Cluster M Mycobacteriophages Bongo, PegLeg, and Rey with Unusually Large Repertoires of tRNA Isotypes

Isolation, genome sequencing, and bioinformatic analysis.

Mycobacteriophage Rey was isolated by direct plating of a soil sample collected in Houston, TX, on lawns of M. smegmatis mc²155; Bongo and PegLeg were isolated by enrichment with M. smegmatis mc²155 from soil samples collected in St. Louis, MO, and Los Angeles, CA, respectively (Table 1). High-titer lysates were prepared and DNA extracted, and the genomes were sequenced using 454 pyrosequencing. Bongo, PegLeg, and Rey were sequenced at the Genome Institute of Washington University in St. Louis, the University of California, Los Angeles, Genotyping and Sequencing Core, and the University of Pittsburgh Genomics and Proteomics Core Laboratories, respectively. Genomes were assembled using the NewBler and Consed programs (22), and the assemblies were reviewed for completion at the Pittsburgh Bacteriophage Institute. Genomes were annotated using the DNA Master (http://cobamide2.bio.pitt.edu/), GLIMMER (23), GeneMark (24), BLAST, HHPred (25), and Phamerator (Mycobacteriophage_285 database) (26) programs. Dot plots were constructed using the Gepard program (27), and further bioinformatic analyses were performed using the DNA Master, FindTerm (Softberry), Splitstree (28), GLAM2, GLAM2SCAN, and MEME (29) programs. tRNAs were identified using the Aragorn (v1.2.36) program with default settings (30) or by the tRNAscan-SE (v1.23) program (31, 32) with default settings, except for bacterial source, relaxed scan modes, and s Cove score of >2. Following comparative analyses, revisions were made to the tRNA genes that included a deletion of mycobacteriophage Rey gene 113.

Lysogeny and immunity assays of cluster M phages.

Serial dilutions of high-titer lysates of Bongo and Rey were spotted onto lawns of mc²155, and turbid infected areas were picked and streaked for single colonies. Colonies were patched onto Middlebrook 7H10 plates and overlaid with soft agar containing mc²155. Putative lysogens were identified by clearing of the overlaid lawn—reflecting phage release—and grown in liquid 7H9 medium. Lysogeny was confirmed by immunity and phage release.

Electron microscopy of cluster M phages.

High-titer lysates of cluster M phages were serially diluted and spotted on a Middlebrook 7H10 plate overlaid with soft agar and M. smegmatis mc²155; plates were incubated for 24 h at 37°C. Spots with confluent plaques were gently washed with 10 μl of phage buffer several times by pipetting up and down to recover phage particles. The liquid phage sample was then diluted 1:2 with phage buffer, and 5 μl was placed on carbon-Formvar-coated glow-discharged copper grids. The sample was allowed to stand for 30 s, rinsed with double-distilled H₂O, stained with 1% uranyl acetate, and visualized with an FEI Morgagni transmission electron microscope (TEM).

Mass spectrometry of Rey virions.

A high-titer lysate of Rey (10¹⁰ PFU/ml) was precipitated overnight with 10% polyethylene glycol 8000, collected by centrifugation (5,500 × g, 10 min), and resuspended in phage buffer for 30 min. The suspension was mixed with CsCl (∼0.85 g/ml) and sealed in a heat-sealed ultracentrifuge tube (Beckman). The sample was spun for 16 h at 85,000 × g in a Ti70.1 rotor, and the visible phage band was collected and dialyzed against phage buffer. Phage particles were pelleted in a microcentrifuge by spinning at top speed for 30 min. The phage pellet was resuspended in 75 μl of 20 mM dithiothreitol, 2 μl of 0.5 M EDTA was added, and the particles were heated to 70°C. The sample was chilled on ice for 10 min, sonicated twice for 10 s, mixed with SDS sample buffer, and loaded onto a 12% SDS-polyacrylamide gel. The proteins were electrophoresed until they just barely entered the resolving portion of the gel, such that all proteins were contained within a narrow region. The gel was stained with Coomassie blue, the protein band was excised, and the proteins were subjected to in-gel trypsin digestion. Tryptic fragments were eluted from the gel, run on a reverse-phase high-pressure liquid chromatography (HPLC) column, and subjected to electrospray and tandem mass spectrometry (MS/MS) in an LTQ ion trap MS. Mass spectra were analyzed using the Scaffold (v4) program.

