Identification and characterization of GRIM-1, a cell-death-associated gene product

Cytokines belonging to the interferon (IFN) group potently suppress cell growth and promote apoptosis (Kimchi, 1992). They exert antiviral, immuno-regulatory and anti-proliferative effects employing the JAK (Janus tyrosine kinase)-STAT (signal transducer and activator of transcription) pathways (Schindler et al., 2007). Certain IFN-insensitive tumors become sensitive to IFNs when in the presence of all-trans retinoic acid (RA) (Moore et al., 1994). RA is a major physiological retinoid that binds to specific nuclear receptors (Chambon, 1996). The RA receptors act as transcription factors to drive expression of genes involved in cell differentiation and growth control. Retinoids inhibit growth of certain leukemias, skin dysplasias in vivo and tumor cell lines in vitro (Altucci and Gronemeyer, 2001). We have demonstrated earlier that the combination of both IFNβ and RA (IFN/RA) is a highly effective inhibitor of tumor growth in vivo and in vitro (Kalvakolanu, 2004; Lindner et al., 1997). Although IFN/RA combination induces apoptosis, the exact molecular mechanisms involved are still unclear.

Apoptosis eliminates superfluous and/or potentially dangerous cells in mammals. It is regulated by cytokines, survival factors, cell-cell and/or cell-ECM interactions, oncogenes, DNA damage and viral proteins (Ashkenazi and Dixit, 1998; Green and Reed, 1998; Stennicke and Salvesen, 2000; Youle and Strasser, 2008). Loss of apoptotic response seems to cause drug resistance in tumor cells (Ashkenazi and Dixit, 1998; Green and Reed, 1998; Logue and Martin, 2008; Lowe et al., 1994; Stennicke and Salvesen, 2000; Youle and Strasser, 2008). Although the roles of central players – such as caspases, Bcl2-like proteins and death receptors – in apoptotic responses have been well defined over the last decade, it is unclear how these proteins control cell death in a signal-specific and cell-type-specific manner. Via a positive-growth selection in the presence of cytotoxic agents, genome-wide expression-knockdown strategies permit the identification of gene products that are indispensable for cell-growth suppression or cell death. Such strategies do not require a prior knowledge about the gene(s) or their product(s) and allow an unbiased identification of activators of apoptosis (Deiss et al., 1995; Hofmann et al., 1998). We have employed one such approach, the antisense technical knockout (TKO), to isolate the GRIM (genes associated with retinoid-IFN-induced mortality) genes. In this approach, endogenous gene expression is knocked down by a library of antisense complementary DNAs (cDNAs) expressing from an episome. In this study, we have characterized a newly identified gene, GRIM-1, whose antisense expression protected cells from IFN/RA-induced cell death. The GRIM-1 mRNA produces three protein isoforms, designated α, β and γ, from a single open reading frame (ORF); these isoforms induce cell death. Whereas GRIM-1γ can readily activate cell death, GRIM-1α and GRIM-1β required a proteolytic activation by caspase-9 to induce cell death. These isoforms suppress ribosomal RNA (rRNA) maturation to ablate cell growth. Together, these results identify a novel mediator and a novel cell-death-regulatory pathway.

We isolated GRIM genes using the antisense TKO strategy (Angell et al., 2000; Hofmann et al., 1998). Briefly, HeLa cells were transfected with an antisense cDNA library and selected with hygromycin B (100 μg/ml) and a lethal dose of human IFNβ (5000 U/ml) and RA (5 μM) for 4 weeks. The surviving colonies were expanded and the episomal DNA bearing the death-associated gene was isolated. One such episome carried an insert of 1.9 kb and was designated GRIM-1. Stable cell-line pools (ñ150 colonies) were generated after transfecting pTKO1 or the same vector expressing antisense GRIM-1 (pTKO1–GRIM-1) and selecting with hygromycin B. Upon exposure of the individual cells to IFN/RA for 4 weeks, the antisense-GRIM-1-episome-transfected but not the empty-vector-transfected cells formed viable colonies (Fig. 1A). Similar results were obtained with two other cell lines, MCF-7 and T47D (data not shown). In a separate experiment, equal numbers of pTKO1 cells and pTKO1–GRIM-1-expressing cells were seeded into 96-well plates, and were grown in the presence of a sub-lethal but growth-inhibitory concentration of human IFN-β (1000 U/ml) and RA (1 μM). In contrast to the pTKO1 cells, cells harboring pTKO1–GRIM-1 continued to grow in the presence of IFN/RA (Fig. 1B). Thus, antisense GRIM-1 blocked IFN/RA-induced growth suppression and cell death.

To determine whether cell protection was indeed due to the production of antisense GRIM-1 message, total RNA from the cells was subjected to a northern blot analysis with ³²P-labeled GRIM-1 probe. In pTKO1 cells and pTKO1–GRIM-1-expressing cells, two GRIM-1 RNAs, of ~2.7 and 3.0 kb, were detected. The basal expression of these RNAs was induced (~fivefold) upon IFN/RA treatment (Fig. 1C). These two RNAs correspond to endogenous GRIM-1 transcripts. In pTKO1–GRIM-1-transfected cells, a new RNA band of 1.9 kb was present. This seems to be the antisense GRIM-1 RNA, given its absence in the empty-vector-transfected cells. Because the antisense construct in the pTKO1 vector is under the control of an IFN-inducible promoter, antisense GRIM-1 message was observed only in IFN/RA-treated cells. Consistent with these results, a loss of GRIM-1 protein expression occurred (see below). Thus, GRIM-1 seems to be a growth suppressor.

