Whole-genome RNAi screening identified factors involved in Dpp signaling
We used nuclear translocation of Mad as the readout in our RNAi screening because this is an early event in Dpp signaling. When Flag-Mad was conditionally expressed in S2R+ cells, it was detected diffusively throughout the cell (). In contrast, when the Dpp receptor kinases Punt and Thickvein (Tkv) were coexpressed, which caused Mad phosphorylation, the bulk of Flag-Mad gradually became predominantly localized to the nucleus (). With this cell line (Mad+R), we performed an RNAi screening in which the cells were treated with a library of ~21,300 dsRNAs individually targeting over 95% of the annotated Drosophila
genome (Armknecht et al., 2005
). dsRNAs against the GFP and the punt
combination were used as negative and positive controls, respectively. After 3 d of incubation with dsRNAs, the Mad+R cell line was induced to express Flag-Mad, Punt, and Tkv, and the subcellular location of Flag-Mad was visualized with anti-Flag immunofluorescence staining followed by high throughput automated microscopy. Upon visual inspection of images obtained from duplicate screenings, we identified 346 dsRNAs that caused diffused distribution of Flag-Mad throughout the cell compared with the negative control dsRNA. Many of the genes corresponding to the 346 dsRNAs contain domains suggestive of their functions, and can be broadly categorized as in . The complete list of strong and weak hits in the primary screening can be accessed at the Drosophila
RNAi Screening Center (DRSC) website (http://www.flyrnai.org
Figure 1. Whole-genome RNAi screening in S2R+ cells uncovered new components in the Dpp–Mad pathway. (A) Phosphorylation-dependent nuclear accumulation of Flag-Mad. S2R+ cells stably transfected with plasmids for Flag-Mad only, or Flag-Mad (more ...)
Preliminary hits and predicted functions
The candidate hits were selected in an anonymous manner (see Materials and methods). Indeed, among the hits that gave strong phenotype were punt
, which confirmed that Mad phosphorylation is prerequisite for its nuclear accumulation and that the screening was robust (). Of particular note among the primary hits was msk
, a karyopherin that was previously suggested to be required for nuclear import of activated Drosophila
ERK (dERK) (Lorenzen et al., 2001
). The RNAi library used here contains dsRNA targeting many molecules known to be involved in nuclear transport, including importins, exportins, and nucleoporins. But besides msk
, none was among the 346 hits identified in the primary screening (). In this study we focus on the analysis of msk
; the validation and bioinformatics analyses of the other hits will be presented elsewhere.
Msk is required for nuclear accumulation of activated Mad and transcriptional activation of Dpp target genes
To verify the effect of msk RNAi in the primary screening, we designed and tested a second non-overlapping dsRNA against msk. Indeed, depletion of Msk by a different dsRNA also led to severely impaired nuclear concentration of Mad, and the effect was as potent as the positive control RNAi targeting Punt and Tkv (). This result strongly suggests that the block of Mad nuclear translocation we observed in the screening was not due to off-target effects of the dsRNA. In contrast to RNAi against Punt and Tkv, RNAi of msk did not affect C-terminal phosphorylation of Mad (), suggesting that Msk functions downstream of Mad phosphorylation, perhaps in transporting Mad into the nucleus.
Because the screening was performed using a S2R+ cell line overexpressing exogenous Flag-Mad, we wanted to determine if endogenous Mad is under the same regulation by msk
in a different Drosophila
cell line. Dpp treatment of Drosophila
S2 cells resulted in predominant nuclear distribution of phosphorylated Mad, as revealed by immunofluorescence staining using a phospho-Mad–specific antibody, PS1 () (Tanimoto et al., 2000
). Depletion of Msk by RNAi clearly resulted in more diffusive distribution of phospho-Mad (changing the nucleus/cytoplasm ratio from 2.9 to 1.1), while not affecting the level of Mad phosphorylation at its C terminus ( and Fig. S1 A, available at http://www.jcb.org/cgi/content/full/jcb.200703106/DC1
Figure 2. msk, but not the importin β homologue ketel, is required for nuclear accumulation of endogenous Mad in Dpp-treated Drosophila S2 cells. (A) S2 cells were treated with indicated dsRNA and then subject to Dpp stimulation (1 nM for 1 h). Distribution (more ...)
