Spen limits intestinal stem cell self-renewal

Precise regulation of stem cell self-renewal and differentiation properties is essential for tissue homeostasis. Using the adult Drosophila intestine to study molecular mechanisms controlling stem cell properties, we identify the gene split-ends (spen) in a genetic screen as a novel regulator of intestinal stem cell fate (ISC). Spen family genes encode conserved RNA recognition motif-containing proteins that are reported to have roles in RNA splicing and transcriptional regulation. We demonstrate that spen acts at multiple points in the ISC lineage with an ISC-intrinsic function in controlling early commitment events of the stem cells and functions in terminally differentiated cells to further limit the proliferation of ISCs. Using two-color cell sorting of stem cells and their daughters, we characterize spen-dependent changes in RNA abundance and exon usage and find potential key regulators downstream of spen. Our work identifies spen as an important regulator of adult stem cells in the Drosophila intestine, provides new insight to Spen-family protein functions, and may also shed light on Spen’s mode of action in other developmental contexts.

Introduction During development, pluripotent stem cells will give rise to all of the different cell types present in the organism. Adult stem cells have more limited plasticity and play essential roles in tissue homeostasis and regeneration by both renewing the differentiated cells as well as maintaining the stem cell pool. Defining the mechanisms governing stem cell self-renewal and differentiation is essential for understanding both organism development as well as tissue maintenance and regeneration.
The adult Drosophila intestine is an attractive model to study adult stem cells as it provides a genetically tractable system with many similarities to other tissues such as the mammalian intestine and lung [1]. The fly intestine is renewed by intestinal stem cells (ISCs), which produce progenitor cells that differentiate into terminally differentiated polyploid absorptive enterocytes (ECs) and diploid secretory enteroendocrine cells (EEs) [2,3] (Fig 1A). A majority of ISC divisions produce EC cells via post-mitotic enteroblast (EB) precursors. Recent findings propose that EEs cells may originate from rare ISC daughter cells that divide once to produce a pair of EE cells [4,5]. Thus, ISCs are the primary dividing cell type in the intestine. In normal homeostasis, the proliferative status of the ISC is largely controlled through signaling of the EGFR, Jak/Stat, Insulin receptor [6][7][8][9][10][11] and reviewed in [12]. In addition, in response to epithelial damage, ISCs receive additional proliferative signals promoting tissue renewal [6,7,[13][14][15][16][17][18][19][20][21][22][23] and see [24] for review). While proliferative control is ensured by the aforementioned signaling pathways, ISC differentiation is largely controlled by Notch signaling [2,3,25,26]. Delta-Notch signaling between stem cell daughters results in Notch activation in EB progenitors that will further promote differentiation into EC cells, whereas inhibition of Notch by Numb and expression of Scute promotes EE fate acquisition [4,25,27]. The ISC itself, has low or no Notch signaling, thereby retaining ISC fate. Thus far, outside of canonical Notch signaling components, little is known about the mechanisms controlling ISC fate acquisition leading to robust asymmetric fate acquisition after ISC division.
In an EMS-based genetic screen, we have identified the gene split-ends (spen) as a regulator of Drosophila adult intestinal stem cells, required to limit their production. The SPEN protein family is broadly conserved in plants and Metazoa and mediates functions in RNA splicing, polyadenylation, and transcriptional regulation via its RNA binding motifs (RRM) and a SPEN Paralog and Ortholog C-terminal domain (SPOC) protein interaction domain ( Fig 1B) [28][29][30]. There are three mammalian SPEN-family paralogs: SPEN/SHARP, and two family members RBM15/OTT1 and RBM15B/OTT3 encoding smaller proteins. SPEN is known to mediate transcriptional silencing via physical association with the nuclear receptor corepressor components including NCoR1, and HDACs [28,31], while the smaller proteins RBM15 and RBM15B have been primarily found to be associated with splicing regulation [32,33]. Interestingly, recent studies in mouse embryonic stem cells have demonstrated that RBM15 and RBM15B interact with the RNA m 6 A methyltransferase complex, which facilitates m6A methylation of the non-coding RNA XIST [34,35]. Consequently, RBM15 and RBM15B are essential for X-inactivation, which also requires SPEN and associated HDAC activities [35][36][37][38][39].
In Drosophila, the Rbm15 paralog Spenito physically associates with components of the RNA m 6 A methylation complex and has been suggested to facilitate m6A RNA methylation and splicing, playing an important role in sex determination [33,40,41]. Drosophila Spen has diverse roles in development, regulating processes such as cell proliferation, differentiation, cell death, axon guidance, and fat metabolism [42][43][44][45][46][47][48][49][50][51]. In mouse hematopoietic stem cells (HSCs), Rbm15 is essential for return to the quiescent state after stress induction [52]. The functions of SPEN family proteins in other adult stem cell populations have not yet been elucidated. While stem cells when compared to control lineages. In addition, spen mutant ISCs showed more abundant levels of Delta protein expression and ectopic localization to the cell surface membrane. Targeted knock-down of spen in different cell types revealed that it has both stem cell autonomous and non-autonomous roles in controlling ISCs. Despite these defects, spen mutant stem cells were still capable of terminal differentiation. We further explored its role in ISCs and EBs by comparing differential gene expression and differential exon usage of wildtype and spen-knocked down conditions. We found that spen altered transcript abundance and exon usage, suggesting that Spen impacts both transcription and splicing of genes, candidates for downstream regulators of ISCs. Importantly, our work identifies spen as a novel regulator acting in multiple cell types in the intestine to control ISC numbers and proliferation.

