Inhibition of the Nuclear Import of Cubitus Interruptus by Roadkill in the Presence of Strong Hedgehog Signal

Hedgehog (Hh) signalling plays an important role in various developmental processes by activating the Cubitus interruptus (Ci)/Glioblastoma (Gli) family of transcription factors. In the process of proper pattern formation, Ci activity is regulated by multiple mechanisms, including processing, trafficking, and degradation. However, it remains elusive how Ci distinctly recognizes the strong and moderate Hh signals. Roadkill (Rdx) induces Ci degradation in the anterior region of the Drosophila wing disc. Here, we report that Rdx inhibited Ci activity by two different mechanisms. In the region abutting the anterior/posterior boundary, which receives strong Hh signal, Rdx inhibited the nuclear import of Ci by releasing importin α3 from Ci. In this region, Rdx negatively regulated the expression of transcription factor Knot/Collier. In farther anterior regions receiving moderate levels of Hh signal, Rdx induced Ci degradation, as reported previously. Thus, two different mechanisms by which Rdx negatively regulates Ci may play an important role in the fine-tuning of Hh responses.

Binding of Hh to the cell surface protein Patched (Ptc) abrogates its inhibition on another cell surface protein, Smoothened (Smo) by inducing a conformational switch [13]. This induces phosphorylation of the cytoplasmic tail of Smo by protein kinase A (PKA) and casein kinase I (CKI), leading to activation of signal transduction [14][15][16][17]. Hh signaling acts through the zinc finger transcription factor Cubitus interruptus (Ci) expressed in A compartment cells [2,3]. Hh signal regulates Ci through multiple levels of regulation, including processing, nuclear-cytoplasmic partitioning, and degradation.
In A compartment cells far from the A/P compartment border which receive no Hh signal, full-length Ci (Ci-155) is phosphorylated by multiple kinases including PKA, which leads to generate a C-terminally truncated form [18][19][20][21][22][23][24]. The F-box/WD40 repeatcontaining protein, Slimb, whose vertebrate homolog FWD1/ bTRCP is a component of the so-called SCF (Skpl, Cdc53, and F-box) ubiquitin ligase complex, targets phosphorylated Ci-155 for ubiquitination followed by proteasome-mediated proteolysis. Ci-75 is localized in the nucleus and suppresses the expression of Hhresponsive genes such as dpp by acting as a repressor [25. 26]. On the other hand, in A compartment cells abutting the A/P boundary, Hh signaling prevents the processing of Ci and stimulates the activity of Ci-155 that has accumulated in the cell [18,27].
Hh signaling also cancels the inhibitory effects of the kinesin-like protein Costal2 (Cos2) [28][29][30][31][32], and Suppressor of fused (Su(fu)) [33,34]. Ci-155 forms complexes with the Cos2, Su(fu), and the Ser/Thr kinase Fused (Fu) that docks at microtubles. Although the tight binding of this complex to microtubles blocks the nuclear entry of Ci, activation of Fu in response to Hh signalling negates the inhibitory role of Cos2 and Su(fu) [35,36]. After its nuclear entry, Ci-155 interacts with the transcriptional coactivator CBP to induce the transcription [37]. A biochemically uncharacterized CiA is released from the regulatory complex and begins to activate transcription, while levels of Ci155 drop [35].
The level of Ci155 is critical for correct responses to Hh, and the Ci turnover is regulated by multiple pathways. In the cells receiving very low Hh signal, Debra induces the poly-ubiquitination of phosphorylated Ci-155 in cooperation with Slimb, leading to its lysosomal degradation [38]. Debra is expressed in the band abutting the border of the Hh-responsive A compartment region of wing disc, indicating that it defines the borders of the Hhresponsive region. In the cells receiving moderate level of Hh signal, Ci-155 level is negatively regulated by Cul3-mediated degradation [39]. Roadkill (Rdx) (also called HIB) is the substrate recognition component of a Cul3 E3 ubiquitin ligase that targets Ci for degradation [40,41]. Rdx contains the BTB and MATH domains that bind to Cul3 and the substrates, respectively [42,43]. Rdx is induced by Hh signaling and is expressed in the Hhresponsive cells in the A compartment of wing disc, suggesting the presence of negative feedback regulation that attenuates Hh responses. Su(fu) inhibits the Rdx-mediated degradation of Ci-155 by competing with Rdx for binding to Ci [41].
Nuclear import and export of Ci-155 also play an important role to regulate Hh signaling. Hh signaling enhances the nuclear import of Ci155 by releasing Ci-155 from the complex with Cos2, Su(fu), and Fu at microtubles [28,29,33,34,44]. In fact, overexpression of Su(fu) reduces the nuclear accumulation of Ci-155. On the other hand, Su(fu) was shown to enter into the nucleus with Ci under some context [45]. Furthermore, Su(fu) inhibits the activity of Ci through a mechanism independent of Ci nuclear translocation [36].
Here, we report that Rdx inhibits Ci nuclear import in the cells receiving strong Hh signal, although it induces Ci degradation in the cells receiving the moderate levels of Hh signal. Dual mechanism of Rdx action may play an important role in finetuning Hh responses.

