E2F function in muscle growth is necessary and sufficient for viability in Drosophila

The E2F transcription factor is a key cell cycle regulator. However, the inactivation of the entire E2F family in Drosophila is permissive throughout most of animal development until pupation when lethality occurs. Here we show that E2F function in the adult skeletal muscle is essential for animal viability since providing E2F function in muscles rescues the lethality of the whole-body E2F-deficient animals. Muscle-specific loss of E2F results in a significant reduction in muscle mass and thinner myofibrils. We demonstrate that E2F is dispensable for proliferation of muscle progenitor cells, but is required during late myogenesis to directly control the expression of a set of muscle-specific genes. Interestingly, E2f1 provides a major contribution to the regulation of myogenic function, while E2f2 appears to be less important. These findings identify a key function of E2F in skeletal muscle required for animal viability, and illustrate how the cell cycle regulator is repurposed in post-mitotic cells.

T he E2F/DP heterodimeric transcription factors (hereafter referred as E2F) are the key cell cycle regulators and downstream targets of the Retinoblastoma tumour suppressor protein (pRB). Besides their prominent role in tumorigenesis, pRB and E2F are also critical in normal animal development; however, the phenotypes of the E2F and Rb knockout animals are not fully understood [1][2][3] . The arrangement of the Drosophila Rb pathway is simpler, yet highly analogous to that in mammals 4 . In flies, there are two E2Fs, an activator E2f1 and a repressor E2f2. Both Drosophila E2Fs heterodimerize with a single obligatory partner Dp, and, therefore, the loss of Dp eliminates all E2F activity and fully phenocopies the E2f1 E2f2 double-mutant phenotype 5,6 . While most E2F targets are occupied by both E2f1 and E2f2, the relative contribution of each E2F family member to the overall level of expression of the target gene varies. The cell cycle E2F targets are regulated primarily by E2f1-dependent activation, while E2f2 represses the expression of differentiation genes and developmental genes in cycling cells [6][7][8] .
The streamlined version of the Drosophila E2F/Dp provides a unique opportunity to examine the impact of the complete genetic loss of E2F during development. Unexpectedly, E2f1 E2f2 double-mutants progress normally throughout most of development, without any obvious defects in cell proliferation, differentiation or apoptosis 9,10 . Eventually, E2F-deficient animals die during pupal stages, and no E2f1 E2f2 double-mutant adults could be recovered, suggesting that E2F is absolutely essential for animal viability. However, the studies on E2F-deficient animals thus far have failed to uncover a defect that could account for their lethality, and, therefore, the essential function of E2F in animal development remains unknown.
Here we show that E2F has an essential function in adult skeletal muscle, which is necessary and sufficient for animal viability. Although E2F is fully dispensable for proliferation of muscle progenitor cells, it is required for muscle growth and myofibrillogenesis. We show that E2F directly regulates the expression of late myogenic genes, with E2f1 providing the major contribution compared with E2f2. Overall, these results illustrate how a cell cycle transcriptional regulator is repurposed in postmitotic cells to carry out an essential function during myogenic differentiation.

Results
Muscle-specific loss of Dp is lethal. To identify the tissue in which E2F function is required for animal viability, we induced tissue-specific knockdown of Dp using RNA interference (RNAi) and the UAS/GAL4 system. The UAS-Dp-RNAi was crossed to a panel of GAL4 drivers, and animal viability was scored (Supplementary Table 1). Unexpectedly, the depletion of Dp using the myogenic Mef2-GAL4 driver resulted in fully penetrant pupal lethality that mimicked the lethal phenotype of the wholebody Dp mutant (Fig. 1a,b and Supplementary Table 1). This result was validated using another muscle-specific GAL4 driver, 24B-GAL4, and two independent nonoverlapping UAS-Dp-RNAi lines. Notably, pupal lethality in Mef24Dp-RNAi animals was partially rescued by co-expression of a UAS-Dp transgene (Fig. 1a).
In Drosophila, the adult skeletal muscles are formed during pupal development 11 . The myoblasts, also known as adult muscle precursors, proliferate in the adepithelial layer at the dorsal proximal side of the wing imaginal disc that give rise to the adult flight muscles 12 . In late third instar larvae, these proliferating myoblasts can be visualized by the expression of Myocyte Enhancer Factor 2 (Mef2) and Twist bHLH (Twi) transcription factors 13,14 , using Mef2-GAL4-driven GFP (green fluorescent protein) and twi-lacZ markers, respectively (Fig. 1c,d). The efficiency of muscle-specific Dp knockdown was confirmed by staining wing imaginal discs of Mef24Dp-RNAi; twi-lacZ animals with Dp and b-Gal antibodies. Dp was depleted specifically in myoblasts (b-Gal-positive) but not in the adjacent epithelial cells of the wing imaginal disc (b-Gal-negative; Fig. 1d).