spen limits intestinal stem cell number in the adult midgut
In ongoing EMS-induced genetic screens in the lab for defects in ISC proliferation, selfrenewal and differentiation, we identified two lines (9E34 and 9C95) which caused altered numbers of diploid cells. They failed to complement each other and mapped to the same genomic region by deficiency mapping. Genes in the region were tested and both lines failed to complement previously identified alleles of gene split-ends (spen): spen 3 and spen 5 [43]. spen encodes 8 isoforms of similar sequence of approximately 5500 amino acids (aa). 3 RRMs (RNA Recognition Motifs) domains are found N-terminally and a SPOC domain C-terminally (Spen paralog and ortholog C-terminal). spen 5 was previously characterized as deleting nucleotides 39548-39601 (2L:192716-2L:192769) resulting in a frame shift and early termination [43]. Sequencing of the mutant lines spen 9E34 and spen 9C95 identified nonsense mutations, affecting nucleotides 2L:185,541 and 2L:190,065 respectively ( Fig 1B). RNAs with non-sense mutations are frequently degraded via non-sense mediated decay, however, if translated, these RNAs would result in truncated proteins. spen 9E34 allele would encode a putative protein of 630 aa, including the RRM domains. spen 9C95 and spen 5 would potentially encode putative truncated proteins of approximately half the wild-type protein size, of respectively 1961 aa, and 2827 aa ( Fig 1B). These alleles all likely function as loss of function alleles as 1) they all result in similar phenotypes, 2) these mutant phenotypes are like those of expressing spen RNAi (see below), and 3) spen 5 mutant phenotypes can be rescued by a single copy of a genomic rescue construct, arguing against dominant neomorphic function of a truncated mutant protein (see below). Published RNAseq data suggest that Spen is likely expressed at low levels in all epithelial cell types of the gut [53] (http://flygutseq.buchonlab.com/). Thus, despite apparent low levels of expression, spen was identified in our genetic screen as being critical to control diploid cell number in the adult intestine.
To further investigate how spen affects the ISC lineage in the Drosophila posterior midgut, we used previously described markers of ISCs to assess the effect of spen inactivation on cell fate using the Mosaic Analysis with Repressible Cell Marker (MARCM [54]) technique in adult ISCs. In this method, a heat shock is given in adult animals to induce expression of a Flipase, which then promotes exchange of homologous chromosomes containing FRT sequences, thereby leading to the production of either wild-type or mutant ISC lineages marked through GFP expression. In the adult intestine, marked lineages are induced preferentially in ISCs, as the ISC is the primary dividing cell type in the adult intestine. In order to characterize the spen phenotype, we first examined the expression of the stem cell marker Delta (Dl) in spen mutant and knockdown stem cell clones (Fig 1C-1I). 5 days after heat shock (AHS), wild-type stem cell clones contained 1.0 Dl+ ISC on average per clone (Fig 1C and 1I). In contrast, spen 5 mutant stem cell clones contained 13.0 Dl+ cells per clone (Fig 1D and 1I). Similar phenotypes for spen 9E34 and spen 9C95 were observed (Fig 1E and 1F). In addition, the aberrant increase of Dl+ cells was rescued upon insertion of a pACMAN BAC construct containing the spen wildtype genomic region, and was phenocopied when clones expressed spen RNAi (Fig 1G and 1H). The Dl+ cells produced in spen mutant clones also expressed additional markers of ISCs: the Notch receptor (Fig 1J-1K), Sanpodo (Spdo; Fig 1L-1M') [26], which marks both ISCs and the EB daughter cells [3]. Thus, spen mutants accumulate Dl+ cells that share additional features with wild-type stem cells. In addition, the level of Dl protein at the cell membrane and in endocytic vesicles was markedly higher in spen mutants than in wild-type (Fig 1N and 1O-1P').
Consistent with an increased number of stem cells upon spen inactivation, mutant clones contained more Phospho Histone H3+ (PH3+) cells than wild-type: whereas only 3.4% of wild-type clones contained one PH3+ cell, 21.3% of spen mutant clones had at least one ( Fig  1Q). This is also consistent with the significant increase in the proportion of dividing stem cells in spen mutants: in the wild-type population 2.2% of Dl+ were PH3+ whereas 5.2% of Dl+ were PH3+ in the spen population ( Fig 1R). In agreement with an alteration in proliferation, spen inactivation led to larger clones containing 27.2 cells per clones compared to wild-type clones, which contained 6.7 GFP+ cells in total ( Fig 1S). Altogether, our data indicated that inactivation of spen results in larger clones and that had an excess number of ISCs, suggesting that Spen is required to limit the number of stem cells.
In contrast to ISC numbers, cell fate specification of terminally differentiated cells was not affected in spen mutant clones: at 5 days AHS, wild-type clones contained 3.3 ECs (marked by nuclear size >7μm and Pdm1 expression) and 0.6 EEs (marked by Prospero expression) per clone on average (Fig 1T and 1T'). Similarly, spen mutant clones contained both types of terminally differentiated cells with 4.9 ECs and 0.7 EEs per clone (Fig 1U and 1V). Therefore, we conclude that terminally differentiated cells are produced in absence of spen function.
spen acts in stem cell autonomously and non-autonomously to regulate stem cells spen is expressed in all epithelial cell types of the gut [53]. To investigate in which cell type spen function is required to control intestinal stem cell production, we expressed spen RNAi in a celltype specific manner using the TARGET system [55]. The downregulation of spen simultaneously in both ISCs and EBs using the esg-GAL4, tub-GAL80 ts line (esg ts ) led to an enrichment of Delta protein at the membrane in ISCs as well as an increased density and proliferation of ISCs (PH3+ cells; Fig 2A-2F'). Quantification revealed a 2.8X increase in the number of ISCs per 1000μm 2 and a 20X increase in the number of PH3+ cells on average per posterior midgut upon spen depletion in ISCs and EBs (Fig 2A-2D).
The knockdown of spen in ISCs-only using esg-GAL4, tub-GAL80 ts , NRE-GAL80 line, similarly led to altered Dl membrane location in ISCs, and increased numbers (2.4X) and proliferation of ISCs (1.3X), measured by the density of Dl+ cells and number of PH3+ cells (Fig 2A-2D and 2G-2H'). Because the effect on ISC proliferation was less pronounced when spen was knocked down in ISCs only as compared to ISCs and EBs, we hypothesized that spen may also function in EBs to limit ISC proliferation. To further explore a function of spen in EBs that could limit non-cell autonomously ISC proliferation, we knocked down spen function in EBs using NRE-GAL4. Consistent with this notion, EB-knockdown of spen led to a 2.2X increased number of Dl+ and a 42X increase in the number of PH3+ cells per posterior midgut (Fig 2A-2D and 2I-2J'). Importantly, there was no impact on Dl membrane localization in ISCs ( Fig  2I-2J'), indicating that 1) this activity of spen is ISC autonomous and 2) that spen has additional effects on limiting ISC proliferation and ISC numbers that are independent of its activity in the ISC.
