Our experiments performed in the Drosophila ovary have shown that impairing P-TEFb activity in the germline results in female sterility (Fig. 27E). This observation contrasts with the fertility of the males expressing Cdk9DN using the same GAL4 driver. Although the germline-specific nos-GAL4:VP16 driver can readily activate expression of our cdk9 transgenes in all female germ cells at all stages of oogenesis (Fig. 27B,C), it may be less efficient in males. Despite the fact that this driver is known to be active in the male germline, we assumed that the levels of Cdk9DN expression in testis is not high enough to impair P-TEFb function. For this reason we concentrated our analysis on the female germline.
Our data indicated that Cdk9 function is required for both, the control of the cystocyte divisions and the proper migration of the pre-follicular cells around the germline cysts. The follicle cells are indispensable for oocyte development. These cells produce yolk proteins and eggshell components and participate in signaling events with the germline that determine the future embryonic axes (reviewed by Horne-Badovinac and Bilder, 2005). The mature eggs produced by cdk9DN germline exhibit multiple defects which correlate with defective functions of the follicle cells, including their lack of yolk deposition and their abnormal morphologies (Fig. 27H). The characterization of mutant follicles produced by cdk9DN germline allowed us to identified two kinds of defective processes affecting the follicle cells.
On the one hand, the presence of two or more individual cysts within a single follicular epithelium strongly suggested a defect in encapsulation (Fig. 29F-H). Compound egg chambers, which arise from mispackaging events, are known to be formed within the germarium when the follicle cell precursors fail to migrate between neighboring germline cysts, without necessarily interfering with the cellular fate of the follicle cells. In agreement with this notion, the ability of the follicle cell precursors to differentiate as polar cells or stalk cells in cdk9DN chambers were unlikely to be affected (Fig. 30B). In cdk9DN germaria, particularly when two copies of the transgene were expressed, we clearly observed a lack of follicle cell migration in region 2 (Fig. 31E), demonstrating that the pre-follicular cells indeed failed to recognize individual cysts. The mechanisms by which the follicle cell precursors precisely recognize and encircle a cyst are largely unknown. Many studies have however
provide evidence that signaling molecules, emanating from both somatic and germline cells, are required for the proper migration of the follicle cells. In this respect, the mutations of a few genes in the germline such as Bicaudal-D (Bic-D), egghead (egh) and brainiac (brn) were shown to disrupt the encapsulation process. These mutations produce compound follicles with a tendency to split 16-cell cysts between adjoining egg chambers (Goode et al., 1996 ; Oh and Steward, 2000). This suggests that, once a complete cyst has been formed, the germ cells must produce a signal facilitating the recognition by the follicle cells of the boundary of the cyst. In cdk9DN germline, it is possible that the timely expression of these genes is impaired, leading to inefficient encapsulation of individual cysts. Beyond the encapsulation defects, cdk9DN expression was also found to cause egg chamber fusions. This phenotype was observed essentially when two copies of cdk9DN were expressed, and can occur within the vitellarium (Fig. 30C). The tendency of egg chambers to fuse often correlates with the absence of interfollicular stalk cells (McGregor et al., 2002 ; Bai and Montell, 2002 ; Torres et al., 2003).
Torres et al. (2003) notably showed that disruption of Notch signaling by depleting Delta from the germline does not impair the migration or the ability of the follicle cells to encapsulate individual cysts. The resulting follicles exhibited however a lack of interfollicular stalks as they leave the germarium, which caused adjacent egg chambers to be tightly apposed. The two layers of follicular cells that separate the germline cysts eventually degenerated to produce compound follicles. Our data suggest that similar defect may occur in cdk9DN ovaries. The polar cells, stalk cells and epithelial follicle cells are thought to differentiate from the follicle cell precursors in a stepwise manner. At the transition between the region 2B and 3 of the germarium, the follicle cells are subdivided by an unknown mechanism into two populations : the stalk/polar cell precursors that express fringe (fng), and the epithelial follicle cells that do not (Grammont and Irvine, 2001). A second spatial organization splits the stalk/polar precursor population into cells that contact the germline and those that do not. Current data suggest that the stalk/polar precursors directly contacting the germline will receive the highest level of Delta signaling from the germ cells inducing their differentiation as polar cells (Lopez-Schier and St. Johnston, 2001 ; Assa-Kunik et al., 2007).
The newly formed polar cells then expressed the JAK/STAT pathway ligand Unpaired, which subsequently induces the remainder precursors to differentiate as stalk cells (McGregor et al., 2002). The fate of polar and stalk cells appears relatively plastic and continuously depends on the balance between Notch/Delta and JAK/STAT signalling (Assa-Kunik et al., 2007).