Msk has been suggested to cooperate with the Drosophila
importin β homologue Ketel in nuclear import of dERK, because mutations in either msk
inhibited nuclear accumulation of dERK (Lorenzen et al., 2001
). Moreover, the mammalian orthologues of Msk have been shown to function in conjunction with importin β (Gorlich et al., 1997
; Jakel et al., 1999
). Thus, we tested if ketel
might also be involved in nuclear translocation of Mad. Knockdown of ketel
by RNAi resulted in reduced phosphorylation of endogenous Mad (Fig. S1 A), but nevertheless phospho-Mad was still detected predominantly within the nucleus (). Quantitation of phospho-Mad staining intensity in the nucleus and cytoplasm confirmed that RNAi against ketel
did not affect nuclear accumulation of phospho-Mad (; changing the nucleus/cytoplasm ratio from 2.9 to 3.1, n
> 50). Similar observations were also made in S2R+ cells (unpublished data). Western blot analysis of cytoplasmic and nuclear fractions of S2 cells further validated that Msk, but not Ketel, is required for nuclear accumulation of phospho-Mad (). As shown in , although depletion of Ketel resulted in reduced amount of phospho-Mad through an unknown mechanism, the majority of phospho-Mad was still present in the nucleus (). Classic NLS–mediated nuclear import is dependent on importin β (Stewart, 2007
). Indeed, RNAi against ketel
clearly impaired nuclear accumulation of classic NLS–fused GFP, while depletion of Msk had no effect (Fig. S1 B). This result verified that RNAi against ketel
was effective, and nuclear transport of Dpp-activated Mad is independent of the importin β homologue Ketel.
In both S2R+ and S2 cells, treatment with Dpp results in transcriptional activation of dad
, a known Smad target gene in mammalian cells as well (Nakao et al., 1997
; Tsuneizumi et al., 1997
). When Msk was depleted by RNAi, the Dpp- induced increase in dad
expression was completely abolished (). The blocking effect of msk
RNAi on dad
expression was as strong as that caused by punt
RNAi (). Thus, consistent with being an essential factor for nuclear import of Mad, Msk is critical for the transcriptional output of Dpp.
Interaction between Msk and Mad
To address the question if Msk is directly involved in transporting phospho-Mad into the nucleus, we tested protein–protein interaction between endogenous Msk and Flag-tagged Mad. Indeed, endogenous Msk coimmunoprecipitated with Flag-Mad from S2 cell extract (). Under our experimental conditions, both basal state and phosphorylated Mad displayed comparable interaction with Msk (). This suggests that binding of Msk is not unique to phospho-Mad, and Msk alone may not account for why only phospho-Mad accumulates in the nucleus. Therefore, although Msk is crucial for phospho-Mad to enter the nucleus, additional factors are involved to retain only phospho-Mad in the nucleus.
The above observations in Drosophila cells identify msk as a new regulator in the Dpp pathway whose function is critical for nuclear accumulation of Dpp-activated Mad.
Analysis of msk mutant cells in the eye imaginal disc
Mutations in the msk
gene resulted in embryonic lethality (Baker et al., 2002
; Vrailas et al., 2006
). Thus, to evaluate the functions of Msk in vivo, we used ey
to generate msk
) clones in the developing eye imaginal disc (Baker et al., 2002
; Vrailas et al., 2006
). The clones were marked as GFP negative. In eye discs of third instar larvae, the PS1 antibody detected two stripes of phospho-Mad–containing cells around the morphogenetic furrow, consistent with the established role of Mad in eye development () (Wiersdorff et al., 1996
). The phospho-Mad signal in the anterior stripe is weaker and more diffused compared with that in the posterior stripe (). In the posterior stripe, phospho-Mad–positive cells span 5–6 cells wide, and quantitation of cell staining showed that 20.8% of the cells (n
= 1,401) had phospho-Mad concentrated in the nucleus (). The rest of the cells within the posterior stripe have either undetectable or diffusive phospho-Mad staining ().
