spict, a cyst cell-specific gene, regulates starvation-induced spermatogonial cell death in the Drosophila testis

Tissues are maintained in a homeostatic state by balancing the constant loss of old cells with the continued production of new cells. Tissue homeostasis can shift between high and low turnover states to cope with environmental changes such as nutrient availability. Recently, we discovered that the elimination of transit-amplifying cells plays a critical role in maintaining the stem cell population during protein starvation in the Drosophila testis. Here, we identify spict, a gene expressed specifically in differentiating cyst cells, as a regulator of spermatogonial death. Spict is upregulated in cyst cells that phagocytose dying spermatogonia. We propose that phagocytosis and subsequent clearance of dead spermatogonia, which is partly promoted by Spict, contribute to stem cell maintenance during prolonged protein starvation.

Tissue homeostasis can shift between high and low turnover states depending on environmental variables. For example, tissues can decrease turnover and/or scale down their overall size when the nutrients are limited [1][2][3] . Stem cells are widely considered the master regulators of tissue homeostasis, and how stem cells respond to changes in their external environment has been heavily studied [4][5][6] . In many tissues, however, the majority of cell proliferation occurs in transit-amplifying cells, which are stem cell progeny that divide mitotically prior to terminal differentiation to lessen the proliferative burden on stem cells [7][8][9][10][11] . Despite this, the response of transit-amplifying cells during shifting tissue homeostasis is poorly explored.
The Drosophila testis is an excellent model system to study the behavior of stem cells and transit-amplifying cells owing to the well-defined anatomy of the tissue and the ample genetic tools available for manipulating gene function in a cell type-specific manner. At the apical tip of the Drosophila testis resides two stem cell populations: germline stem cells (GSCs) and somatic cyst stem cells (CySCs). They are anchored to hub cells that organize the stem cell niche for both stem cell populations (Fig. 1A) 12,13 . In addition, CySCs encapsulate GSCs, and together with the hub cells, function as a part of the GSC niche by contributing to the critical signaling environment 14,15 . Upon stem cell division, GSCs produce gonialblasts (GBs), whereas CySCs produce cyst cells (CCs). GBs undergo four rounds of transit-amplifying divisions as spermatogonia (SGs). As cytokinesis of these divisions is incomplete, these transit-amplifying divisions yield a cluster of 16 interconnected spermatogonia (SGs), which then undergo meiotic divisions and spermiogenesis. Connectivity of SGs (2-cell, 4-cell, 8-cell, 16-cell SGs) serves as a reliable marker for their differentiation stage (Fig. 1A). Throughout this process, a pair of CCs envelop the SGs and help regulate their differentiation. CCs are critical for the survival and differentiation of SGs beyond the 2-cell SG stage (Fig. 1A) 16 .
Recently, we reported that SG death dramatically increases in response to protein starvation 17 . The GSC population, however, is relatively well maintained even during a prolonged period of protein starvation. After an initial drop in GSC number from ~8/testis to ~6/testis after 3-6 days of starvation 2 , the remaining ~6 GSCs can be stably maintained for additional ~20 days while continuing to divide at an unchanged rate compared to fed conditions 17 . This argues that transit-amplifying cells, but not stem cells, may be a major point of regulation in response to changes in nutrient conditions. We have shown that starvation-induced SG death is triggered by Results spict is expressed in differentiating cyst cells. In a small-scale screen to identify genes expressed in the Drosophila testis, we identified a gal4 enhancer trap of spichthyin (spict), which was originally identified as the Drosophila homolog of the human NIPA1 and ichthyin genes 18 . When the expression pattern of spict was visualized by expressing UAS-nlsGFP (nuclear localization signal-containing GFP) with the spict-gal4 driver, we found that GFP was specifically observed in the nuclei of differentiating CCs. Notably, nlsGFP was absent from the nuclei of somatic cells in close contact with hub cells, which most likely represent CySCs. In contrast, the well-established CC driver c587-gal4 19 drove the expression of nlsGFP in all early somatic cells at the apical tip including the CySCs (Fig. 1B,C).
