Deficit of heat shock transcription factor 1-heat shock 70 kDa protein 1A axis determines the cell death vulnerability in a model of spinocerebellar ataxia type 6

Authors


  • Communicated by: Shuh Narumiya

* t-tanabe.mphm@tmd.ac.jp

Abstract

Spinocerebellar ataxia type 6 (SCA6) is caused by a small expansion of polyglutamine (polyQ)-encoding CAG repeat in Cav2.1 calcium channel gene. To gain insights into pathogenic mechanism of SCA6, we used HEK293 cells expressing fusion protein of enhanced green fluorescent protein and Cav2.1 carboxyl terminal fragment (EGFP-Cav2.1CT) [L24 and S13 cells containing 24 polyQ (disease range) and 13 polyQ (normal range), respectively] and examined their responses to some stressors. When exposed to CdCl2, L24 cells showed lower viability than the control S13 cells and caspase-dependent apoptosis was enhanced more in L24 cells. Localization of EGFP-Cav2.1CT was almost confined to the nucleus, where it existed as speckle-like structures. Interestingly, CdCl2 treatment resulted in disruption of more promyelocytic leukemia nuclear bodies (PML-NBs) in L24 cells than in S13 cells and in cells where PML-NBs were disrupted, aggregates of EGFP-Cav2.1CT became larger. Furthermore, a large number of aggregates were formed in L24 cells than in S13 cells. Results of RNAi experiments indicated that HSPA1A determined the difference against CdCl2 toxicity. Furthermore, protein expression of heat shock transcription factor 1 (HSF1), which activates HSPA1A expression, was down-regulated in L24 cells. Therefore, HSF1-HSPA1A axis is critical for the vulnerability in L24 cells.

Introduction

Spinocerebellar ataxia type 6 (SCA6) is an autosomal dominant neurological disorder and it is characterized by late-onset progressive ataxia and selective loss of cerebellar Purkinje cells (Kordasiewicz & Gomez 2007). The disease-causing mutation has been identified as an expansion of CAG trinucleotide repeat (normal range, 4–18; disease range, 19–33), which resides in the exon 47 of CACNA1A gene encoding Cav2.1 subunit of the P/Q type voltage-dependent calcium channels (Zhuchenko et al. 1997). To elucidate the molecular mechanism underlying SCA6 pathogenesis, several groups have examined the properties of Cav2.1 channel carrying SCA6 mutations using heterologous expression systems. But the results of these studies were not consistent: gain-of-function effects of SCA6 mutations were detected in some cases and loss-of-function effects in others (Matsuyama et al. 1999; Restituito et al. 2000; Toru et al. 2000; Piedras-Renteria et al. 2001). Moreover, we have recently reported that properties of human Cav2.1 channel carrying an SCA6 mutation expressed in cerebellar Purkinje cells, the important target for SCA6 pathogenesis, were not significantly different from those of the channel carrying a normal range of glutamine repeat, suggesting that SCA6 may not be a channelopathy (Saegusa et al. 2007). Watase et al. (2008) also reached a similar conclusion by establishing another SCA6 knockin mouse model. Therefore, it may be important to regard SCA6 as a polyglutamine (polyQ) disease. In this sense, it is interesting that the carboxyl terminal region of Cav2.1 (Cav2.1CT) is cleaved from the full-length protein and translocated to the nucleus (Kordasiewicz et al. 2006), because in many of the polyQ diseases, disease-causing proteins are accumulated in the nucleus, where they exert toxic effects (Orr & Zoghbi 2007). Actually, several groups reported that Cav2.1CT showed toxic effects on cultured cells. Cav2.1CTs carrying 13Q and 28Q expressed in HEK293T cells have induced significantly more cell death than full-length Cav2.1, although the toxicity is not dependent upon the length of the polyQ tract (Kubodera et al. 2003). Kordasiewicz et al. (2006) reported that transient expression of Cav2.1CT exerts toxic effects on HEK293-derived cells. Moreover, this toxicity was dependent on the length of the polyQ tract and also on the nuclear localization of the Cav2.1CT fragment. However, Marqueze-Pouey et al. (2008) reported that the polyQ-expanded Cav2.1CT induced cell death in COS-7 cells but that the cell death was not dependent on the nuclear localization of the Cav2.1CT fragment. There seem to be some discrepancies, but results of these studies are consistent in that the Cav2.1CT is toxic to cells. It is therefore possible that pathogenesis of SCA6 may be related to the toxic effects of the Cav2.1CT. However, the detailed mechanism of the toxicity exerted by the Cav2.1CT is totally unknown.

To explore the mechanism of the toxicity of the Cav2.1CT, we have used a cellular SCA6 model that we have established previously (Ishikawa et al. 1999). This cell line is derived from HEK293 cells and expresses fusion protein of enhanced green fluorescent protein (EGFP) and Cav2.1CT carrying an expanded polyQ (designated as L24 cells, carrying 24 repeat of polyQ). As SCA6 is a late-onset chronic disorder, cellular models that express permanently the Cav2.1CT with expanded polyQ may reflect the pathophysiology of SCA6, compared with the transient cellular models. Therefore, we used the permanent cellular model of SCA6. Ideally Purkinje cells are suitable for exploring the pathogenic mechanisms of SCA6. However, Purkinje cells, especially for the screening purpose, are difficult to handle. Actually most of the biochemical studies published previously had been carried out using cell lines. We therefore used HEK293-derived cell lines in the present study. It would be essential to confirm our present results using the SCA6 knockin mouse Purkinje cells in the future.

Under normal culture conditions, L24 cells and control cells that express Cav2.1CT carrying 13Q (named S13 cells) did not show any conspicuous differences, especially in terms of cell death. We therefore treated them with some harmful stimuli and found that CdCl2 treatment yielded a marked difference in cell viability. In the present study, we have analyzed the molecular mechanism that underlies the vulnerability of L24 cells.

