Overexpression of CyclinT in imaginal discs partially rescues the Cdk9 DN

The strong rescue of the Cdk9^DN phenotype by CyclinT expression in larval tissues led us to investigate the effects of the co-expression of Cdk9^DN with CyclinT or CyclinK in other tissues. We found that the effects seem to depend on the tissues tested. For instance, we found no obvious modification of the leg phenotype seen with the c701b driver. In contrast, we found that the phenotype induced by the restricted expression of Cdk9^DN in the eye disc using the GMR driver was completely suppressed when CyclinT was co-expressed (Fig. 37, compare B and D), while the co-expression of CyclinK does not modify the phenotype (Fig.

37C). Similarly, we observed a significant rescue of the wing size and of the vein patterning defect caused by Cdk9^DN expression in the posterior compartment of the wings when CyclinT was co-expressed under control of the en-GAL4 driver (Fig. 37, compare H and J). On the other hand, the co-expression of CyclinK does not strongly modify the wing phenotype (Fig.

37I). The tendency of the overexpressed CyclinT to systematically rescue the Cdk9^DN phenotypes is unexpected. Although these results might indicate a requirement for P-TEFb

Figure 36. CyclinT expression rescues the Cdk9^DN endoreplication defect. (A-F) DAPI staining of salivary glands from third instar larvae expressing (A) Myc:Cdk9, (B) Myc:Cdk9^DN, (C) Myc:Cdk9/Flag:CyclinK, (D) Myc:Cdk9^DN/Flag:CyclinK, (E) Myc:Cdk9/HA:CyclinT and (F) Myc:Cdk9^DN/HA:CyclinT under control of the AB1 driver. The expression of mutant Cdk9 alone (B) or together with CyclinK (D) caused similar growth defects, while the co-expression of CyclinT with Cdk9^DN (F) rescued the phenotype. (G-L) Salivary glands from larvae co-expressing HA:CyclinT and Myc:Cdk9 (G-I) or Myc:Cdk9^DN (J-L) under control of the AB1 driver, stained with anti-Myc (red) and anti-HA (green) antibodies. Nuclei expressing wild-type or mutant kinase were labeled with comparable intensity using one or the other antibody.

(A) : UASp-Myc:cdk9/+ ; AB1/+ ; (B) : UASp-Myc:cdk9^DN/+ ; AB1/+ ; (C) : UASp-Myc:cdk9/+

; UASp-Flag:cyclinK/AB1 ; (D) : UASp-Myc:cdk9^DN/+ ; UASp-Flag:cyclinK/AB1 ; (E,G,H,I) : UASp-Myc:cdk9/+ ; UASp-HA:cyclinT/AB1 ; (F,J,K,L) : UASp-Myc:cdk9^DN/+ ; UASp-HA:cyclinT/AB1.

containing CyclinK during growth and differentiation of the imaginal tissues, this interpretation is difficult to reconcile with our previous observation that CyclinT function is also required during Drosophila development (Table 3 and 4). To further investigate the requirement of one or the other cyclin in the imaginal tissues, we re-examined the Cdk9^DN phenotype in the eye by decreasing the dose of wild-type cyclinT or cyclinK. We found that the rough eye phenotype caused by Cdk9^DN expression using the GMR driver can be significantly enhanced in flies carrying a deficiency uncovering the cyclinT gene, while the phenotype was not modified in flies carrying a mutant cyclinK allele (data not shown). These latest results suggest that CyclinT, rather than CyclinK, is required for normal growth and differentiation of the imaginal discs. At present, we have not investigated in details the requirement for CyclinT and CyclinK in the other somatic tissues and in the germline, and further experiments are thus necessary to determine the role of either P-TEFb complex in cell growth and differentiation in selective tissues.

Discussion

CyclinK seems to be a regulatory partner of Cdk9 in Drosophila.