Figure 3. msk null mutant cells in the developing eye imaginal disc did not have distinct nuclear staining of phospho-Mad. (A) Third instar eye imaginal discs (anterior to the left) were stained with PS1, which specifically recognizes phospho-Mad (Mad-P, red). (more ...)
clones are small in size and number compared with the wild-type clones, consistent with previous reports that msk5
mutation led to growth disadvantages (Baker et al., 2002
; Vrailas et al., 2006
). In msk5
clones that straddle the posterior stripe, we detected a significantly smaller number of cells (4%, n
= 142) with strong phospho-Mad staining concentrated in the nucleus (). In comparison, in wild-type clones generated by ey:FLP
, the number of cells with high phospho-Mad signal distinctively in the nucleus is as high (22%, n
= 491) as in genetically unmodified wild-type cells (20.8%, n
= 1,401; ).
The observed phenotype is consistent with our RNAi results in cell culture, which suggests a defect in nuclear import of phospho-Mad. Because we did not observe a considerable accumulation of cytoplasmic phospho-Mad, it is possible that in vivo the un-imported phospho-Mad is rapidly degraded or dephosphorylated. Although we cannot rule out other possibilities attributing to the observations in , it is clear that in vivo, cells with loss-of-function mutation in msk would have defects in phospho-Mad–mediated signaling.
The Msk orthologues in human are required for nuclear accumulation of Smad1 and transcriptional responses to BMP
Msk has two homologues in mammals, Imp7 and Imp8, each sharing over 50% identity in amino acid sequence with Msk. Imp7 and Imp8 themselves are ~60% identical. Based on in vitro assays, Imp7 has been suggested to import ribosomal proteins, histone H1, HIV reverse transcription complexes, and glucocorticoid receptor into the nucleus (Jakel and Gorlich, 1998
; Jakel et al., 1999
; Fassati et al., 2003
; Freedman and Yamamoto, 2004
). Imp8 was recently shown to support nuclear import of the signal recognition particle 19 (SRP19) in vitro (Dean et al., 2001
To investigate the roles of Imp7 and 8 in nuclear transport of Smads in mammalian cells, we designed siRNA duplexes that are effective in knocking down Imp7 or 8 individually (). Although Imp7 and 8 siRNA duplexes had no effects on BMP2-induced phosphorylation of Smad1 (), immunofluorescent staining with phospho-Smad1–specific antibody showed that knockdown of either Imp7 or 8 resulted in a more diffusive distribution of phospho-Smad1 after BMP2 stimulation, while in control siRNA transfected cells phospho-Smad1 was mostly present in the nucleus (). Corresponding to this defect in nuclear accumulation of Smad1, the BMP2-induced transcriptional activation of Smad6 was also suppressed in cells transfected with siRNA against either Imp7 or 8 (). Therefore, similar to their Drosophila counterpart, Imp7 and 8 are critical for nuclear translocation of BMP-activated Smad1 in mammalian cells.
Figure 4. Imp7 and 8 are required for nuclear accumulation of BMP-activated Smad1 in HeLa cells. (A) HeLa cells transfected with indicated siRNA duplexes (40 nM) were analyzed for mRNA levels of Imp7 or Imp8 by real-time RT-PCR. GAPDH was used as the internal standard (more ...)
Imp7 and 8 are required for nuclear import of TGF-β–activated Smad2/3 and transcriptional activation of their target genes
TGF-β and BMP pathways are similar in the general signaling mechanism, but differ in the receptor kinases and Smads that are used for signaling (Shi and Massagué, 2003
). We thus investigated if Imp7 and 8 are shared by TGF-β and BMP pathways in transporting different receptor-activated Smads into the nucleus. Again, knockdown of either Imp7 or 8 severely inhibited nuclear accumulation of Smad2 and 3 in response to TGF-β stimulation (). Such observation was made in both HeLa and HaCaT cells, and Smad2/3 phosphorylation in response to TGF-β was not affected by the same siRNA against Imp7 or 8 (). The block of Smad2/3 nuclear accumulation was also manifested in substantially reduced transcriptional activation of the TGF-β target gene Smad7 (). Therefore, Smads downstream of TGF-β also depend on Imp7 or 8 for nuclear translocation.
Figure 5. Imp7 and Imp8 are required for TGF-β–activated Smad2/3 to translocate into the nucleus. (A) HeLa or HaCaT cells transfected with indicated siRNAs (40 nM) were analyzed by immunostaining using anti-Smad2/3 antibody, with or without prior (more ...)
