A novel role for the 3′-5′ exoribonuclease Dis3L2 in controlling cell proliferation and tissue growth

ABSTRACT In a complex organism, cell proliferation and apoptosis need to be precisely controlled in order for tissues to develop correctly. Excessive cell proliferation can lead to diseases such as cancer. We have shown that the exoribonuclease Dis3L2 is required for the correct regulation of proliferation in a natural tissue within the model organism Drosophila melanogaster. Dis3L2 is a member of a highly conserved family of exoribonucleases that degrade RNA in a 3′-5′ direction. We show that knockdown of dis3L2 in the Drosophila wing imaginal discs results in substantial wing overgrowth due to increased cellular proliferation rather than an increase in cell size. Imaginal discs are specified in the embryo before proliferating and differentiating to form the adult structures of the fly. Using RNA-seq we identified a small set of mRNAs that are sensitive to Dis3L2 activity. Of the mRNAs which increase in levels and are therefore potential targets of Dis3L2, we identified 2 that change at the post-transcriptional level but not at the transcriptional level, namely CG2678 (a transcription factor) and pyrexia (a TRP cation channel). We also demonstrate a compensatory effect between Dis3L2 and the 5′-3′ exoribonuclease Pacman demonstrating that these 2 exoribonucleases function to regulate opposing pathways within the developing tissue. This work provides the first description of the molecular and developmental consequences of Dis3L2 inactivation in a non-human animal model. The work is directly relevant to the understanding of human overgrowth syndromes such as Perlman syndrome.


Introduction
Regulation of cell proliferation is of crucial importance to all multicellular organisms. Cells must proliferate throughout development in order for the organism to grow to adult size. Proliferation must also occur to repair damaged areas during the process of wound healing. Control of proliferation is therefore vitally important to allow individual animals and their constituent organs to grow to their correct sizes, as well as to maintain symmetry between the left and right sides of an animal. 1 In an epithelial tissue subjected to damage, cell proliferation can be stimulated by apoptosis to heal the wound, demonstrating a balance between apoptosis and proliferation. 2 However, uncontrolled cell proliferation is a hallmark of cancer with many genes involved in growth and proliferation implicated in cancer progression. 3 This paper describes the discovery of a new player in the control of cell proliferation within a tissue, namely the exoribonuclease Dis3L2. Dis3L2 is a 3 0 -5 0 exoribonuclease and is a member of the highly conserved RNaseII/RNB family of 3 0 -5 0 exoribonucleases. 4 Prokaryotes and some eukaryotes such as S. cerevisiae harbour a single member of this family (Dis3) whereas other eukaryotes including S. pombe and Drosophila melanogaster encode 2 enzyme paralogues (Dis3 and Dis3L2). 5 In humans, there is an additional paralogue, Dis3L1. Dis3L2, unlike Dis3 and Dis3L1, does not contain the N-terminal PIN domain 5,6 which confers endoribonucleolytic activity to Dis3, and tethers it to the exosome, a multicomponent protein complex. Recent work has shown that Dis3L2 is primarily cytoplasmic and acts independently of the exosome to degrade both mRNAs and non-coding RNAs. [5][6][7] In humans and S. pombe, this often occurs following the addition of a polyuridine (polyU) tract to the 3 0 end of the transcript, which provides a binding site for Dis3L2 to initiate/reinitiate 3 0 -5 0 decay. 5,8,9 miRNAs such as pre-let-7 have also been shown to be poly-uridylated and then degraded by Dis3L2 in human cells and mouse embryonic stem cells. 10,11 In human cells, Dis3L2 has been shown to degrade miR-27a by target RNA-directed miRNA degradation and to associate with Ago2 in the RNA induced silencing complex (RISC). 7 The clinical importance of Dis3L2 is demonstrated by its association with Perlman syndrome and Wilms' tumor susceptibility. 12,13 Perlman syndrome is a congenital overgrowth condition which is inherited in an autosomal recessive manner. 14 Affected children display foetal gigantism, abnormal enlargement of organs (e.g. kidneys), facial abnormalities, neurodevelopmental delay and high neonatal mortality. Histological examinations reveal nephroblastomatosis, which is an important precursor for Wilms' tumor. Germline mutations in these children are consistent with a loss of function of Dis3L2. 12 In addition, a truncation of the DIS3L2 locus has been associated with a Marfan-like syndrome with skeletal overgrowth. 15 Dis3L2 is also likely to be important in sporadic Wilms' tumor as 30% of these tumors (6/20) show partial or complete DIS3L2 deletion. 12 Understanding the cellular and molecular pathways controlled by Dis3L2 in a well characterized model system such as Drosophila melanogaster is likely to shed light on the molecular basis of these diseases.
In this paper we show, for the first time, a role for Dis3L2 in the proliferation of imaginal disc cells during Drosophila melanogaster development. Imaginal discs are specified in the embryo before proliferating and differentiating during larval and pupal stages respectively to form the adult structures of the fly (e.g., wings, legs). Wing imaginal discs provide an excellent model system for the study of growth homeostasis because the cells being studied are in their natural context of a developing organism, rather than as immortalized tissue culture cells. Here we show that depletion of dis3L2 leads to increased proliferation of cells in the wing imaginal discs, without an increase in cell size, resulting in larger wings in the adult fly. Using RNA-seq, we have evidence that Dis3L2 regulates the expression of a discrete set of transcripts including a mRNA encoding a cation channel. In addition, we show that the increase in proliferation caused by depletion of dis3L2 can be compensated with mutations for the gene encoding the 5 0 -3 0 exoribonuclease Pacman/Xrn1. Therefore, these 2 ribonucleases, which degrade RNA in opposite directions, appear to affect the antagonistic pathways of apoptosis and proliferation. Since Pacman and Dis3L2 are highly conserved throughout evolution, it is likely that they will have similar functions in other multicellular organisms.