The lack of nlsGFP expression in CCs near the hub indicates that spict-gal4 expression might be excluded from CySCs. To test this idea, we examined the relationship of spict-expressing cells with two characteristics of CySCs: 1) attachment to hub cells and 2) ability to undergo mitosis. To examine the attachment to hub cells, UAS-mCD8-GFP (a plasma membrane marker) was expressed using either c587-gal4 or spict-gal4. Consistent with c587-gal4 being expressed in all early CCs including CySCs, we observed mCD8-GFP-labeled cell processes attached to hub cells (Fig. 1D) 12,20 , and 100% of testes contained multiple mCD8-GFP-positive processes attached to hub cells (N = 19). In contrast, when the expression of UAS-mCD8-GFP was driven by spict-gal4, mCD8-GFP-positive processes were rarely associated with the hub (only < 5% of testes contained hub-touching processes, N = 87). These results demonstrate that most spict-expressing cells do not contact the hub, suggesting that spict-gal4-expressing cells are not CySCs (Fig. 1E).
Next, we examined whether spict-gal4-expressing cells can undergo mitosis. In the Drosophila testis, CySCs are the only somatic cell population that undergoes mitosis 20 , and all other somatic cells are post-mitotic. To examine whether spict-gal4-expressing cells undergo mitoses, we labeled mitotic cells with anti-phosphorylated histone H3 (PH3) antibody. When PH3 staining was combined with c587-gal4 > mCD8-GFP, 100% of mitotic somatic cells were mCD8-GFP-positive (N = 19), consistent with the notion that c587-gal4 is expressed in CySCs. In contrast, when spict-gal4 > mCD8-GFP was combined with PH3 staining, only 2.5% of all PH3-positive cells were also positive for mCD8-GFP (N = 119), supporting the idea that spict-gal4-expressing cells are rarely dividing. Taken together, these results strongly argue that spict-gal4 is excluded from CySCs, and that spict expression marks differentiating CCs. spict expression can be used to better identify the CySC population in combination with Zfh-1.
The best marker for labeling CySCs identified to date is Zfh-1 15 . Zfh-1 is a transcriptional repressor, whose function is critical for the maintenance of CySC identity. However, Zfh-1 not only marks CySCs but also their immediate daughters that have been displaced away from the hub and have initiated differentiation as CCs. Accordingly, the number of Zfh-1-positive cells is higher than the expected number of CySCs 12,15 . We reasoned that spict-gal4 could be used to negatively mark CySCs and to better identify the CySC population when combined with Zfh-1. Indeed, we found that the Zfh-1-positive population can be subdivided into spict-negative and spict-positive populations (Fig. 1F). We scored the number of 'Zfh-1-positive, spict-negative' and 'Zfh-1-positive, spict-positive' somatic cells. We observed that there were 33.6 ± 10.3 Zfh-1-positive cells in the wild type flies used in this study (Fig. 1G). Among the Zfh-1-positive cells, 16.1 ± 4.6 were spict-negative, whereas 17.5 ± 11.0 were spict-positive (N = 13 testes). Given that essentially all CySC characteristics (attachment to the hub and the ability to divide) are confined within 'Zfh-1-positive, spict-negative' cells, we conclude that this population (~16 cells/testis) represents the population with the highest CySC concentration identified to date. spict-expressing CCs may become CySCs. It is well established that GSCs divide asymmetrically by orienting their mitotic spindles perpendicular to hub cells and that this spindle orientation is prepared by stereotypical centrosome positioning throughout the cell cycle 21 . In contrast, CySCs do not have consistent centrosome positioning in interphase. They enter mitosis with randomly oriented spindles, but reposition their spindles during anaphase such that one daughter of the CySC division remains attached to the hub, whereas the other daughter is displaced away from the hub and initiates differentiation 20 . Based on these observations, we proposed that CySC divisions are also stereotypically asymmetric. However, a recent lineage tracing study found that CySCs likely undergo stochastic self-renewal and differentiation and suggested that there is no stereotypical asymmetric stem cell division in the CySC population 22 .
