Role of human transcription elongation factor DSIF in the suppression of senescence and apoptosis

Authors

  • Toshiharu Komori,

    1. Graduate School of Bioscience and Biotechnology, Tokyo Institute of Technology, 4259 Nagatsuta, Yokohama 226-8501, Japan
    Search for more papers by this author
  • Naoto Inukai,

    1. Graduate School of Bioscience and Biotechnology, Tokyo Institute of Technology, 4259 Nagatsuta, Yokohama 226-8501, Japan
    Search for more papers by this author
    • a

      Present address: Takeda Pharmaceutical Company Limited, Osaka, Japan.

  • Tomoko Yamada,

    1. Graduate School of Bioscience and Biotechnology, Tokyo Institute of Technology, 4259 Nagatsuta, Yokohama 226-8501, Japan
    Search for more papers by this author
    • b

      Present address: Harvard Medical School, Boston, MA, USA.

  • Yuki Yamaguchi,

    1. Graduate School of Bioscience and Biotechnology, Tokyo Institute of Technology, 4259 Nagatsuta, Yokohama 226-8501, Japan
    Search for more papers by this author
  • Hiroshi Handa

    Corresponding author
    1. Graduate School of Bioscience and Biotechnology, Tokyo Institute of Technology, 4259 Nagatsuta, Yokohama 226-8501, Japan
    2. Integrated Research Institute, Tokyo Institute of Technology, 4259 Nagatsuta, Yokohama 226-8503, Japan
    Search for more papers by this author

  • Communicated by: Hiroshi Hamada

* Correspondence: hhanda@bio.titech.ac.jp

Abstract

DSIF is an evolutionarily conserved, ubiquitously expressed, heterodimeric transcription elongation factor composed of two subunits, Spt4 and Spt5. Previous biochemical studies have shown that DSIF positively and negatively regulates RNA polymerase II elongation in collaboration with other protein factors. While several data suggest that DSIF is a ‘general’ elongation factor, there is also evidence that DSIF exerts a tissue- and gene-specific function. Here we sought to address the question of whether physiological functions of DSIF are general or specific, by using a sophisticated knockdown approach and gene expression microarray analysis. We found that Spt5 is essential for cell growth of various human cell lines and that Spt5 knockdown causes senescence and apoptosis. However, Spt5 knockdown affects a surprisingly small number of genes. In Spt5 knockdown cells, the p53 signaling pathway is activated and mediates part of the knockdown-induced transcriptional change, but apoptotic cell death occurs in the absence of p53. Structure-function analysis of Spt5 shows that the C-terminal approximately 300 amino acid residues are not required to support cell proliferation. These results suggest that one of the functions of Spt5 is to suppress senescence and apoptosis, and that this function is exerted through its association with Spt4 and Pol II.

Introduction

Transcription elongation was once considered a monotonic process, but recent studies revealed multiple layers of regulation governing this process (reviewed in Sims et al. 2004; Saunders et al. 2006). DRB sensitivity-inducing factor (DSIF) is an evolutionarily conserved, ubiquitously expressed, heterodimeric transcription elongation factor that plays a pivotal role in transcription elongation by eukaryotic RNA polymerase II (Pol II). SPT4 and SPT5 genes, which encode the subunits of DSIF, were originally isolated in yeast Saccharomyces cerevisiae by a genetic screen for suppressors of the Ty element insertion into promoter regions of certain metabolic genes (Swanson et al. 1991; Swanson & Winston 1992). Subsequently, through an activity-based biochemical approach, human Spt4 and Spt5 were shown to function as a protein complex termed DSIF during transcription elongation (Wada et al. 1998a,b), and strong support for its role in elongation was provided by genetic studies in yeast (Hartzog et al. 1998). More recently, DSIF has been characterized in various other model organisms, such as nematode, fruit fly, and zebrafish (Andrulis et al. 2000; Guo et al. 2000; Kaplan et al. 2000; Keegan et al. 2002; Shim et al. 2002). Remarkably, part of Spt5 also shows a significant sequence similarity with NusG, a transcriptional terminator/anti-terminator found in bacteria and archaea.

