snoRNAs (small nucleolar RNAs) are key components of snoRNP (small nucleolar ribonucleoprotein) particles involved in modifying specific residues of ribosomal and other RNAs by pseudouridylation (H/ACA snoRNAs) or methylation (C/D snoRNAs). They are encoded within the introns of host genes, which tend to be genes whose products are involved in ribosome biogenesis or function. Although snoRNPs are abundant, ubiquitous and their components highly conserved, information concerning their expression during development or how their expression is altered in diseased states is sparse. To facilitate these studies we have developed a snoRNA microarray platform for the analysis of the abundance of snoRNAs in different RNA samples. In the present study we show that the microarray is sensitive and specific for the detection of snoRNAs. A mouse snoRNA microarray was used to monitor changes in abundance of snoRNAs after ablation of dyskerin, an H/ACA snoRNA protein component, from mouse liver, which causes a decrease in ribosome production. H/ACA snoRNAs were decreased in abundance in these livers while, unexpectedly, C/D snoRNAs were increased. The increase in C/D snoRNAs corresponded with an increase in the abundance of the mRNAs transcribed from snoRNA host genes, suggesting the increase may be part of a cellular response to defective ribosome synthesis.
- C/D small nucleolar RNA (C/D snoRNA)
- dyskerin ablation
- H/ACA small nucleolar RNA (H/ACA snoRNA)
- mouse liver
- ribosome biogenesis
snoRNPs (small nucleolar ribonucleoproteins) are highly conserved ribonucleoprotein complexes essential for ribosome biogenesis and also important for mRNA splicing, telomere maintenance and possibly other functions . There are two classes of snoRNPs: H/ACA snoRNPs which catalyse the formation of pseudouridine (Ψ) from specific uridine nucleotides in newly synthesized ribosomal RNA and spliceosomal snRNAs (small nuclear RNAs)  and C/D snoRNPs which catalyse the methylation of specific residues in these RNAs [1,3]. In both cases the snoRNP complexes consist of a snoRNA (small nucleolar RNA) guide RNA and four proteins. The C/D snoRNAs are associated with fibrillarin, Nop56, Nop58 and the 15.5K/NHPX proteins, and the H/ACA snoRNAs are associated with dyskerin (the pseudouridine synthase), NOP10, NHP2 and GAR1. The latter four proteins are also associated with telomerase RNA, which contains an H/ACA domain [4,5], and mutations in the gene encoding dyskerin are responsible for X-linked dyskeratosis congenita, a disease whose major cause is defective telomere maintenance .
The organization of snoRNA genes in eukaryotes is variable. In vertebrates snoRNAs are mainly intron-encoded and their expression is coupled to the transcription of the host gene. Most host genes encode proteins involved in ribosome biogenesis or function, providing co-ordinated expression of RNA and protein components of the ribosome machinery [7,8]. Some snoRNAs are encoded in introns of genes whose exons do not contain open reading frames and are apparently dedicated solely to the production of snoRNAs .
The process of RNA modification by snoRNAs and, indeed, the proteins associated with snoRNAs are highly conserved in evolution [1,3] suggesting that their role in cellular metabolism is essential. Despite this, very little is known of their role in development or in disease states. This is partly because of the difficulty in identifying snoRNAs using bioinformatics and partly because of the lack of any obvious model of how snoRNAs may influence gene expression. The importance of RNA metabolism in general and ribosome biogenesis in particular have come into focus as important pathways in the control of growth and cell division [10,11], and genetic perturbation of ribosome biogenesis has been found to be the cause of several inherited diseases . Defects in snoRNP production, through mutations in snoRNA genes or in genes encoding their associated proteins, could have a profound influence on cell metabolism and proliferation. To facilitate our studies on snoRNA regulation we have developed a microarray platform for the simultaneous analysis of all known snoRNAs. To characterize and validate this approach we have used a mouse model in which the H/ACA-associated protein dyskerin is specifically ablated from mouse hepatocytes . In this model we have found that, although dyskerin is essential for cell division, hepatocytes lacking dyskerin can survive for weeks, allowing the study of the effects of dyskerin KO (knockout). In the present study we show, using microarray analysis, that dyskerin ablation leads to an overall decrease in the abundance of H/ACA snoRNAs. Surprisingly the abundance of C/D snoRNAs increases. This increase is partly correlated with increased mRNA levels of snoRNA host genes.