Nucleotide sequence accession numbers.

Revised genome annotations were submitted to GenBank (accession numbers JF937105, JN699628, and KC900379).

Bongo, PegLeg, and Rey constitute members of mycobacteriophage cluster M.

Mycobacteriophages Bongo, PegLeg, and Rey were isolated in St. Louis, MO; Los Angeles, CA; and Houston, TX, respectively (6). All three have siphoviral morphologies (Fig. 1). All three cluster M phages make turbid plaques on M. smegmatis mc²155, and we recovered lysogens from Bongo and Rey infections. These lysogens were tested for immunity to infection by all three cluster M phages, as well as to Wildcat and BPs, a member of cluster G (Fig. 2). Bongo and Rey are likely part of the same immunity group but do not confer immunity to either BPs or Wildcat.

The three cluster M genomes vary in length from 80,228 bp to 83,724 bp (Table 1), lengths that are somewhat longer than the average mycobacteriophage genome length of 70 kbp (4). Dot plot analysis shows that they share substantial DNA sequence similarity spanning greater than 50% of their genome lengths (Fig. 3A), and they share high levels of average nucleotide sequence identity (Fig. 3B). A gene content network analysis clearly shows the similarity between Bongo, PegLeg, and Rey (Fig. 4), and alignment of the genome maps indicates closely related architectures (Fig. 5). Taken together, phages Bongo, PegLeg, and Rey are clearly more closely related to each other than to other mycobacteriophages, and we propose that these constitute cluster M. Accession numbers are shown in Table 1.

Several features of the singleton phage Wildcat are shared with the cluster M phages (e.g., tRNA arrays, gene content, and genome architecture; see below), and dot plot analysis showed weak but identifiable sequence similarity (Fig. 3). However, alignment of any of the cluster M phages with Wildcat using Align Two sequences with BLASTN showed that the similarity spanned less than 50% of their genome lengths, and we conclude that Wildcat is insufficiently closely related to Bongo, PegLeg, and Rey to warrant inclusion in cluster M.

Although the three cluster M genomes are clearly related, Bongo and PegLeg are considerably more closely related to each other than either is to Rey, as illustrated by the branching pattern seen in the Splits Tree representation of their gene content relationships (Fig. 4), by dot plot analysis (Fig. 3A), and by the differences in the average nucleotide sequence identity values (Fig. 3B). We therefore propose that cluster M is divided into two subclusters, with Bongo and PegLeg forming subcluster M1 and Rey representing subcluster M2 (Table 1). Alignment of the genome maps further supports this division (Fig. 5).

Unusual genome architecture of cluster M phages.

There are several noncanonical features of the cluster M phages that distinguish them from other mycobacteriophage genomes that have been described. First, although the organization of the virion structure and assembly genes is typical of other phages with siphoviral morphologies—ordered as terminase, portal, protease, scaffolding, major capsid subunit, head-tail connector proteins, major tail subunit, tail assembly chaperones, tapemeasure protein, and minor tail proteins—the terminase is separated from the left cohesive end by 4 to 5 kbp DNA containing 10 to 15 small, leftward-transcribed genes (Fig. 6 to 8). More than 30% of the Rey genes are different from those in Bongo and PegLeg, and most of the genes are of unknown function (Fig. 5). The exceptions are the gp3 genes of Bongo, PegLeg, and Rey that encode Lsr2-like proteins, and although Lsr2 proteins have been described in mycobacteriophages within clusters E, J, and L, they are typically located within right-arm genes, and their specific roles are ill defined. In Mycobacterium spp., Lsr2 proteins are small DNA-bridging proteins that may play a role in DNA compaction. The lysis cassette is located immediately downstream of the virion structural genes, a common location for all the mycobacteriophages except those in cluster A (Fig. 5).

