MicroRNAs are small (about 22 nt) RNAs that play important regulatory roles by targeting mRNAs for degradation or translational repression. MicroRNAs were first identified in Caenorhabditis elegans 1 but high evolutionary conservation eventually led to the identification of microRNAs in other species. This, coupled with conventional sequencing of small RNA libraries, has greatly expanded the list of known microRNAs. The most recent release of the microRNA database, miRBase 10.0 2, contains over 5000 microRNA gene loci in a wide variety of animal, plant and viral genomes.

Conventional sequencing favors identification of highly expressed species, and comparative genomics will not identify nonconserved microRNAs. In order to enhance discovery of small RNA species, massively parallel signature sequencing (MPSS) was used to identify small RNAs in Arabidopsis thaliana 3, and the results showed that the diversity of small RNAs exceeded previous estimates. More recently, newer deep sequencing technologies have been used to profile microRNAs in Arabidopsis DICER and RDR2 mutants 45, and others have applied this technology to various samples including human and chimpanzee brain 6 and Chlamydomonas reinhardtii 7. These approaches have the advantage that they not only provide sequence of low abundance species, but also provide quantitative data since the frequency of sequencing reads reflects the abundance of microRNAs in the population.

We previously reported on the use of deep sequencing technologies for identification of microRNAs encoded by Marek's disease virus (MDV), an economically important pathogenic herpesvirus of chickens 89. In an extension of the pilot study, we sequenced additional reads from both MDV-infected chicken embryo fibroblasts (CEF) and uninfected CEF and now report on the identification of potential novel host microRNAs. In addition, the sequence of several new MDV-encoded microRNAs were discovered by deeper sequencing.

Our deep sequencing approach to microRNA discovery in the chicken confirms the expression of 112 known microRNAs and identifies a pool of 63 candidate novel microRNAs. The majority of the known chicken microRNAs have been identified by sequence comparison with microRNAs from other species 13, and the expression of many has been confirmed by analysis of EST data and in situ hybridization 1415. Cloning from small RNA libraries also validates the expression of microRNAs. A recent study of chicken microRNA cloning 16 used conventional technology to confirm expression of 25 of the known chicken microRNAs and identified one possible novel microRNA. Our study adds to the confirmation of expression of predicted microRNAs, and greatly expands the list of potential novel microRNAs in the chicken.

A large majority (86%) of the reads from chicken CEF small RNA libraries matched known microRNAs. Similar numbers were obtained in deep sequencing of human brain 6, where 80% of the reads matched known human microRNAs. Thus, it could be argued that we are approaching saturation of microRNA discovery. However, it is possible that only highly and ubiquitously expressed microRNAs have been found, and less abundant or tissue-specific microRNAs may still be revealed by deep sequencing of different tissues. This is clearly the case in plants, where a large set of small RNAs have been discovered by deep sequencing 3. In an analysis of the human 'colorectal microRNAome', SAGE was used as a deep sequencing approach, and matches to 200 microRNAs in miRbase were found, as well as 100 previously unrecognized microRNA* strands and 133 candidate novels 17. MPSS analysis of mouse embryo small RNA discovered over 60 potential novel microRNAs, some of which are rodent specific 11. We have identified a pool of 63 candidate novel chicken microRNAs and have confirmed expression of several candidates by northern blot analysis. Some are expressed in a tissue-specific manner, while others are more ubiquitously expressed. Further analysis of these and other candidates with respect to temporal and tissue expression will be a first step to understanding function. Overall, the deep sequencing approach to microRNA discovery suggests that a significant number of novel microRNAs remain to be discovered and characterized.

