NcRNA-microchip analysis

Abstract

Epstein-Barr virus (EBV) infection of human B cells requires the presence of non-coding RNAs (ncRNAs), which regulate expression of viral and host genes. To identify differentially expressed regulatory ncRNAs involved in EBV infection, a specialized cDNA library, enriched for ncRNAs derived from EBV-infected cells, was subjected to deep-sequencing. From the deep-sequencing analysis, we generated a custom-designed ncRNA-microchip to investigate differential expression of ncRNA candidates. By this approach, we identified 25 differentially expressed novel host-encoded ncRNA candidates in EBV-infected cells, comprised of six non-repeat-derived and 19 repeat-derived ncRNAs. Upon EBV infection of B cells, we also observed increased expression levels of oncogenic miRNAs mir-221 and mir-222, which might contribute to EBV-related tumorigenesis, as well as decreased expression levels of RNase P RNA, a ribozyme involved in tRNA maturation. Thus, in this study we demonstrate that our ncRNA-microchip approach serves as a powerful tool to identify novel differentially expressed ncRNAs acting as potential regulators of gene expression during EBV infection.

Introduction

Epstein-Barr virus (EBV), a member of the γ-herpesvirus family, is an abundant human-pathogenic virus with worldwide occurrence and infection rates of more than 95% among adults.¹ EBV infection is etiologically linked to non-malignant diseases, such as infectious mononucleosis (IM) and malignant tumors of lymphoid and epithelial tissues, including Burkitt's lymphoma (BL), Hodgkin's disease (HD), nasopharyngeal carcinoma (NPC) and post-transplant lymphoproliferate disease (PTLD).^1,2 EBV possesses a linear double-stranded DNA genome, exhibiting a genome size of about 172 kb which is embedded in a toroid-shaped protein core.³ Until now, 28 non-protein-coding RNA (ncRNA) genes have been mapped to the EBV genome, including Epstein-Barr virus encoded RNAs (EBER) 1 and 2,⁴ 25 micro-RNAs (miRNAs)^5–7 and a single small nucleolar RNA (snoRNA), designated as v-snoRNA1.⁸

EBV employs viral encoded small ncRNAs from infection of B cells: for example, EBER 1 binds to double-stranded RNA-activated protein kinase (PKR), a key mediator of the antiviral effect of interferon (IFN)-α. Binding of EBER1 to PKR prevents PKR-mediated phosphorylation of the α-subunit of translation initiation factor 2 (eIF-2α), thereby conferring resistance to IFN-α-mediated apoptosis.⁹ In addition, EBV-encoded miRNAs were found to target the 3′-untranslated region (3′-UTR) of viral mRNAs (BALF5, LMP1) as well as host mRNAs (PUMA, CXCL11).^10–13 Decreased expression levels of respective proteins are postulated to modulate EBV infection and persistence.

Recently, the first virus-encoded snoRNA, designated as v-snoRNA1, has been identified.⁸ Interestingly, v-snoRNA1 exhibits sequence complementarity to the 3′-UTR of the viral DNA polymerase BALF5 mRNA. In addition, it has been demonstrated that v-snoRNA1 is processed into a miRNA-like species, which might target BALF5 mRNA, as it was also reported for EBV-encoded miRNA-BART2.¹⁰ Thus, identification of regulatory ncRNAs, associated with EBV infection, will increase our knowledge on the complex interactions between EBV invasion and host defense and might also aid in the therapy of EBV infection and related diseases in the future.

We previously applied an experimental strategy, based on the subtractive hybridization approach, designated as SHORT (subtractive hybridization of ncRNA transcripts), to identify differentially expressed ncRNAs in EBV-infected human B cells.¹⁴ Thereby, the pool of ncRNAs of EBV-immortalized cells, comprised of ncRNAs in the size range from 20 to 500 nt, was subtracted from ncRNAs of non-infected cells. In the present study, by deep sequencing analysis of this SHORT cDNA library, we aimed to identify differentially expressed regulatory ncRNAs, involved in EBV infection.

NcRNAs represent a heterogeneous group of molecules, which vary considerably in size, abundance, sequence and secondary structure. Due to these features, analysis of differential expression of potential novel ncRNAs as well as known ncRNAs by microchip analysis is complicated. Here, we describe the generation of a custom-designed ncRNA-microchip, which was employed to investigate differential expression of novel regulatory ncRNA candidates in EBV-infected human B cells.

Results and Discussion

Deep sequencing analysis of EBV-induced ncRNAs enriched by subtractive hybridization.

The specialized cDNA library, comprised of ncRNAs encoded on the EBV genome or EBV-induced ncRNAs from the host genome, was subjected to deep sequencing analysis (Accession Number SRA010803.⁵). Sequence reads were computationally analyzed and mapped to respective EBV or human genomic loci. After removal of ribosomal, small nuclear and transfer RNA sequences, 7,074 sequence reads were obtained and classified in respect to known ncRNA classes (Fig. 1A). Subsequently, cDNA sequences were grouped into 592 unique ncRNA genes, including 318 entirely novel ncRNA candidates and 274 known ncRNAs (Fig. 1B).

