Report

Impaired Control of IRES-Mediated Translation in X-Linked Dyskeratosis Congenita

See allHide authors and affiliations

Science  12 May 2006:
Vol. 312, Issue 5775, pp. 902-906
DOI: 10.1126/science.1123835

This article has a correction. Please see:

Abstract

The DKC1 gene encodes a pseudouridine synthase that modifies ribosomal RNA (rRNA). DKC1 is mutated in people with X-linked dyskeratosis congenita (X-DC), a disease characterized by bone marrow failure, skin abnormalities, and increased susceptibility to cancer. How alterations in ribosome modification might lead to cancer and other features of the disease remains unknown. Using an unbiased proteomics strategy, we discovered a specific defect in IRES (internal ribosome entry site)–dependent translation in Dkc1m mice and in cells from X-DC patients. This defect results in impaired translation of messenger RNAs containing IRES elements, including those encoding the tumor suppressor p27(Kip1) and the antiapoptotic factors Bcl-xL and XIAP (X-linked Inhibitor of Apoptosis Protein). Moreover, Dkc1m ribosomes were unable to direct translation from IRES elements present in viral messenger RNAs. These findings reveal a potential mechanism by which defective ribosome activity leads to disease and cancer.

X-linked dyskeratosis congenita (X-DC) is a rare and often fatal disease characterized by multiple pathological features, including bone marrow failure and increased susceptibility to cancer (1). X-DC is caused by mutations in the DKC1 gene that encodes dyskerin, a protein associated with small RNAs that share the H+ACA RNA motif, including the telomerase RNA (TR), Cajal body RNAs (scaRNAs), and small nucleolar RNAs (snoRNAs) (2). When associated with snoRNAs, dyskerin acts as a pseudouridine synthase to mediate posttranscriptional modification of ribosomal RNA (rRNA) through the conversion of uridine (U) to pseudouridine (ψ) (3, 4). X-DC patient cell lines and mouse embryonic stem cells harboring knocked-in DKC1 point mutations exhibit reduced rRNA pseudouridylation (2, 5). Hypomorphic Dkc1m mice recapitulate many of the clinical features of X-DC and display reductions in rRNA modification, suggesting that impairments in ribosome function may have a causative effect on X-DC pathogenesis (6). However, the role of rRNA modifications in the control of protein synthesis remains poorly understood. In addition, how alterations in the translational apparatus could lead to specific pathological features associated with human disease remains unknown. We investigated the role of rRNA modifications in the control of protein synthesis in order to understand the molecular basis of X-DC.

We first determined whether decreased rRNA pseudouridylation results in impaired general (cap-dependent) translation, using Dkc1m mice. Total protein synthesis rates were no different in primary Dkc1m cells than in wild-type cells (fig. S1), and therefore decreased rRNA pseudouridylation in Dkc1m ribosomes does not impair general protein synthesis. We next hypothesized that the reduction of modified uridine residues in Dkc1m ribosomes might affect the translation of specific mRNAs and hence may not be readily evident when monitoring for general cap-dependent protein synthesis. We therefore established an unbiased proteomics protocol to screen for candidate mRNAs that rely on Dkc1-dependent rRNA modifications for efficient protein translation. In particular, we optimized a glycerol gradient (7) to purify mRNAs associated with translationally active ribosomes (polysomes) from steady-state and activated primary splenic lymphocytes, one of the hematopoietic lineages affected in X-DC (Fig. 1A). Purified polysome-associated mRNAs from wild-type and Dkc1m lymphocytes, before disease onset, were used to hybridize two commercially available mouse cDNA microarrays (7). These microarrays contained in total 1500 spotted cDNAs with a wide variety of biological functions, including cell signaling, cell differentiation, control of the cell cycle and apoptosis, or genes implicated in cancer initiation (810). Using this protocol, we identified 3 out of 1500 mRNAs that were specifically decreased in polysome association in Dkc1m lymphocytes. In particular, the p27 tumor suppressor and the antiapoptotic proteins XIAP (X-linked Inhibitor of Apoptosis Protein) and Bcl-xL showed a significant decrease (25% or greater) in their association with polysomes in Dkc1m cells as compared to wild-type cells. To validate the microarray results, translational control of p27, XIAP, and Bcl-xL mRNAs was monitored in Dkc1m cells (Fig. 1, B and C). Protein levels of these target mRNAs were significantly down-regulated in Dkc1m lymphocytes (Fig. 1, B and C), whereas no differences were apparent in mRNA transcript levels or protein stability (figs. S2 and S3). In addition, mRNAs showing less than a 25% decrease in polysome association did not reveal differences in gene expression (fig. S2). Altogether, these data reveal defects in the translation of specific mRNAs in Dkc1m cells.