Previous studies have demonstrated that human CyclinK can function as a Cdk9 regulatory subunit, and that Cdk9/CyclinK possesses intrinsic RNAPII kinase activity in vitro (Fu et al., 1999 ; Lin et al., 2002c). In Drosophila, CyclinK is an essential gene whose mutation causes embryonic to early larval lethality. By expressing a transgenic copy of the wild-type cyclinK gene or a tagged version of the transgene that contains a triple Flag-epitope, we completely rescued the lethality and thus demonstrated unequivocally that our transgenes are functional in vivo (Table 4). In similar ways, we showed that the lethality caused by cdk9 and cyclinT mutations can be efficiently rescued through expression of our Cdk9 and CyclinT transgenic constructs, and that their tags do not interfere with the function of the proteins (Table 2 and 3). When expressed in tissue culture cells, Flag:CyclinK and HA:CyclinT form stable heterodimers with Myc:Cdk9 and most of the immunoprecipitated kinase was found in association with either cyclin (Fig. 14). These results indicate that, similarly to CyclinT, CyclinK tightly interacts with Cdk9 in Drosophila cells and that the tags have no detectable effects on the protein interactions. Interestingly, despite the low sequence similarity at the amino acid level between the cyclin box of CyclinT and the one of CyclinK (27% identity), Cdk9 likely interacts with either cyclin with comparable efficiency. This observation does not rule out however that one particular cyclin would have a higher affinity for Cdk9 binding in vivo. This could be tested by competition assays in triple transfected cells. Nevertheless, the tight association of CyclinK with Cdk9 suggests that, as previously proposed in human, CyclinK represents an alternative Cdk9 regulatory subunit in Drosophila.

Transgenic P-TEFb complexes are recruited to transcriptionally active loci.

To investigate the functional significance of having two distinct P-TEFb complexes, we first compared the global distribution of the transgenic proteins by immunofluorescence staining of polytene chromosomes. Particularly, we exploited the inducible Drosophila heat shock genes, which provide a useful model to investigate the recruitment of transcription factors on polytene chromosomes at moderate resolution. In uninduced larvae, endogenous CyclinT has previously been detected at approximately 200 sites along the chromosomes arms

(Lis et al., 2000). In agreement with this, our tagged CyclinT and Cdk9 proteins were broadly distributed on chromosomes and colocalized with comparable intensity at many sites, including chromosomal puffs (Fig. 16). Moreover, the binding pattern of overexpressed Cdk9 overlap completely with RNAPII phosphorylated at Ser-2 (Fig. 18), which is consistent with a general role for P-TEFb in regulating transcription elongation. In contrast to the elongating polymerases, the hypophosphorylated RNAPII (PolIIa) and initiating polymerases (PolIIo^Ser5) have been detected in a widespread pattern on chromosomes and overlap to a large extend to sites of transcriptional quiescence (Weeks et al., 1993 ; O’Brien et al., 1994 ; Wu et al., 2003a

; Srinivasan et al., 2005). This is notably a feature of the uninduced hsp26 and hsp70 heat shock genes, whose promoters are each constitutively occupied by one engaged, but paused, polymerase (Gilmour and Lis, 1986 ; Lis, 1998 ; Lis et al., 2000 ; Kaplan et al., 2000 ; Andrulis et al., 2000 ; Wu et al., 2003a). The trend location of our tagged Cdk9/CyclinT proteins at developmental puffs and the strong correlation of Myc:Cdk9 staining with the elongating polymerase indicate that overexpressed P-TEFb complexes are efficiently recruited to transcriptionally active loci in vivo, but poorly associated with promoter-proximal paused RNAPII. These observations are in complete agreement with data obtained by chromatin IP (Boehm et al., 2003).

To investigate the ability of P-TEFb containing CyclinK to be recruited to chromatin, we co-expressed tagged Cdk9 and CyclinK and examined their distribution on chromosomes.

Interestingly, we found that both proteins broadly colocalized on chromosomes arms with prominent labeling at developmental puffs (Fig. 15). This observation indicates that CyclinK is efficiently targeted to transcriptionally active loci, and that Cdk9/CyclinK heterodimer might functionally substituted for P-TEFb comprised of Cdk9 and CyclinT in vivo.