Dis3L2 is not required for Drosophila melanogaster viability
We have previously shown that ubiquitous knockdown of Dis3 results in complete lethality at or prior to the L2 larval stage. 16 To test whether Dis3L2 (CG16940) is required for viability, ubiquitous GAL4 drivers were used to knockdown dis3L2 throughout the whole organism. Two separate UAS-dis3L2 RNAi lines, each targeting different sequences within both isoforms of the dis3L2 gene were used for each GAL4 driver. Ubiquitous knockdown using tubulin-GAL4 (tub-GAL4), actin-GAL4 (act-GAL4) or daughterless-GAL4 (da-GAL4) drivers resulted in a strong, 9.7-fold knockdown of dis3L2 (to 10% of its level in controls). This was not associated with any lethality (Fig. S1A-B) and shows that Dis3L2, unlike Dis3, is dispensable for Drosophila viability. These findings are consistent with previous data from S. pombe 5 and human cell lines 6,9 which also show no effect on cell viability following dis3L2 depletion.
Loss of dis3L2 in the wing imaginal discs results in wing overgrowth We used the wing imaginal disc as a specific tissue to explore the role of Dis3L2 in tissue growth and development. The wing imaginal disc is divided into distinct regions with the wing pouch region giving rise to the wing blade (Fig. 1A) and the tip of the wing disc forming part of the thorax. Restricted knockdown was achieved using GAL4 drivers with specific expression within the wing imaginal disc (Fig. 1). Depletion of dis3L2 throughout the wing imaginal disc (UAS-dis3L2 RNAi /C ;69B-GAL4/C) resulted in a 5.2-fold knockdown of dis3L2 mRNA (to 20% of its level in controls) (Fig. 1B). This knockdown resulted in wings that are 123% the size of controls (Fig. 1C). However, the 69B-GAL4 driver also drives in the epidermis, together with the eye and ventral thoracic imaginal discs. 24 To ensure the observed overgrowth was due to an effect within the wing disc a more specific GAL4 driver was used. Use of the nubbin-GAl4 driver (nub-GAL4), which drives expression within the wing pouch of the disc, to knockdown dis3L2 also resulted in wings that were 21% larger than parental controls (Fig. 1D). To ensure the apparent overgrowth of the wings was not simply due to an increase in the size of the knockdown flies, individual flies were weighed and the wing area was normalized to the mass of the fly. Normalization of wing area made no difference to the overgrowth observed ( Fig. 1C-D). For the above experiments, driving the expression of an unrelated shRNA (UAS-EGFP RNAi ) did not result in an increase of wing size compared to controls. Use of a second UAS-dis3L2 RNAi line also gave identical results (Fig. S1C-D). These data therefore provide strong evidence that reduced expression of dis3L2 within the wing imaginal discs results in the overgrowth of the adult wings.
To provide further evidence that dis3L2 knockdown results in wing overgrowth, we used an engrailed-GAL4 (en-GAL4) driver, which drives expression in the posterior compartment of the wing, so that the anterior compartment of the wing provides an internal control for wing size. Depletion of dis3L2 using en-GAL4 results in a significant overgrowth of the posterior region of the wing (Fig. 1E). The anterior area also showed a very slight (2.7%) but statistically significant overgrowth. However, this is most likely a knock-on effect of the 11.3% overgrowth observed in the posterior compartment. To ensure it is not only the posterior compartment that is susceptible to overgrowth we measured the anterior and posterior areas of the nub-GAL4 driven knockdown wings and as expected the anterior and posterior areas showed comparable overgrowth of 24.9% and 18.5%, respectively (Fig. S1E). Therefore, it appears that Dis3L2 functions to control the area of the Drosophila wing.
The effect of Dis3L2 on overgrowth is not consistent across the whole fly To assess if wing disc tissue is particularly sensitive to the loss of dis3L2 we further examined the flies with ubiquitous dis3L2 knockdown (RNAi driven by act-GAL4, tub-GAL4 or da-GAL4). Although the knockdown flies had significantly larger wings than parental controls ( Fig. S2A-C), they were not significantly larger in terms of mass (Fig. S2D). This shows that the knockdown of dis3L2 throughout the organism does not induce overgrowth of the whole fly. Therefore, it appears that there is a level of tissue specificity in the activity of Dis3L2, which would be consistent with the phenotypes of Perlman syndrome specifically affecting overgrowth of specific organs such as the liver and kidney. dis3L2 depletion results in increased cell number rather than increased cell size The overgrowth phenotype observed could occur in 2 ways; either an increase in cell number (hyperplasia) or an increase in cell size (hypertrophy). To determine which of these 2 processes was occurring, we counted the total number of cells within the wing of the control and dis3L2 knockdown flies. Since a single hair protrudes from each cell of the wing, the number of hairs can be counted to give an accurate count of the number of cells within the wing. The average number of hairs within 0.1mm 2 regions (representing both the anterior and posterior domains) was calculated ( Fig. 2A). This, along with the area of the wing was then used to estimate the total number of cells as well as their size. We observed that while the density of cells within the 0.1mm 2 regions remained unchanged ( Fig. 2Bi and Ci), indicating no change in cell size, there were 119% the number of cells in the dis3L2-depleted wings compared to controls ( Fig. 2Bii and Cii). This indicates a hyperplasic response rather than hypertrophy of the wing. Therefore, Dis3L2 seems to play a role in controlling the proliferation of cells within the wing imaginal disc.
The proliferative effect of dis3L2 knockdown occurs early in development To determine the developmental time point at which the proliferative effect of dis3L2 knockdown is observed we used the temperature sensitive GAL80 ts system that allows temperature controlled knockdown of dis3L2. 18 Maintaining the flies at the permissive temperature of 19 C allows GAL80 ts mediated repression of the UAS-GAL4 system, therefore inhibiting any Knockdown of dis3L2 throughout the wing imaginal disc using the 69B-GAL4 driver results in 5.2-fold knockdown (to 20% the level of controls) at the RNA level compared to the parental controls. n 10, error bars represent SEM, ÃÃÃÃ D p < 0.0001. (C) Knockdown of dis3L2 throughout the wing imaginal disc using the 69B-GAL4 driver results in significant overgrowth of the wing (23%) compared to controls when wing area is normalized to fly mass. (D) Knockdown of dis3L2 within the wing pouch of the wing imaginal disc using the nub-GAL4 driver results in significant overgrowth (21%) of the wing compared to controls when wing area is normalized to fly mass. For (C) and (D), the "control" is the mean of the parental genotypes plus a UAS-EGFP RNAi driven by the respective GAL4 driver, none of which were significantly different from each other. n 18, error bars represent 95% confidence limits, ÃÃÃÃ D p < 0.0001. (E) Restricting knockdown of dis3L2 to the posterior region of the wing using the en-GAL4 driver results in specific overgrowth of the posterior compartment (11%). n 20, error bars represent 95% confidence limits, ÃÃÃÃ D p < 0.0001, Ã D p < 0.05. Representative images of a male wing are shown in each case, scale bar D 200 mm.
knockdown. However, moving the flies to 29 C results in inhibition of GAL80 ts and subsequent knockdown of dis3L2. Our previous results 19 have shown the perdurance of the system to be between 24 and 32 hours. For these experiments we used the 69B-GAL4 wing disc driver and progeny cultured at 19 C and 29 C as negative and positive controls respectively.
Knockdown of dis3L2 at the start of the 2 nd instar (L2) and 3 rd instar (L3) larval stages resulted in significant overgrowth of the wings (Fig. 3A). However, knockdown during the mid 3 rd instar larval stage and at the start of the pupal stage resulted in wings of wild type size (Fig. 3A). Together, these data indicate that Dis3L2 activity is required during early larval stages to control cell number. Therefore, Dis3L2 is likely to affect transcripts controlling cell proliferation of the wing imaginal discs during the early larval stages, which subsequently affects the later development of the wing.

