In Kwon Chung, Department of Biology, College of Life Science and Biotechnology, Yonsei University, 134 Shinchon-dong, Seoul 120-749, Korea. Tel.: 82 2 2123 2660; fax: 82 2 364 8660; e-mail: firstname.lastname@example.org
Human chromosome ends associate with shelterin, a six-protein complex that protects telomeric DNA from being recognized as sites of DNA damage. The shelterin subunit TRF2 has been implicated in the protection of chromosome ends by facilitating their organization into the protective capping structure and by associating with several accessory proteins involved in various DNA transactions. Here we describe the characterization of DDX39 DEAD-box RNA helicase as a novel TRF2-interacting protein. DDX39 directly interacts with the telomeric repeat binding factor homology domain of TRF2 via the FXLXP motif (where X is any amino acid). DDX39 is also found in association with catalytically competent telomerase in cell lysates through an interaction with hTERT but has no effect on telomerase activity. Whereas overexpression of DDX39 in telomerase-positive human cancer cells led to progressive telomere elongation, depletion of endogenous DDX39 by small hairpin RNA (shRNA) resulted in telomere shortening. Furthermore, depletion of DDX39 induced DNA-damage response foci at internal genome as well as telomeres as evidenced by telomere dysfunction-induced foci. Some of the metaphase chromosomes showed no telomeric signal at chromatid ends, suggesting an aberrant telomere structure. Our findings suggest that DDX39, in addition to its role in mRNA splicing and nuclear export, is required for global genome integrity as well as telomere protection and represents a new pathway for telomere maintenance by modulating telomere length homeostasis.
Telomeres, the specialized nucleoprotein complexes at the ends of eukaryotic chromosomes, are essential for the maintenance of chromosome integrity (Smogorzewska & de Lange, 2004; Palm & de Lange, 2008). Mammalian telomeric DNA consists of long tracts of duplex telomere repeats with 3′ single-stranded G overhang and can form complex higher-order structures (such as a t-loop) that provide telomere protection by hiding chromosome ends from being recognized as sites of DNA damage (Greider, 1999; Griffith et al., 1999). Loss of telomere function results in chromosome end fusions, degradation, and other inappropriate reactions, leading to cell cycle arrest, senescence, and apoptosis (Blasco et al., 1997). In the absence of a telomere maintenance pathway, dividing somatic cells show a progressive loss of telomeric DNA during successive rounds of cell division because of a DNA end replication problem (Lingner et al., 1995; Cerone et al., 2005). Although recombination-mediated telomere elongation has been demonstrated for replenishing telomeric DNA (Bryan et al., 1997; Dunham et al., 2000), the major mechanism to counteract telomere erosion is based on telomerase (Autexier & Lue, 2006; Bianchi & Shore, 2008). In humans, telomerase is strongly repressed in normal somatic tissues but expressed in most cancer cells, suggesting that the activation of telomerase is a critical step in human carcinogenesis (Kim et al., 1994; Bodnar et al., 1998; Hahn et al., 1999).
TRF2 is required to protect chromosome ends by facilitating their organization into the t-loop structure (Stansel et al., 2001; Verdun & Karlseder, 2007). Inhibition of TRF2 function induces a DNA damage response at telomeres. As a result, DNA damage response factors such as 53BP1, MDC1, γ-H2AX, ATM, and the Mre11/Nbs1/Rad50 complex accumulate in telomere dysfunction-induced foci (TIFs) (Takai et al., 2003). Damaged telomeres can activate the ATM kinase signaling, leading to cell cycle arrest mediated by the p53/p21 pathway (Karlseder et al., 1999; Dd’Adda di Fagagna et al., 2003; Guo et al., 2007). The DNA damage signal elicited by TRF2 inhibition is completely abrogated when ATM is absent, suggesting that TRF2 represses ATM signaling pathway (Karlseder et al., 2004; Denchi & de Lange, 2007). Damaged telomeres are subsequently processed by DNA repair pathways, resulting in chromosome end fusions (Dimitrova & de Lange, 2006; Konishi & de Lange, 2008). TRF2 also protects chromosome ends by recruiting the shelterin accessory proteins to telomeres. Apollo, an Artemis-related nuclease, has the ability to localize to telomeres through an interaction with TRF2 (Lenain et al., 2006; van Overbeek & de Lange, 2006). Depletion of Apollo resulted in the activation of a DNA-damage signal at telomeres as evidenced by TIFs. The TIFs occurred in S phase, suggesting that Apollo is a shelterin accessory factor required for complete replication of chromosome ends. Recently, using oriented peptide libraries, PNUTS and MCPH1 have been identified as telomere-associated proteins that directly interact with the telomeric repeat binding factor homology (TRFH) domain of TRF2 (Kim et al., 2009). PNUTS and MCPH1 were shown to regulate telomere length and the telomeric DNA-damage response, respectively.