Initially four to five polar cells are induced at each pole, but several undergo apoptosis during egg chamber maturation in the vitellarium to produce the two characteristic pairs (Besse and
Pret, 2003). In Cdk9DN ovaries, egg chamber fusion seems to occur already in the germarium region 3 (Fig. 30C,D), suggesting that the stalk/polar cell differentiation program is impaired.
Although many of the fused egg chambers possessed two to four pairs of polar cells, some compound follicles did not show any FasIII staining (Fig. 30C). These results suggest that alteration of P-TEFb function in the germ cells may indeed affect the timely expression of genes required for stalk/polar cell differentiation including fringe and Delta. This hypothesis could be examined by immunostaining of cdk9DN ovaries to determine whether these genes are indeed downregulated.
Another category of egg chamber phenotype suggested a defect in the number of cystocyte divisions (Fig. 29I-K). In many cases, we observed a tendency of the germ cells to proliferate beyond the fourth mitotic division, although the extra round of division may concern only a subset of cells. This suggests that impairing P-TEFb activity affects the ability of the germ cells to exit the cell cycle rather than promoting premature cell cycle arrest.
Although the mechanism by which the number of cystocyte divisions is limited to four has not been elucidated, previous studies indicated an important role for the fusome in regulating the synchronization and perhaps the timing of the germ cell mitosis. Mutations of genes encoding various components of the fusome, such as hts, α- and β-spectrin, ankyrin (Yue and Spradling, 1992 ; Lin et al., 1994 ; de Cuevas et al., 1996 ; Deng and Lin, 1997) or Dynein heavy chain (McGrail and Hays, 1997), result in follicles containing less than 16 germ cells.
The proper regulation of proteins involved in cell cycle progression including CyclinA, CyclinB and CyclinE is also critical for accurate cystocyte divisions. Lilly et al. (2000) have shown that overexpression or loss-of-function mutations in CycA and CycB lead to the production of cysts containing exactly twice or half the normal number of germ cells respectively. Similarly, disruption of the ubiquitin-dependent proteasome complex involved in cell cycle protein degradation also results in extra cyst division (Lilly et al. 2000). During mitosis, CyclinA transiently associates with the fusome in all cystocytes suggesting that this structure may coordinates the activity of cell cycle regulators to ensure the synchronous cystocyte divisions (Lilly et al. 2000). In cdk9DN germaria, we occasionally observed a persistence of CyclinA staining in 16-cell cysts (Fig. 31C), indicating that the germ cells may indeed fail to exit the cell cycle after the fourth cystocyte division. In more severe conditions, obtained by expression of two cdk9DN trangenes, CyclinA expression was not anymore restricted to the region 1 of the germarium, but was detected more posteriorly in region 2 or 3 and even within the vitellarium in certain germ cells (Fig. 31D-F). Interestingly, our observation that CyclinA may be detected only in a subset of cells suggests that the
synchronization of the cystocyte divisions was lost. It is possible that some functions of the fusome or proteins that associate with the fusome to regulate this process were impaired. In most cases however, HtsF staining did not revealed breakage of the fusomes. Instead, the extension of the branching fusome in certain cysts indicated that all dividing germ cells remained interconnected and the resulting cysts were normally polarized (a single oocyte was differentiated). We suggest that the extra cystocyte division is due to inefficient turn over of CyclinA and perhaps other cell cycle regulators after the fourth mitosis. Mutations of the encore gene frequently produce egg chambers containing 32 germ cells (Hawkins et al., 1996
; Ohlmeyer and Schupbach, 2003). In these mutants, the germline cysts fail to exit the cell cycle and accumulate high levels of CyclinA and CyclinE proteins (Ohlmeyer and Schupbach, 2003). The authors proposed that Encore function is necessary for the timely degradation of CyclinE after the last cystocyte division, and is required for the proper localization of the proteasome to the fusome. In cdk9DN germaria, it is plausible that a misexpression of encore by the end of the cystsocyte divisions causes the phenotype.