Depletion of dis3L2 results in overgrowth of the wing imaginal discs
The above GAL80 ts experiment predicted that Dis3L2 is required during larval development to regulate cell number in the wing imaginal discs. To test this, we dissected wing imaginal discs from late L3 larvae and measured their sizes relative to parental controls. Knockdown of dis3L2 using the 69B-GAL4 driver resulted in discs that were 132.0% the size of their respective controls (Fig. 3B) while use of the nub-GAL4 driver increased the size of the discs to 121.6% the size of controls (Fig. S3). The difference in the level of overgrowth between the 2 drivers is most likely due to their expression domains; nub-GAL4 drives expression within the wing pouch whereas the 69B-GAL4 expresses throughout the entire disc.
To confirm that the overgrowth was due to an increase in cell proliferation we used phosphohistone H3 (PH3) as a marker for cells undergoing mitosis. Late L3 wing discs were stained with anti-PH3, the nuclei undergoing mitosis were counted and the mitotic index was calculated (see Materials and Methods). This indicates that a significantly higher percentage of cells are undergoing mitosis in the dis3L2 knockdown wing discs (120.5%) compared to parental control discs ( Fig. 3C-D). These data therefore show that the loss of dis3L2 results in an increased rate of cell division within the wing imaginal disc, which in turn leads in an increase in the total area of the tissue and larger adults wings.