Figure 1.

Distribution of cDNA sequences generated from deep sequencing. A cDNA library from human B cells, generated by subtractive hybridization and comprised of EBV-induced ncRNAs, was subjected to deep sequencing. (A) Distribution of total sequence reads from each ncRNA class (host-encoded sequence reads in black, EBV-encoded sequence reads in red). (B) Distribution of unique ncRNA genes after cluster analysis (host-encoded sequence reads in black, EBV-encoded sequence reads in red).

Potential novel ncRNA candidates, comprised of 313 host-encoded and five EBV-encoded ncRNA candidates, could not be assigned to already known ncRNA genes (Fig. 1B and Suppl. Table 1). Thus, they might represent potential novel functional ncRNAs. Surprisingly, 66 host-encoded novel ncRNA genes were derived from genomic loci annotated as genomic repeats, in particular from the class of Alu repeat elements. Most of the Alu-derived RNA sequences were mapped to multiple genomic loci, due to the short length of sequence reads (less than 50 nt) and the high conservation of nucleotide sequences between Alu repeat elements.

We also identified 258 known, host-encoded ncRNAs, comprised of 7SL RNA, 7SK RNA, RNase MRP RNA, RNase P RNA, vault RNAs and Y RNAs, 146 snoRNAs as well as 101 cellular miRNAs (Suppl. Table 1). In addition, identification of 16 known EBV-encoded ncRNAs served as an internal control and confirmed successful enrichment of EBV-encoded ncRNAs by subtractive hybridization. Thereby, EBV-encoded EBER 1 and 2, eight EBV-encoded miRNAs (and five different passenger strands) and v-snoRNA1 were identified (Fig. 1B and Suppl. Table 1). The large number of known ncRNA genes, identified by deep sequencing, indicated an excellent coverage of small ncRNAs in the specialized SHORT cDNA library.

Generation of a custom-designed ncRNA-microchip for analysis of differentially expressed ncRNA genes.

Only a small number of known and novel ncRNAs, selectively enriched by subtractive hybridization, is expected to be differentially expressed.¹⁵ Several methods are available to investigate differential expression of ncRNAs, such as northern blot analysis, Real Time PCR or microchip analysis. In this study, we focused on a ncRNA-microchip approach, because it enables parallel high throughput analysis of differential expression of hundreds of novel ncRNA candidates.

To that aim, we generated a custom-designed ncRNA-microchip for differential expression analysis of novel and known ncRNAs, identified in our study. Additional probes for expression analysis of known human- and EBV-encoded miRNAs¹⁶ as well as human snoRNAs¹⁷ were included. This enabled us to identify also ncRNAs, whose expression was downregulated in EBV-immortalized cells and which were not enriched by our SHORT approach.¹⁴

DNA oligonucleotide probes were designed by employing the Geniom Client software from Febit. For ncRNAs shorter than 30 nt in length, such as miRNAs, the reverse complement sequence was used as probe on the arrays. For all ncRNAs longer than 30 nt, specific 30 nt long probes were calculated to minimize cross-hybridization. In total, 3,956 DNA oligonucleotide probes for expression analysis of 1,015 unique ncRNA genes were synthesized in situ on the microchip (Table 1). The large number of probes resulted from spotting of additional oligonucleotides that served as negative controls in sense orientation to ncRNAs. In addition, several overlapping antisense probes were designed for expression analysis of ncRNAs longer than 30 nt.

Table 1.

NcRNA-microchip design

NcRNA class	Number of ncRNAs	Number of probes
Host-encoded known ncRNAs	11	48
Host-encoded snoRNAs	222	2036
Host-encoded miRNAs	433	866
Host-encoded novel ncRNAs	318	904
EBV-encoded ncRNAs	2	8
EBV-encoded miRNAs	23	78
EBV-encoded snoRNAs	1	6
EBV-encoded novel ncRNAs	5	10

Design of the ncRNA-microchip, including ncRNA classes, number of ncRNAs and number of probes. Probes were designed in antisense and sense orientation to ncRNAs.

Labeling of ncRNAs is crucial for obtaining sensitive and accurate microchip data. Thus, the mirVanaTM miRNA labeling kit, which has previously been employed for successful labeling of miRNAs,¹⁸ was used. In initial experiments, 20 µg of labeled total RNA was hybridized to the microchip, resulting in high background signals, probably due to large amounts of 28S and 18S rRNAs compared to few ncRNA molecules.