Because antisense GRIM-1 expression conferred resistance to IFN/RA-induced death and two GRIM-1 transcripts were seen in HeLa cells, we next examined the effects of IFN/RA on GRIM-1 gene expression in the MCF-7 cell line. Initially, a time course of IFN/RA response in MCF-7 cells was used to determine a suitable time point for analyzing GRIM-1 expression. Total RNA, from untreated and IFN/RA-treated cells, was extracted and subjected to northern blot analysis with GRIM-1 as probe. Similar to HeLa cells, MCF-7 cells also expressed two GRIM-1 transcripts (sizes ~2.7 and 3.0 kb), the expressions of which were highly induced by IFN/RA (Fig. 1D). However, after 72 hours of exposure to IFN/RA, GRIM-1 RNA levels declined (data not shown). This blot was subsequently hybridized to GAPDH probe (Fig. 1D) to ensure a comparable amount of RNA across lanes. This increase in GRIM-1 transcripts also correlated with the time of significant cell death in these cell lines (data not shown). Consistent with these observations, northern blot analyses revealed two IFN/RA-inducible transcripts in five other cell lines (T47D, BT-20, A375, Colo-320 and ACHN) (Fig. 1E). In some cell lines, a higher steady-state level of GRIM-1 was observed. In all cases, IFN/RA induced the expression of these RNAs, although the magnitude of induction was variable among these cell lines.

Because the antisense GRIM-1 (1.9 kb) did not represent the full-length message(s), using a combination of cDNA-library screening and 5′RACE we cloned a cDNA of 2.7 kb, which corresponded to the smaller of the two RNAs detected in northern blots. Attempts to clone the 3.0-kb transcript were not successful. Sequence analysis of the 2.7-kb clone revealed the presence of a 1.7-kb ORF that encodes a 577-amino-acid polypeptide with a theoretical weight of 64.8 kDa. Database searches revealed that GRIM-1 exhibited homology to numerous human SHQ1 entries (see supplementary material Table S1), although no protein(s) have been characterized to date. Despite its similarity with yeast Shq1p (Table 1), GRIM-1 exhibited a significant sequence divergence at its C-terminus. Thus, GRIM-1 seems to be a distant ortholog of yeast shq1. Depletion of Shq1p in yeast causes severe growth retardation, owing to low ribosomal levels (Yang et al., 2002). A modular representation of GRIM-1 with putative domains (predictions made using Expasy, BLOCKS and MOTIF tools) is shown in Fig. 2A. Salient features are: (1) many leucine/isoleucine repeats that are six to eight amino acids apart are found between residues V358 and A438, indicating a potential leucine zipper-like motif; (2) a serine-rich C-terminus, suggesting potential regulation by serine/threonine kinases; and (3) actin crosslinking-like domain between residues S233 and P306. Other regions of interest are putative caspase-cleavage sites and multiple tyrosine residues that, if phosphorylated, could play an important role in the regulation of GRIM-1 function. A detailed characterization of the functional significance of these domains is underway.

In order to ascertain whether GRIM-1 actually encodes a protein, we performed coupled in vitro transcription and translation (Promega). The RNA from the full-length GRIM-1 clone gave three polypeptides of 85, 75 and 45 kDa (Fig. 2B), much higher than the predicted weights of 65, 55 and 39 kDa, respectively. All three proteins were produced from a single ORF and the sizes of the proteins agreed with translation from three potential start sites located within the ORF. This was confirmed by cloning the three respective ORFs into pGEM-7zf and repeating the in vitro analysis (Fig. 2B). We have termed these three peptides as GRIM-1α, GRIM-1β and GRIM-1γ. Thus, GRIM-1 mRNA can produce three proteins with identical C-termini in mammalian cells.

Because two GRIM-1 transcripts were induced by IFN/RA (Fig. 2), we next determined whether the same was true at the protein level. A rabbit polyclonal IgG, raised against the C-terminal 203 amino acids of GRIM-1, was used in western blot analyses to determine the expression of GRIM-1 proteins in MCF-7 (Fig. 2C) and HeLa (data not shown) cells. In whole-cell extracts from both MCF-7 and HeLa cells, four to six bands ranging from 30 to 85 kDa were detected (Fig. 2C, upper panel), and all these proteins were induced upon IFN/RA treatment by 72 hours and continued to accumulate beyond 72 hours (data not shown). We arbitrarily named these bands α1, α2, β1, β2, γ1 and γ2 on the basis of their molecular sizes. The sizes of these bands were different from the predicted molecular weights of α, β or γ or their corresponding in vitro translated proteins, suggesting potential post-translational modification(s). The blot was stripped and reprobed with actin-specific antibody to ensure comparable protein loading (Fig. 2C).

To confirm the effect of antisense GRIM-1 on these proteins, extracts from control and IFN/RA-treated (72 hours) HeLa cells carrying pTKO1 or pTKO1–GRIM-1 were subject to a western blot analysis. This experiment revealed that expression of the majority of GRIM-1-like polypeptides was knocked down in the presence of antisense GRIM-1 (Fig. 2D, upper panel). This blot was stripped and re-probed with actin-specific antibody to ensure comparable protein loading (Fig. 2D, lower panel). Thus, the antisense-mediated knockdown of GRIM-1 proteins correlated with resistance to IFN/RA-mediated growth inhibition (see Fig. 1A,B).

