Mycobacteriophage Marvin: a New Singleton Phage with an Unusual Genome Organization

Abstract

Mycobacteriophages represent a genetically diverse group of viruses that infect mycobacterial hosts. Although more than 80 genomes have been sequenced, these still poorly represent the likely diversity of the broader population of phages that can infect the host, Mycobacterium smegmatis mc²155. We describe here a newly discovered phage, Marvin, which is a singleton phage, having no previously identified close relatives. The 65,100-bp genome contains 107 predicted protein-coding genes arranged in a noncanonical genomic architecture in which a subset of the minor tail protein genes are displaced about 20 kbp from their typical location, situated among nonstructural genes anticipated to be expressed early in lytic growth. Marvin is not temperate, and stable lysogens cannot be recovered from infections, although the presence of a putative xis gene suggests that Marvin could be a relatively recent derivative of a temperate parent. The Marvin genome is replete with novel genes not present in other mycobacteriophage genomes, and although most are of unknown function, the presence of amidoligase and glutamine amidotransferase genes suggests intriguing possibilities for the interactions of Marvin with its mycobacterial hosts.

INTRODUCTION

Bacteriophages are the most numerous biological entities in the biosphere, with an estimated global population of 10³¹ phage particles (56). Bacteriophages appear to have emerged early in evolutionary history and may have been evolving for more than three billion years (28, 29). The population is not only vast and old, but also dynamic, with an estimated 10²³ phage infections per second on a global scale (53). It is, therefore, perhaps no great surprise that the limited genomic information to date reveals a highly diverse and complex population (5, 20, 24). This population is, however, dominated by viruses classified morphologically in the order Caudovirales, double-stranded DNA (dsDNA)-tailed phages whose genomes vary in size from 15 kbp to approximately 500 kbp (6, 20, 25).

Mycobacteriophages are a group of phages that infect mycobacterial hosts such as Mycobacterium tuberculosis and Mycobacterium smegmatis (19). To date, all characterized mycobacteriophages have either siphoviral or myoviral morphotypes (19). Currently, genomic characterization of 83 mycobacteriophages capable of infecting the nonpathogenic host M. smegmatis mc²155 has revealed a large degree of genetic diversity (19, 46, 47). When grouped by gross genomic nucleotide sequence comparisons, mycobacteriophages that infect the common host M. smegmatis mc²155 fall into 12 major groups (“clusters”) designated A to K (47); several of these clusters can be further divided into subclusters according to their gross nucleotide relationships (19, 47). Only the nine phages constituting cluster C have myoviral morphologies, and all of the others morphologically belong to the Siphoviridae. Five of the siphoviral mycobacteriophages (Giles, Corndog, Wildcat, Omega, and LeBron) were classified as singletons (47), although there have been recent findings of close relatives of Corndog, Omega, and LeBron (22).

The expanding collection of sequenced mycobacteriophage genomes continues to throw new light on mycobacteriophage diversity and the evolutionary processes that create these genomes (19, 21, 31, 47). For instance, the group of cluster A phages has increased significantly, representing a growing number of subclusters with information about superinfection immunity (47). Additionally, the presence of an A1 subcluster phage repressor gene in a cluster C phage, LRRHood, suggests that this gene has been recently acquired by LRRHood from a subcluster A1 phage (47). Mycobacteriophage genomes—like bacteriophages of other hosts—carry many genes that mediate their own mobility either within or between genomes, such as transposons (48), homing endonucleases (3, 21), and inteins (47, 54); although introns have been described in other phages (12), none have yet been identified in mycobacteriophages. Overall, the most striking observation to emerge from bacteriophage comparative genomics is that they are pervasively mosaic, with different segments of the genome—commonly containing just a single gene—having distinct evolutionary histories (20, 30).

The grouping of phages into clusters and subclusters is based on gross nucleotide sequence similarity and therefore reflects more recent evolutionary relationships. More distant relationships can be discerned by comparison of the predicted amino acid sequences of genes, and to facilitate this, a program, Phamerator, has been described that assorts genes sharing protein sequence similarity into “phamilies” (“phams”) (10). The 83 published genomes encode a total of 9,308 predicted genes, and these assemble into 2,367 phams, of which 1,120 (47.3%) are “orphams” (phams containing only a single gene member) (46, 47). Of these phams, about 80% have no significant database match to previously published sequences, and the functions of these large numbers of phage genes are unknown (46, 47). Notable exceptions to this are the virion structure and assembly genes that in the siphoviral phages are syntenically conserved, and thus gene location facilitates their functional assignments (19, 24).

Here, we report a novel siphoviral mycobacteriophage, Marvin, isolated and annotated by students at Cabrini College enrolled in an Honors Introductory Biology Laboratory course sponsored and funded by the Howard Hughes Medical Institute (HHMI) Science Education Alliance (SEA) program. Marvin is a new singleton mycobacteriophage with a 65,100-bp genome that is unrelated at the DNA level to any of the other 83 sequenced mycobacteriophages. Marvin has a mosaic genome, and over 70% of the genes have no homologues among known mycobacteriophages or other organisms. Of the 27 genes that are homologous to other mycobacteriophage genes, the matching genes are from genetically diverse mycobacteriophages, and the mosaic structure of the Marvin genome is clear. Surprisingly, a subset of the tail protein genes are displaced about 20 kbp away from their more typical location and are situated among the nonstructural genes in the right arm. The novelty of the Marvin genome and its large number of new genes support the hypothesis that in spite of the growing collection of mycobacteriophages, we are far from having a full understanding of this diverse population.

MATERIALS AND METHODS

Phage isolation and genomic DNA purification.

Mycobacteriophage Marvin was identified by direct plating on lawns of M. smegmatis mc²155 using an extract from soil on the campus of Cabrini College, located in Southeastern Pennsylvania. Phage isolation was accomplished by mixing approximately 1 g of a soil isolate with phage buffer (10 mM Tris-HCl [pH 7.5], 10 mM MgSO₄, 1 mM CaCl₂, 68.5 mM NaCl) for a 30-min incubation period at room temperature. The extract was then filtered through a 0.22-μm-pore filter, and 50 μl of this sample was plated with 0.5 ml of late-log-phase M. smegmatis mc²155 and 4.5 ml of 7H9 agar (Middlebrook 7H9 broth base; Difco Laboratories, Detroit, MI) supplemented with 1 mM CaCl₂. Following several rounds of plaque purification, a high-titer phage stock was prepared by treating 10 ml of a filtered phage crude lysate with RNase A and DNase I for 30 min at 37°C, followed by a 60-min incubation at room temperature. Intact particles were then precipitated with 30% polyethylene glycol 8000 (PEG 8000)–3.3 M NaCl overnight at 4°C and harvested by centrifugation at 10,000 × g for 20 min. DNA was extracted from the phage pellet using a Wizard DNA cleanup kit (Promega) as per the manufacturer's instructions. For other analyses, Marvin particles were purified by equilibrium-density CsCl centrifugation as described previously (23).