The integration cassette of cluster M phages is in an unusual location, positioned ∼12 kbp (15% of the genome length) from the genome right ends (Fig. 5 to 8). Among the mycobacteriophages, the integration cassette is typically located near the center of the genome, irrespective of genome length. The greatest departure from the genome center reported previously is in phage Brujita, in which the integrase is positioned 39% of the genome length from the left end, although we note that there are other examples of nonmycobacteriophages where the integrase is not centrally located, including Streptomyces phages ϕC31 (33) and R4 and their respective relatives (34). Although many of the genes surrounding the integration cassette in the cluster M phages are of unknown function, several genes involved in DNA and RNA metabolism can be identified, and there is an unusually large cluster of tRNA genes. Each of these features is discussed in further detail below.

Virion structure and assembly genes.

The 22 virion structure and assembly genes are tightly linked in a rightward-transcribed operon in the left arms of the genomes. Although some mycobacteriophage genomes—especially those in cluster J (5)—are replete with insertions of nonstructural genes throughout the structural operon, this is not observed in the cluster M phages. The sole possible exception is the first rightward-transcribed gene (e.g., Bongo 12; Fig. 6); there are few bioinformatic clues as to its function or whether it is expressed, but it is not uncommon to find HNH genes in similar locations in other genomes. A striking feature of the operon is the exceptionally long tapemeasure gene (tmp; 6.8 kbp), the longest of any in mycobacteriophages described, reflecting the long tails of the cluster M phages (Fig. 1). The correlation between tmp gene length and tail length (one codon/1.5 Å) reflects that reported for other siphoviral phages (13). Interestingly, the Tmp protein of the subcluster M2 phage Rey has substantially diverged from the Tmp proteins of Bongo and PegLeg and shares only 47% amino acid identity, in contrast to the two tail genes to its right, which share 89% and 86% amino acid identities, respectively, with those of the subcluster M1 phages. However, all three have features that are common to tapemeasure proteins, with a high proportion of glycine plus alanine residues (>20%), strongly predicted coiled-coil domains near their N termini, and a predicted peptidoglycan-hydrolyzing motif in the C-terminal half of the protein. This transglycosylase motif belongs to Tmp motif 1 described previously (13), although it is a relative far distant from those in other mycobacteriophages.

The cluster M major capsid subunits are predicted by HHPred to contain an HK97-like fold and, in addition to being related to the Wildcat capsid subunit (47% amino acid identity), have sequence similarity to the cluster L phages, such as LeBron and Rumpelstiltskin (37% amino acid identity). Like Wildcat and the cluster L phages (4, 14), the cluster M major capsid subunit and major tail proteins contain predicted Ig-like domains in their C-terminal ∼100 residues, and they are weakly related to each other (∼30% amino acid identity), a relationship also seen in Bxb1 and related phages (35). The Ig-like protein domains are widespread in phage proteins and have been proposed to play accessory roles in phage infection by weakly interacting with carbohydrates on the bacterial cell surface and by allowing the phages to bind to glycans from mucosal membranes of metazoans, allowing the phages access to bacterial inhabitants of those layers (36, 37).

Following the tapemeasure protein gene are eight predicted minor tail proteins (Fig. 6 to 8). The first two of these (e.g., Bongo gene 27 and Bongo gene 28) are closely related in all three cluster M phages (>80% amino acid identity), but the others are more distantly related between the subcluster M1 and M2 phages. For example, Rey gp35 and gp36 have little recognizable sequence similarity to their counterparts in Bongo (gp31 and gp32, respectively). Rey gp35 is glycine rich (>16%)—a not uncommon feature of phage tail fiber proteins—and is more closely related to cluster L tail proteins (e.g., LeBron gp21; 27% amino acid identity) than to the cluster M1 phages. The minor tail protein gene cluster extends through to Bongo gene 34, PegLeg gene 34, and Rey gene 39, and the last of these is related to the gene for Marvin gp57, which has been shown to be a virion component (16). In addition, Rey gp39 was identified as a virion component by mass spectrometry (see below; Table 2).

The protein composition of Rey was analyzed using tandem mass spectrometry. Rey virions were purified by CsCl density gradient ultracentrifugation and then electrophoresed into a single band in an SDS-polyacrylamide gel. The particle proteins from the gel slice were digested with trypsin, and the resulting tryptic peptides were concentrated using reverse-phase HPLC, followed by loading into an ion-trap mass spectrometer. Peptides were further fragmented and analyzed for the mass and charge of the composite pieces. Spectra were compared to those in a database generated from the coding sequences of the Rey genome using the program Scaffold (v4) (Table 2).

