Results

ISCs are known to be difficult to kill. Genetic manipulation that induces apoptosis in other cell types does not kill them (Jiang et al, 2009; Lu & Li, 2015; Ma et al, 2016; Singh et al, 2016). Even complete starvation does not kill ISCs, which outlive the organism during starvation (Fig S1A and B). We recently found the first BH3-only protein, Sayonara (synr), in Drosophila (Ikegawa et al, 2023). Synr activates caspases through the Bcl-2 pathway, relatively moderately compared with Rpr-induced DIAP1 inhibition (Fig 1A) (Ikegawa et al, 2023). We found that mild activation of caspases by Synr, while inducing cell death in the wing disc, causes massive expansion of esg>GFP-labelled progenitor cells rather than cell death in the R4 region, where tissue turnover readily occurs (Ciesielski et al, 2022) (Fig 1B–D). Synr induces mitosis (Figs 1D and S1C), which occurs only in ISCs among cells in the gut (Miguel-Aliaga et al, 2018). In addition, the hyperplasia was observed when Synr was expressed specifically in ISCs (Fig S1D). Thus, although we cannot completely rule out an involvement of miss-differentiation in the expansion of esg>GFP cells, ISC proliferation contributes to the phenotype. Other components of the Bcl-2 pathway, such as buffy, debcl, and dark, also induce ISC proliferation and are necessary for Synr-induced ISC proliferation (Fig 1E and F). Caspase inhibition by microRNA for rpr/hid/grim (Siegrist et al, 2010) suppresses Synr-induced ISC proliferation (Fig S1E). The expression of p35 suppresses Synr-induced ISC proliferation, indicating that the phenotype depends on executioner caspases (Fig 1G). Specifically, the executioner caspase Dcp-1, which also works downstream of Synr in wing discs (Ikegawa et al, 2023), was necessary for the phenotype (Fig 1H). Dcp-1 expression for a short period was sufficient to induce ISC proliferation (Fig 1I), whereas its longer expression ablated ISCs (Fig S1F).

What mediates caspase-induced ISC overproliferation? Among pathways that we investigated, we found that JNK is a key: JNK inhibition suppresses Synr-induced ISC proliferation (Fig 2A). Synr activates JNK signalling, which was detected by the AP-1 reporter (Chatterjee & Bohmann, 2012), through Dcp-1 (Fig 2B and C). JNK activation is sufficient to induce ISC proliferation without activation of Dcp-1 (Fig 2D and E), which is consistent with JNK’s role in ISC proliferation (Biteau et al, 2008). Altogether, these data indicate that Synr/caspase signalling induces cell proliferation through JNK. Literature suggests that JNK can induce proliferation through Yki both autonomously and non-autonomously (Sun & Irvine, 2011; La Marca & Richardson, 2020). Indeed, we found that Synr induces transcription of a Yki target DIAP-1 (Fig 2F) and that Yki inhibition suppresses Synr-induced ISC proliferation (Fig 2G).

Thus far, we elucidated that caspases induce proliferation through JNK/Yki signalling in ISCs. As previously mentioned, there is accruing evidence that signalling from JNK to caspases through rpr/hid induces caspase activation (Luo et al, 2007; Shlevkov & Morata, 2012). Most of the previous studies were performed in imaginal discs. In addition to the JNK/caspase signalling, the caspase–JNK pathway, which we showed above, has been previously described, albeit only in the context of cell death (Shlevkov & Morata, 2012). The positive feedback between JNK and caspase was suggested to function as a powerful feedback loop that ensures cell death in the wing disc (Shlevkov & Morata, 2012). We also validated the JNK/caspase feedback in the wing disc by confirming that JNK activates caspase, that caspase activation through either Synr or Rpr activates JNK, and that inhibition of JNK suppresses Synr-induced caspase activation (Fig3A–D). In contrast to the previous model that solely focused on cell death, because JNK clearly induces proliferation through Yki, we propose that this feedback between JNK and caspase has a potential to activate cell death and proliferation simultaneously. In our model, which outcome, death or proliferation, results from the feedback circuit depends on a balance of signals. Clearly, the balance is skewed towards cell proliferation in ISCs.