Fig. 1.

An unbiased proteomics approach reveals specific translational impairments in Dkc1m mice. (A) Polyribosomal profiles of wild-type (WT) and Dkc1m cytoplasmic extracts from lipopolysaccharide-stimulated primary splenic lymphocytes fractionated on 10 to 50% glycerol gradients. The y axis represents the absorbance at 254 nm (A254), and the x axis indicates fractions collected. RNA was purified specifically from the high-molecular-weight polysome fractions and used to hybridize cDNA microarrays. Translationally inactive ribosomal subunits are shown: 40S, 60S, and monosome 80S. (B and C) Identification and validation of p27, XIAP, and Bcl-xL mRNAs, which are translationally impaired in Dkc1m lymphocytes. (B) Representative Western blots for p27, XIAP, and Bcl-xL in wild-type and Dkc1m lymphocytes. (C) Densitometry analyses of p27, XIAP, and Bcl-xL values normalized against β-actin are shown as graphs. Each value represents the mean ± SD of three independent experiments, and statistical significance is indicated.

Having determined that general protein synthesis is unaffected in Dkc1m cells, we hypothesized that a cap-independent mechanism relying on dyskerin activity may account for differences in the translation of these mRNAs. Two of the mRNAs identified in our screen, p27(Kip1) and XIAP, share a common feature in their mode of translation initiation, because they both harbor an internal ribosome entry site (IRES) element, previously shown to be important for their accurate expression (1113). IRES elements, which are present within a subset of cellular mRNAs, are structured RNAs of variable length that bind the ribosome during translation initiation, thereby bypassing the requirement for some of the cap-binding proteins (14). Although IRES elements are active in normal growth conditions, IRES-dependent translation is favored during specific cellular stimuli such as the induction of programmed cell death or during distinct phases of the cell cycle, when cap-dependent translation is decreased (15). We therefore tested whether decreased XIAP and p27 protein expression in Dkc1m cells was due to a defect in IRES-dependent translation. We initially analyzed endogenous levels of XIAP and p27 proteins under conditions in which IRES-mediated translation is stimulated, using mouse embryonic fibroblasts (MEFs). We examined XIAP protein levels when the MEFs were exposed to γ irradiation, a stimulus that specifically increases IRES-dependent translation of this antiapoptotic factor, thereby providing a survival advantage to the cell (15). A 50% increase in XIAP protein levels was evident in wild-type MEFs after this stimulus, whereas no induction was evident in Dkc1m cells (Fig. 2A and fig. S3, C and D). We next examined the translation of the p27 mRNA during the G0/G1 phase of the cell cycle, which has been shown to be increased through its IRES element (12, 13, 16). Levels of p27 protein were up-regulated by 55% in wild-type cells but this induction was markedly reduced in Dkc1m cells (Fig. 2B and fig. S3, A and B). Therefore, the translation of p27 and XIAP mRNAs in Dkc1m MEFs is greatly impaired after specific cellular stimuli that affect endogenous IRES-dependent translation.

Fig. 2.