The most abundant Rey virion peptides correspond to the major capsid protein and the major tail subunit, and other expected virion proteins were identified, including the tapemeasure (gp30), portal (gp19), minor tail component (gp31, gp39), and head-to-tail connector (gp23, gp24, and gp26) proteins. However, several predicted virion proteins were not identified with high confidence, including minor tail proteins gp32, gp33, gp34, gp35, gp36, gp37, and gp38. Peptides for most of these were identified in the mass spectrometry data, but at lower confidence levels, perhaps in part reflecting a lower abundance of these proteins in the particles or a lack of suitable trypsin cleavage sites. A peptide from the large terminase subunit was also detected, a finding which was not expected and could indicate contamination of the sample with incompletely packaged particles. Two independent mass spectrometry runs generated similar profiles. Surprisingly, three proteins encoded by right-arm genes were identified as virion components with high confidence, an Erf-like recombinase (gp61), a single-stranded DNA (ssDNA) binding protein (SSB; gp73), and a ClpP-like protease (gp65) (Table 2). While we cannot rule out the possibility that these are simply contaminants of the initial phage sample, examination of the CsCl-banded phage sample by TEM showed no recognizable host or viral subparticles other than phage virions. The prevalence of the Erf peptides in the mass spectrometry data strongly suggests that the Erf protein in particular is indeed virion associated.

Lysis cassette.

Immediately downstream of the virion structural protein genes is the lysis cassette, which includes both lysin A and lysin B separated by two small genes. We propose that the second of these smaller genes (Bongo gp37, PegLeg gp37, and Rey gp42) is the holin, with each having two predicted transmembrane domains, although they are not related at the sequence level to any other mycobacteriophage proteins. The first small protein (e.g., Bongo gp36, 78 amino acids) contains an N-terminal predicted transmembrane domain, and its role is unclear, although it could perform a chaperone-like role in lysis, as proposed for gp1 of phage Ms6 (38). The lysin A proteins have domain structures corresponding to organization A (Org-A) (39), although the only other similar mycobacteriophage lysin is that encoded by the singleton phage Patience (gp41; 36% amino acid identity). The cluster M lysin B proteins are also distantly related to other lysin B proteins, with the closest non-cluster M relative being DS6A (42% amino acid identity) (6). The last gene in the rightward operon is a predicted NrdH-like glutaredoxin, and although related proteins are quite common in the mycobacteriophages, they are not typically found in this location (with the exception of the cluster J phages). However, it is unlikely that they play a role in cluster M phage lysis.

Immunity functions and the integration cassette.

Although the cluster M phages are temperate and form stable lysogens, the genes responsible for repressor function cannot be easily identified. In other phages, it is common for the repressor to be divergently transcribed from an adjacent Cro-like protein (40), and this is seen in mycobacteriophages such as BPs (18) and Giles (41), but not in others, such as L5 and its relatives (14, 42). In the cluster M phages, there are only two regions with divergent transcription (e.g., Bongo genes 11 and 12 and genes 49 and 50; Fig. 6), but they contain no proteins with predicted DNA binding activity. However, we note that there are examples of other mycobacteriophage repressors that also do not contain predicted DNA binding motifs (41). In some phages, the repressor genes are located near the integration functions (18), but none of the cluster M genes near the integrase has repressor-like features.

As noted above, each of the cluster M phages encodes an integrase of the serine recombinase family. There are many other similar mycobacteriophage integrases, although they are almost exclusively found in cluster A phages. One exception is that encoded by the cluster K phage Larva (6) and is the most closely related to the cluster M integrases (28% amino acid identity). Attachment sites for serine integrases cannot be readily predicted because of the lack of extended sequence similarity between attP and attB sites (43), although regions of imperfectly conserved inverted symmetry immediately to the right of the integrase genes are candidates for providing attP function. The intergenic regions to the right of the int genes are only moderately conserved between Rey and the subcluster M1 phages, raising the question as to whether they use the same attB site. Predicting the location of a putative recombination directionality factor (RDF) is complicated by the variety of different proteins that can perform this function (44–46). However, we note that located near the integrases are genes encoding phosphoesterases (Bongo gp125, PegLeg gp129, Rey gp144), reminiscent of the RDF encoded by the gene for Bxb1 (gp47) (45), which is also predicted to function as a phosphoesterase. However, the phosphoesterase activity of Bxb1 gp47 is not required for its RDF function (47), and as Bxb1 gp47 and the cluster M phosphoesterases are unrelated at the sequence level, an RDF function cannot be confidently predicted.