Digestion with DNA methylation-sensitive and DNA methylation-resistant enzymes.

One microgram of Marvin genomic DNA per reaction was digested overnight at 37°C with 1 U of restriction endonuclease. Products were separated by electrophoresis through a 1.2% agarose gel using Tris-acetate-EDTA buffer.

Phage genome sequencing and gene identification.

Purified phage genomic DNA was sequenced by the Joint Genome Institute (JGI) to a depth of ∼25-fold coverage using 454 sequencing and supplemented by an additional ∼60-fold coverage with SOLiD sequencing. Raw reads were assembled using 454's GS De Novo Assembler; assemblies were then quality controlled using Consed. Six Sanger reads were required to resolve weak areas in the assembly. Finished sequences were analyzed and annotated in genome editors, including DNAMaster (http://cobamide2.bio.pitt.edu), G Browse (52), Apollo (37), Glimmer (11), GeneMark (4), tRNA ScanSE (38), Aragorn (36), and Programmed Frameshift Finder (57) to identify genome features. Genes were assigned to phams, and genome maps and phamily circle diagrams were drawn using Phamerator, with the threshold parameters of 32.5% identity with ClustalW and a BlastP E value of 10⁻⁵⁰, as described previously (10).

Electron microscopy.

A lysate of Marvin with a titer of approximately 10¹⁰ PFU/ml was serially diluted into phage buffer to approximately 10⁴ PFU/ml, and 3 μl of each dilution was spotted onto a soft agar lawn seeded with M. smegmatis mc²155. After overnight incubation at 37°C, the spot that exhibited densely packed yet distinguishable plaques was gently washed with 10 μl of phage buffer by pipetting up and down several times. The 10 μl of buffer was diluted 1:2 in phage buffer, and 5 μl of that dilution was allowed to sit on freshly glow-discharged 400-mesh carbon-Formvar-coated copper grids for approximately 30 s. The grids were then rinsed with distilled water and stained with 1% uranyl acetate. Virus particles were imaged on an FEI Morgagni transmission electron microscope at 80 kV at a magnification of 56,000.

Identification of Marvin virion proteins.

Approximately 100 μl of CsCl-purifed Marvin particles (a total of 10¹² PFU) was collected by centrifugation at 14,000 rpm for 30 min, and the pellet was resuspended in 75 μl of 20 mM dithiothreitol. Two microliters of 0.5 M EDTA was added, and the solution was heated to 65°C for several minutes, when it became viscous. The sample was sonicated on ice for 10 s and allowed to rest on ice for 1 min, and this cycle was repeated six times, at which point, the viscosity was greatly reduced. Finally, 4× SDS sample buffer was added, and the sample was boiled for 2.5 min. Several dilutions were loaded onto a 12% SDS–polyacrylamide gel and electrophoresed at 100 V until the dye front ran off the gel. The gel was stained with Coomassie blue and destained in 10% acetic acid. The visible bands were compared to a standard to determine the approximate molecular mass.

For protein identification by mass spectrometry (MS), 8 μl of sonicated Marvin particles was loaded into a single lane of a different 12% SDS–polyacrylamide gel and electrophoresed only until the sample was approximately 2 cm into the separating portion of the gel. The gel was stained with Coomassie blue and destained in H₂O. The single visible band comprised of all particle proteins was excised, and the proteins were digested in situ with trypsin (at the University of Pittsburgh Genomics and Proteomics Core Labs), followed by peptide elution, chromatography, and tandem MS (MS/MS) on an LTQ Velos Orbitrap mass spectrometer. Peptides were matched against predicted Marvin proteins.

Analysis of the predicted secondary structure and coiled-coil propensity for selected protein sequences was carried out with the Psipred (http://bioinf.cs.ucl.ac.uk/psipred/) and Coils (http://www.ch.embnet.org/software/COILS_form.html) servers, respectively.

Nucleotide sequence accession number.

The GenBank accession number for mycobacteriophage Marvin is JF704100.

RESULTS

Phage isolation and morphological characteristics of mycobacteriophage Marvin.

Mycobacteriophage Marvin was isolated from soil on the campus of Cabrini College, Radnor, PA, by direct plating with M. smegmatis mc²155. Marvin is somewhat unusual among mycobacteriophages in that it propagates slowly and forms tiny barely identifiable plaques after 48 h of growth on a lawn of M. smegmatis at 37°C. The plaques are round and clear, suggesting that under standard growth conditions using M. smegmatis as the host, Marvin is either a lytic phage or a temperate phage that forms lysogens at only a low frequency.

To determine whether lysogens could be recovered from Marvin infections, cells from a spot where Marvin particles had infected a lawn of M. smegmatis were recovered and grown on solid media. Bacterial growth was observed, and two independent colonies were restreaked twice more and then patched onto M. smegmatis lawns to test for phage release; none of the colonies recovered showed phage release (data not shown). Thus, although bacterial survivors can be readily recovered, there is no evidence that Marvin is a temperate mycobacteriophage.

Electron microscopic images show that Marvin has a siphoviral morphotype with a long, flexible noncontractile tail and an isometric head (Fig. 1). The average tail length from several electron micrographic images of Marvin is 250 nm, and the head diameter is 58 nm. The tail length is longer than the average tail length of mycobacteriophages, but not as long as those of the cluster H phages Konstantine, Predator, and Barnyard (18a).

Fig 1.

Morphology of mycobacteriophage Marvin virions. An electron micrograph with uranyl acetate negative stain is shown. The scale bar corresponds to 100 nm.

Genome sequencing and classification.

Marvin DNA was isolated and sequenced using a combination of 454 shotgun and SOLiD sequencing. The Marvin dsDNA genome is 65,100 bp in length, with 11-nucleotide 3′-terminal extensions. This genome length is near the average for the siphoviral mycobacteriophages. The GC% of the Marvin genome is 63.4%, close to both the mycobacteriophage average and to that of the host, M. smegmatis. Comparison of the Marvin genome with examples of each of the mycobacteriophage clusters shows little or no discernible DNA sequence similarity to any of them (Fig. 2), and Marvin has therefore been designated a new singleton phage.