Generation of the Dkc1 conditional KO mice
The generation of the Dkc1lox allele in C57BL/6 mice has been described previously . Mice carrying the Mx1 (myxovirus resistance 1)–Cre transgene were obtained from the Jackson laboratory. The Mx1 promoter is silent in healthy mice, but can be induced to high levels of transcription by administration of interferon-α, interferon-β or synthetic double-stranded RNA. Female Dkc1lox/lox mice were bred on to the Mx1–Cre mice and male Dkc1loxTg:Mx1–Cre mice were selected. These are referred to as KO mice. The Dkc1lox mice were used as controls, and are referred to as WT (wild-type). For induction of the Mx1–Cre transgene, mice (at 5 weeks of age) were injected with poly(I)·poly(C) (250 μg intraperitoneally) every other day for a total of three injections. Mice were killed 4 weeks after the last injection. All experiments on mice were performed under the strict guidelines of the National Institutes of Health and the Institutional Animal Use and Care Committee at Washington University.
RNA and protein analysis
RNA samples from washed livers from 9-week-old WT and KO mice were isolated using TRIzol® (Invitrogen). Total RNA concentration and purity was obtained from an absorbance ratio at 260 nm and 280 nm. Total RNA quality was then determined using an Agilent 2100 bioanalyzer according to the manufacturer's instructions. RNA was used for microarray analysis, qRT-PCR (quantitative real-time PCR) analysis and Northern blotting. For Northern blots, 50 μg of RNA was electrophoresed in 1.5% (w/v) agarose/formaldehyde gels and blotted on to HybondN+ membranes (GE Healthcare) overnight. Oligonucleotides (10 pmol) were end-labelled with [γ-32P]ATP and T4 polynucleotide kinase (Fisher). The temperature of hybridization and washing was determined by the melting temperature of each probe. The snoRNA probes used in Northern blot were: SNO1104, 5′-CAAGGTTGGCTTCCCCACGACGCAGTC-3′; SNO1045, 5′-GAAAGAGGTCCACCCCAGTCT-3′; SNO1023, 5′-TGTTTTTCACTCTGCCCCTTCT-3′; SNO1099, 5′-AGCCAGTGAATAAGGTCAGCAGTT-3′; U39/U55, 5′-CAAGGTTGGCTTCCCCACGACGCAGTC-3′; U30, 5′-TCCAAGTCTCAACAGCAATCATCAGC-3′; SNO1020, 5′-GTCTCTCTCGGGCGCTGTGCCCAG-3′; 5S, 5′-GGTCTCCCATCCAAGTACTAACCAGGCCCGACCCTGCTTAG-3′; and 5.8S, 5′-GCGTTCGAAGTGTCGATGATCAATGTGTCCTGCAATTCAC-3′.
Nuclear protein extraction was performed using NE-PER Nuclear and Cytoplasmic Extraction Reagents (Pierce). Protein (20 μg) was resolved by SDS/PAGE (12 % gels) analysis and transferred on to Hybond™ ECL (enhanced chemiluminescence) nitrocellulose membrane (GE Healthcare). Antibodies were detected by ECL Plus Western blotting detection reagents (GE Healthcare) and visualized by autoradiography. Primary antibodies used in this study were: polyclonal antiserum against dyskerin and anti-TATA-box-binding protein antibody (used at 1:2000; Abcam). The secondary antibody was either horseradish-peroxidase-conjugated goat anti-(mouse Ig) (1:10000; Abcam) or horseradish-peroxidase-conjugated goat anti-(rabbit Ig) (1:10000; Abcam).
The probes for all known mouse snoRNAs compiled from the RNA database (http://jsm-research.imb.uq.edu.au/rnadb/) were designed using Agilent earray software (https://earray.chem.agilent.com/earray/) using the ‘Base Composition Methodology’. Probes were antisense because we hybridized labelled RNA to the arrays. Our customized microarray consists of the 145 snoRNA 60-mer oligonucleotide probes, 40 maize (Zea mays) control 60-mer oligonucleotide probes (see below) and the Agilent-provided control probes. The customized microarray was manufactured on an Agilent custom 8×15K microarray and each designed oligonucleotide probe was replicated at random for an average of 81 times per array.
To test the specificity of the microarray, we used a series of model sequences with varying degrees of similarity to the probes. We designed two set of probes with 0–5 nucleotide mismatches for each of two control maize RNAs. RNA oligonucleotides, complementary to the perfect match sequences, were designed, synthesized by Integrated DNA Technologies and added to the mouse RNA before labelling.