Both Mef2-GAL4 and 24B-GAL4 are broadly expressed in all muscle types throughout development. In contrast, 1151-GAL4 is expressed exclusively in adult muscle precursors, developing adult skeletal muscles and tendon cells for the jump muscles, while it is not expressed in larval, cardiac or visceral muscles [15][16][17][18] . Inactivation of Dp using 1151-GAL4 resulted in a threefold reduction in the number of eclosing adults with lethality occurring at the pupal stage ( Fig. 1b and Supplementary  Table 1). Notably, the 11514Dp-RNAi escapers were significantly defective in their ability to fly and jump (Supplementary Table 1), which indicates impaired adult skeletal muscle function and development 19 . Consistently, depletion of Dp with either Mhc.F3-580-GAL4 or Act88F-GAL4, which are expressed in differentiating indirect flight muscles, resulted in similar flight/ jump defects (Supplementary Table 1). Taken together, these results suggest an essential role for E2F in adult skeletal muscles, which is crucial for animal viability.
Muscle-specific expression of Dp rescues Dp-mutant lethality. Results described above raise the possibility that the lethality in Dp mutants is due to the loss of Dp function in the muscle. If this is the case, then providing Dp function in muscles should be sufficient to rescue the whole-body Dp mutant, and permit the completion of development to adulthood. We expressed a UAS-Dp transgene in the Dp null mutant animals (Dp À / À ) using the Mef2-GAL4 driver. Dp À / À animals were generated by crossing a known Dp null allele, Dp a3 (refs 5,10), to a deficiency that removes the entire Dp gene, Df(2R)Exel7124. The musclespecific expression of Dp in Dp À / À ; Mef24Dp animals was confirmed by staining wing discs with the Dp antibody. Dp expression was indeed restricted to adult myoblasts in the adepithelial layer and was completely absent in the imaginal disc (Fig. 2a).
Strikingly, while all Dp null mutants died during pupal development, 27% of Dp À / À ; Mef24Dp animals survived to adulthood (Fig. 2b). The rescued adult flies did not show any gross morphological defects, although they displayed a subtle rough eye phenotype and sterility, the latter likely reflects the known requirement of Dp in oogenesis 20,21 . Another musclespecific driver, 24B-GAL4, provided an even stronger rescue (68% of rescued Dp À / À ; 24B4Dp adults, Fig. 2b). The lethality of the Dp mutant was also rescued when Dp expression was limited to adult myoblasts and adult skeletal muscles using 1151-GAL4 (31% of rescued Dp À / À ;11514Dp adults, Fig. 2b). We confirmed that the rescue was not due to leaky expression of Mef2-GAL4 and 24B-GAL4 drivers in non-related muscle tissues ( Supplementary Fig. 1).
These results provide strong genetic evidence that the lethality in the Dp mutant was not caused by the requirement for E2F/Dp in the entire organism, but rather was due to the requirement of E2F/Dp function in adult skeletal muscle. Therefore, to understand why E2F/Dp is so important in this tissue, we examined its role during adult skeletal muscle development.
Knockdown of Dp reduces the size of adult skeletal muscles. Staining with anti-Dp antibody and the Dp[GFP] reporter line 22 revealed that Dp is broadly expressed in various muscle types (Supplementary Fig. 2A,B). High Dp levels were detected in the nuclei of larval body wall muscles, in cardiac and pericardial nuclei of the larval cardiac tube, in adult muscle precursors and in the indirect flight muscles of adult flies (Supplementary Fig.  2C-F) and pharate pupa (Fig. 3a). Since Dp mutants die at an early pupal stage, we examined Mef24Dp-RNAi animals at the time point immediately preceding the lethal stage. The efficiency of Dp depletion was confirmed by measuring the expression of a known E2F target gene, Arp53D, which is de-repressed in response to E2F inactivation 7 (Fig. 3b). The indirect flight muscles constitute more than 70% of the entire skeletal muscle mass and comprises two groups of muscles: dorsal longitudinal muscles (DLMs) and dorsal ventral muscles 23 . The architecture of DLMs was analysed by staining the transverse paraffin-embedded tissue sections with haematoxylin and eosin dyes. Most Mef24Dp-RNAi pharates displayed six DLMs per hemithoraces (Fig. 3c), indicating that the main steps of muscle development are unaffected. However, each DLM cross-section area was significantly thinner compared with the control (white asterisks, Fig. 3c,d) suggesting that the overall muscle mass is reduced.
The structure of DLMs closely parallels that of vertebrate skeletal muscles. In DLMs, myofibrils are organized in clusters, with each cluster being surrounded by a membrane (dashed line in Fig. 3d). The area of each myofibril cluster is determined by the number of myofibrils and their packing density within the cluster.
Thus, the small DLM size could be due to a reduction in the number of clusters or the cluster size. Frozen transverse tissue sections of thoraces were co-stained with Phalloidin and b-PS integrin antibody to visualize the myofibrils and the surrounding membrane, respectively (Fig. 3d). The overall structure and organization of DLM3 and DLM4 was examined. The number of clusters per DLM, the area of each cluster and the number of myofibrils per cluster were reduced in Mef24Dp-RNAi compared with control. However, the packing density remained unaffected (Fig. 3d). Consistently, the number of nuclei per DLM was reduced. Importantly, the reduced muscle size was largely rescued by the reexpression of the UAS-Dp transgene in the Dp-depleted muscle of adult animals (Fig. 3e). As expected, Arp53D became fully repressed, confirming that Dp function is restored in Mef24 Dp-RNAi;UAS-Dp (Fig. 3f). Thus, we concluded that Dp-deficient muscles are smaller because of a reduction in the number of myofibrils per DLM without changes in muscle architecture.