Other nonstructural genes.

In the rightmost ∼48 kbp of the cluster M genomes, the genes appear to be organized into three large groups: a leftward-transcribed set of 10 to 12 genes (Fig. 6 to 8) immediately adjacent to the virion genes, a rightward-transcribed group of ∼80 genes that includes many tRNA genes (discussed below), and a leftward-transcribed set of ∼25 genes at the right ends of the genomes. The functions of less than 25% of these genes can be predicted using BLASTP and HHPred searches. Many of these are associated with nucleic acid metabolism and include DNA helicases, a DnaQ-like exonuclease, an ssDNA binding protein, an Erf-like recombinase, Holliday junction resolvase, nucleotide binding proteins, the putative RNA ligase component RtcB (48), and HNH endonucleases. All three genomes also encode proteins containing putative protein tyrosine phosphatase catalytic domain (PTPc) kinase/phosphatase domains, although the genes are located differently in the three genomes (Bongo gene 83, PegLeg gene 82, and Rey gene 135). Additional predicted functions include a Band-7 protein, MazG, NrdC, NrdG, a ClpP-like protease, an oxidoreductase, and a WhiB-like regulator. Gene 152 in Rey has no counterpart in the subcluster M1 phages but is predicted to have a ParB domain-like fold and may play a regulatory role. The specific roles of these genes and their products are not known. However, as noted above, mass spectrometry suggests that the Erf protein and perhaps the SSB protein are virion associated. Inclusion of these proteins in capsids might make sense if they are required for recircularization of terminally redundant genome ends like Erf is in P22 (49), but genomic analysis shows that these phages have cohesive termini with 11-bp ssDNA extensions at the 5′ end (Table 1), and such an involvement of Erf is unexpected.

Transcription signals.

Mycobacteriophage promoters are ill defined, and although some phages—including L5 and BPs and their relatives—use SigA-like promoters (18, 50), others apparently do not (41). A search for SigA-like promoters yields few strong candidates in any of the cluster M genes that are consistent with the gene organization, although we cannot exclude the possibility that regions with weaker matches for the SigA consensus could be functionally important. Nonetheless, we predict that there are likely to be at least five promoters for expression of the five apparent operons. Factor-independent transcription terminators can be confidently predicted, and we have identified six in Bongo and PegLeg (Fig. 6 and 7) and eight in Rey (Fig. 8). The locations in common to all three phages are at the extreme left ends of the genomes following the leftward-transcribed operon, at the end of the rightward-transcribed operons containing the structural genes and lysis cassette, immediately before and immediately following the groups of tRNA genes, and between int and the upstream gene (Fig. 6 to 8). Bongo and PegLeg also have a terminator immediately downstream of int (Fig. 6 and 7), and Rey has one following the whiB gene (gene 79) and two within the leftward-transcribed genes at the right end of the genome (Fig. 8). Identification of the terminator upstream of the tRNA genes suggests that there is likely to be an additional promoter located between the terminator and tRNA genes.

Conserved repeated sequences.

There are at least four short conserved repeated sequences in the cluster M phages that could play regulatory roles. The first—which we refer to as conserved repeat 1 (CR1)—is the 17-bp motif 5′-CTGACCTGCGATTACAG (Fig. 9A). In Bongo and PegLeg there are five identical copies of this asymmetric sequence (CR1-1 to CR1-5), four of which are oriented in one direction and one of which is oriented in the other. Four related sequences are present in Rey, although one shares the identical sequence, and the other three each have a different 1-base departure. The consensus could plausibly be extended another 2 bases at the 3′ end (Fig. 9A). In general, the CR1 repeats are located in analogous positions in Bongo, PegLeg, and Rey (Fig. 6 to 8). An attractive plausible role of these sequences is in the regulation of transcription, for the following reasons. First, in each instance the orientation of these asymmetric sequences corresponds with the direction of transcription of the genes immediately downstream. Second, they are predominantly intergenic, although in both CR1-2 and CR1-4 the termination codon of the upstream gene lies within the conserved sequence (Fig. 9A). Third, two of them (CR1-3 and CR1-5) are positioned immediately downstream of a putative terminator, where they are positioned potentially to initiate transcription of the downstream genes. Whether these sequences function directly in positioning transcription initiation or act as sites for positive or negative regulation of transcription is unclear, but we note that some of the sites are adjacent to plausible weak promoters; however, for two of the repeats (CR1-2 and CR1-4), there is little additional space to accommodate a promoter between the adjacent genes. Although the length of the repeat is similar to that for promoters of phage T7, there is no evidence for a phage-encoded RNA polymerase, and detailed transcriptome analysis is needed to elucidate their roles.