In a search for proteins capable of interacting with TRF2, we identified DDX39 as an interacting partner in a yeast two-hybrid screen. DDX39 was identified as a novel member of the DEAD-box RNA helicases and was shown to be expressed in a growth-regulated manner (Sugiura et al., 2007a,b). DDX39 possesses a high degree of homology to UAP56 (also known as Sub2p) which plays important roles in the splicing reaction as well as in nuclear export of mature mRNA (Pryor et al., 2004). The DEAD-box proteins are highly conserved in a wide spectrum of species and contain the characteristic motifs that act as a RNA helicase (Lüking et al., 1998; Tanner & Linder, 2001). Overexpression of DDX39 accelerates cell proliferation of HeLa cells, suggesting the biological significance of DDX39 in cancer pathogenesis (Sugiura et al., 2007a). Here we report the characterization of DDX39 as a TRF2-binding protein. DDX39 interacts with TRF2 through the TRFH domain in vitro and in vivo. DDX39 also associates with catalytically competent telomerase through an interaction with hTERT. Whereas overexpression of DDX39 led to progressive telomere elongation, depletion of DDX39 resulted in telomere shortening. Furthermore, depletion of DDX39 induced DNA-damage response foci at both internal genome and telomeres. Some of the metaphase chromosomes showed no telomeric signal at chromatid ends. Overall, these results suggest that DDX39, in addition to requirement for global genome integrity, also has a unique and independent role in telomere protection.
Identification of DDX39 as a TRF2-interacting factor
To identify TRF2-interacting factors, we screened a HeLa cell cDNA library using the yeast two-hybrid system. With the full-length TRF2 as bait, two independent clones containing the DDX39 cDNA were obtained and sequenced. Both clones lacked the N-terminal region of the full-length protein. We confirmed that the full-length (DDX39FL) and N-terminal truncation mutant of DDX39 (DDX39ΔN, amino acid residues 161–427) interact with TRF2 but not with TRF1 (Fig. 1A). RAP1, which was known to interact with TRF2, was used as a positive control (Li et al., 2000). DDX39 was identified as a novel member of the DEAD-box RNA helicases (Sugiura et al., 2007a,b). DDX39 is upregulated in lung squamous cell carcinoma and promotes cancer cell growth, suggesting a cancer-associated RNA helicase (Sugiura et al., 2007a).
To confirm the direct interaction between TRF2 and DDX39, we performed GST pull-down experiments. GST-TRF2, but not GST-TRF1 or the control GST, precipitated DDX39-V5 expressed in HT1080 cells, indicating that TRF2 interacts with DDX39 in vitro (Fig. 1B). To determine whether TRF2 and DDX39 associate in vivo, HT1080 cells were cotransfected with Flag-TRF2 (or Flag-TRF1) and DDX39-V5 expression vectors, and subjected to immunoprecipitation. Flag-TRF2, but not Flag-TRF1, was detected in anti-V5 immunoprecipitates (Fig. 1C). Conversely, DDX39-V5 was detected in anti-Flag immunoprecipitates when Flag-TRF2 was expressed, but not in anti-Flag immunoprecipitates when Flag-TRF1 was expressed (Fig. 1C). Endogenous DDX39 was immunoprecipitated with endogenous TRF2 in HT1080 cells, but not with endogenous TRF1 (Fig. 1D). Endogenous TRF2 was also recovered in anti-DDX39 immunoprecipitates, indicating that DDX39 interacts with TRF2 in mammalian cells.
To examine whether DDX39 can associate with telomeres, we determined its subcellular localization by indirect immunofluorescence. Endogenous DDX39 was clearly localized in the nucleus, and about 28% of the DDX39 foci costained with the endogenous TRF2 foci and colocalized with telomere signals as indicated by telomere fluorescence in situ hybridization (FISH) (Fig. 1E,F). DDX39 was also found in many foci that did not colocalize with TRF2 signals. Although the nature of these localization sites was not determined, this observation suggests that DDX39 may be one of the shelterin accessory factors, which have additional nontelomeric functions. Because TRF2 has been implicated in the protection of chromosome ends (van Steensel et al., 1998), we next investigated the effect of DDX39 on TRF2 protein stability. Depletion of endogenous DDX39 by shRNA did not change the levels of TRF2 in two independent clones (Fig. S1A). The effect of DDX39 on TRF2 protein stability was further investigated by determining the nuclear localization of TRF2. When the DDX39 expression was inhibited by shDDX39-1, the fluorescence intensity of TRF2 was not significantly reduced compared to the control cells (Fig. S1B).
DDX39 interacts with the TRFH domain of TRF2
To map the region in DDX39 that is important for TRF2 binding, a series of DDX39 fragments were fused to GST and used in the in vitro binding assay (Fig. 2A). GST-DDX39 fragments encompassing amino acid residues 75–259, 161–259, and 161–427 bound to TRF2, whereas GST-DDX39 fragments encompassing amino acid residues 1–74 and 260–427 failed to associate with TRF2 (Fig. 2B). When HT1080 cells were cotransfected with Myc-DDX39 fragments and Flag-TRF2, Flag-TRF2 was immunoprecipitated by DDX39 fragments containing amino acid residues 161–259 (Fig. 2C). These results indicate that TRF2 interacts with the internal region containing the conserved helicase motifs Ib, II, and III in DDX39 (Sugiura et al., 2007a). Upon examination of the DDX39-binding domain in TRF2, we found that GST-TRFH bound efficiently to endogenous DDX39, whereas GST-Myb had substantially reduced binding activity (Fig. 2D,E). In contrast, GST-GAR and GST-Hinge had no detectable binding activity. These data suggest that TRF2 may bind DDX39 through the two distinct regions, but the TRFH domain has higher affinity for DDX39 binding than the Myb domain.
