Distinct regenerative potential of trunk and appendages of Drosophila mediated by JNK signalling

Regeneration is a widespread phenomenon that has been observed in many different animal groups (reviewed by Tanaka and Reddien, 2011); it refers to the response process triggered by injury or mutilation of organs or tissues that aims to reconstruct the damaged structure. Principal factors in regeneration are the induction of additional cell proliferation and the reprogramming of the resident cells that are necessary to rebuild the lost structures.

Experimental analysis in organisms with high regenerative potential is hampered by the limited number of genetic tools available. In contrast, the sophisticated genetic technology of Drosophila permits genetic manipulations not possible in other model organisms. A large body of classical studies described the regenerative capacity of the imaginal discs (the precursors of the adult skeleton). The experiments consisted of transplantation of fragments of imaginal discs into adult female hosts and the subsequent study of how those fragments regenerate missing parts (Hadorn and Buck, 1962; Nöthiger and Schubiger, 1966; Schubiger, 1971; Haynie and Bryant, 1976; reviewed by Worley et al., 2012). These experiments demonstrated that the wing and leg imaginal discs possess high regenerative potential, as some disc fragments were able to rebuild the whole discs. This method, however, had some limitations: it involved in vivo culture in heterologous medium and also relied on cutting disc fragments by hand under the dissecting microscope, which often resulted in imprecision in the size and shape of the fragments to test.

More recently, the development of new methods has allowed a more reliable and rigorous approach to study regeneration. The Gal4/UAS/Gal80^TS and other binary systems allow targeted ablation of well-defined body regions in time- and stage-dependent manners and the subsequent study of the regenerative response (Smith-Bolton et al., 2009; Bergantiños et al., 2010; Sun and Irvine, 2011; Herrera et al., 2013). Those studies have been focused on the wing disc and particularly the wing pouch, the region that corresponds to the appendage part of the wing disc.

The Drosophila body comprises two major parts: the body trunk and the appendages. The former constitutes the greater part of the body, whereas the appendages appear as additional structures that may have originated as outgrowths of the trunk. Moreover, their developmental processes are very different from those of the trunk. For example, in the case of the wing disc, the spatial positioning and function of major patterning genes such as hedgehog (hh), decapentaplegic (dpp) and wingless (wg) are different in the trunk region and the appendage; they induce different targets and their overexpression has different developmental consequences (reviewed by Morata and Sánchez-Herrero, 1999). Moreover, the function of Hox genes in the wing and haltere discs differs in trunk and appendages; in the former the specificity of Hox function requires critically the concourse of the Homothorax/Extradenticle (Hth/Exd) proteins, which are not active in the appendages. These distinct developmental features may also result in different regenerative responses to damage.

The wing disc is a convenient system with which to study differences between trunk and appendages because it contains a well-delimited trunk region, termed notum (the thoracic part), and an appendage region, which includes the wing pouch and the hinge (Fig. 1E). Recent reports have described the regeneration capacity of the appendage region (Smith-Bolton et al., 2009; Bergantiños et al., 2010; Sun and Irvine, 2011; Herrera et al., 2013) and found that it possesses strong regenerative potential. However, in these studies, the regeneration potential of the notum was not investigated.

Here, we use the wing disc to analyse the response to ablation of different regions of the prospective notum and wing. We find that, in contrast to the appendage regions, the notum shows very limited regenerative response. We also describe the response of notum cells to elimination of the entire wing primordium and that of the wing cells to massive apoptosis of the notum precursors. In the first case, the notum cells acquire a high proliferative rate, but are unable to regenerate a wing; they form instead a notum duplicate. In the second case, the wing cells close to the notum also overproliferate, but they do not regenerate a notum, they produce a wing outgrowth. In both cases, Jun-N terminal kinase (JNK) signalling emanating from dying cells is responsible for the increase in cell proliferation. Observations of the haltere disc indicate that the regenerative response of the trunk region (metanotum) and appendage (haltere) mimics that of the homologous regions in the wing disc.

Our results indicate that, even though they are parts of the same imaginal disc, Drosophila trunk and the appendages become developmentally isolated from each. Our results also highlight the involvement of the JNK pathway in generating the additional cell proliferation required for the regeneration process. They also suggest that the different regenerative potential of the notum and wing may be due to differential function of JNK, for we show that forcing JNK activity has different effects in the notum and in the wing.