We reasoned that spict-gal4′ s ability to specifically mark differentiating CCs might help to reconcile these differences. We used spict-gal4 to drive the expression of FLP recombinase to permanently label CCs that are spict-positive at any point (spict-gal4, UAS-FLP, act > stop > gal4, UAS-GFP, tubP-gal80 ts ) and followed their fates. The flies were raised at 18 °C to repress spict-gal4 expression until adulthood, and the young adult flies of this genotype were shifted to 29 °C after eclosion to allow the lineage tracing of spict-positive cells. Before temperature shift, there was no gal4 activity, thus the entire testis was GFP-negative ( Fig. 2A). 6 hours after temperature shift, GFP-positive, Zfh-1-positive cells were apparent (Fig. 2B). At this point, none of GFP-positive, Zfh-1-positive cells were observed adjacent to the hub cells, consistent with the above results that spict-positive cells are not CySCs. By 24 hours after temperature shift, the number of GFP-positive Zfh-1-positive cells had dramatically increased, and they began to appear adjacent to the hub (Fig. 2C,D). Eventually, almost all Zfh-1-positive cells became GFP-positive (Fig. 2D). This suggests that spict-positive CCs frequently revert back to a spict-negative, CySC state. Although we cannot completely exclude the possibility that spict-gal4 is weakly expressed in CySCs (at a undetectable level), resulting in lineage-marking of CySCs, the fact that the initial lineage-marked cells always appear as CCs indicates that most of lineage-marking events occur in CCs, rather than CySCs. The frequent reversion of spict-positive CCs to spict-negative CySCs is consistent with the neutral competition model proposed by Amoyel et al. 22 . If CySCs divide symmetrically without ever expressing spict, our spict-gal4-mediated lineage tracing would not have resulted in an increase of GFP-positive CySCs. Combined with our previous report that CySCs consistently divide asymmetrically with regard to the attachment to the hub cells (i.e. 96% of CySC division yielded one daughter cell attaching to the hub, the other detaching from the hub) 20 , we propose that CySCs divide asymmetrically with respect to spict expression (and attachment to the hub cells), generating a spict-negative CySC and a spict-positive differentiating CC, and that spict-expressing CCs frequently dedifferentiate to reacquire CySC identity. As a result of frequent dedifferentiation events, the CySC population follows a neutral competition model as demonstrated by Amoyel et al. 22 . In the face of frequent dedifferentiation, which spict does not regulate BMP signaling in the Drosophila testis. It was shown that spict facilitates BMP receptor endocytosis to negatively regulate BMP activity in the Drosophila neuromuscular junction 18 . Spict bears homology to mammalian NIPA1, 2 and Ichthyin proteins, which are predicted to encode transmembrane proteins [23][24][25] . NIPA1 was also shown to be required for downregulating BMP signaling in the cultured neuron 26,27 . Therefore, we sought to test whether spict may be required for BMP signaling in the Drosophila testis.
Upon BMP ligand-receptor binding, Mad (mother against Dpp) is phosphorylated (pMad) and translocates to the nucleus for downstream transcriptional regulation [28][29][30] . By using pMad as a readout of BMP activity, we investigated whether a spict mutant may have altered BMP signaling activity in the testis. In wild type testes, it is well known that GSCs are positive for pMad, as BMP signaling functions in GSC self-renewal 14,31,32 . In addition, it has been observed that differentiating CCs far from the hub also show pMad signal, which was shown to regulate spermatocyte differentiation 33,34 . We observed no detectable difference in pMad levels between control and spict mutant testes, either in the GSC population or in the late CC population (Supplementary Fig. S1). These results suggest that spict might not play a role in regulating BMP activity during spermatogenesis. spict is required for SG death and maintenance of the GSC pool upon starvation. In order to explore the function of spict in Drosophila spermatogenesis, we examined Spict protein localization using UAS-spict-mRFP expressed under the control of spict-gal4 (spict-gal4 > spict-mRFP. UAS-spict-mRFP was shown to complement spict mutant phenotypes) 18 . We noticed that Spict-mRFP extensively overlapped with Lysotracker staining (Fig. 3A,B, see below for detailed Spict localization), which we previously showed to mark dying SGs 17 . These observations prompted us to ask whether spict plays a role in SG death.
