-the genes share their UIRs, which include most of the transcriptional regulatory sequences. Consequently, close HtH genes (UIR<500bp) tend to be co-expressed (Yang et al. 2012), as confirmed by genome-wide analysis using gene expression data across developmental stages and cell lines (all HtH pairs; Fig. 7E). In contrast, we observed that the divergent HtH genes regulated by L(3)mbt in the germline or in somatic tissues showed no correlation of expression (Fig. 7E). This striking difference is not related to the length of UIRs as UIRs between L(3)mbt-regulated divergent pairs were similar to all HtH pairs, peaking around 250 base pairs (Supplemental Fig. S7G). As L(3)mbt binds to the UIRs of differentially-regulated, head-to-head genes and its loss causes mis-expression of such genes, our data favor a model in which L(3)mbt functions as a boundary factor between differentially-expressed closely apposed genes.
To test this model, we analyzed L(3)mbt occupancy in the UIR shared by nos and the adjacent HtH gene CG11779. Using the tdTomato::L(3)mbt fusion protein, we performed tissue-specific chromatin immunoprecipitation followed by quantitative PCR (ChIP-qPCR, Fig. 7F). In the germline, where both nos and CG11779 are expressed, L(3)mbt signal was not detected throughout the region spanning the shared UIR. However, in ovarian somatic tissues, where CG11779 is expressed but nos is not, L(3)mbt signal was robustly detected over the previously identified L(3)mbt-binding site. Given that loss of L(3)mbt in ovarian somatic cells leads to nos ectopic expression, our results support the hypothesis that L(3)mbt operates as an insulator to safeguard tissue identity.
Discussion
By combining developmental and molecular analysis, we show that l(3)mbt mutant ovaries develop aberrantly. L(3)mbt depletion does not result in transdifferentiation, but causes derepression of sets of genes incompatible with original cell identities. Notably the translational repressor nanos becomes derepressed in the somatic cells of the ovary and causes aberrant growth. Molecularly, we demonstrate that L(3)mbt functions through the LINT complex in the somatic ovary. Finally, we show that L(3)mbt exerts boundary activity between differentially-regulated genes transcribed in opposite orientation. We propose that this function of L(3)mbt mediates its role as guardian of cell identity.
Loss of L(3)mbt causes the ectopic expression of a number of genes including instructive regulators that, despite continued presence of initial cell fate determinants, are sufficient to interfere with original cellular function and affect tissue development and homeostasis. A prominent example is the germline specific gene nanos. Ectopic expression of Nanos has been shown to be detrimental in somatic tissues, including the larval brain (Janic et al. 2010) and somatic ovarian cells (this study). Importantly, in contrast to a previously suggested soma-to-germline transdifferentiation, our results favor the hypothesis that l(3)mbt mutation imbalances tissue-homeostasis whereby normally mutually exclusive lineage determinants are simultaneously expressed. In support, L(3)mbt depletion in neuronal, somatic ovarian, and germ cells does not lead to loss of original tissue-specific markers, but genes characteristic of other lineages are derepressed (Richter et al. 2011, this study). Moreover, our results suggest that the role of L(3)mbt is not solely restricted to prevent ectopic expression of germline genes, but instead L(3)mbt represses distinct, broader sets of genes in a tissue-specific manner. Therefore, we propose that L(3)mbt secures tissue identity by stabilizing gene expression profiles established during differentiation.
Our experiments demonstrate that nanos ectopic expression is necessary and sufficient to induce aberrant development of l(3)mbt mutant ovaries. These defects are likely not due to a direct interference at the transcriptional level but are rather caused by Nos’ function as a translational repressor. In support, we find that Pumilio, a sequence-specific translational repressor and co-factor of Nos, is also essential for the l(3)mbt ovarian phenotype. Nos was recently shown to modulate Pum RNA-binding and target-specificity in somatic S2 cells (Weidmann et al. 2016). Since Pum is ubiquitously expressed, we propose that ectopic Nos stabilizes Pum binding at target mRNAs essential for somatic functions. Interestingly, ectopic expression of NANOS1 was found to be required for growth of human pRb deficient tumor cells. In this case, NANOS1 and PUM repress p53 translation allowing cells to bypass apoptosis (Miles et al. 2014). Thus, ectopic Nos-Pum complexes may alter tissue homeostasis at the post-transcriptional level in other systems as well. We would expect to find key regulators of somatic fate among mRNAs aberrantly targeted by Nos-Pum in somatic Drosophila tissues.
