RNAi-Mediated Knockdown Showing Impaired Cell Survival in Drosophila Wing Imaginal Disc

The genetically amenable organism Drosophila melanogaster has been estimated to have 14,076 protein coding genes in the genome, according to the flybase release note R5.13 (http://flybase.bio.indiana.edu/static_pages/docs/release_notes.html). Recent application of RNA interference (RNAi) to the study of developmental biology in Drosophila has enabled us to carry out a systematic investigation of genes affecting various specific phenotypes. In order to search for genes supporting cell survival, we conducted an immunohistochemical examination in which the RNAi of 2,497 genes was independently induced within the dorsal compartment of the wing imaginal disc. Under these conditions, the activities of a stress-activated protein kinase JNK (c-Jun N-terminal kinase) and apoptosis-executing factor Caspase-3 were monitored. Approximately half of the genes displayed a strong JNK or Caspase-3 activation when their RNAi was induced. Most of the JNK activation accompanied Caspase-3 activation, while the opposite did not hold true. Interestingly, the area activating Caspase-3 was more broadly seen than that activating JNK, suggesting that JNK is crucial for induction of non-autonomous apoptosis in many cases. Furthermore, the RNAi of essential factors commonly regulating transcription and translation showed a severe and cell-autonomous apoptosis but also elicited another apoptosis at an adjacent area in a non-autonomous way. We also found that the frequency of apoptosis varies depending on the tissues.


Introduction
In recent years, mechanisms controlling apoptosis have been extensively studied, and various factors are known to be involved in the intrinsic and extrinsic apoptotic pathways. [1][2][3][4] Although these pathways play a pivotal role in the execution of most cases of apoptosis, the apoptosis shown in developing animal tissues is also affected by various growth and differentiation signals to promote or repair organ development. 5 In general, inhibition of apoptosis accompanies growth induction, whereas reduction of growth conversely leads to apoptosis. However, we can often fi nd exceptions showing an opposite relationship, such as overgrowth-induced apoptosis [6][7][8] and apoptosis-induced overgrowth, [9][10][11] indicating that we do not fully understand these cell survival controls between apoptosis and growth. In order to systematically investigate the apoptosis phenotype caused by reducing each gene function in the developing animal tissues, we employed a genetically amenable fruit fl y Drosophila melanogaster, in which each gene can be knocked down by RNAi, [12][13][14] to observe the effect on apoptosis induction. RNAi provides an easy and powerful technique for reducing the quantity of mRNA derived from endogenous specifi c genes, and it has recently been applied in many studies to investigate various gene functions. 15 In this study, we screened 2,497 protein-coding genes of Drosophila to determine whether they were required for prevention of apoptosis in the wing imaginal disc and found that 47% of them showed an apoptosis induction when their functions were knocked down by RNAi in the developing wing disc. Most of the cases (82%) with detectable Caspase activation were associated with JNK activation, which was unexpectedly high because JNK has not been observed as essential for all apoptosis. Alternatively, JNK is known to be involved in inducing nonautonomous apoptosis, 16,17 which occurs in cells distant from the cells associated with the primary cause of apoptosis. Interestingly, a major part of the JNK and Caspase-3 activation found in this study occurred in a non-autonomous manner, suggesting that the non-autonomous pathway is a common way to induce apoptosis. Loss of membrane proteins frequently caused JNK activation, which had also been expected because cell-cell communication is presumed to be important for many developmental processes, including apoptosis in multicellular organisms. These results, as well as the database showing the immunofl uorescent data, provide an archival source for survey of genes and for fi ne analysis of each gene in apoptosis regulation using the Drosophila imaginal discs.