To study the response to damage of defined regions within the notum, we forced expression of the pro-apoptotic gene reaper (rpr) driven by the pnr-Gal4, vein-Gal4 or 248-Gal4 lines (see Materials and Methods). The expression of the Pnr line defines the medial region of the notum and the Vein and 248 lines cover most of the notum, except the more medial region (Fig. 1A-C).

In the Pnr experiments, after 40 h of rpr activity, the domain was remarkably reduced and only a few pnr-GFP cells were detected, most of which were in apoptosis, as indicated by Dcp-1 staining (Fig. 2B,B′). Moreover, the expression of the notum marker Eyegone (Eyg) (Aldaz et al., 2003) was also significantly reduced in the ablated discs (Fig. 2D,D′). Staining with the anti-Wg antibody shows the thoracic Wg stripe just in the proximal border of the notum region (Fig. 2D,D″), whereas in control discs it occupies a more central position, at the border of the normal Pnr domain (Fig. 2C,C″). It indicates the elimination of most of the Pnr domain. After 48 h of recovery, the damaged discs were not capable of reconstructing the part of the Eyg territory lost during ablation; the Eyg domain remains reduced in the proximal region (Fig. 2E,E′). Moreover, the Wg stripe remains localised in the proximal border of the ablated disc (Fig. 2E,E″). Furthermore, all the pnr>rpr adult flies (n=30) that emerge after ablation show an abnormal thorax lacking the medial region (Fig. 2G), precisely the Pnr domain in the adult thorax (Calleja et al., 2000) in wild-type flies (Fig. 2F).

We also assayed cell proliferation levels in pnr>rpr discs after ablation and did not observe any significant difference in EdU label in the areas close to the ablated Pnr domain (Fig. 2H-I′ and Fig. S3A,B,D). These results indicate that the Pnr domain has little or no capacity to regenerate after ablation.

The results of the ablation of the Vein domain also indicate a similar conclusion. At the time we initiate the ablation, the Vein domain covers most of the notum (Fig. S1A). After the regular cell killing protocol, the Vein domain fails to regenerate. Even after 48 h of recovery the notum remains vestigial (Fig. S1C). This lack of regeneration is also visualised by Eyg staining, which is much reduced in comparison with the wild type, and also by the absence of significant proliferative stimulus in the areas around the ablated region (Fig. S1E,E′). Similar results were obtained by forcing rpr activity in the 248 domain, which results in wing discs that lack the corresponding region even after 48 h of recovery (Fig. S1G-I′) and do not show a significant proliferative response to the ablation (Fig. S1J-J″). Putting together the results obtained after ablation of the Pnr, Vein and 248 domains, the general conclusion is that the notum possesses very limited regenerative capacity.

In the following experiments, we analyse in the wing and haltere discs the response to the ablation of the entire appendage regions. These discs form the dorsal cuticular structures corresponding to the second and third thoracic segments. We have used the sd-Gal4 line as driver and UAS-rpr and UAS-hid as pro-apoptotic vectors. The Sd expression domain covers homologous regions in the appendage part in both discs (Fig. 1G,H). In these experiments, in addition to targeting the domain for ablation with either rpr or hid, we have traced the cell lineage of the targeted area using a UAS-Flp act>stop>lacZ cassette combination (see Materials and Methods).

We first describe the results obtained with rpr as pro-apoptotic agent. Wing discs from sd>rpr larvae are grossly altered after 40 h of rpr activity. They are very small and their morphology indicates that the regions corresponding to the wing pouch and hinge are lacking, except for a small group of cells with intense Dcp1 label close to the notum region (Fig. 3B). In effect, the treatment with UAS-rpr for 40 h amounts to a physical amputation of the Sd domain.

Discs allowed to recover for 48 h after ablation are still abnormal; they are small and show no sign of regeneration of wing structures (Fig. 3C). Instead, their morphology and the distribution of the Wg and Eyg proteins (Fig. 3F,G) suggest a duplication of notum patterns, which is especially clear in the case of the Wg stripe (Fig. 3G). Definitive proof of the duplication process is that the great majority of flies (270 out of 300 flies examined) that emerge in this experiment presented duplicated nota in mirror-image fashion (Fig. 3D).

