Gene discovery in the hamster: a comparative genomics approach for gene annotation by sequencing of hamster testis cDNAs

Generation and sequence analysis of a Hamster testis cDNA library

We next considered the overall complexity of the sequences in our database as our cloning strategy made it possible to obtain multiple fragments from the same cDNA. We identified only 40 instances where more than one fragment was derived from the same gene (data not shown). In addition, only 17 fragments appeared more than two times in the database and only 6 fragments appeared more than three times. The most common sequences were derived from genes that are known to be highly expressed in the testis in other organisms and include heat shock protein 90A [20], chitobiase [21], outer dense fiber of sperm tails [22] and kinectin [23]. Therefore, the majority of sequences were represented only once or at most two times in the data set. The fact that most sequences were not identified multiple times suggests that we have not exhaustively sequenced all of the DpnII fragments from cDNAs expressed in the hamster testis. This is not surprising as conservative estimates suggest that there are many thousands of unique transcripts expressed in the various somatic and germ cell lineages in the testis.

The largest group of sequences identified in this project falls into the functionally unclassified group (Table 1). This group includes a small number of named genes for which no function has yet been assigned but primarily contains the results of full-length cDNA sequencing projects or genes predicted by gene finding software. Approximately one third of these clones (78 out of 239) were originally isolated from testis libraries, suggesting that a large number of genes remain to be functionally characterized in the testis.

Identification of novel testis specific genes

200 sequences could not be sorted into the various functional groups described above and represented a potential source of novel genes that had not previously been recognized by EST or genome sequencing. However, there were several other possible origins for such sequences. These included contamination of the original cDNA sample with genomic DNA (note that the testis total RNA was treated with DNase before mRNA purification) or artifactual joining of unrelated sequences into chimeric fragments that resulted in matches that were not detected in our screen. As we synthesized cDNA from total RNA and not cytoplasmic RNA, it was possible that some fragments could be derived from RNAs that were incompletely processed [24]. Furthermore, these fragments could be derived from alternatively spliced exons of known genes or from genes that are only expressed in hamster. Therefore, several clones were selected for further examination based on their sequence similarity to unannotated regions of the human genome (see below) and for Northern analysis using hamster RNAs prepared from various tissues. Eight clones yielded specific signals in RNA from hamster testis while two clones also detected signals in other tissues (figure 1). Subsequent analysis has revealed that each clone is derived form a bona fide gene and the evidence for this conclusion is provided below for each gene (see Table 2). The derived protein structures of the polypeptides encoded by each gene are shown in figure 2.

1030: A new member of the kinesin superfamily

The 1030 clone contained an ORF of 80 amino acids that displayed similarity to members of the kinesin superfamily of molecular motors [25, 26]. Northern analysis using the hamster 1030 probe detected a single band in testis RNA and weak signals in several other tissues including brain, heart and lung (figure 1). Comparison of the hamster nucleotide sequence with genomic sequences revealed highly significant matches within chromosomal positions 9q21.33 in human, 13B2 in mouse and 17 in rat. Partial cDNAs from human and macaque also mapped to the same genetic locus (Table 2). Using the procedures described in materials and methods we mapped an 18 exon gene that was followed by a consensus polyadenylation signal in the human, mouse and rat genome sequences (figure 3A). Exon 1 is located within a strong CpG island, indicating the likely presence of a promoter in this region. PCR cloning of the human cDNA was employed to confirm the exonic structure predicted by the genomic sequences. These studies revealed the existence of three alternatively spliced products, one containing all 17 coding exons (exon 1 is non-coding) and two variants lacking either exon 11 or exons 12 and 13 (figures 3B and 3C).

Conceptual translation of the predicted full-length human mRNA revealed a 1401 amino acid protein that is highly conserved in macaque, mouse and rat (figures 2 and 4). Database searches revealed that it encodes a previously unreported member of the kinesin superfamily. The kinesin superfamily in humans and mice is comprised of at least 45 members, designated with the prefix KIF [26]. Based on this naming system we have assigned the name KIF27 to this new family member with the suffixes A, B and C to designate the splice variants described above (figure 3B). Domain mapping within KIF27 revealed an amino terminal kinesin motor domain and a putative topisomerase domain located in the center of the protein (figure 2 and 3B). Phylogenic analysis revealed that KIF27 belongs to the N-5 phylogenic group of KIFs [26], which includes mouse KIF21A and 21B, and human KIF4 (a group also known as the chromokinesins). Comparisons with the other members of the N-5 subgroup revealed greatest similarity within the motor domain (47–50% identity) (figure 5). In addition, the sequence of the neck domain of KIF27 located at the C-terminal end of the motor domain conformed to the consensus sequence for the neck domain of chromokinesins (figure 3B) [27].