In our previous work, we showed that SG death is significantly upregulated upon protein starvation 17 . SG death is triggered by the apoptosis of an encapsulating CC, which in turn initiates SG death, likely due to the CCs' essential role in SG survival 16 . We further showed that inhibition of SG death by blocking CC apoptosis led to a continuous decline in GSC number and an eventual collapse in tissue homeostasis under protein starvation conditions, suggesting that eliminating SGs plays a critical role in maintaining the GSC population during starvation 17 . Spict-mRFP was observed in dying SGs irrespective of nutrient conditions (Fig. 3A,B). We found that SG death was significantly decreased in spict mutants (spict 65 /spict 41 ) or upon RNAi-mediated knockdown of spict in CC lineage using CC-specific driver tj-gal4 (tj-gal4 > spict RNAi ) 35 compared to control (Fig. 3C,  Supplementary Fig. S2). Although we cannot exclude the possibility that spict may also play a role in germ cells, CC-specific expression pattern of spict (Fig. 1) and the fact that both spict loss-of-function mutant as well as CC-specific knockdown of spict (tj-gal4 > spict RNAi ) result in a similar phenotype (Fig. 3C, Supplementary Fig. S2) argue that spict functions mainly in CCs. Consistent with our previous study, which showed that SG death is required for maintaining the GSC pool 17 , spict mutants failed to maintain GSC number during prolonged protein starvation (Fig. 3D-F). spict mutants maintained similar numbers of GSCs compared to control under fed conditions (data not shown), suggesting that spict's requirement is more profound during protein starvation.
Taken together, these results suggest that spict is required to promote starvation-induced CC/SG death, and consequently for the maintenance of the GSC pool in response to protein starvation.
Spict protein is expressed in CCs associated with dying SGs. To gain further insights into the role of spict in SG death, we expanded our analysis of Spict localization. Spict protein was highly upregulated in CCs near dying SGs and appeared punctate in those cells (Fig. 4A). Spict-mRFP accumulation near dying SGs is likely due to increased mRNA translation or protein accumulation as the spict-gal4 reporter (spict-gal4 > UAS-GFP) (Fig. 1) shows ubiquitous expression of GFP in early CCs (excluding CySCs). In support of this idea, when a CC clone expressing UAS-spict-mRFP was generated using a constitutive actin promoter (hs-Flp, Act > stop > gal4, UAS-GFP, UAS-spict-mRFP), Spict-mRFP was observed specifically when the CC clone was adjacent to dying SGs (Fig. 4B-D): Accumulation of Spict-mRFP was observed in 77% of the CC clones that are associated with dying SGs, whereas Spict-mRFP accumulation was observed only in 14% of the CC clones that are not adjacent to dying SGs (Fig. 4E). Because Spict-mRFP does not accumulate unless associated with the dying SGs, it is unlikely that the use of spict-gal4 causes dominant, over-expression phenotype. Consistently, we did not observe any detectable changes in the GSC number or the frequency of SG death in Spict-mRFP-expressing testis (spict-gal4 > spict-mRFP).
In addition to accumulation in CCs near the dying SGs, Spict-mRFP was observed in the dying SGs themselves (Fig. 4A). Since spict-gal4, which is expressed in CCs but not in germ cells (Fig. 1), was used to drive UAS-spict-mRFP expression, Spict-mRFP might be transferred from the CC to the dying SG. A similar result was obtained using a well-established CC driver, tj-gal4 ( Supplementary Fig. S3), suggesting that the Spict-mRFP observed in dying SGs is not due to upregulation of spict transcription in dying SGs.