RNA-sequencing shows that Dis3L2 affects the expression of a small number of genes
To identify the RNAs that are misregulated in dis3L2 knockdown imaginal discs and are likely to cause the above phenotypes, we used genome-wide RNA-sequencing (RNA-seq). We compared wing imaginal discs where dis3L2 was knocked down to 20% of the level in parental controls, using the 69B-GAL4 driver (dis3L2 RNAi /C ;69B-GAL4/C) ( Fig. 1B) with the 2 parental stocks used to generate the knockdown flies ( ;UAS-dis3L2 RNAi ; and ;;69B-GAL4) used as controls. Total RNA was prepared from 2 biological replicates of 60 dissected wing imaginal discs per genotype and then subjected to RNA-seq. RNAseq reads were subsequently analyzed using the cufflinks pipeline (see Materials and Methods). The number of reads obtained for each sample together with the non-default parameters used during the analysis can be found in Fig. S4.
We first compared the overall expression distributions for the 2 parental controls to ensure it was reasonable to group them together as a total 'control' group. Kernel density plots showed no significant difference (Kolmogorov-Smirnov (KS) test p D 0.555) indicating that the overall expression profiles of the 2 controls was very consistent (Fig. 4A). The four parental replicates were therefore grouped to form the "control" group. The kernel density plots show 3 major peaks comprising transcripts expressed at a very low level (far left), transcripts expressed at a moderate level (center) and a small set of transcripts expressed at a very high level (far right). Comparing the expression distributions within the control discs and the dis3L2 knockdown discs showed a general and significant shift in expression of genes normally expressed at low levels to a level of greater expression (KS test p D 1.041e ¡6 ) (Fig. 4B). This is consistent with the known function of Dis3L2 and suggests that the exoribonuclease Dis3L2 normally functions to maintain a set of discrete transcripts at low levels. Fig. 4C shows a scatter plot comparing the expression levels of individual genes in dis3L2 knockdown discs with those of the grouped parental controls. After stringent filtering (see Materials and Methods), a total of 239 genes (2.7% of total detected) were reported as differentially expressed between the 2 groups (red dots). Therefore, Dis3L2 regulates a very small number of transcripts, which is a similar conclusion to our previous work showing that the 5 0 -3 0 exoribonuclease Pacman has specificity for particular mRNAs. 20,21 Consistent with the expression distributions those transcripts determined as misexpressed showed a low level of expression in parental control discs. We also observed in our sequencing data that the consistently misexpressed genes were distributed across all chromosome arms indicating there is neither a technical bias nor a chromosome bias for Dis3L2 targets (Fig. S5).
qRT-PCR verification of the RNA-seq results shows that pyrexia and CG2678 are post-transcriptionally upregulated in dis3L2 knockdown discs Since Dis3L2 is an exoribonuclease that degrades RNAs, we would expect its candidate targets to increase in expression in the dis3L2 knockdown wing imaginal discs to produce the observed phenotypes. A set of transcripts showed a decrease in expression in the RNA-seq data (Fig. S6). However, of the 7 validated by qRT-PCR only one showed a reproducible decrease in expression (dis3L2) with large biological variation suggesting these are more likely to be due to indirect effects (Fig. S7). Therefore, to validate genes sensitive to Dis3L2 activity we selected those that were the most consistently increased in expression >1.5-fold in the knockdown discs (Fig. 5). From this shortlist we selected those that showed the highest fold changes while maintaining consistency between replicates in addition to having an average FPKM of greater than 0.3. These Figure 3. Knockdown of dis3L2 results in increased proliferation within the wing imaginal disc leading to overgrowth of the disc. (A) Using the GAL80 ts system to control the developmental timings of dis3L2 knockdown throughout the wing imaginal disc revealed that wing overgrowth is due to processes occurring during early Drosophila development.
-ve represents flies that were maintained at the permissive 19 C temperature throughout their development. Cve represents flies that were maintained at the active 29 C temperature throughout their development. The positive control (Cve), start of L2 and start of L3 timings show a significant increase in adult wing area ( ÃÃÃÃ D p < 0.0001) compared to parental controls. The negative control (¡ve), mid L3 and start of pupal timings show no significant increase in wing area of knockdown flies. "Control" represents the mean wing area of the 2 parental controls. n 37, error bars represent 95% confidence limits. (B) Knockdown of dis3L2 in the wing imaginal disc using the 69B-GAL4 driver results in significant overgrowth of the disc compared to controls. Dis3L2 KD late L3 wing discs are 132.0% the size of parental control discs. n 21, error bars represent 95% confidence limits, ÃÃÃÃ D p < 0.0001. (C) Dis3L2 KD wing discs have a significantly higher mitotic index (20.5% increase in the percentage of cells undergoing mitosis) than control discs. n 13, Ã D p D 0.0241, error bars represent 95% confidence limits. For (B) and (C) "Control" shows the mean measurements of the 2 parental controls. (D) Representative images of the DeadEasy MitoGlia output following staining of late L3 wing imaginal discs with anti-Phosphohistone H3. Scale bar D 50 mm.
were taken forward for validation by quantitative RT-PCR (qRT-PCR) on new biological replicates. Fig. 6A shows that the changes in expression between dis3L2 knockdown and control discs were similar for the majority of the selected transcripts according to the 2 techniques. For example, cytochrome-c-distal (cyt-c-d) increased in expression by 7.2fold according to RNA-seq and by 7.4-fold by qRT-PCR. Similarly, CG31808 increased 6.8-fold and 6.0 fold by RNA-seq and qRT-PCR respectively. For other transcripts, such as pyrexia (pyx) and CG2678 the change in expression detected by RT-PCR was in agreement, but more modest than by RNA-seq (Table 1). Only one gene, CG3457, did not show a significant change in expression by qRT-PCR. The verification of the RNA-seq by qRT-PCR demonstrates that we can have confidence in the changes presented by our RNA sequencing experiment.
To identify potential direct targets of Dis3L2 we tested whether the expression changes of 4 upregulated mRNAs (by qRT-PCR) were due to transcriptional or post-transcriptional effects. As Dis3L2 functions to degrade mRNAs at the posttranscriptional level we would expect Dis3L2 targets to be upregulated at the mRNA but not the pre-mRNA level. Those genes upregulated at the pre-mRNA level as well as the mRNA level would be expected to be indirect targets, for example due to Dis3L2 regulating an RNA encoding a transcription factor. Transcriptional changes are observed for cyt-c-d and CG31808 as their pre-mRNAs increase in expression 6.6-fold and 4.3fold respectively (compared to 7.4 and 6.0 fold increases for the mature mRNAs); indicating that the increases in expression are likely to be due to indirect effects (Fig. 6B). However, pyrexia and CG2678 show a clear regulation at the post-transcriptional level as the expression of pre-pyx and pre-CG2678 do not change significantly in dis3L2 knockdown discs compared to controls, but increase in expression by 2.4 and 2.1-fold respectively at the mRNA level ( Fig. 6B and Table 1). Therefore, of the 4 mRNAs tested, 2 (pyx and CG2678) are upregulated at the post-transcriptional level, suggesting that they are targets of Dis3L2.