Because the rabbit antibody produced a lot of background staining, it was not useful for in situ studies. To determine the location of endogenous GRIM-1, we analyzed cells by confocal microscopy using a mouse polyclonal GRIM-1-specific antibody. This antibody recognized the α, β and γ isoforms of GRIM-1, and yielded similar patterns to the rabbit polyclonal antibody in western blots (data not shown). GRIM-1-specific staining was observed in the cytoplasm and nucleus (Fig. 2E). To determine the effect of IFN/RA on GRIM-1, cells were treated with IFN/RA for 24 hours before processing. Upon IFN/RA treatment there was an increase in nuclear and cytoplasmic levels of GRIM-1 (Fig. 2E). In both cases, a focal staining was seen in the nuclei, indicating the association of GRIM-1 with sub-nuclear complexes and/or structures. The isotypic control IgG did not produce any signals in untreated and/or IFN/RA-treated cells (Fig. 2E). Because the antibody recognizes all GRIM-1 isoforms and isoform-specific antibodies are not available at this stage, we employed epitope-tagged GRIM-1 constructs for determining their cellular localization. Cells were transfected with an empty vector (pCXN2) or individual Myc-tagged GRIM-1 isoforms. Cells were permeabilized, fixed and incubated with Myc-tag-specific antibody 24 hours post transfection. They were then incubated with an anti-mouse IgG labeled with Alexa Fluor 488. Like the native antibodies, the anti-Myc-tag antibody detected nuclear and cytoplasmic GRIM-1 staining. Interestingly, the GRIM-1γ-expressing cells had lobing of nuclei, which is an early sign of apoptosis. The empty-vector-transfected cells did not show any signals, indicating the specificity of detection (Fig. 2F). Although GRIM-1 was present in the cytoplasm diffusely, in the nucleus it localized primarily as distinct foci.

As mentioned earlier, antisense-mediated downregulation of GRIM-1 conferred resistance to IFN/RA-induced death (see Figs 1 and 2). Because three GRIM-1 peptides can be produced from a single RNA (see Fig. 2B), we next determined whether there were differences in their death-inductive properties. MCF-7 cells were transfected to express individual GRIM-1 isoforms as Myc-tagged proteins. Significantly fewer stable GRIM-1α (P<0.05), GRIM-1β (P<0.01) and GRIM-1γ (P<0.01) colonies formed, compared with the control-vector transfectant (Fig. 3A). Cells expressing moderate levels of GRIM-1 isoforms grew slower compared with control-vector-transfected cells (Fig. 3B). We next examined the sensitivity of GRIM-1-expressing cells to IFN/RA-induced apoptosis by using TRITC-labeled-annexin-V binding as a tool. Indeed, cells expressing GRIM-1 isoforms were hypersensitive to IFN/RA-induced apoptosis compared with those expressing the control vector alone (Fig. 3C). Importantly, a high baseline apoptosis occurred upon GRIM-1 overexpression compared with vector control. In all three experiments above, the growth-suppressive and/or apoptotic effects of GRIM-1 isoforms were in the order of γ>β>α. The finding that IFN/RA further enhances GRIM-1-dependent death suggested that post-translation modifications and/or activation of other pathways that control GRIM-1 activity might play a role in promoting cell death. We also determined the expression of GRIM-1 isoforms by a western blot analysis of cellular lysates with a Myc-tag-specific antibody (Fig. 3D, upper panel). All three isoforms were expressed in MCF-7 cells. However, GRIM-1γ expressed at the lowest level compared with the others. The blot was stripped and re-probed with actin-specific antibody to ensure comparable protein loading (Fig. 3D, lower panel). We consistently observed a lower level of GRIM-1γ expression in other cell lines compared with GRIM-1α and GRIM-1β (data not shown). Although we have used annexin V as the marker for apoptosis, other characteristics such as caspase activation and nuclear fragmentation were noted in these cells (data not shown). However, MCF-7 cells do not undergo nuclear fragmentation, owing to a lack of caspase-3 gene (Janicke et al., 1998).

Cellular sensitivity to GRIM-1-induced cell death was assessed using another transient assay. For this purpose, we infected HeLa cells with lentiviral particles coding for GFP-tagged GRIM-1 isoforms. Twenty-four hours later, cells were fixed, stained with DAPI and observed, using a fluorescent microscope, for dying cells. In initial studies, we observed that cells expressing GFP– GRIM-1 isoforms rounded up overtime, unlike those expressing GFP alone (supplementary material Fig. S1). These cells also had fragmented and/or condensed nuclei, whereas no such changes were observed in control cells (supplementary material Fig. S2). We also ensured the expression of these constructs by performing a western blot analysis with a GFP-specific antibody (Fig. 3E); this corresponded to a shift towards higher molecular size. We counted green fluorescent cells from multiple independent fields with fragmented and/or condensed chromatin and expressed them as fraction of total GFP-positive cells (Fig. 3F, white bars). The potency of GRIM-1 isoforms to induce cell death by this criterion was in the order of γ>β>α. This experiment was further supported by a FACS analysis of GFP–GRIM-1-expressing cells stained with TRITC-labeled annexin V. GFP and TRITC double-positive cells were calculated from total GFP-positive cells (Fig. 3F, black bars). The data obtained using these two methods were very similar. In summary, all three GRIM-1 isoforms induced apoptosis, with GRIM-1γ being the most potent.