7012

This clone contained an ORF of 73 amino acids that detected a strong signal by Northern blotting in testis RNA (figure 1). Genome analysis permitted the mapping of a common set of 10 exons on human, mouse and rat chromosomes 5, 15 and 2, respectively (Table 2), assembly of which resulted in translation products of 579, 555 and 563 amino acids, respectively (figure 6). These predictions were also corroborated by a macaque cDNA (AB070167). The encoded proteins did not contain any recognizable functional motifs (figure 3) and were most similar to an uncharacterized human protein named B29 (figure 2). B29 was originally identified as a gene located on human chromosome 18q21 in a search for candidate tumor suppressor proteins in lung [28]. However, the function of B29 is unknown and it does not contain any obvious functional domains. Interestingly, the apparent size of the detected mRNA is significantly larger than that of the assembled sequences, suggesting that additional exons are likely to remain to be discovered. In this regard, gene prediction software identified an additional 35 exons in the mouse genomic sequence that were partly conserved in the rat but not present in the human sequence. Final mapping of the genomic structure of 7012 will await further refineme nt of each genomic sequence.

9004

This clone encoded an ORF of 87 amino acids that did not display significant similarity to any known proteins in public databases. Northern blotting revealed a specific signal in hamster testis RNA (figure 1). The clone sequence mapped to human chromosome 1p13.3, mouse chromosome 3F3 and rat chromosome 2 (Table 2). Human and macaque cDNAs have recently appeared in the database that encompass this clone and the sequences of their encoded proteins are compared in figure 7. These polypeptides contain a predicted coiled coil region (figure 2) but do not contain other functional domains that might indicate their possible function(s).

13043

This clone encodes an ORF of 56 amino acids that did not display significant similarity to known proteins. Northern blotting revealed a strong signal in testis (figure 1). Genome comparisons revealed matches on human chromosome 3, mouse chromosome 16 and rat chromosome 11 (Table 2). Human, mouse and rat coding sequences were assembled using a macaque cDNA (AB070087) as template (figure 8). Domain mapping identified several conserved transmembrane domains in each protein as well as a putative cyclic nucleotide binding site close to the C-terminus (figure 2). In addition a putative cation channel was identified in the center of the protein. Further analysis will be necessary to determine whether this protein functions as a regulatable cation channel.

15014

This clone encodes an ORF of 67 amino acids whose mRNA was specifically detected in hamster testis RNA (figure 1). Genome comparisons revealed strong similarities with regions of human chromosome 11, mouse chromosome 9 and rat chromosome 8 (Table 2). This clone matches to hypothetical proteins recently added to the annotation of the human and mouse genomes (FLJ13386 and XP_134746) as well as a protein encoded by a macaque cDNA (BAB63125). The alignment in figure 9 accounts for each of these clones as well as additional exons predicted in the mouse from inter-genome comparisons. The encoded protein contains several predicted coiled-coil regions but no other identifiable functional domains (figure 2).

15018

This clone encodes an ORF of 47 amino acids that detected a specific signal in hamster testis RNA (figure 1). It mapped to a predic ted 20 exon gene on human chromosome 4, with orthologous (but incomplete) sequences on mouse chromosome 3 and rat chromosome 2 (Table 2). An orthologous protein has recently been reported from rat and named sodium channel associated protein 1A (SCAP1A). Alignment of the human and rat proteins is shown in figure 10. Although the specific function of this protein is currently unknown, it contains a potential V-type ATPase domain as well as several predicted coiled-coil regions (figure 2).

15037

This clone encodes an ORF of 59 amino acids that was detected specifically in hamster testis (figure 1). It mapped to specific regions on chromosomes 10, 19 and 1 in human, mouse and rat, respectively. A comparison of the predicted human protein based on several partial human cDNAs and a mouse protein named oocyte-testis gene 1 (Otg1) is shown in figure 11. The protein sequence is predicted to contain several coiled coil regions but no other potential functional domains (figure 2).

