Endosymbiont DNA in Endobacteria-Free Filarial Nematodes Indicates Ancient Horizontal Genetic Transfer

Background Wolbachia are among the most abundant symbiotic microbes on earth; they are present in about 66% of all insect species, some spiders, mites and crustaceans, and most filarial nematode species. Infected filarial nematodes, including many pathogens of medical and veterinary importance, depend on Wolbachia for proper development and survival. The mechanisms behind this interdependence are not understood. Interestingly, a minority of filarial species examined to date are naturally Wolbachia-free. Methodology/Principal Findings We used 454 pyrosequencing to survey the genomes of two distantly related Wolbachia-free filarial species, Acanthocheilonema viteae and Onchocerca flexuosa. This screen identified 49 Wolbachia-like DNA sequences in A. viteae and 114 in O. flexuosa. qRT-PCR reactions detected expression of 30 Wolbachia-like sequences in A. viteae and 56 in O. flexuosa. Approximately half of these appear to be transcribed from pseudogenes. In situ hybridization showed that two of these pseudogene transcripts were specifically expressed in developing embryos and testes of both species. Conclusions/Significance These results strongly suggest that the last common ancestor of extant filarial nematodes was infected with Wolbachia and that this former endosymbiont contributed to their genome evolution. Horizontally transferred Wolbachia DNA may explain the ability of some filarial species to live and reproduce without the endosymbiont while other species cannot.


Introduction
Several important evolutionary milestones, such as the emergence of eukaryotes and the development of intracellular organelles (e.g. mitochondria and plastids), have involved disparate species uniting to form composite organisms. Filarial nematodes and their Wolbachia endobacteria are an interesting example of such a composite [1]. Although antibiotics can be used to cure arthropods of their Wolbachia infection, similar treatments lead to infertility, improper development and sometimes death of Wolbachia-dependent filarial nematodes [2,3]. Likewise, attempts to maintain filarial Wolbachia in culture have failed (B.E. Slatko, P.U. Fischer, R.U. Rao, pers. comm.). This interdependence may be a consequence of reductive evolution in both partners, as several biosynthetic pathways (e.g. synthesis of heme, riboflavin, nucleotides, etc.) seem to require both genomes for complete functionality [4].
Wolbachia are vertically transmitted through infected oocytes [5]. Their presence in the germline allows for heritable DNA transfer from the bacteria to the metazoan host. Horizontal genetic transfers (HGTs) have been reported in several Wolbachia-infected filarial and arthropod species [6,7,8,9], and evidence for transcription of transferred sequences has been reported in Drosophila ananassae that have been cured of Wolbachia by antibiotic treatment [7]. Similar expression studies in filarial nematodes are difficult because the infection cannot be cleared without damaging the worms. In any case, the transferred Wolbachia DNA sequences found in the Wolbachia-dependent filarial species are most likely degenerate [7] and therefore unable to produce functional proteins.
In this study, we examined two filarial nematode species from different clades [10] that are naturally Wolbachia-free, namely Acanthocheilonema viteae (a rodent parasite whose life cycle can be maintained in the laboratory) and Onchocerca flexuosa (a parasite of European red deer and a close relative of the agent of African river blindness) [11,12,13]. It has been suggested that the ancestors of these species were colonized in the distant past, as some 90% of filarial nematode species examined to date contain the bacteria [10,11]. We hypothesized that if this is true, HGT may have brought Wolbachia DNA into the nuclear genomes of these species prior to endosymbiont loss. We used massively parallel sequencing to survey the genomes of A. viteae and O. flexuosa in search of Wolbachia-like DNA sequences. The presence of such Wolbachialike sequences in their nuclear genomes provides the first direct evidence that the ancestors of these species harbored Wolbachia endosymbionts that were subsequently lost. Transferred Wolbachia genes and/or regulatory elements may help explain the ability of uninfected species to survive without a bacterial partner. Further analysis of transferred genes will provide insight into the nature of the symbiotic relationship between Wolbachia and its filarial nematode hosts.

