Marek's Disease Virus Encodes MicroRNAs That Map to meq and the Latency-Associated Transcript

Abstract

MicroRNAs (miRNAs) are a class of small (∼22-nucleotide) regulatory molecules that block translation or induce degradation of target mRNAs. These have been identified in a wide range of organisms, including viruses. In particular, the oncogenic gammaherpesviruses Kaposi's sarcoma herpesvirus and Epstein-Barr virus encode miRNAs that could potentially regulate either viral or host genes. To determine if Marek's disease virus (MDV), an oncogenic alphaherpesvirus of chickens, encodes miRNAs, we isolated small RNAs from MDV-infected chicken embryo fibroblasts (CEF) and used the 454 Life Sciences sequencing technology to obtain the sequences of 13,679 candidate host and viral small RNAs. Eight miRNAs were found, five of which flank the meq oncogene and three that map to the latency-associated transcript (LAT) region of the genome. The meq gene is unique to pathogenic serotypes of MDV and is transcriptionally active during latency and transformation, and the LAT region of the MDV genome is antisense to the immediate-early gene ICP4. Secondary structure analysis predicted that the regions flanking the miRNAs could form hairpin precursors. Northern blot analysis confirmed expression of all miRNAs in MDV-infected CEF, MDV-induced tumors, and MDV lymphoblastoid cell lines. We propose that the MDV miRNAs function to enable MDV pathogenesis and contribute to MDV-induced transformation of chicken T cells.

Marek's disease (MD) is a lymphoproliferative disorder in which aggressive T-cell lymphomas result from infection of susceptible chickens with Marek's disease virus (MDV) (76). The complete nucleotide sequences of two strains of serotype 1 (oncogenic) MDV (Md5 [74] and GA [10, 38]), one serotype 2 (nononcogenic) strain (HPRS24 [29]), and one serotype 3 strain (herpesvirus of turkeys [1, 34]) have been published. These viruses all have a general genome structure that resembles the herpes simplex virus genome; that is, unique long (UL) and unique short (US) regions are flanked by terminal (T) and inverted (I), long (L), and short (S) repeat regions (TRL, TRS, IRL, and IRS). The US and UL regions of these herpesviruses are generally conserved and colinear with the corresponding regions of other alphaherpesviruses. However, the repeat regions of the genome differ among the alphaherpesviruses and the MDV serotypes.

MDV is generally considered to be a cell-associated herpesvirus (76). Productive infection occurs in epithelial cells and in B lymphocytes. Fully productive infection, which results in the release of infectious virus particles, occurs only in feather follicle epithelium. MDV assumes a latent posture in T lymphocytes, and in transformed cells, MDV is generally considered to be latent. The latency-associated transcripts (LATs) for MDV have been characterized and map antisense to the ICP4 gene (57). These consist of a long 10-kb LAT, one or more small MDV RNAs (MSRs), and a group of other spliced variants. No protein gene products that could be derived from translation of these transcripts have been reported.

MDV-induced lymphomas are complex, and the molecular details of MDV-induced transformation are not fully understood. Genes that are transcribed from the serotype 1-specific regions in the I/TRL are believed to play key roles, and transcription in transformed cells appears to be limited to these regions (68, 72). Known genes transcribed from these regions include the meq gene (32, 36, 44, 45, 47), pp38 gene (17, 19, 79), 1.8-kb gene family (8, 9), RLORF1 (62), L1/RLORF5a (59), RLORF8 (62), and the vIL-8 gene (46, 60, 63).

meq is the best studied among these genes, and the Meq protein is the strongest candidate oncoprotein described so far for MDV. The gene encodes a 339-amino-acid bZIP protein that resembles the Fos/Jun family of oncoproteins. In addition to expression in lymphoblastoid cells (47), the meq gene is expressed in lymphomas (66) and, to a lesser extent, in lytically infected cells (32). Meq expression is important for the maintenance of transformation of lymphoblastoid cells (77), and overexpression of Meq in combination with a complementing oncoprotein such as v-Ras results in transformation of Rat-2 cells (47). Recently, Meq has been shown to transform DF-1 cells, a spontaneously immortalized chicken embryo fibroblast (CEF) cell line that does not express transformed phenotypes in cell culture or cause tumors in animals (27, 41). Additional evidence that Meq is involved in MDV oncogenesis stems from the analysis of Meq mutants, which fail to cause tumors in chickens (51). A splice variant of Meq, termed Meq-sp, has been reported to be the major form of Meq expressed during lytic infection (60, 61).