Alternatively, it is possible that a particular cyclin subunit is actually required to activate transcription of selective targets. To test the latter possibility, we co-expressed both cyclins and compared their distribution profiles on chromosomes. In this condition, the overexpressed CyclinT and CyclinK subunits compete for assembly with the endogenous kinase, and thus a preferential interaction of Cdk9 with one or the other cyclin should be detectable at the chromatin level. Instead, we found a striking co-localization of the two proteins at almost all developmental puffs and many additional loci, and their staining intensity were roughly equivalent (Fig. 17). This result does not support the idea that distinct P-TEFb are recruited to specific targets, at least in this particular tissue and in the condition of similar levels of expression. In human cells, it is unlikely that P-TEFb binds spontaneously to chromatin, even when P-TEFb is overexpressed. Instead, a wide variety of transcriptional cofactors binds to

and recruits the kinase to selective genes (Table 1). Many of these recruiting factors were shown to interact with the cyclin box of CyclinT1, while no interactions with CyclinK, except for the particular case of Tat transactivation, have been reported to date. It is conceivable that CyclinK might substitute, at least partially, some functions of the T-type cyclins since both complexes Cdk9/CyclinK and Cdk9/CyclinT1 were shown to phosphorylate the RNAPII CTD in vitro with comparable efficiency (Fu et al., 1999). The mechanisms of P-TEFb recruitment in Drosophila remain unknown. Although the targeting of P-TEFb is controlled by multiple factors in mammalian system, other strategies may have evolved in other organisms. In fission yeast, a recent study suggests that P-TEFb is recruited to the elongation complex through interaction with the capping enzymes, a mechanism that could be commonly used for all protein-coding genes in this species (Guiguen et al., 2007). Similar strategy could exist in Drosophila to bring the kinase in close proximity to the RNAPII CTD, and might be independent of the regulatory cyclin. In any case, our results clearly showed that both cyclins are efficiently recruited to the same loci, suggesting that both P-TEFb complexes share common mechanisms for chromatin binding.

Heat shock treatment of larvae causes a rapid activation of the heat shock genes and a concomitant reduction in transcription of many normally expressed genes (Lis et al. 1981).

This feature of the heat shock genes has been largely exploited to investigate the requirement for transcription elongation factors in vivo as it provokes a dramatic relocation of proteins at the heat shock loci on chromosomes. Many transcription factors are recruited with comparable kinetics to these loci including RNAPII (Greenleaf et al., 1978 ; Weeks et al., 1993), heat-shock factor (HSF, Westwood et al. 1991), P-TEFb (Lis et al., 2000), DSIF (Wu et al., 2003a) and the chromatin remodelers Chd1 (Stokes et al., 1996), Spt6 (Kaplan et al., 2000 ; Andrulis et al., 2000) and FACT (Saunders et al., 2003). To determine whether our overexpressed P-TEFb complexes are also actively recruited to the heat shock loci, we compared the binding of all subunits immediately after heat shock. Strikingly, while both complexes are normally located at hundreds loci on polytene chromosomes, all proteins redistribute rapidly to the heat shock loci and colocalize with the elongating polymerases (Fig.

19). This recruitment of exogenous P-TEFb containing CyclinT or CyclinK to the heat shock puffs is comparable to what was previously observed for endogenous CyclinT (Lis et al., 2000), and thus strongly suggests that our ectopically expressed proteins are targeted to active genes with similar kinetics to endogenous P-TEFb. However, these data does not provide evidence that overexpressed P-TEFb can functionally substituted for endogenous complexes once recruited to active promoters. Nevertheless, the fact that the tagged proteins can rescue

the lethality caused by their respective mutations, and that ectopically expressed P-TEFb actively compete with the endogenous complexes for chromatin binding, it is likely that the overexpressed Cdk9/CyclinT and Cdk9/CyclinK also stimulate transcription elongation through their kinase activities.

Disruption of Cdk9 function alters larval growth and organ patterning.