Results

Expression of EGFP-Cav2.1CT containing polyQ tract in permanent cell lines

We have previously established HEK293-derived cell lines (named C13, S13 and L24) that express fusion proteins of EGFP and human Cav2.1 carboxyl terminal region (EGFP-Cav2.1CT) under the control of CMV promoter (Ishikawa et al. 1999). S13 and L24 cells express EGFP-Cav2.1CT carrying 13 and 24 polyQ repeat, respectively. C13 cells express EGFP-Cav2.1CT lacking a polyQ tract, owing to the lack of five nucleotide GGCAG insertion at the sequence corresponding to the exon 46/47 boundary of CACNA1A (Zhuchenko et al. 1997). HEK-EGFP is a control cell line that expresses non-fusion EGFP. When we established these cell lines, we obtained several lines for each construct. However, in all the S13 and L24 lines, expression levels of the fusion proteins were lower compared with C13 and HEK-EGFP (Fig. 1). Compared with the expression level of the EGFP in the HEK-EGFP cells, expression level of EGFP-Cav2.1CT in S13 and L24 cells was as low as 1/30. This situation may be consistent with the previous studies, where transient expression of the Cav2.1CT containing a polyQ tract in cultured cells exerted toxic effects on the cells (Kubodera et al. 2003; Kordasiewicz et al. 2006; Marqueze-Pouey et al. 2008). Owing to the toxic effects of the Cav2.1CT, only the cells that express low level of EGFP-Cav2.1CT seemed to be able to survive.

Figure 1.

 Expression levels of EGFP-Cav2.1CT fusion proteins in C13, S13 and L24 cells. (a) Proteins (40 μg/lane) from all the cell lines were analyzed by western blotting using anti-GFP antibody. The blot was reprobed with anti-GAPDH antibody for normalization. (b) Intensities of the specific bands shown by arrowheads in (a) were quantified with FluorImager 595. Note that protein expression levels in S13 and L24 cells are almost the same. Results are expressed as mean ± SEM (n = 4).

Cadmium toxicity is enhanced in cells expressing Cav2.1CT

A recent report suggests that carboxyl terminal region of Cav2.1 channel is cleaved and translocates to the nucleus, where the cleaved fragment is speculated to be involved in nuclear functions such as regulation of gene expression (Kordasiewicz et al. 2006; Kordasiewicz & Gomez 2007). As there is a polyhistidine repeat sequence, a well-known metal binding motif, in the Cav2.1CT (Mori et al. 1991; Terpe 2003), we aimed at examining whether the function of the Cav2.1CT was altered in the presence of the metal ions. It is because transition metal compounds such as those of cadmium, cobalt and nickel are known to interfere with transcription (Hartwig 2001). First, we have checked whether Cav2.1CT actually binds metal ions. The results showed that Cav2.1CT can actually bind metal ions such as Ni2+ and Cd2+ but that the affinity of the Cav2.1CT to these metal ions is not significantly altered by the polyQ tract length (Fig. S1 in Supporting Information). We then cultured the cells in the presence of various metal compounds and noticed that CdCl2 treatment led to marked differences in cell death among the cell lines. When C13, S13, L24, HEK-EGFP and parental HEK293 cells were cultured in medium containing 1 μm CdCl2 for 24 h, cell viability of each line was more than 80% with no significant differences among the cell lines. But when the cadmium concentration was increased to 3–7 μm, dose-dependent cell death was observed in all the cell lines. Among them, L24 cells were most sensitive to Cd2+ and the viability was significantly lower than the other cell lines at 3–7 μm CdCl2 (Fig. 2a,b). S13 cells, expressing EGFP-Cav2.1CT carrying a normal range of polyQ repeat, also showed enhanced CdCl2 toxicity. The viability of S13 cells was significantly lower than that of HEK-EGFP and HEK293 cells, when the CdCl2 concentration was 3–7 μm. But it was significantly higher than the viability of L24 cells in this range of CdCl2 concentration (Fig. 2b). C13 cells also showed significantly lower viability compared with HEK-EGFP cells (at 7 μm CdCl2) and to HEK293 cells (at 5 and 7 μm CdCl2). Therefore, the Cav2.1CT fragments enhance cadmium toxicity in a polyQ length-dependent manner.

Figure 2.

 Cadmium-induced cell death is enhanced with the increments of polyQ tract length. (a) Differential interference contrast images of cells treated with various concentrations (1–7 μm) of CdCl2. Scale bar, 20 μm. (b) Cell viability determined by trypan blue exclusion. Four independent experiments were performed and data are presented as mean ± SEM. Tukey–Kramer test was used to evaluate the statistical significance.

As we aimed at comparing effects of expanded and normal polyQ on various parameters, S13 and L24 were mainly used in the further experiments, with the S13 cells as a control.

First, to rule out the possibility that the different viabilities observed in S13 and L24 cells are because of non-specific clonal variations, we examined several independent clones. We used another independent S13 line (named S13-2) and an L24 line (L24-1) in addition to the above-mentioned S13 and L24 cells (original name was S13-1 and L24-5, respectively). S13-2 cells expressed almost the same amount of EGFP-Cav2.1CT(Q13) as S13-1 and also showed similar viabilities, when treated with CdCl2, compared with S13-1 (data not shown). As for L24, expression level of EGFP-Cav2.1CT is lower in L24-1 than L24-5. Consistent with this, L24-1 cells showed higher degree of survival against Cd toxicity compared with L24-5 cells but still lower viabilities, compared with S13 cells (data not shown). We selected L24-5 cells as L24 cells in this study, because expression levels of the exogenous proteins in L24-5 and S13-1 (S13-2) were similar.

We also performed RNAi experiments to knock down the EGFP-Cav2.1CT expression in these cells (Fig. S2 in Supporting Information). S13 and L24 cells were transfected with the targeted or negative control siRNA and the silencing effects were confirmed by Western blot analyses (Fig. S2A in Supporting Information). Then we examined the viabilities of siRNA-treated S13 and L24 cells 24 h after treatment with 3 μm CdCl2. The viability of L24 cells transfected with the targeted siRNA was significantly increased compared with the cells treated with negative control siRNA, reaching the same level of viability of S13 cells (Fig. S2B in Supporting Information). These results suggest that the lower viability in CdCl2-treated L24 cells was caused by toxic effects of Cav2.1CT(Q24).