Because GRIM-1α and GRIM-1β were weaker inducers of apoptosis compared with GRIM-1γ, we next checked whether there was an inhibitory region in these two proteins that prevents their spontaneous apoptotic activity. Given that GRIM-1α and GRIM-1β shared similar apoptotic properties and a common N-terminal region that is absent in GRIM-1γ, we generated serial N-terminal deletions to the GRIM-1β ORF (Fig. 4A) and cloned them downstream of the GFP tag in the pEGFP-C2 vector. We first ensured the expression of these deletion mutants. All chimeric proteins expressed with the expected sizes (Fig. 4B). The GRIM-1β deletions NΔ97 and NΔ121 expressed to a lower extent compared with GRIM-1β. Upon expression in MCF-7 cells, NΔ82, NΔ97 and NΔ121 robustly induced apoptosis, compared with the GRIM-1β and NΔ27 constructs (Fig. 4C). Similar results were obtained in HeLa and Cos-7 cells (data not shown). Thus, it seems that the proapoptotic activity of GRIM-1β is restrained by a domain located between amino acids 27 and 82 of its N-terminus. Interestingly, this region harbored potential caspase-cleavage sites. Therefore, in the next set of experiments, we examined the impact of caspase activities in stimulating the death-promoting activity of GRIM-1.

To determine the role of caspases in regulating GRIM-1 cleavage, we expressed GRIM-1α and GRIM-1β as N-terminally FLAG-tagged proteins (Fig. 5A), and then treated cells with IFN/RA. Fig. 5B shows a typical IFN/RA-induced cleavage of GRIM-1α and GRIM-1β proteins. IFN/RA treatment caused a decline of full-length protein level, which was accompanied by an appearance of a short N-terminal fragment that corresponds to the N-terminus. No such cleavage was observed in steady state. Because the N-terminal fragment, which bears the FLAG tag, is cleaved off, the larger processed GRIM-1 product could not be seen in these blots. In the next set of experiments, we investigated whether Z-VAD-fmk, a pan-caspase inhibitor, could inhibit IFN/RA-induced cleavage of GRIM-1α and GRIM-1β. As controls, we used empty vector and FLAG–GRIM-1γ. Because there are no other cleavage products (see Fig. 5B), we showed only the relevant portions of the blots in these experiments. As expected, in all cases no cleavage of GRIM-1α and GRIM-1β occurred in the untreated controls or in the presence of Z-VAD-fmk (Fig. 5C). However, IFN/RA treatment activated cleavage of GRIM-1α and GRIM-1β, which was inhibited in presence of Z-VAD-fmk. As expected, IFN/RA treatment did not induce the cleavage of GRIM-1γ. GRIM-1 cleavage occurred in a delayed manner, indicating the activation of additional processes, such as mitochondrial damage, prior to GRIM-1 activation.

Our previous studies indicated a role for caspases, in particular caspase-9, in the regulation of IFN/RA-induced cell death (Angell et al., 2000). Therefore, we next checked whether knockdown of caspase-9 affected IFN/RA-induced cleavage of GRIM-1α and GRIM-1β. Using specific lentiviral short hairpin RNAs (shRNAs), we knocked down the expression of caspase-9 in HeLa cells and measured its effects on IFN/RA-induced cleavage of GRIM-1α and GRIM-1β. The CASP9-specific shRNA knocked down >85% of endogenous protein, compared with the controls (Fig. 5D, top panel). IFN/RA treatment activated a normal cleavage of GRIM-1α and GRIM-1β in the control cells but it failed to do so in cells expressing CASP9-specific shRNA (Fig. 5D, middle panels). A comparable protein loading was confirmed by probing the blots with an actin-specific antibody (Fig. 5D, bottom panel). The importance of caspase-9 for the cleavage of GRIM-1α and GRIM-1β was further ascertained in a complementary experiment. Casp9^−/− MEFs were transfected with expression vectors coding wild-type caspase-9 or a catalytically inactive mutant along with GRIM-1α or GRIM-1β. First, the expression of caspase-9 was confirmed by a western blot analysis (Fig. 5E, top panel). No IFN/RA-induced cleavage of GRIM-1α and GRIM-1β occurred in cells complemented with empty vector and/or mutant caspase-9 (Fig. 5E, middle panels). However, expression of wild-type caspase-9 complemented this defect and restored IFN/RA-induced cleavage of GRIM-1α and GRIM-1β. A comparable protein loading was ascertained by probing these blots with an actin-specific antibody (Fig. 5E, bottom panel). The above experiments indicated the involvement of mitochondrial damage in regulating the cleavage of GRIM-1. Therefore, we transfected GRIM-1β-expressing HeLa cells individually with expression vectors coding for Bcl2 (which blocks mitochondrial damage), Bax (which activates mitochondrial damage), wild-type caspase-9 or a catalytically inactive mutant and measured cell death. Cell death via GRIM-1β was robustly induced in the presence of Bax and caspase-9 (Fig. 5F, bars 3 and 4), whereas it was significantly lower in the presence of Bcl2 and mutant caspase-9 (Fig. 5F, bars 2 and 5). Similar results were obtained with GRIM-1α (data not shown). We also ensured the expression of the transfected Bcl2, Bax and caspase-9 by performing a western blot analysis of the extracts with specific antibodies. In all cases, corresponding protein bands were intense (compared with the controls) when the proteins were expressed (Fig. 5G). These results suggested that caspase-9 and mitochondrial damage are important for the induction of GRIM-1β and IFN/RA-driven apoptosis.