A second set of conserved sequences contains a 12-bp sequence (5′-GTGATTACAGGA) that is closely related to the 3′ portion of CR1 (Fig. 9B). However, it differs from CR1, in that position 2 of CR2 is a nearly invariant T residue (with the exception of Rey CR2-4), and in CR1 this is a nearly invariant C base (with the exception of CR1-5 in Rey). In addition, the penultimate 3′ base of CR2 is an invariant G base, which corresponds to the variable position (purine) in CR1 (Fig. 9A). Like CR1, CR2 is predominantly intergenic (the only exception being CR2-9 in Rey, which is within the coding sequence of gene 85). However, the correspondence with the orientation of transcription of the surrounding genes is inverted to that of CR1 and its surrounding genes. For example, in Bongo, CR2-4, CR2-5, and CR2-6 all contain the motif on the bottom strand, but the genes around CR2-4, CR2-5, and CR2-6 are all transcribed rightward; CR1-1, CR1-2, CR1-3, and CR1-4 are also among the rightward-transcribed genes, but the conserved motif is on the top strand (Fig. 6 and 9C). The role of CR2 and its relationship to CR1 are unclear. However, we note that the motif is not common in other mycobacteriophage genomes.

A third set of conserved repeats may play a role in regulation of translation initiation. In Bongo and PegLeg, there are six identical copies of the 16-bp asymmetric sequence 5′-ACACACGGAGAAGGGA (Fig. 10A), all located within the leftward-transcribed operon at the right end of the genome (Fig. 6 and 7). We refer to these as conserved repeat 3 (CR3). In Rey, there are three identical copies of the repeat, another five with a single-base mismatch, and one with a 2-base mismatch (there are no other sites with two mismatches or less in the genome; Fig. 10A); all are located in the leftward-transcribed operon at the right genome end (Fig. 9). These repeats are different but reminiscent of the start-associated sequences (SASs) reported previously in the cluster K mycobacteriophages (19) and have the following features. First, they are all located immediately upstream of putative translation initiation codons, spaced 6 to 8 bp from an ATG start (Fig. 10A). Second, they contain a 5′-AGAAGGGA motif that corresponds to a strong ribosome binding site and is perfectly complementary to the 8 bases at the extreme 3′ end of the 16S rRNA (Fig. 10A). Nonetheless, the conserved sequences extend beyond the 3′ end of the rRNA by an additional 8 bases, and it is unlikely that the sequence is simply conserved for ribosome binding through interactions with rRNA. The observation that these are constrained to a small region of the genome is consistent with them playing a regulatory role, perhaps through a phage-encoded function that modulates translation initiation at these genes. The finding that both cluster M and cluster K phages share these features—albeit with different conserved sequences—suggests that this may be a more widespread regulatory system than previously recognized.

A fourth set of conserved repeats (CR4) is also predicted to play a role in translation initiation (Fig. 10B). In Bongo and PegLeg, there are nine occurrences of the sequence 5′-AGAAGGGAA, and there are six in Rey. Like CR3, CR4 is positioned upstream of start codons (Fig. 10B), is 9 bp long, and is a partial subset of CR3, containing the rightmost 8 bp of CR3 but extended for 1 additional base (Fig. 10B). The 8 bases in common with CR3 are perfectly complementary to the 3′ end of the 16S rRNA, but the additional invariant A residue at the extreme right end is not (Fig. 10B). While the conservation of this 1 base does not provide compelling evidence for a regulatory system in addition to simply promoting ribosome binding for translation initiation, we note that CR4 is located primarily within the leftward-transcribed operon at the genome right end, with the one exception being upstream of Bongo gene 40. We also note that CR4 is distinct from CR3, in that the spacing from the start codon is more variable (3 to 9 bp) and the start codon is not always ATG, as seen with CR3 (Fig. 10).