The results in the haltere disc mimic in every aspect those found in the wing disc. After ablation it becomes very small and does not regenerate the appendage part after recovery. The flies that emerge lack the haltere appendage, although the response of the metanotum is hard to describe due to lack of morphological landmarks. Nevertheless, in mature haltere discs we observe a mirror-image disposition of the Wg stripe (Fig. 3H), clearly pointing to a duplication of the metanotum.

An important issue was to determine the origin of the cells that form the duplicated trunk structure. Duplicated notum structures have been reported to be associated with the loss of wg function in the wing primordium (Ng et al., 1996; Morata and Lawrence, 1977); the lack of wg activity causes the cells of the wing primordium to acquire notum fate. In our experiments the wing primordium is effectively eliminated by rpr activity; thus, we expected that the duplicate notum should derive from cells of the original notum. This issue was addressed by tracing the lineage of the wing territory subjected to ablation by using a UAS-Flp transgene in combination with the act>stop>lacZ cassette. This cassette recombines in sd-Gal4-expressing cells, thus generating multiple lacZ-expressing clones that are marked by β-galactosidase (β-gal) activity. In controls, all cells of the Sd territory were marked with β-gal (Fig. 3I,I′). In sharp contrast, sd>rpr discs subjected to 40 h ablation and allowed 48 h of recovery showed very few cells labelled with β-gal (Fig. 3J,J′). The conclusion from these experiments is that the duplicated structures do not derive from cells of Sd lineage but from cells of the original notum.

The experiments using hid as pro-apoptotic agent yielded different results. After 40 h of treatment, the morphology of the Sd domain, wing pouch and hinge, remains normal, despite the presence of many Dcp-1-labelled cells (Fig. S2B). After 48 h of recovery the Dcp-1 label disappears and most of the discs present with near-normal morphology in the appendage region, suggesting they have regenerated the ablated territory. Moreover, lineage tracking of the ablated Sd domain with β-galactosidase label after 48 h of recovery indicates that the regenerated wing territory derives from cells of the original Sd domain (Fig. S2D,D′). This regeneration is visualised in adult flies that differentiate wing structures of variable size and near-normal pattern (Fig. S2E,F). These results reinforce previous observations indicating the strong regenerative capacity of wing cells.

The different results obtained with UAS-rpr and UAS-hid as pro-apoptotic factors most likely derive from their different cell killing capacities. UAS-rpr appears to be a very effective cell death inducer and essentially eliminates the Sd region. UAS-hid is less effective and allows the survival of a sufficient number of Sd cells that regenerate wing structures.

The key observation from the preceding experiments is that after the complete elimination of the wing or the haltere primordium, the trunk (notum or metanotum) cells cannot regenerate the corresponding appendage. Instead, they form mirror-image duplicates of trunk structures. This is significant, as it suggests a limitation in the reprogramming of the trunk cells. In what follows, we have analysed developmental events associated with the appearance of these duplicates.

In the sd>rpr experiment, we measured the proliferative response to the elimination of the entire presumptive wing region. After 40 h of rpr activity, EdU levels are increased in cells close to the ablated tissue (Fig. 4B,B′, Fig. S3C and Fig. S4B,B″), identified by the high Dcp-1 levels. To check the possibility that some of the overproliferating cells are of hinge identity, we performed a double staining of EdU and the hinge marker Zinc finger homeodomain 2 (Zfh2) (Whitworth and Russell, 2003; Terriente et al., 2008). As illustrated in Fig. S4B-B″, the great majority of the over-proliferating cells do not express Zfh2, indicating that the duplicated nota derive from the original notum. This zone of high proliferation also shows elevated levels of dMyc (Fig. 4D,D′) and of the Yorkie targets expanded-lacZ (ex-lacZ) (Hamaratoglu et al., 2006) (Fig. 4F,F′) and diap1-lacZ (Huang et al., 2005) (Fig. 4H,H′). Similar observations were also made in homologous regions of the haltere disc (Fig. 4F,F′,H,H′).