Rationale for RNAi-mediated screening for genes regulating apoptosis
We induced RNAi in the dorsal compartment of the wing disc and monitored the activities of Caspase-3 and JNK. Caspase-3 plays a central role in most apoptosis, while JNK leads to a subgroup of stress-induced apoptosis. 18 In the Drosophila wing disc, JNK activation is usually linked to the activation of Caspase-3. 16 Puc is a protein phosphatase specifically inactivating JNK, and its transcription occurs in response to the JNK signal, thereby making a negative-regulatory circuit. 19 Thus, the expression of puc reflects the JNK activity and can be used for monitoring it.
Before expanding the RNAi analyses to the entire genome, we checked whether the mutant phenotypes caused by previously known apoptosisregulating genes, such as diap1 20 and dark, [21][22][23] were reproduced by their RNAi. When diap1 (Drosophila Inhibitor of Apoptosis Protein 1) was knocked down within the dorsal compartment of the imaginal disc, a local but prominent activation of Caspase-3 was detected (Fig. 1B). The positionspecifi city may be dependent on the difference in sensitivity in the induction of apoptosis, as described later. In contrast, when dark (Drosophila Apaf-1-Related Killer) was knocked down, no apoptosis induction was observed (Fig. 1C). Furthermore, the use of this collection of RNAi strains has already been validated, since they were screened for apoptosis phenotype in the compound eye. 24 For the non-autonomously induced apoptosis during restoration of morphogenesis, for example, we tested whether the apoptosis shown in several RNAi samples really refl ected the non-autonomous apoptosis by conventional gene manipulation in previously studies. 16 We manipulated signaling factors for a diffusible extracellular ligand Dpp, a homolog of mammalian BMP (Bone Morphogenetic Protein)−2/4. Mad (Mothers against Dpp), 25 a Drosophila homolog of mammalian r-Smad, transmits the intracellular signal caused by Dpp. As shown in Figure 1E, the RNAi of mad within the dorsal compartment activated JNK and Caspase-3 in both dorsal and ventral compartments of the central wing disc region, which is a typical example of non-cellautonomous induction of apoptosis. These features are also quite similar to those seen in the case of overexpression of Dad (Daughters against Dpp), a homolog of anti-Smad, (Fig. 1F) that can repress Dpp signaling. 26  , and classified the strength of JNK and Caspase-3 activities in each area into three grades (+ −, +, and ++). The database was constructed by using FileMaker Pro 7 (FileMaker, Inc.) and contains each immunofl uorescence image with the above classifi cation of JNK/Caspase-3 activities in each gene page.
First, we noticed that a narrow area in the dorsocentral (DC) wing region showed JNK and Caspase-3 activation too sensitively (e.g. Fig. 2), which was not always correlated with RNAi. Consequently, the results that simply refl ected this feature were excluded from all of the analyses.
When a survey of all of the RNAi experiments was carried out, both weak and strong activation of JNK and Caspase-3 (shown by + and ++ signs in the database) was observed in 41% and 87% of the cases, respectively (Fig. 3A). These proportions seemed much higher than expected because the imaginal disc cells may not actually express such a large number of genes, and they are thought to express several thousand genes. [27][28][29]   number of strains with weak JNK and/or Caspase-3 activation may refl ect false-positive results due to an off-target effect (OTE) or other non-specifi c effects of dsRNA expression. A part of these IR lines was known to have possible OTs and the frequency of such suspected OTE lines was calculated to be 47% as a maximal estimation. Accordingly, we did not give further consideration to this class of strains and hereafter focused on the results showing high levels of JNK and Caspase-3 activation (shown only by the ++ sign in the database), which amounted to 10% and 47% of cases, respectively (Fig. 3A). The immunofl uorescence data can be accessed on the website: http://www.shigen.nig.ac.jp/fl y/nigfl y/ index.jsp (see Experimental Procedures).

Screening summary
Among these strong cases of JNK and/or Caspase-3 activation, more than 80% showed a coupling of JNK and Caspase-3 activation to various extents ( Fig. 3C). Conversely, there were only two cases in which JNK activation did not accompany Caspase-3 activation (calculated as 0.17%). Therefore, these fi ndings are consistent with the previous observation that the JNK activation precedes Caspase-3 activation and is strongly linked to apoptosis in the Drosophila wing. 17 Furthermore, we carefully assessed the non-cell autonomous effect of RNAi by examining the phenotypes in the vicinity of the DV boundary. Most of the RNAi-induced JNK/ Caspase-3 activation showed a striking non-cell autonomy (Fig. 3D), which is similar to the previously known feature in non-cell-autonomous activation of JNK by altered Dpp signaling. 17 This suggests that the non-cell-autonomous induction of apoptosis is one of the common patterns in cell death induction.
It has previously been shown that LRR (Leucine Rich Repeat) family cell adhesion proteins contribute  to unique cell affi nity. Furthermore, alteration of LRR protein functions causes JNK activation followed by Caspase-3 activation. 30,31 Therefore, transmembrane and secreted proteins are good candidates for mediators of non-cell-autonomous apoptosis. We thus tested the apoptosis-inducing activity by RNAi of putative secreted proteins selected based on the presence of N-terminal signal sequences. Consistently, the RNAi of such genes caused a higher frequency of strong JNK activation (74 out of 296 cases, 25%) when compared with the frequency of strong JNK activation in all of the RNAi cases (248 out of 2,497 cases, 10%). Moreover, the RNAi of various transcription factors frequently showed non-autonomous apoptosis, suggesting that they are highly involved in regulation of morphogenesis and that their aberration likely induces non-autonomous apoptosis. These results suggest that a relatively large number of secreted proteins and transcription factors are involved in the prevention of non-cell-autonomous apoptosis. However, functional redundancy may have prevented identifi cation of such molecules. We also noted that there was a significant tendency for various examples of RNAi-mediated apoptosis to be preferentially found in the wing blade region but not outside of this region. For example, the RNAi of taf6 (TBP-Associated Factor 6) leads to an autonomous activation of Caspase-3 at high levels in the wing blade region but at lower levels in the wing hinge region (Fig. 4). Similar traits have also been reported in the case of the apoptosis induced by a reduced-Dpp signal. 17 Furthermore, when we surveyed all of the results, the activation of JNK/Caspase-3 in the wing blade was found to be much more frequent than those in the notum (Figs. 3E and F). Consequently, localization of apoptosis in the wing blade region may not be a feature specifi c to the alteration of a particular cell signal (such as Dpp) but instead a general  feature found in all cases of apoptosis in the wing disc. This suggests that the number of stimuli to activate JNK or apoptosis varies depending on the tissue, or that the sensitivities to alterations of gene expression in induction of apoptosis are quite different between tissues.