To determine whether Dp depletion affects the size of other types of muscles in Drosophila, body wall preparations of third instar larvae were stained with phalloidin, and the size of the ventral longitudinal 3 (VL3) muscle was measured. As in the flight muscles, the Dp-depleted larval muscles were thinner (Fig. 3g). Importantly, the size of VL3 was reduced in Dp À / À  mutant to a similar extent to Mef24Dp-RNAi larvae (Fig. 3h). Thus, we concluded that the reduced muscle size was specifically the result of Dp depletion in Mef24Dp-RNAi larvae because it was also observed in mutant animals. However, the reduction in larval muscle size was relatively subtle, thus suggesting that E2F plays a minor role in larval skeletal muscle development in contrast to its requirement in adult skeletal muscles.
Dp depletion does not alter myoblast proliferation or fusion. Drosophila muscle development occurs in a highly ordered manner that is tightly coordinated with the myogenic transcription programme. Adult myoblasts proliferate during larval stages within the adepithelial layer of the wing imaginal disc. In early pupae, myoblasts migrate and fuse into multinucleated myotubes. Fusion is completed by 36 h APF (after pupa formation), and, during the remainder of pupal development, DLMs increase in size because of the addition of myofibrils, and the assembly of the sarcomeres [23][24][25] . This late stage of myogenesis is characterized by a high level of expression of genes encoding structural and contractile proteins, which supports the increase in muscle mass. Thus, the final muscle size is determined by (1) the number of myoblasts and their fusion competence, (2) the proper assembly of myofibrils and (3) the ability of the muscle to grow. To understand why the loss of Dp results in reduced muscle size, we examined the effect of Dp depletion on each step of myogenesis.

Dp -/-; Mef2>Dp
Dp -/-; 24B>Dp Dp -/-; 1151>Dp  We began by analysing proliferation in Dp-depleted adult myoblasts. The adult myoblasts within the adepithelial layer of Mef24Dp-RNAi and control discs were blindly quantified using twi-lacZ reporter and Mef2 antibody as myoblast-specific markers (Fig. 4a). The number of Dp-depleted myoblasts and the corresponding area were indistinguishable from that of control  Fig. 4a). Consistently, the expression of genes encoding myogenic regulators, twi, Holes-in-muscles (Him) and Mef2, was not significantly different in Dp-depleted adult myoblasts compared with control ( Fig. 4b). Dp depletion was efficient because the E2F target Arp53D was strongly de-repressed ( Fig. 4b). Thus, the small size of Dp-deficient adult muscles was not because of a reduction in the pool of adult myoblasts. Next, developing myotubes were visualized with the vestigial adult muscle enhancer (vg AME -lacZ) reporter 26 at 21 h APF when wild-type migrating myoblasts are undergoing fusion. Dp- depleted myotubes properly activated vg AME -lacZ expression (white asterisks, Fig. 4c) and, in several cases, expression of vg AME -lacZ was detected in individual myoblasts immediately before fusion similarly to that observed in control (arrowhead, Fig. 4c). Counting the number of Mef2-positive nuclei in the nascent myotubes and surrounding myoblasts revealed a subtle, yet not statistically significant, reduction in Mef24Dp-RNAi (Fig. 4c). These results suggest that at 21 h APF both migration and fusion proceeded normally. This conclusion was corroborated using quantitative reverse transcriptase-PCR (qRT-PCR). The expression of genes encoding fusion proteins (blow, sing, mbc), myogenic regulators (Him, twi, lame duck (lmd), Mef2, erect wing (ewg), vg), muscle progenitor specification genes (heartless (htl), stumps) and the structural gene Actin 88F (Act88F) was normal at 21 h APF in Dp-depleted muscles (Fig. 4d).
To determine whether all migrating Dp-depleted myoblasts properly completed fusion to developing DLMs, we examined DLMs at 40 h APF, when fusion is complete. The multinucleated DLMs were stained with Phalloidin to reveal the overall architecture, and with Mef2 antibody to mark individual nuclei. The number of nuclei in Dp-depleted DLMs was slightly but not significantly reduced (Fig. 4e). Thus, knockdown of Dp does not affect the overall pool of myoblasts or their ability to migrate and fuse properly.
Growth is defective in Dp-depleted adult skeletal muscles. The development of the indirect flight muscles culminates in myofibrillogenesis, during which actin and myosin filaments are assembled to form myofibrils, which in turn contain the highly ordered repetitive contractile units known as sarcomeres. Sarcomeres are composed of parallel thick and thin filaments. Thin filaments from neighbouring sarcomeres are crosslinked in the Z-disc, and thick filaments are crosslinked in the M-line. The indirect flight muscles express specific proteins, such as Flightin (Fln), and muscle-specific isoforms of structural proteins, such as Act88F and Troponin C isoform 4 (refs 27-29). Myofibrillogenesis is accompanied by a dramatic increase in muscle size. The robust expression of muscle structural genes is needed to support muscle growth during the remainder of pupal development.