Results
Negative regulation of kn expression by Rdx in the region close to the A/P boundary Using Rdx-lacZ, we have first monitored the precise pattern of Rdx expression in wing discs. The expression levels of Rdx were high in 2-4 cells anterior to the A/P boundary, while they were moderate in a more anterior region that is several cells-wide ( Figure 1A, a and d; Figure S1) as reported previously [40,41]. In the region abutting the A/P boundary, where Rdx is expressed at its highest level, Ci-155 levels were low ( Figure 1A, b-d). In contrast, Ci-155 levels were high in the more anterior regions, which express moderate levels of Rdx ( Figure 1A, b-d).
We isolated Rdx mutants using ethyl methanesulfonate treatment. One isolated allele, rdx 11 , had a nonsense mutation at Q120 in the MATH domain ( Figure 1B). Since this mutant encoded the N-terminal 119 amino acids protein fragment, which lacks both the intact MATH and BTB domains, this mutation was thought to be a null allele. The classical mutant/deficiency test suggests that rdx 11 is a null allele ( Figure 2A). Furthermore, accumulation of Ci-155 was observed in the clones of rdx 11 mutant cells ( Figure 2B). Similar accumulation of Ci-155 was also observed in the clones of hib D6 , which was a null mutant of rdx isolated by Jiang's group, and also in the clones of Df(3R)Exel6171, a deficiency line [41]. These results also support the notion that rdx 11 is a null mutant. In the rdx 11 homozygous clones, expression of decapentaplegic (dpp) and patched (ptc), both of which are Ci target genes [1,2], was not up-regulated in 2-4 cells anterior to the A/P boundary ( Figure S2A, B), although dpp expression levels were enhanced in the more anterior region, where Rdx is expressed at moderate levels ( Figure S2A-b, arrow). Kent et al. described that the rdx loss-of-function clones had no discernible effect on expression of ptc and dpp, although no data was shown [40]. The difference between their observations and ours could be due to the use of different alleles. Kent et al. used rdx 5 , which had two missense mutations, Q193H and D653V, and also rdx 6 , which contained a point mutation in the splice donor between exons 12 and 13. The rdx 6 allele generated mRNAs retaining intron 12/13 as well as wild-type mRNA. On the other hand, the rdx 11 allele we used had a nonsence mutation at Q120 in the MATH domain. Thus, rdx 5 and rdx 6 seem to be a hypomorphic allele, while rdx 11 is thought to be a null allele. Zhang et al., also reported that ptc was overexpressed in the hib/rdx mutant clones [41]. They used the hib D6 allele, which was generated by replacing the hib/rdx coding region by the white gene, indicating this allele is a nulle allele. However, they observed ptc overexpression only in the presence of overexpressed Hh. So, there was no data indicating that ptc was overexpressed in the hib/ rdx clones without overexpressed Hh.
In contrast to ptc, the expression level of knot/collier (kn/col), another Ci target gene in the cells abutting the A/P boundary [12], was apparently enhanced in the rdx 11 clones ( Figure 1C and S3). Knot (Kn) is a member of the COE (Co/Olf-1/EBF) family of transcription factors [46], and controls the formation of the central intervein by inducing the expression of Blistered (the Drosophila homolog of serum response factor) in cooperation with Ci [12]; thus, Rdx negatively regulated the expression of kn in the cells close to the A/P boundary which receive the strong Hh signal.
Rdx does not induce Ci-155 degradation in the presence of a strong Hh signal, but enhances the cytosolic retention of Ci-155 Low levels of Ci-155 in 2-4 cells anterior to the A/P boundary were not further downregulated in the clones overexpressing Rdx ( Figure 3A and S4A, arrow). In contrast, Ci-155 levels were reduced in cells ectopically expressing high levels of Rdx in the more anterior region, as reported previously [40,41]. Ci-155 levels were not affected in the rdx 11 clones in 2-4 cells anterior to the A/ P boundary ( Figure 3B and S4B, arrow), although Ci-155 levels were increased by the loss of Rdx in the more anterior region, which is consistent with a previous report [41]. These results suggest that Rdx did not induce Ci degradation in the anterior narrow region abutting the A/P boundary, although it induced Ci degradation in the more anterior region.
To assess the role of Rdx in the presence of strong Hh signals, effect of Rdx on the subcellular localization of Ci was examined. Coexpression of Ci-155 with Rdx and Hh in Hh-responsive clone-8 cells led to an increase in the population of cells harbouring Ci-155 predominantly in the cytoplasm ( Figure 4A). In addition, treatment of transfected cells with leptomycin B (LMB), an inhibitor of nuclear export, brought about an evident enhancement of the ability of Rdx to retain Ci in the cytoplasm ( Figure 4B). In the absence of Hh, Rdx did not dramatically increase the population of cells harbouring Ci-155 predominantly in the cytoplasm ( Figure 4C). Although Rdx induces degradation of nuclear Ci-155 in the absence of Hh [40,41], significant amounts of nuclear Ci may be retained in the presence of LMB, so that the subcellular localization of Ci might not be affected. Rdx more efficiently reduced the population of cells harbouring Ci-155 predominantly in the nucleus in the presence of higher levels of Hh ( Figure 4D). Similar results were observed using a Ci mutant, in which the nuclear export sequence was mutated ( Figure 4E); thus, in the presence of strong Hh signal, Rdx enhanced the cytosolic retention of Ci-155.