As the ablation of the Pnr and Vein notum regions (Fig. 2 and Fig. S1) did not elicit a proliferative response, we surmised that the proliferative response in the notum cells of sd>rpr discs could be due to signals emanating from the dying wing cells. It is known that apoptotic cells secrete proliferative signals that may cause overgrowths and that the signalling process requires JNK activity (Ryoo et al., 2004; Pérez-Garijo et al., 2004, 2009). Consequently, we examined ablated sd>rpr wing discs for JNK activity, which in the wild type is restricted to the most-proximal region (Agnès et al., 1999).

To monitor JNK activity, we examined the levels of two well-known JNK targets: the metalloprotease 1 protein (Mmp1) (Srivastava et al., 2007; Stevens and Page-McCaw, 2012) and the puckered (puc) gene (Martín-Blanco et al., 2000). The results are illustrated in Fig. 5. In non-ablated discs, puc expression is restricted to some proximal cells (Fig. 5A,A′), but in ablated discs both puc expression and Mmp1 protein levels accumulate in Dcp-1-expressing cells (Fig. 5B,B′,D,D′), indicating upregulation of JNK in the dying wing cells.

To assay the functional connection between JNK activity in the dying cells and the increase of cell proliferation of their neighbours, we compromised JNK activity in the ablated Sd domain by overexpressing puc, which functions as an inhibitor of the pathway (Martín-Blanco et al., 1998). In sd>rpr UAS-puc discs, cell proliferation levels in the proximity of the dying wing cells were reduced (Fig. 5F,F′) in comparison with those in sd>rpr discs in which JNK is functional (Fig. 5E,E′). We also observed that the inhibition of JNK results in reduction in the amount of tissue eliminated in the Sd domain. Although there is strong Dcp-1 label in the domain (Fig. 5F,F″), there are more cells surviving 40 h of rpr treatment. Our explanation for this difference is that inhibiting JNK interrupts the apoptosis amplification loop (Shlevkov and Morata, 2012), which results in lower apoptotic levels. This result illustrates the functional significance of the loop, which amplifies the strong apoptotic response generated by the UAS-rpr transgene.

Ablated sd>rpr and sd>rpr UAS-puc larvae can develop to adulthood; thus, we could compare their adult phenotypes. The results are shown in Fig. 5G: 87% of sd>rpr flies have duplicated nota compared with only 18% in sd>rpr UAS-puc flies. Interestingly, 45% of the latter show lack of wings but no notum duplication, and 37% are able to form some wing tissue.

The preceding results identify the JNK pathway as a principal factor responsible for the proliferative signalling originated in the wing cells in apoptosis. They also allow discrimination of the identity of signal-secreting cells (wing) and signal-responding cells (notum); notum cells respond to the signals emanating from dying wing cells. The fact that they cannot regenerate a wing but only form a duplicate of the notum suggests a restriction in the reprogramming capacity between notum (body trunk) and wing (appendage) structures.

The proliferative response of the notum cells to JNK signalling emanating from dying wing cells contrasts with their lack of response to dying notum cells. We checked whether in the pnr>rpr, 248>rpr and vn>rpr experiments JNK is activated after ablation in the notum regions and we found upregulation of Mmp1 and puc-lacZ in the corresponding notum domains (Fig. 6A-C′). These results show that the JNK pathway is activated in the notum in response to injury, but it appears incapable of inducing a proliferative response.

As JNK activity appears to have different effects in the notum and in the wing, we sought JNK targets that are expressed differentially after ablation. Recent work (Harris et al., 2016), has identified a wg enhancer (BRV118) in the wing disc that is specific to damage; within this enhancer the BRV-B module directs the strongest response. This fragment behaves as a target of JNK signalling and contains consensus sites that match the Drosophila AP-1 motif (Harris et al., 2016). We have compared the expression of the BRV-B module after ablation of the appendage cells (sd>rpr discs) and after ablation of notum regions (pnr>rpr, vein>rpr and 248>rpr). The results, illustrated in Fig. 6D-H′, indicate that although BRV-B is strongly expressed in the ablated wing region (Fig. 6H,H′), it is not induced in the ablated notum (Fig. 6E-G′).