A series of biochemical studies have shown that DSIF positively and negatively regulates Pol II elongation in collaboration with other protein factors. Thus, DSIF collaborates with the negative elongation factor NELF to inhibit Pol II elongation at a promoter–proximal region, a phenomenon known as promoter–proximal pausing (Gilmour Lis 1984; Yamaguchi et al. 1999a; Cheng & Price 2007). This inhibition is reversed by the positive elongation factor P-TEFb, which phosphorylates Pol II at the C-terminal domain and thereby facilitates the release of NELF (Wada et al. 1998b; Yamaguchi et al. 1999a). P-TEFb also phosphorylates the Spt5 subunit of DSIF, after which DSIF further activates elongation in collaboration with other protein factors (Ivanov et al. 2000; Yamada et al. 2006; and our unpublished data).

As the name suggests, DSIF was first defined biochemically as a factor mediating the action of the transcriptional inhibitor DRB (Wada et al. 1998a). It turned out that DRB inhibits P-TEFb kinase and thereby enhances a transcriptional block caused by DSIF and NELF. As mRNA synthesis by Pol II is generally suppressed when cells are exposed to DRB (Sehgal et al. 1976), repression/anti-repression by these elongation factors may be viewed as a general checkpoint during the transcription cycle. In line with this idea, immunostaining of Drosophila polytene chromosomes indicated that DSIF is associated with numerous protein-coding genes with a pattern similar to that of Pol II (Andrulis et al. 2000; Kaplan et al. 2000; Wu et al. 2003). Moreover, high resolution chromatin immunoprecipitation (ChIP) analysis of known DSIF target genes, such as Drosophila hsp70 and human junB, showed that in an uninduced state, Pol II and DSIF are localized to the promoter–proximal region, whereas following induction, their distribution is extended to the coding and 3′ regions, the observation that fits nicely with earlier biochemical data (Andrulis et al. 2000; Aida et al. 2006). Collectively, these findings are consistent with the idea that DSIF is a ‘general’ transcription elongation factor.

However, there is also evidence that DSIF exerts a tissue- and gene-specific function. In zebrafish, a certain point mutation in the SPT5 gene causes very specific defects in neuronal differentiation, including a decrease in the number of dorpaminergic neurons and a concomitant increase of serotonergic neurons in hypothalamus. (Guo et al. 2000). These phenotypes appear to be allele-specific, as a subsequently study showed that SPT5-null mutant zebrafish displays more severe and extensive defects (Keegan et al. 2002). Nevertheless, these findings raise the question of whether physiological functions of DSIF are general or specific. In this study, we sought to address this issue by using a sophisticated knockdown approach and gene expression microarray analysis.

Results

Spt5 is essential for HeLa cell growth

Previously, we described inducible knockdown of Spt5 in HeLa cells (Yamada et al. 2006). In this system, HeLa cells expressing Flag-Spt5 under the control of a tetracycline-repressible promoter (HeLa/Flag-Spt5(RNAir) cells) were used to transduce lentiviral vectors expressing shRNA against Spt5. As the Flag-Spt5 sequence contains silent mutations within the shRNA target sequence, the shRNA causes selective depletion of endogenous Spt5 mRNA. Hence, in the engineered cell line, the intracellular pool of Spt5 is provided almost entirely by exogenous Flag-Spt5, and the remaining Spt5 protein can be depleted by simply adding tetracycline to culture media. Another advantage of this strategy is that, compared to direct down-regulation by siRNA, tetracycline-mediated knockdown is not accompanied by potential off-target effects of siRNA.

Here, we first examined knockdown efficiencies at different time points after tetracycline addition. Expression of Flag-Spt5 was significantly reduced on day 1, and reduced to approximately 5% of the original level on day 3 (Fig. 1A). In the cells in which endogenous Spt5 was not knocked down, the protein level of total Spt5 was only partially reduced by tetracycline addition (Fig. 1A, left panels). In the cells in which endogenous Spt5 was knocked down, however, the level of total Spt5 was strongly reduced as well as that of Flag-Spt5 following tetracycline addition (Fig. 1A, right panels). Under these conditions, the control cells proliferated actively regardless of tetracycline addition (Fig. 1A, left panels). As for the cells in which endogenous Spt5 was knocked down, they proliferated as well as the control cells in the absence of tetracycline, suggesting that Flag-Spt5 successfully complemented the deficiency of endogenous Spt5. In the presence of tetracycline, however, growth retardation was apparent on day 5, and the cells eventually died completely (Fig. 1A, right panels). These data clearly demonstrate that Spt5 is essential for cell growth at the single-cell level.

Figure 1.