In contrast to the widespread effects of nos derepression observed in multiple somatic tissues upon loss of L(3)mbt, ectopic expression of additional genes may define the exact phenotypic consequences, which depend on tissue type. For example, piRNA pathway genes are ectopically expressed in l(3)mbt larval brain tumors and somatic ovarian cells, however, depleting them ameliorates the brain tumor but not the ovarian phenotype. This difference may be explained by the fact that the somatic ovary uses core components of the piRNA pathway to regulate transposable elements (Handler et al. 2013). Consistent with our finding that in l(3)mbt mutants new and original tissue identities are co-expressed, these results suggest that the phenotypic consequences of l(3)mbt mutation depend on the context of the original tissue identity.
Our genetic analysis revealed that L(3)mbt functions with Lint1 in the somatic ovary. Lint1 likely contributes to L(3)mbt insulating functions. It co-occupies many L(3)mbt-bound high occupancy insulator sites and co-purifies with L(3)mbt. Further, like L(3)mbt, derepressed genes in Lint1 depleted Kc167 cells (Meier et al. 2012) show a bias towards head-to-head conformation. We propose that Lint1 functions as an architectural protein and contributes to L(3)mbt boundary properties in the somatic ovary. However, the phenotype of lint1 loss of function is weaker than the l(3)mbt phenotype suggesting that other protein complexes contribute to L(3)mbt boundary function. L(3)mbt also associates with the dREAM complex, some components of which have been shown to cooperate with insulator proteins to maintain transcriptional integrity at a subset of divergently-paired genes in Kc167 cells (Bohla et al. 2014; Korenjak et al. 2014). Further, E2f2 binding sites that possess enhancer-blocking properties are all co-occupied by L(3)mbt in Kc cells (Korenjak et al. 2014). Nevertheless, we find that the dREAM complex is dispensable for L(3)mbt boundary activities in the ovary. Thus composition and functionality of complexes associated with L(3)mbt vary in different cell types. It will be interesting to determine how L(3)mbt is recruited to different insulator complexes and how functionality of these complexes is regulated site- and tissue-specifically.
Our results support the idea that L(3)mbt functions as an insulator-associated protein and further show that it exerts insulating functions, in particular at differentially expressed, juxtaposed genes. However, L(3)mbt properties differ from canonical insulator proteins. For example, L(3)mbt does not mediate nucleosome remodeling at its binding sites, indicating that additional factors independently organize nucleosomes at these regions. These factors are not sufficient to maintain gene expression status in the absence of L(3)mbt. Further, the insulator proteins BEAF-32 and Cp190 co-occupy most of L(3)mbt’s binding sites but they do not share transcriptional targets. This suggests that L(3)mbt has insulating properties that might be independent of these insulator complexes. While the principles governing which genes are regulated by L(3)mbt in each tissue remain elusive, we find that L(3)mbt-regulated genes are preferentially found near TAD boundaries. Notably, the L(3)mbt targets nanos and vasa are localized at the border of TADs in multiple cell types (Ulianov et al. 2016). Interestingly, housekeeping genes cluster at TAD boundaries, and TfIIs and CG11779 – the HtH genes upstream of nos and vas fall in that category (Hou et al. 2012; Hug et al. 2017). We therefore propose that L(3)mbt function is to locally insulate developmental genes from adjacent, highly and promiscuously expressed genes. Considering that CTCF is the only insulator protein identified in mammalian genomes to this date, it will be interesting to investigate whether L(3)mbt mammalian homologues also have the ability to mediate insulating functions.