We then further explored potential non-cell autonomous functions of spen by examining the effects of the knockdown of spen in terminally differentiated EE and EC cells. Expression of spen RNAi in enteroendocrine cells using the pros voila -GAL4, tub-GAL80 ts line, or in EC cells using the Myo1A-Gal4, tub-Gal80 ts did not have a major impact on the density of Dl+ ISCs or Delta membrane localization in ISCs (Fig 2A-2D and 2K-2N'). Interestingly, we found that while ISC numbers were not altered, the knockdown of spen in ECs and EEs did lead to increased ISC proliferation suggesting a non-autonomous function of spen in controlling ISC proliferation from EEs and EC cells (Fig 2A-2D and 2K-2N'). Indeed, spen depletion in EEs led to a 3X increase, whereas inactivation in ECs led to a 4X increase in PH3+ cells. (Fig 2B).
We did notice, however, that both of these drivers, showed some leakiness of expression in Dl+ ISC cells (S1 Fig). Consistent with this, we observed that upon spen knockdown with pros voila -GAL4 and Myo1A-Gal4, as well as NRE-Gal4 line, rare Dl+ ISCs had strong Dl signal at the membrane and showed GFP expression indicating driver leakiness (S1 Fig). However, we believe that the limited number of ISC cells with leaky expression cannot explain the dramatic effect on ISC proliferation noted above. Thus, spen activity in ECs and EEs limits ISC proliferation. Therefore, through this series of experiments, we conclude that spen has multiple functions in controlling ISCs: an ISC-intrinsic role in controlling Dl protein accumulation at the membrane, ISC fate and ISC proliferation; a function in the EB to non-autonomously regulate ISC fate and proliferation; and roles in terminally differentiated cells that limit ISC proliferation.

spen acts upstream or parallel to Notch activation in control of ISC numbers
Our data thus far indicated that spen activity has different functions: controlling Dl accumulation on the cell membrane (in ISCs), limiting ISC number (in ISCs and EBs), and restraining proliferation (in ISCs, EBs, EEs and ECs). Importantly, these separate activities can occur independently of each other, suggesting that they involve distinct mechanisms.
In order to better understand the role that spen plays in controlling ISC fate and limiting ISC numbers, we then focused on a potential interaction with the Notch pathway, the major pathway implicated in controlling cell fate decisions in the ISC lineage. We hypothesized that spen may function through Notch regulation to mediate cell fate decisions and asymmetric fate outcomes. The inactivation of the Notch pathway causes an accumulation of ISC cells and excess numbers of EE cells [2,3]. Nevertheless, our previous work has shown that lowering, but not completely inactivating Notch signaling can increase the number of ISCs without impacting EE and EC cell numbers [26], similar to the phenotype observed here for spen mutant clones. In addition, links between the Spen family proteins and Notch signaling have been demonstrated in other tissues [31,49,50,56]. Thus, spen could promote asymmetric outcomes of the cell fate decisions following stem cell divisions by influencing Notch pathway signaling. We therefore explored the possibility that deregulated Notch signaling occurred upon spen inactivation by assessing a reporter of Notch transcriptional activity (NRE-LacZ, NRE+ cells) in spen mutants. The NRE reporter is a synthetic construct with three copies of the SPS site (for Su(H) Paired Sites, binding sites for the Notch activity-dependent transcription factor, Su(H)) that is taken from the E(spl)m8 regulatory region and combined with three copies of the Grh binding element (Gbe); it likely reports Notch signaling above a given threshold [57]. While in spen mutant clones there was an overabundance of Dl+ ISCs that did not have activated Notch, consistent with an increase in stem cells, many cells in these clones still activated the Notch pathway ( Fig 3A-3D), further consistent with our results above that terminal differentiation was unaffected. However, these data do not rule out spen acting to enhance Notch signaling but not being required for it, affecting a subset of cells receiving Notch signaling. Thus, while spen mutant stem cells overproliferated and did not produce as many Notch active EB progenitors per stem cell as wild-type stem cells, they could still derive enough progenitors to give rise to normal numbers of EC cells. The lack of defects detected in EE cell production also imply that spen does not alter the production of EE cell precursors derived from the ISC.
Upon cell division, both ISC daughter cells inherit Dl protein present in the ISC mother cell. In a majority of ISC divisions, one of these daughter cells will undergo commitment through activation of Notch signaling and extinguish its expression of Dl and become an EB cell committed towards EC fate [58]. It is possible that ISCs accumulate in spen mutants because of a delay in the commitment process, which would result in a decreased number of EBs relative to ISCs. We assessed this by comparing the number of ISCs (Dl+) to the number of EB cells (NRE+). In wild-type lineages, there were 0.79 Dl+ISCs per NRE+ EB cells 5 days AHS. In contrast, upon spen inactivation, there were 2.1 ISCs per EB (Fig 3D), suggesting that spen mutant ISCs have more symmetric ISC divisions that wild-type ISCs. Thus, while most wild-type stem cell divisions result in asymmetric daughter cell fates [59,60], in spen mutants stem cell divisions, ISC division more frequently leads to symmetric stem cell fate acquisition. As Spen family members have been implicated in transcriptional regulation, it is possible that spen acts to promote reliable activation of Notch target genes either directly or indirectly, as has been suggested in the hematopoietic system [50]. We therefore tested whether spen mutant cells could differentiate in response to cleaved active Notch. If spen were required for the activation of Notch target genes, the expression of activated Notch would have no effect in spen mutants. As previously demonstrated [3], when a cleaved active form of Notch was expressed in wild-type stem cells, it caused their differentiation into ECs generating 1 or 2 cell clones ( Fig 3E-3H). spen mutant clones contained 4.8 Delta+ cells per clone, while the expression of an active Notch receptor (UAS-N Act ) suppressed this to 0.05 Delta+ cells per clone in spen and led to smaller clones composed largely of ECs (Fig 3E-3J). To exclude the possibility that overexpression of activated Notch simply forced differentiation of the ISC prior to establishment of the spen loss of function phenotype due to residual Spen protein, we repeated the experiment with altered kinetics: spen was first inactivated in MARCM clones, then UAS-N Act was expressed in these clones by releasing Gal80ts inhibition at 29˚C (Fig 3K-3P). 14d after spen mutant clone induction, indeed stem cells with Delta enrichment at the membrane accumulated ( Fig 3L). Note that in this experiment no GFP was detected due to the activity of Gal80ts at 18˚C. However, large clones composed of ECs were produced when Notch pathway overactivation was induced for 4d, 10d after spen clones had been induced ( Fig 3P). Thus, the extra ISC-like cells resulting from spen inactivation were forced to differentiate into ECs, when the active Notch form was expressed (Fig 3L-3P). Therefore, spen mutant ISCs can respond to Notch pathway activation and differentiate, indicating that spen is not essential for Notch target gene activation. This is consistent with our findings above that some spen mutant cells can activate Notch signaling.