In certain Cdk9DN follicles, we noticed the presence of a large number of small germ cells, which exhibited a round spectrosome and occasionally expressed CycA (Fig. 31F). Such phenotype has been associated with mutation of genes required for germ cell proliferation and differentiation during early oogenesis. Mutations of bam, benig gonial cell neoplasm (bgcn) and fused affect cystoblast determination and lead to an overproliferation of germline stem cells (GSCs) (McKearin and Ohlstein, 1995 ; Lavoie et al., 1999 ; Ohlstein et al., 2000 ; Narbonne-Reveau et al., 2006). Our data suggest that the “tumorous” germ cells observed in certain cdk9DN follicles did not reach the state of cystoblast differentiation as none of these cells expressed the determination factor Bam. Although this latter phenotype was observed at low frequency and only when cdk9DN was expressed at high level (i.e. expression of two cdk9DN transgenes), we suggest that impairing P-TEFb function in the GSCs might occasionally affect their maintenance in the niche. At the tip of the germarium, the GSCs are notably maintained via localized adherent junctions with the cap cells rich in E-cadherin (reviewed by Fuller and Spradling, 2007). The disruption of the E-cadherin-mediated cell adhesion, through mutation of shotgun in the germline, eventually results in the loss of the GSCs from the niche (Song et al., 2002 ; Jin et al., 2008). In our experiments, Cdk9DN could sufficiently reduce the level of E-cadherin expression in certain GSCs to promote their departure. These cells could thus proliferate within the germarium without differentiating as cystoblast, and be incorporated in certain follicles (Fig. 31F).
Distinct transcriptional activity for P-TEFb containing CyclinT or CyclinK.
Our observations that both cyclins can be recruited along with Cdk9 to the same chromosomal locations raised the possibility that P-TEFb complexes containing CyclinT or CyclinK may mediate similar transcriptional activities. However, the inability of ubiquitously expressed CyclinK to rescue cyclinT mutation and reciprocally clearly indicates that the two cyclins have non-redundant activities in vivo (Table 5 and 6). This result may imply that the two P-TEFb complexes regulate distinct steps of the transcription process and both complexes may actually be required for the proper expression of many common targets. It is also possible that both complexes display some redundant activity but that one particular cyclin is required for the timely expression of specific genes at the onset of cell growth or cell differentiation. To determine whether both complexes were able to stimulate transcription elongation, we artificially tethered either complex to sites upstream of a reporter construct in tissue culture cells (Fig. 33 and 34). Our results show that, while Cdk9/CyclinT heterodimer can efficiently stimulate expression of the reporter, the recruitment of Cdk9/CyclinK to enhancer elements had no effects. Our data corroborate with previous studies on human P-TEFb, which showed that T-type cyclins chimeras can stimulate transcription when targeted to various enhancer elements while CyclinK is almost inefficient (Napolitano et al., 2000 ; Taube et al., 2002). Importantly, is was also shown that the C-terminal part of T-type cyclins is necessary for the transactivation of P-TEFb from the enhancer in these assays, and that the C-terminal half of CyclinT1 or CyclinT2 fused to CyclinK confers its ability to stimulate transcription from the enhancer (Taube et al., 2002 ; Lin et al., 2002c). These findings suggest that the inability of P-TEFb containing CyclinK to stimulate transcription is due to the absence of RNAPII-interacting domain within the cyclin, and that the complex has to be recruited differently to the CTD to mediates its effect. This notion is further supported by the observation that direct recruitment of CyclinK to the nascent RNA can well stimulate transcription (Lin et al., 2002c). In our assays, we hypothesise that the binding of CyclinK to UAS sites may not facilitate the interaction of the kinase with its substrate and thus cannot activate transcription. Intriguingly, we noticed that the targeting of inactive P-TEFb composed of Cdk9DN and GBD:CyclinT can still activate transcription although at intermediate levels (Fig. 33). We propose that this effect could happen if the induced Cdk9DN and GBD:CyclinT proteins can makes complexes with endogenous CyclinT and Cdk9. Although S2 cells are unlikely to contain a large excess of endogenous proteins, it is possible that a fraction of
GBD:CyclinT can be assembled with the endogenous wild-type kinase and activate the reporter, while some P-TEFb made of Cdk9DN and endogenous CyclinT could reduce the expression of the renilla luciferase reporter used for standardization. Alternatively, it is also possible that the multiple UAS sites of the reporter allowed binding of both active and inactive complexes and that recruitment of a single active P-TEFb could be sufficient for transcription activation. Another possibility comes from the observations that several intron-less genes such as the human histone (H2a, H3 and H4) and U2 snRNA genes are efficiently transcribed in absence of P-TEFb function and did not showed a strong requirement for Ser-2 phosphorylation of the RNAPII CTD (Medlin et al., 2005). Similarly, P-TEFb inhibition rather affects mRNA 3’ end processing than transcription elongation in expression of the intron-less Drosophila hsp70 gene (Ni et al, 2004 ; Ni et al., 2008). Therefore, although P-TEFb is critically required in early elongation for the majority of protein-coding genes, its function appeared partially dispensable for expression of several small intron-less genes. In our transcription assays, the luciferase reporter construct did not possessed any intron and could be transcribed, although at lower rate, in absence of P-TEFb kinase activity. The lack of Ser-2-phosphorylated polymerases would also result in defective 3’ end processing of the mRNAs and non-polyadenylated RNAs are rapidly degraded. Interestingly, Ni et al. (2004) found that Drosophila hsp26 and to a lesser extent hsp70 mRNAs can still be polyadenylated in P-TEFb inhibited cells, suggesting that other mechanisms than Ser-2 phosphorylation participate in the recruitment of the 3’ end processing machinery and could partially compensate the loss of TEFb activity. This observation also raises the possibility that P-TEFb could stimulate mRNA processing independently of its kinase activity. In this respect, P-TEFb was shown to interact with several splicing factors suggesting that the kinase participates in the coupling of transcriptional elongation with pre-mRNA processing by recruiting these factors to the elongation complex (Fong and Zhou, 2000 ; Bres et al., 2005). It is possible that P-TEFb also facilitates the recruitment of the 3’ end processing complexes through protein interactions, which might explain the intermediate level of luciferase activity reported in our assays.