Impaired IRES-dependent translation of specific cellular mRNAs identified in the proteomics screen in Dkc1m mice. (A) Representative Western blot of XIAP in the steady state (S.S.) and after γ irradiation (with 0.5 Gy). Induction of XIAP expression (%) is impaired in Dkc1m MEFs after γ irradiation. A densitometry analysis after induction (mean ± SEM) from at least three independent experiments is shown. (B) Representative Western blot of p27 in the steady state and after serum starvation (in 0.1% fetal bovine serum for 16 hours). Induction of p27 expression (%) is impaired in Dkc1m MEFs after serum starvation. A densitometry analysis after induction (mean ± SEM) from at least three independent experiments is shown. (C) Wild-type and Dkc1m MEFs transfected with the XIAP bicistronic reporter mRNA. The ratio of firefly luciferase to Renilla luciferase (Fluc/Rluc) (IRES/cap) activity was analyzed in the steady state or after γ irradiation (with 0.5 Gy). Each value is relative to the wild-type Fluc/Rluc ratio in the steady state, which was set to 1 and represents the mean ± SD of three independent experiments performed in triplicate. Statistical significance is indicated. (D) Wild-type and Dkc1m MEFs transfected with the p27 bicistronic reporter mRNA. The ratio of Fluc/Rluc activity was analyzed in the steady state or after serum starvation. Each value is relative to the wild-type Fluc/Rluc ratio in the steady state, which was set to 1 and represents the mean ± SD of three independent experiments performed in triplicate. (E) Schematic diagram of the full-length (FL) Bcl-xL 5′ UTR in the reverse orientation (R) and two deletions (del 1 and del 2) used in bicistronic assays. M7 Gppp, 7-Methylguanasine; pRF, bicistronic reporter plasmid. (F) MEFs were transfected with the indicated plasmids. The Fluc/Rluc ratio was calculated, and the expression of the pRF empty vector was set as 1. Each value represents the mean ± SD of three independent experiments in triplicate. IRES-dependent translation is still retained when only 250 nucleotides of the 5′ UTR are used (Bcl-xL del1), but its activity is 35% that of the full-length UTR. The Bcl-xL (del 2) mutant indicates that an important functional domain resides with the first 100 nucleotides of the Bcl-xL IRES. (G) Wild-type and Dkc1m MEFs transfected with the Bcl-xL bicistronic reporter mRNA. The ratio of Fluc/Rluc activity was analyzed in the steady state or after γ irradiation (with 0.5 Gy). Each value is relative to the wild-type Fluc/Rluc ratio in the steady state, which was set to 1 and represents the mean ± SD of three independent experiments performed in triplicate.

We next tested whether Dkc1m ribosomes could promote the translation of a reporter mRNA directed by the XIAP and p27 IRES elements. To this end, we used the well-established bicistronic expression system to detect IRES activity (17, 18), in which the first cistron is translated by a cap-dependent initiation mechanism and the second is translated by the preceding IRES element (Fig. 2, C and D). IRES-dependent translation of p27 and XIAP mRNAs has been documented with the use of this bicistronic assay (11, 13, 19). We confirmed IRES-dependent activity of these mRNAs in the primary cells used in our studies and by employing previously reported mutated IRES sequences (fig. S6). To analyze IRES function, we transfected bicistronic reporter mRNAs in wild-type and Dkc1m MEFs and used specific stimuli that favor IRES-dependent translation (Fig. 2, C and D, and figs. S4, A to D and S5, A and B). At first, we confirmed that both wild-type and Dkc1m cells responded equally to these stress stimuli (fig. S4). IRES-dependent translation of XIAP and p27 was specifically affected in Dkc1m cells, and IRES-mediated translational induction of these mRNAs by γ irradiation and serum starvation was reduced (Fig. 2, C and D). Therefore, the translation impairment of XIAP and p27 mRNAs in Dkc1m cells is at the level of IRES-dependent translation.

Because Bcl-xL was the third mRNA identified in our proteomics screen, we next determined the molecular mechanisms for its translational impairment in Dkc1m cells (Fig. 1B). Although Bcl-xL had not previously been reported to contain an IRES element, we tested whether the Bcl-xL 5′ untranslated region (UTR) possesses IRES activity when cloned in a bicistronic vector. These experiments revealed that Bcl-xL contained a functional IRES that could direct the translation of a second cistron and that IRES-dependent translation was abolished when the Bcl-xL 5′ UTR was cloned in a reverse orientation or when deletion mutants were used (Fig. 2, E and F). These findings indicated that a functional IRES element is present –446 to –1 nucleotides upstream of the initiation codon and that IRES-dependent translation is still retained when only 250 nucleotides of the 5′ UTR are used (Bcl-xL del1), but its activity is 35% that of the full-length UTR. We ruled out the presence of cryptic promoters in the Bcl-xL 5′ UTR sequence because RNA transfection of the bicistronic vector retains IRES activity (Fig. 2G), and we confirmed the integrity of the transfected vectors (fig. S5, D and E). Thus, the unbiased proteomics approach identified a cellular mRNA harboring an IRES element in its 5′ UTR, affected in Dkc1m cells. We confirmed that the defect in Bcl-xL translation was at the level of IRES-dependent translation, because Bcl-xL IRES activity in bicistronic assays was impaired in Dkc1m cells as compared to wild-type cells (Fig. 2G). Moreover, the ratio of Bcl-xL IRES/cap activity was increased after γ irradiation in wild-type cells, whereas this was reduced in Dkc1m cells. Taken together, these data demonstrate that Dkc1m cells are impaired in promoting the translation of a subset of mRNAs that share a common mode of translation initiation, directed by an IRES element.