Four total RNAs (two KO and two WT) from liver tissue, were chemically labelled at two input masses (10 μg and 1.5 μg) with the Kreatech ULS (universal linkage system) RNA-labelling kit. RNA was mixed with the maize control RNA oligonucleotides (1 μg each), Kreatech 10× labelling buffer and 2 μl of Kreatech Cy3 (indocarbocyanine)–ULS reagent. The reactions were incubated at 85 °C for 15 min in the dark and placed on ice for 3 min. Labelled RNAs were purified with Kreapure gel columns according to the manufacturer's protocol. The labelled RNAs were then quantified on a NanoDrop fluorometer at A260, to determine RNA concentration, and A530, to measure for Cy3 dye incorporation.
Labelled RNAs (with the volume reduced to 7.5 μl with a SpeedVac) were mixed with 12.5 μl of Kreablock, 5 μl of Agilent 10× Block and 25 μl of Agilent 2× gene expression Hi-RPM buffer. The hybridization solution was heated to 70 °C for 5 min and allowed to cool to room temperature (23 °C). Each hybridization solution (45 μl) was then added to the customized mouse snoRNA microarrays and hybridized for 18 h at 65 °C. The Agilent hybridization oven rotator was set at 10 rev./min. Arrays were washed according Agilent Gene Expression wash protocol.
Slides were scanned on a Molecular Devices Genepix 4000B microarray scanner to detect Cy3 fluorescence. Laser power was kept constant and the photomultiplier voltage was established for all eight arrays on the glass such that the high-input hybridizations had less than 1% signal saturation. Images were quantified with Genepix Pro, version 6.1 (Molecular Devices).
Raw, background-subtracted intensity data was imported into GeneSpring software (Agilent) in order to derive the mean value of replicate spots. This data was then exported and imported into Partek Genomic Suite (St. Louis) and quantile-normalized and log-transformed. Two-way ANOVA analysis was performed with the snoRNA-derived elements, RNA mass and genotype as variables. A false discovery rate of 5% was applied to identify genes differentially expressed between livers from KO and WT mice.
The SYBR Green master mix (Applied Biosystems) was used to follow RNA amplification in real-time. Reverse transcription and PCRs were performed according to the manufacturer's protocols using the SuperScript III first strand synthesis system (Invitrogen). The primers were specifically designed by Primer Express 3.0 software (Applied Biosystems) (Table 1). 28S and β-actin were used as internal positive controls. Fluorescence was quantified using the Applied Biosystems 7500 fast real-time PCR system.
RESULTS AND DISCUSSION
Specificity of the snoRNA microarray
The microarray was composed of 60-mer oligonucleotides complementary to the snoRNA sequences. Total RNA was uniformly labelled as the probe. SnoRNAs in higher organisms sometimes contain families of related sequences that may differ by one or a few nucleotides. It was therefore important to determine the specificity of the hybridization and the sensitivity to nucleotide changes. For this purpose we included on the array two sets of 60-mers designed from a maize genomic DNA sequence that were not homologous with any mouse sequence. These sets consisted of WT oligonucleotides and oligonucleotides with mutations in different positions along the 60-mer (Figure 1). Two synthetic RNA oligonucleotides, which contained the complement of the WT probes, were spiked into the samples and labelled along with the liver RNA. The results from this analysis show hybridization efficiency is sensitive to the number, proximity and location of mismatches within the probe. Hybridization was relatively insensitive to changes close to (within 13 base pairs of) the 3′-end of the immobilized oligonucleotide probe, a phenomenon observed previously with a microRNA array . This is the end of the probe that is attached to the glass, leading us to suspect a possible steric effect. We therefore suggest including a non-hybridizing linker sequence at the 3′-end of the oligonucleotides when it is desirable to distinguish between highly homologous targets. Of the 145 snoRNA 60-mers used in our microarray only three were partially complementary to a second sequence within the mouse transcriptome in which mismatches were confined to the 3′-end of the sequence and none of these alternative sequences have so far been annotated as snoRNAs. We conclude that, based on sequence homology, the specificity of our microarray was suitable for our purpose and the array can effectively differentiate between snoRNAs and other sequences.
Dkc1 ablation in mouse liver
We described previously a line of mice with lox elements inserted into intron 11 and into the 3′-UTR (untranslated region)  of the dyskerin gene. These mice were bred so as to contain the Mx1–Cre transgene, which can by induced by interferon or double-stranded RNA [poly(I)·poly(C)] to produce the Cre recombinase protein . This poly(I)·poly(C) treatment of these mice causes efficient excision of exons 12–15 of the Dkc1 gene, resulting in the loss of detectable full-length or truncated dyskerin protein in all tissues we examined. However, after 1 month, most tissues have near normal levels of dyskerin protein as they have become populated by proliferating cells that escaped the deletion event. In the liver, however, dyskerin levels remain very low even 4 weeks after the final poly(I)·poly(C) dose . In the livers of two Dkc1loxTg:Mx1Cre (KO) mice, one month after poly(I)·poly(C) injection, dyskerin levels were 10% (KO1) and 40% (KO2) of WT (Figure 2).