Caspase-dependent apoptotic cell death was enhanced in cadmium-treated L24 cells

It has been already reported that HEK293 cells undergo apoptosis when treated with CdCl2 by way of caspase-dependent and -independent pathways (Mao et al. 2007). We then carried out immunocytochemical analyses using antibodies against apoptosis-inducing factor (AIF, a marker of caspase-independent apoptosis) and caspase 3 (a marker of caspase-dependent apoptosis) to examine what type of apoptosis occurred in S13 and L24 cells. S13 and L24 cells treated with 3 μm CdCl2 for 4 h were subjected to immunocytochemistry using these antibodies. AIF is normally localized to mitochondria but it translocates to nucleus under conditions that induce apoptotic cell death (Krantic et al. 2007). When S13 and L24 cells were stained with the AIF antibody, many of the cells showed small dot-like signals, suggesting mitochondrial localization of AIF. But small portion of cells showed AIF signals accumulated in the nucleus (Fig. 3a), suggesting an activated state of caspase-independent apoptosis pathway. However, there seemed to be no difference in the number of S13 and L24 cells that showed nuclear accumulation of AIF (Fig. 3d, left). In contrast, immunocytochemistry using an anti-caspase 3 antibody, which recognized cleaved caspase 3 specifically, showed a marked difference between S13 and L24 cells (Fig. 3b,c). The number of L24 cells stained by anti-cleaved caspase 3 antibody was four times higher than that of S13 (Fig. 3d, right). This result would suggest that the caspase-dependent apoptosis pathway was more activated in CdCl2-treated L24 cells, although several other apoptosis markers may also need to be tested in the future to conclude this.

Figure 3.

 Enhanced caspase-dependent apoptosis pathway by cadmium treatment and its suppression by heat pretreatment in L24 cells. Confocal images of S13 and L24 cells subjected to immunocytochemistry using anti-AIF (a) or anti-cleaved caspase 3 antibody (b) after they were treated with CdCl2 (3 μm, 4 h). (c) Cell populations undergoing or destined to caspase-dependent apoptosis as shown by immunocytochemistry using the antibody to cleaved caspase 3. S13 and L24 cells were treated with CdCl2 (3 μm, 4 h, Cd), heat treated (42 °C, 2 h, H) and CdCl2 treated (3 μm, 4 h) after heat pretreatment (42 °C, 2 h, H+Cd). (d) Quantification of cells positive for nuclear translocation of AIF (left) and cleaved caspase 3 (right). For the statistical analysis, Tukey–Kramer test was used (n = 3, *P < 0.05). Scale bars, 10 μm in (a) and (b): 100 μm in (c).

Subcellular localiztion of EGFP-Cav2.1CT

As already mentioned, carboxyl terminal region of Cav2.1 channel is suggested to be cleaved and translocate to the nucleus (Kordasiewicz et al. 2006). However, the exact pattern of the localization of the Cav2.1CT is not known. To examine the subcellular or subnuclear localization of EGFP-Cav2.1CTs and their possible difference in the localization patterns in relation to the polyQ length, we observed EGFP signals in each cell line. In all of the cell lines, EGFP fluorescence was detected in both cytoplasmic and nuclear regions, with the EGFP signal much stronger in the nucleus than in cytoplasmic region (Fig. 4a). This is in contrast with the situation of HEK-EGFP cells, where EGFP is present almost uniformly in both nuclear and cytoplasmic regions. Furthermore, EGFP signals in the nucleus of C13, S13 and L24 were observed as dot-like structures and the size and the number of the dots within a nucleus were different among these cell lines: dots in C13 cells in the nuclear matrix were bigger than those in S13 and L24, and the number of dots was smaller in C13 than in S13 and L24. The size and the number of the dot-like structures in S13 and L24 were almost the same (Fig. 4a).

Figure 4.

 Nuclear localization of EGFP-Cav2.1CT and aggregate formation of EGFP-Cav2.1CT. (a) Confocal images of cells immunostained with anti-histone H3 (H3) antibody to show that EGFP fluorescence in C13, S13 and L24 cells was localized almost to the nuclear region. Note the different pattern of EGFP signals in HEK-EGFP cells. (b) Immunostaining using anti-PML antibody of S13 and L24 cells under normal conditions (Ctrl), in the presence of 3 μm CdCl2 for 24 h (Cd), after heat treatment (42 °C, 2 h, H), and in the presence of CdCl2 (3 μm, 24 h) after heat treatment (42 °C, 2 h, H+Cd). Note that after cadmium treatment the dots of EGFP signals in the nuclear region were bigger, compared with controls, in cells where the PML-NBs were disrupted (arrows). However, heat treatment before cadmium exposure prevented EGFP-Cav2.1 aggregation compared with cadmium treatment alone (note the smaller EGFP-Cav2.1 aggregations in cells shown by arrows in the H+Cd panel). (c) Percentages of cells without visible PML-NBs in S13 and L24 cells after cadmium treatment, heat treatment or heat treatment before cadmium exposure (left panel). Comparison of sizes of the dots of EGFP-Cav2.1 in the nuclear region of cells positive (+) or negative (−) for the PML immunofluorescence under various conditions (right panel). Under normal conditions, there are no cells negative for the PML immunofluorescence. (d) Percentages of cells with large aggregates in S13 and L24 cells after cadmium treatment or heat treatment before cadmium exposure (left panel). Comparison of the number of dots of EGFP-Cav2.1CT in the nuclear region of cells negative (−) for the PML immunofluorescence under various conditions (right panel). For the statistical analysis, Tukey–Kramer test was used (n = 4, *P < 0.05). Scale bars, 10 μm.

The dot-like structures of the Cav2.1CTs reminded us of the promyelocytic leukemia (PML) protein and PML-nuclear bodies (PML-NBs) (Bernardi & Pandolfi 2007). PML-NBs are thought to be worth examining, because polyQ-containing proteins which cause SCAs, such as ataxin-1 and ataxin-7, are known to colocalize with PML-NBs (Dovey et al. 2004; Janer et al. 2006). When the S13 and L24 cells under normal culture conditions were immunostained with an anti-PML antibody, PML protein was detected as speckles with variable sizes in the nucleus (Fig. 4b). As shown in Fig. 4b, many of the speckles of EGFP-Cav2.1CT in both cell lines did not colocalize with PML. Intriguingly, however, the EGFP-Cav2.1CT speckles of larger sizes colocalized with PML-NBs of larger sizes (not shown).