The preceding experiments suggested that although the JNK pathway is induced upon ablation in the notum, the consequences are different from those observed in the wing. To explore the idea of differential function of JNK in different territories of the wing disc, we performed an experiment using the Gal4/UAS system to force constitutive activity of JNK in the Pnr domain of the notum and in the Spalt^EPV (Barrio and de Celis, 2004) domain of the wing. We have made use of the UAS-hep^CA transgene that induces high levels of constitutive JNK activity (Adachi-Yamada et al., 1999). Because one major property of JNK is to induce apoptosis (McEwen and Peifer, 2005; Igaki, 2009), the experiments were carried out in dronc mutant background so that affected cells cannot enter apoptosis (Xu et al., 2005). The results are illustrated in Fig. 7. We find that the overexpression of hep causes a large overgrowth in the Sal domain whereas it does not significantly affect the size of the Pnr domain (Fig. 7A-F). This increase in size is associated with upregulation of the Yki pathway, as indicated by the targets expanded and cyclin E (Fig. 7G-L′). This enlarged Sal domain also shows an increase in the proliferation levels, as indicated by the EdU levels (Fig. 7C). This distinct behaviour of JNK in notum and wing may be related to their different regeneration capacity.

To study the response to the ablation of notum precursors, we used the 41D11-Gal4 line: an insertion at the apterous gene (Bieli et al., 2015) that drives expression in nearly all the notum and many hinge cells. Fig. 1D shows the expression of the line at the early larval third period, when we initiate ablation.

In discs fixed just after 40 h of ablation, we observe a clear diminution in the size of the disc, in particular in the notum region (Fig. 8B,B′), most of which is eliminated. There is also strong Dcp-1 label in this region and in the hinge. The loss of notum tissue is also visualised by the expression of Eyg, which becomes reduced to a small patch (Fig. 8D). In discs allowed to recover for 48 h after ablation, the notum region does not regenerate; it remains as a rudiment, as the Eyg label indicates (Fig. 8E). Interestingly, these discs show a large overgrowth of the zone close to the notum cells that alters their normal morphology (Fig. 8E). The position of the overgrown region suggested that it corresponds to hinge structures, as indicated by the expression of the hinge markers Stat (Bach et al., 2007; Ayala-Camargo et al., 2013) and Zfh2, expressed in the overgrown tissue (Fig. 8F-H). We cannot ascertain the identity of this tissue in cuticular structures because no adult flies emerge in this experiment.

The overgrowth of the hinge tissue is associated with an increase in proliferation in the corresponding zone of the disc, as indicated by the accumulation of EdU in the wing and haltere discs (Fig. 8I,I′). The two discs respond similarly to the ablation of the notum/hinge zone.

Following the observation in the sd>rpr experiment that JNK is responsible for the increased cell proliferation close to the ablated zone, we investigated the implication of the JNK pathway on the high proliferation levels observed in the hinge region in the 41D11 experiment. Indeed, staining for puc-lacZ indicates high JNK activity in the ablated region (Fig. 8J,J′). Moreover, the suppression of JNK in the zone, by overexpressing puc, results in almost suppression of the hinge overgrowths (Fig. 8K-L′). Furthermore, unlike the 41D11>rpr experiment, some few 41D11>rpr, UAS-puc flies emerge and differentiate wings that, although abnormal, show few signs of overgrowth.

This result reinforces the conclusion above that the increased cell proliferation triggered by ablation of the disc is mediated by activation of the JNK pathway in the dying cells. In this experiment, the ablated region includes notum and hinge cells, and because in the experiments ablating only notum there is no proliferative response, the additional cell proliferation observed here is most likely due to JNK activity in the hinge cells.

One major subdivision in the segmented body of Drosophila is that between body trunk and appendages, which is especially visible in the adult cuticle. In the case of the second thoracic segment, the dorsal cuticle is differentiated by the wing imaginal disc, which contains a trunk region (the notum) and an appendage, which includes the wing blade and associated hinge structures. The homologous haltere disc also contains a trunk region, metanotum and an appendage, the haltere. In this work, we have compared the regenerative potential of the different regions of those discs, with special attention to the distinction between trunk and appendage structures. The bulk of our results are based on the wing disc but we present evidence that the haltere disc behaves similarly.

