- Email: [email protected]
An Alternative to the Accepted Phylogeny of Purple Bacteria Based on 16S rRNA: Analyses of the Amino Acid Sequences of Cytochromes C2 and C556 from Rhodobacter (Rhodovulum) sulfidophilus
Archives of Biochemistry and Biophysics Vol. 388, No. 1, April 1, pp. 25–33, 2001 doi:10.1006/abbi.2000.2221, available online at http://www.idealibrary.com on
An Alternative to the Accepted Phylogeny of Purple Bacteria Based on 16S rRNA: Analyses of the Amino Acid Sequences of Cytochromes C2 and C556 from Rhodobacter (Rhodovulum) sulfidophilus Richard P. Ambler,* ,1 Terry E. Meyer,† Robert G. Bartsch,‡ and Michael A. Cusanovich† *Institute of Cell and Molecular Biology, University of Edinburgh, Edinburgh, EH9 3JR, Scotland; †Department of Biochemistry, University of Arizona, Tucson, Arizona 85721; and ‡Department of Biochemistry 0506, University of California at San Diego, La Jolla, California 92093
Received June 14, 2000, and in revised form November 20, 2000; published online March 9, 2001
It is becoming increasingly apparent from complete genome sequences that 16S rRNA data, as currently interpreted, does not provide an unambiguous picture of bacterial phylogeny. In contrast, we have found that analysis of insertions and deletions in the amino acid sequences of cytochrome c 2 has some advantages in establishing relationships and that this approach may have broad utility in acquiring a better understanding of bacterial relationships. The amino acid sequences of cytochromes c 2 and c556 have been determined in whole or in part from four strains of Rhodobacter sulfidophilus. The cytochrome c 2 contains three- and eight-residue insertions as well as a single-residue deletion in common with the large cytochromes c 2 but in contrast to the small cytochromes c2 and mitochondrial cytochromes. In addition, the Rb. sulfidophilus protein shares a rare six- to seven-residue insertion with other Rhodobacter cytochromes c 2. The cytochrome c556 is a low-spin class II cytochrome c homologous to the greater family of cytochromes cⴕ, which are usually high-spin. The similarity of cytochrome c556 to other species of class II cytochromes is consistent with the relationships deduced from comparisons of cytochromes c 2. Thus, our results do not support placement of Rb. sulfidophilus in a separate genus, Rhodovulum, which was proposed primarily on the basis of 16S rRNA sequences. Instead, the Rhodobacter cytochromes c 2 are distinct from those of other genera and species of purple bacteria and show a different pattern of relationships among species than reported for 16S rRNA. © 2001 Academic Press
1 To whom correspondence and reprint requests should be addressed. Fax: 44-131-668-3870. E-mail: [email protected].
0003-9861/01 $35.00 Copyright © 2001 by Academic Press All rights of reproduction in any form reserved.
Key Words: cytochrome; sequence; Rhodobacter sulfidophilus; RNA; phylogeny; purple bacteria; evolution.
Molecular sequence comparisons provide a powerful means for quantifying relationships among species. As we enter the new era of whole genome sequence analysis, these comparisons take on new meaning as thousands of genes may now be separately analyzed and compared. Nevertheless, there is no agreement on the best way to compare species. The most popular gene for evolutionary studies has been 16S rRNA (1), although it is now becoming apparent that there is only partial agreement with proteins or whole genome analysis (2– 6). The lack of complete agreement is likely to have little to do with the actual genes that are being compared but has more to do with the methodology (7). Thus, both phylogenetics and genomics appear to require development of new tools for quantifying relationships among genes and species which we propose may be provided in part by analysis of insertions and deletions. There is a strong case to be made for the analysis of insertions and deletions in cytochrome c 2 as well as other proteins for which an extensive structural database is available (8) although c-type cytochromes are more characteristic of the gram-negative bacteria and eucaryotes than of gram-positive species and therefore somewhat restricted in their utility. In a general sense, any protein, for which there are insertions and deletions and a number of three-dimensional structures, may be used for such analysis. At present, this ex25
26
AMBLER ET AL.