To determine the relative contribution of impaired IRES-dependent translation to X-DC pathogenesis, we undertook a direct genetic approach with one of the target mRNAs affected in Dkc1m cells. The tumor suppressor p27 is a Cdk inhibitory protein that coordinates accurate cell-cycle progression. Haploinsufficiency in p27 expression results in susceptibility to cancer, and p27+/– mice are tumor-prone upon oncogenic challenge, without biallelic inactivation due to loss of heterozygosity (20, 21). We reasoned that reduced p27 translation in Dkc1m mice might contribute to the tumor-prone phenotype of these animals (6). We therefore tested for a genetic interaction between Dkc1 and p27 and focused our analysis on the cell-cycle status of thymocytes, the predominant cell type that shows increased proliferation in p27–/– mice (2224). Neither p27+/– nor Dkc1m thymocytes showed differences in cell proliferation as compared to wild-type mice. Thymocytes from Dkc1m;p27+/– mice displayed a marked increase in S phase progression, similar to p27–/– cells (Fig. 3). Thus, these findings suggest that reductions in p27 IRES-dependent translation may render Dkc1m mice tumor-prone, which is consistent with the cancer susceptibility phenotype manifested in X-DC pathogenesis.

Fig. 3.

Dkc1 cooperates genetically with p27 in cell-cycle control. The cell cycle of thymocytes from indicated genotypes (8 weeks old) was analyzed by flow cytometry after propidium iodide staining. Each bar represents the mean ± SD from six mice. P values are indicated.

To gain additional insight into the molecular mechanisms by which reductions in rRNA modifications affect IRES-dependent translation, we next investigated whether Dkc1m ribosomes are intrinsically impaired in their ability to translate IRES-dependent viral mRNAs via an IRES-dependent mechanism. Many viruses do not possess capped mRNAs and require IRES elements to promote translation initiation. Moreover, the translation of certain viral mRNAs occurs independently of some or all eukaryotic initiation factors (eIFs) of translation employed in cap-dependent translation (14). For example, the cricket paralysis virus (CrPV) IRES directly recruits the ribosome on an initiation codon without any canonical eIFs (25). It has previously been shown that CrPV IRES is active in mammalian cells, and we confirmed those findings in the primary cells used in our studies and by employing previously published CrPV IRES mutants (2527) (fig. S6). To test whether the defect in IRES-dependent translation in Dkc1m cells resides in the ribosome, we used the CrPV IRES as a molecular tool. To this end, we transfected a bicistronic reporter mRNA in which the translation of Fluc is driven by the CrPV IRES element (Fig. 4A and fig. S5C). The translation of the CrPV IRES was severely impaired in Dkc1m cells (Fig. 4A), strongly suggesting that impairments in IRES-dependent translation in Dkc1m cells are attributable to an intrinsic defect in Dkc1m ribosomes.

Fig. 4.

Molecular role of dyskerin in IRES-dependent translation in X-DC. (A) Wild-type and Dkc1m MEFs transfected with CrPV bicistronic vector in mRNA form. Each value is relative to the wild-type Fluc/Rluc ratio, which was set to 1 and represents the mean ± SD of three independent experiments performed in triplicate. (B) Human X-DC B-lymphoblast and fibroblast cell lines electroporated with the bicistronic vector mRNAs as indicated. Each value is relative to the control steady state, which was set to 1 and represents the mean ± SD of three independent experiments performed in triplicate. Statistical significance is indicated. The lymphoblast cells lines were derived from two affected brothers with X-DC [DKC1 (T66A)a (DKC1a) and DKC1 (T66A)b (DKC1b)]. As controls, both a normal lymphoblast cell line derived from the unaffected carrier mother as well as two other normal lymphoblast cell lines were used (7). The DKC1 mutant fibroblast cell line [DKC1 (A386T)] was, in turn, compared to three normal human fibroblast lines. All normal and/or carrier cell lines indicated as controls displayed similar values of Fluc IRES-mediated translation, which were averaged and set to 1. Each value represents the mean ± SD of three independent experiments performed in triplicate.