Fig 2.

Dot plot comparison of mycobacteriophage Marvin with representative mycobacteriophages. A sequence file containing the four singleton (sin) phages Corndog, Giles, Wildcat, and Marvin was compared against a file containing a single representative of each cluster or subcluster (as indicated) using Gepard (35). Marvin is classified as a singleton phage because of its lack of identifiable sequence similarity to other known mycobacteriophages. Omega is not shown as a singleton phage here because it has recently been grouped with unpublished phages as cluster J.

Analysis of the Marvin genome identified 107 putative open reading frames (ORFs), but no tRNA or other small RNA genes (Fig. 3 and Table 1). The ORF density is relatively high (92.85%), and there are only four noncoding intergenic gaps larger than 300 bp. Ninety-three of the ORFs are expressed from the top strand (shown rightwards in Fig. 3) spanning the leftmost 56 kbp of the genome. Twelve of the leftwards-transcribed ORFs (genes 92 to 103) are closely linked and situated about 10% of the genome length from the right end; the other two are interspersed with the rightwards-transcribed genes (Fig. 3). This overall organization is unlike any other mycobacteriophage genome (19), consistent with its assignment as a new singleton phage.

Fig 3.

Annotated genome map of mycobacteriophage Marvin. The viral 65,100-bp genome of Marvin is represented in four tiers with markers spaced at 1-kbp and 100-bp intervals. The predicted genes are shown as boxes either above or below the genome, depending on whether they are rightwards or leftwards transcribed, respectively. Gene numbers are shown within each box, and the phamily to which that gene belongs is shown above with the number of phamily members shown in parentheses; genes are color coordinated according to their phamily identity. Putative functions are shown above the genes. Other sequences, including putative promoters (P), a terminator (t), and a long palindromic sequence, are shown.

Table 1.

Mycobacteriophage Marvin predicted genes and gene products

Gene	F/Ra	Position		Product mass (kDa)	Pham no.b	Size (no. of members)c	Putative function
		Start	Stop
1	F	424	768	12.98	2844	1
2	F	802	1281	18.18	2845	1
3	F	1278	1538	9.47	2846	1
4	F	1511	1789	9.02	2847	1
5	F	1789	2058	10.08	2848	1
6	F	2048	2533	18.63	2849	1
7	F	2530	2967	15.74	1566	109
8	F	2964	3248	10.66	1566	109
9	F	3241	3600	13.04	1566	109
10	F	3597	3977	13.2	1566	109
11	F	3974	4192	8.03	1566	109
12	F	4189	4404	8.02	2850	1
13	F	4433	4720	10.91	2851	1
14	F	4843	5079	8.37	2852	1
15	F	5076	5366	10.49	2853	1
16	F	5359	5625	9.47	2854	1
17	F	5612	5806	7.1	2855	1
18	F	5935	6060	4.01	2856	1
19	F	6072	6209	5.12	2858	1
20	F	6250	6405	5.74	2416	5
21	F	6402	6551	5.38	2087	9
22	R	6781	6548	8.7	2859	1	Repressor?
23	F	6901	7233	11.75	2861	1
24	F	7217	7768	19.89	2862	1
25	F	7765	8046	11.04	2863	1
26	F	8018	8338	11.36	2864	1
27	F	8335	8568	9.46	2865	1
28	F	8630	8884	8.91	2866	1
29	F	8881	9276	15.34	2867	1
30	R	10398	9748	24.6	973	6	DNA methylase
31	F	10504	10857	13.43	2689	3	Terminase small subunit
32	F	10858	11115	9.22	2869	1
33	F	11120	12862	64.79	1469	18	Terminase large subunit
34	F	12859	14424	58.35	2870	1	Portal
35	F	14421	15065	24	2871	1	Protease
36	F	15195	15827	22	2872	1	Scaffold
37	F	15831	16235	14.49	2873	1	Virion protein
38	F	16247	17317	38.96	2874	1	Capsid
39	F	17422	17823	15.04	2875	1	Virion protein
40	F	17823	18194	14.21	2876	1	Virion protein
41	F	18194	18517	12.23	2877	1	Virion protein
42	F	18521	19027	19.54	2878	1	Virion protein
43	F	19043	19708	23.78	2879	1	Major tail subunit
44	F	19842	20414	20.93	2880	2	Tail assembly
45	F	19824	20911	40.42	2880	2	Tail assembly
46	F	20889	26438	195.74	2808	18	Tapemeasure
47	F	26438	27400	36.52	2881	1	Minor tail protein
48	F	27400	29112	63.28	2821	21	Minor tail protein
49	F	29144	29491	12.47	475	3	Minor tail protein?
50	F	29517	30194	25.41	2882	1	Minor tail protein?
51	F	30191	31570	50.63	2803	30	Lysin A
52	F	31567	31842	10.25	2883	1
53	F	31842	32996	42.52	2884	1	Lysin B
54	F	33016	33306	10.26	876	2	Holin
55	F	33713	34222	19.5	2885	1
56	F	34209	34547	12.07	2886	1
57	F	34557	35534	35.63	2918	1	Virion protein
58	F	35578	35871	10.87	2920	1	DNA-binding protein
59	F	35885	36265	13.89	2921	1
60	F	36210	36548	11.57	2922	1
61	F	36548	36874	12.5	2923	1
62	F	36874	37824	35.66	2924	1
63	F	37821	38873	37.75	2925	1
64	F	38873	39298	15.76	2926	1
65	F	39295	39648	13.6	2927	1	HNH protein
66	F	39924	40205	10.34	2928	1
67	F	40202	40567	13.6	2929	1
68	F	40564	40827	10.12	2930	1
69	F	40187	41308	18.4	236	15
70	F	41313	41516	8	2931	1
71	F	41650	42576	34.32	2932	1
72	F	42578	43198	22.92	2933	1
73	F	43195	43524	11.98	2935	1
74	F	43591	43734	5.11	2936	1
75	F	43731	44018	10.86	2937	1
76	F	44158	44355	7.5	1146	4
77	F	44352	44612	9.94	2938	1
78	F	44566	44754	7.48	2939	1
79	F	44750	45333	22.19	1835	2	Methyltransferase
80	F	45330	46217	33.77	458	4	Glycosyltransferase
81	F	46217	46945	27.77	1598	6	Glycosyltransferase
82	F	46955	47125	6.37	2940	1
83	F	47128	47790	24.71	87	22	Glycosyltransferase
84	F	47787	48341	20.71	2941	1
85	F	48363	48776	15.7	2943	1
86	F	48763	49089	12.62	2944	1	HNH protein
87	F	49116	53072	140.89	2822	86	Minor tail protein
88	F	53069	55204	72.95	2441	2	Minor tail protein
89	F	55227	55634	14.19	20	15	Minor tail protein
90	F	55647	55841	7.08	2603	2	Minor tail protein?
91	F	55838	56083	9.23	2945	1
92	R	56158	56018	5.27	2946	1
93	R	56575	56453	4.55	2948	1
94	R	57642	56572	40.15	2949	1
95	R	57970	57740	8.29	2950	1
96	R	58218	58015	7.55	2951	1
97	R	58417	58226	7.89	2952	1
98	R	58574	58404	6.46	2953	1
99	R	58798	58571	8.45	2954	1
100	R	59232	58801	15.68	2955	1
101	R	59393	59235	6.27	2956	1
102	R	59958	59536	15.4	2957	1
103	R	60379	60122	10.13	2958	1
104	F	61145	61567	15.37	2959	1
105	F	61626	62987	51.84	2960	1	Amidoligase
106	F	63067	63966	33.65	2961	1	Glutamine amidotransferase
107	F	64205	64453	9.04	2962	1