These data suggest that spen may be upstream of, or parallel to, cleavage and activation of the Notch receptor. One important step upstream of Notch activation is the trafficking and endocytosis of its ligand Delta, raising the possibility that the increased numbers of stem cells may result from cell fate defects due to alteration of Delta trafficking or endocytosis. Interestingly, we noticed that Dl protein, but not Sanpodo protein, was more abundant and more strongly localized to the plasma membrane in spen mutant stem cells than in wild-type stem cells (Fig 1D and 1L-1N). To test whether this effect on Delta localization was specific to the intestine, we generated spen mutant clones in the sensory organ cells of the peripheral nervous system (PNS). We did not find obvious effects on Delta protein levels nor subcellular localization, or lineage specification defects in the PNS (Fig 3Q-3R), or in the ovary (Fig 3S and 3T). Spen's function on Dl levels is therefore tissue specific.
The membrane accumulation of Delta is known to inhibit, in cis, the activity of the receptor Notch present in the same cell [61,62]. We thus tested whether simply elevating the levels of Dl by overexpression was sufficient to promote accumulation of stem cells using the marker Spdo to assess the number of ISCs. Consistent with observations of others [8,13,63], overexpression of Dl in clones led to an enrichment of Delta protein at the membrane, but did not drive accumulation of ISCs or affect stem cell proliferation (Fig 3U-3X). This suggests that the spen phenotype is not due to a simple increase in Dl levels, though it could result from altered Dl trafficking dynamics.
All together, these data suggest spen has a tissue-specific effect on Dl protein levels and/or trafficking to the cell membrane, and that spen limits symmetric cell fate acquisition upon ISC division. This activity impinges upstream or in parallel to cleavage and activation of the Notch receptor.
spen mutant ISCs are less sensitive to lowered EGFR signaling, but are dependent on Akt/Insulin signaling To gain insight into how spen regulates ISC proliferation, we tested whether proliferation of spen 5 mutant ISCs was dependent on EGFR signaling, a major pathway controlling ISC cell division. As previously reported [9,64], we found that the expression of a dominant-negative version of EGFR (EGFR-DN) reduced the size of wild-type clones (19.4 cells per clone in controls to 10.7 cells per clone on average in EGFR expressing clones). Interestingly, expression of EGFR-DN did not decrease the size of spen mutant clones (Fig 4A-4E). Furthermore, in spen mutants expressing EGFR-DN neither the number, percentage of Dl+ ISCs per clone, nor the strong accumulation of Dl at the membrane were suppressed compared to spen mutants ( Fig  4A-4D, 4F and 4G). Thus, we conclude that spen mutant clones are less sensitive than wildtype clones to reduction in EGFR activity. However, as the EGFR-DN likely does not totally inactivate EGFR signaling, further studies combining complete inactivation of EGFR with spen will be required to determine whether spen mutant stem cells can proliferate in its absence.
In mammalian cells, Spen was recently suggested to influence the activity status of Akt [65], a downstream component of the Insulin pathway known to regulate ISC proliferation [6,59,66,67]. We therefore wanted to test whether inactivating Akt or the Insulin receptor could suppress spen-associated phenotypes. We found that reducing the activity of Akt decreased both the size of spen mutant clones and the accumulation of Dl+ ISCs per clone (Fig 4H-4N). Interestingly, the strong accumulation of Dl at the membrane was not observed in small spen 5 ; Akt RNAi cells (Fig 4J and 4K) suggesting that Akt is required for the accumulation of Dl in spen mutants. Very few clones were recovered of spen 5 mutant expressing a dominant negative Insulin receptor, suggesting clone loss. However, a majority of those that were recovered were very small (6.9 spen 5 , InR-DN cells on average compared to 70 cells in spen 5 mutant clones; S2 Fig), consistent with the spen 5 , Akt RNAi phenotype described above.
Altogether, these data suggest that the proliferation of spen mutant stem cells is less sensitive to levels of EGFR signaling but dependent on Akt and Insulin receptor signaling.

spen controls transcript abundance and differential exon usage
We then wanted to better characterize downstream effectors of Spen in ISCs and EBs that may contribute to adult stem cell regulation. Spen family members have been shown to regulate gene expression at transcriptional and post-transcriptional levels [30,33,34,36,[68][69][70]. We thus assessed the impact of spen loss of function on gene expression at transcriptomic and exonic level in ISCs and compared this with EBs. We developed a two-color FACS sorting approach to simultaneously sort ISCs and EBs (NRE+ cells). The esgGal4, Gal80ts system to express spen RNAi in both ISCs and EBs, along with RFP to mark these cell populations was used. The EBs were further discriminated using a NRE-Venus reporter [71]. Thus, ISCs were RED only and Notch active EBs were RED and GREEN (Fig 5A). Wild-type and spen RNAi ISCs and EBs were sorted by FACS (Fig 5B). polyA+ RNAs were isolated 2 days after spen RNAi expression and sequenced. RNAseq analysis was then performed to assess the effect of spen inactivation on differential gene expression and exon usage.
In order to validate our RNAseq protocol, we first assessed the similarities between our samples. Principal component analysis (PCA) revealed that our samples clustered in four distinct groups according to their genotype and cell type: WT-ISC, SPEN-ISC, WT-EB and SPE-N-EB (Fig 5C). Secondly, we analyzed differential gene expression between sorted wild-type stem cells and wild-type EB cells using DEseq2 [72]. Among the 366 differentially expressed genes identified with a False Discovery Rate (FDR) of 0.05, 299 genes were upregulated in ISCs compared to EBs, including known ISC markers and cell cycle related-genes: polo [73], Cdc2  https://doi.org/10.1371/journal.pgen.1007773.g005 [73], Delta [2,3], sanpodo [26], charlatan [74] (Fig 5D, S1 and S2 Tables). In addition, Gene Ontology terminology (GO term) enrichment analysis of the ISC-enriched genes revealed an over-representation of functions and processes associated with cell division regulation and cytoskeleton regulation, reflecting stem cell properties (S3 Table). Interestingly, other overrepresented functions were also found associated with regulation of protein localization to nucleus, enzyme activity, signal transduction, and negative regulation of cell differentiation, among others (S3 Table). No specific GO term enrichment could be found with EB-enriched genes. Thus, ISCs and EBs cells were separated by our two-color FACS approach, and this allowed us to reliably identify differentially expressed genes.