In the cluster K phages, a subset of the genes with start-associated sequences is accompanied by an extended start-associated sequence (ESAS) characterized by a conserved inverted repeat (19). Its role is unclear, and the lack of potential G·U base pairs precludes a clear indication of whether it represents a DNA recognition site for a regulatory protein or plays a role in RNA folding. We thus examined the cluster M phages for similar features upstream of CR3 and CR4. Although an ESAS as such was not observed, all six of the CR3 motifs in Bongo are associated with an upstream inverted repeat located within the intergenic space 20 to 40 bp upstream of CR3 (Fig. 11). Six of the CR4 repeats are associated with a similar inverted repeat, and the three that do not (CR4-1, CR4-2, and CR4-3) have only short intergenic upstream spaces. The inverted repeats have the potential to form 12- to 16-bp paired stems in RNA, accommodating G·U pairing; most form perfectly paired stems, and only three have any mispaired bases in the stem (Fig. 11). None of these repeats contains a run of thymines to their 3′ sides and thus are not predicted to function as transcriptional terminators. We propose that they may play a role in conjunction with the CR3 and CR4 motifs in regulation of the downstream genes, but the mechanism is unclear.

Clusters of tRNA genes in Bongo, Rey, and PegLeg.

The cluster M phages have a set of 21 to 24 predicted tRNA genes together with a transfer-messenger RNA (tmRNA) gene located within a 5- to 7-kbp span and interspersed with small open reading frames (Fig. 12A; Tables 3 to 5). Most of these tRNAs are predicted with a reasonably high degree of confidence, identified by both the Aragorn and tRNAscan-SE programs with a Cove score of >20, although 4 to 6 are predicted with lower confidence because they either were not reported by Aragorn or have lower tRNAscan-SE Cove scores (Tables 3 to 5). Although these could be false-positive predictions, we have included them because phage tRNAs may have noncanonical structures or atypical modifications to escape host-mediated destruction (51). This is illustrated by comparison of the cluster M phage tRNAs with those of M. smegmatis mc²155. Using tRNAscan-SE and including only those tRNAs with Cove scores above a threshold value of 20, the 47 predicted M. smegmatis tRNAs have a mean Cove score of 74.30. In contrast, the Bongo, PegLeg, and Rey tRNAs have mean Cove scores of 50.25, 47.91, and 45.70, respectively. The lower overall Cove scores for phage tRNAs may be a reflection of the lower selective pressure on the phages compared to that on the hosts, a phenomenon that has been noted with respect to conserved residues of phage/host protein homologs. It remains to be determined if all the phage-encoded tRNAs are functional, but atypical tRNA structures appear to be common for phage-encoded tRNAs. Some of the tRNAs have noncanonical base positions or mismatches in stem pairings (Tables 3 to 5), but a comparison of the three genomes suggests that many of these are indeed functional (Fig. 12). For example, a tRNA^Tyr(GUA) is conserved in all three cluster M genomes at syntenically related positions, even though these tRNAs are quite divergent at the nucleotide sequence level (Fig. 12B). Each has different noncanonical bases or base pairings, but the conservation of isotype and gene order suggests that they are nonetheless functional. Similar patterns are seen with other predicted tRNA genes (Fig. 12B; Tables 3 to 5).