MicroRNAs (miRNAs) represent a major class of small (∼22-nucleotide [nt]) regulatory noncoding RNAs that are encoded by the genomes of all multicellular organisms (37). miRNAs are derived from primary transcripts containing hairpin structures (primary miRNA) which are processed in the nucleus by the endonuclease Drosha to precursor miRNAs (pre-miRNA) (39). Pre-miRNAs are transported to the cytoplasm, where they are processed by another endonuclease, Dicer (39), to generate double-stranded 22-nt miRNA duplexes, one strand of which becomes incorporated into an RNA-induced silencing complex (20, 28, 58). miRNAs generally act as repressors of gene expression by either decreasing mRNA stability, blocking translation, or both (75, 78).

Several recent studies have reported the presence of miRNAs in viruses (73). Among the herpesviruses, 23 miRNAs have been validated experimentally for Epstein-Barr virus (EBV) (24, 65), 12 for Kaposi's sarcoma herpesvirus (KSHV) (11, 24, 64, 67), 9 for murine herpesvirus 68 (MHV68) (64), and 9 for human cytomegalovirus (HCMV) (23, 64). For EBV, KSHV, and MHV68, the miRNAs are clustered in one to two regions of the genome, whereas the HCMV miRNAs are scattered throughout the genome. For KSHV, 11 miRNAs are derived from a 4-kb region of the genome expressed during latency and transformation that encodes kaposin. Likewise, for EBV, one cluster lies in a latency-active region of the genome that encodes BHRF1 as well as EBNA 2 and EBNA-LP. Herpesvirus miRNAs have been predicted to target a variety of host genes, including genes encoding cytokines, chemokines, apoptotic genes, cell growth factors, and signaling factors, but none of these targets have been experimentally validated. Some viral genes may also be miRNA targets, such as the KSHV kaposins (73) and the EBV DNA polymerase (65).

Until recently, traditional cloning/sequencing technologies have been applied to the identification of small RNAs. These approaches tend to sample only highly expressed species and can generate only limited numbers of sequences. Recently, we applied massively parallel signature sequencing technology to identify small RNAs in Arabidopsis thaliana and found that this component of the genome is much more complex than previously thought (49, 54). In this study, we report the use of another parallel sequencing methodology, the 454 Life Sciences pyrosequencing technology, to identify MDV-encoded miRNAs. This approach provided sampling without the biases associated with cloning and resulted in the identification of novel miRNAs. Five unique MDV serotype 1-specific miRNAs were identified, mapped near the meq gene, and found to display an expression pattern that parallels meq gene expression. Three additional MDV-specific miRNAs were mapped to the LAT region of the genome. We propose that these MDV serotype 1-specific miRNAs play a role in MDV-induced transformation of T cells. In addition, host miRNAs revealed by the sequencing have been catalogued.

MATERIALS AND METHODS

Cloning and sequencing of chicken miRNAs.

Secondary CEFs, prepared by routine techniques, were infected with the RB1B strain of MDV for 24 h at a multiplicity of infection of ∼25,000 PFU/10⁶ cells. Protocols developed previously in one of our laboratories were used to construct the libraries (49). Briefly, RNA was isolated using TRIzol and size fractionated using polyethylene glycol (PEG) precipitation. The low-molecular-weight fraction was electrophoresed on a 15% polyacrylamide-8 M urea gel, and small RNAs (∼20 to 27 nt) were extracted from the gel and purified. Both 5′ and 3′ RNA adapters (Dharmacon Research, Boulder, CO) were sequentially ligated onto the RNA using T4 RNA ligase (Ambion, Austin, TX). The 5′ RNA adapter (5′-OH-GGUCUUAGUCGCAUCCUGUAGAUGGAUC-OH 3′) and 3′ RNA adapter (5′-P-CACUGAUGCUGACACCUGC-idT-3′; idT is inverted deoxythymidine) were designed to prevent self-ligation, and the ligation products were purified following each step. The RNA was then reverse transcribed (Superscript reverse transcriptase [RT]; Invitrogen) using a primer complementary to the 3′ adapter. cDNA inserts were amplified by PCR using primers corresponding to both adapters. Amplicons were sequenced by using the 454 Life Sciences system that utilizes pyrosequencing-based, sequence-by-synthesis, high-throughput, parallel sequencing (52). Sequence data were filtered for adapter sequences and clustered (allowing a 4-base overhang or mismatch at either end), and the insert sequence was analyzed by comparing it to the chicken and MDV genomes and to the miRNA database (55) using BLAST (3).