SPEN-family proteins play prominent roles in transcription as well as post-transcriptional processes. Therefore, we analyzed the impact of spen inactivation on differential gene expression (DEseq2) [72] and differential exon usage (DEXseq) [75] in ISCs and EBs. In ISCs, we found 55 genes were differentially expressed with a FDR of 0.05, of which 26 were up and 29 were down upon spen inactivation (Fig 5E, S4 Table). In EBs, 440 genes had significantly altered expression with 298 up and 142 down in spen RNAi (Fig 5F, S4 and S2 Tables). We validated a number of these genes using RT-qPCR analysis. Indeed, genes that were found up-regulated or down-regulated in ISC and/or EB in our RNAseq data, showed similar relative expression pattern when spen was knockdown in whole midgut (compare Fig 5G to 5H and 5I). Interestingly, 29 (13 DOWN and 16 UP) spen-dependent genes were similarly de-regulated in the two cell types (Fig 5J and S2 Table). Of note, the observed upregulation of spen in ISCs and EBs expressing spen RNAi was due to the overexpression of the dsRNA (742bp) that targets the exon E14 of spen in this experimental set up. Indeed, sequencing reads accumulated in both sense and anti-sense orientations in this region when spen RNAi is expressed, but not in WT. Importantly, we did not detect alteration of Dl mRNA levels or exon usage, suggesting that the increase in Dl protein detected in spen mutant clones is likely due to regulation of Dl protein itself, perhaps during its endocytic trafficking steps.
While the shorter SPEN-family members (Spenito in flies; RBM15, RBM15B in mouse) impact splicing regulation [33,35,[37][38][39]41], the larger SPEN-family proteins (SPEN/SHARP in mouse), to date, have primarily been implicated in transcriptional regulation [28,31,34,76]. Nevertheless, mammalian SPEN was found as an interactor of the spliceosome, raising the possibility that SPEN also impacts splicing [29]. We therefore assessed exon usage in wild-type and spen RNAi ISCs and EBs, and identified 141 transcripts that had one or more exon(s) with significantly different usage in ISCs and 193 in EBs (Fig 5K-5M, S5 and S6 Tables). We also found that the usage of 47 exons located in 37 transcripts, was similarly affected in the two cell types (Fig 5K, S5 and S6 Tables). Of note, the alterations in exon usage that we detected upon spen knockdown are comparable to the number of transcripts with altered splicing events upon inactivation of the Spenito-associated m 6 A methylation complex. Upon the inactivation of ime4, encoding a methyltransferase component of this complex, 163 transcripts were affected. Inactivation of, YT521, encoding an m 6 A reader altered 103 transcripts [33,41]. Interestingly, our data also revealed that in EBs expressing spen RNAi , spenito exon E4, located in its 3'UTR, was significantly less used than in WT (Fig 6A). This may reflect an increase in the long 3'UTR spenito isoform due to E4 exon skipping. These observations suggested that spen affects exon usages of many RNAs including that of its related gene, spenito.
We then wanted to understand whether the majority of Spen's activity in controlling gene expression was mediated by alteration in exon usage. Of the 55 genes altered upon spen knockdown in ISCs, 7 (16.7%) were found to have altered expression levels and exon usage. Similarly, of the 440 genes differentially expressed between wild-type and spen RNAi in EBs, 25 genes (5.7%) were both differentially expressed and had altered exon usage (Fig 5L and 5M and S7 Table). This suggests that in absence of spen function, most of the changes in differential gene expression (83.3% in ISC and 94.3% in EBs) are not due to a difference in exon usage. Thus, we conclude that Spen controls transcript abundance and exon usage in ISCs as well as in EBs. In addition, Spen has common effects on transcripts independently of the cell type, which may reflect more general Spen target genes. Finally, Spen also has cell-specific effects on RNAs in ISCs and EBs.

spen regulates genes involved in similar biological processes
In order to assess the function of the spen-dependent genes, we performed a GO term enrichment analysis on the differentially expressed (DE) genes and genes with alternative exon usage (AEU) in ISCs and in EBs respectively, upon spen inactivation (S3 Table). In ISCs, our analysis revealed 3 over-represented GO term categories associated with the symporter activity. Of note, these included Ca 2+ [5,77,78], we hypothesize that the disruption of their expression upon spen inactivation may contribute to the defects in spen mutants of ISC numbers and proliferation. Furthermore, cac affects lysosomal transport [79], raising the possibility that its alteration could change the dynamics of Dl protein in spen mutant ISCs.
In EBs, the GO term enrichment analysis of the spen-dependent differentially expressed genes showed 10 over-represented GO term categories associated with enzymatic activity (S3 Table). It also revealed an over-representation of the GO term category "maintenance of imaginal disc-derived wing hair orientation" (S3 Table), which was supported by the upregulated genes cora (encoding an actin binding protein) and Gli (encoding a transmembrane protein known to localized at tricellular junctions in the intestine [80]). This function is affected by spen during Drosophila development, where it is required for the formation and the correct positioning of vein and bristle cells, as well as the maintenance of planar polarity in the wing imaginal discs [81]. Interestingly, spen knockdown led to alteration of transcript levels and exon usage in numerous ribosome associated genes. 4 over-represented GO term categories associated with cytoplasmic ribosomal processes were found of genes with alternative exon usage in EB upon spen knockdown. Indeed, upon spen knockdown in EBs, 13 ribosomal protein-coding genes had changes in their exon usage (Fig 6B-6I, S5 Table). In addition, 19 ribosomal protein-coding genes had differential gene expression and were found downregulated in EBs upon spen RNAi (S4 Table). Among them, RpS21 and RpL19 were downregulated and had an alternative exon usage dependent of spen (S3 and S4 Tables). In ISCs, spen knockdown affected the exon usage of 8 ribosomal protein coding transcripts, including the RpS21 and RpL19 transcripts (S4 Table). Therefore, spen regulates a variety of genes encoding ribosomal proteins.