To reduce background signals, size-fractionated RNA was employed for further hybridizations. To that end, total RNA, isolated from EBV-immortalized and non-infected B cells, was size-fragmented from 18 to 500 nt by 8% denaturating polyacrylamide gel electrophoresis (PAGE). No all abundant ncRNAs are eliminated by size-selection (e.g., 5S and 5.8S rRNA, tRNAs). However, the background was significantly reduced, also probably due to purification via PAGE. Size-fractionated RNA was eluted from the gel and labeled as indicated above. Subsequently, 2.5 µg of the size-fractionated, labeled RNA, in five biological replica, of EBV-immortalized and non-infected B cells was hybridized to the microchip (Fig. 2A).

Figure 2.

NcRNA-microchip analysis of differentially expressed ncRNAs. (A) Cluster analysis of ncRNA expression of five biological replica of EBV-infected versus five non-infected control samples by ncRNA-microchip analysis. Downregulated expression is indicated by blue, expression by yellow and upregulated expression by red color. (B) Distribution of differentially expressed ncRNA genes as deduced by ncRNA-microchip analysis. The number of differently expressed ncRNA genes is indicated in respect to their fold change (FC) determined by microchip analysis. The FC value represents the ratio of the hybridization signal of the ncRNA gene expression in EBV-infected compared to non-infected cells. The black dotted line separates the expression of upregulated (left site) and downregulated (right site) ncRNA genes. The non-linear y-axis indicates the number of differentially expressed ncRNA genes. (C) Classification of differentially expressed ncRNAs. A schematic overview of EBV-encoded and host-encoded differentially expressed ncRNAs identified by microchip analysis (numbers of differentially expressed ncRNAs are indicated).

Due to the heterogeneity of ncRNAs, differential expression of ncRNAs by microchip analysis is technically demanding. So far, detection of differentially expressed ncRNAs by microchip analysis has been limited to a single class of ncRNAs, i.e., miRNAs, which exhibit a length of 21–24 nt.¹⁹ We anticipated that secondary structures of longer ncRNAs might impair hybridization of ncRNAs to complementary probes of the microchip. To enable efficient hybridization of highly-structured ncRNAs to oligonucleotide probes, a temperature gradient was employed. The hybridization temperature, starting at 70°C, was reduced in 2–3°C increments per hour to a final hybridization temperature of 39°C over a period of 16 h. Microchip hybridization was performed on the Febit Geniom technology system.^20,21

NcRNA genes that exhibited signal intensities three times above hybridization background, in at least three of five biological replica, were included in this analysis. Thereby, differential expression of ncRNA genes, indicated as fold change (FC), was calculated by the ratio of hybridization signals of EBV-immortalized versus non-infected ncRNAs. The data of the custom-designed ncRNA-microchip analysis were deposited at the Gene Expression Omnibus (GEO; Accession Number GSE20441).

NcRNA-microchip analysis of differentially expressed virus- and host-encoded ncRNAs.

The distribution of differentially expressed, classified ncRNAs, as deduced from ncRNA-microchip analysis, is indicated in Fig. 2B and C. In total, 151 of 1,015 spotted, unique ncRNA genes were identified to be differentially expressed. From those differentially expressed ncRNAs, eleven ncRNAs were found to be encoded on the EBV genome as well as 140 ncRNAs were identified which mapped to the host genome (Fig. 2C).

Differential expression of virus-encoded ncRNAs.

EBV-encoded ncRNAs are expected to be expressed in EBV-immortalized cells, only. Thus, known EBV-encoded ncRNAs served as internal controls to analyze the efficiency of differential expression of the custom-designed ncRNA-microchip and indeed they were detected in EBV-immortalized cells, only (Fig. 2B). In contrast, the majority of host-encoded ncRNAs was not found to be differentially expressed (Fig. 2B).

By the microchip analysis approach, we could not confirm differential expression of five EBV-encoded novel ncRNA candidates due to very low hybridization signals. Northern blot analysis did also not verify differential expression of these ncRNA candidates, which might be explained by expression levels below the detection limits of both methods. In contrast, microchip analysis identified differential expression of EBER 1 and 2, v-snoRNA1, a virus-encoded snoRNA, as well as eight EBV-encoded miRNAs (Supp. Table 2). Differential expression of selected EBV-encoded ncRNAs was subsequently confirmed by northern blotting (Fig. 4A and B).

Figure 4.

Expression analysis of known ncRNAs as deduced by northern blotting. Differential expression of selected known ncRNAs (A), miRNAs (B) as well as snoRNAs (C), as analyzed by northern blotting. Quantification of northern blot and microchip hybridization signals of EBV-infected versus non-infected cells is indicated. Black arrows show differential expression (↑: upregulation of expression, ↓: downregulation of expression).

Differentially expressed host-encoded ncRNAs.