Classical transplantation experiments demonstrated the developmental plasticity of the imaginal discs (Hadorn and Buck, 1962; Nöthiger and Schubiger, 1966; Schubiger, 1971; Haynie and Bryant, 1976; reviewed by Worley et al., 2012), and recent reports using the Gal4/UAS/Gal80^TS method have confirmed the strong regenerative capacity of the wing disc (Smith-Bolton et al., 2009; Bergantiños et al., 2010; Sun and Irvine, 2011; Herrera et al., 2013). In addition, in our experiments forcing the pro-apoptotic gene hid in the Sd domain that covers the entire wing/hinge primordium, we find that the surviving wing/hinge cells are able to regenerate much of the lost wing tissue. Those experiments also identified genes and transduction pathways, wg, dMyc, Hippo pathway, JNK, etc., that appear to be associated with the regeneration process.

However, those studies paid very little attention to the trunk regions, one of the aspects we have explored in this work. One unexpected result is that the notum exhibits very limited regenerative potential. The ablation of the Pnr, Vein or 248 domains of the wing disc during the early third instar period leaves a permanent loss that cannot be recovered; the adult flies that emerge lack the corresponding structures (Fig. 2 and Fig. S1). This is in sharp contrast with the response of the wing regions, as mentioned above. The failure of the notum to regenerate matches the observation that, after the ablation of the Pnr and Vn domains, there is no detectable proliferative response in the neighbouring zones (Fig. 2, Fig. S1).

The finding that the notum and appendage respond differently to injury is novel and significant, for it describes a situation in which the same tissue, the wing imaginal disc (which cells originate from a common lineage), has developed two regions with very different regenerative potential. The study of the mechanism responsible for this difference may illuminate the distinct regeneration abilities found in diverse species of animals and in different tissues (Tanaka, 2003). We discuss this issue in connection with the function of the JNK pathway – see below.

We have assayed the regenerative potential of the wing disc after ablation of the entire wing/hinge region – the sd>rpr experiments, or after ablating the entire notum territory, the 41D11>rpr experiments. It is worth emphasising that, in our treatments using rpr, the targeted region is virtually eliminated (Figs 3 and 7). Thus, those experiments amount to a physical amputation.

The sd>rpr experiments yielded unexpected results: the cells close to the dying wing cells exhibit high proliferation levels, but they fail to regenerate the ablated wing/hinge structures and generate instead a notum duplicate. This is an intriguing observation, because it shows that, unlike in the Pnr, Vein and 248 experiments, notum cells can proliferate in response to injury (see below). The notum duplication can be visualised in mature discs after recovery (Fig. 3F,G), and is especially clear in the adult flies that emerge after the treatment, most of which show duplicated notum structures (Fig. 3D). The fact that notum cells in spite of their proliferative response can only develop notum structures is of interest, as it suggests that trunk cells cannot be reprogrammed to develop appendage structures. The fact that haltere discs from this experiment also produces metanotum duplicates (Fig. 3H) reinforces this latter point and makes it more general.

Notum duplications have been reported in classical transplantation experiments (Bryant, 1975), but the interpretation was very different. In those experiments, duplications were not limited to the notum; hinge and wing regions could also duplicate, depending on the size of the regenerating fragment; the behaviour of the fragments would depend on the positional values present in the fragment that would generate new positional values by the ‘shortest intercalation’ and the ‘complete circle’ rules (French et al., 1976).

The 41D11>rpr experiment is complementary to the sd>rpr experiment. Here, the entire notum territory is eliminated and the wing/hinge cells respond to the injury. After ablation of the 41D11 domain, the notum territory is reduced to a minimum and does not regenerate, even after 48 or 72 h of recovery time (Fig. 7E). This failure to regenerate the notum contrasts with the increased cell proliferation in its vicinity. However, as shown by the molecular markers Stat and Zfh2, the overproliferating cells are mostly of hinge identity (Fig. 7G,H). Thus, the appendage cells cannot form notum structures, in spite of their proliferative response.