^aF/R, forward or reverse transcription.

^bPham number derived using the Phamerator database “Marvin.”

^cNumber of gene members of that pham.

Marvin genome architecture.

Each of the Marvin open reading frames was compared with all other mycobacteriophage genes using the program Phamerator (10) (using the database “Marvin”), and the ORFs were assorted into phamilies according to their amino acid sequence similarities. The “Marvin” Phamerator database contains 84 genomes, 9,415 genes, 2,446 phamilies, and 1,196 orphams. A striking outcome of this analysis is that 75 (70%) of the predicted Marvin protein coding genes are orphams, although this is not uncommon for a singleton phage for which there are no close relatives (19) (Fig. 3). Searching against the NCBI database revealed very few significant matches with any of these orphams, and only 19 of the predicted Marvin genes gave any informative matches (Table 1).

Although relatively few Marvin genes can be assigned putative functions, an overall architecture can be proposed. The virion structure and assembly genes likely span genes 33 to 50, deduced from the observations that the terminase genes are typically the leftmost of the operon, and the putative Marvin lysis genes lie to the right of gene 50 (Fig. 3). However, this segment spans only about 20 kbp, which would make this among the smallest of the virion structure and assembly operons of any of the mycobacteriophages. For example, although BPs and related cluster G phages have the smallest mycobacteriophage genomes (48), their 25 virion structural genes span more than 24 kbp of the genome. An explanation for this lies in the observation that Marvin's “missing” minor tail protein genes (87 to 90) are located elsewhere in the genome among nonstructural genes, displaced by more than 20 kbp from their typical position (see below). We also note that the terminase large subunit gene is separated from the physical end of the genome by more than 10.5 kbp. This is atypical but not unprecedented and is also seen in the cluster A phage genomes (14, 17, 23). However, in those examples, the lysis cassette also lies within this region, whereas in Marvin it is to the right of the structural operon (Fig. 3).

Temperate phages typically encode either a serine- or tyrosine-integrase that mediates prophage integration, and these genes are usually positioned near the center of their genomes (18). However, there are no Marvin ORFs with recognizable similarity to either type of integrase, and no apparent relatives of the ParAB functions that some mycobacteriophages use to stabilize extrachromosomally replicating prophages (47). This is consistent with the conclusion from the lysogen analysis described above that Marvin does not appear to be a temperate phage. The presence of putative transcriptional regulator genes and their potential roles are discussed below.

Nonstructural genes 1 to 30.

Marvin genes 1 to 30 occupy the space between the physical left end of the genome and the terminase genes. Twenty-two of these genes are orphams and have no close relatives in other mycobacteriophages; most also have no database matches, although protein gp2 (gene product 2′) has weak similarity (31% identity) to gp59 of Tsukamurella phage TPA2 (44). We note though that they are all small, and none is longer than 600 bp. Two of the 29 genes in this region have functionally informative database matches, and gene 22 encodes a 78-residue helix-turn-helix putative DNA-binding protein with similarity to putative repressors of the XRE class. Although some members of this family of proteins are predicted to be phage repressors, others are components of toxin-antitoxin systems. We note, for example, that Marvin gp22 shares 33% identity with the putative antitoxin component of Escherichia coli TA271. Because Marvin does not appear to form stable lysogens, gp22 seems unlikely to be a phage repressor, and an antitoxin component of a toxin-antitoxin system is an attractive role. It is possible that a closely linked gene, such as gene 23, encodes the toxin component, although gp23 has no close relatives. We note that toxin-antitoxin systems have been implemented in phage resistance mechanisms (13), and it is reasonable to expect these also to be carried by phage genomes. Indeed, the previously reported mycobacteriophage Fruitloop also encodes a putative toxin-antitoxin system (47).

Marvin gp30 matches known proteins, suggesting that it functions as a cytosine-C5-specific DNA methylase. Related proteins are found in other mycobacteriophages, including U2, DD5, Jasper, Lockley, and Pukovnik (all cluster A phages). The specific role of this protein is not known, although it could act to modify Marvin DNA nonspecifically, or alternatively act as a component of a restriction modification system. Because Marvin DNA is readily digested by several restriction enzymes that are typically inhibited by cytosine methylation (Fig. 4), we favor the second explanation, although we have not been successful in identifying a restriction enzyme partner in the Marvin genome.

Fig 4.

Restriction enzyme sensitivity of Marvin DNA. Marvin DNA was digested with the enzymes DpnI, Dpn II, HpaII, MspI, and McRBC as indicated, and the products were separated by agarose gel electrophoresis. Lane 2 contains undigested genomic DNA; Lane M is a 1-kbp size marker. Note that DpnI and Dpn II are isoschizomers (recognizing 5′-GATC), and DpnI only cuts DNA if the recognition site is methlylated, whereas DpnII is blocked by dam methylation. Likewise, HpaII and MspI are isochizomers (recognizing 5′-CCGG), and HpaII is blocked by CpG methylation, whereas MspI is insensitive to site methylation. McrBC (lane 7) that recognizes 5′-Pu^mC(N_40–3000)Pu^mC only cuts methylated DNA.