We have analyzed the physiological role of Cdk9 kinase by expressing a dominant-negative form of the protein at different stages of Drosophila development. Cdk9^DN was generated by substitution of the well conserved aspartate residue D199 within the Drosophila protein active site, which corresponds to D167 of the human homolog, a mutation known to abolish the kinase activity in both species (Garriga et al, 1996a ; Peng et al, 1998a). An ectopic expression of Cdk9^DN in the eye discs or in the wing discs have dominant effects on both the morphology and growth of the adult structures (Fig. 20 and 25). These dominant phenotypes were either enhanced or suppressed by respectively decreasing or increasing the dose of wild-type cdk9 demonstrating that, in these conditions, we could alter Cdk9 function specifically. To assess whether the Cdk9^DN phenotypes were due to competition with the endogenous kinase to bind to its substrates, we analyzed the levels of RNAPII phosphorylation at Ser-2 in both tissue culture cells and salivary glands after a pulse of Cdk9^DN expression. By immunostaining of polytene chromosomes, we observed a remarkable decrease of Ser-2 phosphorylation by 10 hours following the first induction of cdk9^DN expression, and a wide-spread binding of the mutant protein (Fig. 21D-F). This result indicates that Cdk9^DN efficiently competes with the endogenous kinase for promoter binding and blocks the accessibility of the wild-type kinase to the RNAPII CTD. Consistently, P-TEFb inhibition using the kinase-specific inhibitor flavopiridol has been found to provoke a complete disappearance of Ser-2 phosphorylation on polytene chromosomes (Ni et al. 2004).

Intriguingly, we were not able to detect such a reduction of RNAPII phosphorylation by western blotting analysis from S2 cell extracts (Fig. 21A) or from various Drosophila tissues expressing cdk9^DN including salivary glands (data not shown). Previous studies have pointed to the fact that H5 antibody, despite being widely used and likely specific for the Ser-2 phosphorylated polymerase in immunostaining or chromatin IP experiments, may actually hybridize both Ser-2 and/or Ser-5 phosphorylated CTD in some protein assays (Patturajan et al., 1999 ; Cho et al., 2001 ; Licatalosi et al., 2002 ; Jones et al., 2004). It is therefore plausible

that our failure in detecting decreased levels of Ser-2 phosphorylation by immunoblotting reflects a lack of specificity of the H5 antibody in denaturing conditions.

In order to identify the cellular functions or the tissues that are more sensitive to Cdk9 activity, we tested the effect of Cdk9^DN expression in various Drosophila tissues. Consistent with a general requirement for P-TEFb function in developing Drosophila, we found that constant ubiquitous expression of Cdk9^DN, starting during embryogenesis, is lethal during the late embryonic to the first larval stage at 25°C. This precocious lethality contrasts with the tardive effect reported in RNAi knockdown of Cdk9 in flies, where death was delayed to the pupal stage (Eissenberg et al., 2006). Cdk9 is however unambiguously an essential gene in metazoans and mutation of the endogenous gene in fly (this work) or RNAi knockdown of Cdk9 in C. elegans also results in embryonic lethality (Shim et al., 2002). Our transgenic approach allowed us to modulate the levels of GAL4 activity in flies and thus the level of Cdk9^DN by increasing or decreasing the temperature. In less severe condition (18°C), the lethality caused by ubiquitous expression of Cdk9^DN can be delayed to the third-instar larvae revealing a strong effect on the endoreplication process. During larval development, endoreplication is responsible for the massive growth of the larvae that take place in all larval tissues except imaginal discs and nervous system (reviewed by Edgar and Orr-Weaver, 2001).

Consistent with a role for P-TEFb in regulating endoreplication, we did not detect significant effect of Cdk9^DN on growth of the brain and imaginal tissues. The failure of Cdk9^DN -expressing larvae to grow beyond the first instar at 25°C suggests that the lethality is primarily due to a larval growth defect. Larval growth involves both increase cell size and DNA endoreplication (Edgar and Orr-Weaver, 2001). Our BrdU incorporation experiments in salivary glands clearly showed that, during the whole third-instar larvae, Cdk9^DN-expressing cells failed to synthesis DNA and remained small (Fig. 23). It is not yet clear however whether the limited size of the cells is a consequence of the reduced rate of endoreplication.

The mechanisms controlling the endocycles in larval cells are not well understood.