Genome sequencing and identification of Wolbachialike sequences
To survey the genomes of A. viteae and O. flexuosa for transferred Wolbachia sequences, fragment and paired-end genomic libraries were sequenced using 454 GS-FLX technology. Two orthologous approaches were undertaken to remove redundancy or capture longer contigs containing Wolbachia homologs (see Materials and Methods). B. malayi, the only filarial nematode for which draft genome information is available, has an ,95 Mb genome containing ,14% repetitive sequences [14]. Assuming a similar size and structure, we estimate that assembled contigs provide a ,38% coverage of the nonrepetitive portions of the A. viteae genome. It is not possible to estimate coverage of the O. flexuosa genome, as the coverage was insufficient for assembly of paired-end reads. BLASTN analyses identified 45 and 92 genomic fragments containing Wolbachialike sequences in A. viteae and O. flexuosa, respectively (Tables 1, S1 and S2). Subsequent similarity searches found that 14 of the 45 genomic fragments in A. viteae and 32 of the 92 genomic fragments in O. flexuosa also contain filarial nematode gene homologs. This demonstrates that the Wolbachia homologs residing on these fragments are physically integrated into the filarial genomes.

Analysis of Wolbachia-like sequences
BLAST analysis was used to annotate the genomic DNA fragments identified in this screen (Tables S1 and S2). A total of 49 and 114 Wolbachia-like DNA sequences were identified in A. viteae and O. flexuosa, respectively. The average identity (% 6 standard deviation) of the Wolbachia homologs to their top BLAST hit was 7866% in A. viteae and 8166% in O. flexuosa, and the average alignment length (bp 6 standard deviation) was 159683bp and 1746192 bp for A. viteae and O. flexuosa, respectively. For comparison, the average identity to a filarial nematode gene was 79616% in A. viteae and 8366% in O. flexuosa. Despite low-level sequence coverage, seven Wolbachia genes were represented by sequence fragments in both A. viteae and O. flexuosa. Some of the transferred fragments present in A. viteae or O. flexuosa also correspond to Wolbachia gene fragments present in the nuclear genome of B. malayi [7]. So far, none have been identified in all three species (Table 2).

Cellular processes represented by the transferred DNA
Each of the Wolbachia-like gene fragments identified in this study was assigned to a COG functional category in order to determine which cellular processes and pathways were most heavily represented in our transferred fragment collection (

Potential source of transferred fragments
Of the 49 Wolbachia homologs found in A. viteae, 19 (39%) align best to a filarial Wolbachia sequence and 30 align best to an insect Wolbachia sequence. Likewise, 47 (41%) of the Wolbachia homologs in O. flexuosa align best to a filarial Wolbachia sequence and 67 align best to an insect Wolbachia sequence (Tables S1 and S2). Additionally, there was no apparent clustering pattern when transferred sequences were aligned to the circular genome of the Wolbachia endosymbiont of B. malayi (Fig. 1). Alignment of transferred fragments to the genomes of the Wolbachia endosymbionts of Drosophila melanogaster, Drosophila simulans, and Culex pipiens showed a similar lack of clustering. Therefore, we cannot determine whether a large piece of Wolbachia DNA (or even an entire bacterial genome) was inserted and subsequently fragmented and scattered in the filarial genomes over time, or if small fragments were shuttled into the genome separately.

Mechanism of transfer
The mechanism responsible for DNA transfer from Wolbachia is unknown, but the sequence data provide some interesting clues  Arthropod Wolbachia contain prophage sequences thought to enhance DNA rearrangements [15]; one phage sequence was detected in A. viteae (wAv16332).

Coding potential of transferred fragments
Most of the sequences identified in this screen represent only small portions of Wolbachia genes. Some of these sequences, 23 in A. viteae and 61 in O. flexuosa, are truncated at the end of a contig, so further sequencing will determine their actual length. The fragments that fall entirely within a sequenced contig (i.e. .

Expression of the transferred fragments
Sybr Green qRT-PCR was used to assess expression of Wolbachia-like sequences regardless of the presence of an open reading frame. Stringent controls were used to rule out DNA contamination. Thirty of 42 A. viteae and 56 of 87 O. flexuosa Wolbachia-like sequences tested were expressed at the RNA level (Table S5). 14 of the expressed Wolbachia homologs in A. viteae and 34 in O. flexuosa appear to be transcribed from pseudogenes relative to what is known from the endobacterial genome.