One general conclusion from these experiments is that both the notum and the wing primordia appear to be incapable of regenerating each other. The genetic programs specific to notum and wing become established during early larval period (Zecca and Struhl, 2002) from a pool of cells of common lineage. Our results clearly indicate that once the notum or the appendage program is established, it cannot be reversed, even after major injury to the disc. Additional observations of the haltere disc (Figs 3, 4 and 8) suggest that the irreversible trunk/appendage restriction is a general feature of the body.

The irreversibility of trunk/appendage identity even after massive damage is intriguing in view of our previous observation (Herrera and Morata, 2014) that anterior and posterior identity can be altered during regeneration: the trunk/appendage subdivision appears much later than the A/P subdivision and yet it is more rigid. We do not know the reasons for this difference, but it appears that the timing of the events is less crucial than the nature of the developmental decision.

The JNK pathway has been related with regeneration processes in Drosophila (Bosch et al., 2005; Bergantiños et al., 2010; Sun and Irvine, 2011; Repiso et al., 2011; Herrera et al., 2013; Herrera and Morata, 2014; Santabárbara-Ruiz et al., 2015) and also in planarian regeneration (Almuedo-Castillo et al., 2014). A relevant feature of JNK in Drosophila is that in many situations it promotes apoptosis (Igaki, 2009), and it is known that apoptotic cells generate proliferative signalling (reviewed by Bergmann and Steller, 2010 and Morata et al., 2011), which in turn is mediated by JNK (Pérez-Garijo et al., 2009).

Our experiments here clearly establish the involvement of JNK in the additional cell proliferation caused by the ablation process in Drosophila. The notum duplications observed in the sd>rpr experiment and the hinge outgrowth in the 41D11>rpr experiment are both associated with and dependant on upregulation of the JNK pathway in the cells in apoptosis. It is worth pointing out that, as clearly shown in the sd>rpr experiment, the JNK signalling is originated in the dying wing cells and received in the notum cells, which have different identity. These results highlight the relevant role of apoptosis in the regeneration phenomena.

The overproliferating cells exhibit upregulated expression of Yki targets such as expanded or diap1 (Fig. 4), indicating downregulation of the Hippo pathway as the vehicle to induce high cell proliferation. The regulation of Hippo signalling by JNK during regeneration has been previously reported (Sun and Irvine, 2011, 2013).

One intriguing aspect of our findings is why the JNK pathway is unable to increase cell proliferation after ablation of notum regions such as the Pnr or Vein domains. This cannot be due to inability of notum cells to respond to proliferative signalling, as shown by the sd>rpr experiment where notum cells overproliferate. Nor is it due to lack of activation of JNK in the notum cells after ablation, as we show it is upregulated (Fig. 6A-C′). It is possible that JNK is activated in the notum at lower levels or that even if activated it cannot generate proliferative signalling. Along this line, we have observed that a JNK target, the BRV-B wg enhancer (Harris et al., 2016), is upregulated in wing cells upon ablation but not in notum cells, and a related observation has been reported in eiger-overexpressing clones (Harris et al., 2016). This already indicates a distinct response of trunk and appendage cells to JNK that may be reflected in their different abilities to induce cell proliferation. The proximal notum region contains resident JNK activity (Agnès et al., 1999) but it does not induce overproliferation.

Moreover, the experiments forcing the expression of JNK in the wing and the notum further support the notion of a differential response to JNK activity in notum and wing: the overexpression of JNK causes a large overgrowth in the Sal domain, correlated with upregulation of the Yki pathway. However, this effect is not observed in the Pnr domain (Fig. 7), which shows no increase in size. This distinct response exhibited by these two territories may be related to their different regeneration capacities.

The Drosophila stocks used in this study were sd-Gal4, pnr-Gal4 and 248-Gal4 (Calleja et al., 1996); vn-Gal4 (Austin et al., 2014) and 41D11-Gal4 (Bieli et al., 2015); sal^EPV-Gal4 (a gift from J. F. de Celis, CBMSO, Madrid, Spain); the BRV-B-GFP line (Harris et al., 2016); tub-Gal80^TS (McGuire et al., 2003); puc-lacZ line (puc^E69) and UAS-puc^14C line (Martín-Blanco et al., 1998); UAS-GFP and UAS-Flp (BDSC); act>stop>lacZ (a gift from G. Struhl, Columbia University, USA); ex-lacZ (ex⁶⁹⁷); CycE-lacZ (BDSC 10384) and Diap-1-lacZ (Diap1^j5C8); and STAT-10xGFP (Bach et al., 2007). STAT-lacZ; UAS-rpr, UAS-hid and UAS-hep^CA, lines are described in FlyBase. dronc^I24 and dronc^I29 are considered null alleles (Xu et al., 2005) and were provided by A. Bergmann (University of Massachusetts, USA).

