Moreover, there is increasing indication for the existence of lineage- or species-specific transcripts from numerous studies (e.g., [11]). Recent studies [12-15] provided conclusive evidence for the abundance of lineage-, tissue-, development-, or spatial-temporal-specific transcripts encoded in the cattle genome. For instance, Kumar and colleagues [12] discovered a number of novel transcripts predominantly expressed in the cattle placenta, which not had been found previously in primates or rodents. These results indicate towards the molecular basis of the relative anatomical variation in placental architecture of ruminants and suggest that loss and gain of genes and transcript variants are important mechanisms of genome evolution in mammals. Especially in ruminants, there is already unequivocal evidence for the occurrence of highly divergent genes encoding proteins having adaptive functions in the biology and metabolism of reproduction (e.g., IFN tau, [16]).

Given the unique adaptations for the ruminant reproductive system, it is conceivable that also other highly divergent processes associated with rumination and dramatic adaptive changes in ruminant metabolism, for instance, at the onset of lactation and continuous maintenance of a high milk production performance [17-21] could be caused by the action of individual species-specific genes or transcripts. This would provide important rationale for the identification of genes or transcripts being located in QTL intervals associated with milk production traits.

To complete the bovine gene catalogue on the targeted BTA6 region carrying the QTL, alternative methods will be required to identify novel transcripts and genes independently of their individual tissue-specific or spatial-temporal expression pattern. In our study, we focused on targeted isolation of transcribed sequences from defined subchromosomal regions on BTA6 poorly covered with gene information. Therefore, we exploited bovine bacterial artificial chromosome (BAC) clones mapped to the QTL interval using exon trapping, a technique that has been developed to identify genes in cloned eukaryotic DNA (e.g., [22-24]). This technique enables the experimental identification of potential exons in a fragment of eukaryotic DNA of unknown gene content or intron-exon structure based on the presence of functional splice sites; thus making the exon trapping assay independent of the expression pattern of the transcripts to be identified. This feature represents the main advantage of the exon trapping method compared to other methods of gene identification (cDNA selection [25-27] or cDNA hybridization approaches [28-30]).

Sequence comparison identified 92 sequences as being derived from the trapping vector caused by vector self-splicing and from the BAC vector sequence or due to cloning artefacts, which indicated 23% false positive splicing events. Of the remaining 304 putative ETS, 92 (30%) were unique sequences. The number of unique sequences detected per BAC clone varied between three and fifteen. There was a relatively high percentage (70%) of ETS observed redundantly. Some sequences were found with a frequency up to fourteen-fold. This redundancy is due to the fact that for each BAC clone two separate exon trapping libraries based on two different restriction enzyme digestion of BAC DNA were constructed. However, the set of ETS detected in the two single libraries constructed per BAC clone was not completely identical. Furthermore, it has also to be considered that the exon trapping technique itself includes several steps based on PCR amplification, a feature that intrinsically should be associated with the abundance of multiple copies of specific DNA targets.

The size of unique putative ETS ranged from 35 to 317 bp with a medium of about 125 bp. Locus-specific primers were derived for unique putative ETS with a size exceeding 60 bp (see Additional file 1). These primers were applied to remapping of ETS to their parent BAC clones by PCR. The chromosomal localization of exons on BTA6 could be inferred from physical mapping of the BACs carrying the trapped sequences by FISH or by in silico sequence similarity mapping of the corresponding BAC end sequences on the genome sequence assembly of BTA6 in combination with our high-resolution RH map [9].

All 92 unique ETS were further analyzed by screening the bovine genome sequence assembly (NCBI, Btau4.0) for similarity to known sequences, genes or transcripts. The results are summarized in Figure ​Figure2A2A and presented in detail in Additional file 2. Sequence identity to sequence scaffolds on BTA6 was detected for the majority of unique ETS (91%). Two ETS were identified on sequence scaffolds not yet assigned. A further two sequences contained repetitive sequence motifs and matched to more than one scaffold located on different chromosomes. We found that five ETS (5%) could not be assigned to the bovine genome sequence resources by in silico sequence similarity search.

As illustrated in Figure ​Figure2B2B and presented in Additional file 2 in more detail, for a number of ETS (28%) sequence similarity was found to genomic sequence scaffolds from alternative sequence assembly of the human genome (based on Celera assembly, for instance ETS_25316, ETS_25598 etc.). Furthermore, a few ETS (11%) revealed similarity to genomic sequences located on the syntenic chromosomes HSA4 and MMU5 of the human and mouse reference genome assemblies. The remaining ETS did not detect comparative genomic counterparts in human and mouse.