By ncRNA-microchip analysis, from 984 host-encoded ncRNAs in total, 140 were found to be differentially expressed in EBV-immortalized cells (Fig. 2C), comprised of 101 known and 39 novel ncRNAs. Expression levels of novel ncRNAs were also investigated by northern blotting (Fig. 3A). By this method, we confirmed differential expression of 25 of the 39 novel ncRNA candidates, including six uncharacterized ncRNA genes (Table 2) and 19 novel ncRNAs derived from repeat loci of the host genome. From the repeat-derived ncRNAs, 18 ncRNAs originated from Alu repetitive elements and one ncRNA from a long terminal repeat (LTR). Sizes of ncRNAs, as estimated by northern blotting, were larger than sizes deduced from cDNA sequencing, indicating that these sequence reads might correspond to fragments of larger ncRNA transcripts. PCR amplification with hybrid primers might be responsible for truncated sequence reads, because previously employed small-scale Sanger sequencing has yielded full length sequences.

Figure 3.

Expression analysis of novel ncRNAs as deduced by northern blotting. Northern blot analysis of novel non-repeat derived ncRNAs (A) and selected repeat-derived ncRNAs (B). Quantification of northern blot and microchip hybridization signals of EBV-infected versus non-infected cells is indicated. Black arrows show differential expression (↑: upregulation of expression, ↓: downregulation of expression). (C) Predicted secondary structures of novel ncRNAs. Structure predictions were generated with the program RNAfold from the Vienna RNA package.²³ The 0 to 1 color code scale represents base-pair probabilities.

Table 2.

Novel differentially expressed host-encoded ncRNAs

Name	Sequence	SR length (nt)	SR Orientation	Gene information
c15429-A	CC G GGG ACA GGC TGA GGT CGT TGA GGC C	28	1	+/+	chr 6, +str, intron 15–16 of ARID1B
c15308-A	CAG CTG AGG CTC AGC CAT TTC TGC GGG TGG CAG AGC CGG TGA TGA C	46	3	+/−	chr19, -str, intron 1–2 of ZNF 787
c14750-S	CCC TCT GCT GGG GAT GCC ACG TCC GGC TGC CCT GTC CTG TGG CCT CCT CCC GCT GAC GGC TGC GAT CGC TCG GCC GGT CAC	81	4	+/+	chr16, +str, intron 1–2 of HERPUD1
c14715-A	CAT TCC AGG GTG CTG TGG CCG CCT CAC GTA TCC AGA GTG ATG CAG CTC CCT GGG GAC ACA G	61	4	+/+	chr22, +str, intron 5–6 of RANBP1
c18397-A	AGG GCG GGA CGG GGC TGT CTG GGC CGC CTG CGA AGG GCG GAG GGC GCT GCA AGT CCC GCG G	61	2		chr2, -str, intergenic
c15488-A	TGC TGC ACA GAG GGC GGC GG	20	1	+/+	chr14, +str, intron 2–3 of FUT8

Differentially expressed novel ncRNAs, indicating clone names, cDNA sequences, sequence read (SR) lengths (nt), number of sequence reads (SR), their orientation in respect to mRNA strand (+/+, cDNA clones in sense orientation to mRNAs; +/−, cDNA clones in antisense orientation to mRNAs) and gene information.

Differentially expressed novel host-encoded ncRNAs. From the six differentially expressed, non-repeat derived ncRNAs, we identified three up- and three downregulated ncRNAs (Fig. 3A). Identification of downregulated ncRNA candidates is not unexpected: the elimination of abundantly expressed ncRNAs by SHORT enables identification of low abundant, even downregulated ncRNAs. Novel host-encoded ncRNA candidates were predominantly encoded within intergenic or intronic genomic regions which have previously been shown to contain the majority of functional ncRNA species within eukaryal genomes (Table 2). Most of the intron-derived transcripts mapped in sense orientation to the pre-mRNA transcripts, except ncRNA candidate c15308-A, which is located in antisense orientation to the mRNA of zinc finger protein 787 (ZNF 787). Thus, c15308-A might be involved in post-transcriptional regulation of ZNF787 mRNA upon infection by EBV.²²

We also investigated whether novel ncRNA candidates fold into stable secondary structures, a hallmark of many functional regulatory ncRNA species. Secondary structures of novel ncRNA candidates were predicted using the in silico prediction program RNAfold (Fig. 3C).²³ Thereby, we demonstrated that selected novel ncRNAs indeed were able to fold into stable secondary structures and might represent potential novel regulatory ncRNAs (Fig. 3C).

Differentially expressed ncRNA candidates derived from repeat loci. By computational analysis, 66 of the 313 novel host-encoded ncRNA candidates mapped to genomic loci, which were annotated as Alu repetitive elements, indicating that transcription occurs from these repeat gene loci. NcRNA-microchip analysis identified 2- to 5-fold upregulated expression of 22 of these repeat-derived ncRNAs in EBV-immortalized cells (Supp. Table 1). We also confirmed differential expression of Alu-derived ncRNAs by northern blotting. Thereby, we verified that expression levels of 18 Alu-derived ncRNAs were 2- to 5-fold upregulated (Fig. 3B). This is in agreement with threefold upregulated expression levels of 7 SL RNA which have previously been reported by our group upon EBV infection.¹⁴ In general, an excellent correlation between northern blot and microchip analysis was observed.