Because both Imp7 and Imp8 are required for nuclear transport of Smads, we next examined the relative contributions from these two. The efficiency of Smads nuclear translocation could be quantitated by counting the number of cells exhibiting “nucleus only” versus “cytoplasmic” distribution of Smads. We found that indeed when transfected at the same final concentration, combining siRNAs targeting Imp7 and 8 was more effective in inhibiting nuclear accumulation of Smad2/3 than individual siRNA against either Imp7 or 8 alone (). This suggests that Imp7 and 8 are likely to act in parallel in mediating nuclear translocation of TGF-β–activated Smad2/3.
Overexpression of Imp7 and 8 rescued siRNA effects
Because siRNAs against Imp7 and 8 are highly effective in blocking nuclear accumulation of Smad2/3, we examined if reintroducing Imp7 and 8 cDNAs would rescue the RNAi phenotype. To this end, we generated silent mutations in Imp7 and 8 sequences and verified that the mutants are no longer targeted by the Imp7 or Imp8 siRNA, respectively (not depicted; see Materials and methods). Such mutant cDNAs were transfected into HeLa cells 2 d after the siRNA transfection. Indeed, upon TGF-β stimulation, Smad2/3 accumulation in the nucleus was restored only in cells that expressed the rescuing HA-tagged Imp7 or 8 plasmids (). This result validated that the defects in Smads nuclear translocation observed in and were specifically due to depletion of endogenous Imp7 and 8.
Figure 6. Effects of Imp7 and 8 overexpression on Smad localization. (A) Overexpression of siRNA-insensitive Imp7 and 8 rescued the defect in Smad2/3 nuclear accumulation in response to TGF-β. HeLa cells were transfected with siRNAs targeting endogenous (more ...)
When overexpressed in HeLa cells, Imp7 was diffusively distributed throughout the cell while more Imp8 was detected in the nucleus than in the cytoplasm (). Such patterns of Imp7 and 8 localization remained the same upon TGF-β stimulation (). Overexpression of Imp7 or 8 had no detectable effects on the distribution of endogenous Smad2/3 at both basal and TGF-β–stimulated states (). This suggests that in HeLa cells, endogenous Imp7 and 8 are not limited in quantity to support nuclear translocation of Smad2/3.
Smads are direct nuclear transport substrates of Imp7 and 8
We next investigated Smads interactions with Imp7 and 8 by coimmunoprecipitation experiments. Flag-tagged Smad1 or Smad2 were overexpressed in 293T cells and immunoprecipitated with anti-Flag antibody. In both cases, HA-tagged Imp7 or Imp8 coimmunoprecipitated with either Smad1 or Smad2 (). Constitutively active BMP receptor (ALK3-QD) or TGF-β receptor kinase (ALK5-TD) was cotransfected to induce C-terminal phosphorylation of Smad1 and Smad2, respectively. Such phosphorylation of Smads did not affect their interaction with Imp7 or 8 (). Because phosphorylated Smad1, 2, and 3 readily assemble into complexes, our results suggest that monomeric and multimeric forms of Smads have similar interactions with Imp7 or 8 (Wu et al., 2001
; Chacko et al., 2004
). These observations are consistent with our finding in Drosophila
Figure 7. Interaction of Smads with Imp7 and Imp8, and the regulation by Ran-GTP. (A) Co-immunoprecipitation of Smad1 with Imp7 and Imp8. 293T cells were transfected with indicated expression plasmids and the whole-cell extract (WCE) was immunoprecipitated with (more ...)