Northern blot analysis of miRNAs.

RNA from uninfected and MDV-infected CEFs and normal spleen and MDV-induced splenic tumor cells was fractionated using PEG (25) or through the use of the FlashPage system (Ambion). The low-molecular-weight fractions were electrophoresed on a 15% denaturing polyacrylamide gel, electroblotted to charged nylon, and hybridized to ³²P-labeled antisense primers that cover the entire length of the miRNAs. Hybridization to an antisense primer for two abundant chicken miRNAs, gga-miR-21 (5′-GTCAACATCAGTCTGATGAACTA) and gga-miR-221 (5′-AGCTACATTGTCTGCTGGGTTTC), was used as a loading control. Signals were detected by phosphorimaging. A 10-bp DNA ladder was used to approximate size.

RT-PCR.

Total RNA was treated with DNase (1 U/μg RNA) and reverse transcribed with a gene-specific primer, using Invitrogen's two-step PCR kit according to the manufacturer's directions.

Cell lines and tumors.

MSBI, CU91 (70), and UA30 cells (provided by Mark Parcells) were grown at 37°C, 5% CO₂ in RPMI 1640 supplemented with 10% fetal bovine serum. Tumors were isolated from spleens of 8-week-old specific-pathogen-free birds inoculated with the RB1B strain of MDV (∼1,000 PFU) at 2 weeks of age. Normal spleens were collected from 2-week-old specific-pathogen-free birds.

RESULTS

MDV grows as a cell-associated virus, and initial infections take place in a limited number of cells. In order to increase the likelihood of identifying MDV-encoded miRNAs, we collected RNA from heavily infected CEFs (25,000 PFU/10⁶ cells). Following size selection, adapter ligation, and amplification, the resulting amplicons were sequenced in parallel using 454 Life Sciences pyrosequencing technology. A total of 13,679 high-quality reads containing both the 5′ and 3′ adapters used in cloning (see Materials and Methods) were obtained (see Table S1 in the supplemental material). The median insert length was 23 nucleotides (range, 15 to 33), and these formed about 1,727 distinct clusters of nonredundant sequences. After comparison to the miRNA database (55) (Table 1), we found 9,925 matches to known miRNAs, representing a total of 101 distinct species. Among the most abundant were miR-21, miR-221, and miR-222 (24, 8, and 13%, respectively). Of the 3,197 reads that did not show a match to known miRNAs, 360 were sequenced more than once, and many of these are candidates for novel miRNAs.

TABLE 1.

Distribution of small RNAs from CEFs infected with MDV

Type	No. (%)
MDV microRNA	141 (1)
Known microRNAa	9,925 (73)
tRNA, rRNA, and mtRNAb	416 (3)
No matchc	3,197 (23)
Total	13,679 (100)

^aIncludes 101 unique clusters.

^bmtRNA, mitochondrial RNA.

^cNo match to MDV or small noncoding RNAs.

One hundred forty-one reads showed 100% identity to the MDV genome and clustered into 10 distinct candidate miRNAs that were sequenced more than once. Seven of these flank the meq gene, while three species map to the 5′ end of the 10-kb LAT RNA that is antisense to ICP4 (16). None of these sequences matched the chicken genome.

The MDV miRNAs are listed in Table 2 and are designated MDV-miR-1 to -8. MDV-miR-1, MDV-miR-2, and MDV-miR-4 were the most frequently sequenced (32, 28, and 40 reads, respectively), while MDV-miR-3 and MDV-miR-5 were less abundant (4 and 12 reads, respectively). MDV-miR-6, -7, and -8 were also found in low abundance (10, 2, and 3 reads, respectively). The sequences flanking the miRNAs can form hairpin structures (Fig. 1) (80), one of the criteria of authentic miRNAs (4). This analysis also revealed that two of the sequences are complementary to MDV-miR-2 and -4 and form the stems of the hairpins. The low frequency of cloning and the local thermodynamic stability of the duplexes indicate that these are likely the miRNA* strands that are not incorporated into an RNA-induced silencing complex. Therefore, these are designated MDV-miR-2* and MDV-miR-4*, in accordance with established nomenclature (43). As is typical, the miRNAs within both duplexes start with U or G and have less stably paired 5′ ends compared to those of the miRNA*s (35, 71).