Alu repeat elements are ancestrally derived from the 7SL RNA gene and exhibit a size of approximately 300 bp in length. With about 1.1 million copies, they represent the most abundant repetitive DNA elements in the human genome.^24,25 Alu repeats belong to the subclass of short interspersed nuclear elements (SINEs), which are members of the class of interspersed repeats and represent transposable DNA segments. As previously reported, Alu repeat elements are highly conserved on the nucleotide level.²⁵ Surprisingly, most cDNA clones of novel Alu-derived ncRNAs in our study deviated from the consensus nucleotide sequences of Alu repeats. It is tempting to speculate that Alu-derived RNAs might serve as a source for the evolution of novel ncRNAs.

In addition, we identified one novel ncRNA candidate, c15817-A, which mapped to a genomic locus annotated as a long terminal repeat (LTRs). Similar to Alu repeats, LTRs from an endogenous retrovirus also represent a class of interspersed repeats, derived from a transposable element, however, LTRs differ from Alu repeat elements due to characteristic nucleotide sequence features. Though differential expression of c15817-A could not be verified by microchip analysis, its expression was found to be upregulated by two-fold in EBV-immortalized cells by northern blotting, indicating a size of approximately 170 nt (Fig. 3B).

Human Alu-derived RNAs are usually transcribed by RNA polymerase III at low levels,^25,26 however, their expression can be stimulated by various stress conditions.^27,28 Therefore, we addressed the question whether increased expression of Alu repeat-derived ncRNAs might represent a general stress response or might be specific for EBV infection. To that end, non-infected B cells were exposed to different stress stimuli (Table 3) and differential expression of two selected Alu-derived ncRNAs, c14061 and c15475, was subsequently investigated by northern blotting. Treatment with stress stimuli did not increase expression levels of Alu-derived ncRNAs in stress-treated B cells to a level comparable to EBV-immortalized B cells (data not shown). It is thus tempting to speculate that EBV infection promotes transcription of Alu-derived RNA transcripts, a hypothesis that has to be further investigated in the future.

Table 3.

Stress treatment parameters

Stress treatment	Dose	Recovery
Heat-shock	30 min/45°C	30 min/37°C
Cold-shock	30 min/45°C	30 min/37°C
UVC	0.120 J/254 nm/1x crosslinking	-
NaCl	500 mOsM/4 h	-
H2O2	100 µM/4 h	-
Cycloheximide	355 µM/4 h	-
Serum deprivation	no FCS/48 h	-

Different stress stimuli, dose and recovery time used for stress treatment of B cells.

Differentially expressed host-encoded miRNAs. We also investigated differential expression of host-encoded miRNAs and their potential impact on EBV infection. Thereby, ncRNA-microchip analysis identified three miRNAs, mir-221, mir-222 and mir-34a, to be more than 20-fold upregulated in EBV-immortalized cells (Supp. Table 1). In addition, expression levels of three miRNAs, mir-193b, mir-193b* and mir-22, were more than 5-fold and twelve miRNAs were more than 2-fold increased, respectively. Differential expression of several of these miRNAs has previously been described by Mrázek et al.¹⁴ In addition, we confirmed elevated expression levels for miRNAs mir-222, mir-22 and mir-24 by northern blotting (Fig. 4B).

Recently, by employing ncRNA-microchip analysis, Cameron et al. have demonstrated differential expression of several host-encoded miRNAs by comparing type III versus type I EBV latency.²⁹ In contrast, in our study we identified differentially expressed, host-encoded miRNAs in EBV-infected versus non-infected cells. Thereby, we observed elevated expression levels of mir-221 and mir-222, which have not previously been reported upon EBV infection. Several recent studies have implicated involvement of mir-221 and mir-222 in tumorigenesis of various types of cancer.^30–32 Thereby, a negative regulation of a tumor suppressor and inhibitor of the cell cycle, p27(Kip1), by mir-221 and mir-222 was described. According to our study, it is tempting to speculate that both miRNAs might also be involved in tumorigenesis of lymphatic tumors, which probably is caused by EBV infection. The question whether p27(Kip1) is regulated by mir-221 and mir-222 in human lymphoid tumors has to be proven in future experiments.

In addition, we identified 32 host-encoded miRNAs, comprised of six miRNAs whose expression was more than 5-fold and 26 miRNAs whose expression was more than 2-fold downregulated in EBV-immortalized cells, respectively (Suppl. Table 2). Expression levels of selected miRNAs, including mir-15b and mir-182, were further verified by northern blot analysis (Fig. 4B). In general, the identification of deregulated expression levels of numerous oncogenic miRNAs upon EBV infection is consistent with the hypothesis that EBV infection is associated with tumorigenesis.