Because caspase activation seems to be a crucial step, we next examined whether mutation of potential caspase-cleavage sites would lead to suppression of GRIM-1β-induced apoptosis. On the basis of the observation that the NΔ97 mutant maximally induced cell death (see Fig. 4), similar to GRIM-1γ, and of the size of N-terminal protein generated by caspase cleavage (see Fig. 5), we decided to mutate the most probable caspase-cleavage site, located between positions 110 and 113, in the GRIM-1β protein: YLAD to YLAA. The effect of caspase-9 on the cleavage of the N-terminus of this mutant was determined by performing a western blot analysis with FLAG-tag-specific antibodies (Fig. 6A). Empty pCXN2 vector or the same vector carrying wild-type or mutant GRIM-1β were transfected along with wild-type caspase-9 into Casp-9^−/− MEFs. As expected, wild-type GRIM-1β, but not the mutant GRIM-1β, was cleaved only in the presence of caspase-9. Such a result was not due to a difference in expression levels of either mutant GRIM-1β or caspase-9. The incomplete cleavage might be due to low caspase activity in the transfectant, unlike under the conditions of IFN/RA treatment (see Fig. 5C), which unleashes many cooperative pathways. The negative control, empty-vector transfection, did not yield any signals, showing the specificity of detection.

To test the biological relevance of these observations, Casp-9^−/− MEFs were transfected with wild-type and mutant GRIM-1β expression vectors in the absence and presence of caspase 9 (Fig. 6B). Although the magnitude of apoptosis was low in these MEFs, there were clear differences. A small but significant (P<0.05) apoptosis was observed upon transfection of either wild-type GRIM-1β or caspase-9; the level of apoptosis was synergistically induced when both proteins were present in the cells. Wild-type GRIM-1β potently induced apoptosis in the presence of caspase-9 (P<0.01). Unlike wild type, the mutant GRIM-1β failed to promote apoptosis. Thus, caspase-9 and the YLAD site in GRIM-1β were required for promoting apoptosis. Consistent with these results, GRIM-1γ, which lacks these sites, equivalently induced apoptosis in the absence and presence of caspase-9 in a control experiment (Fig. 6C).

As said earlier, GRIM-1 is homologous to the yeast protein, Shq1p. Yeast mutants depleted of Shq1p exhibit growth retardation owing to defects in rRNA maturation (Yang et al., 2002). Because GRIM-1 isoforms seem to be potent growth suppressors, we next examined their effects on rRNA processing. We used 18S RNA as a model for these experiments. Total RNA purified from HeLa cells expressing Myc-tagged GRIM-1 isoforms was converted to cDNA using a specific primer in the invariant region of 18S rRNA. To distinguish precursor versus mature 18S rRNA, we used primers that overlapped the processed region and the invariant 18S region, representing unprocessed and total rRNA, respectively, in PCR (Fig. 7A). In these reactions, a higher level of PCR signal, i.e. an early threshold cycle (C_t), with junction-specific primer(s) would be indicative of more unprocessed rRNA. Raw data from the junction-specific primer was normalized using the invariant primer region and represented as the fraction of unprocessed rRNA. Expression of GRIM-1 isoforms increased the fraction of 5′ETS (external transcribed spacer)-18S unprocessed rRNA (Fig. 7B). As observed earlier, a differential effect of GRIM-1 isoforms was noted in rRNA processing, with GRIM-1γ being the most potent inhibitor of processing followed by GRIM-1β and GRIM-1α, compared with vector control. Thus, the growth-suppressive ability of GRIM-1 isoforms can be, in part, caused by the suppression of rRNA processing.

Lastly, to determine the relevance of GRIM-1 to tumor growth, MCF-7 cells expressing individual Myc-tagged GRIM-1 isoforms (see Fig. 3) were transplanted into athymic nude mice and tumor growth was monitored over a period of 12 weeks (Fig. 8A). The GRIM-1α, GRIM-1β and GRIM-1γ-expressing tumors grew significantly (the respective P values are <0.05, <0.01 and <0.005) slower than control tumors. At the end of 12 weeks, the average size of the control tumor was 128 mm³. When compared with the control tumors, those expressing GRIM-1α, GRIM-1β or GRIM-1γ were significantly smaller, with mean tumor volumes of 95 mm³ (P<0.05), 71 mm³ (P<0.01) and 41 mm³ (P<0.001), respectively. In a complementary experiment, we transplanted cells expressing empty vector (pTKO1) or antisense GRIM-1 (see Fig. 1) into nude mice and tumor growth was monitored (Fig. 8B). Tumors expressing antisense GRIM-1 grew significantly faster than those expressing empty vector (P<0.01). These differences in tumor growth were seen at 6 weeks and continued until the end of the study.