While not found as an enriched GO term, we also noticed that numerous spen-dependent genes in ISCs and EBs were related to chitin binding and chitin metabolism. In insects, chitin is a major component of the cuticle or exoskeleton and forms a barrier with the external environment. In Drosophila, the intestine is protected by both type I and type II peritrophic matrices with type I being secreted throughout the midgut and type II forming a sleeve-like Spen limits intestinal stem cell self-renewal membrane that is produced by the proventriculous. In spen knocked-down ISCs, the chitin binding proteins coding genes CG14989, CG33258, Cpr51A, and the Chitinase 2 coding gene (Cht2), were found downregulated, whereas the gene yellow, associated with cuticle pigmentation, was upregulated (S4 Table). In absence of spen function in EBs, CG33258, Cpr51A genes were downregulated as well, whereas the gene ebony, also associated with cuticle pigmentation, was upregulated (S4 Table). This may explain defects of spen mutants in cuticle integrity noted by others leading to defects in the formation of the sclerites (a cuticular structure), and in the trachea [43, 82,83].
It is also of note that the expression of several genes related to immune response was altered upon spen inactivation. The Imd pathway is activated in the adult gut in response to bacterial infection [84] and inactivation of Imd pathway can stimulate ISC proliferation [7,85,86]. In spen knockdown ISCs, the immune response genes AttC and pirk were upregulated, as well as Tg, a transglutaminase that acts to inhibit the Imd component Relish [87]. Interestingly, the Pvr ligand Pvf1 was found to be upregulated in spen knockdown EBs. The PDGF and VEGF (Pvr) pathway regulates immune signaling through the IMD pathway [88]. Previous work has established that the overexpression of the secreted Pvf1 ligand in ISCs and EBs activates Pvr in the ISC and drives increased proliferation and expansion of the ISC population [89]. Therefore, the upregulation of Pvf1 in spen mutant EBs may promote increased ISC proliferation and accumulation. In addition, other immune response genes were also upregulated, including bactericidal factor-encoding genes LysB, LysE and LysD from the lysozyme locus, the gene coding the peptidoglycan recognition protein PGRP-SC2, and also additional downstream effectors of the Imd pathway: relish/NF-kB and pirk (S4 Table). Moreover, alternative exon choice of PGRP-LB was detected in spen knockdown EBs (S5 Table). Thus, the Imd pathway may be affected by the inactivation of spen function in the intestine. This could explain some phenotypes of spen mutant larvae that have defects in the immune response [50,90]. Altogether, these data suggest that Spen regulates genes associated with similar functions that may impact on intestinal homeostasis and physiology.

Discussion
The mechanisms by which ISCs undergo asymmetric and symmetric divisions are still not completely understood. Here, in a genetic screen, we have identified spen as an essential regulator of adult stem cells in Drosophila. Our data indicate that in the absence of spen activity, stem cells have aberrant high levels of Delta (Dl) protein and fail to properly commit into EB daughter cells resulting in large increase in numbers of ISC-like cells and that this activity is cell autonomous roles in the ISC. Furthermore, spen acts in EBs to limit ISC numbers, and in EBs, EEs and ECs cells to suppress ISC proliferation. Therefore, in spen mutant tissue, stem cell autonomous and non-autonomous mechanisms that act to drive ISC proliferation, increased ISC numbers, and aberrant accumulation of Dl protein in ISCs (Fig 7). While spen mutants are dependent on Akt and InR for their growth, they are not as sensitive to a reduction in EGFR activity as wild-type control clones are. Our findings argue that Spen does not act as a general regulator of Notch signaling in all tissues, and is upstream of, or parallel to, Notch pathway activation to promote intestinal stem cell commitment. Through a transcriptomic analysis, we have identified genes controlled by spen in ISCs and EBs, both at the transcript and exon level, providing candidate ISC regulators downstream of spen. Critically, this work shows that the RNA binding protein Spen is an important regulator of asymmetric fate outcomes of ISC division and its proliferation.
It is interesting that spen has both stem cell-autonomous and non-autonomous activities to regulate stem cell numbers and proliferation (Fig 7). The knockdown of spen in ISCs and EBs can expand stem cell numbers, whereas the knockdown in EEs and ECs led to a dramatic increase in the number of dividing ISCs, but did not change the density of ISCs. We believe, therefore, that the growth of spen mutant clones is influenced by both cell autonomous and non-autonomous processes: spen inactivation in the stem cell leads to more ISC symmetric self-renewing divisions, however, EB, EE and EC cells are still produced. The mutant EB, EE, and EC cells can further drive proliferation of the extra ISCs through non-autonomous regulation. It is possible that the non-autonomous regulation of ISC proliferation detected upon downregulation of spen in EEs and ECs could be due to stress induced in the tissue owing to disruption of numerous spen targets genes in these cells. Importantly, our findings highlight the complex process of stem cell deregulation arising in tumor contexts whereby inactivation of a tumor suppressor genes may have numerous functions in different cells within a tumor that collaborate to drive tumorigenesis. In the intestinal stem cell (ISC), spen acts autonomously to limit their proliferation and stem cell self-renewal. Its inhibition leads to an accumulation of the Delta protein at the membrane. This is likely due to an effect on Dl trafficking, which may be direct or indirect. spen also acts in a nonautonomous manner in the Enteroblast (EB) to limit ISC numbers. Furthermore, spen inactivation in EBs, Enteroendocrine cells (EE) and Enterocytes (EC) leads to enhanced ISC proliferation. Further studies will be required to better understand Spen's molecular functions. https://doi.org/10.1371/journal.pgen.1007773.g007 One intriguing effect of spen inactivation is the high levels of accumulation of the Notch ligand Dl at the plasma membrane and in intracellular vesicles. Our RNAseq data revealed that Dl mRNA levels or exon usage are not affected, suggesting a regulation at the protein level, perhaps during its endocytic trafficking steps. This phenotype of Dl protein accumulation is reminiscent of those occurring upon Dl trafficking perturbation, such as in neuralized mutants [58, 91,92]. A previous study has shown genetic interaction of spen and regulators of trafficking of Notch ligands [93]. Indeed, in the developing Drosophila eye disc, spen genetically interacts with the endocytic adaptor Epsin/liquid facets (lqf), which promotes Dl trafficking and internalization facilitating Dl activation [94,95]. Spen, however, is likely not a general regulator of protein trafficking as the localization of the membrane protein Sanpodo occurred normally in spen mutant clones in the intestine. Therefore, Spen likely regulates Dl trafficking in a tissue-specific manner and indirectly through transcriptional or post-transcriptional control of a downstream gene.