Rdx inhibits the nuclear import of Ci-155 only in the presence of strong Hh signal
To understand the mechanism by which Rdx enhanced the cytosolic retention of Ci-155, we have examined the subcellular localization of Rdx in clone-8 cells. Rdx was detected in the nuclear periphery region in an immunostaining experiment using Triton X-100-treated clone-8 cells ( Figure 5A, lower). However, Rdx was not detected in the cells treated with digitonin, which can permeabilize the plasma membrane but not the nuclear membrane ( Figure 5A, upper), indicating that Rdx is localized at the inside of the nuclear membrane. Rdx signal was detected in speckle-like structures located on the inside of the nuclear membrane, and overlapped with nuclear pores immunostained using the monoclonal antibody 414 ( Figure 5B), which recognizes a common nucleoporin epitope [47,48]. This suggests that Rdx may regulate nuclear import of Ci.
To test this hypothesis, fluorescence recovery after photo bleaching (FRAP) experiments were performed using Ci-155 fused to GFP (Ci-GFP). After bleaching Ci-GFP fluorescence in the nucleus, compartmental equilibration (which occurs via nuclear import) was measured by time-lapse fluorescence microscopy. When Ci-GFP was coexpressed with Hh in clone-8 cells, the rate of nuclear import of Ci-GFP was 2.460.5 s 21 /100 ( Figure 5C, left and Figure S5A, B). Coexpression of Ci-GFP with Rdx apparently reduced the rate of nuclear import of Ci-GFP, and the rate of nuclear import was 1.960.5 s 21 /100. The significant effect of Rdx on the amount of nuclear Ci-GFP was observed within 10 sec after photpbleaching, indicating that this was not due to the Rdx-induced degradation of Ci, because the turnover rate of Ci is not so rapid. Thus, Rdx decreased the rate of nuclear import of Ci-GFP in the presence of strong Hh signal (P,0.05). When the FRAP experiments were performed in the absence of the Hh expression vector, the rate of nuclear import of Ci-GFP was similar in the presence and absence of Rdx ( Figure 5C, right and Figure S5C), indicating that Rdx did not affect the nuclear import of Ci-GFP; thus, Rdx negatively regulates the nuclear import of Ci-155, but only in the presence of a strong Hh signal. To further confirm that the effect of Rdx in the FRAP experiment is not due to the Rdx-induced degradation of Ci-155, similar experiments were performed in the presence of MG132, an inhibitor of proteasome. Even in the presence of MG132, Rdx decreased the rate of nuclear import of Ci-GFP in the presence of strong Hh signal (P,0.05) ( Figure 5D). The decrease in the saturated level of relative nuclear Ci-GFP intensity may be due to increased levels of cytosolic Ci-GFP by MG132. To examine the effect of Rdx on the nuclear export of Ci-155, fluorescence loss in photobleaching (FLIP) was used. Similar Ci-GFP nuclear export patterns were observed in the presence or absence of Rdx ( Figure 5E and Figure  S6), indicating that Rdx does not affect the nuclear export of Ci-155.
Rdx inhibits the nuclear import of Ci-155 by competing with importin a3 for binding to Ci-155 When we screened the Ci-binding proteins using a yeast twohybrid assay, we have also identified importin a3, in addition to Rdx (see Materials and methods). Furthermore, another group also identified importin a3 in the two-hybrid screening using Ci as a bait [49]. Importin a3 mediates the nuclear import of multiple proteins [50]. To examine whether importin a3 interacts with Ci in vivo, co-immunoprecipitation assays were performed. Importin a3 was co-immunoprecipitated with Ci-155 in lysates of clone-8 cells ( Figure 6A).
To investigate whether knock down of importin a3 affects the nuclear import of Ci, we first tried a down-regulation of importin a3 using double-starnded RNA (ds-RNA). Transfection of the importin a3 ds-RNA into clone-8 cells decresaed the levels of importin a3 to about 1/10 of the control ( Figure 6B). Knock down  Figure 6C). These results indcate that importin a3 binds to Ci and acts as a nuclear import receptor for Ci.
Importin a3 overexpression did not enhance the nuclear accumulation of Ci-155 ( Figure 7A, left). This may be due to the presence of saturated amount of endogenous importin a3 in clone-8 cells. When importin a3 was coexpressed with Rdx, however, the Rdx-dependent inhibition of the nuclear import of Ci-155 was counteracted by importin a3 (Figure 7A, right). These results suggest a competitive action of Rdx and importin a3. To directly assess the competitive action of Rdx and importin a3, the effect of Rdx on the Ci-importin a3 interaction was examined. Importin a3 was co-immunoprecipitated with C-155 in the absence of exogenously expressed Rdx, whereas this interaction was abrogated by overexpressed Rdx ( Figure 7B). Thus, Rdx competes with importin a3 for binding to Ci-155. These results suggest that Rdx may inhibit the nuclear import of Ci by competing with importin a3 in the presence of a strong Hh signal. If Rdx inhibits the nuclear entry of Ci-155 by competing with importin a3, furthermore, Rdx would also block the nuclear entry of importin a3 in the presence of Ci-155. To test this, we have also examined whether Rdx affects the nuclear entry of importin a3 itself. Rdx also inhibited the nuclear accumulation of importin a3 when importin a3 was coexpressed with Ci-155 ( Figure 7C and 7D).