Recently, it has been reported that the F/Y-X-L-X-P motif (where X is any amino acid) on shelterin-associated proteins is critical for their recognition by the TRFH domain of TRF2 (Chen et al., 2008; Kim et al., 2009). Interestingly, we identified the sequence R208-D-V-Q-E-I-F-R-L-T-P-H-E220 within the TRF2-binding domain of DDX39. This sequence closely resembles the TRFH-binding motifs of Apollo, PNUTS, and MCPH1 (Fig. 2F). Alanine substitutions of F214 and L216 residues (DDX39/2A) resulted in its loss of TRF2 interaction (Fig. 2G,H), suggesting that DDX39 interacts with the TRFH domain of TRF2 in the same fashion as do other TRF2-binding shelterin accessory factors. To investigate whether this mutant still gets recruited to the telomeres, HT1080 cells were transfected with Myc-DDX39 or Myc-DDX39/2A and subjected to indirect immunofluorescence staining. Although ectopically expressed DDX39 proteins showed a diffused staining pattern, we found that the percentage of telomeric sites costained with DDX39/2A was significantly reduced compared to wild-type DDX39 (Fig. S2A,B). These results suggest that DDX39 interaction with the TRFH domain of TRF2 is required for DDX39 recruitment to the telomeres.
DDX39 directly binds to the C-terminal region of hTERT
RNA helicases have been shown to associate with biological processes involving RNA, such as splicing, ribosome biogenesis, translation, RNA transport, and RNA editing (Rocak & Linder, 2004). Because telomerase contains a template RNA for reverse transcriptase activity, we asked whether DDX39 is involved in biogenesis and/or regulation of the telomerase holoenzyme. First, we examined whether DDX39 interacts with hTERT. HT1080 cells were cotransfected with Flag-hTERT and DDX39-V5 and subjected to immunoprecipitation. Flag-hTERT was detected in anti-V5 immunoprecipitates when DDX39-V5 was expressed (Fig. 3A). KIP has been previously shown to interact with hTERT (Lee et al., 2004) and was used as a positive control. Because expressed hTERT is assembled with hTERC, we next examined whether hTERC mediates the interaction between hTERT and DDX39. We observed that RNase treatment did not affect the interaction of DDX39 with hTERT (Fig. 3B), indicating that DDX39 binds directly to hTERT in the absence of telomerase RNA component.
To map the region in DDX39 that interacts with hTERT, we expressed hTERT-HA in HT1080 cells and subjected to GST pull-down experiments using a series of DDX39 fragments. GST-DDX39 fragments encompassing amino acid residues 161–427 and 260–427 bound to hTERT, whereas other GST-DDX39 fragments failed to interact with hTERT (Fig. 3C). These results indicate that hTERT binds the C-terminal domain of DDX39 (residues 260–427), whereas TRF2 associates with the internal region of DDX39 (residues 161–259) (see Fig. 2A). To map the domain of hTERT that mediates the interaction with DDX39, we accessed binding of Myc-DDX39 by immunoprecipitating a series of Flag-tagged fragments of hTERT in HT1080 cells (Fig. 3D). The C-terminal fragment of hTERT encompassing amino acid residues 946–1132 immunoprecipitated DDX39 (Fig. 3E).
DDX39 physically associates with catalytically competent telomerase
Because DDX39 interacts with hTERT, we examined whether DDX39 binds to catalytically competent telomerase. Lysates from HT1080 cells were immunoprecipitated with a DDX39-specific antibody and analyzed for telomerase activity by the TRAP assay. Anti-DDX39 antibody immunoprecipitated telomerase activity (Fig. 3F). Anti-KIP antibody was used as a telomerase-binding positive control (Lee et al., 2004). Telomerase activity was not immunoprecipitated with control antibodies against TRF2 and p53. Identical results were observed in H1299 cells (Fig. S3A). To exclude the possibility that anti-DDX39 antibody may bind nonspecifically to telomerase, HT1080 cells were transfected with DDX39-V5 and immunoprecipitated with anti-V5 antibody. Telomerase activity was detected in anti-V5 immunoprecipitates from cells expressing DDX39-V5 but not from the control cells (Fig. S3B). We note that telomerase activity was not significantly altered by overexpression of DDX39.
To determine the domain in DDX39 that interacts with active telomerase, a series of GST-DDX39 fragments was used in the in vitro binding assay. Consistent with hTERT-binding domain in DDX39, telomerase activity was recovered with GST-DDX39 containing residues 260–427 (Fig. 3G). The extent of telomerase activity recovered by GST-DDX39161–427 was reproducibly lower than that recovered by GST-DDX39260–427. These results suggest that DDX39 associates with catalytically competent telomerase through an interaction with hTERT and that this association could be specific.