Although the ‘core’ apoptotic machinery, consisting of the members of the BCL2 and caspase families, have been well characterized, it is far from being clear how disparate exogenous and endogenous death stimuli control cell death in a signal- and cell-specific manner. In this report, we isolated a newly identified regulator of apoptosis, GRIM-1, using a genetic technique. The crucial role of GRIM-1 in mediating IFN/RA-induced cell death was highlighted by the following observations: (1) antisense expression of GRIM-1 conferred resistance to IFN/RA-induced death (Fig. 1) and promoted tumor growth (Fig. 8B); and (2) its overexpression promoted cell death (Figs 3, 4 and 5). We have found that IFN/RA was able to induce the expression of two GRIM-1 transcripts in multiple cell types (Fig. 1). In HeLa cells, IFN-β alone induced expression of these mRNAs, but weaker than did IFN/RA, whereas RA had no effect on mRNA levels (data not shown). However, the fact that RA stimulated a similar mRNA (NCBI entry AK001401) in NT2 neuronal precursor cells, although the protein sequence differed at the N-terminus, was noteworthy.

GRIM-1 seems to be orthologous to shq1 of Saccharomyces cerevisiae and Schizosaccharomyces pombe, and to other undefined proteins coded by Drosophila melanogaster and Caenorhabditiselegans (Table 1). The depletion of Shq1p in yeast causes growth retardation due to a defect in rRNA processing, thus highlighting its importance for cell growth (Yang et al., 2002). By contrast, we found that human GRIM-1 isoforms differentially suppress rRNA processing by acting as inhibitors (Fig. 7). The differential behavior of these two proteins could be due to intrinsic differences in their structure and/or regulation by other factors. These issues are currently being investigated.

Surprisingly, translation of a transcript derived from GRIM-1 cDNA yielded three proteins in vitro (Fig. 2C). These observations are consistent with the expression of multiple proteins in cells as detected by the polyclonal antibodies (Fig. 2). Analysis of the DNA sequences around the putative start codons corresponding to GRIM-1 isoforms revealed the presence of a suboptimal Kozak sequences. Two crucial bases required for optimal initiation are the A in the −3 position and the G in the +4 position (Kozak, 1999). For GRIM-1α, neither the A–3 nor the G+4 were present (Fig. 2); although GRIM-1β and GRIM-1γ both have an A–3, neither have the G+4 base. Therefore, multiple GRIM-1 proteins observed in vivo can be produced either by translational control and/or by differential post-translational modification(s).

Cells expressing GRIM-1 isoforms grew significantly slower than the controls in vitro (Fig. 3) and in vivo (Fig. 8), thus highlighting their anti-tumor property. The anti-tumor activity in GRIM-1α- and GRIM-1β-expressing tumors might not be due to the production of GRIM-1γ as we did not observe a form consistent with GRIM-1γ in these tumors (data not shown). Such lack of expression of GRIM-1γ from GRIM-1α and GRIM-1β mRNAs could also be due to the presence of an ideal Kozak sequence upstream of the translational start site and/or the 3′UTR of β-globin in these expression vectors, unlike the native mRNA. In summary, each of these proteins seems to exhibit differential anti-tumor properties. GRIM-1 seemed to activate apoptosis independently of other death regulators, such as p53, and this was supported by several observations. HeLa and Cos-7 cells lack endogenous p53 protein owing to degradation or physical binding by the viral proteins HPV-E6 and SV40 T-antigen, respectively (Finlay et al., 1989; Levine, 2009; Scheffner et al., 1990). Although MCF-7 cells possessed a wild-type p53, its inactivation by an overexpressed HPV-E6 did not substantially affect GRIM-1-driven apoptosis (data not shown). MCF-7 cells also lack caspase-3 owing to a genetic deletion (Janicke et al., 1998). On the basis of these observations, we propose that GRIM-1 induces apoptosis independently of p53 and caspase-3. Therefore, we investigated a mechanism of its activation.

Another mechanism that generates high levels of GRIM-1 isoforms in vivo occurs via a caspase-9-dependent conversion of high-molecular-weight forms of GRIM-1 to shorter death-activating forms in response to IFN/RA (Figs 5 and 6). An N-terminal sequence, present in GRIM-1α and GRIM-1β, seems to act as a negative regulator of its apoptotic activity. Deletion of the N-terminal sequences dramatically induced apoptosis (Fig. 4). It is possible that this domain folds back on to the C-terminus to prevent the apoptosis-inducing capacity of GRIM-1α and GRIM-1β. Indeed, it was recently shown that yeast Shq1p folds into two independent domains that contain sites of casein kinase 1 phosphorylation (Godin et al., 2009). Alternatively, it might associate with other undefined protein(s), which holds GRIM-1α or GRIM-1β in an inhibitory conformation. This region harbored three potential caspase-cleavage sites. It seems that one of the three potential cleavage sites present in this area, the third site (YLAD), is preferred for cleavage. Mutation of this site not only resulted in a failure to undergo caspase-9-dependent cleavage of GRIM-1β but also suppressed apoptosis (Fig. 6). Although it is theoretically possible for caspase-9 to play a role in GRIM-1γ-induced apoptosis, currently we do not have any data to that extent.