CdCl2 treatment disrupts PML-NBs and enhances aggregation of Cav2.1CT

Recent studies suggest that treatment of cells with Cd2+ leads to disruption of PML-NBs in the nucleus and affects aggregation formation of some polyQ-containing proteins (Nefkens et al. 2003; Janer et al. 2006). We then performed immunocytochemistry using an anti-PML antibody. The S13 and L24 cells were treated with 3 μm CdCl2 for 4 or 24 h and then they were immunostained with an anti-PML antibody. Four hours after the CdCl2 treatment, S13 cells showed almost no changes in the PML staining and EGFP-Cav2.1CT. However, in small number of L24 cells, PML-NBs were already disrupted and in those cells EGFP-Cav2.1CT aggregations were larger than normal control. After 24 h, this tendency was more obvious. Twenty-four hours after the CdCl2 treatment, PML bodies were disrupted and became invisible in some of the cells. The number of cells where PML-NBs were invisible in L24 was three times larger than that in S13 (Fig. 4c, left). In this kind of cells (both S13 and L24), EGFP-Cav2.1CT proteins tended to form larger aggregates (Fig. 4c right). In S13, EGFP-Cav2.1CT protein aggregations became 3.7 times larger by 3 μm CdCl2 treatment compared with the control. In L24 as well, EGFP-Cav2.1CT protein aggregations became four times larger. The number of cells where large aggregates were formed in L24 cells was three times larger than S13 cells (Fig. 4d, left). Furthermore, the number of aggregates formed in L24 cells was four times larger than that in S13 cells (Fig. 4d, right). However, the cadmium-treated cells that were positive for the PML immunofluorescence in S13 and L24 cells, the size of the ‘dot’ did not become larger compared with the control (Fig. 4b,c).

cDNA microarray analysis and knock-down of the differentially expressed genes

As mentioned, Cav2.1CT is suggested to be involved in nuclear functions including control of gene expression (Kordasiewicz & Gomez 2007). It is possible that the Cav2.1CT that bound metal ion alters its functions. Furthermore, it will be important to clarify the alteration in general gene expression profiles during the cadmium treatment to gain insights into the mechanism underlying the enhanced cell death in L24. We searched for the genes expressed differentially in S13 and L24 cells in the presence of CdCl2. First, S13 and L24 cells were treated with 3 μm CdCl2 for 4 h and then the RNA samples were prepared together with the controls. The 4-h duration for the treatment was selected, because results of the preliminary experiments suggested that cells were still alive at approximately 4 h and began to die afterward. Then the RNA samples were subjected to the cDNA microarray analysis. Four sets of microarray analyses were conducted: (i) L24 with vs without CdCl2 treatment, (ii) S13 with vs without CdCl2 treatment, (iii) L24 vs S13 in the presence of CdCl2, (iv) L24 vs S13 under control conditions. We searched for the genes, whose expression was almost the same in S13 and L24 cells under normal conditions and showed different patterns in both cells when treated with CdCl2. We selected genes that displayed more than twofold change in the comparison set nos 1–3 mentioned above and less than twofold change in the comparison set no. 4. Table 1 shows the list of the selected 21 genes. These are categorized into several groups. Group 1 includes the genes, whose expression is increased in S13 moderately and in L24 greatly by CdCl2 treatment (EGR1, EGR2, EGR3, EGR4, TNFRSF11B, HPX2). EGR1, EGR3 and EGR4 were selected, even though they showed more than twofold changes in the comparison set no. 4. In group 2, contrary to the group 1, gene expression is enhanced in S13 strongly and in L24 cells moderately (HSPA1A, FOSL1, KCNK3). Group 3 includes the genes whose expression is increased by CdCl2 treatment in S13 but not in L24 cells (SLC30A2, ORM1, LAMA2, GLP2R, AGMAT, NCF2, IL11). Group 4 includes the others not categorized into the above groups (REG3G, MED18, GABARAPL1, GINS4). FOSL2 was selected as a kind of control, because expression of FOSL2 was increased by cadmium treatment almost the same way in both S13 and L24 cells.

Table 1.   cDNA microarray analysis of S13 and L24 cells treated with CdCl2
GroupGene symbolLog10 ratioL24 CL24 CdLog10 ratioS13 CS13 CdLog10 ratioS13 CdL24 CdLog10 ratioS13 CL24C
  1. C, control; Cd, treated with 3 μm CdCl2 for 4 h.

1EGR10.682498023 9000.69219010 7000.41810 10026 3000.44823206520
EGR20.3161152370.19267.21050.4251203200.22471.6120
EGR30.64166.72920.60333.11330.4671283760.4543188.4
EGR40.79229.51830.9875.0949.30.65549.52240.361125.2
TNFRSF11B0.49531.599.40.283057.20.29865.51300.15428.340.4
HPX-20.39518.345.50.23617.530.20.31828.358.70.03817.519.1
2HSPA1A0.824246016 4000.968269025 000−0.41231 60012 200−0.02824202260
FOSL10.46660.71770.64698.5436−0.318478230−0.09377.862.8
KCNK30.3075.3710.90.6836.9533.5−0.53244.513.1−0.1496.764.79
3SLC30A20.21532.9541.0181.7836−0.856848118−0.18565.742.9
ORM10.0343503780.303215433−0.6136631610.022253266
LAMA20.14756.779.50.3964.4158−0.30114572.7−0.04165.559.6
GLP2R−0.0067775.90.37477.3183−0.34418985.5−0.07298.983.9
AGMAT0.18322.434.20.35926.159.6−0.57168.818.50.09719.824.7
NCF20.1665.918.650.8796.650−0.76650.98.72−0.1367.035.14
IL110.141211670.54116402−0.497460146−0.0811293.5
4REG3G0.795.7435.405.445.20.9025.4843.805.085
MED18−0.57619150.8−0.071210178−0.74918633.1−0.034227209
GABARAPL10.3859314200.4295591500−0.4981870594−0.109641499
GINS4−0.416856328−0.401911362−0.428778291−0.051994885
FOSL20.63423310000.6532291030−0.041127011500.012200206

We then examined whether these selected genes were causally related to the increased cell death in L24 cells. To facilitate the screening step, we first performed functional screening, RNAi experiments in this case, and then performed quantitative RT-PCR only on the candidate genes identified from the screening to confirm the expression pattern obtained from the microarray analyses.