Discussion
In the region close to the A/P boundary, strong Hh signal induces the high level of Rdx expression, and Rdx inhibits the nuclear import of Ci-155 ( Figure 8). In the more anterior region, moderate level of Hh signal induces the moderate level of Rdx, and Rdx induces the proteasome-dependent degradation of Ci-155. Why Rdx uses different mechanism to inhibit Ci activity . Similar accumulation of Ci-155 was also observed in the clones of hib D6 , which was a null mutant of rdx isolated by Jiang's group, and also in the clones of Df(3R)Exel6171, a deficiency line (Zhang et al., 2006). These results also support the notion that rdx 11 is a null mutant. doi:10.1371/journal.pone.0015365.g002 depending on the Hh signal strength? In the region close to the A/ P boundary, Ci level is low and a limited number of Ci target genes, such as kn, is regulated by Rdx. Kn is a member of the COE (Co/Olf-1/EBF) family of transcription factors [46], and controls the formation of the central intervein by inducing the expression of Blistered in cooperation with Ci [12]. To selectively regulate a limited number of target genes, maintaining nuclear Ci at low levels may be important. To do this, inhibition of nuclear import could be more efficient than degradation in the nucleus. Thus, switching the dual function of Rdx depending on the Hh signal strength may exhibit the elegant regulatory system in fine-tuning Hh response. Ci-155 binds to importin a3, which forms a heterodimer with importin b [51], suggesting that the ternary complex containing Ci-155 and importin a3/b is transported into the nucleus via the nuclear pore complex (NPC). The NPC that is composed of about 30 different proteins termed nucleoporins (nups) is a cylindrical structure [52]. The nups containing phenylalanine-glycine (FG) repeats make up about half the mass of the NPC [52] and the hydrophobic FG repeats interact weakly with each other to form a ''selective sieve'' [53]. The hydrophobic nature of importins enables them to enter the sieve and travel via diffusion along an affinity gradient of nup-binding sites, encountering nups of increasing affinity during their translocation. After translocation through the NPC, the Ci-155-importin a3/b complex encounters RanGTP, leading to cargo association and release for termination of the transport cycle. Rdx is colocalized with nups at the NPC, suggesting that it binds to Ci-155 during the Ci-155-importin a3/ b complex travels in the FG repeats sieve. Binding of Rdx to Ci-155 would release importin a3 from Ci-155, resulting to exclude Ci-155 from the FG repeats sieve. Thus, localization of Rdx at the NPC is essential to block the nuclear import of Ci-155. Although majority of Rdx molecules are at the NPC in clone-8 cells, Rdx was also detected in the specific region of nucleoplasm. Therefore, Rdx localized in the nucleoplasm may lead to the proteasomedependent degradation of Ci-155, while Rdx at the NPC may inhibit the nuclear import of Ci-155, although we cannot exclude the possibility that Rdx at the NPC regulates both the degradation and nuclear import of Ci-155. Zhang et al. showed that Su(fu) competes with Rdx for binding to Ci-155, and negatively regulates the Rdx-dependent degradation of nuclear Ci-155 [41]. Su(fu) was previously shown to retain Ci-155 in the cytoplasm [33,34]. However, we have observed that in the presence of Rdx, Su(fu) enhances the nuclear entry of Ci-155 (data not shown). This suggests that Su(fu) could enhances the nuclear entry of Ci-155 by competing with Rdx for binding to Ci-155. Further analyses will be required for understanding the role of Su(fu) in the Rdx-dependent regulation of Ci-155.