Localization of transcripts
In situ RNA hybridization was used to localize two transcripts predicted to arise from pseudogenes of 2-methylthioadenine synthase (2-MAS), which contains a frameshift mutation, and DNA polymerase I (pol I), which contains several premature stop codons. The Wolbachia homologs of both of these sequences are involved in nucleic acid synthesis. The 321bp 2-MAS and 415bp polA probe sequences had 83% and 73% identity to their homologs in wBm, respectively. The 2-MAS probe labeled the lateral chords and intrauterine stretched microfilaria, tissues containing Wolbachia endobacteria, as well as the intestinal and uterine epithelium in B. malayi (Fig. 3B). The staining of the intestine and uterine epithelium could be indicative of an expressed HGT fragment in B. malayi, as these tissues do not contain Wolbachia. In A. viteae, this probe labeled developing embryos in females and late spermatogonia in males ( Fig. 3D and 3F). No staining was seen in the lateral chords. Most of the examined O. flexuosa nodule sections did not contain females with developing embryos. However, one O. flexuosa nodule containing a young male showed intensive labeling of the testes similar to that seen in male A. viteae (Fig. 3F, H). The pol I probe weakly labeled Wolbachia in the lateral chords (arrows) of B. malayi (Fig. 3J). Some Wolbachia inserts in the nuclear genome of B. malayi were described previously [7]. Shared homologs were identified by alignment to Wolbachia sequences with the same locus tag by BLASTN. Presence of sequences in the nuclear genome or among transcripts (+), lack of expression at RNA level (2), and an inability to test for expression (n/a) are noted. doi:10.1371/journal.pone.0011029.t002 background staining was observed in the ovaries and uterus using the sense probe ( Fig. 3 I, K), but this was very weak compared to the strong signal obtained with the anti-sense probe in female B. malayi and A. viteae (Fig. 3J, L). These results show that Wolbachialike pseudogene transcripts can be detected in Wolbachia-free filarial species. Expression appears to be tightly regulated, because not all stages, body regions and tissue types were stained.