Three sets of siRNAs were designed for each gene, and S13 and L24 cells were transfected with these siRNAs (Table S1 in Supporting Information). Three days after transfection, the cells were replated and then treated with CdCl2 for another 24 h and the cell viability was determined. We suggested that higher expression level in L24 may lead to the lower viability and that therefore the knock down of this type of gene may result in higher viability. On the contrary, we suggested that the higher expression level in S13 exerts more protective effects compared with L24. Hence, we assume that knock down of this kind of genes lead to lower viability. According to the results of initial preliminary experiments, ten of the 21 genes showed expected results. We therefore analyzed the ten genes further. The summary of the results of the RNAi experiments on these ten genes is shown in Fig. 5. Treatment of S13 and L24 cells with siRNA for TNFRSF11B and REG3G resulted in higher viability of both cells after cadmium treatment. In these cases, significantly higher viability of S13 cells, compared with L24 cells, was still observed. Therefore, knock down of these genes lead to protective effects against cadmium toxicity but these genes can not account for the different vulnerability of S13 and L24 cells. Knock down of EGR1 and EGR2 yielded similar results. In these cases, viability of S13 cells decreased and that of L24 increased compared with the respective controls. These results are difficult to explain, because microarray analyses suggested that expression of these genes increased in both S13 and L24 cells in response to CdCl2, although the increment of expression in L24 was greater. Knock down of the other genes yielded not significant difference in the viability of S13 and L24 cells, suggesting possible involvement of these genes in the mechanism that generates the different vulnerability of S13 and L24 cells. However, the results of the knock down of EGR3, LAMA2 and FOSL2 genes are difficult to interpret, based on the respective expression pattern deduced from the results of the microarray analyses. Furthermore, although knock down of HPX-2 and EGR4 resulted in not significantly different viabilities of S13 and L24 cells, the viability of S13 still tended to be higher than that of L24. The most clear-cut results were obtained for HSPA1A gene. siRNA for HSPA1A greatly reduced the viability of S13 cells after treatment with CdCl2, whereas the viability of L24 cells remains almost unchanged, with no significant difference compared with S13. Therefore, HSPA1A can well account for the vulnerability of L24 cells. Moreover, as HSPs are thought to be related to some polyQ diseases (Sakahira et al. 2002; Tagawa et al. 2007), we further analyzed the HSPA1A.

Figure 5.

 Effects of siRNA for differentially expressed genes in S13 and L24 cells on viability of S13 and L24 cells treated with CdCl2. S13 and L24 cells were transfected with siRNA for the genes selected by microarray analyses. Three days after transfection, cells were replated. One day after that they were treated with 3 μm CdCl2 for another 24 h. At the end of the cadmium treatment, cell viabilities were determined by trypan blue exclusion assay (black columns). Negative control siRNA was used as control (white columns). Tukey–Kramer test was used for statistical analysis (n = 3, *P < 0.05, **P < 0.01).

Protein expression of heat shock transcription factor 1 (HSF1) is down-regulated in L24 cells

To explore the mechanism underlying the lower expression level of HSPA1A gene in CdCl2-treated L24 cells, we quantified the expression level of the gene encoding HSF1, which controls the expression of heat shock protein genes including HSPA1A (Akerfelt et al. 2007). We also quantified the HSPA1A gene expression in S13 and L24 cells to confirm the results of microarray analyses.

DNA binding ability of HSF1, correlated with oligomerization, has been reported to reach a peak between 60 and 120 min after the onset of heat treatment (42 °C) in HeLa cells (Sarge et al. 1993). We therefore treated S13 and L24 cells at 42 °C for 2 h, to examine whether the HSPA1A expression is also lower in L24 cells, when they were exposed to other type of stress. We also treated the cells with 3 or 7 μm CdCl2 for 4 h. Then the RNA samples from the cells treated as above were subjected to quantitative RT-PCR. Basal expression level of HSPA1A gene was very low in both S13 and L24 cells with no significant difference. CdCl2 treatment induced HSPA1A gene expression dose-dependently in both S13 and L24 cells. However, the expression level was significantly lower in L24 compared with the S13 (Fig. 6a). Heat shock at 42 °C enhances greatly the expression of this gene, but the expression level was still significantly lower in L24 cells (Fig. 6a). Therefore, consistent with the microarray analysis, similar low level expression of HSPA1A under normal culture conditions in both S13 and L24 cells and significantly lower enhancement of HSPA1A expression in L24 cells after CdCl2 treatment were confirmed.

Figure 6.

 Quantitative RT-PCR for HSPA1A and HSF1. S13 and L24 cells were treated with 3 μm CdCl2 for 4 h (Cd3), or 7 μm CdCl2 for 4 h (Cd7), or exposed to high temperature (42 °C) for 2 h (H). C, control. Quantity of HSPA1A mRNA (a) or HSF1 mRNA (b) determined by quantitative RT-PCR. All the data are normalized by quantity of GAPDH mRNA. Tukey–Kramer test was used for statistical analysis (n = 3, *P < 0.05, **P < 0.01, ***P < 0.001).

However, expression level of HSF1 gene was not significantly altered, irrespective of the stressor stimuli in both of the cell lines. However, expression level tended to be lower in L24 cells than in S13 cells. Especially, when they were treated with 3 μm CdCl2, the difference reached a significant level (Fig. 6b).

We then examined the protein expression level of HSF1. Under normal conditions, majority of the HSF1 existed as higher mobility form, suggestive of lower phosphorylation state. When the cells were heat shocked at 42 °C for 2 h, majority of the HSF1 was hyperphosphorylated and the mobility in the SDS-PAGE gel was retarded. CdCl2 treatment yielded the moderate effects: HSF1 seemed to be partially phosphorylated (Fig. 7a). In S13 cells, the band intensity of the HSF1 from each lane was almost constant, whereas in L24 cells, samples other than the heat-treated one showed weaker band intensities compared with those of S13 cells and the intensity was increased after heat-treatment to the level comparable with that of S13 cells (Fig. 7a,b). It is noteworthy that the small expansion of polyQ so greatly affected the HSF1 protein expression level, whereas leaving the mRNA level similar.

Figure 7.