Non-autonomously induced apoptosis
As we reported previously, aberrations of some morphogenetic signaling induce JNK activation followed by Caspase-3 activation at the boundary between cell populations with different levels of signaling intensities. This non-autonomous apoptosis is thought to be important for restoration of abnormally developing tissues. 17 Various examples of apoptosis are probably induced in a similar nonautonomous way.
To determine to what extent this applies to non-autonomous apoptosis, we surveyed the relationship in cell autonomy between JNK and Caspase-3 activation by focusing on the DV boundary at which the two cell populations come in contact (Fig. 3D). Around this position, there is an apparent tendency for JNK and Caspase-3 activation to occur simultaneously, which was observed in 471 cases, as shown in the 9 upper-left boxes in Fig. 3D's grid. As stated above, the most frequent pattern is that both JNK and Caspase-3 are both autonomously and non-autonomously activated. However, 112 out of 196 cases with autonomous JNK activation (57%) displayed a non-autonomous Caspase-3 activation (upper-most row in Fig. 3D grid). In contrast, except for 6 cases, autonomous Caspase-3 activation (331 cases) did not show non-autonomous JNK activation (leftmost column in Fig. 3D grid). Accordingly, these data strongly suggest that JNK activation is also crucial for priming non-autonomous apoptosis, whereas Caspase-3 is not. Furthermore, when observed throughout the wing disc, the activation patterns of JNK and Caspase-3 are different (Figs. 3E and F). JNK activation seems to be found unevenly in the dorsal region (2 + 42 + 13 = 57%), whereas the Caspase-3 activation only within the dorsal region is less (2 + 5 + 35 = 42%). On the other hand, JNK with nonautonomous activation is minor (0 + 0 + 16 + 27 = 43%), whereas non-autonomous Caspase-3 is major (0 + 0 + 4 + 54 = 58%).
We were interested in the fact that some RNAi examples resulted in a non-autonomous apoptosis similar to that seen previously. 17 The RNAi of basal transcription factor Taf6 showed an autonomous activation of Caspase-3 in the blade region and a non-autonomous activation of JNK in the hinge region (Figs. 4A, B), the latter of which was unexpected because Taf6 is known to be necessary for the function of TBP (TATA-Binding Protein), suggesting its ubiquitous requirement for most of the transcription by RNA polymerase II. Accordingly, the RNAi of taf6 is expected to induce a severe autonomous apoptosis in all of the tissues, as is the case for the RNAi of ribosomal protein genes (e.g. RpS14, Fig. 4C). Thus, the non-autonomy in JNK activation in the hinge region in taf6 RNAi suggests a morphogenetic function rather than its common transcriptional function and/or a difference in the sensitivity of decreased transcription leading to apoptosis between the tissues. The difference in responses between the blade and hinge regions was previously described in the apoptosis associated with homeotic transformation by overexpression of spineless. 30 Through these RNAi experiments, we discovered numerous cases of non-autonomous apoptosis cases. Among these, there are particular cases in which an autonomous apoptosis must be induced as a primary response (e.g. RpL17, shotgun) while an additional non-autonomous apoptosis may be further induced as a secondary response, which is probably caused by a juxtaposition of a normal area and a wide apoptotic area, as proposed previously. 32 Therefore, we tested a model case in which the proapoptotic gene reaper (rpr) is temporarily induced within the dorsal compartment by combination with a temperature sensitive-GAL80 33 (Fig. 5). As a result, at around 24 hours after rpr expression, non-autonomous apoptosis could be observed in the ventral compartment (Fig. 5B), although no similar non-autonomy was observed earlier or later (Figs. 5A and C). Thus, a wide area of autonomous apoptosis induction can cause a secondary non-autonomous apoptosis. This phenomenon seems to be a repair mechanism for fi tting the adjacent tissue size. 32