We further investigated whether point mutations in the DKC1 gene, present in X-DC patients, would result in the same translational defects as in hypomorphic Dkc1m mice. At first, we analyzed the efficiency of CrPV IRES translation, which relies solely on ribosomal subunits to initiate translation in human X-DC lymphoblasts and fibroblast cell lines (7). We observed a specific decrease in CrPV IRES activity in human X-DC patient cells as compared to normal controls (Fig. 4B). We next tested whether translational impairments of particular cellular IRES mRNAs, identified in our proteomics screen, were also evident in X-DC patient cells. A marked decrease in XIAP and p27 IRES-dependent translation was evident (Fig. 4B). Therefore, human X-DC patient cells show a specific defect in IRES-dependent translation. Moreover, mRNAs identified by an unbiased proteomics protocol in Dkc1m cells also show a translational impairment in X-DC human cells, thereby representing the first target genes for X-DC pathogenesis.

The full inactivation of Dkc1 results in a lethal phenotype in yeast, fly, and mouse, supporting the notion that the complete loss of pseudouridine modifications in rRNA may not be compatible with life (2). How then could reductions in the activity of a housekeeping gene required for ribosome modification lead to a complex phenotype associated with X-DC disease? In this study, we addressed this question using an unbiased proteomics approach to uncover translational defects in primary cells derived from hypomorphic Dkc1m mice, which faithfully recapitulate X-DC pathogenesis (6). Our findings indicate that the activity of IRES elements, which regulate the translation of a subset of mRNAs, is affected in Dkc1m cells and X-DC human patient cells, thereby providing a molecular mechanism through which decreased rRNA modifications may result in specific phenotypic consequences. In this respect, our proteomics screen has identified two important cellular mRNAs previously shown to possess an IRES element: the tumor suppressor p27 and the antiapoptotic factor XIAP, which are translationally impaired in Dkc1m mice and X-DC human patient cells. In addition, this screen led to the unbiased identification of a previously uncharacterized IRES element present in the antiapoptotic factor Bcl-xL, and IRES-dependent translation of this mRNA was defective in Dkc1m cells. The deregulated IRES-dependent translation of mRNAs identified in our screen may account for two specific pathological features of X-DC: susceptibility to cancer and bone marrow failure. Because decreases in p27 protein expression are sufficient to result in tumor susceptibility (20), reductions in p27 IRES-dependent translation in Dkc1m mice and X-DC patient samples is likely to account, at least in part, for the cancer susceptibility phenotype present in X-DC. To this end, we have demonstrated a genetic interaction between Dkc1 and p27 required for restricted cell-cycle progression. In addition, bone marrow failure, a hallmark of X-DC pathogenesis, is characterized by increased apoptosis of hematopoietic progenitors and stem cells (28). Many antiapoptotic factors possess an IRES element in their 5′ UTRs, which promotes translation during stress conditions, thereby providing a survival advantage to the cell (15). Therefore, it is tempting to speculate that a combinatorial defect in IRES translation of antiapoptotic factors such as XIAP and Bcl-xL may underlie the bone marrow phenotype in X-DC. Bcl-xL chimeric embryos display increased cell death of hematopoitic progenitors associated with anemia and lymphopenia (29, 30), which are hallmarks of X-DC pathogenesis.

The fact that translation initiation directed from CrPV IRES, which relies solely on the ribosome itself, is severely impaired in Dkc1m cells strongly implies that the molecular defect intrinsically resides in the inability of Dkc1m ribosomes to efficiently engage the IRES element. Taken together, these findings allow us to propose a model whereby reductions in rRNA modifications due to dyskerin malfunction affect the translation of important cellular IRES mRNAs, which may require more direct interactions with the ribosome for translation initiation, thereby contributing to specific pathological features of X-DC. Although we cannot determine how defects in other cellular functions attributed to dyskerin activity may contribute to X-DC, these findings indicate a previously unknown molecular mechanism by which impairments in rRNA modifications affect translation control and lead to disease pathogenesis.

Supporting Online Material

www.sciencemag.org/cgi/content/full/312/5775/902/DC1

Materials and Methods

Figs. S1 to S6

References and Notes

View Abstract

Navigate This Article