19045

This clone detected two distinct transcripts in testis, brain and heart by Northern blotting (figure 1). The sequence mapped to regions on chromosomes 1, 1 and 13 in human, mouse and rat, respectively and we have assembled the orthologous sequences from human, mouse, rat and Fugu (figure 12). The protein is 433 amino acids in length in human, mouse and rat (434 in Fugu) and contains two recognizable functional domains characteristic of proteins with chromatin remodeling activity (figure 2). The first is an interrupted SET (S u(var)3–9, E(z), trithorax) domain [29] and the second is a MYND (m yeloid transcription factor, n ervy, D EAF-1) domain [30]. A similar domain organization is also found in the BOP (CD8b op posite, recently renamed the SET and MYND domain protein (SMYD)) family of proteins and alignment of the human protein identified here with SMYD1 (BOP1) from mouse and chicken is shown in figure 13. The SMYD1 proteins are transcriptional regulatory proteins with chromatin-remodeling activities important in cardiac tissue, muscle and T lymphocytes [31–33]. Therefore, the new protein identified in this report is likely to play as yet undefined roles in transcription and has been assigned the official name SMYD2 to designate its relatedness to SMYD1.

KIF27: a new kinesin family member

Our studies revealed the existence of a new member of the kinesin superfamily of microtubule-based molecular motors, which we have named KIF27. KIF27 RNA was detected primarily in hamster testis but weaker signals were also present in several other tissues, suggesting that this protein may function in numerous cell types. Significant characteristics of the KIF27 genes mapped in human, mouse and rat included a conserved 18 exon arrangement and the existence of at least three mRNAs that resulted from alternative splicing, at least in human. Although partial cDNAs existed for the human and macaque genes in public databases, we were able to construct full length cDNA sequences for human, mouse and rat from genomic comparisons, and to assemble a corresponding sequence from macaque by joining two previously reported cDNAs. Kinesins are characterized by a conserved motor domain of approximately 350 amino acids that may be located at the amino terminus (KIN-N), carboxy-terminus (KIN-C) or within the polypeptide (KIN-I) [27] and the motor domain of KIF27 is located in the N-terminus of the protein. Based on this arrangement and the sequence of the adjacent neck domain, we assigned KIF27 to the N-5 phyogenetic group of kinesins [27]. The N-5 subgroup is defined by the human KIF4 protein, which is primarily localized to the nuclear matrix and associates with chromosomes during mitosis [35]. In cell division, nuclear kinesins of this chromokinesin class appear to be important for the maintenance of sister chromatids on the metaphase plate [36]. However, additional functions for nuclear kinesins have recently been uncovered. For example, KIF17a, a member of the N-4 subgroup that is predominantly expressed in germ cells, was recently shown to possess transcriptional regulatory properties by controlling access of the transcriptional activator protein CREM to a coactivator, ACT [37]. Clearly, functional analysis of the KIF27 polypeptide will be needed to determine the subcellular location of this protein to determine whether it may add to the growing number of kinesins that function in the nucleus. In this regard, although the carboxy terminal regions of KIF27 and KIF4 display little significant sequence similarity, both contain a putative topoisomerase domain that may be important for nuclear functions (figure 2). In addition, several clusters of basic amino acids that may function as nuclear localization signals are located in the KIF27 polypeptide.

SMYD2: a putative chromatin remodeling protein

The final clone isolated in our search encodes a 433 amino acid protein whose sequence was highly conserved from human to pufferfish. This protein, now named SMYD2, contains SET and MYND domains that are characteristic of proteins with chromatin remodeling capabilities. The SET domain is a common feature of proteins with histone lysine methyl transferase (HKMT) activity and has been identified in hundreds of proteins in organisms ranging from bacteria and viruses to humans [38]. The SMYD2 SET domain is separated into two parts (i.e. a S-ET domain) and is followed by a short cyteine-rich region that is common in many SET domain proteins (figures 2 and 12). The MYND domain is located between the two halves of the SET domain and similar domains have been identified in a number of proteins that function as transcriptional repressors, including the ETO protein that is fused to the AML-1 transcription factor in the t(8;21) translocation in acute myeloid leukemias [39]. The MYND contains two zinc finger motifs and is a protein-protein interaction interface responsible for the recruitment of corepressors [39–41].

This domain organization is conserved in several proteins in public databases, including the recently described SMYD1 (aka BOP) family of proteins [31]. Three isoforms of SMYD1 have been reported thus far (referred to by their original names, m-BOP1, m-BOP2 and t-BOP) that are products of a single gene that result from either alternative splicing or promoter usage [31–33]. m-BOP is essential for cardiac differentiation and morphogenesis while t-BOP is expressed in cytotoxic T lymphoctes. Studies are currently underway to examine the function(s) of SMYD2 in testis and heart.