Yeast two-hybrid screening
The yeast two-hybrid assay was carried out according to the methods described by Vojtek et al. [54]. The "bait" was a Ci fragment (amino acids 1-441) fused to the LexA DNA-binding domain. The Drosophila whole adult cDNA library (Clontech) was used. Among the 81 clones isolated, 32 clones were derived from the rdx gene, while 7 clones enclosed importin a3.
Isolation of rdx mutants w 1118 males were fed 25 mM ethyl methanesulfonate (EMS) and then mated with w; Pr, Dr/TM6B, Tb, Sb virgin females. The TM6, Tb balanced F1 males were used to cross single male lines with virgin females of genotype w; Df(3R)red-P52/TM3, Ser, Sb. Eleven thousand four hundred mutagenized third chromosomes were used for the screening and 20 recessive lethal candidate lines were established. Candidate lines were also crossed with meiP19 M2 / TM6B, Tb, Sb (rdx enhancer trap line and recessive lethal mutant of rdx) [41] and Df(3R)293gamma 7 /TM6B, Tb, Sb for complementation tests, and seven lines were identified as recessive lethal mutants of the rdx locus. These lines were backcrossed to w 1118 six times.

Plasmid construction
The full-length rdx cDNA, which encodes the B isoform of CG9924, was cloned from the wh Drosophila ole adult cDNA library (Clontech). The importin a3 cDNA was obtained from S. Cotterill [55]. The plasmids to express the following proteins were constructed using a vector containing the Drosophila act5C promoter: Hh, HA-Ci-155, 3xFlag-Ci-155, HA-Ci-155 (L768A), HA-Rdx, 2xFlag-Rdx and Myc-Importin a3. In these constructs, the HA-, Flag-, or Myc-tag was placed at N-terminus. The mutant of the nuclear export sequence (NES) of Ci (L768A) was constructed using a PCR-based method.

UAS-rdx construct and germ line transformation
The 2.1 Kb full-length rdx cDNA was cloned between the EcoRI and NotI sites of the pUAST vector [62]. This construct was co-injected with the D2-3 helper plasmid into the w 1118 host line.