The DLMs were dissected from pharate adults (96 h APF), and longitudinal sections were stained with phalloidin to visualize myofibril organization. Dp-depleted muscle showed a mild disorganization in the myofibril structure. The assembly of sarcomere was assessed using a GFP-tagged Mhc isoform, weeP26, that localizes to two foci flanking the M-line at the core of the sarcomere 30 . The localization of weeP26 in Dp-depleted muscles was more diffuse and the intensity of the signal was reduced compared with control (Fig. 5a). The Z-line constitutes a critical anchoring point for thin filaments and can be visualized using the Sls-GFP reporter 30 . Notably, the pattern of Z-lines in Dp-depleted DLMs was less definitive than in control (Fig. 5b). In addition, myofibril width was significantly reduced in Dp-depleted DLMs, as shown by the plot profile of phalloidin intensity over distance (Fig. 5c), suggesting an impaired recruitment of thick filaments. Interestingly, nuclei were larger in Dp-depleted muscles than in control (Fig. 5d). Thus, Dp-depleted muscles formed sarcomeres properly; however, the sarcomere structure was less pronounced, which might result from stoichiometric imbalance among sarcomere components 31 .
Since myofibrillogenesis and the accompanying muscle growth depend on robust expression of structural genes, we examined the level of myogenic transcripts in flight muscles of pharate using qRT-PCR. The expression of genes encoding structural and contractile proteins, such as Myosin heavy chain (Mhc), fln, Tropomyosin 1 (Tm1), Tropomyosin 2 (Tm2), Myosin light chain 2 (Mlc2), sarcomere length short (sals) and Act88F, and myogenic regulators, such as held out wings (how), Limpet (Lmpt), Myocyte enhancer factor 2 (Mef2) and spalt major (salm), were markedly reduced (Fig. 5e). We confirmed the decreased expression of these genes with an independent UAS-Dp-RNAi transgene (Mef24Dp-RNAi-TRiP, Supplementary Fig. 3A). Importantly, the defect in gene expression was fully rescued by muscle-specific re-expression of the UAS-Dp transgene in the Dp À / À mutant background (Dp À / À ; Mef24Dp, Fig. 5f, grey box) and was partially rescued in the Dp-RNAi background (Mef24Dp-RNAi;UAS-Dp, Supplementary Fig. 4). In each experiment, we monitored the expression of Arp53D to ensure the efficient depletion of Dp in RNAi experiments and the rescue following Dp re-expression (Supplementary Fig. 3B and Figs 3f and 5f). Not every structural gene was reduced following Dp knockdown, such as upheld (or Troponin T) and a-actinin (Actn; Fig. 5e). Similarly, the myogenic transcription factor Chorion factor 2 (Cf2), which represses the expression of Act88F in indirect flight muscles 32 , was unaffected (Fig. 5e).
These results suggest that the loss of Dp does not alter muscle development, but rather affects the growth of muscles, which is likely to be due to the reduced expression of a subset of late myogenic genes.
Mitochondria are abnormal in Dp-depleted muscles. During DLM growth, mitochondria undergo extensive fusion, increase in size 33 and become large and tubular as revealed by mitogFP Confocal images of late third instar wing discs stained with anti-b-GAL (twi-lacZ), anti-Mef2 antibody and DAPI. Z-stack-projected images of adepithelial area were magnified. Quantification of total number of nuclei in Twi-positive area using DAPI staining, and Twi-positive area relative to total wing area, mean ± s.e.m., n ¼ 14, 10 wing discs per genotype, t-test with Welch's correction, P40.01. Two independent experiments. (b) The expression of myoblastspecific genes in wing discs is not affected. Expression of twi, Holes-in-muscles (Him) and Mef2 genes quantified by RT-qPCR. Arp53D was de-repressed in Dp-depleted wing discs. Mean ± s.e.m.,  (Fig. 6a). Dp-depleted muscles failed to properly remodel mitochondria, which remained thin and elongated, resembling mitochondria at earlier time points of muscle development 33 .
E2F/Dp was shown to regulate mitochondrial function by directly controlling the expression of mitochondria-associated genes in the eye 34 . In Dp-depleted muscle, the expression of mitochondria-associated E2F target genes, COX5A, ND-B17.2 and Idh (ref. 34), were downregulated (Fig. 6b). Thus, the role of E2F in the regulation of mitochondria is conserved in Drosophila skeletal muscle.
We analysed mitochondrial morphology in rare Mef24 Dp-RNAi;UAS-Dp animals that were rescued to adulthood by muscle-specific expression of Dp (Fig. 1a) and no longer displayed the muscle growth defect (Fig. 3e). Interestingly, the mitochondrial network in Mef24Dp-RNAi;UAS-Dp adults remained abnormal (Fig. 6c). Consistent with persistent mitochondrial defects, the expression of mitochondria-associated genes also remained low (Fig. 6d). These data suggest that both the muscle growth defect and the lethality of Mef24Dp-RNAi animals can be rescued, at least partially, without fully suppressing the mitochondrial defect. Notably, in the Dp À / À mutants that were rescued to adulthood by expression of UAS-Dp with Mef2-GAL4, the expression of mitochondria-associated genes was fully restored (Fig. 6e). The rescue of mitochondria network correlated with animal viability rescue, as B30% of Dp À / À ; Mef24Dp reached the adult stage versus 10% of Mef24Dp-RNAi; Dp (Figs 1a and 2b). Thus, the rescue of muscle growth defects and mitochondrial function is needed in Dp-deficient animals to survive to adulthood.
E2F directly regulates the expression of late myogenic genes.