Differentially expressed host-encoded snoRNAs. Surprisingly, we also observed decreased expression levels of 45 host-encoded snoRNAs whose expression was approximately 2- to 3-fold downregulated in EBV-immortalized cells, as deduced from ncRNA-microchip analysis (Supp. Table 2). Reduced expression of several snoRNAs, including U69, ACA43 and HBII-85, was further confirmed by northern blotting (Fig. 4C).

In general, snoRNA species are involved in guiding rRNA and snRNA modifications in eukaryotes. However, the precise roles of these RNA modifications has not been elucidated as of now.³³ A reduced number of rRNA or snRNA modifications, caused by decreased expression levels of numerous host-encoded snoRNAs, might affect efficient folding of rRNA molecules and regulate protein synthesis in EBV-immortalized cells. Alternatively, decreased snoRNA expression levels could also reduce snoRNA-mediated spliceosomal RNA modifications and thus modulate splicing of viral as well as host genes.

Upregulated expression levels of vault RNAs. Though expression of most host-encoded ncRNAs was not affected upon EBV infection (Fig. 2B), we identified four vault RNAs (vtRNAs), i.e., vtRNA1-1, vtRNA1-2, vtRNA1-3 and the recently identified novel human vtRNA2-1, whose expression was significantly upregulated by ncRNA-microchip analysis. The results of microchip analysis are consistent with previous work that demonstrated differential expression of vtRNAs by northern blotting (Fig. 4A).^14,34 VtRNAs have been reported to be components of a large RNP, the vault complex,¹⁴ however, their biological role is still in the focus of current investigations.

Decreased expression levels of RNase P RNA upon EBV infection. Intriguing was the observation that expression of RNase P RNA was 3-fold downregulated in EBV-immortalized cells, as deduced from ncRNA-microchip and northern blot analysis (Fig. 4A). RNase P RNA is the RNA component of the ribonucleo-protein endonuclease that generates the mature 5′-terminus of tRNAs.³⁵ The generation of functional tRNAs is crucial for protein synthesis. Recently, a novel mechanisms of translational gene regulation has been discovered: cleavage of full-length tRNAs resulted in tRNA depletion and impaired translation during sporulation of Aspergillus fumigatus.³⁶ Thus, we addressed the question whether decreased RNase P RNA expression might affect gene expression at the translational level upon EBV infection. Reduced expression levels of RNase P RNA might impair processing of functional 5′-tRNA ends, resulting in accumulation of tRNA precursor transcripts. However, by northern blot analysis, we neither did observe decreased expression levels of full-length tRNAs nor accumulation of pre-tRNA transcripts in EBV-immortalized cells (data not shown).

Material and Methods

Cell lines and EBV strains.

The cell line BL2 is an EBV-negative Burkitt's Lymphoma cell line. LCL B95.8, was obtained after in vitro transformation of primary human blood lymphocytes (LCLs) with the B95.8 strain of EBV.⁴⁰ Both cell lines were cultured in RPMI 1640 supplemented with 10% FCS, L-glutamine (2 mMml^-1) and antibiotics (100 U penicillin ml^-1 and 100 µg streptomycin ml^-1). The EBV deletion strain B95.8, used in this study, lacks a 12 kb large portion of the genome.⁴¹

Stress induction.

BL2 cells were exponentially grown as described above and treated at a concentration of 1 × 10⁶ cells/ml with the following different stress conditions: For heat and cold shock treatment, cells were incubated for 30 min at the corresponding temperatures (45°C for heat shock and 15°C for cold shock) and subsequently incubated for 30 min at 37°C. For UV irritation, cells were pelleted by centrifugation, suspended in 1x PBS and irradiated with 120 mJ/cm² for 1 min at 254 nm (BLX-254, Bio-Link), afterwards resuspended in fresh medium and incubated for 30 min at 37°C. Osmotic stress treatment was performed at 37°C for 4 h after the addition of NaCl to the medium and adaptation of osmolarity to 500 mOsM. Oxidative stress was induced by the addition of 100 µM H₂O₂ to the medium and incubation for 4 h at 37°C. For inhibition of protein synthesis, 355 µM was added to the medium and incubated at 37°C for 4 h. Stress induction by serum deprivation was performed in medium devoid of FCS at 37°C for 48 h. After stress treatment, RNA was isolated and expression of selected Alu-derived RNA transcripts was analyzed by northern blotting. To apply optimal stress conditions, time- and dose-response experiments for a small selection of Alu-derived transcripts were determined.

Northern blot analysis.