DDX39 modulates telomere length homeostasis
Because DDX39 associates with telomerase in vivo, we wished to determine whether DDX39 affects telomerase activity and telomere length. We established HT1080 cell lines stably expressing DDX39 or the control vector. Multiple independent clones were isolated to rule out the effect of clonal variation. As measured by immunoblotting analysis with anti-DDX39 antibody, the amount of exogenously expressed DDX39 was approximately 5-fold greater than that of endogenous protein (Fig. 4A). Cells expressing DDX39 and the control vector grew normally and exhibited no detectable differences in growth rates or morphology over 46 population doubling (PD). At 46 PD, we compared telomerase activity in stable cell lines expressing DDX39 and the control vector. Telomerase activity was not significantly changed by overexpression of DDX39 (Fig. 4B and Fig. S4A). We next measured the terminal restriction fragment (TRF) length to investigate whether overexpression of DDX39 affects telomere length. HT1080 cells expressing the control vector maintained a stable TRF length over 46 PD (Fig. 4C). A similar TRF length was detected in at least two HT1080 cell lines stably expressing DDX39 at early PD (5 PD), but the TRF length progressively increased at late PDs (46 PD). These results suggest a role of DDX39 as a positive regulator of telomere length.
To examine the role of DDX39 in a more physiological setting, the expression of endogenous DDX39 was depleted using shRNA produced from a lentiviral vector. HT1080 cells were infected with two different DDX39 shRNA viral constructs, and multiple independent stable clones were isolated for each construct. Two different shRNA constructs substantially reduced DDX39 expression as measured by reverse transcription–polymerase chain reaction (Fig. 4D). DDX39 protein levels were also significantly reduced in DDX39-depleted cells as determined by immunoblot analysis, although the extent of the DDX39 reduction was variable among individual clones (Fig. 4D). Cells harboring two different shRNAs maintained the reduced level of DDX39 protein over 45 PD. Depletion of DDX39 did not affect telomerase activity as detected by the TRAP assay (Fig. 4E and Fig. S4B). Whereas cells expressing the control shRNA maintained a stable TRF length over 45 PD, depletion of DDX39 resulted in modest telomere shortening in two independent clones (Fig. 4F). To rule out the effect of clonal variation, we isolated multiple independent clones expressing shDDX39-1 or shDDX39-2 and measured the TRF length. Reduction in telomere length was found in all clones examined (Fig. S5). These results suggest that although DDX39 has no direct regulatory effect on telomerase activity, the endogenous level of DDX39 is required to maintain a stable telomere length.
Because DDX39 modulates telomere length homeostasis in telomerase-positive HT1080 cells, we wished to determine whether the effect of DDX39 on telomere length is telomerase dependent. The expression of endogenous DDX39 was inhibited using shRNA in U2OS cells that lack detectable telomerase activity (Oh et al., 2010) (Fig. S6C). As shown in Fig 4G, U2OS cells stably expressing two different shRNAs maintained a stable TRF length up to 42 PD. These findings support the idea that the interaction of DDX39 with active telomerase is important for DDX39 function on telomere length regulation.
Finally, we examined the effects of DDX39 depletion on cell growth. DDX39 knockdown HT1080 cells grew slower than the control cells but with constant rates (Fig. S6A). To examine whether this growth arrest of DDX39 knockdown cells was accompanied by cellular senescence, we performed the senescence-associated β-galactosidase assay. We found that DDX39 knockdown cells were all β-galactosidase negative and displayed no senescence-associated phenotype compared to the control cells (data not shown). Thus, the minor growth defect observed is not likely to be connected to telomere shortening, in which case the DDX39 knockdown cells would be expected to grow with rates similar to the control cells until their telomeres become critically short. We also found that depletion of DDX39 in U2OS cells had no effect on cell growth (Fig. S6B).
Depletion of DDX39 induces DNA-damage foci at internal genome as well as telomeres
In the absence of a functional telomere maintenance pathway, dividing cells show a progressive loss of telomeric DNA during successive rounds of cell division (Lingner et al., 1995; Cerone et al., 2005). Therefore, it is possible that telomere shortening induced by DDX39 depletion may be because of telomere dysfunction. Dysfunctional telomeres resemble damaged DNA and are recognized by the canonical DNA-damage signaling pathway (Karlseder et al., 1999; Dd’Adda di Fagagna et al., 2003). The resulting TIFs represent the foci of DNA damage response factors that coincide with telomeres (Takai et al., 2003). The telomeric foci for 53BP1 and γ-H2AX were examined in cells expressing two different DDX39 shRNAs. Depletion of DDX39 induced numerous 53BP1 and γ-H2AX foci in the nucleus, resulting in about 24% cells showing ten or more DNA-damage foci per nucleus, whereas these events were rare in control cells (Fig. 5A). Representative fluorescence images of nuclei showing a large number of 53BP1 and γ-H2AX foci are shown in Fig. 5B,C, and some of these foci colocalized with telomeres as indicated by dual staining with telomeric TTAGGG-specific FISH probe. When the TIF response was quantified in 100 nuclei showing ten or more DNA-damage foci, about 21% of cells expressing DDX39 shRNA contained more than five foci at telomeres and were scored as TIF positive (Fig. 5D). DDX39 depletion also resulted in 53BP1 and γ-H2AX foci that were not obviously associated with telomeres, indicating that DDX39 is involved in nontelomeric DNA-damage. These results suggest that DDX39 has a novel function in controlling a general DNA damage response. While there is a general increase in 53BP1 and γ-H2AX foci following DDX39 knockdown, there is a statistically significant re-localization of those foci to telomeres (Fig. 5E), indicating that DDX39 deficiency preferentially affected telomeres. HT1080 cells contained a considerable basal level of 53BP1 and γ-H2AX foci even in the absence of DDX39 inhibition. 53BP1 functions have been shown to be constitutively activated in several tumor cell lines (DiTullio et al., 2002). However, these foci rarely colocalized with telomeres (Fig. 5B,C).