To better understand Spict-mRFP's localization during the progression of SG death, we first conducted a detailed characterization of the SG death process by combining various markers (Vasa, LaminDm0, DAPI and Lysotracker, Supplementary Fig. S4), extending our previous characterization 17 . Based on these markers, we now divided SG death into four phases. During phase 1, as SGs initiate cell death, the level of Vasa (a germ cell marker) decreases compared to neighboring SGs that are not dying, coinciding with the appearance of Lysotracker in the dying SG at a very low level. Lysotracker intensity increases significantly during phase 2, indicating acidification of the germ cell cyst and thus progression of the cell death process. This phase is also characterized by the complete disappearance of Vasa staining, whereas nuclear envelope marker LaminDm0 and DAPI staining remains distinct at this phase, suggesting that nuclear compartment is still relatively intact. This is followed by phase 3, when LaminDm0 staining disappears, while DAPI staining remains, indicating that digestion of nuclear compartment is progressing. Finally during phase 4, all Vasa, LaminDm0 and DAPI staining disappears, and the remnants of dead SGs are visible only with Lysotracker, which sometimes becomes weaker than phases 2 and 3. The temporal order of SG death was confirmed by inducing synchronized SG death through expression of the proapoptotic gene grim in CCs, as described previously 16 . Indeed, after induction of grim expression, phases 1 through 4 appeared in the expected order, suggesting that phase 1-4 characterization is accurate. (Supplementary Fig. S4C).
We adapted these criteria to further characterize Spict localization in detail during SG death ( Fig. 4F-J). Because the patterns of Vasa, LaminDm0 and DAPI during SG death were fairly consistent, cell death phases can be determined even without the use of Lysotracker. Thus, we combined Spict-mRFP with Vasa, LaminDm0, and DAPI to follow the localization of Spict-mRFP during SG death. We found that dying SGs are not yet associated with Spict-mRFP during phase 1 (Fig. 4F). During phase 2, Spict-mRFP starts to accumulate near dying SGs (Fig. 4G), followed by phase 3, when Spict-mRFP is highly upregulated in the CCs near dying SGs and is also observed in dying SGs (Fig. 4H). During phase 4, Spict-mRFP in the CC is mostly gone and is mainly observed in dying SGs (Fig. 4I).
The above results suggest that Spict-mRFP protein accumulates in CCs and dying SGs after the initiation of SG death, i.e. in phases 2 through 4. This raised the question regarding in which CCs Spict-mRFP might be upregulated. Our previous study showed that SG death is triggered by the apoptosis of CCs 17 , indicating that CC apoptosis must occur before/during phase 1. If this is the case, Spict-mRFP is not expressed in the CCs that apoptose to trigger SG death. We speculate that only one of two CCs that encapsulate the SG cyst dies to trigger SG death, and the other CC might survive and upregulate Spict protein. Alternatively, it is possible that other CCs, such as those encapsulating other (alive) SGs, might upregulate Spict protein. However, we do not believe the latter possibility likely, as we have never observed CC clones that encapsulate multiple SG clusters, including living SGs and dying SGs, in our various experiments in the course of this study. Irrespective of its origin, dying SGs are clearly encapsulated by CCs with high Spict expression (Fig. 4B-E). Taken together, we propose that the death of SGs is triggered by apoptosis of encapsulating CC, and that Spict is subsequently upregulated in the neighboring, surviving CC (Fig. 4K).

Spict localizes to the late endosome/phagosome during engulfment of dying SGs by CCs.