Genomic studies in the hamster

The impetus for this study arose from the need to perform microarray experiments to investigate the molecular changes elicited by environmental toxicants on male reproduction function in hamster. Comparisons performed between the limited numbers of hamster sequences in public databases with ortho logous sequences from the mouse suggested that, despite the close taxonomic relationship between hamster and mouse, evolutionary divergence in coding sequences was sufficiently great in certain cases that reagents developed for the mouse would be of limited use for genomic studies in the hamster. This conclusion was supported by experimental observations in our laboratories indicating that probes derived from rat and mouse cDNAs yield inconsistent results in Northern blotting under stringent hybridization conditions (data not shown). The reagents described here will now permit the initiation of genomic studies in the hamster.

RNA, cDNA and plasmid preparation

Total RNA was prepared from testes of adult Syrian golden hamsters (Mesocricetus auratus). Poly A⁺ RNA was isolated using the poly A Spin™ mRNA isolation kit (New England Biolabs, Beverly, MA). 5 μg of polyA⁺ RNA was converted to double stranded cDNA using the cDNA Synthesis Kit (Life Technologies, Gaithersburg, MD). The cDNA was then digested to completion with DpnII and electrophoresed on a 1.5% agarose gel. Five populations of digested cDNAs in size ranges between 100 and 800 bp were excised from the gel and purified using Qiaex II resin (Qiagen, Valencia, CA). Each population of cDNA was ligated into a pBluescript vector that had been digested with BamHI and alkaline phosphatased to decrease the rate of self-ligation. The ligations were transformed into DH5α supercompetent cells (Life Technologies) and positive clones were identified by blue-white selection. The success of the cloning procedure was initially monitored by picking 5 clones from each group, preparing plasmid DNA and sequencing. These preliminary studies indicated that the most useful clone sets were those derived from cDNAs in the 100–300 bp range. White colonies from these two sets were carefully picked and used to inoculate single cells of a 96 well culture block with each well containing 1.2 ml of TB (1.2% Tryptone, 2.4% yeast extract, 0.4% glycerol) supplemented with 50 μg/ml carbenecillin. The bacteria were grown for 20 hours at 37°C and glycerol stocks were subsequently prepared using a Biomek 3000 robot (Beckman-Coulter, Inc., Fullerton, CA). Fresh cultures (in 1 ml of 2X LB (2% Tryptone, 1% yeast extract, 1% NaCl) plus 50 μg/ml carbenecillin) were inoculated from the glycerol stocks using a 96-well needle transfer device and grown as before. Plasmid DNA was prepared using the full lysate protocol of the Montage™ Plasmid Miniprep96 Kit (Millipore, Bedford, MA). Typical yields of plasmid DNA ranged from 5–10 μg in 50 μl volume.

Sequencing

Plasmid DNA was diluted four-fold in water and 2 μl of the diluted sample was sequenced in 96-well format using the Dye Terminator Cycle Sequencing Kit (Beckman Coulter, Inc.). The reactions were cleaned up by an ethanol precipitation step in which 35 μl of 10 mg/ml glycogen, 70 μl of 0.5 M EDTA, 70 μl of 1.5 M Sodium Acetate (pH 4.8) was first mixed with the sequencing reaction (6.6 μl). 20 μl of 95% ethanol was added and the plate was centrifuged at 4,000 rpm for 45 minutes. The pellets were washed twice with 100 μl of 70% ethanol, air dried and dissolved in 25 μl of molecular biology grade formamide. A drop of mineral oil was overlaid on each reaction and these were then run on a CEQ2000 capillary array sequencer (Beckman Coulter, Inc.).