Immunohistochemistry of liver tissue showed that for each cell dyskerin ablation was all or none, with some cells having no detectable dyskerin and others, having escaped the deletion, having normal amounts . In two control Dkc1lox (WT) animals that did not contain the Mx1–Cre transgene there was no reduction in dyskerin levels.
SnoRNA expression profile in the dyskerin KO liver tissue
Labelled RNAs from KO and WT livers were hybridized to our microarray containing 60-mer oligonucleotides representing all 145 known mouse snoRNAs. Interestingly, all snoRNA signals were robustly strong compared with those of the Agilent negative control spots (negative control at 39.7±7.1 compared with snoRNAs at 456.4±174.7), suggesting that all known snoRNAs are expressed in liver. After normalization of the raw data, PCA (principal component analysis), which detects global trends of gene expression patterns in microarray analysis, showed that the gene expression in the two WT samples was closely clustered and that of the two KO samples was well separated from WT (Figure 3). In PCA, the KO1 sample, in which the dyskerin level was lowest, was more divergent from WT than the KO2 sample. A total of 59 snoRNAs were detected at a higher expression level in the KO samples than the WT and 42 snoRNAs were detected at a lower expression level in the KO samples (multiple testing corrected values of P<0.05) (Figure 4). The significantly altered snoRNAs, along with their host genes and target RNAs, are listed in Table 1.
Eight differentially expressed snoRNAs were randomly selected to validate microarray results by qRT-PCR. The relative change of snoRNA expression assayed using microarray analysis and qRT-PCR were consistent (Figure 5A). The correlation coefficient of the eight snoRNAs between microarray and qRT-PCR was 0.912. Whereas qRT-PCR is regarded to be superior to microarrays for comparative analysis, microarrays offer a high-throughput method that generally captured changes in snoRNA expression.
The microarray results were also validated by Northern blotting (Figure 5B). As the expression level of snoRNA is low in the liver tissue, seven abundant snoRNAs were chosen for Northern blot analysis to observe the change of expression in the KO mice compared with the WT mice, and the size of the product was investigated. Four H/ACA snoRNAs (SNO1104, 1045, 1023 and 1099) significantly decreased, and three C/D snoRNA (U39/U55, U30 and SNO1020) increased, as suggested by the microarray results. Ribosomal 5.8S and 5S RNAs were used as loading controls. Moreover, the intron-containing pre-mRNA was not detected in these Northern blots (Supplementary Figure S1 at http://www.BiochemJ.org/bj/429/bj4290033add.htm), indicating that the amount of the precursor is very small compared with that of the mature snoRNA.
Decreased levels of H/ACA snoRNAs, but increased levels of C/D snoRNAs in the absence of dyskerin
Of the 50 H/ACA snoRNAs represented on the microarray, 40 significantly decreased in abundance in the KO livers. This is not surprising as dyskerin is a crucial component of H/ACA snoRNPs and is involved in the post-transcriptional assembly of the snoRNP, as well as in providing the pseudouridylation activity. In the absence of dyskerin snoRNP assembly cannot occur and presumably snoRNAs are degraded. The level of the reduction in H/ACA snoRNA abundance is much less than the decrease in dyskerin levels. This could be explained if there was a rapidly turned over component of dyskerin, which is in excess of snoRNAs, and a second more stable component in mature snoRNPs. Further experiments will be necessary to determine the exact relationship between dyskerin levels and those of individual snoRNAs.
We did not expect the relative abundance of C/D snoRNAs, which do not associate with dyskerin to increase. However, of 95 C/D snoRNAs represented on the microarray, 59 were significantly up-regulated and two were down-regulated in the dyskerin KO livers. A possible explanation is that the transcription of the snoRNA host genes, which are almost all involved in ribosome biogenesis, is higher in the KO hepatocytes. This would constitute a mechanism for increasing ribosome biogenesis in response to its attenuation caused by lack of dyskerin. Whereas many of the host genes for C/D snoRNAs encode ribosomal proteins, the host genes for nine C/D snoRNAs are non-protein-coding multiple snoRNA genes, namely UHG (U22 host gene) and gas5 (growth arrest-specific transcript 5) . UHG and gas5, as well as most other snoRNA host genes, belong to a class of genes with a specific sequence in the 5′-UTR, known as the 5′-TOP (terminal oligopyrimidine tract) sequence . Although the 5′-TOP sequence is known as a translational control element it may be that the increase in C/D snoRNAs is due to increased transcription of host genes or alterations in splicing , mediated by the 5′-TOP sequence.