Dis3L2 and Pacman function to regulate opposing pathways of apoptosis and proliferation
In previous work, we showed that mutations in the gene encoding the 5 0 -3 0 exoribonuclease pacman resulted in increased apoptosis. 21 In contrast, the work outlined above shows that knockdown of the 3 0 -5 0 exoribonuclease dis3L2 results in increased proliferation. These results suggested to us that knockdown of dis3L2 may rescue the effects of the pacman mutations. To test this hypothesis, we again utilised the UAS-GAL4 system to knockdown dis3L2 in the wing imaginal discs of hypomorphic (pcm 5 ) and null (pcm 14 ) pacman mutants and observed the phenotypic effects.
The hypomorphic pcm 5 mutation has been shown to have 66% of the activity of the wild-type allele resulting in a small wing phenotype. 21,22 Knockdown of dis3L2 throughout the wing imaginal discs in a pcm 5 mutant background (pcm 5 /Y ; UAS-dis3L2 RNAi /C ; 69B-GAL4/C) results in wings that are not significantly different in size from control wings (Fig. 7A). This shows that the phenotypic effects of the dis3L2 knockdown (i.e. proliferation) compensate for the phenotypic effects of the pcm 5 mutation (i.e., apoptosis). To test this striking effect further, we also knocked down dis3L2 throughout the wing imaginal discs in the pacman null mutant, pcm 14 . The pcm 14 mutation results in small wing imaginal discs leading to a developmental delay and eventual pupal lethality. 21 In confirmation of the above results, dis3L2 knockdown in a pcm 14 mutant background (pcm 14 /Y ; UAS-dis3L2 RNAi /C ; 69B-GAL4/C) results in late L3 wing imaginal discs of normal size, again showing a compensatory effect between Dis3L2 depletion and the pcm 14 mutation (Fig. 7B).
In a previous paper, 21 we showed that the pcm 14 null mutation results in an increase in apoptosis due to the post-transcriptional upregulation of the pro-apoptotic genes hid and reaper (rpr). In Drosophila, the key activators of the intrinsic apoptosis pathway are the Hid, Grim and Reaper proteins. 23 To determine whether Dis3L2 derives its compensatory effects from downregulation of rpr expression we compared the levels of this RNA in the wing imaginal discs of pcm 14 mutants with or without additional dis3L2 knockdown. As shown in Fig. 7C, levels of rpr mRNA are very similar in both cases indicating Dis3L2 does not play a role in the regulation of rpr expression. Confirmation is provided by results showing that the dis3L2 knockdown alone has no significant effect on either rpr pre-mRNA or mature mRNA according to qRT-PCR (Fig. 7C) or RNA-seq (1.07 fold change). Similarly, the knockdown of dis3L2 within the wing imaginal disc has no effect on the expression of hid or grim mRNA (Fig. S8).
To confirm that knockdown of dis3L2 does not affect apoptosis via the Hid, Grim or Reaper pathway, we made use of the Df(3L)H99 deficiency, where these 3 genes (which are located together on the chromosome) are deleted from the genome. Heterozygous Df(3L)H99/C have larger wings than wild-type controls due to reduced apoptosis in the wing discs ( Fig. S9). If dis3L2 knockdown was to inhibit the Hid, Grim and Rpr apoptotic pathway we would expect knockdown of dis3L2 in a heterozygous Df(3L)H99 background to result in wings no different in area to the Df(3L)H99/C controls or when dis3L2 is knocked down in a wild-type background. In contrast, this knockdown (; UAS-dis3L2 RNAi /C ; Df(3L)H99/69B-GAL4) results in a significant increase in wing area compared to either Df(3L)H99/C controls or when dis3L2 is knocked down in a wild-type background (UAS-dis3L2 RNAi /C ; 69B-GAL4/C) (Fig. 7D). Therefore, Dis3L2 acts independently of the Hid, Grim and Reaper induced Caspase pathway and controls proliferation rather than preventing apoptosis. The lack of redundancy between these exoribonucleases is further illustrated by the fact that following the loss of one decay enzyme (Dis3L2) the cell does not increase the transcription of mRNAs encoding either of the other major RNA decay enzymes, Pacman and Dis3 to compensate (Fig. S10).
Although depletion of dis3L2 rescues wing disc area in pcm 14 mutants, the dis3L2 KD , pcm 14 mutant larvae are still developmentally delayed. We have previously shown that the 5 0 -3 0 exoribonuclease Pacman post-transcriptionally regulates the expression of the insulin-like peptide Dilp8 resulting in a developmental delay. 20 To determine whether dilp8 levels are also regulated by Dis3L2, we used qRT-PCR to assess the levels of dilp8 pre-mRNA and mature mRNA after dis3L2 knockdown with or without the pcm 14 mutant background. Fig. 7E shows that dilp8 mRNA levels increase approximately 8000 fold in the dis3L2 knockdown pcm 14 mutant discs (pcm 14 /Y ; UAS-dis3L2 RNAi /C ; 69B-GAL4/C) which is not significantly different to the increase in pcm 14 discs (pcm 14 /Y ; UAS-dis3L2 RNAi ;). These data further emphasizes the specificity of transcripts being targeted to particular exoribonucleases. Taken together these data indicate that Dis3L2 and Pacman have opposing functions of proliferation and apoptosis within cells, which are achieved through the degradation of different and specific RNA targets.