To investigate the role of the E2F/Rb pathway in regulating the expression of late myogenic genes, we selected genes that were downregulated in Dp-deficient muscles and contained putative E2F-binding sites in the proximity of the transcription start site. Chromatin was isolated from pharate pupa and was subjected to chromatin immunoprecipitation (ChIP)-qPCR. Significantly, upstream regions of myogenic genes, such as how, fln, Lmpt, sals, Tm1 and Mef2, were specifically enriched with E2f1, E2f2, Dp and Rbf antibodies in comparison with a nonspecific antibody (IgG, Fig. 7a). No enrichment was detected for the negative site, which did not contain predicted E2F-binding sites. Some myogenic genes, such as Tm2, did not show any enrichment upstream the transcription start site (Fig. 7a), suggesting that either their expression is indirectly regulated by E2F or E2F binds to a different region from the one tested using ChIP-qPCR. Next, we performed ChIP-qPCR using chromatin isolated from Mef24Dp-RNAi pharate adults, in which Dp was depleted exclusively in muscles, while the remainder of the animals expressed normal levels of endogenous Dp. The muscle-specific depletion of Dp resulted in significantly reduced enrichment of E2f1, E2f2, Dp and Rbf on the myogenic genes (Fig. 7a). This suggests that E2F/Dp is present as a complex, and that the recruitment of Rbf is E2F-dependent.
Since the expression of the contractile and structural genes is restricted to a narrow window during mid to late pupal development, we examined the occupancy of E2F/Dp/Rbf on these genes during different time points in development. Chromatin was isolated from third instar larva, early pupa (24 h APF) and pharate (96 h APF) animals. Notably, E2f1, E2f2, Dp and Rbf were highly enriched at these promoters even before the activation of fln, Lmpt and Tm1 gene expression (Fig. 7b). These results suggest that, although E2F/Dp is required for the expression of several myogenic genes, it is insufficient to activate their expression by itself. Therefore, other myogenic transcription factors are operating at this point to induce their expression.
E2f1 provides a major contribution to skeletal muscle growth. During cell proliferation, E2f1 and E2f2 act antagonistically to each other, with E2f1 being an activator and E2f2 acting as a repressor 7,9 . The loss of E2f1 blocks cell proliferation and results in larval lethality 35,36 ; however, this defect is rescued by the E2f2 mutation as E2f1 E2f2 double mutants proliferate normally and survive until the pupal stage 9 . To determine the relative contribution of E2f1 and E2f2 in muscle, each E2F family member was inactivated singularly, and the impact on myogenesis was  investigated. Since E2f1 mutants are lethal at larval stages, we used UAS-E2f1-RNAi 37 . As expected, depletion of E2f1 in proliferating myoblasts with Mef2-GAL4 significantly impaired proliferation and resulted in early pupal lethality before muscle differentiation. Therefore, E2f1 was depleted specifically in the post-fusion indirect flight muscles with a late myogenic Act88F-GAL4 driver 17 . The efficiency of E2f1 depletion was confirmed using qRT-PCR (Fig. 8a).  (Fig. 8a). The area of DLM4 cross-sections, as well as DLM3, was significantly reduced in Act88F4E2f1-RNAi (Fig. 8a), which was highly reminiscent of the muscle defect in Mef24Dp-RNAi (Fig. 3d). Notably, the analysis of DLM sagittal sections revealed an altered myofibril assembly (Fig. 8b), like in Dp-depleted muscles (Fig. 5a-c). Importantly, Act88F4E2f1-RNAi adult flies were severely impaired in their ability to fly, implying dysfunctional flight muscles (Fig. 8c). In this test, adult flies with fully functional flight muscles were able to fly to the highest region of the 1,000-ml cylinder when flipped into the column, while flightless flies usually fall to the bottom of the cylinder 19 . The relative distribution of flies over the height of the cylinder and the median distribution are shown in Fig. 8c. In contrast, the loss of E2f2 had no effect on muscle size, myofibril width or the ability to fly (Fig. 8d-f).

E2f2 +/-
To understand why the inactivation of E2f1 exerts a stronger effect on muscle morphology and function than the loss of E2f2, we evaluated the expression of myogenic genes in the flight muscles of each genotype. In agreement with the muscle defects observed above, the expression of myogenic genes was strongly deregulated in the absence of E2f1 (Fig. 8g). Depletion of E2f1 resulted in the downregulation of Tm2, Act88F and Mlc2 genes. Interestingly, we observed higher levels of the myogenic regulators Mef2, how and salm that regulate the expression of muscle genes. This may explain why some structural genes were expressed at normal or higher levels in E2f1 depletion. Unlike E2f1, the loss of E2f2 exerted a much weaker effect on the myogenic transcriptional programme (Fig. 8h). Although some of the muscle-specific genes were slightly downregulated in the flight muscles of the E2f2 mutant, none of the changes were statistically significant.
Collectively, these results suggest that E2f1 provides greater contribution to adult skeletal muscle development than E2f2.

Discussion
In this report, we addressed a long-standing question in the E2F field: why do E2F-deficient animals die? Our goal was to understand what is the essential function of E2F in animal development. By expressing a Dp transgene in a spatially restricted pattern in Dp null mutant animals, we identified adult skeletal muscles as a critical tissue in which E2F function is necessary and sufficient for animal viability. Surprisingly, E2F requirement was restricted to late myogenesis when muscles undergo extensive growth. E2F occupies the promoters of musclespecific genes, and their expression is markedly reduced in the absence of E2F, suggesting that E2F is directly involved in the control of myofibrillogenesis (Fig. 9). Thus, the role of E2F in regulating skeletal muscle growth is essential for animal viability.