 Quantification of HSF1 protein by western blot analysis. (a) Abbreviations are the same as those in Fig. 6. Proteins (30 μg/lane) were resolved and blotted onto PVDF membrane. The blot was first probed with anti-HSF1 and then reprobed with anti-GAPDH for normalization. (b) Band intensities of HSF1 normalized by those of GAPDH in HEK-EGFP, S13 and L24. For statistical analysis, Tukey–Kramer test was used (n = 3, *P < 0.05, **P < 0.01, ***P < 0.001).

Expression level of HSF1 protein is critical for the cadmium-induced cell death

In L24 cells, the HSF1 protein expression level induced by the heat treatment was much higher, compared with the cadmium treatment. If the HSF1 protein expression level determines the vulnerability of L24 cells after cadmium exposure, we can expect that more L24 cells can survive with similar survival rates between L24 and S13 cells after cadmium exposure if they are heat shocked before the cadmium exposure. We first confirmed that the heat shock treatment (42 °C for 2 h) did not significantly alter the expression levels of EGFP-Cav2.1CT by Western blot analyses (Fig. S3 in Supporting Information). Then we measured the survival rate of these heat-treated S13 and L24 cells after exposure to 1-7 μm of CdCl2. As expected, L24 cells showed higher viability at all the concentrations of CdCl2 tested compared with the cadmium treatment alone and also showed almost the same viability as that of S13 cells at all the concentrations (Fig. 8).

Figure 8.

 Heat shock before cadmium treatment can improve the viability of L24 cells. Viability of S13 and L24 cells which were heat-treated at 42 °C for 2 h before treatment with 1–7 μm of CdCl2 for 24 h. At the end of the cadmium treatment, cell viabilities were determined by trypan blue exclusion assay. Tukey–Kramer test was used for statistical analysis (n = 3, *P < 0.05, **P < 0.01, ***P < 0.001).

Effect of heat shock on caspase-dependent apoptosis and aggregate formation

To investigate whether caspase-dependent apoptosis was suppressed by the heat pretreatment, we performed immunocytochemical analyses of CdCl2-treated cells that had been heat shocked, using the antibody that recognizes activated caspase 3. As shown in Fig. 3c,d, marked decrease in the number of L24 cells positive for caspase 3 was observed.

Next to test the effect of heat shock on aggregate formation, we first performed immunocytochemical analyses of S13 and L24 cells using the antibody against PML. Heat treatment (42 °C, 2 h) was found to dramatically increase the cell population lacking PML-NBs in both S13 and L24 cells (Fig. 4c, left). However, heat treatment by itself did not affect aggregate formation (Fig. 4b,c, right).

Next, we have tested the effect of heat shock on CdCl2-induced aggregate formation and found that the number of cells with large aggregates, and the dot size and the number of EGFP-Cav2.1CT aggregates in L24 cells lacking PML-NBs were all significantly decreased by heat shock pretreatment (Fig. 4b–d).

Discussion

Nuclear localization of Cav2.1CT

A recent study has reported that Cav2.1CT is cleaved from the full-length protein and translocated to the nucleus (Kordasiewicz et al. 2006). To confirm this and to examine the subnuclear localization of Cav2.1CT, we observed EGFP-Cav2.1CT fusion proteins in the HEK293-derived cell lines that express the fusion proteins carrying the carboxyl terminal sequence variations (Ishikawa et al. 1999). Consistent with the previous report, EGFP-Cav2.1CTs were localized almost exclusively to the nucleus, irrespective of whether they are short form or long form with respect to the carboxyl terminal tail, and of the polyQ tract length. However, the fusion proteins were localized not uniformly in the nucleus but as punctate structures. The size and the number of the ‘puncta’ in one nucleus were different among the cell lines. So far the biological significance of this difference is not clear, but it is intriguing that the size and the number of the puncta became larger in the cells which showed no visible PML protein expression as assessed by immunofluorescence. Similar enlargement of aggregation of polyQ-expanded protein was reported for ataxin-7, the causative protein for SCA7, when COS7 cells expressing this protein were treated with CdCl2 (Janer et al. 2006). In this case, polyQ-expanded ataxin-7 is normally degraded by clastosomes assembled by PML-NBs and the disruption of PML-NBs by cadmium treatment led to inhibition of the degradation by the clastosomes. Similar degradation mechanism may be present in HEK293 cells and Cav2.1CT is likely to be degraded by a PML containing machinery, although the polyQ tract length in Cav2.1 is much shorter than that of ataxin-7.

The number of cells where PML-NBs were disrupted by the cadmium treatment was larger in L24 than in S13 cells. It is suggested that Cd2+-dependent disruption of PML-NBs requires p38 MAPK or ERK1/2 signaling (Nefkens et al. 2003). Therefore, it is possible that Cav2.1CT(Q24) may affect the p38 MAPK or ERK1/2 signaling so that the disruption of PML-NBs is enhanced. It would also be interesting to know whether the polyQ-expanded ataxin-7 also accelerates the cadmium-induced disruption of PML-NBs as we have found in the present study. PML-NBs were suggested to work as sensors of foreign or misfolded proteins in the nucleus (Muratani et al. 2002). Therefore, the fact that Cav2.1CT aggregates become larger when the PML-NBs are disrupted might mean that Cav2.1CT(Q24) and Cav2.1CT(Q13) proteins are equally good substrates for PML-containing subnuclear degradation system. However, L24 cells are easier to lose PML-NBs and the number of aggregates is increased in these cells by the toxic stimuli, exposure to cadmium in this case, than S13 cells. This fact may determine the difference of cell death vulnerability between L24 and S13 cells.