To examine the specificity of DDX39 shRNA, we performed the rescue assay with the version of DDX39 resistant to shDDX39-1 (RNAi-R-DDX39). When DDX39 knockdown cells were transfected with Myc-DDX39 or Myc-RNAi-R-DDX39, the expression of Myc-DDX39 was almost completely reduced (Fig. S7A). However, the expression of Myc-RNAi-R-DDX39 was not inhibited. The DNA-damage foci phenotype associated with shRNA-1 was not affected by the expression of wild-type DDX39 but significantly reduced when overexpressed with RNAi-R-DDX39 (Fig. S7B,C), indicating that the DNA-damage signal is the result of DDX39 depletion. We note that the basal levels of the DNA-damage foci in HT1080 cells were not changed by the overexpression of wild-type DDX39 or RNAi-R-DDX39 (Fig. S8).
Although there is a significant increase in DNA-damage foci located at telomeres in DDX39 knockdown cells, there is also a massive increase in the number of DNA-damage foci that did not colocalize with telomeres. It is possible, therefore, that the observed TIFs may represent fortuitous colocalization of the telomere and 53BP1 (or γ-H2AX) signals rather than true colocalization. To test this possibility, we compared TIF formation induced by DDX39 depletion with DNA-damage foci induced by camptothecin that is not expected to affect telomeres preferentially (Fig. 6A,B). We found that 53BP1 signals were markedly increased by camptothecin treatment in both control and DDX39-depleted cells, such that approximately 60% of camptothecin-treated cells showed ten or more 53BP1 foci per nucleus (Fig. 6C). However, only about 25% of the DDX39-depleted cells and about 2% of the control cells, respectively, exhibited ten or more 53BP1 foci per nucleus. We also found that approximately 20% of the DDX39-depleted cells contained more than five TIFs per nucleus as detected by 53BP1 staining (Fig. 6D). However, the number of TIF-positive DDX39-depleted cells was not significantly increased by camptothecin treatment, such that approximately 23% of the cells were positive for 53BP1 TIFs. Taken together, these results suggest a specific effect of DDX39 depletion at telomeres.
We next evaluated the effects of DDX39 depletion on telomere DNA structure by metaphase chromosome FISH. The analysis of metaphase spreads derived from DDX39-depleted cells did not show significant levels of end-to-end chromosome fusions (Fig. 7A). However, depletion of DDX39 resulted in a significant increase of chromatid ends without telomeric FISH signal compared to chromosomal structures observed in control nuclei (Fig. 7A,B). We found that no significant levels of telomere aberrations were induced by overexpression of DDX39 (Fig. 7B).
DDX39 mutant lacking helicase activity retains the ability to represses the formation of DNA-damage foci
As DDX39 is an RNA helicase (Sugiura et al., 2007a), we determined the requirement of helicase activity in the DNA-damage foci phenotype associated with DDX39 depletion. We generated DDX39 point mutant (S227L) in which the serine (S) residue at position 227 in the ‘SAT’ helicase motif was substituted by leucine (L) (Fig. 8A). To test the DDX39′s ability to unwind double-stranded RNA, wild-type and mutant DDX39 proteins were incubated with duplex RNA. Whereas DDX39 unwound double-stranded RNA in an ATP-dependent manner, a mutation of the helicase motif abolished RNA unwinding activity (Fig. 8B). Although the TRF2-binding domain (amino acid residues 161–259) in DDX39 includes the helicase motif (see Fig 2A), the S227L mutation did not impair its ability to interact with TRF2 (Fig. 8C). To examine whether RNA helicase activity is involved in the DNA damage foci phenotype associated with DDX39 depletion, we created the version of the S227L protein resistant to shDDX39-2 (RNAi-R-S227L). The expression of Myc-RNAi-R-S227L was not inhibited in DDX39-depleted cells (Fig. 8D). The DNA-damage foci phenotype associated with DDX39 depletion was not significantly changed by the expression of S227L but significantly reduced by the expression of RNAi-R-S227L (Fig. 8E), suggesting that DDX39 mutant lacking helicase activity is still able to repress the formation of DNA-damage foci.