The Gal4/UAS/Gal80^TS system (Smith-Bolton et al., 2009) was used to induce the expression of the pro-apoptotic genes rpr or hid in the wing disc domains defined by the driver lines: pnr-Gal4, vn-Gal4, 248-Gal4, 41D11-Gal4 and sd-Gal4. The expression of rpr or hid was temporally regulated by the thermo-sensitive Gal80^TS inhibitor, which is active at 17°C and degraded at 29°C.

The standard protocol consisted of growing larvae at 17°C until day 7 after egg-laying (transition from 2nd to 3rd instar), except for the Vn experiments (in which larvae were grown at 17°C until day 6), and then shifting them to 29°C to allow for Gal4 function driving rpr or hid activity. After 40 h at 29°C, larvae were shifted back to 17°C to stop ablation and allow for recovery for an additional 48 h (Fig. 1I) or until the emergence as adult flies. In some experiments, the experimental larvae were kept at 29°C for the rest of development to allow continuous ablation of the targeted domain.

The UAS-Flp, act>stop>lacZ cassette was used to mark and track the progeny of the cells of the targeted domain. As previously described (Herrera et al., 2013), the UAS-Flp transgene becomes activated and produces large amounts of Flipase that recombine the act>stop>lacZ cassette. As a consequence, the great majority of cells in the target domain, and their progeny, are indelibly labelled by β-galactosidase activity.

Immunostaining was performed as described previously (Shlevkov and Morata, 2012). Images were captured using a Leica DB5500 B confocal microscope.

The following primary antibodies were used: rabbit anti-Dcp1 (Cell Signaling, 9578), 1:200; mouse anti-β-galactosidase (DSHB 40-1a), 1:50; rabbit anti-β-galactosidase (Cappel, 55976), 1:2000; mouse anti-Wingless (DSHB 4D4), 1:50; mouse anti-Mmp-1 (DSHB, a combination of 3B8D12, 3A6B4 and 5H7B11), 1:50; guinea pig anti-Eyegone, 1:200 (a gift from N. Azpiazu, CBMSO, Madrid, Spain); mouse anti-Zfh2, 1:150 (gift of F. Diaz-Benjumea, CBMSO, Madrid, Spain); guinea pig anti-dMyc, 1:100 (in-house); and mouse anti-GFP (Roche, 11814460001), 1:200.

Fluorescently labelled secondary antibodies (Molecular Probes Alexa) were used at 1:200 dilution. TO-PRO3 (Invitrogen) was used at 1:600 dilution to label nuclei.

For EdU incorporation, wing imaginal discs were cultured in 1 ml of EdU labelling solution for 20 min at room temperature. Discs were washed once with 1× PBS and then fixed in 4% paraformaldehyde, 0.1% Triton for 1 h at room temperature. EdU detection was performed according to the manufacturer instructions (Click-iT EdU Alexa Fluor 555 Imaging Kit, ThermoFisher Scientific). Antibody labelling of the samples was performed after EdU detection following the immunostaining protocol described above.

Adult wings were mounted in Euparal Mounting medium after having dissected the flies in a mixture of alcohol/glycerine. Images were captured with a Leica microscope.

Images were processed using the ImageJ software. In the hep^CA experiments, GFP-labelled area and total area of the discs were measured by using the Area Fraction option in set Measurements. In the pnr>rpr and sd>rpr experiments, the proliferation index was measured by calculating the ratio of the EdU-labelled area in two regions of the ablated discs. The number of discs analysed in each experiment is given in the figure legend. The P-values were calculated using the two-tailed Student's t-test.

We thank C. Estella for useful reagents and helpful suggestions, and I. Hariharan for fly stocks. We also thank R. González and A. Cantarero for technical support and E. Sánchez-Herrero for critical reading of the manuscript.