TABLE 2.

Sequence and location of cloned MDV miRNAs and miRNA*s

Name, sequence (5′-3′)	Length (nt)	No. of readsa	MDV/IRL positionb
MDV-miR-1, UGCUUGUUCACUGUGCGGCA	20	32	136873
MDV-miR-2, GUUGUAUUCUGCCCGGUAGUCCG	23	28	134232
MDV-miR-2*, CGGACUGCCGCAGAAUAGCUU	21	5	134270
MDV-miR-3, AUGAAAAUGUGAAACCUCUCCCGC	24	4	134080
MDV-miR-4, UUAAUGCUGUAUCGGAACCCUUC	23	40	134368
MDV-miR-4*, AAUGGUUCUGACAGCAUGACC	21	3	134405
MDV-miR-5, UGUGUAUCGUGGUCGUCUACUGU	23	12	133647
MDV-miR-6, GAGAUCCCUGCGAAAUGACAGU	22	10	176052
MDV-miR-7, UCGAGAUCUCUACGAGAUUACAG	23	2	175874
MDV-miR-8, GUGACCUCUACGGAACAAUAGU	22	3	176164

^aOut of a total of 13,679 reads.

^bIdentical sequences are located in the TRL and TRS. Numbering is based on AF243438.

FIG. 1.

Secondary structure predictions of MDV pre-miRNAs. Folding and minimum free energy calculations were performed with mfold (80) without constraints. The mature cloned microRNAs are shown in bold, and the * strands are in italics (MDV-miR*).

The sequences of these miRNAs in the context of the MDV genome are shown in Fig. 2. MDV-miR-1 lies downstream of meq and is embedded within the open reading frame (ORF) of the L1/RLORF5a transcript (59, 69) and within the intron of Meq-sp, a splice variant of Meq (61). MDV-miR-2 to -5 are immediately upstream of meq and are antisense to RLORF8. MDV-miR-6 to -8 are located between the a-like sequences and the ICP4 gene and within the large intron of the MDV latency-associated MSR (16).

FIG. 2.

Genomic location of MDV miRNAs. (A) Schematic and sequence of the MDV miRNAs flanking meq. (B) Schematic and sequence of the MDV miRNAs downstream of ICP4. Nucleotide and MDV gene numbering are according to Md5 (AF243438). In the schematics, small arrowheads indicate the locations of MDV miRNAs, and in the sequence, the individual miRNAs are shown in red and by an overline. A bent arrow depicts the transcriptional start site of the meq gene, and the TATAA and ATG for meq are boxed. Large orange arrows correspond to MDV transcripts, and the open reading frames are the hatched areas. The green arrow indicates the a-like sequence. Spliced introns are shown as solid lines linking exons. Additional spliced variants of the MSRs are described in reference 16.

Northern analysis confirmed that the MDV miRNAs are indeed expressed in MDV-infected cells and MDV-induced tumors, but not in uninfected cells or normal tissue. MDV-miRNA expression in CEFs infected with MDV (4,000 PFU/10⁶ cells) was low but could be readily detected in older cultures (Fig. 3). As a loading control, blots were stripped and probed for gga-miR-21 to show the presence of miRNA in all samples. Figure 4 shows hybridization of oligonucleotide probes antisense to MDV-miRs to normal spleen cells and MDV-induced splenic tumors. Relative to miR-21, MDV-mir-1 to -5 and MDV-miR-8 miRNAs were generally expressed at higher levels in the tumors compared to expression in MDV-infected CEFs (Fig. 3). MDV-miR-6 was expressed at lower levels in MDV-induced tumors, and MDV-miR-7 was not detected in tumors. MDV-miR-2* and -4* can also be detected by Northern blotting, albeit at lower levels (not shown). MDV miRNAs are also expressed in MSB1 cells, an MDV-transformed lymphoblastoid cell line (2), but not in CU91 cells, a reticuloendotheliosis virus-transformed avian T-cell line (70). UA30 cells, which are CU91 cells infected with the RB1B strain of MDV, also express low levels of these miRNAs (Fig. 5).

FIG. 3.