Two C/D snoRNAs, MBII-163 (SNO1003) and MBII-170 (SNO 1004), were exceptional in that they decreased in abundance in the absence of dyskerin. Interestingly these snoRNAs are similar in that they do not have an identified methylation target and that their host genes are not genes whose products, huntingtin-interacting protein 1 and amiloride-sensitive cation channel 1 respectively, are predicted to be involved in ribosome biogenesis . In fact both of these genes are expressed predominantly in the brain. Nevertheless, they produce detectable amounts of snoRNA, according to the microarray analysis, in the liver.
Transcription of host genes
To gain insight into the mechanism whereby expression of H/ACA snoRNAs decrease and that of C/D snoRNAs increase in dyskerin-deficient cells, we measured the mRNA expression level of three host genes of each type by qRT-PCR (Table 2). The mRNA of host genes of the three C/D snoRNAs increased 20–75%. Surprisingly in the light of decreased H/ACA snoRNA levels, the mRNA of the host genes of the three H/ACA snoRNAs also increased 15–25%. This might be expected if there was a global increase in ribosome-related gene transcription and mRNA is more stable than free snoRNA. As unspliced mRNA precursors give rise to both mRNA and snoRNA, we next investigated levels of mRNA precursors for two H/ACA snoRNA host genes and two C/D snoRNA host genes. The results (Table 3) show that precursor levels in three cases were not altered, whereas the RPS8 precursor was twice as abundant in the liver samples after dyskerin ablation. These results indicate that though dyskerin ablation may affect the transcription of host genes, this is not consistently reflected in increased levels of rapidly turned over pre-mRNA. Alternatively the increase in C/D snoRNA abundance may be due to post-transcriptional events. mRNAs encoding proteins involved in ribosome biogenesis can be preferentially recruited to polysomes via the 5′-TOP sequences in their 5′-UTR. This can result in increased stability of the mRNAs , but does not obviously account for the increase in C/D snoRNA level.
As we observed differences between KO1 and KO2 in the PCA analysis, we examined the correlation of the expression of dyskerin and snoRNA. We compared the expression level of snoRNAs in KO1- and KO2-mice as the dyskerin protein level in the KO1 liver was much lower (90% reduction) than that in KO2 (60% reduction) (Figure 2).
Although the dyskerin level of KO1 was 4-fold lower as that of KO2 mice, the expression of most H/ACA snoRNAs in KO1 was approx. 10–20% lower compared with those in KO2 mice. However, the increased expression of C/D snoRNAs in KO1 and KO2 did not show a similar correlation (Figure 6). This may be related to the fact that the level of H/ACA snoRNAs is directly affected by the absence of dyskerin, whereas the increase in C/D snoRNA abundance may be an indirect consequence of dyskerin ablation, though the precise mechanism remains obscure.
We conclude that the absence of dyskerin in mouse liver decreases the level of H/ACA snoRNAs and increases the level of C/D snoRNAs. The snoRNA microarray is an excellent tool to study the expression of snoRNA, and this innovative technical approach will provide valuable insights into the molecular and biological consequences of pathogenic dyskerin mutations. It may also prove useful in exploring the unknown regulation of snoRNA levels in development and disease.
This study was devised by Jingping Ge, Monica Bessler and Philip Mason as part of a long-term study of the pathogenesis of dyskeratosis congenita. All the practical molecular biology was done by Jingping Ge. Seth Crosby and Mike Heinz devised and designed the snoRNA microarray and carried out the hybridization. The data were analysed by Jingping Ge and Seth Crosby. Jingping Ge, Monica Bessler, Philip Mason and Seth Crosby interpreted the results. All authors contributed to the writing of the paper.
This work was supported by the National Cancer Institute and National Institutes of Health [grant numbers CA106995 (to P.J.M.) and RFA-HL079556 (to M.B.)].
We thank Debbie Laflamme for looking after the mice used in the study.
Abbreviations: Cy3, indocarbocyanine; ECL, enhanced chemiluminescence; gas5, growth arrest-specific transcript 5; KO, knockout; Mx1, myxovirus resistance 1; PCA, principle component analysis; qRT-PCR, quantitative real-time PCR; snRNA, small nuclear RNA; snoRNA, small nucleolar RNA; snoRNP, small nucleolar ribonucleoprotein; TOP, terminal oligopyrimidine tract; UHG, U22 host gene; ULS, universal linkage system; UTR, untranslated region; WT, wild-type
- © The Authors Journal compilation © 2010 Biochemical Society