We hypothesized that if the feedback from JNK to caspases, which should operate independently from JNK-Yki–induced proliferation, is defective, it could weaken caspase-induced cell death, potentially favouring proliferation more. First, we tested whether Synr can induce transcription of rpr, which mediates JNK-induced caspase activation, in ISCs by performing qRT-PCR. We found that Synr cannot induce rpr efficiently in ISCs, in contrast to its induction in wing discs (Fig 4A and B). We also clarified that Synr or JNK activation cannot induce rpr expression in ISCs using rpr-lacZ (Fig 4C and D). Because Synr activates JNK itself (Fig 2B), we postulated that the chromatin status around rpr might be closed in ISCs. To test this hypothesis, we collected esg-GFP cells by FACS and performed ATAC-seq, a method to identify open chromatin regions. ATAC-seq demonstrated that the chromatin structure around rpr is closed (Fig 4E), which is in clear contrast to the open chromatin of rpr in the wing discs (Fig S2), which was analysed from the previous literature (Vizcaya-Molina et al, 2018). We also noted that the previous literature had suggested that ectopic Rpr cannot induce apoptosis in ISCs (Jiang et al, 2009; Lu & Li, 2015; Ma et al, 2016; Singh et al, 2016) and confirmed it ourselves: Rpr, which readily causes apoptosis in the wing disc, induces no effect or a very mild effect on ISCs, never leading to their complete depletion (Figs 5A and B and S3A).

These data indicate that there is a two-tiered mechanism that inhibits rpr in ISCs: rpr induction and its protein function. This can be considered as a fail-safe mechanism. We further pursued why overexpressed Rpr, which is usually very potent to induce cell death in other tissues (Goyal et al, 2000; Shlevkov & Morata, 2012), was not functional in ISCs. Rpr becomes functional when it localizes to mitochondria (Sandu et al, 2010). We found that Rpr is not specifically localized to mitochondria in ISCs (Fig 5C), which is contrasting to its localization in mitochondria of the wing disc cells (Sandu et al, 2010), suggesting a possibility that ectopic Rpr cannot induce cell death efficiently because of its failed localization to mitochondria. To test this idea, we forced localization of Rpr to mitochondria by expressing rpr that has a mitochondrion localization sequence (Sandu et al, 2010). Forced localization of Rpr to mitochondria depleted ISCs completely (Fig 5B and D) in the whole gut, except the R3 region (Fig S3B). To exclude a possibility that the difference between Rpr WT and Rpr-MTS is due to some genetic backgrounds, we newly generated UAS-Rpr WT and UAS-Rpr-MTS by inserting each on the same landing site (attP2). In this setting, Rpr-MTS (attP2) was more potent than Rpr WT (attP2) (Fig S3C). Furthermore, we also compared Rpr targeted to either mitochondria, ER, or Golgi and found that its targeting only to mitochondria ablates ISCs (Fig 5D). To localize WT Rpr to mitochondria using a different approach, we combined Rpr and the weak expression of Hid, which can help localize Rpr to mitochondria (Sandu et al, 2010). The low expression of Hid itself did not kill ISCs, but its combination with Rpr killed ISCs (Fig 5E), supporting the idea that localization of Rpr to mitochondria is defective in ISCs.

Our data indicate that caspase signalling leads to proliferation in ISCs because the feedback from JNK to caspase is defective. If this is the case, we should be able to induce massive proliferation in other tissues by creating a situation like ISCs. As a proof of principle, we mimicked the situation of ISCs by expressing microRNA for rpr/hid/grim (Siegrist et al, 2010) in wing disc cells. This arrangement should inhibit the feedback from JNK to caspase, similar to the status of ISCs. We found that a combination of JNK activation and the miRNA leads to cell proliferation (Fig 6A), validating our hypothesis based on the feedback loop model.

Discussion

Here, taking advantage of the phenomenon that moderate activation of caspase leads to ISC proliferation, we demonstrate the existence of the positive feedback loop between JNK and caspases, which can promote both cell death and proliferation. This modifies the previous feedback loop that was suggested to promote only cell death (Shlevkov & Morata, 2012). In ISCs, JNK-mediated caspase activation is defective because of the two-tiered fail-safe mechanism, leading to cell proliferation.