Northern blot of FlashPage-fractionated RNA (1 μg/lane) from uninfected CEFs (-) or CEFs infected with MDV (+) for 1, 3, or 5 days. RNA was electrophoresed on a 15% denaturing polyacrylamide gel, blotted, and hybridized to ³²P-labeled oligos antisense to MDV miRNAs. Blots were stripped and hybridized to the miR-21 antisense probe to show the presence of microRNA in all lanes (a representative blot of this set is shown).

FIG. 4.

Northern blot analysis of FlashPage-fractionated RNA (1 μg/lane) from normal chicken spleen cells and MDV-induced tumors. Blots were prepared as described in the legend to Fig. 3.

FIG. 5.

Northern blot analysis of PEG-purified RNA (1 μg/lane) from MSB1, CU91, and UA30 cells. Blots were prepared as described in the legend to Fig. 3, except that blots were hybridized to the miR-221 antisense probe as a loading control (only a representative blot is shown). Positions of DNA ladder markers are shown on the left.

A single amplicon containing MDV-miR-2 to -5 was detected using RT-PCR (not shown). However, even though MDV-miR-2 to -5 could be derived from a common primary transcript, the differences in the abundance (number of reads) suggest that these miRNAs differ with regard to processing or stability. In MSB1 cells, the precursor miRNAs for MDV-miR-6 to -8 were either equal to or more abundant than the mature miRNA, indicating less efficient processing of these miRNAs.

DISCUSSION

We have identified eight miRNAs encoded by the MDV genome. The MDV miRNAs conform to the criteria set by Ambros et al. (4) for valid miRNAs. These have been identified in a cDNA library and detected by Northern blot analysis. In addition, the sequences flanking the miRNAs form hairpin structures with at least 16 nucleotides participating in Watson-Crick or G/U base pairing. There is no match of the MDV-miRNAs to any entry in the miRNA database and, thus, these represent novel miRNAs.

Five miRNAs flank the meq oncogene of the MDV genome. While most of the MDV genome is transcriptionally silent during latency and transformation, these miRNAs map to an unusual region of the genome in the I/TRL that is transcribed in tumors and transformed cells (68, 72). Many discrete transcripts derived from this locus and ranging in size from 700 to 4,000 nt have been reported using conventional Northern blots hybridized with strand-specific or double-stranded probes, and transcripts encoded by both strands have been detected in MDV-induced tumors, lymphoblastoid cell lines, and in cells lytically infected with serotype 1 MDV (32, 59, 61, 63, 72). In addition to revealing discrete species, Northern blots probed for RNAs mapping near the meq locus typically exhibit a smear of hybridization and include relatively low-molecular-weight species, suggesting that transcription from this region of the genome is complex and that many more transcripts in excess of those that have been well-characterized may exist.

MDV-miR-1 maps downstream of meq and is embedded in the ORF of the 600-nt L1/RLORF5a transcript, which potentially encodes a 107-amino-acid protein showing no significant homology with other known proteins (59). MDV-miR-1 is also present in the intron of Meq-sp, an alternatively spliced product of the meq and vIL8 genes (61). L1/RLORF5a is expressed at higher levels in CU41 cells (reticuloendotheliosis virus-transformed lymphoblastoid cells that are latently infected with MDV) than in lytically infected CEFs (59), but Meq-sp is expressed at higher levels in lytically infected cells (60). Both or either could serve as the primary transcript under different conditions. Since the profile of L1/RLORF5a expression is similar to MDV-miR-1, it is possible that the L1/RLORF5a transcript could be the primary MDV-miR-1 transcript. It should be noted that a deletion of L1/RLORF5a does not detectably affect virus replication, establishment of latency, or oncogenesis (30, 69); therefore, it is unlikely that MDV-miR-1 is essential for these functions.

MDV-miR-2 to -5 are all located upstream of the meq promoter and appear to be expressed as a single transcriptional unit in the same orientation as the meq gene. Processing of miRNAs from polycistronic primary transcripts has been reported previously (5, 40, 67), and it has been suggested that selection pressure acts to group miRNAs into coregulated clusters (6). This cluster of miRNAs is antisense to another MDV transcript, RLORF8, which encodes a potential ORF of 135 amino acids that has no homology with other known sequences. The RLORF8 mRNA is expressed in lymphoblastoid cells and in infected CEFs (62). It is possible that these miRNAs could regulate expression of an RLORF8 protein; however, expression of a protein from this locus has not been reported. To our knowledge, there is only one other example of a herpesvirus miRNA that is antisense to a known viral transcript. In EBV, miR-BART2 maps to the intron of BART2 and is antisense to the coding region of BALF5, the EBV DNA polymerase (65). There is some evidence that EBV-miR-BART2 targets the DNA polymerase for degradation, since the 5.0-kb BALF5 mRNA has a 3.7-kb short form with a 3′ terminus that maps to the predicted miR-BART2 cleavage site.