Discussion
We created a database to show the immunofl uorescent images for JNK and Caspase-3 activities in each RNAi experiment. In addition to our purpose of surveying genes affecting cell survival, the database may also be useful for searching for genes regulating tissue growth and patterning. For example, the RNAi of mad displays an apparent shrinkage of the compartment size without showing severe apoptosis except at the DV boundary (Fig. 1E). This phenotype strongly suggests the involvement of this gene in tissue growth and/or patterning. In contrast, the wing disc in which the dorsal compartment cells overexpress rpr showed a wide and severe apoptosis so that most of the dorsal cells disappeared (Fig. 5D). This case represents a typical phenotype, showing that the gene plays a role exclusively for apoptosis. These fi ndings provide insight into the roles of genes for regulating various developmental processes.
As is the case for the above-mentioned possibility of weak activation of JNK and/or Caspase-3, the OTE should also be considered for all phenotypes. 34 As an effective initial examination, the RNAi phenotype must be ameliorated by addition of the wild type transgene that is targeted by RNAi. Furthermore, two ways to distinguish the real RNAi effect from OTE have been proposed. 35 One is a test of a dsRNA corresponding to the other part of the same mRNA for displaying the  same phenotype. The other is a test of an artifi cially altered transgene that is not targeted by the dsRNA but that encodes the same amino acid sequence for complete rescue of the phenotype. Of course, classical analysis using the loss-of-function mutant may be another reliable way to judge the involvement of the gene in each phenotype. In either case, further analyses are required to demonstrate each phenotype as a real loss-of-function phenotype of the gene under focus. The database will be updated when we check the phenotype by the above examinations or accumulate the data from other RNAi constructs.

Materials
Various fl y strains harboring a transcribable inverted repeat sequence (IR) driven by UAS (Upstream Activation Sequence) were prepared in the National Institute of Genetics (NIG) in Japan, as previously described. 36 Briefl y, a cDNA fragment with nucleotide position 1-500 of the coding sequence was obtained by PCR and was inserted as an IR in a head-to-head manner into a modifi ed Bluescript vector, pSC1. Then IR-fragments were excised by NotI and were subcloned into pUAST, a germline transformation vector containing UAS. 37 In the earlier stages of this research, we did not select the IR strains but randomly employed them according to the order in which NIG collected them. In the later stages, we preferentially focused on 302 genes that were predicted to encode secretory proteins. 38 Each UAS-IR fl y strain was crossed with another strain carrying ap-GAL4, UAS-GFP and puc-lacZ. The offspring larvae possessing these four transgenes were reared at 25 °C on a standard diet and then dissected at the late third instar larval stage for immunological staining.

Prediction of secreted proteins in the Drosophila genome
Putative secreted proteins were searched based on the presence of hydrophobic residues in the N-terminal amino acids. When the average hydrophobicity index in the 25 amino acids between positions 6 and 30 exceeded 0.953, the protein was assumed to be secreted. The predicted protein data set from BDGP release 4.2 was searched using the program "Ahiru", 38 which was written based on the algorithm described in 40 (http://bioinformatics. oxfordjournals.org/cgi/reprint/18/2/298); this yielded a list of 2,184 candidate genes encoding secreted proteins. Among them, 296 genes available in the RNAi strains in NIG were used for the screen.

How to access the immunofl uorescence images
To access the fl uorescent images, go to the middle of the right column of the web page (http://www. shigen.nig.ac.jp/fl y/nigfl y/index.jsp), and click the line of "Browse All RNAi Stocks". In the newly appeared page, you can see all of the IR strains. When you click each Stock ID name which has the "wing disc" icon on the right column, you can see a set of immunofl uorescence images at the bottom of the further next page.
for fl y stock, and to Masayuki Miura and Erina Kuranaga for anti-DIAP1 antibody. We also thank Shigeo Hayashi and Kagayaki Kato for providing us the information on putative secretory proteins before its publication, and Shigeo Hayashi for his critical reading of the manuscript. This work was supported by grants from the Japan Science and Technology Agency and the Kato Memorial Bioscience Foundation.