Spt5 is essential for HeLa cell growth. (A) Control virus (left) or RNAi virus (right) was transduced into HeLa/Flag-Spt5(RNAir) cells. These cells were cultured for several days in the absence or presence of 2 µg/mL tetracycline and harvested for cell counting and immunoblotting with anti-Flag (Flag-Spt5), anti-Spt5 (total Spt5), and anti-PSF antibodies. PSF serves as loading control. (B) RNAi virus-transduced HeLa/Flag-Spt5(RNAir) cells were analyzed as in (A) except that various concentrations of tetracycline were used. (C) RNAi virus-transduced HeLa/Flag-Spt5 (RNAir) cells were analyzed as in (A) except that tetracycline was removed at the indicated times.

To determine the Spt5 protein level that is required to support cell growth, we examined Spt5 expression and cell numbers at various concentrations of tetracycline (Fig. 1B). Whereas the tetracycline concentration used in Fig. 1A was 2 µg/mL, tetracycline caused a similar inhibition of both Spt5 expression and cell growth at as low as 16 ng/mL. At 3 ng/mL, however, Spt5 expression and cell growth were not affected appreciably (Fig. 1B). It was difficult to reproducibly reduce Spt5 expression to an intermediate level at tetracycline concentrations in between (data not shown). Nevertheless, these data show a good correlation between Spt5 expression and cell growth.

Next, we were interested in knowing the effect of transient knockdown of Spt5 on cell growth. When cells were treated with tetracycline for up to 2 days and then cultured in tetracycline-free medium, Spt5 expression was transiently reduced (Fig. 1C, lanes 1–19), but cells proliferated normally during and after tetracycline treatment (data not shown). In contrast, when cells were treated with tetracycline for 3 days, inhibition of Spt5 expression was more persistent (Fig. 1C, lanes 20–23), and cells underwent cell death before the recovery of Spt5 expression with a similar time course to the cell death observed in the continuous presence of tetracycline (data not shown). These data illustrate that HeLa cells are unable to proliferate when Spt5 expression is reduced below a certain level for a certain period of time.

Knockdown of Spt5 causes G1 arrest and apoptotic cell death

We then examined the phenotypes associated with Spt5 knockdown-induced cell death more carefully. Cell cycle analysis revealed that Spt5 depletion causes G1 arrest. As shown in Fig. 2A, tetracycline addition resulted in a gradual decrease in the populations corresponding to S and G2/M phases (P4 and P5) and a concomitant increase in the population corresponding to G1 phase (P3). Then we examined whether the observed cell death is as a result of apoptosis. Microscopic observation revealed that cells undergoing cell death shows blebbing of plasma membranes and fragmentation of nuclei, which are hallmarks of apoptosis (Fig. 2B). In addition, activation of caspases 3 and 7 were indicated from cleavage of the nuclear protein PARP (Fig. 2C). From these data, we concluded that Spt5 depletion causes G1 arrest and apoptotic cell death.

Figure 2.

Knockdown of Spt5 causes G1 arrest and apoptotic cell death. (A) Following the indicated days of tetracycline treatment, RNAi virus- or control virus-transduced HeLa/Flag-Spt5(RNAir) cells were harvested and stained with propidium iodide for cell cycle analysis. The percent of cells present in each phase of the cell cycle was calculated and shown to the right. P3, P4 and P5 correspond to G1, S and G2/M, respectively. (B) Following 6 days of tetracycline treatment, RNAi virus- or control virus-transduced HeLa/Flag-Spt5(RNAir) cells were stained with DAPI and examined under a fluorescence microscope. PC, phase contrast. (C) Following the indicated days of tetracycline treatment, RNAi virus-transduced HeLa/Flag-Spt5(RNAir) cells were harvested and immunoblotted with anti-PARP antibody. Treatment with anti-Fas antibody was used as positive control (PC).

Genome-wide profiling of Spt5 target genes

Although many DSIF target genes have already been identified, to our knowledge, any genome-wide expression study on DSIF target genes has not been carried out. To understand more about in vivo targets of DSIF and to gain insight into the underlying cause of apoptosis induction, we investigated genome-wide transcriptional changes before the onset of apoptotic cell death. RNAs were isolated in triplicate from the HeLa derivative cells that were treated with tetracycline for 0, 2 or 4 days, and the resulting RNAs were subjected to microarray analysis using the Affymetrix Human Genome Focus Array, which represents over 8500 well-annotated genes (Fig. 3). On day 4, depletion of Spt5 significantly altered the levels of 65 transcripts; 34 transcripts were down-regulated, and 31 transcripts were up-regulated. Not surprisingly, Spt5 was the most strongly down-regulated gene and was the only gene whose expression was significantly altered on day 2. The rest of the genes were concomitantly up-regulated or down-regulated between days 2 and 4, suggesting that these genes are regulated either directly or indirectly by DSIF. Considering the similar time course changes, we speculate that many of the 64 genes are direct targets of DSIF.

