Expression of small nucleolar RNAs in leukemic cells

Introduction

The discovery that the majority of the mammalian genome is transcribed into non-protein-coding RNAs (ncRNAs) has changed the landscape of human genetics [1]. Recent work has established that these RNAs are involved in normal cellular processes as well as pathological conditions such as cancer, heart diseases, immune disorders, and various neurodegenerative and metabolic diseases [2]. Several classes of ncRNA have been identified, including small interfering RNAs (siRNA), microRNAs (miRNA), PIWI-associated RNAs (piRNA), small nucleolar RNAs (snoRNA), and various species of long ncRNA.

The small non-coding RNA molecules of the nucleolus, snoRNAs, can be grouped into two main classes. One class comprises the C/D box snoRNAs, which share the evolutionarily conserved C (RUGAUGA) and D (uCUGA) box motifs. The function of these RNAs is to guide methylation of specific target RNAs, and they associate with the nucleolar protein fibrillarin, which catalyzes the transfer of 2′-O-methyl groups [3]. Members of the other class, the H/ACA box snoRNAs, share the conserved box H (AnAnnA) and ACA motifs. These RNAs guide pseudouridylation by associating with dyskerin, the enzyme responsible for catalyzing the uridine-pseudouridine isomerization reaction [3]. An interesting group of snoRNAs, the small Cajal body-specific RNAs (scaRNAs), contains composite RNA molecules that guide both methylation and pseudouridylation of the spliceosomal small nuclear RNAs (snRNAs) [4, 5].

snoRNAs associate with a number of protein components to form the functional small nucleolar ribonucleoprotein (snoRNP) complex, which then carries out the RNA modification reactions. The snoRNAs are responsible for the selection of target sequences by transient complementary base-pairing with the target RNA molecules, which are in most cases ribosomal RNAs (rRNAs). In addition to directing the modification of rRNA, snoRNPs modify other RNAs, including spliceosomal snRNAs, some transfer RNAs (tRNA), and messenger RNAs (mRNA) [6].

In vertebrates, the majority of both C/D and H/ACA box snoRNAs are encoded within introns of protein-coding genes [7–10], whose protein products in many cases have some role in protein synthesis [11]. A common feature of the genes that host intronic snoRNAs is that most of them encode 5′ terminal oligopyrimidine (5′TOP) mRNAs, which are classically thought to produce proteins associated with translation [12, 13]. Intronic encoding provides an economical mechanism for simultaneous production of both the protein and RNA components needed in protein synthesis. After splicing of the pre-mRNA, the snoRNAs are removed from the host introns by exonucleases and processed further [10, 14–16]. While most snoRNAs have been found in protein-coding host genes, some reside in genes with no apparent protein-coding potential [12, 17, 18]. Such genes are presumed to act as vehicles for snoRNA production, although a recent report described additional snoRNA-independent functions for the snoRNA-host RNA, Zfas1 [19]. Indeed, the expression strategies of snoRNAs show great diversity, ranging from different types of independent coding units with specialized promoters to intronic coding units without independent promoters. Encoding of snoRNAs can be organized in clusters or individually, and frequently a single host gene may encode several snoRNAs in adjacent introns, according to the “one snoRNA per intron” principle [20]. Thus, the expression strategies and genomic organization of snoRNAs are variable and complex, and our understanding of them is far from complete.

The general features of the archetypal snoRNAs have been known for some time. However, in addition to the stereotypic snoRNAs, recent work has identified several snoRNAs with structural and/or functional features typically associated with other types of ncRNA, adding a further layer of complexity. For example, mature snoRNAs can be further processed to produce short RNAs with miRNA-like functions [21–23], suggesting a connection between processing and silencing of RNAs. As new studies shed light on the relationships between different ncRNAs, unexpected complexities and unorthodox features are gradually emerging in the interplay between these RNA molecules.

The Next-generation sequencing technologies have played a crucial role in the high-throughput detection of ncRNAs. In the present work, we analyzed the expression patterns of snoRNAs in several leukemic cell lines using SOLiD, a massive parallel sequencing platform. Our results show that the expression patterns of snoRNAs comply with the classical subtyping of leukemias based on morphology and flow cytometry. The expression levels of individual snoRNAs analyzed by RT-qPCR confirm those revealed by SOLiD sequencing. Our results suggest that snoRNA profiling bears potential as a diagnostic tool for leukemia. Revealing the mechanisms that regulate snoRNA expression, and their possible role in pathogenesis, is of considerable interest and warrants further studies.