Materials and methods
Crosses schemes and genotypes are detailed in Supplemental File S1
Fly stocks
FRT82B, l(3)mbtGM76, e/TM6b was generated in the Lehmann lab (Yohn et al. 2003) and secondary mutations removed (Richter et al. 2011). The following stocks were obtained from the Kyoto Stock Center: w1118;; Df(3R)ED19066/TM6c (#150208) and y* w*;P{GawB}NP1624/CyO, P{UAS-lacZ.UW14}UW14 (Tj-Gal4, #104055); from Bloomington Drosophila Stock Center: w1118; P{neoFRT}82B P{Ubi-GFP(S65T)nls}3R/TM6B, Tb1 (BDSC #32655), y1 w*; P{UAS-FLP.D}JD1 (BDSC #4539), P{otu-GAL4::VP16.R}1, w*; P{GAL4-nos.NGT}40; P{GAL4::VP16-nos.UTR}CG6325MVD1 (Maternal Triple Driver, BDSC #31777), and w*; P{tubP-GAL80ts}2/TM2 (BDSC #7017). The following transgenes C587-Gal4, UAS-nos-tub (Clark et al. 2002; Ye et al. 2004), UAS-vas (Sengoku et al. 2006), UASp-mCherry-myr, UAS-Aub-GFP (Harris and Macdonald 2001), and TdTomato::l(3)mbt (Blanchard et al. 2014) were obtained from the Xie, Jan, Nakamura, Zallen, Macdonald, and Botchan labs, respectively. The mip12067.9A.9 (Beall et al. 2007) and FRT40A e2f2c03344 (Ambrus et al. 2007) mutations were generated by the Botchan and Frolov labs. The following mutations are from the Lehmann lab stocks: tud1, tudB42 (Arkov et al. 2006), aubHN2, aubQC42(Schupbach and Wieschaus 1991), nosL7 (Wang and Lehmann 1991), and nosBN (Wang et al. 1994). All stocks were maintained at 18°C and crosses were performed at 25°C unless otherwise stated.
Generation of transgenic lines
To generate UASp-l(3)mbt:: myc transgenic flies, l(3)mbt coding sequence was amplified from LD05287 gold cDNA (Drosophila Gemonics Resource Center), cloned using the p-ENTR/D-TOPO system and recombined into the pPWM destination vector (Drosophila Gateway Vector Collection) using Gateway technology (Invitrogen). pPWM-l(3)mbt was randomly inserted on the 2nd chromosome through P-transposition. The UAS-tdTomato short hairpin RNA (tdTomatoshRNA) was generated by Transgenic RNAi Project (SH08523.N) and inserted in the attP2 site (BDSC #8622). To generate the lint11 mutation, the target sequence (chrX:11044844-11044866) was identified by the flyCRISPR Optimal Target Finder tool (Gratz et al. 2014), amplified from genomic DNA using lint-CRISPR oligos (Supplemental Table S4) and ligated in the pU6-BbsI-gRNA plasmid (Gratz et al. 2013). The resulting construct was injected in FRT19A;; vas-Cas9 embryos and progeny were screened by PCR and sequencing.
Immunofluorescence
Adult ovaries from 2-3 day-old fatten females were dissected and stained following standard procedures (Arkov et al. 2006). Ovaries were mounted in SlowFade® Gold mountant (Invitrogen) and imaged on Zeiss LSM780 or LSM800 confocal microscopes, using 10, 20 or 43x objectives. The following primary antibodies were used: rabbit α-Vasa (1:5000, Lehmann lab); goat α-Vasa (1:200, Santa Cruz Biotechnology sc-26877); mouse α-spectrin (1:200, DSHB); mouse α-Orb (4H8, 1:200, DHSB); chicken α-GFP (1:1000, Aves GFP-1020); goat α-tj (1:7000, kind gift of Dorothea Godt (Li et al. 2003)); rabbit α-zfh1 (1:5000, Lehmann lab); rat α-RFP (1:500, Chromotek 5F8); rabbit α-Nanos (1:200, kind gift of Prof. Nakamura); rabbit α-Aub (1:1000, Lehmann lab); rabbit α-DsRed (1:500, Living Colors #632496). Alexa Fluor 647 Phalloidin (1:500 Life Technologies) and rabbit α-myc Alexa fluor 555 conjugated (Millipore 16-225) were used as secondary antibodies. Alexa Fluor 488-(Life Technologies), Cy3-or Alexa Fluor 647-conjugated (Jackson Immunoresearch) secondary antibodies were used at a 1:1000 dilution.