Discussion
In this study we show that the 3 0 -5 0 exoribonuclease Dis3L2 controls organ size in Drosophila melanogaster. Knockdown of dis3L2 in the Drosophila wing imaginal disc results in substantial wing overgrowth due to increased cellular proliferation during the larval stages. Using RNA-seq we have identified a small set of mRNAs that are sensitive to Dis3L2 activity, with only 239 genes changing in expression >1.5-fold out of the 8837 detected. Of the mRNAs that increase in expression upon dis3L2 knockdown, and are therefore likely to be targets of Dis3L2, we identified 2 that change at the post-transcriptional level but not at the transcriptional level, namely CG2678 and pyrexia. We also demonstrate a compensatory effect between Dis3L2 and the 5 0 -3 0 exoribonuclease Pacman where knockdown of dis3L2 in a pacman mutant background results in a normal wing size. Loss of Pacman results in increased apoptosis 21 while dis3L2-depletion promotes proliferation demonstrating that these 2 exoribonucleases function to regulate opposing pathways within the developing tissue.
The results described in this paper are in agreement with previous reports that Dis3L2, unlike its paralogue Dis3, is dispensable for organism viability, 5,6,9 indicating the specific roles of these highly related exoribonucleases. However, we cannot at present rule out that Dis3L2 is required during very early developmental stages, as ubiquitous knockdown with the GAL4 drivers we have used is not effective prior to the syncytial blastoderm stage when the majority of zygotic transcription  Table 1 for p-values. (B) pyrexia (pyx) and CG2678 show post-transcriptional changes in gene expression as the mature mRNA (dark gray) but not the pre-mRNA (light gray) show significant increases in expression. cyt-c-d and CG31808 show transcriptional changes as both the pre-(light gray) and mature (dark gray) mRNA increase in expression. The parental control (left) and dis3L2 knockdown (right) values are presented for each transcript for both the pre-and mature mRNA. n 4, error bars represent standard error. See Table 1 for p-values. commences. Our data are also in line with a previous paper showing that Dis3L2 depletion in HeLa cells results in increased cell proliferation. 12 However, we do not have any evidence that Dis3L2 knockdown inhibits apoptosis, as has previously been reported in HeLa cells. 9 Since our study uses normal cells in the context of a natural tissue rather than as immortalised cells, it is not surprising that our results differ from the above study. The use of a developing organism has also allowed us to show that Dis3L2 normally represses cell proliferation during the early developmental stages of wing disc development (in L2 and early L3 larvae) as depletion in later stages (mid L3 and early pupal) has no effect on proliferation, even though Dis3L2 is normally expressed throughout wing disc development. 24 In addition, in common with the effect of Dis3L2 in human diseases such as Perlman syndrome, our data show that Dis3L2 has specific effects on certain tissues as ubiquitous knockdown affects wing size but does not increase the mass of the whole fly. It is possible that Dis3L2 predominantly affects tissues that are regenerative and/or particularly proliferative such as the kidneys and liver in humans and wing imaginal discs in flies.
The use of RNA-seq as an unbiased global approach allowed us to characterize the ways that Dis3L2 functions to control gene expression. Three main groups of transcripts in L3 wing imaginal discs were identified: those that are expressed at very high levels, those at moderate levels and the Dis3L2 sensitive genes at a very low level. This pattern of gene expression is consistent with our previous work. 20 The general low level of expression in control wing imaginal discs for the Dis3L2 sensitive genes suggests a role for Dis3L2 in maintaining the expression of these specific genes at low levels, presumably to prevent the overgrowth of the developing wing imaginal discs and wings. Our results show that the depletion of dis3L2 results in the deregulation of a small number of transcripts. The small number may be due to silencing efficiency, but is more likely to represent the substrates most sensitive to dis3l2 depletion. We have shown that the upregulation of the transcripts that increase in expression occurs either at transcriptional level or at the post-transcriptional level. The increase in levels of cyt-c-d and CG31808 at the transcriptional level suggest that depletion of dis3L2 results in an increase in activity of one or more transcriptional activators, which in turn enhance expression of these transcripts. Interestingly, cyt-c-d and CG31808 neighbor each other on the left arm of chromosome 2 with their 5 0 UTRs overlapping. It is therefore possible that their increased expression is due to transcriptional co-regulation. Chromatin immunoprecipitation experiments from the modENCODE consortium 25 have identified binding sites for transcription factors such as Trithorax-like (Trl), Dichaete (D), Distal-less (Dll) and Dorsal (dl) within the overlapping region of cyt-c-d and CG31808. Although the expression of these transcription factors at the RNA level remains unchanged following dis3L2 knockdown (Fig. S11) it is possible that their protein activity is altered following the loss of Dis3L2 in the discs, which in turn drives the observed increase in expression of cyt-c-d and CG31808. Expression of cyt-c-d and CG31808 do not change significantly in pacman mutants, 20 again demonstrating the specificity of Dis3L2 on these cellular pathways.
The finding that cyt-c-d is upregulated in dis3L2 knockdown tissue is interesting and suggests a mechanism whereby dis3L2 controls proliferation. Cytochrome C (Cyt c) is an essential component of the mitochondrial electron transport chain, which is involved in the generation of ATP. The role of cyt-c-d in the generation of ATP in Drosophila cells is confirmed by experiments where knockdown of cyt-c-d using RNAi leads to reduced ATP levels. 26 In mammalian cells, it has been shown that, under certain cellular conditions, Cyt c can be released from mitochondria and bind to the apoptosome complex for activation of initiator caspase-9. 27 Although cyt-c-d mutations (e.g. the point mutation cyt-c-d Z2-1091 ) result in inhibition of apoptosis during development of the retina in Drosophila, 28,29 the mechanism whereby this may occur in Drosophila is controversial. 27 Since our experiments show no effect of dis3l2 knockdown on apoptosis, a simple hypothesis is that increased cyt-c-d levels result in an increase of ATP in the cell. Since ATP is required for the generation of ribosomes, and ribosomes are required for cell proliferation, the increase in cyt-c-d is consistent with an increase in cell proliferation.
Our experiments have identified potential Dis3L2 targets, CG2678 and pyx, which increase post-transcriptionally upon dis3L2 depletion. These changes in expression have been shown to be specific to the loss of Dis3L2 as they are not seen in pcm mutant wing imaginal discs. 20 CG2678 is predicted to encode a transcription factor with homology to the human Kr€ uppel C2H2-type zinc finger proteins ZNF37A and ZNF768. 24 The biological functions of these proteins are not known. pyrexia (pyx; dTRPA2; CG17142) encodes a transient receptor potential (TRP) channel which mediates temperature response in Drosophila. 30 The TRP family of receptors are evolutionarily conserved throughout metazoans and comprise ankyrin repeats plus a cation transport membrane domain, which is particularly permeable to potassium ions. 31 Although the relationship between pyx and proliferation has not been studied, it is known that potassium ions are involved in the cell cycle and proliferation 32 and that mutations in the gene encoding the TRP channel TRPP2 (PKD2) can result in polycystic kidney disease. 33 Interestingly, the Drosophila retinoblastoma protein RBF1 and the transcription factor E2F, both of which play important roles in cell division and cell cycle regulation, bind to the promoter of pyx thus linking pyx to cell proliferation. 34 Regulation of the potassium channel protein Pyx by Dis3L2 may represent a novel way of controlling tissue proliferation.
The findings that the 3 0 -5 0 exoribonuclease Dis3L2 regulates proliferation while the 5 0 -3 0 exoribonuclease Pacman controls apoptosis are novel and intriguing. Our data strongly suggest that Dis3L2 and Pacman regulate different and opposing pathways in proliferation and apoptosis, rather than both of them affecting the same pathway in opposite directions. This is demonstrated in our experiments showing that Dis3L2 is not functioning to prevent overgrowth through the regulation of the expression of the proapoptotic genes hid, grim and rpr. In contrast, our previous results 21 showed that Pacman post-transcriptionally regulates these RNAs to inhibit apoptosis. Therefore, control of apoptosis and proliferation by these 2 exoribonucleases is achieved by regulation of independent cellular pathways. Although Pacman and Dis3L2 can control the balance of apoptosis/proliferation to determine the final size of the wing, they do not control compensatory pathways to control developmental delay. Knockdown of Dis3L2 does not affect the expression of dilp8 or developmental delay in wild-type or pcm mutant backgrounds further emphasizing the distinct target specificity of these 2 exoribonucleases. This result would appear to conflict with previous experiments on human tissue culture cells, which have identified an RNA-dependent interaction between human Dis3L2 and XRN1, suggesting they target the same RNAs for degradation. 6 A possible explanation is that many RNAs can be degraded by both the 5 0 -3 0 and 3 0 -5 0 pathways; however, there are a small number of transcripts which are specifically targeted by Dis3L2 or Pacman.
How is this specificity achieved? Recent research has demonstrated that Dis3L2 preferentially targets RNAs and pre-miRNAs that are tailed by the addition of uridines. 5,[9][10][11] This tailing is performed by uridyltranferases (TUTases) which in Drosophila include the uridyl transferase Tailor. Tailor is known to uridylate intron derived pre-miRNAs showing preference for substrates with a 3 0 terminal Guanine. 35 Tailor may also play a role in the uridylation of mRNA substrates and therefore may be responsible for targeting specific mRNAs for Dis3L2 degradation. A further possibility is that Dis3L2 targets could have specific instability elements within their 3 0 UTRs, similar to AU rich elements used by the exosome to target mRNAs for decay. 36,37 Finally, specificity may be conferred by binding of specific miRNAs which then induce RNA degradation. In summary, our work has led to the discovery of the importance of the 3 0 -5 0 exoribonuclease Dis3L2 in cell proliferation plus tissue growth and shows that it controls a specific set of RNAs independent of the 5 0 -3 0 degradation pathway. This work also demonstrates that Drosophila provides an excellent model system to understand the molecular basis of human overgrowth syndromes such as Perlman syndrome and Wilms' tumor.