Immunohistochemistry
Imaginal discs from Drosophila third instar larvae were fixed and stained using standard techniques. Primary antibodies used in this study were: rat anti-Ci (2A1) (

Fluorescence Recovery after Photobleaching (FRAP) and Fluorescence Loss in Photobleaching (FLIP) analysis
Clone-8 cells (2610 6 cells/2.5 cm dish) were transfected with a mixture of the following plasmids: for Figure 5C (left), 5D and S5B, act5c-hh (500 ng), UAS-Ci-GFP (23) (200 ng), act5C-GAL4 (50 ng), act5C-HA-rdx (0 or 500 ng), and UAS-Su(fu) (0 or 400 ng); for Figure 5E and S6, act5C-hh (500 ng), UAS-Ci-GFP (200 ng), act5C-GAL4 (100 ng), and act5C-HA-rdx (0 or 500 ng); for Figure 5C (right) and S5C, UAS-Ci-GFP (100 ng), act5C-GAL4 (100 ng), and act5C-HA-rdx (0 or 500 ng). The total amount of DNA was adjusted to 1.65 mg by adding the empty vector. Cells were observed 48 h after transfection. FRAP and FLIP analyses were performed using a 1006 oil immersion objective, the 36zoom on the Zeiss LSM510 confocal microscope and the 488 nm line of a Kr/Ar laser operating at 85% laser power and 2% transmission. The stage temperature was maintained at 25uC. For FRAP experiments, a specific region of the nucleus was selected and bleached using 100 iterations with 100% transmittance. Images were recorded at 1 sec intervals. Cells were imaged 50 times before photobleaching and 50 times afterwards, with maximum scan speed. For FLIP analysis, a region of the cytoplasm around the nucleus was selected and bleached by 50 iterations after imaging scans (10 times before photobleaching and 20 times after photobleaching). Images were acquired at 10 sec intervals. For every protein, or combination of proteins, at least 8 FLIP or FRAP curves were generated per transfection. For data analysis, the LSM software (Zeiss) was used to calculate the mean fluorescence intensity of the nuclear region. A fluorescence recovery curve was drawn using the Excel software, according to Rabut and Ellenberg [63]. The nuclear import rate of Ci-GFP was calculated as the slope of a linear regression fit of the initial phase of the flux curve (at least 5 post-bleach points).

Western blotting and co-immunoprecipitation
To study the interaction of Importin a3 with Ci-155, clone-8 cells (4610 6 cells/6 cm dish) were transfected with the plasmid to express Hh (1 mg), HA-Ci-155 (1.5 mg), Myc-Importin a3 (1 mg), and 2xFlag-Rdx (1 mg for competition assay) using Cellfectin. Forty-eight hours after transfection, cells were lysed in 300 ml lysis buffer (50 mM HEPES (pH 7.4), 250 mM NaCl, 0.5% NP-40, 0.2 mM EDTA, protease inhibitor mixture). The lysates were diluted to 150 mM NaCl, and then immunoprecipitated with rat anti-HA (3F10) antibody or control normal rat IgG. The immunoprecipitates were analyzed by In region 1 abutting the A/P boundary, which receives strong Hh signal, the Rdx levels are high. In this region, Rdx inhibits the nuclear import of Ci-155 by competing with importin a3 for binding to Ci-155 at the nuclear pore complex. This maintains the nuclear Ci-155 at low levels. Low levels of nuclear Ci-155 selectively activate the transcription of kn. Low levels of nuclear Ci-155 may be central to the selective induction of specific Hh target genes, such as kn. On the other hand, in region 2 distant from the A/P boundary, which exhibits moderate levels of Hh signal, the Rdx levels are moderate. In this region, the nuclear import of Ci-155 is allowded, which maintains the nuclear Ci-155 at relatively high levels; however, the nuclear Ci-155 level is negatively regulated by Rdx via a proteasome-dependent degradation. The relationship between Hh signal, Rdx and Ci is shown below. Strong Hh signal induces Rdx expression, which inhibits the nuclear entry of Ci-155. doi:10.1371/journal.pone.0015365.g008 immunoblotting with either a rat or mouse anti-HA antibody, a mouse anti-Myc antibody and a mouse anti-Flag antibody.