The role of caspases was based on our additional observations that, in the presence of Z-VAD-fmk and/or mutant caspase-9, cell death and IFN/RA-induced cleavage of GRIM-1α and GRIM-1β was blocked (Fig. 5). Consistent with these observations, Bcl2 blocked and Bax enhanced IFN/RA-induced and GRIM-1β-dependent apoptosis. Indeed, our earlier studies have shown caspase-9 activation and release of cytochrome c in response to IFN/RA (Angell et al., 2000; Ma et al., 2001). Caspases specifically cleave peptides after aspartic-acid residues (Nicholson and Thornberry, 1997). Recognition of at least four amino acids N-terminal to this site is also required. Several putative caspase-cleavage sites are present in GRIM-1 protein. Interestingly, many of the putative caspase-cleavage sites are conserved only in mammalian and C. elegans proteins, but not in yeast proteins. This is consistent with the fact that yeast does not have a known caspase (Kang et al., 1999). The presence of putative protein-interaction domains, phosphorylation sites and caspase-cleavage sites in the GRIM-1 protein suggests a highly ordered regulation of cell execution by GRIM-1, wherein several regulators converge on a single substrate. Indeed, deletion of an N-terminal region in GRIM-1β converts it into a GRIM-1γ-like death activator (Fig. 4). Activation of pro-death proteins via caspase-dependent cleavage has been reported in other cases, e.g. death-promoting activity of mitochondrial regulator BID occurs via caspase-8-dependent cleavage (Li et al., 1998; Luo et al., 1998) and apoptotic activity of CAD, an endonuclease that fragments nuclear DNA, occurs via a caspase-3-dependent mechanism (Enari et al., 1998; Sakahira et al., 1998). Importantly, we have shown a newly identified IFN-inducible gene product that promotes IFN action by slowing rRNA maturation. However, a complete understanding of this regulation requires additional studies, which are currently being pursued.

We have previously reported that antisense-mediated inactivation of two other proteins, GRIM-12 (also known as TR; thioredoxin reductase) (Hofmann et al., 1998) and GRIM-19 (Angell et al., 2000), also suppressed cell death in response to IFN/RA. GRIM-12 was required for keeping the active sites of caspases in reduced state through its substrate thioredoxin (Ma et al., 2001). The second protein, GRIM-19, inhibits transcription factor STAT3 (Zhang et al., 2003), a protein known to upregulate the expression of mitochondrial antiapoptotic regulators Bcl2, Bcl-X_L and Mcl-1. It also represses the cell-death inducer FAS (Ivanov et al., 2001). We have shown that GRIM-19 antagonizes these functions of STAT3 to promote tumor suppression (Kalakonda et al., 2007). Indeed, we have recently documented loss of GRIM-19 in a number of primary tumors (Alchanati et al., 2006; Zhang et al., 2007). Loss of permeability of the mitochondrial membrane results in a release of apoptogenic proteins – including cytochrome c (Kluck et al., 1997; Liu et al., 1996), which is required for the activation of caspase-9. In our earlier studies, we have shown activation of caspase-9 (Angell et al., 2000; Ma et al., 2007) and cytochrome-c release in response to IFN/RA (Ma et al., 2001). In this study, we have shown that mutant caspase-9 and Bcl2 block the processing of GRIM-1. On the basis of our current data, we propose that GRIM-19 and GRIM-12 act upstream of GRIM-1 in the cell-death pathway regulated in response to IFN/RA treatment (Fig. 9). In summary, we identified a novel death-regulatory protein, whose activation by caspase-dependent mechanism(s) seems to contribute to apoptosis.

Human IFNβ (Biogen), murine Ifnβ (R&D Systems), RA (Sigma), Z-VAD-fmk (Calbiochem), Ni-chelation sepharose (Novagen), ECL reagents (Pierce), HRP-coupled secondary antibodies (Amersham), hygromycin B (Boehringer Mannheim), DAPI (Sigma), and GFP-specific and caspase-9-specific antibodies (Santacruz Biotech) were employed in these studies. Lipofectamine-Plus (Invitrogen) was used for routine transfections as per the manufacturer's recommendation. Fresh stocks of RA were prepared in ethanol and added to cultures under subdued light.

Individual GRIM-1 isoforms were expressed as Myc-tagged (at the C-terminus) proteins using pCXN2-Myc vector (Hofmann et al., 1998), as FLAG-tagged (N-terminus) proteins from pCXN2-FLAG vector and/or as GFP-tagged proteins using pEGFP-C2 (Clontech); expression vectors for wild-type and catalytically inactive caspase-9 were provided by Srinivasa M. Srinivasula (NCI, Bethesda, MD), and Bax and Bcl2 were provided by Richard J. Youle (NIH-Bethesda, MD). To knock down endogenous GRIM-1 and caspase-9, antisense GRIM-1 in pTKO1 (Hofmann et al., 1998) and CASP9-specific shRNA in pLKO1 (Open Biosystems) were used, respectively. Lentiviral particles coding for GFP-tagged GRIM-1 isoforms were cloned into pLVX-Puro (Clontech) and shRNA constructs were produced as in our earlier publication (Gade et al., 2008). Deletion and site-directed mutagenesis, and northern and western blot analyses were performed as described earlier (Hofmann et al., 1998).