We propose that this feedback between JNK and caspases could be a general mechanism that regulates cell death and proliferation. The resulting outcome depends on the balance of this feedback. We note that if the feedback mechanism from JNK to caspase is defective as in ISCs, not only cell death signals but also proliferation signals should be suppressed (Fig 6B). Thus, for this model to be consistent with our observation, proliferation needs to be more sensitive to the feedback circuit than cell death. This is consistent with the idea that moderate caspase activation induces a variety of non-lethal phenotypes, whereas high caspase activity induces apoptosis (Florentin & Arama, 2012; Nakajima & Kuranaga, 2017; Eskandari & Eaves, 2022). We also found that Dcp-1 activation initially induces proliferation in ISCs (Fig 1I), which is followed by cell death if Dcp-1 is expressed for a longer time (Fig S1F). This is consistent with the previous study that implicated Dcp-1 in ISC death induced by inhibition of lipolysis (Aggarwal et al, 2022). We reason that Dcp-1 does not require the feedback mechanism to induce apoptosis when it is highly expressed. In general, if cell death signalling is strong enough, that is, if activated caspases cleave many important substrates, cells should die no matter how strong the proliferation signal is.

Previously, it was noted that JNK activation or its upstream activator Eiger/TNF induces either cell death or proliferation. Although Ras was suggested to affect JNK signalling during Ras-mediated transformation, there has been no conceptual framework to interpret seemingly contradictory phenotypes under the physiological condition (Enomoto et al, 2015; La Marca & Richardson, 2020). We provide a conceptual framework to interpret these dichotomous observations.

JNK is known to induce cell proliferation both autonomously and non-autonomously (Biteau et al, 2008; Sun & Irvine, 2011; Enomoto et al, 2015; La Marca & Richardson, 2020). Although we cannot rule out a possibility that JNK-Yki–mediated ISC proliferation involves non–cell-autonomous signals, we propose that the feedback loop likely occurs cell-autonomously in ISCs because ISCs are sparsely located, we manipulate genes autonomously, and multiple reports previously demonstrated JNK induces ISC proliferation autonomously (Biteau et al, 2008; Buchon et al, 2009; Biteau & Jasper, 2011; Meng & Biteau, 2015; Loudhaief et al, 2017; Hu & Jasper, 2019; Mundorf et al, 2019; Herrera & Bach, 2021). Thus, most likely this is different from the non-autonomous compensatory proliferation or apoptosis-induced proliferation, which is driven by cell death of surrounding cells (Perez-Garijo et al, 2004; Ryoo et al, 2004; Kondo et al, 2006; Fan & Bergmann, 2008; Mollereau et al, 2013).

Physiologically, why does the feedback system that simultaneously promotes death and proliferation exist? Previously, we demonstrated that oncogenic signalling that induces proliferation often causes cell death (Nishida et al, 2021). A similar concept could be applied here. We consider that the ancestral role of caspases is in cell proliferation, which is supported by literature (Shinoda et al, 2019; Eskandari & Eaves, 2022). This also intuitively makes sense: the scenario that a mechanism of cell proliferation acquired cell death seems more probable than the other way around. Thus, we reason that even in the modern time, caspases could play the ancestral role in regulating cell proliferation. To support this idea, the endogenous expression of Dcp-1 was shown to be important for imaginal disc growth previously (Shinoda et al, 2019).

Regarding ISC resistance to cell death, we speculate that such a system exists because tissue stem cells are important for organisms and should not die easily. Among ISCs, it is interesting to note that ISCs in the R3, gastric stem cells in the copper region, are the most resistant to apoptosis. We speculate that the harsh acidic environment of the R3 (Strand & Micchelli, 2011) might make gastric stem cells especially resistant to cell death. Resistance to cell death can be accomplished through a variety of mechanisms. In ISCs, suppression of rpr is mediated by the two-tiered fail-safe mechanism at the chromatin and protein levels. The epigenetic regulation of rpr is similar to its regulation in the salivary gland (Zhang et al, 2014), and distinct from apoptosis resistance in cell cycle–arrested cells (Ruiz-Losada et al, 2022). The role that caspases play in ISCs is distinct from their apoptotic or non-apoptotic roles in enteroblasts (Reiff et al, 2019; Arthurton et al, 2020; Lindblad et al, 2021). Because the ancestral status such as the preservation of genes that predate animal origin is a key feature of tissue stem cells (Alie et al, 2015), we speculate that ISCs may accomplish their relative immortality by keeping the ancestral role of caspases.

Materials and Methods