Materials and methods

Results and discussion

In order to study the expression of snoRNAs in leukemic cells, fractions representing small RNAs from various leukemic cell lines were analyzed by massive parallel sequencing (SOLiD 4, Life Technologies), (see Supplementary Fig. 1). The chosen cell lines included representatives of acute myeloid leukemia (AML), acute lymphoblastic leukemias (T-ALL and pre-B-ALL), Burkitt’s lymphoma/leukemia (BLL), and a solid tumor (HeLa cell line) (see Table 1). The data were analyzed to detect global differences in the expression values of snoRNAs. In our dataset, hierarchical clustering of samples based on tag count values grouped the cell lines according to the leukemia subgroup (Fig. 1). These results show that different cancers and subgroups of leukemia have unique expression patterns of snoRNAs, and that various subgroups of leukemia are more closely related to each other than to a solid tumor (cervical carcinoma represented by HeLa cells).

In order to find differences among various leukemia subgroups and not individual cell lines, we compared the differentially expressed (DE) features between the subgroups (i.e. AML, pre-B-ALL, T-ALL) by using the fold change (FC) values and p-values. For filtering out differentially expressed genes, the minimum fold change was set at 1.5, and the p-value significance level at 0.05. This yielded 46 differentially expressed snoRNAs in the comparison of T-ALL with pre-B-ALL, of which 29 were up-regulated and 17 down-regulated. C/D box snoRNAs were affected the most: 19 of the up-regulated and 15 of the down-regulated snoRNAs were of the C/D box type. Interestingly, there were several scaRNAs among the up-regulated snoRNAs, the most prominent of which was scaRNA6. In the comparison of ALL (including T-ALL and pre-B-ALL cells) with AML, 40 differentially expressed snoRNAs were found, of which 22 were up-regulated and 18 down-regulated. In this comparison, the C/D box snoRNAs also stood out: 17 of the up-regulated and all of the down-regulated were C/D box snoRNAs. A heat map clustering of the DE features is shown in Fig. 2. These results indicate that the expression pattern of snoRNAs is dependent on the subgroup of leukemia, and that various subgroups of leukemia show marked differences in their snoRNA expression patterns.

Next, we focused on the differentially expressed snoRNAs from the T-ALL/pre-B-ALL comparison. In order to validate the expression profiles obtained by SOLiD analysis, the expression of nine individual snoRNAs was analyzed by RT-qPCR. These nine snoRNAs included both up- and down-regulated snoRNAs, and two scaRNAs (Table 2). Independent small RNA fractions were prepared from the T- and pre-B-ALL cell lines. First, we analyzed the leukemic cell lines used in the original SOLiD analysis. The RT-qPCR analyses showed changes in expression similar to those in the original analyses for eight out of nine snoRNAs (Supplementary Fig. 2), corroborating the results of the SOLiD analysis. In order to see if these results can be substantiated on a larger scale, we chose additional leukemic cell lines of T cell origin (Jurkat, P12-Ichikawa, Peer) and pre-B cell origin (Kasumi-2, Nalm-6, REH) (see Table 1). The expression of the nine DE snoRNAs was tested by RT-qPCR in the five T-ALL and five pre-B-ALL cell lines, and as shown in Fig. 3, six out of nine snoRNAs showed changes in expression similar to those in the original SOLiD data set. Among these were four box C/D snoRNAs (SNORD24, SNORD44, SNORD82, SNORD105) and two scaRNAs (scaRNA6 and scaRNA9). Statistically significant differences in expression between the pre-B-ALL and T-ALL subgroups could be seen in cases of scaRNA9 and SNORD24. It is worth noting that there was a marked variation in the expression of individual snoRNAs even within the leukemia subgroups, probably reflecting the unique characteristics of individual cell lines. We also analyzed the expression of the six snoRNAs by RT-qPCR using cDNA derived from total RNA (Supplementary Fig. 3). While scaRNA6 and scaRNA9 showed similar differences in expression, SNORD105 showed opposite difference in expression, and SNORD24, SNORD44 and SNORD82 were expressed at a very low level, making the interpretation unreliable. These latter results suggest that some low-expressed snoRNAs may be missed when using total RNA as a template.