Discussion
The nuclear genomes of two distantly related, Wolbachia-free filarial nematode species contain Wolbachia-like DNA sequences that were obtained from a former endosymbiont via HGT. We detected transcription of several Wolbachia homologs present in A. viteae and O. flexuosa despite the fact that many of these sequences are degenerate. In situ hybridization showed that two pseudogene transcripts had tissue-specific expression patterns in three filarial species. Our results provide strong evidence to support the hypothesis that the ancestor of extant Wolbachia-free filarial species was infected with Wolbachia.
Low coverage genome sequencing was used as a cost-effective approach to identify Wolbachia-like sequences. Providing draft genomes and a full inventory of all transferred DNA fragments from A. viteae and O. flexuosa was beyond the scope of this project.
Rather, our aim was to provide evidence that the absence of Wolbachia in presently uninfected filarial species is due to secondary loss and that DNA had been passed from Wolbachia to these species prior to this loss. At present, draft genomes are only available for three parasitic nematode species; over 20 more are in progress [16]. Although priority is still given to pathogens of socioeconomic importance, future studies may provide more complete drafts of these genomes at a later date.
The genetic screens described in this report identified 49 and 114 Wolbachia homologs in A. viteae and O. flexuosa, respectively. It will not be possible to determine if O. flexuosa and A. viteae have the same number of transferred genes until both genomes are fully sequenced. However, it is likely that we have identified more sequences in O. flexuosa due to technical issues like the use of a paired-end sequencing approach as opposed to the traditional fragment library that was used for A. viteae.
Several methods are commonly used to detect HGT events [17]. Approaches such as analysis of GC content or codon bias were not feasible because the GC content of filarial worms is similar to that of Wolbachia [4], and because we do not have enough information about the genomes of A. viteae and O. flexuosa to evaluate their codon usage. Our project relied heavily on homology-based analyses. This was ideal for our application because the suspected source of the transfer could be inferred based on the presence of Wolbachia in other filarial species and based on the propensity of Wolbachia to transfer DNA to the nuclear genomes of its hosts [7].
It is believed that mutualistic Wolbachia provide metabolites that are essential for host reproduction, development and survival. The combined genomes of B. malayi and its Wolbachia endosymbiont encode complete pathways for the biosynthesis of purines, pyrimidines, riboflavin, flavin adenine dinucleotide and heme, pathways that are missing or incomplete in the filarial genome alone [4]. Some of the sequence fragments we identified are homologous to genes involved in these processes (Tables S1 and  S2). For instance, several genes related to heme synthesis and export were present in the HGT fragment list. Nematodes, including C. elegans, are not known to synthesize their own heme [18], so the presences of a complete heme biosynthetic pathway in either A. viteae or O. flexuosa would be unique. Since only portions of these genomes have been scanned, further studies are needed to determine the full insert length for many of the known Wolbachia homologs and to reveal more transferred sequences, some of which may encode full-length gene products.
The data provide intriguing clues about the donor strain and how transferred DNA was incorporated into the nuclear genome of filarial nematodes. If the transfer came from an extant Wolbachia species, the majority of the fragments would show a top BLAST score to this species. We would also expect that the transferred fragments would fall within certain portions of the donor genome if they were incorporated in one or few events. These characteristics were not seen when we compared the transferred fragments to sequenced Wolbachia genomes. The variability in the strain of the top BLAST hit and the average percent identity to that hit suggests that transfer(s) either came from a species that is not represented in public databases and/or that the transferred sequences have mutated over time with respect to their parent gene. Either way, these sequence changes must have taken millions of years to accumulate and stabilize in filarial and Wolbachia populations [19]. The initial transfer event(s) probably took place between an ancestral Wolbachia strain and an ancestral filarial species. It is likely that neither of these exists in the same form today, but we see the ''fossilized'' evidence of their interaction in the genomes of their descendents.  Questions remain as to how the DNA escaped the bacterial compartment and entered the nucleus of the host cell. It is possible that this may have involved bacterial Type IV secretion, a system conserved in Wolbachia. Type IV secretion systems are capable of shuttling proteins and nucleoprotein complexes across membranes and are known to facilitate gene transfer [20,21]. After translocation, DNA may have been inserted into the host genome by transposable elements. Furthermore, the remnants of retrotransposons and large poly(A) stretches found in our sequences suggest that some of the sequences may be processed pseudogenes derived from mature mRNAs [19]. We have yet to identify the parent copies of these putative processed pseudogenes.
Many of the Wolbachia homologs identified in this study show insertions and/or truncations, frameshift mutations and premature stop codons relative to homologs from sequenced Wolbachia genomes. Gene fragmentation and degradation suggests a lack of selective pressure to maintain coding capacity, but full length transcripts and proteins are not always required for biological function. For example, transfer of DNA from mitochondria to the nuclear genome can generate novel exons that alter protein function [22]. Likewise, truncated Wolbachia sequences inserted into filarial genes may act as new protein domains that alter or enhance the function of existing nematode proteins. Pseudogenes containing frameshifts and premature stops are abundant in many genomes [23], and widespread transcription of pseudogenes has been reported [24,25]. mRNA recoding mechanisms, which are employed infrequently in most organisms, could allow the translational machinery to produce a protein despite these coding errors [26]. Even if the pseudogene sequences are not translated, recent studies have shown that expressed pseudogenes sequences can regulate expression of other genes through RNA interference [27,28,29,30]. Further studies will determine which of these mechanisms might allow the transferred Wolbachia sequences reported here to contribute to filarial biology.
We have provided strong evidence that two distantly related Wolbachia-free filarial nematodes contain Wolbachia-like DNA in their nuclear genomes. HGT from bacteria may be a relatively common phenomenon in nematode phylogeny. As previously mentioned, Wolbachia have inserted DNA in the nuclear genomes of endosymbiont-dependent filarial nematode hosts [7]. In these cases, the source of the transfer is obvious because the mutualistic relationship between the two organisms has been maintained, but the functions of the inserted sequences are still unknown. The converse situation exists in plant parasitic nematodes. Cellulase enzymes that support plant parasitism were probably obtained from bacteria via HGT [31,32]. These cellulase genes show a high degree of homology to genes from a wide range of bacterial species. The exact source of the transferred DNA is difficult to pinpoint because the relationship between the DNA donor and recipient has not been maintained. The case of A. viteae and O. flexuosa is special because the relationship with the DNA donor has not been maintained, but the donor can be easily identified by BLAST homology and by the presence of Wolbachia in most filarial species. Future studies will generate a comprehensive list of transferred fragments in several Wolbachia-free filarial species. This information may identify key genes and pathways that explain the fascinating and medically important symbiotic relationship between filarial worms and Wolbachia endobacteria.