We next examined whether DDX39 binding to TRF2 is required for suppression of the DNA damage phenotype. We generated the version of the DDX39/2A protein resistant to shDDX39-1 (RNAi-R-DDX399/2A) (see Fig. 2G for DDX39/2A). The expression of Myc-RNAi-R-DDX39/2A was not inhibited in DDX39-depleted cells (Fig. 8F). As shown in Fig. 8G, DDX39-depleted cells expressing RNAi-R-DDX399/2A still showed the DNA-damage foci phenotype although the number of the foci was slightly reduced compared to vector control cells. These results suggest that DDX39 binding to TRF2, at least in part, is responsible for suppression of a DNA-damage response.
DDX39 was recently identified as a novel member of the DEAD-box RNA helicases and was shown to stimulate cancer cell growth (Sugiura et al., 2007a,b). Because of a high sequence similarity to UAP56 (Zhao et al., 2004; Shen et al., 2007, 2008), DDX39 has been suggested to play a role in pre-mRNA splicing and mRNA export (Sugiura et al., 2007a,b). Here we describe the characterization of DDX39 as a novel protein interacting with TRF2. The following results support that DDX39 is a shelterin accessory factor required for telomere protection. First, TRF2 interacts with DDX39 via the TRFH domain in vitro and in vivo. A motif scan of the amino acid sequence revealed the F214-R-L-T-P218 motif within the TRF2-binding region of DDX39. Alanine substitutions of F214 and L216 completely abolished its interaction with TRF2, indicating that DDX39, like other TRF2-binding shelterin accessory factors, binds TRF2 through the FXLXP consensus motif (Chen et al., 2008). Second, immunolocalization experiments demonstrate that DDX39 is a nuclear protein that colocalizes with TRF2 and telomere signals as indicated by dual staining with telomere-specific FISH probe. These observations suggest that DDX39 can be targeted to telomeres through an interaction with TRF2. However, DDX39 was also found in many small foci that did not colocalize with TRF2. Although the nature of these localization sites was not determined, its low abundance at telomeres indicates that DDX39 is not a core component of shelterin. Third, whereas overexpression of DDX39 in telomerase-positive HT1080 cells significantly lengthened telomeres, depletion of endogenous DDX39 resulted in modest telomere shortening without affecting telomerase activity. Fourth, depletion of DDX39 resulted in numerous 53BP1 and γ-H2AX foci in the nucleus, and some of the foci colocalized with telomeres. However, significant numbers of the foci were not obviously associated with telomeres, suggesting that DDX39 has additional nontelomeric function that is required for a general DNA-damage response. Finally, depletion of DDX39 resulted in an increase of chromatid ends without telomeric FISH signal. Such telomere aberration was not induced by overexpression of DDX39. Taken together, these results suggest that DDX39 is a novel shelterin accessory factor which modulates telomere length homeostasis and regulates the telomere-associated DNA damage response. In addition to its telomeric function, DDX39 also has a nontelomeric function in DNA-damage signaling pathway, contributing to global genome integrity. Although the precise role of DDX39 in DNA-damage response remains to be seen, this novel function may be relevant to its role in cancer progression.
The functional significance of DDX39 was investigated by overexpression and depletion of DDX39. Although DDX39 depletion was only partial (approximately 20% of control), it resulted in modest telomere shortening and telomere dysfunction, including an increase in 53BP1- and γ-H2AX-associated telomere foci and an increase in metaphase chromosome aberrations. In contrast, overexpression of DDX39 did not affect the basal levels of TIFs in HT1080 cells. Because overexpression of DDX39 resulted in telomere elongation, these observations are consistent with the recent finding that increased telomere length reduces the probability of telomere dysfunction in telomerase-positive cells (Cesare et al., 2009). The ability of TRF2 to participate in t-loop formation suggests that this protein is also required for repression of a DNA-damage response at telomeres (Stansel et al., 2001; Verdun & Karlseder, 2007). Because depletion of DDX39 neither affects the level of TRF2 nor removes TRF2 from telomeres, the TIF phenotype associated with DDX39 depletion could not be because of diminished TRF2 function. We found that disruption of DDX39 binding to TRF2 resulted in failure of DDX39 to prevent the formation of DNA-damage foci. These results suggest that the interaction between DDX39 and TRF2, at least in part, is required for suppression of a DNA-damage phenotype. The telomere FISH analysis on metaphase spreads from DDX39-depleted cells showed a significant increase of telomeric signal-free chromatid ends, but did not show end-to-end chromosome fusions. These results suggest that dysfunctional telomeres associated with DDX39 depletion still retain the capacity to suppress end-to-end fusions through the retention of TRF2.