Previous genome-wide studies in Drosophila revealed two broad categories of E2F target genes 6-8 . One group is regulated primarily by E2f1 activation and consists of cell cycle genes. Another group comprises genes related to gametogenesis, development and differentiation. These targets are regulated exclusively by E2f2-dependent repression, while E2f1 is never present on these targets. Therefore, the role of Drosophila E2F in differentiation was thought to prevent the inappropriate expression of differentiation programmes in cycling cells. Here we show that E2f1 has an important function in late muscle differentiation, where it controls the expression of muscle-specific genes.
Even though both E2f1 and E2f2 can be detected using ChIP on myogenic genes at different developmental stages, E2f1 is more important than E2f2 during muscle growth. This conclusion is supported by the strong impact of E2f1 knockdown on the size of the indirect flight muscles, and the poor performance of E2f1-depleted muscles in the flight ability test. In contrast, the E2f2 mutation did not affect muscle size and function nor myogenic gene expression. These observations point to the functional importance of E2f1 during myogenic differentiation. However, we noted that Act88F4E2f1-RNAi results in a less severe phenotype than the knockdown of Dp with Mef2-GAL4. The differences in the severity of the phenotypes could be due to the knockdown efficiency of E2f1 and Dp, which is always a concern with any RNAi experiment, and the different strength, tissue specificity and timing of expression of two GAL4 drivers. Alternatively, this difference may reflect a partial redundancy between E2f1 and E2f2 since Dp depletion functionally inactivates both E2fs. In agreement, the knockdown of E2f1 exerted less severe effect on target gene expression in comparison with Dp depletion; thus, E2f2 could partially compensate for the loss of E2f1. Such an interpretation also parallels the relationship between the mammalian activator E2f3a and the repressor E2f4 during neuronal differentiation. In neural precursor cells (NPCs), both E2f3a and E2f4 bind and regulate the expression of differentiation genes; however, only E2f3 and not E2f4-deficient NPCs exhibit differentiation defects 38,39 .
The knockdown of Dp in the myogenic lineage revealed that the major defect associated with the loss of E2F function is the dramatic reduction in the size of the adult skeletal muscles. The most obvious explanation is that this is due to the reduced proliferation of Dp-deficient myoblasts, which would decrease  Although E2F/Dp complexes occupy promoters of myogenic genes throughout development, they are needed only within a narrow temporal window during muscle growth to directly control the expression of several muscle genes. This novel role for E2F in late muscle differentiation is necessary and sufficient for adult fly viability.
the pool of myoblasts available for the formation of myotubes. A reduction in the number of myoblasts was shown to result in smaller muscles 24,40,41 . However, Dp-depleted myoblasts proliferated normally within the adepithelial layer in the wing imaginal disc, and there was no clear reduction in the number of migrating Dp-depleted myoblasts during fusion or in the number of nuclei in myotubes after the completion of fusion. The apparent lack of proliferation defect is in agreement with the normal packing density of myofibrils in Dp-depleted muscles, since reducing the number of myoblasts would result in loosely packed myofibrils 24 . In addition, altered myoblast proliferation was shown to reduce the total number of DLMs 24,40,41 , which was not the case in Dp-depleted muscles. Despite the lack of proliferation defects, we noted a small decrease in the number of nuclei in fully differentiated Dp-depleted fibres at 96 h APF, indicating that some nuclei are eliminated during muscle growth. However, the reduction in the number of nuclei does not precede the muscle growth defect, and therefore it is likely to be a secondary event. Thus, the knockdown of Dp does not interfere with the normal progression of development, and E2F function is dispensable for cell proliferation in flight muscle development.
During late myogenesis, muscle volume increases because of the continuous addition of filaments to myofibrils and accompanied muscle growth. These processes rely on high levels of expression of muscle-specific structural proteins. How the expression of late myogenic genes is maintained in adult muscles is not completely understood. The Cf2 transcription factor was previously linked to the control of structural gene expression; however, it acts as a repressor in adult flight muscles 32 . Other factors that may operate at this point to maintain muscle gene expression have not yet been identified, but might include Salm, Exd, Hth and the alternative splicing regulator Arrest, which are all involved in muscle fibre specification [42][43][44][45] . One implication of our work is that E2F contributes to the high level of expression of late muscle genes, and this contribution is needed to support myofibrillogenesis and muscle growth (Fig. 9). Among myogenic E2F targets are the muscle structural genes, including Act88F, Mhc and fln. Intriguingly, myofibrils were shown to be thinner in Act88F and Mhc heterozygotes, suggesting that even small changes in the levels of structural proteins, as found in Dpdepleted muscles, alter the stoichiometry of contractile proteins and could affect myofibril assembly 46,47 . Consistently, the sarcomere structure was less pronounced and the myofibril width was reduced in Dp-depleted muscles, indicating the presence of subtle defects in myofibril assembly. However, E2F is detected only on a subset of promoters of downregulated myogenic genes, suggesting that some of the transcriptional changes observed in Mef24Dp-RNAi could be indirect. This could be the consequence of the altered expression of the key myogenic regulators, such as Mef2, how and Lmpt, which are misexpressed in Dp-depleted muscle and are direct targets of E2F.