The tRNA locus is the region of the greatest divergence between Bongo and PegLeg, and both are only weakly related to Rey in this region (Fig. 12). PegLeg differs largely due to four HNH genes, three of which are absent from Bongo and Rey. Interestingly, there are five tRNAs [tRNA^Tyr(GUA), tRNA^Met(CAU), tRNA^Gly(TCC), tRNA^Val(UAC), and tRNA^Glu(CUC)] where the PegLeg sequences have diverged greatly from those in Bongo (<35% identity)—although the isotypes are retained—and have likely been acquired by horizontal genetic exchange; in addition, PegLeg has a tRNA^Gln(UUG) that Bongo lacks. Four of these [tRNA^Tyr(GUA), tRNA^Met(CAU), tRNA^Glu(CUC), and tRNA^Gln(UUG)] are immediately adjacent to an HNH gene, and the other two are nearby (Fig. 12). These HNH endonucleases are thus directly implicated in generating diversity within the tRNA repertoire, perhaps facilitating rapid adaptation to growth in new bacterial hosts that have poorly matched host tRNA abundances with the phage requirements. This is reminiscent of the role of SegB in tRNA inheritance in phage T4 (52). We note than in spite of the low overall sequence similarity between Rey and the subcluster M1 phages in this region, 17 of the tRNA isotypes are present and in syntenically conserved positions. Although Rey lacks the M1 phages' tRNA^Ser(CGA), tRNA^Ser(GCU), and tRNA^Leu(CAG), it has acquired tRNA^Ser(GGA), tRNA^Ala(CGC), tRNA^Leu(UAG), and tRNA^Leu(CAA) (Fig. 12C).

Interestingly, each genome encodes tRNAs for all 20 isotypes, with the exception of PegLeg, for which no tRNA^Cys is predicted (Fig. 13). However, this distribution is somewhat odd, as there appears to be less than a complete coding set of tRNAs. Even if wobble paring, superwobble pairing (53), and both canonical and atypical tRNA modifications facilitate the reading of each two-box, three-box, and four-box codon set, at least three additional tRNAs would be required per genome to read the six-box codon sets (Leu, Arg; Ser; Fig. 13). Both Bongo and PegLeg encode two serine isotype tRNAs [tRNA^Ser(CGA) and tRNA^Ser(GCU)], but they do not encode tRNA for the Leu and Arg codons AGR and UUR, respectively (Fig. 13). Rey differs in that it lacks the tRNA^Ser(GCU), but it also encodes a tRNA^Leu(CAA) (Fig. 13). We also note that even decoding of all two-box, three-box, and four-box codon sets would require unusual tRNA modifications. For example, the Bongo and PegLeg tRNA^Gly(UCC) and tRNA^Ala(UGC) would not typically be expected to read all four glycine and alanine codons, respectively, and there is no tRNA for decoding the isoleucine AUA codon [Rey has a tRNA^Ala(CGC)].

Comparison of the normalized synonymous codon usage frequencies of the cluster M phages and M. smegmatis shows similar codon usage preferences (Fig. 14). However, there is a notable difference in codon usage between the rightward-transcribed operon containing virion structure and assembly genes (e.g., Bongo genes 12 to 39; Fig. 13) and the rest of the genome (Fig. 13). For example, although this operon contains 42% of the total codons, in Bongo (Fig. 13) and PegLeg (Fig. 13), it is almost completely depleted of UUA (Leu), AUA (Ile), and the AGR (Arg) codons (in Bongo, the only AUA codon is found in gene 12, the first gene transcribed in the forward direction and of unknown function, and the two AGG codons are in an early and possibly untranslated part of gene 22 and in an untranslated part of gene 25, a reading frame expressed as the downstream part of the programmed translational frameshift). Thus, late gene expression may be almost entirely dependent on the phage-encoded tRNAs, and codons for which phage-encoded tRNAs are lacking are absent. In Rey, the codon distribution is somewhat different (Fig. 13), in that there are 14 AGG codons in the late operon, and either it presumably uses a host tRNA for these or there may be an unidentified noncanonical Rey-encoded tRNA for this function.

The tmRNAs encoded by Bongo and PegLeg are identical but differ substantially from the Rey tmRNA. The 5′ 133 nucleotides share 94% nucleotide sequence identity, and the 3′ 66 nucleotides are identical, but the central ∼200-bp segments are virtually unrelated (Fig. 12D). The degradation tag is encoded within the common 5′ region, and the sequence departures give rise to slightly different tags (AAFVDADYAVAA in Bongo and PegLeg and AAFVAADYAVGA in Rey), each of which differs from the host tmRNA-encoded tag (ADSNQRDYALAA). We note that all three cluster M phages also encode a Clp-like protease (e.g., Bongo gp58), which could be specifically involved in degrading stalled tmRNA-tagged proteins and releasing ribosomes for translation of other mRNAs.