According to Additional file 2 and Figure ​Figure2C,2C, we found that only a small number of ETS showed identity to known bovine cDNA sequences or expressed sequence tagged sites (ESTs). On BAC clone BBI_750F0243 carrying ETS_20835, ETS_20849 and ETS_20866, we identified four exons of the bovine KCNIP4 gene (NM_001076935), which has an orthologous counterpart on the syntenic HSA4 region. The ETS_20866 sequence was not yet identified on scaffolds of the current bovine sequence assembly, although the mRNA of the bovine KCNIP4 gene revealed 100% identity to this sequence, which is also supported by sequence identity of ETS_20866 to several bovine ESTs. Strong sequence similarity of ETS_20866 was also identified to the adjacent exons 11 and 12 of the orthologous human gene. This result indicates that the bovine KCNIP4 gene is not completely annotated in the current bovine sequence assembly because at least two exons are missed in the predicted gene model. Besides to human, high comparative sequence similarity of these ETS was also observed to exons of the KCNIP4 gene from other species (rat, mouse, chicken and horse).

Sequence identity to known bovine ESTs was detected for ETS_25609 on BAC BBI_750F06108, ETS_15246 on BAC BBI_750B0172, and ETS_25294 on BAC BBI_750J07162. The ESTs were all assigned to sequence scaffolds on BTA6 (NW_001495159, NW_001495179, and NW_001495167, respectively), where the corresponding parental BAC clones were assigned to. In addition, ten ETS captured from BAC clone BBI_75006108 revealed sequence similarity to two ovine mRNAs and one ETS from BAC clone BBI_750B0172 showed similarity to a porcine transcript.

To analyze their potential protein-coding capacity, the unique ETS were screened for existing open reading frames (ORFs) using the program ORF Finder at NCBI. For most ETS, the program could predict ORFs. However, no ORF spanned the sequence length of the analyzed ETS completely. Protein coding sequences predicted by the program were subjected to sequence alignment with TBLASTN in the nucleotide collection and the ESTs databases at NCBI. Again, only the sequences ETS_20835, ETS_20849 and ETS_20866 showed high similarity matches pointing to the KCNIP4 gene from several species.

Including the ab initio map at the NCBI Bos taurus MapViewer [31] into our analysis we identified fifteen ETS with sequence similarity to bovine gene models predicted by the gnomon gene prediction program. One of them (ETS_16667) pointed to a hypothetical locus (LOC100140872) predicted to be a pseudogene similar to the PHYHIPL gene (see Additional file 2, Figure ​Figure2C2C).

The results obtained by the different sequence similarity analyses suggested that the majority of the isolated transcripts in our study presumably represent bovine noncoding transcripts located in intergenic regions. By screening the comprehensive mammalian noncoding RNA database (RNAdb 2.0) we could not identify similarity to known noncoding RNAs in the available data sets [32].

To elucidate, if the identified ETS were really expressed in the bovine transcriptome, multi-tissue expression analysis by RT-PCR using locus-specific primers on total RNA from nine different bovine reference tissues (liver, small intestine, kidney, thyroid gland, mammary gland, pituitary gland, skeletal muscle, subcutaneous and intestinal fat) was performed. For each BAC clone we exemplarily analyzed at least three ETS, if available and specific primers could be designed.

The results are summarized in Figure ​Figure2D2D and are displayed in detail in Figure ​Figure3.3. They show that 43 ETS out of 51 analyzed (84%) were expressed at least in one tissue isolated from a lactating cow. Specifically, 39 ETS displayed a different expression pattern in the tissues analyzed, and four revealed expression in all tissues examined.

In Figure 4(A–D), examples of different expression patterns of ETS ranging from expression in only a few tissues (A, C, D) to ubiquitous expression (B) observed in the experiment are shown. Figure ​Figure4E4E displays expression of the house-keeping gene GAPDH in the tissues analyzed in our study. Generally, ETS revealed amplification of the expected amplicon size in bovine genomic DNA (see Figure 4A–C, lane BT). However, in the case of ETS_20866 amplification in bovine genomic DNA failed (Figure ​(Figure4D,4D, lane BT), which is due to the fact that this ETS consists of two adjacent exons. This was supported by sequence comparison to the human KCNIP4 gene sequence revealing that the primers of ETS_20866 span an intronic region on genome level. There were a further 6 unique ETS captured by exon trapping (ETS_21696, ETS_25582, ETS_16037, ETS_15328, ETS_25303, ETS_26263), which also seem to contain at least two adjacent transcripts and are interrupted by intergenic DNA sequence on genomic sequence level. This result is supported by both alignment of the identified sequences to sequence scaffolds and experimentally, by amplification of the corresponding longer amplicons on bovine genomic DNA.