The size of the indirect flight muscles also depends on mitochondrial fusion 24 . E2F/Dp was previously shown to regulate the expression of mitochondria-associated genes in Dp-mutant animals. Therefore, Dp-dependent mitochondrial morphology defects could also account for the muscle growth phenotype in Dp-deficient adult skeletal muscles.
An unexpected finding of our work is that the lethality of E2Fdeficient animals is largely due to the requirement of E2F in adult skeletal muscles. Thus, during Drosophila development the essential function of E2F is not to control cell proliferation but rather to regulate terminal differentiation by engaging a myogenic transcriptional programme. Intriguingly, mammalian E2F factors were shown to regulate differentiation of various cell types by promoting tissue-specific transcriptional programmes. During myogenic differentiation, E2f3 is recruited to the promoters of muscle-specific genes and plays a critical role in their regulation 48 . In NPCs, E2f3 and E2f4 govern cell fate decisions and directly regulate the expression of differentiation genes 38 . Thus, the role of E2F in the regulation of tissue-specific transcriptional programmes is highly conserved, and the relevance of our findings extends beyond the fly model.
Fly viability experiment. Fly viability was assessed by collecting and transferring third instar larvae to fresh vials at 25°C. The total number of larva, pupa, as well as pharate pupa, and adult flies able to eclose out of the pupal case was scored over time. The pupal developmental stages were assessed by following markers of metamorphosis as described in ref. 60. Images of 2-to 5-day-old adult flies were taken with the microscope Zeiss Discovery V8, Â 0.63. In the case of the rescue experiments, at least 65 flies per genotype were scored in three independent experiments, except for the rescue using 1151-GAL4 driver, which was performed twice. The genotypes were confirmed using PCR and sequencing in the rescue experiments using Mef2-GAL4 and 1151-GAL4 drivers in the Dp-mutant background. The viability experiment using different GAL4 drivers and Dp-RNAi was scored at least twice if Dp knockdown induced lethality.
Flight test. The test of flight ability was adapted from ref. 19. Adult females were collected on eclosion and recovered at 25°C for 4-6 days. A 1,000-ml cylinder was coated with mineral oil. Flies were flipped to the top of the column. Flies able to fly land on the top section of the column, while flightless flies fall to the bottom of the cylinder. The landing spot along the height of the cylinder was scored for each fly, and the frequency of landing spots per height of the cylinder was plotted. Each experiment was carried out three times. At least 20 flies per genotype were tested in each replicate.
Histology. Pharate animals were removed from their pupal case at 96 h APF and were fixed with 4% formaldehyde at 4°C overnight. Thoraces were embedded in paraffin, and sections of 8-mm thickness were cut using a microtome 61 . Haemotoxylin and eosin staining was performed using standard procedures. Images were taken using the bright-field microscope Zeiss ApoTome. At least 16 flies per genotype were scored.
Immunofluorescence and confocal microscopy. Tissues were dissected and immediately fixed in 4% formaldehyde in PBS for 30  Triton X-100 in PBS twice for 10 min each and blocked in 2% bovine serum albumin (BSA) with 0.3% Triton X-100 in PBS for 60 min. Samples were incubated with antibodies overnight at 4°C in 2% BSA and 0.1% Triton X-100 in PBS. After washing four times for 10 min each in 0.1% Triton X-100 (in PBS), samples were incubated with appropriate Cy3-or Cy5-conjugated anti-mouse and anti-rabbit secondary antibodies (Jackson Immunoresearch Laboratories, used at 1:300) for 90 min in 10% normal goat serum and 0.1% Triton X-100 in PBS. After washing with 0.1% Triton X-100 (in PBS), tissues were stored in glycerol with antifade, and then mounted on glass slides. All steps were performed at room temperature, unless otherwise stated. The DLMs of pupa staged at 21 h APF were dissected as in ref. 62, fixed for 20-30 min, and washed in PBS with 0.1% Triton X-100. Thoraces of 2-to 5-day-old adult or pharate pupa staged at B96 h APF were fixed for 30 min in relaxing buffer. Thoraces were bisected in the sagittal plane, and were then fixed for an additional 15 min as described in ref. 63. In the case of transverse plane sectioning, flies were snap-frozen in liquid nitrogen and cut twice with a razor. The transverse sections were fixed for 1 h (ref. 24). For larval body wall musculature staining, larva was dorsally opened, pinned in a Sylgard dish and fixed for 20 min. A minimum of five to eight animals per genotype was dissected per experiment, except for the experiment that required quantitative analysis, in which case the staining was carried out two or three times.
Fluorescent images were acquired with the laser-scanning confocal microscope (Zeiss LSM Observer.Z1) using Â 20/0.8, Â 40/1.2 and Â 100/1.45 objectives. Images were processed using Photoshop (Adobe Systems). All images are confocal single-plane images and are oriented anterior to the left, unless otherwise stated. Only representative images are shown.
Quantitative and statistical analysis. The total number of myoblasts in the adepithelial cells of the wing disc in late third instar larvae (120 h after egg-laying) was determined by marking the adult myoblast with twi-lacZ and Mef2 antibodies. Images were taken with confocal microscopy in a z-stack to cover all nuclei in the adepithelial layer, and then were projected into a single plane. The area covered by the adult myoblasts, as well as the disc area, was quantified using the ImageJ software. The plugin Image-based Tool for Counting Nuclei was used to determine the myonuclear number in the myoblast-positive area of the adepithelial layer. A minimum of 10 wing discs per genotype was scored.