Previously, it was shown that Spict colocalizes with Rab5-positive early endosomes 18 . Likewise, it was also shown that a mammalian homologue of Spict, NIPA1, colocalizes with various endosomal compartments 36 . Their localization to endosomal compartments was linked to their role in the regulation of BMP receptors. Although our results indicate that spict is likely not regulating BMP signaling in the Drosophila testis ( Supplementary Fig. S1), colocalization of Spict-mRFP with Lysotracker (Figs 3 and 4) indicates that Spict is involved in the endocytic pathway. Thus, we examined the potential colocalization of Spict-mRFP with various endosomal compartments by expressing Spict-mRFP together with EYFP-tagged endogenous Rab small GTPases 37 , or UAS-GFP tagged Rab GTPases 38,39 using spict-gal4. We did not observe obvious colocalization of Spict-mRFP with early endosomal markers, such as Rab5, or with recycling endosome markers, such as Rab4, Rab11 and Rab35 in the testis (Fig. 5A and Supplementary Fig. S5) 40 . However, we found that Spict often colocalizes with Rab7, a marker for the late endosome (Fig. 5B, 68% of dying SG cyst was encapsulated by Rab7-positive vesicle, N = 183 dying SG cysts in 51 testes scored). Given that Spict also colocalizes with Lysosomes, it appears that Spict is associated with the late endosome/lysosome compartment in the Drosophila testis.
While studying the colocalization, we noticed that Spict-positive dying SGs were frequently contained within a large vesicle positive for Rab7-EYFP (Fig. 5B). Such Rab7-positive 'vesicles' could reach almost ~20 μ m in diameter, which is the size of the entire cyst of dying SGs containing up to 16 cells. Since Rab7 is known to be required for phagosome maturation 41,42 , we suspected that the entire dying SG might be phagocytosed by the neighboring cells (i.e. CCs). Indeed, when a single CC clone expressing GFP and a membrane marker (mCD4-tdTomato) (hs-FLP; act > stop > gal4, UAS-mCD4-tdTomato) was induced, the plasma membrane of the clone marked by mCD4-tdTomato continuously wrapped around the Lysotracker-positive dying SGs (Fig. 5C). This CC is most likely the surviving CC that does not undergo apoptosis to initiate SG death. Moreover, staining with another membrane dye, FM4-64, corroborated that the Rab7-EYFP-positive vesicle was contained within a single CC (Fig. 5D). Taken together, these data suggest that dying SGs are encapsulated by a single CC, likely via phagocytosis (Fig. 5E). spict is required for normal progression of SG death. Based on our observation that Spict is highly upregulated in the CCs that phagocytose dying SGs, we speculated that spict might be required for processing/clearing the dead SGs. However, we did not see obvious differences between control and spict mutants in the degree of acidification of the dying SG (not shown). Moreover, although overall SG death was decreased in spict mutant as described above (Fig. 3C), once SGs initiate death process, the frequency of dying SGs being encapsulated by Rab7-positive phagosome or not was not different between control and spict mutant: 68% of dying SG cysts was encapsulated by Rab7-positive vesicle (N = 183 dying SG cysts) in control, 84% of dying SG cysts was encapsulated by Rab7-positive vesicle (N = 204 dying SG cysts from 66 testes scored) in spict 65 / spict 41 mutant. The SG cell death process from initiation to completion appears to take a long time (> 24 hours, based on indirect inference from our observations), making it difficult to assess the cell death process via live observation. Therefore, to study the process of SG death progression, we developed an ex vivo testis culture system: first, dissected testes were stained with Lysotracker for 30 minutes to label SGs that were already dying. Testes were then transferred to Lysotracker-free medium and cultured for 8 hours (Fig. 6A). At the end of the culture period, testes were fixed and stained for Vasa, LamDm0, and DAPI to identify dying SGs as described above (Supplementary Fig. S4). With this method, SGs that were already dying before the culture period can be identified as Lysotracker-positive dying SGs, whereas SGs that initiated the cell death process during the culture period can be identified as Lysotracker-negative dying SGs (Fig. 6B). We restricted our analysis to the Lysotracker-negative dying SGs, because these allowed us to precisely follow how SGs initiated the death process and progressed through the cell death stages during the 8-hour chase period.