For detailed analysis of Smad-Imp7/8 interaction, we focused on Smad3. We produced GST-fusions of the MH1, MH2, and linker plus MH2 domains of Smad3 in Escherichia coli
and tested their ability to pull down endogenous Imp7 and 8 in Hela cells. When comparable amount of GST fusion proteins were used, both the MH1 (aa 1–155) and the linker plus MH2 (aa 146–425) domains were able to bind endogenous Imp7/8, with the MH1 domain exhibiting stronger interaction (). The same assay barely detected any interaction between the Smad3 MH2 (aa 231–425) domain and Imp7/8 (). Therefore, interaction with Imp7/8 appears to involve multiple interfaces in the MH1 and linker regions of Smad3. The MH1 domains of Smad2 (aa 1–185) and Smad3 are highly similar except for two insertions in Smad2 that prevent Smad2 from binding to DNA (Zawel et al., 1998
). But apparently such differences did not affect Smad2 binding to Imp7/8 through the MH1 domain (). Bacterially produced GST-Imp8 was able to pull down purified recombinant Smad1 or Smad3, suggesting that Imp8 could directly interact with Smad1 or Smad3 ( and Fig. S2, available at http://www.jcb.org/cgi/content/full/jcb.200703106/DC1
One characteristic among importins is that the interaction with their cargoes is regulated by Ran in its GTP-bound form (Gorlich et al., 1996
; Mattaj and Englmeier, 1998
). To test if this is true between Smad3 and Imp8, we first pulled down HA-Imp8 using GST fusion of full-length Smad3. After washing off the unbound proteins, RanQ69L-GTP (the Q69L mutation locks Ran in its GTP-bound state) or BSA was added to the GST beads for further incubation (). Indeed we found that comparing to the BSA control, RanQ69L-GTP caused more release of Imp8 into the supernatant and correspondingly resulted in a decrease of Imp8 remaining bound to GST-Smad3 on the beads (). This suggested that association of Smad3 with Imp8 was disrupted upon binding of Ran-GTP, supporting the notion that Smad3 is a nuclear import cargo of Imp8.
Basal state Smads can enter the nucleus via Msk/Imp7/8-independent mechanisms
Without TGF-β stimulation, Smads undergo spontaneous nucleocytoplasmic shuttling and are distributed evenly in both the nucleus and cytoplasm in many types of cells (Inman et al., 2002
; Xu et al., 2002
; Nicolas et al., 2004
). Prompted by the observation that unphosphorylated Smads interact with Imp7 and 8, we examined if Imp7 and 8 are also required for basal state Smads import into the nucleus. In Drosophila
S2R+ cells, Flag-Mad was detected throughout the cells without exogenous Dpp (). The presence of Flag-Mad in the nucleus is not likely due to autocrine Dpp secreted by the cells, because further blocking any residual Mad phosphorylation by RNAi against punt
did not eliminate the presence of Mad in the nucleus (). Treatment with dsRNA targeting msk
also had no effect on the presence of Mad in the nucleus at basal state (). Because RNAi against punt
have been validated to be highly potent (), we concluded that nuclear import of basal state Mad does not rely on Msk, in contrast to Dpp-activated Mad.
Figure 8. Roles of Msk, Imp7, and Imp8 in nuclear import of Smads at basal state. (A) S2R+ cells transfected with Flag-Mad expression vector were subject to RNAi as indicated. After inducing the expression of Flag-Mad, the cells were analyzed by anti-Flag (more ...)
Similar to Drosophila cells, knockdown of Imp7 and Imp8 individually or in combination did not reduce the amount of Smad2/3 in the nucleus of unstimulated HeLa cells (). Therefore, in both Drosophila and mammalian cells, although Msk and Imp7/8 interact with basal state Smads, the presence of Smads in the nucleus without TGF-β stimulation is not critically dependent on Msk or Imp7/8.
Overexpressing Fox H1 in HeLa cells is another way to drive endogenous Smad2/3 into the nucleus without TGF-β stimulation (). Such nuclear accumulation of Smad2/3 could be due to nuclear sequestration of the shuttling Smad by the nucleus-bound Fox H1, and has been observed with other Smad-interacting transcription factors such as ATF3 (Kang et al., 2003
). Transfection of siRNA targeting Imp7 or Imp8, individually or combined, did not alter the “nucleus only” pattern of Fox H1, and did not affect Fox H1–induced nuclear concentration of Smad2/3 (). In control siRNA-transfected cells, 82.3% (n
= 34) of Fox H1–positive cells contained endogenous Smad2/3 predominantly in the nuclei, whereas in cells with double-knockdown of Imp7 and Imp8, 84.3% (n
= 32) of Fox H1–expressing cells have Smad2/3 in the nucleus. Thus, the above observations, from both Drosophila
and mammalian cells, led us to conclude that unphosphorylated Smads may be able to enter the nucleus through additional mechanisms, such as those described in previous studies (Xiao et al., 2000
; Kurisaki et al., 2001
; Xu et al., 2002