Figure 3.

Cluster analysis of transcripts significantly affected by Spt5 depletion. Triplicate samples for each condition were averaged, and pair-wise comparisons were carried out. Shown are 65 transcripts that changed > 1.5-fold by > 40 units in day 4 samples. Red indicates an increase in RNA levels, whereas green indicates a decrease. Yellow rectangles in the gene list indicate the following classifications: GO-Biological Process, GO:0006350, transcription (a); GO-Biological Process, GO:0006334, nucleosome assembly (b); GO-Molecular Function, GO:0003723, RNA binding (c); GO-Biological Process, GO:0007049, cell cycle (d); and KEGG-Pathway, has04115, p53 signaling pathway (e). A query of the Gene Ontology Database with a list of the 65 genes reveals several classes that are significantly over-represented in the gene list. The number of genes in each Gene Ontology class and the associated P-values are shown.

To validate our microarray data, expression levels of 10 putative Spt5 target genes were determined by quantitative real-time RT-PCR (Fig. 4A). For all the genes examined, the direction and magnitude of fold change in expression levels following DSIF depletion agreed well between microarray and real-time RT-PCR analyses. For some genes, such as CDKN1A and CCNA2, statistically significant changes in the mRNA levels were observed on day 2. This observation is probably because real-time RT-PCR is more sensitive for detecting small changes in gene expression than microarray analysis. It is also noteworthy that some of the up-regulated transcripts appear to peak approximately 4–5 days after tetracycline addition, whereas others remain high up to 6 days. This may be because of different roles played by Spt5 or simply due to different stability of the mRNAs.

Figure 4.

p53-dependent and -independent transcriptional changes induced by Sp5 depletion. (A) Expression levels of 10 putative Spt5 target genes were determined by real-time RT-PCR. Data represent means ± SEM from three experiments. Statistical significance at each time point (relative to the basal level) was determined by two sample t-test. *P < 0.05; **P < 0.01. (B) HeLa, IMR-90, and NCI-H1299 cells were infected with the RNAi virus, cultured for the indicated times, and harvested for immunoblotting. (C) MCF7 cells were infected with control or RNAi viruses, cultured for 4 days, and harvested for real-time RT-PCR analysis.

Gene Ontology (GO) classification of Spt5 target genes revealed a significant enrichment in genes whose products are involved in RNA processing, and remarkably, most of the suspected or proven RNA processing factors in the gene list (denoted by ‘c’ ) were down-regulated rather than up-regulated after Spt5 depletion (Fig. 3). In addition, several cell cycle regulators (denoted by ‘d’ ) were found to be down-regulated, four histone genes (denoted by ‘b’ ) up-regulated, and many transcriptional regulators (denoted by ‘a’ ) up-regulated or down-regulated. Furthermore, the levels of some p53-inducible transcripts such as CDNK1A/p21 and DDB2 were found to increase in Spt5-depleted cells (denoted by ‘e’).

Activation of the p53 signaling pathway following Spt5 depletion

The above findings prompted us to investigate the possible role of the p53 signaling pathway in transcriptional change and apoptosis associated with Spt5 depletion. As activation of the p53 signaling pathway is often achieved through stabilization of the p53 protein, the protein level was examined by immunoblotting. In cervical carcinoma HeLa cells, wild-type p53 is expressed, but the protein level is normally quite low as a result of its enhanced ubiquitination by the human papilloma virus E6 protein. We found that Spt5 depletion resulted in a significant elevation of the p53 protein level and a concomitant increase in the CDNK1A/p21 protein level (Fig. 4B). As the p53 mRNA level did not change substantially in this period (data not shown), it is very likely that p53 induction occurs by stabilization of the p53 protein.