Using this approach, we compared control and spict mutant testes to see whether spict mutants might show any defects in the process of SG death progression. After the 8-hour chase period, control testes had 0.77 ± 0.15 Lysotracker-negative dying SGs/testis (N = 119), whereas spict mutant testes had 0.55 ± 0.16 dying SGs/testis (N = 123 testes, p = 0.028). This result suggests that spict mutants are defective in initiating SG death, which is consistent with the results described in Fig. 3. In addition, 38% of dying SGs in control testes were observed to be in phase 2 and 3, indicating that some dying SGs in control testes progressed through the SG death phases after initiation during the 8-hour culture period (Fig. 6C). In contrast, in spict mutant testes, only 21% of Lysotrackernegative dying SGs progressed beyond phase 1 (p = 0.032). These results indicate that spict mutants are also defective in progressing through the SG death phases after the initiation of death (Fig. 6C). This further indicates that it takes longer for SGs to complete the cell death process in spict mutants. If it takes longer for each dying SG to complete the death process in spict mutants, the frequency of SG death in spict mutants is likely an overestimate, when estimated based on the number of dying SGs in a fixed sample (Fig. 3C). Thus, the decrease in SG death in spict mutants described above is likely more profound than is shown in Fig. 3C. In summary, we conclude that spict is required for initiation and progression of SG death in the Drosophila testis.

Discussion
In this study, we identified spict as a gene that is specifically expressed in the differentiating cyst cells (CCs) of the Drosophila testis and that it can serve as a novel marker to better identify cyst stem cells (CySCs). We showed that Spict protein is specifically stabilized in the CCs that envelope dying SGs, and that spict is required for the normal progression of SG death. Our study indicates that CC-mediated SG death plays a critical role in allowing the tissue to cope with protein starvation in order to maintain the germline stem cell (GSC) pool.
Our data show that dying SGs are encapsulated entirely by Rab7-positive vesicles, and these vesicles are contained within a single CC. Thus, we propose that dying SGs are phagocytosed by neighboring CCs. Our earlier study showed that the apoptosis of a CC is required to trigger SG death 17 . Now, our study suggests that the remaining CC encapsulates the dying SGs, further elucidating how SG death occurs in the Drosophila testis. We propose the following scenario of CC-mediated SG death: first, one of the two CCs that encapsulate the SGs undergoes apoptosis in response to certain stimuli, such as protein starvation. Death of one CC breaks the 'blood-testis-barrier' generated by the pair of CCs 43 and leads to SG death 16 . The remaining CC ('surviving CC'), or possibly another neighboring CC, now engulfs the dying CC and the SGs to clear the dead cells via phagocytosis. Spict protein was specifically upregulated in surviving CCs and apparently transferred to dying SGs. Considering that the progression of SG death is slower in spict mutants compared to the control, we speculate that spict may be required for the progression of phagocytosis, the subsequent digestion of engulfed dead cells, and/or the recycling of digested SG material in the surviving CCs. However, we did not observe any differences in the frequency of large, Rab7-positive phagosome formation between control and spict mutant testes. In addition, we did not detect any differences in the degree of lysosome/phagosome acidification in control vs. spict mutant testes. Therefore, it remains unclear exactly how spict may promote the progression of SG death.
Programmed cell death can be induced in a cell-intrinsic (suicide) or -extrinsic (murder) manner 44 . Recently, engulfment genes were shown to be required for the developmentally programmed death of the B.alapaav cell in C. elegans by assisting with the cell death processes 45 . In addition, engulfment genes were also shown to promote the unequal segregation of apoptotic potential to induce NSMsc cell death in C. elegans 46 . Likewise, it was shown that in Drosophila oogenesis, the phagocytic machinery of follicle cells is required for developmentally programmed death and removal of nurse cells 47 . Starvation-induced SG death in the Drosophila testis described in this and our previous study 17 holds a striking similarity with these cell death processes in that the cell death precedes via a complex interaction of multiple cell types. SG death depends on the apoptosis of a CC, and later, its removal/clearance depends on the phagocytic activity of the remaining CC. Our data suggest that spict is likely required for promoting the process of SG removal after phagocytoses by the remaining CC.