Total RNA from EBV-immortalized and non-infected B cells was isolated employing the Tri Reagent method (Ambion, Austin, USA) according to the manufacturer's protocol. Subsequently, 2–30 µg of total RNA was size fractionated on an 8% denaturing polyacrylamide gel (PAGE; 7 M urea, 1x TBE buffer) and transferred onto a nylon membrane (Hybond N⁺, Amersham, GE Healthcare, Little Chalfont, UK) using the Bio-Rad semi-dry blotting apparatus (Trans-blot SDBio-Rad, Vienna, Austria). Immobilization of transferred RNAs on membranes was performed employing the STRATAGENE UV crosslinker (Stratagene, La Jolla, CA, USA). Oligonucleotides (Microsynth, Balgach, Swiss) complementary to potential novel RNA candidates were end-labeled with [γ-³²P]-ATP (Hartmann Analytic, Vienna, Austria) and T4 polynucleotide kinase (Promega, Mannheim, Germany). Depending on the T_m of the respective oligonucleotides, hybridization was carried out from 42°C to 52°C in hybridization buffer (178 mM Na₂HPO₄, 822 mM NaH₂PO₄, 7% SDS) over night. Blots were washed twice, i.e., at room temperature in 2x SSC buffer, 0.1% SDS for 5–7 min and subsequently in 0.1x SSC, 0.1% SDS for 5–7 min. Ethidium bromide-stained 5.8S rRNA or 5.8S rRNA hybridization signals were used as loading control for normalization after polyacrylamide gel electrophoresis. Northern blots were either exposed to a Kodak MS-1 film (Kodak, Bagnolet, France), using an intensifier screen or analyzed by a Molecular Dynamics Storm PhosphorImager (Image quant software version 5.0).

Deep sequencing analysis.

Analysis of cDNAs of the previously established EBV SHORT library¹⁴ by deep sequencing (GS-FLX system, Roche, GATC company, Konstanz, Germany) required amplification by two hybrid primers (hybrid A: 5′-GCC TCC CTC GCG CCA TCA GGT CAG CAA TCC CTA ACG AG-3′ and hybrid B: 5′-GCC TTG CCA GCC CGC TCA GAG GAG CCA TCG TAT GTC G-3′). For this purpose, subtracted cDNAs from the small fraction were directly amplified by hybrid primers. In contrast, subtracted cDNAs from the large fraction could not sufficiently be PCR amplified. Therefore, the PCR product of the large fraction was used as template for the PCR amplification with hybrid primers. Both, PCR products of the small and large fraction were purified by 8% native PAGE, followed by gel elution. The lyophilized and pooled cDNA library was subjected to deep sequencing. Raw data were deposited at the short read archive (SRA) with the accession number SRA010803.⁵.

Sequence analysis of cDNA library.

Computational sequence analysis of 37,682 sequence reads was done employing the LASERGENE sequence analysis program package (DNASTAR). Removal of the 5′-adaptor sequence and the 3′-C-tail resulted in a significant reduction from 37,682 to 25,045 sequences (66.5%), probably due to library amplification by hybrid primers. From 25,045 analyzed sequences, 78% exhibited read lengths shorter than 40 nt (Suppl. Fig. 1). Subsequently, cDNA sequences were clustered using the Lasergene Seqman II program package to identify identical sequences (DNASTAR),

The clustered cDNA sequences were subjected to the BLASTN alignment tool (NCBI, http://www.ncbi.nlm.nih.gov/BLAST), BLAT alignment tool (UCSC, http://genome.ucsc.edu/), snoRNA-LBME-database (http://www-snorna.biotoul.fr/) and Sanger miRNA database (http://www.mirbase.org/) to scan for known ncRNA genes. In total, 19,528 sequences (81.2%) of the computationally analyzed cDNA sequences were successfully mapped to genomic loci. Removal of ribosomal RNA, small nuclear RNA and transfer RNA sequence reads resulted in 7,074 sequences that were grouped into 592 unique ncRNA genes and classified in respect to known ncRNA classes. Sequences of new ncRNA candidates were additionally analyzed for repeat-like properties (http://www.girinst.org/censor/index.php).⁴² Secondary structures of novel ncRNA candidates were generated employing the program RNAfold from the Vienna RNA package.²³

Custom-designed ncRNA-microchip analysis.

A custom microchip for expression analysis of novel identified ncRNA candidates was generated in collaboration with the company Febit Biomed GmbH (Heidelberg, Germany). In addition to novel ncRNA candidates, probes for expression analysis of following ncRNAs were included: host-encoded rRNAs (5S, 5.8S), snRNAs, 7SL RNA, 7SK RNA, RNase MRP RNA, RNase P RNA, vault RNAs, Y RNAs, human snoRNAs from snoRNA-LBME-database, version 3,¹⁷ human miRNAs as well as EBV-encoded miRNAs (from miRBase; release 11.0).¹⁶ In total, 3,956 DNA oligonucleotide probes were designed using the Febit Geniom Client software (Table 1).^20,21

Preparation of size-fragmented RNAs is described below (see Results). Labeling was performed using the mirVana™ miRNA labeling kit from Ambion (Austin, USA). By this biotinylation method, a tail consisting of a mixture of unmodified and amino allyl-modified nucleotides, is added to the 3′-end of RNA molecules. In a second step, the amino allyl moieties react with the N-hydroxysuccinimide (NHS) ester group of a biotin-NHS ester and form a stable linkage between the biotin and the RNA molecule.