To examine the role of p53 induction in the Spt5 knockdown-induced transcriptional change, we used breast adenocarcinoma MCF7 cells, in which p53 is expressed at a normal level. Here, Spt5 and p53 were knocked down either individually or in combination using lentiviral vectors, and RNAs isolated from the resulting cells were analyzed by real-time RT-PCR (Fig. 4C). As expected, Spt5 shRNA had little effect on the p53 mRNA level, whereas p53 shRNA efficiently reduced the mRNA level. Consistent with the findings in HeLa cells, DDB2, CDNK1A/p21 and MAP1LC3B transcripts were all induced by Spt5 knockdown in MCF7 cells. Of these transcripts, induction of DDB2 and CDNK1A/p21 was abrogated by depletion of p53, while induction of MAP1LC3B was observed in the absence of detectable levels of p53. These results illustrate that some of the transcripts induced by Spt5 knockdown are p53-dependent, whereas the others are not.

To confirm and extend the above findings, we used p53 wild-type and null cell lines, IMR-90 and NCI-H1299 (Fig. 4B). In agreement with the above findings, Spt5 knockdown resulted in an increase of the protein levels of p53 and CDNK1A/p21 in IMR-90 cells, but not in NCI-H1299 cells. Moreover, Spt5 knockdown resulted in cell death in both cell types with a similar time course to that observed in HeLa cells (data not shown). These results suggest that although p53 constitutes an important part of the knockdown-induced transcriptional change, apoptotic cell death occurs in the absence of p53.

C-terminal region of Spt5 is not required to support cell proliferation

Given that p53 is not the cause of knockdown-induced cell death, we were interested in knowing the actual cause of cell death. To provide a clue to this issue, we carried out structure-function analysis of Spt5 by introducing a slight modification to our knockdown-rescue system. Thus, Spt5 shRNA and a rescue construct were coexpressed in HeLa cells using two lentiviral vectors carrying different selection markers. Spt5 has a modular structure as shown in Fig. 5A, and previous studies revealed, to some extent, functional significance of these motifs (Yamaguchi et al. 1999b; Yamada et al. 2006). Thus, we prepared a series of deletion mutants, some of which were previously characterized in vitro, and asked whether they are capable of supporting cell proliferation instead of endogenous wild-type Spt5.

Figure 5.

Expression of a series of Spt5 deletion mutants in HeLa cells. (A) Schematic drawings of the domain structure of Spt5 and the Spt5 mutants used in this study. Ability of each mutant to support cell proliferation, deduced from the data shown in Fig. 6, is summarized (right). (B) HeLa cells expressing wild-type or mutant versions of Flag-Spt5 were stained with anti-Flag and Alexa 488-labeled secondary antibodies, and counterstained with DAPI. Shown to the right is a putative nuclear localization signal of Spt5.

We started by examining subcellular localizations of these mutants (Fig. 5B). As a result, wild-type Spt5 and all the mutants except Δacidic were found in the nucleus, whereas Δacidic was largely cytoplasmic, suggesting that a functional nuclear localization signal exists within the first 174 amino acids. In fact, sequence analysis revealed the presence of a putative nuclear localization signal starting at amino acid residue 66.

Complementation assays were then carried out. Expression levels of different Flag-Spt5 constructs were adjusted so that they were similar to each other and to the normal level of endogenous Spt5 (Fig. 6A). Figure 6A also shows that endogenous Spt5 was efficiently knocked down when RNAi virus was transduced. Under these conditions, growth curves of each cell type were determined. When endogenous Spt5 was expressed, all the cell types proliferated at similar rates, indicating that these Spt5 mutants do not show dominant-negative effects under the conditions used (Fig. 6B, left panel). When endogenous Spt5 was knocked down, however, only cell types expressing Flag-Spt5 wild-type, Δrepeat, or ΔCT, proliferated exponentially, and the rest of the cell types underwent growth arrest and cell death (Fig. 6B, right panel). These results suggest that the most C-terminal approximately 300 amino acid residues are not required to support cell proliferation.

Figure 6.

The C-terminal region of Spt5 is not required to support cell proliferation. (A) Cells that were transduced with one of the indicated Flag-Spt5 constructs and either control or RNAi virus were subjected to immunoblotting. Note that Flag-Spt5 ΔCT is not detected by anti-Spt5 blot, because it lacks the epitope. (B) HeLa cells expressing wild-type or mutant Flag-Spt5 were transduced with either control or RNAi virus on day 0, reseeded on day 2, and harvested for cell counting on days 3, 5, 7 and 9.