Fly husbandry
Fly stocks were cultivated on standard media in uncrowded conditions at 25 C unless otherwise stated. All stocks used were obtained from the Bloomington stock center unless stated otherwise.

Genetic crosses and genotypes
To knockdown dis3L2 throughout the wing imaginal disc ;;69B-GAL4 females were crossed to ;UAS-dis3L2 RNAi ; males producing knockdown offspring with the genotype ;UAS-dis3L2 RNAi /C ; 69B-GAL4/C. In addition to the parental controls ;UAS-EGFP RNAi ; males were also crossed to ;;69B-GAL4 females to produce control offspring ;UAS-EGFP RNAi /C ; 69B-GAL4/C which accounted for the expression of a hairpin RNAi construct.
To knockdown dis3L2 within the wing pouch of the wing imaginal disc ;nub-GAL4; females were crossed to ;UAS-dis3-L2 RNAi ; males producing knockdown offspring with the genotype ;UAS-dis3L2 RNAi /nub-GAL4;. In addition to the parental controls ;UAS-EGFP RNAi ; males were also crossed to ;nub-GAL4; females to produce control offspring (;UAS-EGFP RNAinub-GAL4;) which accounted for the expression of a hairpin RNAi construct.
dis3L2 knockdown was achieved in the posterior compartment of the wing imaginal disc by crossing ;en-GAL4/CyO; females to ;UAS-dis3L2 RNAi ; males to produce ;en-GAL4/UAS-dis3L2 RNAi ; offspring. Due to the en-GAL4 parents containing the CyO marker making them unsuitable for measurements the ;UAS-dis3L2 RNAi ; parents were used as controls.