Only a few reports discussing these snoRNAs have been published, and detailed descriptions are available for only SNORD24 [27] and SNORD82 [28]. SNORD44, encoded in the multi-snoRNA host gene Gas5 (growth arrest-specific transcript 5) [18], has been studied in cancer patients and a low expression of this snoRNA was shown to correlate with a poor prognosis [26]. Gas5 was also found to play an important role in normal and transformed T-lymphocytes [29]. scaRNAs are known to be involved in post-transcriptional modification of snRNAs [4, 5].

While most snoRNAs are located in introns and co-transcribed with the host gene transcripts, a small number appears to be under the control of independent promoters [20]. Expression of snoRNAs may also be regulated by differential splicing of their host introns, since removal and disbranching of introns is essential for snoRNA production.

Valleron et al. [30] recently reported a global down-regulation of snoRNA expression in acute leukemias as compared to healthy blood samples. Specific snoRNA signatures were identified in leukemias of different subgroups, a finding that mirrors our results. Furthermore, they identified a locus containing SNORD112-114 as an ectopic expression cluster specific to acute promyeloid leukemias (APL). Expression of SNORD112-114 was found to be regulated by the APL-associated fusion gene PML-RARalpha, but the expression of individual snoRNAs in the cluster was independently regulated. Another recent report described down-regulation of a set of H/ACA box snoRNAs by the growth suppressor GRIM-1, which led to disruption of normal rRNA maturation [31].

There are only a handful of reports dealing with the association of snoRNAs and cancer. The U50 snoRNA has been implicated in prostate and breast cancers [32, 33]. In prostate cancer, a specific mutation in U50 was found to be associated with clinical manifestation of the disease [32]. In surgical specimens of non-small cell lung cancer (NSCLC), six snoRNAs were found to be overexpressed in the tumor tissue [34]. One of them, SNORA42, was reported to act as an oncogene in NSCLC, and its high expression in clinical samples correlated with poor survival [35]. Mourtada-Maarabouni et al. [36] discovered that Gas5, an RNA transcript hosting 10 snoRNAs (see above), regulated growth and apoptosis in mammalian cell lines, and that levels of the Gas5 transcript were significantly reduced in breast cancer samples.

Two classes of small ncRNA, miRNAs and snoRNAs, have both been associated with cancer. However, there are marked differences in the mechanisms of action and the molecular sizes between snoRNAs and miRNAs. While snoRNAs mainly direct modifications of rRNA, miRNAs usually bind to the 3′ untranslated region (3′-UTR) of a mRNA, resulting in its degradation or a block in its translation. Mature miRNA molecules are very short (~22 nucleotides), whereas C/D box snoRNAs are approximately 60–200 nucleotides, and H/ACA box snoRNAs 120–250 nucleotides, in length [22]. Nevertheless, similarities have also emerged between these two classes of ncRNA. Some snoRNAs have been shown to undergo further processing into smaller molecules with miRNA-like properties [23, 37], and some miRNA precursors have been found to resemble snoRNAs in their mode of action [38]. The involvement of miRNAs in normal as well as aberrant hematopoiesis is now well established [39, 40]. Several reports of miRNA profiling from patient serum, plasma, saliva, and other tissues highlight the diagnostic potential of these RNAs in the detection and monitoring of small tumors or those located deep in tissues [41–47]. Of particular relevance to the present work is the finding that snoRNAs can be detected in plasma [34].

Conclusions

We have shown here that various leukemia-derived cell lines carry a specific snoRNA expression signature, and we demonstrated specific differences in snoRNA expression between ALL subgroups. These results indicate that snoRNA expression profiles can potentially be used for the identification of novel subgroups of leukemia, and also hint at the possible involvement of snoRNAs in pathogenesis of leukemia. This work provides an impetus for further expression profiling of snoRNAs in normal and malignant processes.

Electronic supplementary material

Below is the link to the electronic supplementary material.