RNA sequencing
60-70 ovaries from females of maternal tud1/tudB42 and zygotic l(3)mbtGM76/+ or l(3)mbtGM76/Df genotypes (see Supplemental File 1) were dissected in cold PBS and RNA was extracted using TRIzol (Invitrogen) following the manufacturer’s protocol. For early embryos, TJ>UAS-l(3)mbt::myc;l(3)mbtGM76/+ and TJ>UAS-l(3)mbt::myc;l(3)mbtGM76/Df females were allowed to lay for 30 minutes to 1 hour on agar plates. Embryos were dechorionated in 50% bleach for 5 minutes, rinsed with PBS, and then lysed in TRIzol. Libraries were generated from 1μg of total RNA using the NEBNext Poly(A) magnetic Isolation Module (NEB #7490) and the NEBNext Ultra Directional RNA Library Prep Kit for Illumina (NEB # E7420). Libraries from biological replicates (three for ovaries, two for embryos) were sequenced on an Illumina Hi-Seq2000, single-end 50 run.
RNA-seq data analysis
Sequencing results were demultiplexed and converted to FASTQ format using Illumina Bcl2FastQ software (Illumina). Reads were aligned to the fly genome (build dm3/BDGP5) using the splice-aware STAR aligner (Dobin et al. 2013). PCR duplicates were removed using the Picard toolkit (https://broadinstitute.github.io/picard/). HTSeq package was used to generate counts for each gene based on how many aligned reads overlap its exons (Anders et al. 2015). These counts were then used to test for differential expression using negative binomial generalized linear models implemented by the DESeq2 R package.
Tissue expression clustering
Expression of deregulated genes was extracted from FlyAtlas (Chintapalli et al. 2007) using FlyBaseIDs, normalized to fly average and log2 transformed. Distance matrix was calculated using the “Manhattan” method and data clustered using “ward.D2”. Heatmap was generated using the heatmap.2 function of the gplots R package.
Tissue-specific Chromatin Immuno-Precipitation followed by quantitative PCR (ChIP-qPCR)
The tdTomato::l(3)mbt transgene (in a heterozygous l(3)mbt background) was knocked-down in the soma using TJ>UAS-tdTomatoshRNA for germline-specific L(3)mbt ChIP. Reciprocally, we used MTD>UAS-tdTomatoshRNA to perform L(3)mbt somatic ChIP. Chromatin Immuno-Precipitation was carried as previously described (Soshnev et al. 2008) with modifications: 60-70 1-to-2 day-old MTD>UAS-tdTomatoshRNA or 45-50 TJ>UAS-tdTomatoshRNA females were fattened overnight, dissected in cold PBS and their ovaries were fixed in 1.8% formaldehyde in PBS for 20 minutes at room temperature. Chromatin was sheared at 4°C using 16 cycles of sonication (30s ON – 30s OFF) in a Bioruptor Pico (Diagenode). 150μg of chromatin was used per IP reaction with 15μg of goat anti-tdTomato polyclonal antibody (SICGEN AB8181-200). QPCRs were performed on a Roche LightCycler 480 machine using standard procedures, primers described in Supplemental Table S4. Binding was calculated as a percentage of Input: %input = 100*2^(Ct(input)-Ct(IP))* 1/Dilution Factor).
Expression correlation of divergently paired genes
Pearson correlation was calculated for the expression of both genes of head-to-head pairs (UIRs<500 bp) across multiple cell lines and developmental stages (modENCODE dataset #3305, http://www.modencode.org/) using R correlation function and “pairwise.complete.obs” method. Correlations of expression were plotted as probability density for all, L(3)mbt-regulated somatic, and germline HtH gene pairs using the ggplot2 R package.
Cell culture
Ovarian Somatic Cell (OSCs), kind gifts of the Siomi and Brennecke labs, were cultured using standard procedures (Sumiyoshi et al. 2016).