The density and size of nuclei at 21 and 40 h APF were determined by staining nuclei of both myoblasts and developing DLMs using Mef2 antibody. Both the number and area of nuclei were measured using the analyse particle tool of the ImageJ software. The area of the selected region of interest was measured to calculate nuclear density. Two or three independent experiments were conducted per time point. A minimum of eight animals per genotype was scored.
The number of myofibrils was counted as the number of peaks of grey value (that is, Phalloidin intensity) over a distance of 20 or 40 mm using the plot profile tool from the ImageJ software. Three independent lines perpendicular to myofibril orientation were drawn over a set of myofibril clusters to count the number of peaks (that is, myofibrils). An average was calculated for each image (that is, hemithorax). Myofibril width was calculated as the distance divided by the total number of myofibrils within the selected region. At least six hemithoraces per genotype was quantified.
ImageJ was also used to quantify DLM 3 and DLM 4 cross-section area, as well as the myofibril cluster area in the transverse sections of indirect flight muscles, and the total area of the body wall muscle VL3. The numbers of muscle fibres, myofibrils and nuclei were manually counted per DLM cross-section in pharate animals. At least two independent experiments were conducted. A minimum of eight flies per genotype was quantified.
Images of the DLM sagittal section were blindly scored as either tubular or thin/ fragmented regarding mitochondrial shape. At least 15 hemithoraces stained with mitoGFP were quantified per genotype. Two independent experiments were conducted.
Graphs were generated with the GraphPad Prism version 6.0 (Graphpad Software). The group means were analysed for overall statistical significance using the unpaired Student's t-test with Welch's correction, two-way analysis of variance and non-parametric analysis (Mann-Whitney and Kruskal-Wallis).
RT-qPCR and ChIP-qPCR. Tissue was dissected as described above from at least three to five different animals. Total RNA was isolated with TRIzol (Invitrogen). Reverse transcription to measure standard mRNAs was performed using the SensiFAST cDNA Synthesis Kit (Bioline) according to the manufacturer's specifications. qPCR was performed with the SensiFast SYBR No-ROX Mix (Bioline) on a LightCycler 480 (Roche). The reference genes RpL32, RpL30 and b-tubulin were validated as stable control genes 65 . Normalization was calculated using the geometric mean of these reference genes. Primer sequences are in Supplementary  Table 2. At least three independent biological samples were collected for each genotype and developmental stage. Each sample was measured twice. Raw data for ChIP-qPCR presented in Fig. 7a are in Supplementary Tables 3 and 4, and raw data included in Fig. 7b are in Supplementary Table 5.
The degenerate E2F-binding sites NWTSSCSS 7 and the Drosophila E2F motif TTGGCGC 6 were identified from searches carried out with the Regulatory Sequence Analysis Tools 66 . The evolutionary conservation between D. melanogaster and D. erecta was considered to predict E2F-binding sites.
At least 15 animals staged at third instar larvae, and pupae atB24 and 96 h APF were dissected out of their pupal cases and homogenized. Chromatin was extracted using homogenizer with 60 mM KCl, 15 mM NaCl, 4 mM MgCl 2 , 15 mM HEPES (pH 7.6), 0.5% Triton X-100, 0.5 mM dithiothreitol (DTT), 10 mM sodium butyrate and protease inhibitor cocktail (Complete, Roche) as in ref. 67. Samples were crosslinked for 15 min at room temperature in 1.8% formaldehyde, and 225 mM Glycine was added. Then, cells were lysed with 15 mM HEPES at pH 7.6, 140 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 0.1% sodium deoxycholate, 1% Triton X-100, 0.5 mM DTT, 0.1% SDS, 0.5% lauroylsarcosine and 10 mM sodium butyrate with protease inhibitor cocktail (Complete, Roche), and were sheared using a Branson 450 Sonifier. Chromatin was immunoprecipitated with rabbit anti-DP (#212 (ref. 7)), mouse anti-Rbf (DX3/DX5, ratio 1:1 (ref. 68)), rabbit anti-E2f1 (#210 (refs 7,69), ratio 1:1) and rabbit anti-E2f2 antibodies (#79 ( 9 )). Rabbit IgG (Sigma) was used as nonspecific antibodies. Complexes were pulled down with Protein G Dynabeads (Invitrogen), washed with lysis buffer four times, washed twice with TE (pH 8), eluted and decrosslinked overnight at 65°C; RNA was degraded with RNase A (Sigma) for 1 h at 37°C and protein was digested with proteinase K for 2 h at 50°C. DNA was purified by phenol-chloroform extraction, followed by overnight ethanol precipitation. The input genomic DNA (before precipitation), as well as immunoprecipitated DNA, was quantified using qPCR as described here. Primer sequences are in Supplementary Table 2. A negative sequence site that does not contain any predicted E2F-binding sites, and the positive target genes, Arp53D, was used as controls (Supplementary Table 2). The protein enrichment was calculated as the percentage of immunoprecipitated DNA relative to input DNA (prior DNA precipitation) for each antibody. Data presented are relative to the negative binding site for each ChIP. Each sample was measured twice.