This study represents a further step in the ongoing molecular and genetic analysis of complex traits and annotation of genes and transcripts localized in the region on BTA6 containing QTL for milk and meat production, health and conformation traits.

As an initial step towards systematic analysis of transcripts and genes in this region, we carried out exon trapping using selected bovine BAC clones previously mapped to QTL intervals on BTA6. Mining a genomic interval comprising about 1 Mb for transcribed sequences using this technique, we identified a total of 92 unique exon trapping sequences. Genome similarity searches revealed sequence identity matches to sequence scaffolds on BTA6 for most unique ETS (91%). With two ETS, which were identified on sequence scaffolds not yet assigned to the current NCBI sequence assembly of the bovine genome, gaps could be closed on the genome sequence level. There were about 2% of ETS, which could not be unambiguously assigned to the bovine genome sequence, because of matches to multiple chromosomes due to repetitive sequence motifs. We found only 5% of ETS without any hits to known sequences contained in the archive of bovine sequence databases. These ETS can provide new additional sequence information to complete the current genome sequence assembly. This result further indicates that targeted deep sequencing within the corresponding genomic regions would be required to improve the accuracy of the BTA6 sequence assembly.

Comparative sequence similarity search to human and mouse genome sequences revealed that 11% of the isolated ETS displayed high similarity to genomic sequences located on the syntenic chromosomes HSA4 and MMU5 of the human and mouse reference genome assemblies pointing to highly conserved genome regions in these species. Almost a third of the ETS identified similar equivalent sequences in genomic sequence scaffolds from the alternative Celera-based sequence assembly of the human genome. The residual 62% of ETS without comparative genomic sequence counterparts in human and mouse refer to presumably species-specific genomic regions in the bovine genome.

Screening the gene, ESTs and protein databases at NCBI detected only a few known transcribed sequences revealing identity to the ETS isolated by exon trapping in our study (17%, Figure ​Figure2C2C).

Whereas 6% of all ETS identified known bovine transcripts, a further 16% of ETS pinpointed to bovine gene models predicted ab initio.

For the evaluation of the relatively low proportion of ETS identifying known transcripts, we have to consider, that the BAC clones subjected to exon trapping in our study had been selected from regions on BTA6 poorly covered with protein-coding genes, which could still be noted on the current sequence assembly Btau4.0 (Figure ​(Figure1).1). Eleven BACs were assigned to gene desert regions [9], which have been found conserved in mammals and birds [34,35] and were thought to be transcriptionally silent. Hence, we could not expect a priori that many sequences would have been annotated as known genes. It should also to be considered that even the current annotation of the bovine genome is still limited and consequently, the set of functional elements is not completely identified to date. Additionally, transcripts from the most lowly expressed genes, or genes specifically expressed in important but relatively minor cell types may very likely be under-represented in the ESTs database predominantly established by large scale ESTs projects.

At the beginning of our experiments it was known that only one BAC clone (BBI_750F0243) was assigned near a human gene (KCNIP4) on the syntenic region on HSA4 by in silico comparative mapping. Thus, this BAC clone could serve as a proof for the efficacy of the exon trapping procedure. Indeed, we identified ETS in this BAC clone pinpointing to four exons of the bovine KCNIP4 gene, which underlined and validated the usefulness of the exon trapping method for targeted mining of transcribed sequences in defined chromosomal regions based on genomic DNA from BAC clones. In addition, by identifying two additional exons of the bovine KCNIP4 gene, which are not present in the current bovine genome assembly Btau4.0, it was exemplarily demonstrated that this experimental approach is a useful complement for the annotation of the bovine genome sequence.