A striking feature of this region is the segment containing genes 7 to 11 (Fig. 5). These are all members of the same Pham (Pham1566), although distant relatives of each other. However, this is a large Pham, with 109 members in the current Phamerator database (database “Marvin”), and there are representatives in virtually every other mycobacteriophage cluster, the exceptions being clusters G, H, and K. Moreover, the sequence similarity extends to the nucleotide sequence level, with short but significant matches of similarity to many phages, including Che8 (subcluster F1) and SkiPole (subcluster A1) (Fig. 5A). For example, Marvin gene 8 has 95% nucleotide identity with Che8 gene 86, spanning a region of about 300 bp (Fig. 5A); 10 other mycobacteriophages contain genes with similar levels of sequence similarity. Within a genome, these related genes form short arrays, and in Che8, there are seven Pham1566 genes; however, the order of genes varies between genomes (Fig. 5). Although the genes within the array are related at the level of the protein sequences, there is little evidence of nucleotide sequence similarity between them (Fig. 5B), in sharp contrast to the intergenome relationships. Thus, while the arrays may have arisen through gene duplications, these must have been far distant evolutionary events, and individual members appear to have been exchanged between genomes during very recent evolutionary times. It is tempting to speculate that perhaps these represent novel mobile elements, although we have been unable to find any significant similarity to transposases or homing endonucleases using Psi-BLAST or HHPred.

Fig 5.

Marvin genes 7 to 11 and their homologues. Marvin genes 7 to 11 form a group of genes that are related to each other (Pham1566) and which are related to other similar groups in other mycobacteriophages. (A) Alignment of the Marvin genome with those of Che8 (subcluster F1) and SkiPole (subcluster A1) illustrates the nucleotide sequence similarities between the genomes. Genome representations are made in Phamerator, and the gene annotations are as described for Fig. 3. Segments of nucleotide sequence similarity are shown by colored regions between pairs of genomes and spectrum colored, with violet being the most similar and red the most dissimilar. (B) Although gp7 to gp11 are related at the amino acid sequence level, they are not related at the DNA sequence level as shown by a dot plot of genes 7 to 11 against themselves.

Within the region from genes 1 to 30, there are three plausible promoters, each of which contains a canonical −35 sequence (5′-TTGACA) of the σ⁷⁰ class of promoters; promoters of this class have previously been described in mycobacteriophage L5 (41). These are located between genes 1 and 2, between genes 13 and 14, and between genes 22 and 23 (Fig. 3). Between genes 13 and 14—but located upstream of the putative promoter—there is a putative rightwards-facing stem-loop transcription terminator (Fig. 3). The activity and role of these putative transcription signals remain unclear.

Marvin virion structural and assembly genes.

A putative operon of virion structure and assembly genes (31 to 50) shares the canonical organization and common synteny seen in phages with siphoviral morphologies: terminase, portal, protease, scaffold, capsid, head and tail completion proteins, major tail subunit, tail assembly chaperones, tapemeasure protein, and minor tail proteins (Fig. 3). The genes are generally tightly packed, with overlapping or minimal gaps between start and stop codons, with the exceptions of three larger gaps (100 to 130 bp) between genes 35 and 36, 38 and 39, and 43 and 44. The latter two contain putative transcriptional terminators that presumably modulate transcription levels throughout the operon (Fig. 3); there is little space also to accommodate promoters between these putative terminators and the downstream genes. The gene assignments within the operon correlate well with proteins present in intact virions, as determined by SDS-PAGE separation of virion proteins (Fig. 6) and identification of virion proteins by mass spectrometry (Table 2). These gene assignments are discussed in further detail below.

Fig 6.

SDS-PAGE analysis of Marvin virion proteins. Illustrated is SDS gel electrophoresis of Marvin virion proteins, showing the predicted gene products. Molecular mass markers (M) are shown in kDa.

Table 2.

Identification of virion-associated proteins

No. of PSMsa	Coverage (%)b	No. of peptidesc	Productd	Mol mass (kDa)	Scoree
197	71.91	19	gp38	39.0	767.70
101	53.39	8	gp43	23.8	474.64
120	40.29	56	gp46	195.6	470.89
81	43.57	19	gp34	58.3	270.10
52	73.85	17	gp57	35.6	200.46
63	64.06	13	gp47	36.5	183.93
41	55.70	19	gp88	72.9	175.92
29	44.21	16	gp48	63.3	116.50
18	15.86	14	gp87	140.9	60.93
16	66.42	9	gp37	14.5	56.33
4	33.08	3	gp39	15.0	16.13
5	34.29	5	gp36	22.0	15.74
5	27.98	4	gp42	19.5	15.36
5	19.11	4	gp50	25.4	15.29
3	20.09	3	gp35	24.0	10.87
2	19.51	2	gp40	14.2	9.98

^aPSMs, peptide spectrum matches.

^bPercentage of predicted protein sequence identified in peptides.

^cNumber of different peptides identified corresponding to the protein.

^dPredicted gene product of mycobacteriophage Marvin.

^eThe sum of the matching scores of individual peptides to the predicted sequence.

In Marvin, gene 33 encodes the terminase large subunit with relatives in other mycobacteriophages (Fig. 3), the closest being Bxz2 gp13 (32% identity). However, there are closer relatives in nonmycobacteriophage phage genomes, and the closest match is to the terminase of a putative prophage in Corynebacterium kroppenstedtii (47% identity). Curiously, Marvin gp33 has a short (32 residues) but significant (E value, 5 × 10⁻³) match to a conserved domain (pfam02459) in the adenoviral terminal protein, which is of interest given the related functionalities of these proteins. Marvin gp31 is a strong candidate for a terminase small subunit with homologues in the subcluster I1 phages (e.g., Brujita gp1 and Island3 gp1), where it is positioned close to the genome physical end and immediately upstream of the terminase large subunit gene. Marvin gp32 is of unknown function and has no database matches.

Marvin gp34 is a strong candidate for the portal protein and contains a DUF1484 domain common to phage portal proteins. It is well represented in the peptides identified by mass spectrometry (Table 2), and a product of the expected size is seen by SDS-PAGE (Fig. 6).