In summary, our study identified spict as a gene expressed in the differentiating CCs in the Drosophila testis. spict is required for the proper progression of SG death and for maintaining the GSC population during protein starvation. Through a detailed characterization of Spict localization, we revealed a highly regulated process of SG death that involves CC death and phagocytosis by the surviving CC, and our data suggest that spict may be involved in this process. We propose that carefully regulating the death of transit-amplifying cells during starvation is a critical mechanism to preserve the stem cell population and that the transit-amplifying cell population serves as a major point of regulation in shifting tissue homeostasis.

Materials and Methods
Fly strains and husbandry. Flies were cultured in standard Bloomington medium at 25 °C. For protein starvation experiments, newly eclosed adult flies were transferred within 24 hours (day 0) onto either standard food (fed) or 16% sucrose/0.7% agar (starved) at a density of 20-40 flies per vial. Flies were transferred to fresh vials every three days. The following fly stocks were used: c587-gal4 19 , nos-gal4 48 . UAS-spict-mRFP, spict 41  Generation of spict 65 allele. Two target sequences (AACAGAGC|AAGTGAGTCATA AGG and GCAAGGGA|TGTAACTAGACC TGG) for CRISPR-mediated knockout were selected to delete the whole second exon of spict. These constructs were co-injected into vas-Cas9 ZH-2A flies. The genotype of potential mutants was determined by PCR and sequenced to confirm the deletion. spict 65 contains the deletion with imprecise repair at the junction resulting in a short, in-frame insertion (underlined, CAGAAC-AGAAACAGA-ACA) and internal deletion of spict.
Scientific RepoRts | 7:40245 | DOI: 10.1038/srep40245 Immunofluorescent staining and microscopy. Immunofluorescent staining of testes was performed as described previously 51 . Briefly, testes were dissected in PBS, transferred to 4% formaldehyde in PBS and fixed for 30-60 minutes. The testes were then washed in PBS-T (PBS containing 0.1% Triton-X) for at least 30 minutes, followed by incubation with primary antibody in 3% bovine serum albumin (BSA) in PBS-T at 4 °C overnight. Samples were washed for 60 minutes (three 20-minute washes) in PBS-T, incubated with secondary antibody in 3% BSA in PBS-T at 4 °C overnight, washed as above, and mounted in VECTASHIELD with DAPI (Vector Labs). The following primary antibodies were used: mouse anti-Adducin-like (hu-li tai shao - Fly

Lineage tracing of spict-expressing cells.
To lineage-label cells that once expressed spict, spict-gal4 was used to drive UAS-FLP, which removes the stop codon between the actin promoter and gal4 (spict-gal4, UAS-FLP, Act > stop > gal4, UAS-GFP, tubP-gal80 ts ). Cells that once expressed spict will be labeled with GFP. spict-gal4 activity was repressed by tubP-gal80 ts when flies are cultured at 18 °C, ensuring that no labeling occurred before adulthood. Newly eclosed flies were shifted to 29 °C to activate spict-gal4 expression.
Pulse-chase experiment to track SG death progression. Freshly eclosed control (spict 65 /CyO or spict 41 /CyO) and spict mutant (spict 64 /spict 41 ) flies were starved for three days. Testes were dissected directly into Schneider's Drosophila medium (Thermo Fisher Scientific) on a glass dissection dish and soaked in Lysotracker Red (1:1000 in Schneider's media) in the dark for 30 minutes (tubes were rotated to ensure even staining). The tissues were then rinsed with media three times prior to the chase period to avoid carry over of extra Lysotracker. < 5 testis were transferred into 20 μ l-droplets of media on a 35 mm petri dish. To prevent evaporation, extra media droplets were added on the dish near the tissue-containing droplets, and the dish was sealed with the parafilm. The samples were then kept in dark at room temperature for 8 hours (chase period) prior to fixation and staining.