In addition, our data indicated that the knockdown of Akt and InR activity could reduce the number of Dl+ ISCs produced in the spen mutant background. The InR pathway has previously been implicated in regulating symmetric cell divisions of the ISC during adaptive growth [59]. Thus, it appears as if Akt and InR are also facilitating symmetric cell division in the spen mutant context. Our data also show that spen mutant stem cells are insensitive to a reduction of EGFR signaling upon overexpression of EGFR-DN, suggesting that spen and EGFR pathway have opposing roles in regulating ISC proliferation. While a decrease in EGFR signaling reduces ISC proliferation, spen inactivation, in contrast, leads to an increase in proliferation. spen has been shown to genetically interact with the EGFR signaling pathway during Drosophila embryogenesis and during eye development [44,46,48,51,96,97]. Nevertheless, in these contexts, spen potentiates EGFR signaling. Thus, spen may have a tissue specific effect regarding EGFR pathway regulation. The precise link between spen and EGFR signaling regulation is nevertheless still unclear. Indeed, while a constitutive active form of Ras can rescue the lethality caused by a spen dominant negative form in embryo [48], spen seems to act downstream or in parallel of Ras activation during photoreceptor specification [44,46,48,49,96,97]. Additionally, spen loss of function in cone cells during pupal eye development affects the EGF receptor ligand Spitz level, which suggest a role at the level of the EGF receptor activation [51]. Further studies will be required to better understand the link between spen and the EGFR signaling pathway in the regulation of intestinal stem cell regulation, as well as in other developmental contexts.
In addition to describing a new function of spen in adult stem cell regulation, our data raise a number of interesting questions about the function of spen and its downstream target genes in stem cell regulation. Importantly, we demonstrate that inactivation of a large SPEN family member also results in considerable alteration of exon usage, like the smaller family members (spenito in flies, RBM15 and RBM15b in mouse). Thus, spen activity is important for transcriptional and post-transcriptional regulation in Drosophila, although we cannot exclude that some of these effects might be indirect. Interestingly, in absence of spen function, altered exon usage of spenito was detected in EBs, raising the possibility of feedback control between family members, reminiscent of the plant SPEN family member FPA [30].
Little is known about the impact on intestinal stem cell self-renewal of RNA binding proteins. A recent study found that inactivation of Tis11, encoding a regulator of RNA stability, led to larger clones than controls [98]. The Tis11 phenotype differs markedly from that of spen as it appears not to affect cell fate decisions since ISCs do not accumulate as in spen mutant clones. Importantly, our study provides a data set that opens new perspectives for future investigation on the relationship between Spen and RNA regulation at transcriptional and posttranscription levels. Interestingly, our data also reveals that Spen co-regulates several genes encoding proteins that are involved in similar biological processes, such as cytoplasmic ribosomal proteins, immune response pathway and Imd pathway genes, and genes involved in Chitin metabolism. This may explain previously reported spen phenotypes in other tissues [43,46,48,49,56,82,90,99].
Recent studies have highlighted important, essential functions of Spen family proteins from X-inactivation in mammals to sex determination and fat metabolism in flies [33,35,41,47,100]. This family of proteins is also frequently mutated in cancers, though mechanistically the role SPEN family proteins in cancers is not understood [101][102][103][104][105][106]. Our work demonstrates that spen inactivation causes the deregulation of numerous genes and results in alteration of the stem cell fate acquisition. In mammals, RBM15 has a critical function in promoting hematopoietic stem cell return to quiescence [52]. These findings raise the possibility that conserved functions of SPEN family proteins may help restrict stem cell activity required for cancer prevention.

Fly stocks
Fly stocks used in this study were the following: FRT40A spen 9E34 [allele presents a nonsense mutation (C>T at genome position 2L:185541), which leads to a putative 630 aa truncated protein]; FRT40A spen 9C95 [allele presents a nonsense substitution (C>T, 2L:190065), which leads to a putative 1961aa truncated protein]. The 9E34 and 9C95 mutations were mapped with Sanger and Illumina wholegenome sequencing, respectively. Of note, our fly stock containing the spen 9E34 allele was lost from our collection in 2017, though this allele produced identical phenotypes to the spen 9C95 and spen 5 [111] kindly received from B. Edgar).
MARCM clones were generated using the following stock:

Mosaic analysis with a repressible cell marker (MARCM) and RNAi experiments
Adult flies and crosses were kept at 25˚C in rich yeasted medium tubes unless otherwise noted. MARCM clones were induced by a 35 min. heat shock at 36.5˚C in 3 day old adults. Flies were then dissected either at 5 days, 6 days or 10 days AHS. For MARCM ts experiments in Fig 3K-3P were done with the following modifications: Crosses and adults were kept at 18˚C before and after heat shock. 10 days AHS, adults were shift at 29˚C for 4 days and then dissected. RNAi thermosensible experiments were kept at 18˚C and shifted at 29˚C during indicating period before dissection. For GeneSwitch experiments (Fig 5G, S3 Fig), crosses and newly eclosed siblings were raised in a standard food. 3 days old adults were sorted and shifted on food supplemented with 50 μg/ml of RU486 in EtOH or EtOH alone, at 25˚C during 2 days prior dissection. City, IA 52242. The fixation protocol previously described in [25] was used in this study. A methanol fixation protocol described in [112], was used for anti-Notch immunofluorescence in Fig 1K and 1L.

Quantification and statistical analysis
All experiments were performed on female flies. For clonal analyses of this study, posterior midguts were analyzed and "stem cell clones" (i.e. clones deriving from ISCs and containing two or more cells) were scored. In Fig 3E and 3F, single cell clones were also included in the analysis since stem cell clones differentiate under expression of active Notch. In Fig 2, Delta+ cell density referred to the number of Delta+ cell per pictures area (1000μm 2 ) and the number of PH3+ cells was quantified in the posterior midgut. Statistical analyses were performed with GraphPad Prism version 5 for Mac, GraphPad Software (La Jolla California USA, www. graphpad.com) using the nonparametric Mann-Whitney two-way ANOVA, when not specified. In Fig 1Q, a Fisher's test was performed. For Fig 4, data from 2-4 experiments were combined, therefore some of the data points from wild-type and spen mutant clones are the same in graphs of Fig 4E-4J and 4L-4N. We note that in the experiments at 10d in Fig 4, while spen clones had a greater mean size than wild-type, this difference was not significant in a Mann-Whitney test which measures median values, in contrast to 5d clones in Fig 1 where this difference was significant. This may be due to the fact that at 10d, cells within clones are undergoing turnover. Nevertheless, the number of clones >29 cells, was significantly different in spen compared to controls at 10d (p>0.05, Mann-Whitney).