DDX39, in addition to its binding to TRF2, can associate with catalytically active telomerase through an interaction with hTERT. DDX39 interacts with hTERT via its C-terminal region which does not overlap with TRF2-binding region, suggesting that DDX39 binding to TRF2 and hTERT may not be mutually exclusive. However, we did not detect formation of a ternary protein complex (DDX39/TRF2/hTERT). The absence of evidence does not mean that such complexes do not exist in the nucleus. They may exist only in chromatin context or may be transiently associated. The critical question that remains to be answered is how DDX39 regulates telomere length. One possibility is that DDX39 may shuttle between shelterin complex and telomerase. Differential binding of DDX39 may suggest a mechanism to recruit telomerase to telomeres, pointing to a crucial role of DDX39 in telomerase-mediated telomere maintenance. Through the interactions of DDX39 with TRF2 and hTERT, telomerase could be enriched at telomeres, positioning it in the vicinity of the 3′ telomere terminus. Indeed, we found that DDX39 depletion had no effect on telomere length in telomerase-negative U2OS cells, suggesting that the effect of DDX39 on telomere length is telomerase dependent. This model is supported by the previous finding that TRF2 preferentially binds to the end of telomeric repeat array if it contains a G-strand overhang (Stansel et al., 2001). Although DDX39 does not affect the catalytic activity of telomerase as measured in cell lysates, two interactions of DDX39 may influence the functions of TRF2 and hTERT at telomeres. Alternatively, DDX39 may play multifunctional roles in diverse pathway in the cells because it binds other proteins as well (Sugiura et al., 2007b).
Under conditions in which the DDX39 concentration is maintained below the threshold level, TRF2 may lose its ability to repress DNA damage response at telomeres. In parallel, telomerase may not be efficiently recruited to telomeres, subsequently resulting in gradual telomere shortening until a new length setting is achieved. We observed that the telomere erosion rate in DDX39-depleted cells (approximately 6–7 bp per PD) is much lower than the speed of telomere erosion in the absence of telomerase (25–200 bp per PD) (Lingner et al., 1995). This might occur because the residual level of DDX39 is still available in knockdown cells. Conversely, overexpression of DDX39 did not affect the telomere protection function but resulted in progressive elongation of the telomere to a new equilibrium length, suggesting that DDX39 exerts a positive role in telomere length control. Because the DDX39 concentration is important to the cells in maintaining telomere length within a given size range, this protein may represent a potential molecular target in cancer.
It has been recently reported that mammalian telomeres are transcribed into telomeric repeat-containing RNA (TERRA) (Azzalin et al., 2007; Schoeftner & Blasco, 2008). TERRA RNA localizes to telomeres through direct interaction with TRF2 GAR domain and facilitates origin recognition complex (ORC) recruitment and heterochromatin formation at telomeres (Deng et al., 2009). Furthermore, TERRA RNA depletion caused telomere dysfunction, including an increase in TIFs and metaphase chromosome aberrations. Given that DDX39 contains the characteristic motifs required for RNA helicase activity and plays an essential role in many steps of RNA metabolism (Sugiura et al., 2007a,b), we speculate that DDX39 might be involved in regulating TERRA function. Because the TRF2 GAR domain has a high binding affinity for the G-rich RNA capable of forming G quadruplex structures (Azzalin et al., 2007), it will be of interest to determine whether DDX39 resolves the higher order of TERRA RNA structures.
Overall, our results provide an insight into the new cellular function of DDX39 in addition to its role in RNA metabolism. In this work, we propose that DDX39 is a novel shelterin accessory factor that helps maintain the telomere protection function and may play an important role in recruitment of telomerase to telomeres. Although important questions about the physiological role of a ternary protein complex (DDX39/TRF2/hTERT) at telomeres and how such a complex is regulated remain to be resolved, our results suggest that DDX39 represents a new pathway for maintaining the functional telomeres and modulating telomere length homeostasis.
Yeast two-hybrid screening
Yeast two-hybrid screening was performed as described previously (Lee et al., 2004). Briefly, the TRF2 cDNA was fused to the LexA DNA-binding domain and transformed by the lithium acetate method into the EGY48 yeast strain. Expression of the LexA-TRF2 fusion protein was verified by Western blotting using anti-LexA antibody. The stable strain was transformed with a HeLa cDNA library fused to the activation domain vector pB42AD (Clontech).
Recombinant protein expression and antibody production
The DDX39 expression vectors were constructed by inserting the EcoRI and XhoI fragments from the full-length DDX39 cDNA into pcDNA3.1/V5-His or pcDNA3/Myc (Invitrogen, Carlsbad, CA, USA). The expression vectors for GST-DDX39 were constructed by cloning the full-length and truncated fragments from the DDX39 cDNA into pGEX-6X-1, and the GST fusion proteins were purified by glutathione-Sepharose beads according to the manufacturer’s instructions (Amersham Biosciences, Seoul, Korea). To raise antibodies against DDX39, rabbits were immunized with recombinant DDX39 prepared by cleaving GST-DDX39 with Prescission protease followed by the removal of the cleaved GST and uncleaved GST-DDX39 with glutathione-Sepharose. Anti-DDX39 antibodies were affinity-purified on DDX39 coupled to CNBr-activated Sepharose 4B (Amersham Biosciences).