ATAC-seq
ATAC-seq libraries were generated from 1x106 WT and ΔΔl(3)mbt OSC cells using the Greenleaf lab protocol (Buenrostro et al. 2013) modified as in (Iwasaki et al. 2016). Experiment was performed in triplicates, and libraries were sequenced on Illumina Hiseq 2000 using Paired-End 50 runs.
ATAC-seq data analysis
Sequencing results were demultiplexed and converted to FASTQ format using Illumina Bcl2FastQ software. Reads were aligned to the Drosophila melanogaster genome (build dm3/BDGP5) with Bowtie2 (Langmead and Salzberg 2012) using local alignment and allowing maximum fragment length up to 2 kb. Mitochondrial and poorly mapped reads (mapping quality <30) were discarded. Duplicate reads were removed using Sambamba (Tarasov et al. 2015). Broad open chromatin regions were detected with MACS2 (Zhang et al. 2008). Triplicate libraries yielding around 42x107 demultiplexed reads were pooled, and nucleosome positioning was generated using NucleoATAC (Schep et al. 2015) based on the broad peaks extended by 100 bp.
Data processing and plotting
Metaplots were generated using SeqPlots (https://github.com/Przemol/seqplots). L(3)mbt, CP190 and BEAF32 ChIP-seq profiles used to generate metagene plots originate from ((Li et al. 2015), GEO GSE63518). H2Av (modENCODE dataset #2991) ChIP-chip data in S2 cells was used to plot profiles across L(3)mbt (larval brain ChIPseq, (Richter et al. 2011), GEO GSE29206), and Cp190 and BEAF32 sites (0-12h embryos, ChIP-chip, modENCODE datasets #21 and #22). The coordinates of the Upstream Intergenic Regions (+/-1kb to include UTRs) of genes upregulated in l(3)mbt depleted Kc167 cells ((Meier et al. 2012), ArrayExpress MTAB-849) were intersected with the Architectural Protein Binding Sites matrix generated by ChIP-seq in Kc167 cells (Additional File 3 from (Van Bortle et al. 2014)) using Bedtools (Quinlan 2014). Hi-C and TADs predictions in OSC cells were generated by ((Ulianov et al. 2016), GEO GSE69013) and plotted using the WashU EpiGenome Browser (Zhou et al. 2011). Occurrence of derepressed genes with respect to TADs was calculated using a custom R script. Briefly, TADs were divided in 20 windows and occurrence of derepressed genes was calculated for each bin as well as 6 bins up and downstream.
Author Contributions
R.X.C. and R.L. conceived the project. R.X.C. F.K.T. and R.L. designed experiments and wrote the manuscript. R.X.C. performed the experiments and analyzed the data.
List of Supplemental files:
Supplemental File S1: fly genotypes and crosses
Supplemental Table S1: tudM, l(3)mbt (somatic) ovaries RNA seq
Supplemental Table S2: l(3)mbt (germline) emrbyos RNA seq
Supplemental Table S3: Insulator and accessory proteins occupancy at L(3)mbt binding sites
upstream regulated genes in Kc167 cells.
Supplemental Table S4: list of oligos
Acknowledgements
We thank the entire Lehmann lab for discussion and input; C. Desplan, D. Keefe, E. Mazzoni, N. Dyson and J. Treisman for discussions and advice; A. Soshnev for advice on ChIP, P. Rocha, R. Raviram and I. Dogalev for bioinformatics input. We would like to thank J. Knoblich and C. Richter, M. Siomi, J. Brennecke, N. Dyson, M. Botchan and D. Blanchard, M.V. Frolov, D. Godt, S. Kobayashi and A. Nakamura for sharing reagents. We are grateful to the Bloomington Drosophila Stock Center (NIH P40OD018537) for fly strains; the Developmental Studies Hybridoma Bank, created by the NICHD of the NIH and maintained at The University of Iowa for antibodies and the NYUMC Genome Technology Center (NIH P30CA016087), for sequencing. R.X.C was supported by the New York State Stem Cell Program (NYSTEM) of the New York State Health Department (#C026880) and FKT was funded by EMBO and HFSP fellowships. RL is supported by NIH R01/R37HD41900 and is an HHMI investigator.