GAL80 ts experiment
; tub-GAL80 ts ; 69B-GAL4 homozygous females were crossed to ; UAS-dis3L2 RNAi ; males and left to mate for 2 d at the appropriate temperature in addition to parental controls for all time points (19 C for all developmental timings except the positive control at 29 C). Flies were tipped into new vials for 8 hour egg lays. Vials for both the parental controls and the ; UAS-dis3-L2 RNAi /GAL80 ts ; 69B-GAL4/C flies were moved from 19 C to 29 C at the required times for each developmental stage taking into account a 24-32hr perdurance. Each comparison between knockdown and controls was made specifically between those for each time point.

Wing mounting and measurement
Flies were aged to between 1 and 2 days old for all measurement experiments. A single wing (the 'left-hand' wing) was cut from each fly using microscissors and stored in isopropanol for 1 hour before being mounted in DPX (Sigma cat. no. 06522). For all experiments the controls represent the mean wing area of the parental controls together with an EGFP RNAi driven by the same GAL4 drivers. All results shown were obtained with male wings, however female wings also showed the same phenotypes. Mounted wings were measured using Axiovision 4.7 on an Axioplan microscope (Carl Zeiss). UAS-EGFP RNAi was driven by both 69B-GAL4 and nub-GAL4 and when normalized to fly mass they were not significantly larger the parental controls.

Disc dissection, mounting and measurement
Wing imaginal discs were dissected from late L3 larvae in PBS. Late L3 larvae were staged by the addition of bromophenol blue to the food (0.05%) followed by selection of larvae that had cleared the dyed food from their gut. Dissected discs were mounted on Poly-L-Lysine slides and mounted in 85% glycerol. Disc areas were measured using Axiovision 4.7 on an Axioplan microscope (Carl Zeiss).

Calculation of the mitotic index
The number of nuclei undergoing mitosis were counted using the ImageJ plugin DeadEasy MitoGlia. 39 All discs were stained with anti-Phosphohistone H3 as above under the same conditions. The mitotic index was then calculated for each disc by dividing the number of cells in M phase by the area of the disc.

RNA extraction and qRT-PCR
RNA extractions were performed using a miRNeasy RNA extraction kit (Qiagen, cat. no. 217084), with on-column DNAse digestion (Qiagen, cat. no. 79254). RNA concentrations were measured on a NanoDrop1000 spectrophotometer (Thermo Scientific). When examining wing imaginal discs at least 30 imaginal discs were collected for each biological replicate.
For qRT-PCR, 500ng of total RNA was converted to cDNA in duplicate using a High Capacity cDNA Reverse Transcription Kit (Life Technologies, cat. no. 4368814) with random primers or oligo(dT) primers. A control "no RT" reaction (without enzyme) was performed in parallel to confirm that all genomic DNA had been degraded. qRT-PCR was performed on each cDNA replicate in duplicate (i.e. Four technical replicates), using TaqMan Universal PCR Master Mix, No AmpErase UNG (Life Technologies, cat. no. 4324018) and an appropriate TaqMan assay (Life Technologies). For custom pre-mRNA assays, the pre-mRNA sequence of the desired target area was submitted to Life Technologies' web-based Custom TaqMan Assay Design Tool as in 22 (Fig. S12). RpL32 (Rp49) gene was used for normalization.

RNA-seq analysis and selection of consistently changed genes
RNA extraction was performed as above from 60 wandering L3 wing imaginal discs. RNA integrity was checked using an Agilent 2100 Bioanalyzer. 3mg of total RNA per replicate was sent to Oxford Gene Technology for sample preparation and sequencing. Libraries were poly(A) selected and then were run on a single HiSeq2000 lane using TruSeq v3 chemistry (Illumina) generating between 12 and 15 million paired end reads of 100bp per sample. The RNA-seq data has been deposited in the ArrayExpress repository under the accession number E-MTAB-4545.
Due to the knockdown replicates being in duplicate the statistical output from Cufflinks could not be used. Therefore stringent filtering was applied. Only genes showing a fold change of 1.5-fold when compared between each knockdown and parental control replicate were selected as misexpressed genes. The Cuffdiff output for the misexpressed genes can be found in Supplemental File 1. Only those with an average fragments per kilobase per million (FPKM) of 0.3 where considered for validation by qRT-PCR due to the difficulty of reliably detecting very lowly expressed genes by qRT-PCR. The FPKM cut-off of 0.3 was chosen because this threshold has been used previously 43,44 and it provided results which are reliable enough to be verified by qRT-PCR. Those showing the highest reliable fold changes together with a suitable average FPKM were selected for verification.

Statistical tests
All statistical analyses were performed in either R v3.1.2 or GraphPad Prism 6. Two-sided 2-sample t-tests were used to compare the means of single test groups to single control groups. If multiple comparisons were required, a one-way ANOVA was performed with a post-test to compare the means of each possible pair of samples.

Disclosure of potential conflicts of interest
No potential conflicts of interest were disclosed.