During a 16 hours hybridization step, a temperature gradient was employed and samples were moved within the array channels to enable faster diffusion. Upon hybridization of the biotinylated samples, a stringent washing step with 0.5x SSPE buffer (75 mM NaCl, 5 mM NaH₂PO₄, 5 mM EDTA, pH 7.4) at 45°C was performed. The hybridized biotin-labeled RNAs were stained with streptavidin-phycoerythrin (SAPE) in 6x SSPE buffer (900 mM NaCl, 60 mM NaH₂PO₄, 60 mM EDTA, pH 7.4) at 25°C and excessive SAPE was removed by five additional rounds of washing with 6x SSPE. Subsequently, the signal was enhanced by employing biotinylated anti-steptavidin antibodies and staining a second time with SAPE. All steps, including hybridization, washing, staining, signal enhancement and analysis of fluorescent signals, were carried out on the Febit Geniom technology system (Febit Biomed GmbH, Heidelberg, Germany). Microchip data analysis was performed as previously described.²¹

Novel ncRNA candidates, which exhibited a signal intensity above 1,000, i.e., three times over the background signal, in at least three of five biological replica, were selected and considered to be expressed. The fold change (FC) of probes of EBV-infected versus non-infected RNA preparations was calculated as a logarithmic value (log2).

Conclusion

Our study aimed at identifying differentially expressed novel regulatory ncRNAs in B cells infected by EBV. Thereby, deep sequencing analysis of a specialized cDNA library resulted in identification of 274 known ncRNAs and 318 novel ncRNA candidates, whose differential expression was subsequently investigated by custom microchip analysis. A technical challenge was to investigate differential expression of very heterogeneous classes of ncRNAs by a single microchip analysis. NcRNA classes, i.e., miRNAs, snoRNAs and novel ncRNA candidates, vary considerably in their structural features and sizes. Thus, improved hybridization of fluorescently labeled ncRNAs to complementary probes on the microchip was enabled by employing a temperature gradient during hybridization.

In this study, we identified 25 novel host-encoded ncRNAs to be differentially expressed in EBV-immortalized cells, comprised of six non-repeat-derived ncRNAs and 19 ncRNAs derived from repeat loci of the host genome. One novel non-repeat-derived ncRNA, c15308-A, is located in antisense orientation to the transcription factor ZNF787 mRNA and thus might be involved in post-transcriptional regulation of ZNF787.

In addition, most repeat-derived ncRNAs, whose expression was significantly upregulated upon EBV infection, originated from Alu repetitive elements. It has previously been described that different stress conditions, i.e., viral infections, heat shock, etc., induce transcription of Alu repeat-derived RNAs.^27,37 However, expression levels of Alu repeat-derived ncRNAs, identified in our study, were not found to be elevated after exposure of non-infected B cells to other stress stimuli.

By employing ncRNA-microchip analysis, differential expression of previously described EBV-encoded ncRNAs was verified, thus validating our approach. However, we could not confirm differential expression of five novel EBV-encoded ncRNA candidates from the deep sequencing analysis due to very low hybridization signals, indicating their low abundance. This might suggest that by now all abundant ncRNAs, mapping to the EBV genome, have been identified.

We also identified numerous host-encoded miRNAs whose expression was deregulated upon EBV infection. Deregulated miRNA expression levels were demonstrated to promote tumorigenesis.^19,38 In our study, expression of oncogenic mir-221 and mir-222 was significantly upregulated in EBV-infected cells. Since EBV infection is associated with tumorigenesis, increased expression levels of both oncogenic miRNAs might contribute to development of lymphatic cancers. It is tempting to speculate that mir-221 and mir-222 negatively regulate expression of tumor suppressor and cell cycle inhibitor p27(Kip1), as has previously been described for various types of cancer.^31,39

In conclusion, in this study we have established a novel approach to investigate differential expression of various classes of ncRNAs by a custom-designed ncRNA-microchip analysis. In particular, subsequently to deep sequencing our method might serve as a powerful tool to screen for differentially expressed tissue-, developmental-, pathogen- or disease-specific ncRNAs. Novel specifically expressed ncRNAs could be used as diagnostic markers or serve as potential drug targets in future therapeutical applications.

Abbreviations

ncRNA

non-coding RNA

EBV

Epstein-Barr virus

SINE

short interspersed nuclear element

LTR

long terminal repeat

miRNA

microRNA

EBERs

Epstein-Barr virus encoded RNAs

SHORT

subtractive hybridization of ncRNA transcripts

Accession numbers

Raw data were deposited at the Short Read Archive (SRA) with the Accession Number SRA010803.⁵. Data of custom microchip analysis were deposited at the Gene Expression Omnibus (GEO) with the Accession Number GSE20441.