Intestinal cell FACS sorting
The following genotype was used: esg-GAL4, tub-GAL80ts/ (UAS-spen KK RNAi or +); UAS-RFP, NRE-Venus expression was induced by a 2 day temperature shift from 18˚C to 29˚C. Entire intestines were dissected and collected in cold PBS. Dissociation was performed as described in [113] using PBS-Elastase (1mg/ml) solution and mechanical pipetting. Dissociated cells were resuspended in PBS, 1% BSA and FACS (FACS Aria BD Biosciences) using the following parameters: 20psi, 100μm nozzle. Diploid RFP positive fraction were separated from the diploid RFP, GFP double positive fraction. Cells were directly collected in TRIzol (Life Technologies).

RNA-seq
RNAs was extracted by performing two series of Phenol-Chloroform extractions at 4˚C, followed by DNase treatment according to the manufacturer's instructions. ExpressArt Pico RNA Carrier compounds for very small samples was used to precipitate the RNA. Around 100ng total RNA was collected from 100,000 sorted cells. Total RNA quality and quantification were assessed on an Agilent Bioanalyzer using the Agilent RNA 6000 nano kit. An Illumina TruSeq Stranded RNA (LS) protocol was used to build the libraries, from polyA RNAs enriched by Oligo-dT bead polyA tail capture. Due to a failure of one of the EB-WT samples to cluster with the other two samples during the Principal component analysis; only two biological replicates were used for the EB-WT whereas three biological replicates for the others conditions. Sequencing was performed on the Illumina HiSeq 2500.

RNA-seq analysis
For the following analyses, default parameters were used if not indicated, associated with a paired-end and first strand library type modes. Alignment-Sequencing resulted in about 50 million 100bp paired-reads from each library, which were then aligned against the Drosophila melanogaster r5.54 genome reference using TopHat (version 2.0.6) with the following parameters (-N 3-read-edit-dist 3 -p 2 -g 1) [114]. Unmapped reads were discarded. Counting reads -Mapped reads were counted on gene using SummarizOverlaps method from the Genomi-cAlignments package (version 1.8.0) [115] with a "Union" mode, based on the ENSEMBL GENES 75, BDGP5 genome reference.
Principal component: Principal component analysis was performed using the PCAplot function from DEseq2 R package on transformed normalized read counts (rlog function) of all samples. This analysis considers the 300 top-ranked genes (ntop = 300) with the highest row variance across all samples. PCA showed that our samples were distinguishable by the first and second principal component (PC1, 44% and PC2, 19% of the total variation). Interestingly, when doing the PCA analysis by selecting the 500 top-ranked genes based on their row variance, ISC and EB populations were clearly discriminated by the first principal component (PC1, 39% of the total variation), but not the difference between WT and SPEN samples by the second principal component (PC2, 19% of the total variation).
Differential expression analysis: Differential gene expression was analyzed using DEseq2 (version 1.12.2) [116]. The following parameters: DESeq(minReplicatesForReplace = Inf) and results(cooksCutoff = FALSE) were used in order to disable the Cook's filtering and flagging. Genes were defined as differentially expressed according to their padj<0.5.
Differential exon usage: Differential exon usage was analyzed using DEXseq (version 1.18.4) with default parameters [75]. With this method developed by Anders et al 2012, we assessed the changes in relative exon usage independently of the overall differential gene expression. During the read mapping process, in the case where exons do not share the same boundary in all transcripts, DEXseq splits the exon in two or more parts to create "counting bin". Thus, this term refers to exons or parts of exons generated from this manipulation. Exons with an adjusted p values (padj) less than 0.5 were considered as significantly differentially used. In order to visualized the relative exon usage without the impact of the change in gene expression level, we used the function "plotDEXSeq" with the option "splicing = TRUE" provided by DEXseq, which removes the effects due to changes in gene expression from the plots. Venn diagrams were generated using Venny 2.1.0 tool (Oliveros, J.C. [2007][2008][2009][2010][2011][2012][2013][2014][2015].
The RNAseq data produced from this publication have been deposited to the NCBI GEO and are available under accession number GSE84367.

GO enrichment analysis
The GO term gene annotation was done with the PANTHER Classification system [117] developed by the Gene Ontology Consortium [118].
GO term enrichment test-The R package GOseq version 1.28.0 (released 3.5, 2016-05-30) [119] was used to perform a GO term enrichment test. Each set of differentially expressed genes were compared with the list of 11,103 expressed genes in the ISC and EB as a background. This list of background genes was obtained by considering as an expressed gene, a gene with non-zero FPKM in at least 1 sample in either WT-EB or WT-ISC conditions. Thus, all unexpressed genes were removed from the list of total number of gene in the Drosophila melanogaster genome (r5.54). Only pval<0.05 are shown.
Supporting information S1 Table. Differentially expressed genes: ISC versus EB. Genes that were differentially expressed between ISC and EB cells. (XLSX) S2 Table. Raw data DEseq2 analysis. Raw data of analysis of differential gene expression using DEseq2 comparing wild-type ISCs vs EBs, spen knocked-down ISC versus wild-type ISC, and spen knocked-down EB versus wild-type EB. (XLSX) S3 Table. Go term enrichment. Go term analysis of differentially expressed genes in wild-type ISC-enriched genes; of differentially expressed genes in spen knocked-down ISC versus wildtype ISC; of differentially expressed genes in spen knocked-down EB versus wild-type EB; and of altered exon usage in spen knocked-down EB versus wild-type EB. (XLSX) S4 Table. Differentially expressed genes: spen-dependent genes. Genes with significant changes (FDR of 0.05) upon spen knockdown with indicated Log2FC and associated biological term. Differences between spen knocked-down ISC versus wild-type ISC, and spen knockeddown EB versus wild-type EB are shown. (XLSX) S5 Table. spen-dependent altered exon usage. Genes resulting from DEXseq analysis that demonstrated significant altered exon use upon knock down of spen in ISCs and EBs. (XLSX) S6 Table. Raw data DEXseq analysis. Raw data of analysis of altered exon usage using DEXseq comparing wild-type ISCs vs EBs, spen knocked-down ISC versus wild-type ISC, and spen knocked-down EB versus wild-type EB. (XLSX) S7 Table. Genes with differential gene expression and altered exon usage. Genes that were both differentially expressed and had altered exon usage in ISCs and EBs, wild-type versus spen knockdown. (XLSX) S1 Fig. pros voila -GAL4 and myo1A-GAL4