Parasite material and DNA isolation
Adult O. flexuosa worms were collected from red deer (Cervus elaphus) in Germany (Schleswig-Holstein). Adult worms were dissected from nodules collected from freshly shot deer [13]. Adult A. viteae and B. malayi were obtained from experimentally infected Mongolian gerbils as previously described [33,34]. DNA was isolated from adult worms using the DNeasy Blood and Tissue Kit (Qiagen, Valencia, CA).

Library construction, sequencing and BLAST analysis
A fragment library was created from 15 mg of A. viteae DNA. Two runs on a Genome Sequencer FLX (454 Life Sciences/ Roche Diagnostics, Branford, CT) using standard FLX chemistry generated 768,909 reads and a total of 181 Mbp of sequence. Sff files were deposited in the NCBI short read archive (SRX001994). Newbler 1.1.03.24 assembled reads into 68,805 contigs containing ,31 Mbp of sequence using default parameters. The largest contig was 6.7 kbp and the average size was 0.9 kbp. Contigs were analyzed by BLASTN against the Genbank nucleotide collection (nt) and by BLASTX against the Genbank non-redundant protein database (nr) using NCBI BLASTALL to identify regions with a top hit to Wolbachia with an e-value less than 1610 25 . This cutoff was used for all BLAST searches. All A. viteae contigs larger than 600 bp were also split into smaller fragments (optimal maximum and minimum were set to 600 and 300 bp, respectively) using a custom Perl script and re-analyzed by BLASTN (nt_05292009) and BLASTX (nr_05292009) to ensure that all Wolbachia homologs were identified despite strong homology to different genes/proteins in other regions of the sequence. Contigs of interest were PCR amplified and cloned using the TOPO TA Cloning Kit for Sequencing or the TOPO-XL PCR Cloning Kit (Invitrogen, Carlsbad, CA, USA), and sequenced for verification. Sequences are posted on the Whole Genome Shotgun FTP site on Nematode.net (www.nematode.net/FTP/index.php).
O. flexuosa paired-end genomic DNA libraries were constructed as previously described [35] using 5.0 mg of DNA. One run on a Genome Sequencer FLX using standard FLX chemistry produced 516,745 reads containing 135 Mbp of sequence including adapter sequences. Sff files were deposited in the NCBI short read archive (SRX015550). A custom Perl script was designed for removal of the 44 bp adapter and separation of the paired end (PE) sequences which were then analyzed by BLASTX (nr_05292009) and BLASTN (nt_05292009). Complete, unsorted reads were also analyzed by BLASTX (nr_09182007) and BLASTN (nt_04062009) using WU-BLAST 2.0. 137 reads containing Wolbachia-like sequences were identified. Read sequences were used to amplify larger portions of the O. flexuosa genome. After sequencing the larger genomic fragments, redundant/overlapping fragments were collapsed using Contig Express (Invitrogen). This produced 42 large fragments (those amplified using primers matching to PE sequences) ranging in size from 1124 to 7725 bp with an average size of 302861386 bp and 50 smaller sequences (those representing assembled or unassembled portions of a PE read) ranging in size from 55 to 548 bp with an average size of 243683 bp. All fragments sequences are posted on the Data FTP page of Nematode.net.
3. Assessment of COG functional roles and mapping to the genome of the Wolbachia endosymbiont of B. malayi Wolbachia-like sequences were compared against the genome of the Wolbachia endosymbiont strain TRS of B. malayi (wBm) using BLASTN, and the locus tag of the top hit was recorded. The Wolbachia-like sequence was assigned to the same COG functional category as its wBm homolog as reported in NCBI Entrez. The Open Source Python (http://www.python.org/) library ReportLab (http://www.reportlab.com/software/opensource/rl-toolkit/) was used to generate a figure that marks the position of the homologous locus. Tick marks represent the midpoint of the coordinates for each locus tag as extracted from the GenBank data file for the wBm genome (NC_006833).