Because there was no identity detected to known genes and ESTs by in silico sequence comparison for the majority of ETS identified in our study, these sequences are assumed to be novel and could be predicted to originate from unknown bovine transcripts. Expression analysis was performed to validate this hypothesis. Examination of a subset of the trapped putative transcripts showed exemplarily that in a lactating cow numerous ETS displayed a divergent, tissue-specific expression pattern (Figure ​(Figure2D,2D, Figure ​Figure3).3). Most expression signals were observed in liver, thyroid gland, small intestine, kidney, and pituitary gland. Tissue-dependent expression pattern of the ETS may indicate to potentially specific functions in the corresponding tissues of the lactating cow. Some of the ETS were found to be expressed in all tissues examined, indicating a ubiquitous expression pattern and suggesting them being probably part of housekeeping genes or conserved structural genes, if similarity to repetitive sequences could be excluded. As shown in Figure ​Figure3,3, 16% of the analyzed ETS did not display expression signals in the multi-tissue panel. This could likely be due to the fact, that these transcripts were not expressed in the tissues contained in the panel of a lactating cow analyzed here. In this context it should be mentioned that the advantage of the exon trapping approach is that the method is independent of spatio-temporal expression patterns due to the identification of transcripts based only on intrinsic characteristics of the genuine genomic sequence. But non-expressed ETS, possibly, could also represent false-positive sequences isolated from regions of the genomic DNA due to existing sequence similarities to splice site consensus sequences (e.g., splice donor/acceptor, branch point region), which the exon trapping technique is based on.

From the results of the expression analysis it could be inferred that the ETS revealing expression in the bovine multi-tissue panel may represent bona fide transcribed sequences. However, the ETS have to be characterized in further studies with regard to their functional significance. Based on the presented data, we can not exclude that a part of the identified ETS may be attributed to the class of pseudogenes or to non-functional RNA. Pseudogene transcription has been observed in small-scale gene-centred studies and genome-scale unbiased mapping of transcriptionally active regions in the human and mouse genomes. Surveys of Gerstein and Zheng [36,37] have revealed that for example, 5–20% of human pseudogenes can be transcriptionally active. However, considering the relatively high percentage of ETS (84%), for which expression has been demonstrated in our study, it could be assumed that a number of them might attain to another category of transcribed sequence elements as for example noncoding RNA.

Continued submission of ESTs and other sequence information in a variety of species points to the existence of transcripts that do not map to currently annotated genes [12,38-40]. These transcripts may possibly correspond to novel protein coding genes, genes encoding small unknown peptides, pseudogenes or noncoding RNA. Evidence of transcription had increasingly been found in unannotated intergenic genome regions of the human genome, which were thought to be transcriptionally silent (e.g., [41-46]). The ENCODE consortium reported that a vast amount of DNA, not annotated as known genes, is transcribed into RNA. While the majority of the genome appears to be transcribed at the level of primary transcripts, only about the half of the processed transcripts is mapped as currently annotated genes [36,42,46]. Particularly, a high number of new transcriptionally active regions (more than 50%) were detected in non-annotated intergenic regions. These studies indicated that genomic regions previously considered as "junk" encode for multiple polyadenylated and non-polyadenylated transcripts of unknown function. According to Gerstein et al. [36] the ENCODE project provided evidence that there is much activity between annotated genes and intergenic space in the human genome contributed by transcribed non-protein-coding RNAs and transcribed pseudogenes. The authors highlighted that a number of these transcribed pseudogenes and noncoding RNA genes are located even within introns of protein-coding genes and assumed that these components may possibly influence the expression of their host genes. It is also possible that these transcripts themselves do not have a direct function, but rather are important for a particular process (e.g., chromatin accessibility for transcription factor binding). Continuously, numerous noncoding RNA sequences are recognized in the transcriptomes of different eukaryotes as having important regulatory functions in controlling various levels of gene expression in physiological and developmental processes and diseases of complex organisms (e.g., [44,47-49]). The detailed investigation of the functional relevance of the numerous unknown transcripts was postulated as a prospective task in the post ENCODE era.

The findings of our study provide experimental support for transcripts lacking ESTs or other cDNA evidence in the targeted regions of BTA6. The majority of unknown ETS presumably identified novel noncoding transcripts located in intergenic regions of the chromosome. However, prospective studies should be performed to further characterize the transcripts with regard to their putative functional significance. In this respect it has to be proven, if these transcripts belong to non-functional RNA or if they have any specific regulatory function in the bovine genome. Currently, there is scare information on the function of bovine noncoding RNA genes compared to the state in mouse and human. The prevalence of bovine noncoding RNAs, their regulatory impact on gene expression and their physiological effects are not yet examined in detail.

While the mammalian genomes contain nearly similar repertoires of protein-coding gene sequences comprising only a fraction of about 1.5% of the whole genome, the majority of the mammalian genome is obviously transcribed. Our results support the increasingly accepted concept suggesting that the physiological complexity and the unique phenotypes of species-specific or individual genomes might evolve from combinatorial features contributed by the entire genome sequence including previously neglected genome regions [33,42,45,46,50]. Consequently, variation in noncoding sequences might be important effectors of phenotypic variation in complex traits and diseases (see reviews [51,52]) in livestock.