Discussion

Identification of a set of candidate DSIF target genes

Growing evidence suggests that DSIF plays a general role in the regulation of Pol II transcription. In addition to the previous studies mentioned in Introduction, recent data also support this view. For example, ChIP analysis of 11 randomly selected protein-coding genes showed that DSIF is associated with most of these genes with a pattern similar to that of Pol II (our unpublished data). ChIP–chip analysis also showed that DSIF is broadly distributed among protein-coding genes (our unpublished data). In clear contrast, genome-wide transcriptional profiling revealed that Spt5 knockdown leads to significant changes in 64 transcripts, less than 1% of the approximately 8500 transcripts examined (Fig. 3). Even considering the presence of false negatives that may have been overlooked for various reasons, the number of the affected genes is considerably smaller than that expected from the above assumption.

We offer three possible explanations for this discrepancy. First, residual Spt5 after knockdown may have weakened the resulting phenotypes and prevented the identification of a full spectrum of DSIF target genes. Although such a possibility does exit, it seems unlikely to us because under the conditions employed, Spt5 knockdown caused such a severe phenotype as cell death.

Second, DSIF may be partially redundant with other factors, and as such, only a limited transcriptional effect may have been seen after Spt5 knockdown. So far, over a dozen transcription elongation factors have been identified, and many of them have been shown to bind Pol II and enhance the elongation step in vitro (Sims et al. 2004). However, their physiological functions remain largely unknown, and whether they have distinct or redundant functions also remains unclear. It is therefore possible that the loss of DSIF is partially rescued by the presence of other functionally similar elongation factors.

Third, elongation defects may not necessarily lead to alterations in mRNA levels. mRNA synthesis is controlled at various steps, including chromatin remodeling, transcription initiation, elongation, termination, and mRNA processing, and therefore, transcription elongation may be a rate-limiting step for only a subset of genes. One can imagine a situation in which every step of mRNA synthesis is critical, that is, exceptionally high-level mRNA synthesis. On Drosophila hsp70 gene, for example, Pol II initiates transcription at a rate of once every few seconds during heat shock, resulting in a train of elongation complexes traversing the gene (Wu et al. 2003). Expression of such genes is expected to be sensitive to the reduction of elongation factors, and indeed, many of the so-called immediate-early genes that undergo transient high-level expression, such as hsp70, c-fos, and junB, are known targets of DSIF (Wu et al. 2003; Aida et al. 2006; Yamada et al. 2006). It is not surprising that these genes were not identified by our microarray analysis, because the current analysis was carried out without any extracellular stimuli. Considering such genes whose expression depends on DSIF only under certain circumstances, the total number of DSIF target genes would be greater than that shown here.

Mechanisms underlying knockdown-induced transcriptional changes

We speculate that knockdown-induced transcriptional changes may be classified into the following categories: (i) direct targets of DSIF, (ii) transcriptional changes that are associated with knockdown-induced physiological changes, and (iii) cellular actions to confront abnormal changes induced by knockdown. The second and third types look similar, but we view them as opposite cellular responses to knockdown-induced phenotypic changes.

As for the first category, because DSIF is able to both activate and repress transcription elongation, both up-regulated and down-regulated genes could be direct targets of DSIF. Among the gene list, the presence of histone genes is intriguing because it reminds us of our recent findings on NELF, an elongation factor that works together with DSIF (Narita et al. 2007). Replication-dependent histone genes normally produce mRNAs lacking poly(A) tail, and this mechanism is partly responsible for S phase-specific expression of histone genes. We previously found that knockdown of NELF causes a defect in histone-specific 3′ processing and leads to the synthesis of abnormal polyadenylated histone transcripts. The current microarray data suggest that Spt5 knockdown also causes a similar defect and that DSIF and NELF work together to facilitate histone-specific 3′ processing. Note that only polyadenylated transcripts are amplified and quantified by Affymetrix microarray analysis.

Some genes in the list may be classified into the second category. For example, it is reasonable to assume that p53 target genes such as CDKN1A/p21 and DDB2 were up-regulated because of p53 activation by Spt5 knockdown, without direct transcriptional regulation by Spt5. Besides, the cyclin A2 gene CCNA2, whose expression was down-regulated by Spt5 knockdown, may also belong to this category. As CCNA2 expression is high in S and G2 phases and low in G1 phase (Whitfield et al. 2002), its reduced expression is most likely because of G1 arrest associated with Spt5 knockdown.