RNA isolation, cDNA synthesis and qRT-PCR
Adult worms were homogenized in 1 mL Trizol (Invitrogen), and RNA was isolated using organic extraction followed by column purification using an RNeasy Mini Kit (Qiagen) including the oncolumn DNase digest. A second DNase treatment was performed using Ambion's DNA-free DNase kit (Applied Biosystems, Austin, TX, USA). rRNA was depleted using the RiboMinus Eukaryote Kit for RNA-Seq (Invitrogen). Samples were tested for DNA contamination by conventional PCR using 35 cycles to ensure that no products were amplified. cDNA was made from 500 ng mRNA using qScript cDNA super mix (Quanta Biosciences, Gaithersburg, MD, USA). Conventional PCR assays using primers designed to span introns were used to test for DNA contamination again. Intron-spanning primer sequences are provided in Table S6. These target sequences from O. flexuosa with homology to hypothetical proteins in B. malayi that were readily amplifiable from O. flexuosa genomic and cDNA. Only Of_IC 1 worked well with A. viteae mRNA, which was also rigorously tested with wAv primers used in this study prior to cDNA synthesis (see Table S6). Specific Wolbachia sequences were detected in cDNA by SybrGreen qRT-PCR using PerfeC T a SYBR Green FastMix, ROX (Quanta Biosciences). Primer sequences are reported in Table S6. Reactions were done in duplicate to ensure accuracy, and all qRT-PCR experiments included a 10 ng DNA positive control, 1020.1 ng ten-fold dilutions of cDNA, and a 10 ng RNA negative control. Dissociation curves were examined to rule out non-specific amplification, and all products were assessed by agarose gel electrophoresis. Sequences were reported as expressed at the transcript level when signal for the 0.1 ng cDNA template crossed the cycle threshold at least 3 cycles before the mRNA negative control.

In situ hybridization
Adult B. malayi and A. viteae worms and O. flexuosa nodules were fixed for 24-72 h in DEPC-treated 4% buffered formaldehyde and embedded in paraffin using standard histological procedures. Two sequences were amplified from O. flexuosa cDNA and cloned into a dual promoter PCRII plasmid (Invitrogen). Primer sequences are reported in Table S6. After linearization of the plasmid, biotinylated antisense probes and sense negative controls were prepared with Ambion Megascript T7 and Sp6 high yield transcription kits (Applied Biosystems). Following DNase digestion (Roche, Indianapolis, IN, USA), probes were concentrated by ethanol precipitation, re-suspended in DEPC-treated water, and stored at 220uC. Sections (5 mm) were deparaffinized, digested with pepsin HCl for 7 min, and hybridized at 37uC overnight in a humid chamber with 1 mg of RNA probe in hybridization buffer (50% formamide, 5XSSC, 0.3 mg/ml yeast tRNA, 100 mg/ml heparin, 16 Denhardts Solution, 0.1% Tween 20, 0.1% CHAPS and 5 mM EDTA). A stringency wash was performed at 60uC for 30 min, and detection was performed using the 'In situ Hybridization Detection System' (K0601, DakoCytomation, Hamburg, Germany). Sections were incubated for 20 min with streptavidin-AP conjugate at room temperature. BCIP/NBT substrate solution was used for 10-30 min to visualize the RNA target. Sections were examined using an Olympus-BX40 microscope (Olympus, Tokyo, Japan) and photographed with an Olympus DP70 microscope digital camera.  13.8%. According to Student's t-test, this is significantly lower than the average percent identity to a nematode protein, 79.1 plus or minus 15.6% (p-value = .0001). The Student's t-test indicates that the average percent identity to a Wolbachia protein is also significantly lower than the percent identity of a sequence to a Wolbachia gene on the nucleotide level (p-value = 8e-10). Found at: doi:10.1371/journal.pone.0011029.s003 (0.13 MB DOC)  13.2%. According to Student's t-test, this is significantly lower than the average percent identity to a nematode protein, 76.3 plus or minus 16.0% (p-value = .00014). The Student's t-test indicates that the average percent identity to a Wolbachia protein is also significantly lower than the percent identity of a sequence to a Wolbachia gene on the nucleotide level (p-value = 2.03e-15).   Table describes the results of qRT-PCR studies used to determine whether Wolbachia homologs described in Tables S1 and S2 are expressed at the transcript level. + indicates that the sequenced is expressed at the RNA level while -indicates that it was not. n/a indicates that the sequence could not be tested using this method (for example, the sequence was too short or AT rich to design qRT-PCR primers). Sequences more than 25bp from the end of a genomic fragment are considered internal and sequences lacking premature stop codons and frameshift mutations are reported as potential open reading frames. Defined start and stop codons were not required for classification as a potential open reading frame.