MDV-miR-6 to -8 map to a large intron in the 5′ end of the latency-associated MSR transcript, which is presumably derived from a large 10-kb transcript that maps antisense to the ICP4 gene (16). Transcripts from this region are abundantly expressed during latency, in lymphoblastoid cell lines, and in lymphomas (15, 16). The miRNAs mapping to this region were detected in MSB1 cells, tumors, and lytically infected CEFs. The 5′ end of the 10-kb LAT and the MSR have been mapped (16), and the results are consistent with the 10-kb LAT being the primary transcript of MDV-miR-6, -7, and -8. It will be interesting to see if MDV-miR-6, -7, and -8 play a role in the balance of lytic replication and latency for MDV.

It is possible that some or all of the MDV-miRNAs play a role in MDV oncogenesis. MDV miRNAs appear to be expressed under similar conditions as the meq oncogene and the LATs. All are expressed in tumors and lymphoblastoid cells. Both the MDV-miRNAs, meq (32) and LATs (16) are expressed at low levels in infected CEFs. The facts that MDV miRNAs map adjacent to the meq locus and their expression patterns generally parallel that of meq suggest that they contribute to T-cell transformation. These miRNAs may target host genes that play a role in MDV-induced transformation. It is tempting to speculate that some of the host genes that must be regulated to allow viral infection are also genes that are involved in transformation pathways.

The role of miRNAs in tumor biology in general is just beginning to be understood, and the emerging story is complex. miRNAs can regulate differentiation, proliferation, and apoptosis, all of which are important processes in neoplastic transformation (56). Many miRNAs map to fragile sites or cancer-associated regions of chromosomes (14). In some contexts, miRNAs resemble tumor suppressors, and in other situations, they appear to have oncogenic potential. Both up- (18, 22, 26, 31, 53) and downregulation (12, 50, 73) of miRNAs have been found in neoplasms. In addition, miRNA expression patterns among tumors reflect the developmental history and lineage of the neoplasms and are likely to have utility in diagnostics (13, 50).

miRNAs appear to recognize their targets through base pairing, generally in the 3′ untranslated region of mRNAs. In plants, miRNAs show a high level of complementarity to their targets, making target prediction a relatively straightforward computational task, and multiple target sites in multiple genes for individual miRNAs have been predicted computationally (33) and detected experimentally (48). However, in animals, base pairing is incomplete, making target prediction more challenging. It appears that a 6- to 7-base seed signature at the 5′ end of the miRNA defines the interaction with targets (42), while the role of the 3′ end of the miRNAs is not yet understood. There are numerous algorithms for target prediction, but only a few targets have been actually validated (reviewed in reference 7). Since many miRNAs show strong conservation across species, computational approaches to target identification include cross-species conservation in the 3′ untranslated regions of homologous genes. However, there is no apparent conservation of miRNAs among the herpesviruses, and each herpesvirus group appears to have evolved miRNAs that are unique to its particular biological niche (64). For HCMV, conservation of the miRNAs among various virus isolates appears to be greater than conservation of viral coding sequences (21), suggesting that viral miRNAs tend to repress expression of conserved or slowly evolving host genes. Identification and validation of the targets for MDV-encoded miRNAs will be an important future challenge in MDV biology.

The miRNAs identified in these studies were identified in lytically infected CEFs by using the 454 Life Sciences parallel sequencing technology. This high-throughput approach has distinct advantages over traditional cloning techniques. Aside from the very large sampling number that allows identification of low-copy-number species, it requires relatively little manipulation of samples, is readily available, and is particularly well suited to identification of small RNAs since the entire sequence is obtained in each read. Even though we did not sequence to maximal depths in this pilot study, we greatly exceeded that of most traditional sequencing efforts to identify viral miRNAs and were able to find low-abundance species. Most other viral miRNAs have been identified in latently infected cells, and many more MDV miRNAs may be discovered when latently infected cells and other cell types are examined using this technology.