Sequence analysis

Raw sequence data was first imported into the Contig Express component of the Vector NTI suite of sequence analysis programs (InforMax, Inc. Bethesda, MD). Each clone was named according to its position in the original 96-well plate, for example, clone 1030 came from position 30 in plate 1. Vector sequences were first scanned in batch mode for the presence of vector-derived sequences using Contig Express and these sequences were trimmed before proceeding. In some cases, manual processing of the sequences was necessary using the VecScreen program at the National Canter for Biotechnology Information (NCBI) http://www.ncbi.nlm.nih.gov/VecScreen/VecScreen.html. The presence of DpnII sites at either end of the clone simplified the removal of vector sequences. The sequences were concatenated in batches of 96 into a text file using FASTA format and submitted for a batch BLASTN search using the interface program BLASTCL3. This search returned the three top matches into an output file. No further analysis was performed in those cases where a match with a known gene was clearly established, and the clone was annotated as the hamster orthologue of a known gene. A series of additional comparisons were performed if a clear relationship between a hamster clone and known sequence was not established by this initial search. The most useful additional comparison was a BLASTX search, which compares the translated sequence of the input clone in all six frames to the peptide sequences in GenBank. This search clarified the identity of several additional hamster clones. 200 clones did not match to known sequences and were analyzed to determine whether they might represent potentially novel genes. First, the output sequences were translated in all frames to determine whether a complete open reading frame (ORF) was present. If so, the clones were resequenced on both strands to ensure that the sequence was correct. Promising sequences were compared to the public version of the annotated human sequence (v 9.30a.1) at Ensembl http://www.ensembl.org to find matches with human chromosomal sequences. Further comparisons were performed against the mouse (v 9.3a.1), rat (v 9.1.1), zebrafish (v 9.08.1), pufferfish (v 9.1.1) and mosquito (v 9.1a.1) genome sequence in the same database. When matching genomic loci were identified, approximately 100,000 bp of genome sequence surrounding the match was imported into Vector NTI and submitted to gene prediction software programs to determine if the hamster sequences were located within exons of predicted genes. Two gene prediction programs were used FGENESH http://www.softberry.com/berry.phtml and GENEMARK http://opal.biology.gatech.edu/GeneMark/eukhmm.cgi?org=H.sapiens. Predicted cDNA sequences were then assembled into contiguous files and subjected to further comparisons against cDNA and EST databases at NCBI. Genome structures were further examined using PipMaker, a program that supports comparisons of large DNA fragments and identifies short conserved regions of similarity, such as exons. The predicted protein sequences of the derived cDNAs were analyzed for the presence of functional domains using the conserved domain function of BLAST at NCBI as well as the Simple Modular Architecture Research Tool at http://smart.embl-heidelberg.de. Protein structures were annotated in Vector NTI and published in Canvas v7.0 (Deneba Systems, Inc., Miami, FL). Sequences were submitted to the appropriate databases at the National Center for Biotechnology Information (NCBI). Specifically, EST sequences were submitted to dbEST under accession numbers BI431001-BI431008 and CB884447-CB885166. Human KIF27 sequences were submitted to Genbank under accession numbers AY237536-AY237538. Sequences defined by annotation of previously available sequences were submitted to the Third Party Annotation database under accession numbers BK001053-BK001057 and BK001326-BK001332. The Human Genome Organization (HUGO) nomenclature committee has approved proposed gene names.

Northern analysis

Total RNA was purified from various hamster tissues using TriZol reagent (Life Technologies, Inc.). 20 μg of each RNA was electrophoresed through a 1% agarose-formaldehyde gel and transferred to a Nylon membrane (Micron Separations, Westborough, MA). The RNA was cross-linked to the membrane by exposure to UV light and hybridized with specific probes labeled with ³²P-dCTP (Perkin Elmer Life Sciences, Boston, MA) using the Prime-It random prime labeling kit (Stratagene, La Jolla, CA). The hybridization procedure was performed as described before [42]. Several duplicate membranes were prepared and the radioactive probe was stripped after each round of hybridization by boiling for 10 minutes in 10 mM Tris-HCl (pH 7.5), 1 mM EDTA, 1% SDS. The specific probes were prepared by digesting the appropriate pBluescript plasmid with XbaI and EcoRI, separating the fragment on a 1.5% agarose gel followed by extraction using the Qiaex II extraction kit (Qiagen).

PCR cloning of human KIF27 cDNAs

PCR primers based on the predicted human KIF27 genomic sequence were designed for PCR cloning of overlapping regions of human KIF27 cDNAs. PCR reactions were performed using human testis Marathon RACE-Ready cDNA (Clontech, Palo Alto, CA) using conditions described previously [43]. PCR products were subcloned in to the pGEM-Teasy plasmid and sequenced. The sequences were then assembled into full length cDNAs using the Vector NTI sequence analysis suite of programs. The primers used for the reaction shown in figure 2 were: 5' AACTAGATGTAGAAGTCGTTCATGGATTC 3' and 5'TTCCAGTAAGTTCAGGCGAGTTG 3'.