Endocycle consists of distinct S and gap phases bypassing mitosis, regulated notably by the oscillatory activity of CyclinE/Cdk2 and its kinase inhibitor Dacapo (de Nooij et al., 2000 ; Edgar and Orr-Weaver, 2001). More recently, dMyc function has also been involved in regulating endoreplication in larval tissues (Pierce et al., 2004) as well as in the Drosophila ovary (Maines et al., 2004). The dMyc protein and its ortholog c-Myc in vertebrates are thought to regulate a wide range of cellular processes through transcriptional control of genes required for growth and proliferation. In agreement with this notion, a large set of Myc target genes identified in Drosophila and in mammals are growth related factors (Orian et al., 2003 ;

Fernandez et al., 2003). In human cells, P-TEFb is critically required for Myc-mediated transcription and P-TEFb inhibition eventually impairs the Myc-induced proliferation in vivo (Eberhardy and Farnham, 2002 ; Kanazawa et al., 2003 ; Gargano et al., 2007). In Drosophila, dm loss-of-function has phenotypes similar to cdk9^DN expression, where both larval growth and DNA replication are impaired (Pierce et al., 2004). These observations led us to ask whether the growth defects induced by Cdk9^DN were due to the misfunction of dMyc. The ability of ectopically expressed dMyc to rescue the Cdk9^DN phenotype suggests that it may indeed be the case (Fig. 24). At this time however, we can not say whether the Cdk9^DN phenotype is a direct consequence of a decreased expression of dm itself and/or of the dMyc target genes involved in endoreplication. This hypothesis could be tested by quantitative RT-PCR to compare the levels of dMyc transcripts in control and cdk9^DN expressing larvae. In endoreplicating tissues, dMyc function seems mainly required for cell growth and endocycle, but not for mitotic proliferation prior to the onset of endoreplication (Maines et al., 2004 ; Pierce et al., 2004). Although the mechanisms by which dMyc regulates endocycle are currently unknown, a small number of its target genes are involved in DNA replication suggesting that dMyc may directly stimulate endocycle (Orian et al., 2003). Moreover, Pierce et al. (2004) provides evidence that the cell cycle control exerted by the oscillating activity of CyclinE/Cdk2 is downstream of dMyc function, even though CycE is not directly regulated by Myc (Orian et al., 2003). These data, together with the observation that dMyc expression is sufficient to promote endoreplication (Fig. 24), suggest that Myc is an important upstream regulator of the endocycle.

Impairing P-TEFb activity in mammals has profound effect on c-Myc function, as Myc was shown to recruit the kinase directly. If a similar strategy is used by the Drosophila dMyc protein, we would also expect a more sensitive effect of Cdk9^DN on dMyc target genes transcription. We have tested a putative genetic interaction between dm and cdk9 alone or together with either cyclin in the eye using a dm mutant strain that contain a rescuing dm cDNA flanked by FRT sites (Bellosta et al., 2005 ; see materials and methods). The dm cDNA is expressed in all parts of the body except for the head capsule, where the Flp-mediated recombination system eliminates the cDNA and allows expression of GAL4 instead. In these experiments, we found that the eye phenotype causes by hypomorphic or amorphic dm alleles were not strikingly enhanced by our cdk9^DN transgenes, nor rescued by an overexpression of wild-type cdk9 (data not shown). The absence of genetic interaction between dm and cdk9 does not favour the possibility that dMyc directly interacts with and recruits P-TEFb to activate transcription. Instead, as discussed previously, we believe that P-TEFb is recruited to

promoter independently to the dMyc transactivator. As a consequence, the dMyc targets would not be more sensitive to cdk9 mutation than any other genes that required P-TEFb activity, whatever the tissue examined including salivary glands. Taken together, these data suggest that the growth defect caused by Cdk9^DN in larval tissues is primarily due to a decreased expression of dm, and possibly others currently unknown endocycle regulators, resulting in lower activation of the dMyc target genes.

A restricted expression of Cdk9^DN in imaginal tissues has also severe consequence on

A restricted expression of Cdk9^DN in imaginal tissues has also severe consequence on