Several genes implicated in mRNA metabolism, such as mRNA transcription and processing, are found in the gene list. We speculate that some of these genes may be classified into the third category. For instance, expression of the mRNA export factor NXF1 was increased by Spt5 knockdown. This may be accounted for as a cellular reaction to compensate for knockdown-induced defects in mRNA synthesis and possibly export. Coincidently, it is reported that a yeast mutant strain lacking Spt4, the non-essential subunit of DSIF, is defective in mRNA export (Burckin et al. 2005).

Actual cause of Spt5 knockdown-induced senescence and apoptosis

What is the actual cause of Spt5 knockdown-induced senescence and apoptosis? We are interested in the observation that, whereas the protein level of Spt5 is significantly reduced on day 2 or 3 following knockdown, apoptosis becomes evident on day 5. This contrasts with the observation that, when Rpb1, an integral subunit of Pol II, is knocked down, a reduction in the protein level is immediately followed by cell death (our unpublished data). As a reason for the timing difference, we speculate that stochastic or cumulative events mediate the knockdown-induced senescence and apoptosis.

A few findings that may be relevant to this issue have been obtained from genetic studies in yeast; it has been shown that genetic mutations of Spt4 or Spt6, an elongation factor playing a similar role, affect genomic stability and result in a hyper-recombination phenotype (Basrai et al. 1996; Malagon & Aguilera 1996). The underlying cause may be the formation of R loops. An R loop is a structure in which an RNA molecule is partially hybridized with one strand of a double-stranded DNA, leaving the other strand unpaired. Whereas the presence of an 8 to 9-bp DNA : RNA hybrid in the transcription bubble is well-documented, extensive formation of R loops is also observed during transcription under certain conditions, for example, when a component of the Pol II elongation machinery is defective (reviewed in Li & Manley 2006). Transcriptional R looping induces genomic instability, possibly because single-stranded DNA is more vulnerable to mutations than double-stranded DNA (reviewed in Li & Manley 2006). Thus, one of the functions of Spt5 may be to protect chromosomes from damage by preventing R loop formation during Pol II elongation. This idea is consistent with the observation that Spt5 knockdown resulted in the activation of the p53 signaling pathway.

It was found that the most C-terminal approximately 300 amino acid residues of Spt5 are not required to support cell proliferation. However, Δacidic, ΔSpt4BD, ΔKOW1/2 and ΔKOW3/4 could not rescue the deficiency of wild-type Spt5. The failure of Δacidic in complementation is probably because of its mislocalization. Moreover, the results of ΔSpt4BD, ΔKOW1/2 and ΔKOW3/4 suggest that Spt4-binding and Pol II-binding activities of Spt5 are required to support cell proliferation. Dispensability of the C-terminal approximately 300 amino acids was somewhat surprising to us because we previously showed that the C-terminal segment contains a repetitive structure that is phosphorylated by P-TEFb and is critical for elongation activation activity (Yamada et al. 2006). It may be that the N-terminal two-thirds of Spt5 exert a core function through its association with Spt4 and Pol II, whereas the C-terminal segment plays a regulatory role that is dispensable under certain conditions in vivo.

Despite this conjecture, this work will serve as an important foundation for further understanding of the cell biology of DSIF and elongation control.

Experimental procedures

Immunological analyses

The following primary antibodies were used for immunoblotting: anti-Flag (Sigma, F3165), anti-topoisomerase I (Santa Cruz, sc-10783), anti-PARP (Affinity BioReagents, MA3–950), anti-p53 (Oncogene, OP03), anti-p21 (PharMingen, 65951A), anti-actin (Chemicon, MAB1501) and anti-Spt5 (Wada et al. 1998a). Samples for immunoblotting were prepared using High-salt lysis buffer (50 mm Tris–HCl, pH 8.0, 500 mm NaCl, 1% NP-40). For fluorescence microscopy, formaldehyde-fixed cells were stained using anti-Flag and Alexa Fluor 488-conjugated anti-mouse IgG (Molecular Probes, A11029) as primary and secondary antibodies, respectively. Subsequently, cells were counterstained with 4′,6-diamidino-2-phenylindole (DAPI) and observed under a BX51 fluorescence microscope (Olympus) equipped with an ORCA-ER digital camera (Hamamatsu).