To date, it is not known how Drosophila P-TEFb is recruited to the CTD. Although Cdk9/CyclinT can stimulate transcription from enhancers, it is probable that other DNA-binding proteins are required for bending of the enhancer towards the elongation complex and to facilitate P-TEFb interaction with the CTD. The inability of Cdk9/CyclinK to activate transcription from UAS sites suggests that additional factors or alternative mechanisms are involved to bring the kinase in close proximity to the CTD. In fission yeast, P-TEFb is
recruited to the nascent transcripts by the cap-methyltransferase enzyme (Guiguen et al., 2007), a mechanism that might facilitate interaction of P-TEFb with the CTD. Similarly, the Tat transactivator stimulates transcription by recruiting P-TEFb to the TAR element of the nascent HIV transcript. We hypothesise that such a mechanism could be used for the recruitment of Drosophila Cdk9/CyclinK complex to active genes. Following RNase treatment of polytene chromosomes however, we did not notice significant changes of CyclinK binding to chromatin suggesting that stable association of Cdk9/CyclinK to chromatin does not require nascent mRNAs, unless these are protected from degradation (Fig.
35). This observation does not exclude however the possibility that capping enzymes may recruit P-TEFb to the polymerase as these proteins can directly interact with the CTD in yeast and mammals.
CyclinT and CyclinK may regulate different cellular processes during Drosophila development.
During the early phase of Drosophila development, both cyclins are maternally deposited in embryos and likely ubiquitously expressed until the end of embryogenesis as indicated by RNA in situ hybridization and microarray time-course analysis (data available at www.fruitfly.org, BDGP expression pattern). Consistent with a requirement for both proteins in developing Drosophila, we found that mutation in either cyclinT or cyclinK results in early larval lethality. Interestingly, we noticed that cyclinK homozygous mutant died mostly during the first instar larvae while the lethality caused by cyclinT mutation was delayed to the second stage and occasionally during the third larval and pupal stages. It is not yet clear however whether the cyclinT mutant allele, generated by P element insertion, is a null allele or a hypomorphic allele. It would be thus informative to compare the viability of cyclinT homozygous mutant to cyclinT trans-heterozygous carrying a deficiency uncovering the cyclinT gene. Although both cyclins are required during embryogenesis, we speculate that cyclinT and cyclinK may be differently expressed in certain Drosophila tissues to regulate different cellular tasks. In mammals, in situ experiments indicated that T-type cyclins and CyclinK are ubiquitously expressed in most tissues. However, it is also apparent that these cyclins are upregulated in selective tissues, which might account for the distinct activity of the proteins. In our experiments, we attempted to identify tissues and/or cellular functions that are more dependent of one or the other P-TEFb complex by co-expressing Cdk9DN with CyclinT
or CyclinK. Intriguingly, while the ubiquitous co-expression of CyclinK with Cdk9DN produced phenotype similar to that of Cdk9DN alone, the co-expression of CyclinT with Cdk9DN rescued the larval growth defect until the end of the third stage. It is thus tempting to speculate that CyclinK function may be critically required for larval growth, perhaps by promoting endoreplication (Fig. 36). However, because the phenotype is not enhanced nor suppressed by respectively increasing or decreasing the dose of wild-type cyclinK (data not shown), we excluded this hypothesis. The other possibility would be that CyclinT, rather than
or CyclinK. Intriguingly, while the ubiquitous co-expression of CyclinK with Cdk9DN produced phenotype similar to that of Cdk9DN alone, the co-expression of CyclinT with Cdk9DN rescued the larval growth defect until the end of the third stage. It is thus tempting to speculate that CyclinK function may be critically required for larval growth, perhaps by promoting endoreplication (Fig. 36). However, because the phenotype is not enhanced nor suppressed by respectively increasing or decreasing the dose of wild-type cyclinK (data not shown), we excluded this hypothesis. The other possibility would be that CyclinT, rather than