Gene 35, for which only a small number of peptides are represented in the mass spectrometry data, likely codes for a protease, with weak sequence matches to other putative mycobacteriophage proteases, including gp5 of both phages Ramsey and Boomer. A similar small number of peptides are found corresponding to the putative scaffolding protein, gp36. Marvin gp36 has poor (29% identity) but significant similarity to a putative scaffolding protein encoded in the Caldicellulosiruptor owensensis genome. Analysis of the gp36 sequence predicts that it has several alpha-helical regions, some with high propensity to form coiled coils, joined by regions of unstructured sequence, and little or no beta structure: these are all features of known scaffolding proteins, and this analysis strengthens the identification of gp36 as the scaffolding protein. It is somewhat unexpected to find peptides from the protease and the scaffolding protein in mature virions, as these proteins are thought to be lost from the structure during capsid maturation in most phages. However, there is evidence for residual amounts of both protease and scaffolding proteins being retained in virions of coliphage T4 (8, 50, 51), and the protease of coliphage P2 is retained in mature virions (9). Our results suggest that some of both protease and scaffolding proteins, or fragments of them, are similarly retained in the Marvin virions. An alternative explanation—that these proteins came from contaminating procapsids that had not packaged DNA—seems unlikely for these virions that were purified in a CsCl density gradient.

Marvin gp37 has weak matches to nonmycobacteriophage proteins, including gp37 of Streptomyces phage VWB, but its specific role has not been established. However, it is present in virions with greater than 66% coverage in the mass spectrometry analysis (Table 2). Many phages have abundant “decoration” proteins on the surface of the capsid which typically stabilize the capsid structure, and in some of those phages (e.g., coliphage lambda [26] and Bacillus phage G [GenBank accession no. JN638751.1]), the gene encoding the decoration protein is known to lie between the scaffolding protein and major capsid protein genes. We accordingly speculate that Marvin gp37 may be such a decoration protein.

Marvin gp38 contains a pfam03864 domain associated with major capsid subunits, and gp38 is presumably the capsid protein. It is the most abundant protein represented in the mass spectrometry analysis and a major band of the predicted size is seen by SDS-PAGE; we note that the Marvin capsid does not engage in wholesale covalent cross-linking, as seen in some other mycobacteriophages (14, 15, 23). Although it has no identifiable relatives among other mycobacteriophages, it has sequence similarity (35% identity) to the gp38 putative capsid subunit of Streptomyces phage VWB (1, 55). We suggest that genes 39 to 42 encode the head and tail completion proteins, and all of the products except gp41 are present in virions (Table 2), albeit in low abundance. Gene 43 encodes the major tail subunit, with weak sequence similarity (35% identity) to the putative major tail subunit (gp13) of mycobacteriophage LeBron (47). The gp43 product is the second most abundant protein seen by mass spectrometry, but it separates indistinctly by SDS-PAGE and migrates slower than anticipated by its predicted molecular mass (Fig. 6). However, this aberrant migration is not unusual among major tail subunit proteins, including those of mycobacteriophages (15, 23). Immediately downstream of gene 43 are two genes that are likely expressed by a −1 programmed translational frameshift, a highly conserved feature of phage genomes (57), with the protein products acting as tail assembly chaperones (Fig. 3); the predicted position of the frameshift is 15 to 16 codons prior to the termination codon of gene 44. These are not expected to be components of intact virions, and corresponding peptides are not observed (Table 2).

We identify Marvin gene 46 as encoding the tapemeasure protein (Tmp) based initially on its position in the gene order and its very large size (5,550 bp). Analysis of the predicted amino acid sequence shows a high propensity for alpha-helical and coiled-coil structure; these properties are characteristic of Tmp's. There is typically a correlation between the length of a phage tail and the length of the Tmp in the alpha-helical form that it is thought to assume during tail length determination (33, 34, 43). In the case of Marvin, the measured length of the tail (Fig. 1) is 250 nm, and the predicted length of the Tmp as an alpha-helix is 277.5 nm (1,850 amino acids × 0.15-nm rise per amino acid in an alpha helix), and it is plausible that some processing occurs prior to tail assembly. Although the product corresponding to gp46 cannot be unambiguously assigned by SDS-PAGE, there is a possible candidate at approximately 130 kDa (Fig. 6). This protein is too big to be encoded by any of the Marvin genes, except 46 (Tmp) and 87 (putative tail fiber), and there is a different band at the expected position for gp87. We therefore propose that the 130-kDa protein is derived from the gp46 Tmp. It is considerably smaller than the predicted 196 kDa of full-length gp46 and would therefore necessarily be a posttranslationally processed form of the Tmp; we note that such processing of Tmp's is seen quite commonly (27, 43, 58). Interestingly, the Marvin Tmp also contains two small motifs implicated in peptidoglycan hydrolysis. One of these is motif 3, described previously (43, 45), but the other is a putative lytic transglycosylase domain (cd00254), the first such motif to be identified in mycobacteriophage Tmp's. The roles of such domains in Tmp's have not been not fully resolved, but the motif 3 domain in the Tmp of phage TM4 enhances the ability of the phage to productively infect cells in the late stages of growth (45). The motifs in the Marvin Tmp may provide similar or related functions.

The arrangement of the minor tail protein genes—encoding the structure at the very tip of the tail and therefore important for host recognition and triggering the DNA injection process—in Marvin is one of its more unusual features. In all other mycobacteriophage genomes analyzed to date, the minor tail proteins are encoded by a group of 4 to 10 genes immediately downstream of the tapemeasure protein gene (19). However, in Marvin, this group of genes is split such that genes 47 to 50 likely encode four tail proteins, and the remaining proteins are encoded by genes 87 to 90 (Fig. 3 and 7), located among nonstructural genes and displaced by about 20 kbp from their normal location. Marvin gp47 and gp48 have sequence similarity to LeBron gp17 and gp18 (36% and 54% identity, respectively) and, more distantly, to Wildcat gp39 and gp40 (Fig. 7). Marvin gp49 shares 43% identity with Corndog gp72. Marvin gp50 has no database matches but is proline rich (12%), a feature sometimes found in minor tail proteins. Virion analysis confirms that gp87 and gp88, as well as gp47, gp48, and gp50, are structural components (Table 2 and Fig. 6); in a separate mass spectrometry experiment, gp89 was also identified as a virion protein (data not shown).

Fig 7.

Noncanonical arrangement of tail protein genes. The genomes of Marvin, Wildcat (singleton), Konstantine (subcluster H1), and Predator (subcluster H1) are represented with gene annotations as described for Fig. 3. Pairwise nucleotide sequence similarities are displayed using Phamerator and are colored as described for Fig. 5. Note that there are segments of DNA sequence similarity between Konstantine and Predator, but none between Marvin and Wildcat or between Wildcat and Konstantine. Two relevant segments of the Marvin genome are shown, encompassing genes 46 to 50 and 85 to 91, whereas contiguous regions of the other phages are displayed. The Pham designations of the minor tail protein genes shared between the phages are shown in large bold type.