General cell culture techniques

HeLa, MCF7, and 293FT cells were cultured in DMEM supplemented with 10% FCS. IMR-90 and NCI-H1299 cells were obtained from ATCC and cultured according to the distributor's instruction. Counting of cell numbers was carried out essentially as described (Inukai et al. 2004), except for the data shown in Fig. 6. For cell cycle analysis, cells were stained with propidium iodide and analyzed using a FACSCalibur flow cytometer and the Cell Quest software (Becton Dickinson). Data were gated using pulse width and pulse area to exclude debris and doublets, and the percent of cells present in each phase of the cell cycle was calculated.

Knockdown and rescue experiments

Knockdown experiments were carried out essentially as described (Yamada et al. 2006). shRNA target sequences for Spt5 and p53 are 5′-gaactgggcgagtattacatt-3′ and 5′-gactccagtggtaatctactt-3′. Briefly, recombinant lentiviruses used for knockdown experiments were produced by cotransfection of 293FT cells with ViraPower packaging mix (Invitrogen) and pLenti-U6-hSpt5-RNAi#1 or pLenti-U6-control. In some experiments, such as those shown in Fig. 4B, knockdown was carried out by directly infecting recombinant lentiviruses into cells of interest. In other experiments, the RNAi virus was infected into HeLa/Flag-Spt5(RNAir) cells and selected in the presence of 4 µg/mL blasticidin. In the resulting cells, which we called F-WT cells in our previous study (Yamada et al. 2006), endogenous Spt5 was knocked down, and RNAi-resistant Flag-Spt5 was expressed under the control of a tetracycline-repressible promoter.

For rescue experiments, HeLa cells were first infected with one of the lentiviral vectors expressing wild-type or mutant versions of Flag-Spt5 and selected for 2 weeks in the presence of 75 µg/mL zeocin. The resulting cells were infected with either control or RNAi virus (day 0), and on day 2, 5 × 104 cells were seeded and cultured in the presence of 1 µg/mL blasticidin. On days 3, 5, 7 and 9, cell numbers were counted using a hemocytometer.

Rescue constructs were prepared such that various Spt5-derived fragments were inserted into pLenti4/V5-GW/lacZ (Invitrogen) in place of its lacZ sequence. For construction of pLenti4-Flag-Spt5(RNAir) WT, Flag-Spt5 sequence containing silent mutations that confer RNAi resistance was obtained from the pcDNA3 version (Yamada et al. 2006) and subcloned into pLenti4. For Δacidic(lacking 1–175 amino acids), ΔSpt4BD (lacking 176–313 amino acids), ΔKOW1/2 (lacking 314–516 amino acids), and Δrepeat (lacking 758–936 amino acids), fragments encoding these deletion mutants were obtained from respective pET-14b-based vectors (Yamaguchi et al. 1999b) and used to replace part of pcDNA3-Flag-Spt5(RNAir). The inserts of the resulting plasmids (containing an N-terminal Flag-tag and the silent mutations) were then subcloned into pLenti4. For ΔKOW3/4 (lacking 517–758 amino acids), an internal SmaI–SmaI fragment was removed from pET-Spt5 WT, the resulting DNA was religated in frame, and the insert sequence was subcloned into pLenti4. ΔCT (lacking 937–1087 amino acids) was prepared by PCR using appropriate mutagenic primers.

Microarray and RNA analyses

RNA samples were prepared in triplicate and subjected to microarray analysis using the Affymetrix Human Genome Focus Array (Affymetrix) according to the manufacture's instructions. For cluster analysis, triplicate samples were averaged, pair-wise comparisons were carried out using the dChip software (Li & Wong 2001), and transcripts that changed > 1.5-fold by > 40 units in day 4 samples were chosen. Gene Ontology analysis was carried out using the L2L tool (Newman & Weiner 2005).

Real-time RT-PCR analysis was carried out using total RNAs and QuantiTect SYBR Green RT-PCR master mix (Qiagen). Primer sets used in this study are shown in Fig. S1 in Supporting Information.

Acknowledgements

Authors thank S. Okamoto and J. Kato for technical assistance. This work was supported by Special Coordination Funds for Promoting Science and Technology from the Japan Science and Technology Agency (to H.H.) and by a Grant-in-Aid for Scientific Research on Priority Areas from MEXT (to Y.Y.). This work was also supported in part by a Grant of the Global COE Program from MEXT.

Ancillary