However, heat treatment (42 °C, 2 h) promoted disruption of PML-NBs to much higher extent compared with cadmium treatment, consistent with previous reports (see review Eskiw & Bazett-Jones 2002). Furthermore, in contrast to Cd2+-dependent disruption, extent of disappearance of PML-NBs in heat shock-dependent pathway is similar between S13 and L24 cells. Moreover, cadmium treatment after heat shock enhanced further the disruption of PML-NBs. However, heat treatment was found to prevent large aggregate formation in S13 and L24 cells lacking PML-NBs and reduce the number of aggregates in L24 cells lacking PML-NBs. Therefore, it seems that there are two (PML-NBs-dependent and -independent) mechanisms preventing the Cav2.1CT proteins from forming large aggregates. Under normal conditions, Cav2.1CT aggregate formation may be effectively inhibited by the PML-NBs-dependent pathways. PML-NBs-independent pathway harboring HSF1-HSPA1A axis may also be present. However, because all the L24 and S13 cells contain PML-NBs, PML-NBs-independent pathway may not be necessary to contribute to the blockade of large aggregate formation. Oxidative stress like CdCl2 stimuli promoted the degradation of PML-NBs in the cells expressing Cav2.1CT. Cav2.1CT(Q24) seems to be more effective in inducing Cd2+-dependent PML-NBs loss than Cav2.1CT(Q13). Significantly higher percentage of L24 cells contained large aggregates than S13 cells under Cd2+ exposure. Furthermore, a larger number of aggregates were found in L24 cells lacking PML-NBs than S13 cells lacking PML-NBs. Therefore, more enhanced activation of HSF1-HSPA1A axis found in S13 cells than in L24 cells may contribute to the reduction of aggregate formation. Heat shock treatment seems to effectively block PML-NBs-dependent pathway by promoting the degradation and possibly by blocking the production of PML-NBs. However, in contrast to Cd2+-dependent activation, heat shock seems to activate PML-NBs-independent (HSF1-HSPA1A-dependent) pathway very strongly to block aggregate formation. Further rigorous study would be necessary to clarify the relationship between aggregate formation and cell death.

Lower viability of cadmium-treated L24 cells

RNAi experiments to knock down the Cav2.1 CT(Q24) in L24 cells showed that this vulnerability is causally related to the expression of this protein.

Enhanced cadmium toxicity was dependent on polyQ tract length and even the modestly expanded 24Q stretch was clearly toxic to HEK293 cells, because S13 cells expressing almost the same level of Cav2.1CT protein were more resistant to the CdCl2 stimuli. EGFP expression in HEK-EGFP cells was 30 times higher than the EGFP-Cav2.1CT expression in S13 and L24 cells. However, the viability of HEK-EGFP cells was essentially the same as that of parental HEK293 cells, when they were treated with 1–7 μm CdCl2, suggesting no protective effects of EGFP against CdCl2 toxicity. Low viability of cadmium-treated L24 cells is thus not ascribed to lack of the protective effects of EGFP.

It is suggested that cadmium treatment activates two different cell death pathways in HEK293 cells (Mao et al. 2007). One is caspase-independent cell death: cadmium induces AIF release from mitochondria and its translocation to cytoplasm and nuclei. The other is caspase-dependent cell death: cadmium induces mitochondrial cytochrome c release dramatically, which ultimately cleaves pro-caspase 3 to form active caspase 3. In the present study, we found that the number of cells stained by anti-cleaved caspase 3 antibody in L24 cells was significantly larger than in S13. In contrast, the number of cells, which showed nuclear AIF localization, was almost the same in L24 and S13. These results suggest that caspase-dependent cell death pathway was more preferentially activated in L24 cells compared with S13 cells. Then, what makes this difference? One plausible explanation is that HSP functions are attenuated in L24 cells. HSPs are known to inhibit apoptosis (Lanneau et al. 2008). Especially, HSP70 is known to be able to inhibit almost all the cell death pathways. For example, HSP70 inhibits kinases which induce apoptosis such as JNK and Ask1, and also inhibits transcription factors such as p53, whose expression leads to apoptosis in response to DNA damage, by masking the nuclear localization signal to prevent p53 from being imported to the nucleus (Zylicz et al. 2001; Park et al. 2002; Lee et al. 2005). In the present study, expression of HSPA1A gene encoding HSP70 was found to be greatly attenuated in L24 cells. Therefore, the attenuation of HSPA1A expression in L24 cells is likely to contribute to the vulnerability in these cells. Results of RNAi experiments for this gene support this notion.

HSF1 is a critical transcription factor for the expression of HSP genes and thus plays a pivotal role in stress responses (Akerfelt et al. 2007). In the present study, mRNA expression of HSF1 tended to be lower in L24 cells compared with S13 cells and the protein expression was greatly decreased in L24 cells under normal conditions and in the presence of Cd2+. However, when L24 cells were heat shocked at 42 °C for 2 h, the protein expression level was highly increased to the level comparable with that of S13 cells. We then tested whether more L24 cells become able to survive the cadmium insult after heat treatment. As expected, the viability of L24 cells was increased to the level comparable with that of S13 cells. Furthermore, heat shock treatment reduced the number of activated caspase 3-positive cells. These results may suggest that this increase in the viability was caused by suppression of caspase-dependent apoptosis. In other words, the low viability of L24 cells when treated with CdCl2 was because of the low level of HSF1 protein, which eventually led to enhanced caspase-dependent apoptosis. To go furthermore, it is tempting to speculate that reduced expression of HSF1 protein induced by the polyQ expansion of Cav2.1CT may be the basis of cell death vulnerability of SCA6. Further rigorous studies including the use of SCA6 knockin mouse Purkinje cells are necessary to conclude this issue.

Why HSF1 protein levels are greatly different, although the difference in the mRNA expression levels between S13 and L24 cells is small, except when they were exposed to high temperature? It may be possible that protein degradation system which is heat-sensitive is activated more highly in L24 cells. In some models of Huntington’s disease, one of the most studied polyQ diseases, activation of proteasome pathway has been reported (Diaz-Hernandez et al. 2003; Bett et al. 2006). Otherwise, we can also speculate that posttranscriptional control of HSF1, such as miRNA, may be up-regulated in L24 cells. However, further rigorous studies will be necessary to clarify this issue.

Recently, drugs that activate HSF1 are reported to be efficient in ameliorating abnormal phenotypes in some of the polyQ disease models, even though the changes in HSF1 expression is not causally related (Fujikake et al. 2008). Therefore, it seems well-grounded to speculate that HSF1-activating drugs may be useful for SCA6 as well.

In conclusion, the present study shows that L24 cells expressing Cav2.1CT carrying an SCA6 mutation are more sensitive to toxic effect of cadmium than control S13 cells. Caspase-dependent apoptotic cell death pathway is more highly activated in L24 cells. These are thought to be due to the lower expression of HSF1 protein, which leads to attenuation of HSPA1A gene expression. Therefore, the mechanism underlying the lower expression of HSF1 protein in L24 cells will be of importance to gain further insight into the pathogenesis of SCA6. Furthermore, it would be essential to examine whether HSF1-HSPA1A plays a critical role in SCA6 pathogenesis using cerebellar Purkinje cells from SCA6 knockin mice.