One of the displaced genes, gene 87, encodes a large protein (1,318 residues) corresponding to gp43 and gp44 of Wildcat (Fig. 7), which are combined into a single open reading frame (Fig. 7). Wildcat gp44 contains a putative β-lactamase domain, and related proteins are widespread throughout mycobacteriophage genomes, although in each instance, the genes coding for these proteins are located among the minor tail protein genes (39). This is observed in Wildcat, as well as in the cluster H1 genome, Konstantine (Fig. 7), where they are positioned just downstream of tapemeasure protein genes; another H1 phage, Predator, lacks this function (Fig. 7). Marvin gp88 is a member of Pham2441, along with Wildcat gp45, although BlastP searches suggest that Konstantine gp36 is a more distantly related homologue (Fig. 7). Marvin gp89 and gp90 are homologues of Konstantine gp37 and gp38, respectively, and there is also a relative of Marvin gp89 in Wildcat (gp46). These relationships suggest that all four Marvin genes (87 to 90) code for minor tail proteins, although gp90 was not found by mass spectrometry.

Surprisingly, an additional protein, gp57, is found associated with virions (Table 2) although gene 57 lies to the right of the lysis cassette and outside the regions described above (Fig. 3). Marvin gp57 has weak sequence similarity to LeBron gp24 (27% identity), which is encoded at the extreme right end of the LeBron tail gene cluster and which therefore is also a candidate for a virion protein. None of the protein products of any of the surrounding genes were identified as virion-associated proteins.

Marvin lysis cassette.

The lysis cassette of Marvin is coded near the middle of the genome, a common location for mycobacteriophage genomes, and includes lysin A (gp51), the holin protein (gp54), and a putative lysin B (gp53) (16, 42) (Fig. 3). The lysin A is most closely related to the cluster B phages Pacc40 and Cooper (55% and 51% identity, respectively) and contains a PGRP domain associated with N-acetylmuramoyl-l-alanine amidase activity. Marvin gp53 is only a distant relative of other mycobacteriophage lysin B proteins, with the central portion having weak sequence similarity to Giles gp32, extending the considerable sequence diversity of this phamily of proteins (42). The 97-residue Marvin gp54 is a good candidate for the holin protein, containing two strongly predicted transmembrane domains at residues 8 to 30 and 51 to 73. The only other mycobacteriophage protein with significant sequence similarity is Barnyard gp41 (47% identity). The small protein encoded between lysins A and B (gp52) has no close relatives and is of unknown function. Although it is not related to the gp1 protein of mycobacteriophage Ms6 (7), it is plausible that it plays a similar chaperone-like role in the functioning of the lysis system.

Nonstructural genes 55 and 56, 58 to 86, and 92 to 107.

The 32-kbp right half of the Marvin genome (from 33.4 kbp to the right end) encodes mostly nonstructural proteins gp55 to gp86 and gp92 to gp107 (Fig. 3), the only exception being gp57, which is virion associated. These genes form three distinct groups: 55 to 91 in a rightwards-transcribed group that also includes the putative minor tail protein genes 87 to 90, the leftwards-transcribed genes 92 to 103, and four rightwards-transcribed genes at the right end, 104 to 107. Genes 55 to 92 may constitute a single operon, and most genes are closely linked, although there are intergenic gaps of >100 bp between genes 65 and 66, 70 and 71, and 75 and 76. Because of the difficulty in accurately predicting mycobacteriophage promoter sequences (other than canonical σ⁷⁰-like candidates) it is unclear if these genes are transcribed from a single upstream promoter (presumably upstream of gene 55) or if there are additional promoters in the intergenic gaps. Twenty-seven of the 37 genes are orphams and have no close mycobacteriophage relatives (Fig. 3). However, several of these have either a weak match to other mycobacteriophage proteins or to other database matches. These include gp62, gp63, and gp72, which have weak sequence similarities to LeBron gp58 (34% identity), Tweety gp64 (46% identity), and Pacc40 gp68 (36% identity), respectively, all of which also match bacterial proteins of unknown function. Marvin gp64, gp71, and gp75 also have similarities to bacterial proteins of unknown function. Marvin gp65 and gp86 have similarity to HNH homing endonucleases (Fig. 3).

Perhaps the most informative of the database matches of genes in this region is to the gene coding for Marvin gp58, which has significant similarity to phage-encoded Xis proteins, including the putative Xis of the M. tuberculosis prophage-like element ϕRv2 (Rv2657c; 43% amino acid identity) (30). This is surprising because there is no evidence of an integrase gene in the Marvin genome. However, this is reminiscent of the genome structure observed in mycobacteriophage TM4. Until recently, TM4 was also a singleton phage, but it is now a member of cluster K, for which there are four other relatives (46). Although TM4 is not temperate and does not form stable lysogens, all of the other cluster K phages are temperate and contain easily recognizable integrase genes. The simple explanation is that TM4 is a derivative of a temperate parent in which the integrase and presumably the repressor genes have been lost (46); a similar event has been proposed for mycobacteriophage D29 (14). It is plausible that Marvin is also a derivative of a temperate parent that has lost its integrase gene but retained the Xis gene, 58.

Of the 10 putative gene products with mycobacteriophage relatives (Fig. 3), several have informative database matches to nonmycobacteriophage proteins or to conserved domains. For example, gp69 is predicted to have a domain of the family cl00695 that is associated with the SMF family of proteins, including Helicobacter pylori DprA, which binds to single-stranded DNA (ssDNA) to facilitate transformation. The gene segment from 79 to 83 is of particular note in that HHPred predicts that four of these genes' products (all except the small orpham gp82) are transferases, with gp79 being a methyltransferase and gp80, gp81, and gp83 being glycosyltransferases. Marvin gp80 and gp81 are predicted specifically to be polypeptide N-acetylgalactosaminyltransferases, and gp83 is predicted specifically to be an α-1,3-mannosyl-glycoprotein β-1,2-N-acetylglucosaminyltransferase. The four genes coding for these proteins perhaps contribute to a common biochemical pathway because they are conserved with a common synteny in phage Corndog (gp35 to gp38), although other mycobacteriophages have just a subset of the genes in mosaic relationships (see below). It is unclear whether the presumed protein targets of modification are phage or bacterial in nature.

The 11 leftwards-transcribed genes 93 to 103 have no close relatives in other mycobacteriophages, and only one's product, gp94, has weak matches to other mycobacteriophages as well as nonmycobacteriophage proteins. The closest mycobacteriophage relative is Konstantine gp57 (30% identity), and there are numerous related proteins of unknown functions. The role of this segment of the Marvin genome is therefore unclear. We note, however, that there are three putative σ⁷⁰-like promoters positioned between genes 94 and 95, between genes 101 and 102, and upstream of 103 (Fig. 3).