HeLa cells were cultured in DMEM containing 5% charcoal-stripped FBS, non-essential amino acids, L-glutamine and antibiotics. MCF-7 and T47D cell lines were cultured in Phenol-Red-free EMEM containing 5% charcoal-stripped FBS, non-essential amino acids, L-glutamine and 10⁻¹¹ M estradiol during IFN/RA treatment. The BT-20 cell line was cultured in similar media with Phenol Red but supplemented with 5% charcoal-stripped FBS prior to IFN/RA treatment. Because Phenol Red in culture media exerts estrogenic effects, cells were grown in Phenol-Red-free media 24 hours before treatments were initiated. Casp9^−/− MEFs, provided by S. M. Srinivasula (NCI, Bethesda, MD), were grown in DMEM containing 5% FBS.

Cell growth was measured using a colorimetric assay (Skehan et al., 1990). Briefly, cells (2000 cells/well) were treated with human IFN-β (1000 U/ml) and RA (1 μM) in EMEM with 2.5% charcoal-stripped FBS in 96-well plates, and fixed with 10% tri-chloro acetic acid (TCA) at the indicated time points. One control plate was fixed 8 hours after plating to determine the starting cell numbers. Plates were stained with 0.4% Sulforhodamine B (SRB; Sigma) prepared in 1% acetic acid for 1 hour; washed, dried and bound dye was eluted by adding 10 mM Tris-Cl (pH 10). The absorbance at 570 nm was quantified using a micro plate reader.

A cDNA library was generated using poly-A⁺ RNA derived from the BT-20 cell line treated with human IFNβ (500 U/ml) and RA (1 μM) for 0, 2, 4, 8, 16, 24, 48 and 72 hours, pooled, converted to cDNA and inserted in antisense orientation and expressed from an episomal vector, pTKO1. This library was electroporated into cells and transfectants were selected with IFN/RA as described (Hofmann et al., 1998). The surviving colonies were expanded, and episomal DNA was extracted and transformed into Escherichia coli XL-10 to isolate the potential GRIM genes. Individual antisense GRIM genes were transfected into several cell lines to ensure protection against IFN/RA-induced apoptosis. One antisense clone identified in this manner contained a 1.9-kb fragment corresponding to the 3′ region of GRIM-1 cDNA. This fragment was labeled with ³²P and used as a probe to screen a phage library. After three rounds of screening, two clones (~2.1 kb) were isolated. These clones, however, did not contain the 5′ end of the cDNA. Therefore, a 5′RACE was performed using a commercially available kit (Life Technologies). The RACE product (560 bp) was sequenced and then ligated to the 2.1-kb clone to generate the near-full-length cDNA (~2.7 kb).

GRIM-1 cDNA and the indicated ORFs were subcloned into pGEM-7zf (Promega) under the control of T7 promoter. After linearizing the plasmid DNA (1 μg), with HindIII, it was programmed into rabbit reticulocyte lysate in a coupled in vitro transcription-translation system (Promega) in the presence of ³⁵S-methionine. The resultant products were separated by SDS-PAGE, dried and fluorographed.

Initial attempts to express full-length GRIM-1 ORF did not yield sufficient quantity of the protein. Hence, a cDNA fragment corresponding to the C-terminal 203 amino acids was cloned into pET-32b (Novagen) to generate the recombinant protein in E. coli BL21(DE3). Cells were lysed by sonication and GRIM-1 protein was purified from clarified supernatant using Ni-chelation Sepharose (Novagen) as recommended by the manufacturer. The purified GRIM-1 protein was digested with enterokinase, to remove the tag, and resolved by SDS-PAGE. The band corresponding to purified GRIM-1 peptide was used for antibody production in rabbits and mice.

Cells cultured on cover glass in a 24-well tissue-culture plate were fixed for 15 minutes using 4% paraformaldehyde, permeabilized with 0.5% Triton X-100 in PBS and blocked in 5% BSA before additional processing. DAPI was used to visualize nuclei. Direct or indirect fluorescence was employed to visualize tagged GRIM-1 isoforms. Images were captured using a fluorescence microscope (Olympus BX-FLA, Osaka) fitted with a digital camera (QICAM), and processed by Q-capture Pro 5.1 (Q-Imaging Corporation) or using a confocal microscope (Zeiss LSM 510). Cell numbers were determined using immunofluorescent images from ten randomly selected fields, with each field containing ~60 cells and subjected to statistical analysis with Student's t-test.

Three- to four-week-old athymic nude (nu/nu) NCr mice (Taconic) were used in the study (Lindner et al., 1997). Procedures involving animals and their care were conducted in conformity with the institutional guidelines that comply with national and international laws and policies (EEC Council Directive 86/609, OJL 358, 1 Dec. 1987, and the National Institutes of Health Guide for the Care and Use of Laboratory Animals, NIH Publication No. 85-23, 1985). Cells (2×10⁶) were injected into flanks in the mid-axillary line and tumor growth was monitored over a period of 12 weeks. Tumor volume (V) was calculated using caliper measurements and the formula: V=(⁴/₃)πa²b, where 2a = minor axis, 2b = major axis of the prolate spheroid. Student's t-test was used to assess the statistical significance of difference between pairs of samples.

Total RNA was converted to cDNA using an rRNA-specific primer (supplementary material Table S2). Unprocessed 5′ETS-18S and processed 18S rRNA fraction were obtained using specific primer pairs (supplementary material Table S2) as shown in Fig. 8. Unprocessed 5′ETS-18S rRNA was represented as fraction of total 18S rRNA pool using the Δ^Ct method (Nolan et al., 2006).

D.V.K. is supported by NIH grants CA105005 and CA78282. Deposited in PMC for release after 12 months.