Experimental procedures

Cell culture

Cells were cultured in Dulbecco’s Modified Eagle Medium (Invitrogen) supplemented with 10% fetal bovine serum (Biowest) and G418 (500 μg/mL, Sigma) at 37 °C and 5% CO2. Cells were passaged every 3–4 days, when they became near confluent. For CdCl2 toxicity experiments, cells were trypsinized and plated at a density of 2 × 105 cells/well in 12-well plates. After 24 h, medium containing CdCl2 was added to make the final concentration 1–7 μm. After another 24 h, cells were trypsinized and the number of living cells, assessed by trypan blue exclusion, was determined.

We found that the expression of the EGFP-Cav2.1CT fusion proteins became lower, as the cells were passaged further in the absence of G418. However, addition of G418 prevented the loss of expression of the fusion proteins. We therefore added G418 to the culture medium and used cells with early passage numbers throughout the present study.

Antibodies

Antibodies used in the present study and their dilutions are as follows: rabbit anti-GFP antibody (Medical & Biological Laboratories (MBL), used at 1 : 1000 dilution); rabbit anti-PML antibody (MBL, 1 : 500); rabbit anti-histone H3 antibody (Cell signaling, 1 : 50); mouse anti-glyceraldehyde-3-phophate dehydrogenase (GAPDH) monoclonal antibody (Chemicon, 1 : 1000); rabbit anti-AIF (CT) (ANASPEC, 1 : 100); rabbit anti-cleaved caspase 3(Asp175) antibody (Cell Signaling, 1 : 1000); rabbit anti-HSF1 antibody (Cell Signaling, 1 : 1000); Cy3-conjugated affiniPure goat anti-mouse IgG(H+L) (Jackson immunoresearch, 1 : 1000); Cy3-conjugated affiniPure goat anti-rabbit IgG(H+L) (Jackson immunoresearch, 1 : 1000); anti-rabbit Ig, horseradish peroxidase linked whole antibody from donkey (GE Healthcare, 1 : 10 000); anti-mouse Ig, horseradish peroxidase linked whole antibody from donkey (GE Healthcare, 1 : 1000); biotin-conjugated donkey anti-rabbit IgG (Chemicon, 1 : 100 000).

Immunocytochemsitry

Immunocytochemistry was performed essentially the same way as previously described (Hu et al. 2005). Images were taken using the LSM5 Pascal confocal imaging system (Zeiss).

Quantification of the cells positive for cleaved caspase 3 and nuclear translocation of AIF

First, cells were double labeled with Hoechst 33258 and either anti-caspase 3 or anti-AIF antibody. The number of cleaved caspase 3-positive cells and the total number of nuclei in four independent randomly chosen areas with a fixed size were determined by using the ImageJ software (http://rsb.info.nih.gov/ij/index.html). The percentage of the caspase 3 positive cells was calculated and averaged. This was performed for 3 independent coverslips.

Cells with nuclear translocation of AIF were counted by observing an 18 × 18 mm coverslip by an experimenter. The total number of the cells in the coverslip was estimated by counting the number of total nuclei in five independent randomly chosen areas with a fixed size (usually containing 500–800 cells) using the ImageJ software. The permillage of cells with nuclear translocation of AIF was calculated and this was performed for three independent coverslips.

Immunoblotting

Cells were harvested by treating them with 1mM EDTA/HEPES-buffered saline for 20 min at room temperature, and they were thoroughly washed by cold phosphate-buffered saline. Cell pellets were briefly sonicated in RIPA buffer containing 20 mm Tris, 150 mm NaCl, 1mm Na2EDTA and 1 mm EGTA, 1% NP-40, 1% sodium deoxycholate and protease inhibitors (Sigma). Total extracts were centrifuged at 16 000× g for 5 min at 4 °C. Supernatants (40 μg protein) were subjected to SDS-PAGE and resolved proteins were transferred to polyvinylidene difluoride (PVDF) membrane (Immobilon P, Millipore). The blot was probed with an anti-GFP antibody and reprobed with anti-GAPDH antibody for normalization. The bound antibody was detected with biotinylated secondary antibody, alkaline-phosphatase-labeled streptavidin (Vector, used at 1 : 10 000 dilution), and ECF (GE Healthcare) as a substrate. FluorImager 595 (GE Healthcare) was used to quantify the band intensities.

For the detection of HSF1 protein, ECL kit (GE Healthcare) was used. In this case, we used the ImageJ software to quantify the band intensities.

Microarray analysis

Cells were treated with CdCl2 for 4 h and total RNA was prepared from cells by using RNeasy mini kit (Qiagen). cDNA microarray analysis was performed using Whole Human Genome OligoDNA Microarray (Agilent) following the manufacturer’s instructions.

Quantitative real-time PCR

Reverse transcription was performed using SuperScript III (Invitrogen). After treatment with RNaseH, the resultant cDNA was subjected to the quantitative PCR on a 7500 real-time PCR system (Applied Biosystems). Primer and Taqman probe sets were from Applied Biosystems (Hs00232134_A1 for HSF1, Hs00339163_S1 for HSPA1A and part No. 4352934E for GAPDH).

RNAi

Three sets of siRNAs designed for each gene (Stealth™ RNAi; Invitrogen) or the equal amount of negative control siRNA (Stealth™ RNAi Negative Control Medium GC Duplex #2, Invitrogen) were transfected into S13 or L24 cells by reverse transfection method using Lipofectamine RNAiMAX (Invitrogen) according to the manufacturer’s instruction. Three days after transfection, the cells were replated at the density of 2 × 105 cells/well in 12-well plates. The next day, cells were treated with 3 μm CdCl2 for 24 h. At the end of the cadmium treatment, cell viability was assessed by trypan blue exclusion. Detailed information on the siRNAs used in this study is supplied (Table S1 in Supporting Information).

Acknowledgements

This study was supported by research grants from the Ministry of Education, Culture, Sports, Science and Technology of Japan, by the Grant-in-Aid for Scientific Research of the Japan Society for the Promotion of Science, and by the Naito Foundation to T. Tanabe. L. Li was supported by a grant from the 21st Century COE Program on Brain Integration and its Disorders to Tokyo Medical and Dental University.

Ancillary