The four rightwards-transcribed genes at the right end of the genome, 104 to 107, have no mycobacteriophage homologues, but gp105 and gp106 are closely related to families of host-encoded proteins. Marvin gp105 contains an amidoligase-2 (COOH-NH₂ ligase superfamily) domain similar to that found in RflaF proteins of Ruminococcus flavefaciens (32% identity), and gp106 is related to glutamine amidotransferases of the type II class. The specific role of these genes is not known but could be involved in the synthesis of novel metabolites or peptide-tagging systems (32). We note that a pair of genes encoding related functions but only very distantly related are also present in phage phiEco32 (49), and these have been postulated to modify the bacterial cell wall to prevent infection by other bacteriophages (32). Curiously, located between Marvin genes 105 and 106 is a 26-bp palindrome composed of identical 13-bp inverted repeats (Fig. 3). The role of this is unclear, but it is a candidate for a binding site of a regulatory protein.

Mosaicism of the Marvin genome.

The prominent architectural feature of mycobacteriophage genomes is that they are mosaic, with different segments having distinctly different evolutionary origins (30, 43). Marvin is likely to be no exception to this, although the small number of genes with relatives in other mycobacteriophages makes this less obvious (Fig. 3). However, a particularly good example of genome mosaicism is seen in genes 78 to 84 (Fig. 8). Genes 78, 82, and 84 have no relatives, although 79, 80, 81, and 83 are related to other mycobacteriophage genes (Fig. 8). Phamily circle representations of the latter four genes show which mycobacteriophage genomes have the related genes and which do not (Fig. 8). For example, although all four have a related gene in Corndog, the presence in other genomes varies greatly. Pham 87 (containing Marvin gp83) has the largest number of members, none of which are in the subcluster I1 genomes Brujita or Island3. In contrast, Pham 458 and Pham 1598 have fewer members, but both Brujita and Island3 are included in both of them. These genes therefore have distinct phylogenies and have arrived in the Marvin genome through different evolutionary journeys.

Fig 8.

Marvin genome mosaicism. Mosaicism of the Marvin genome is illustrated by a segment of genes 78 to 84 and their relatives. Genes 78, 82, and 84 are orphams and have no relatives in other mycobacteriophages. In contrast, genes 79, 80, 81, and 83 have various numbers of relatives that are present in a wide variety of other mycobacteriophage genomes, as illustrated by the phamily circles. Each phamily circle has all 84 genomes in the “Marvin” phamerator database around the circumference, and arcs are drawn between genomes that contain members of that particular phamily. Blue arcs correspond to relationships revealed by BlastP comparison and red arcs to those by Clustal comparison.

DISCUSSION

We have described here a new singleton mycobacteriophage, Marvin, that reveals a number of new insights into the diversity and evolution of bacteriophages. Although the number of sequenced mycobacteriophage genomes has increased sharply over the past 10 years (19, 47), the continued discovery of new singleton phages such as Marvin demonstrates that our current collection is far from being a representative sample of the population at large. As the mycobacteriophage collection expands further, we anticipate that relatives of Marvin will be discovered, although we note that phages such as Giles, Corndog, and Wildcat (40, 43) persist as singleton phages many years after their initial isolation.

Marvin is the first mycobacteriophage in which we have observed an obvious interruption in the group of minor tail protein genes that are typically positioned immediately downstream of the tapemeasure protein gene. There are several examples of gene insertions within the structural gene operon, such as in Wildcat or Corndog (43), and the integration cassette appears to be “misplaced” within the Giles genome, such that it is flanked by tail genes (40). There are additional examples of interruptions within the head genes of siphoviral phages, including a large insertion between the head accessory protein and capsid protease genes in Vibrio phage SIO-2 (2). The Marvin genomic architecture is unusual, however, with a contiguous segment of the minor tail protein genes positioned about 20 kbp away from the other tail genes and within nonstructural genes. There is also a lone virion gene, 57, situated among nonstructural genes. In the absence of any close relatives of Marvin, it is not clear whether the evolutionary events giving rise to this are relatively recent or ones that are older and well established. The organization raises substantial questions as to how the structural genes are expressed, and if there are promoters for late gene expression upstream of genes 57 and 87.

Marvin is not a temperate phage, and we have been unable to recover stable lysogens. It does not contain an identifiable integrase gene, and although there are at least two candidate DNA-binding proteins (gp22 and gp58), we doubt that either acts as a phage repressor; Marvin gp22 may for example be an antitoxin component of a toxin-antitoxin system. Marvin gp58 is strongly predicted to contain a helix-turn-helix DNA binding motif and shows strong sequence similarity to Xis family proteins, including the RDF of the M. tuberculosis prophage-like element ϕRv2. This is a curious gene to find in a lytic phage and we therefore predict that Marvin is a derivative of a temperate parent and has lost—perhaps recently—its immunity and integration functions. This is not unprecedented, and similar conclusions can be drawn about the origins of mycobacteriophage D29 (14) as well as TM4 (15, 46). We thus predict that future phage discovery efforts will identify close relatives of Marvin but which are temperate, just as occurred with the finding of relatives of TM4 (46).

The Marvin genome contains several groups of genes that are not found in other mycobacteriophages. Although many of these have no known function, the presence of amidoligase and glutamine amidotransferase genes (105 and 106) suggests the possibility of intriguing new functions. Genes with these putative functions have been observed in the unrelated phage phiEco32, and it has been suggested that they could play a role in modifying the cell wall and thus preventing superinfection by other phages (32). This is certainly a plausible role in Marvin too, although they could also play roles in synthesis of secondary metabolites or in modulating expression of either phage or host genes. If Marvin is indeed derived from a temperate parent, then these genes could be expressed during lysogeny so as to influence the physiological state of the bacterial host.

Finally, although Marvin has no close relatives, comparisons with other mycobacteriophages clearly show its mosaic nature. This is observed with genes 79 to 83 encoding predicted transferases (Fig. 8) but also with the curious array of genes 7 to 11. This is the one segment of the Marvin genome that appears to have been acquired relatively recently and must be in rather rapid exchange among the mycobacteriophage genomes. We note that although Marvin gene 8 has 95% or greater nucleotide sequence similarity to at least 10 other mycobacteriophages, there do not appear to be any closely related host genes. Acquisition presumably therefore came from other mycobacteriophages, rather than from the host chromosome, and it is possible that these are new types of self-mobile elements.