Establishment of stable cell lines and RNA interference
Stable HT1080 cells expressing DDX39 were established as previously described (Her & Chung, 2009). Briefly, the full-length DDX39 cDNA was subcloned into a pLentiM1.4 expression vector (Macrogen, Seoul, Korea) and transfected into 293T cells with VSV-G (vesicular stomatitis virus glycoprotein) expression vector and gag-pol expression vector. The culture supernatants containing viral vector particles were harvested 48 h after transfection, and HT1080 cells were transduced with the DDX39-expressing lentivirus. After selection with puromycin, multiple independent single clones were isolated and checked for protein expression by immunoblotting with anti-DDX39 antibody. Two different shRNA expression lentiviral vectors for targeting DDX39 (5′-CCGGGCGAGTCAACATCGTCTTTAACTCGAGTTAAAGACGATGTTGACTCGCTTTTTG-3′ for shDDX39-1; 5′-CCGGCCAGGTGATAATCTTCGTCAACTCGAGTTGACGAAGATTATCACCTGGTTTTTG-3′ for shDDX39-2) and the pLKO.1-puro control vector were purchased from Sigma-Aldrich (St. Louis, MO, USA). For virus production, the pLKO.1-puro vectors were transfected into HEK293T packaging cells with compatible packaging plasmids according to the manufacturer’s instructions (Sigma). The version of DDX39 resistant to shRNAs was silently mutated in the shDDX39-1 and shDDX39-2 target sequences (underlined nucleotide) without altering the amino acid sequence with the following primers: 5′-GGGATGGACATCGAGCGAGTTAACATAGTCTTCAACTACGACATG-3′ for shDDX39-1; 5′-GTGCTGGAGTTTAATCAGGTGATAATATTTGTCAAGTCAGTGCAGCGCTGCATGGC CCTGG-3′ for shDDX39-2.
GST pull-down, immunoprecipitation, and immunoblot
The expression vectors were transfected into HT1080 cells using LipofectAMINE 2000 (Invitrogen), and GST pull-down and immunoprecipitation were performed as previously described (Lee et al., 2004). Immunoprecipitation and immunoblot analyses were performed using anti-V5 (Invitrogen), anti-Flag (Sigma), anti-HA (Santa Cruz Biotechnology, Santa Cruz, CA, USA), anti-TRF2 (Upstate Biotechnology, Waltham, MA, USA), anti-TRF1 (Sigma), and anti-DDX39 antibodies. All the immunoblots are representatives of at least three experiments that demonstrated the similar results.
Telomerase assay and terminal restriction fragment (TRF) analysis
Immunoprecipitated proteins were added to telomerase extension reactions, and the TRAP assay was used as previously described (Kim et al., 2003). To measure the telomere length, genomic DNA was digested with RsaI and HinfI and separated on 1% agarose gel. DNA samples were transferred to a nylon membrane (Hybond N+; Amersham Biosciences) and hybridized with a 32P-labeled (TTAGGG)20 probe.
Immunofluorescence and telomere FISH staining
Cells grown on glass coverslips were fixed with 4% paraformaldehyde at room temperature for 10 min and permeabilized with 0.1% Triton X-100 in phosphate-buffered saline (PBS) for 20 min. Cells were then blocked in PBS containing 5% bovine serum albumin and incubated with rabbit anti-53BP1 (Novus Biologicals, Littleton, CO, USA) or rabbit anti-γ-H2AX (Cell Signaling Technology, Danvers, MA, USA) overnight at 4 °C. After thorough washing with PBS, cells were incubated with FITC-conjugated anti-rabbit secondary antibody (Animal Genetics Inc., Suwon, Korea). Telomere FISH staining was performed with Cy3-(CCCTAA)3 PNA probe (Panagene, Seoul, Korea) as previously described (van Steensel et al., 1998). DNA was counterstained with 4,6-diamino-2-phenylindole (DAPI) (Vectashield; Vector Laboratories, Burlingame, CA, USA). Immunofluorescence images were captured using a confocal laser-scanning microscope (Carl Zeiss, Jena, Germany). For telomere FISH on metaphase spreads, cells were treated with demecolcine (10 μg mL−1) for 6 h, and metaphase chromosomes were prepared from growing cell cultures by standard methods (van Steensel et al., 1998). Telomere FISH with a Cy3-(CCCTAA)3 PNA probe was performed using telomere PNA FISH kit (Dako).
RNA unwinding assay
RNA substrates used in the RNA unwinding assay were generated with pBluescript vectors as previously described (Iost et al., 1999). pBluescript was linearized with Kpn1 and transcribed with T3 polymerase to generate a 120-nucleotide transcript and linearized with EcoR1 and transcribed with T7 polymerase in the presence of [α-32P]UTP to generate a 50-nucleotide transcript. These transcripts were annealed, and the resulting partial duplex was purified by native polyacrylamide gel electrophoresis. The reaction mixtures contained 20 mM Tris–HCl, pH 8.0, 70 mM KCl, 2 mM MgCl2, 2 mM dithiothreitol, 15 units RNasin, 2 mM ATP, and the labeled partial duplex. Reactions were initiated by addition of purified DDX39. After incubation for 30 min at 37°C, the reactions were stopped and resolved in a 10% native polyacrylamide gel. Labeled RNAs were visualized by autoradiography and quantified using a Fuji phosphorimager.
This work was supported in part by Korea Research Foundation Grants (KRF-M1075604000107N560400110 and KRF-20090084897) and by the WCU Fund (R31-2009-000-10086-0) from the Korean Ministry of Education, Science, and Technology.