In addition to identifying novel chicken microRNAs, deep sequencing of MDV infected CEF has revealed previously uncharacterized MDV-encoded microRNAs, bringing the total of MDV microRNAs to 18. Other herpesviruses encode microRNAs, and these are thought to play a role in immune evasion, apoptosis and cell cycle control 1819. MDV causes a well- characterized, virally induced T cell lymphoma of chickens and represents an excellent model system for analyzing the function of viral microRNAs in the pathogenesis of cancer. Many recent studies implicate microRNAs as either tumor suppressors or oncogenes 20, and host encoded microRNAs can act in cis (on viral target genes) or trans (on host encoded genes) to affect phenotypic changes 21. Moreover, virally infected cells are stressed, and it has been proposed that microRNAs play an important role in the stress response 22. A comparison of the reads of MDV-infected CEF versus uninfected CEF indicates that the majority of the microRNAs are expressed at similar levels. However, CEF are used to propagate virus, and viral infection occurs in a very small percentage of cells, thus making it difficult to observe changes when analyzing whole cultures. In addition, CEF are not the in vivo target of the virus, and we might expect a more critical layer of regulation in T cells. A small set of microRNAs appeared to be differentially expressed in infected vs. uninfected cells, but this was not confirmed by northern blot analysis. This lack of concordance between the two techniques is not uncommon when expression levels are very low, or when cross-hybridization with similar species can confound the results. Additionally, in our system, MDV infects only a small percentage of the cells and infections vary considerably from culture to culture. This biological noise could also hamper the ability to reproduce differences noted in the sequencing data set.

The two most frequently sequenced microRNAs were gga-miR-222 and -221, which also share sequence identity in the seed region. These are located within a 3000 nt region of Chromosome 1, where there are two copies of gga-miR-222 followed by one copy of gga-miR-221. In human, miR-221 and -222 are coordinately expressed from a single primary transcript 23. We see about 1.7-fold higher abundance of ggg-miR-222 compared to gga-miR-221, consistent with their sharing a transcript that is highly expressed in CEF. Computational analysis has predicted that p27Kip1 protein, a key inhibitory regulator of the cell cycle 24, is a potential target for this cluster. Down regulation of p27Kip1 by miR-221/222 promotes growth and proliferation of cancer cells, and could play a similar role in dividing CEF 24. miR-125b and 21 were also abundant in our libraries. miR-125b is critical for the proliferation of some human cell lines 25, and mir-21 is thought to function as an oncogene by decreasing apoptosis 26. The high levels in rapidly dividing CEF could play a permissive role in the cell cycle in CEF.

Our northern analysis of MDV-induced tumors shows several host microRNAs that were noticeably less abundant in MDV-induced tumor tissue compared to normal spleen, consistent with previous reports of a general down-regulation of microRNAs in tumors 27. Among those down regulated, let7 was particularly interesting. The let7 microRNA is known to down-regulate Hmga2 28, which is a small, non-histone, chromatin-associated protein that is believed to influence chromatin remodeling 29. Hmga2 is expressed robustly in undifferentiated proliferating cells, and its expression during embryogenesis and in a variety of benign and malignant tumors has been characterized 30. Down-regulation of let7 should lead to increased expression of Hmga-2, and such a scenario would be consistent with the cell proliferation that characterizes tumors. miR-16 is considered a tumor suppressor 31, which acts by targeting BCL-2, and repressed expression is consistent with tumorigenesis. MiR-181s were down-regulated in gliobastoma compared to normal brain controls 32, and miR-199a was down-regulated in hepatocelluar carcinoma 33. Thus, it is likely that in MDV-induced tumors, as in other tumors, many microRNAs act collectively to facilitate cellular transformation and proliferation. More information on the perturbations of host microRNAs will come from a deep sequencing analysis of microRNAs in tumors.

Understanding the biological function of microRNAs first requires identification of all microRNAs within a genome. Here we have described the application of deep sequencing technology for the identification of many candidate novel chicken microRNAs from a single tissue source. The application of this technology to other tissues will no doubt lead to the identification of other novel microRNAs, which will improve the annotation of the chicken genome and further our understanding of this important class of regulatory molecules.

JB and RM designed and coordinated the study and wrote the manuscript; MO performed the analysis of the sequencing data; AA, EB and GI prepared samples for sequencing and for northern blots; CL prepared the libraries; PG and BM assisted in design, library construction and sequence analysis.