cludes the majority of genes within a given genome because of insufficient structural information. We have chosen Rb. sulfidophilus for analysis of insertions and deletions because it was reported that the 16S rRNA was sufficiently different from that of other Rhodobacter species to warrant creation of a new genus (9). Thus, one would expect that these differences would also be apparent with other molecules in general and with cytochrome c 2 in particular because we have previously found large sequence differences for this protein among species and genera of purple bacteria (10, 11). With virtually no comparable sequence data currently available for Rb. sulfidophilus, cytochrome c 2 may provide a test of how well the two approaches might agree. As will be shown, the published conclusions of the 16S rRNA analyses do not agree with those based upon insertions and deletions in the cytochrome c 2 from Rhodobacter sulfidophilus. Rb. sulfidophilus was first described by Hansen and Veldkamp (12) as a marine bacterium, related to the fresh-water bacteria, Rb. sphaeroides and Rb. capsulatus, but which would tolerate and utilize high concentrations of hydrogen sulfide similar to that of the purple sulfur bacteria. Imhoff et al. (13) transferred these species from Rhodopseudomonas and placed them in a separate genus, Rhodobacter, along with Rb. adriaticus. More recently, Rb. sulfidophilus was renamed Rhodovulum sulfidophilum on the basis of 16S rRNA sequences and marine habitat (9). We had previously determined the amino acid sequences of the cytochromes c 2 and c⬘ from Rb. sphaeroides and Rb. capsulatus, which showed that they were significantly more closely related to one another than to other species consistent with their inclusion in the same genus, Rhodobacter (10, 14). The cytochrome c 2 from the nonphototrophic Paracoccus denitrificans also belongs to this group and is closest to Rb. capsulatus (15, 16). Moreover, the cytochromes c 2 from Thiobacillus versutus and Thiosphaera pantotropha are 95% similar to that of Paracoccus denitrificans (17), which suggests that they are actually Paracoccus species. In fact, they have recently been transferred to that genus on the basis of 16S rRNA sequence data (18) (an instance where cytochrome c 2 and rRNA agree, presumably due to the small amount of change observed in these molecules). We have isolated the soluble cytochromes from four strains of Rb. sulfidophilus and have found that the principal soluble cytochrome is a c556 and there are lesser amounts of c 2 and c⬘ (19). All cytochromes from the Rb. sulfidophilus strains studied are acidic, which is characteristic of marine and halophilic species (20). We have now determined the amino acid sequences of the Rb. sulfidophilus cytochromes and find that insertions and deletions establish that the cytochromes c 2 are most similar to those of Rb. sphaeroides.
MATERIALS AND METHODS Rb. sulfidophilus strains 2.8.1 and 2.10.1 were grown photosynthetically on a complex medium including yeast extract, peptone, malate, glutamate, acetate, and succinate plus mineral salts. Strains W12 and BSW8 were grown on a simpler medium composed of yeast extract, malate, and mineral salts. Accumulated frozen cells of strain 2.8.1 (the preparation was similar for strains 2.10.1 and W12) were suspended in 100 mM Tris–Cl, pH 8, they were broken in an automated French press, membranes were separated by ultracentrifugation, the clarified extract was desalted on Sephadex G-25, and then adsorbed to DEAE-cellulose from 20 mM Tris–Cl, pH 8. After the column was washed with 20 mM Tris–Cl plus 80 mM NaCl, a mixture of cytochromes c 2 and c⬘ was eluted in steps at 120 to 160 mM NaCl in buffer, followed closely by cytochrome c556 at 200 –240 mM NaCl. Pooled protein fractions were concentrated by pressure dialysis and fractionated by ammonium sulfate. Cytochrome c556 precipitated primarily at 50 – 80% saturation, cytochrome c 2 at 70 – 90% saturation, and cytochrome c⬘ at 50 –70% saturation ammonium sulfate. The cytochrome fractions were further purified by Sephadex G-75 and chromatographed on DEAE-cellulose using a linear gradient. For cytochrome c 2, the gradient was from 20 mM Tris–Cl to 20 mM Tris–Cl plus 100 mM NaCl. The absorption spectra were similar to those of other species. The purest fraction had a ratio of 280 nm to oxidized Soret absorbance of 0.18. For cytochrome c556, the gradient was 100 to 200 mM NaCl in 20 mM buffer. The oxidized absorption spectrum displayed a prominent peak at 630 nm, indicating that there was a significant amount of high-spin protein present in addition to the predominant low-spin form. The purest protein fraction had a ratio of 280 nm to oxidized Soret absorbance of 0.14. For cytochrome c⬘, the gradient was from 0 to 100 mM NaCl in buffer, resulting in elution at slightly lower concentration than for cytochrome c 2 . The absorption spectra were typical of the cytochromes c⬘. In strain BSW8, the cytochrome c 2 was slightly more acidic than the cytochrome c556, but the two were eluted from the initial DEAE-cellulose column in a single band at about 300 mM NaCl. There was much less cytochrome c⬘ than the other two proteins and further purification of this protein was not attempted. The cytochrome c556 was poorly resolved from cytochrome c 2 using a linear gradient from 160 to 320 mM NaCl, but further purification of leading and trailing edges of the elution band was similar to that for the other strains. The complete amino acid sequences of strain 2.8.1 cytochromes c 2 and c556 were determined by a combination of standard manual and automated Edman degradation plus mass spectroscopy (20, 21). Prior to sequence determination, the heme was removed by overnight treatment with acidified mercuric chloride. Details of enzyme digestions, peptide purification, amino acid analyses, and sequence determination may be obtained from the corresponding author. Peptides were isolated from the cytochromes c 2 and c556 of strains 2.10.1 and W12 and their mobilities, compositions, and N-terminal sequences were determined to detect differences from strain 2.8.1. The cytochromes c 2 and c556 from strain BSW8 were sequenced nearly as thoroughly as from strain 2.8.1 and the cytochrome c⬘ from strain W12 was completely sequenced except for a missing peptide in the center. Internal insertions and deletions were initially located by sequence alignment and their precise sizes and locations were determined by superposition of the three-dimensional structures of horse cytochrome c (22), Rb. capsulatus cytochrome c 2 (23), Paracoccus denitrificans cytochrome c 2 (24), and Rb. sphaeroides cytochrome c 2 (25). Usually, automatic structural alignment is not possible where the proteins do not have the same numbers of amino acid residues. Insertions and deletions normally occur outside regions of secondary structure. Therefore, regions of conserved secondary structure such as N-terminal and C-terminal helices were manually aligned in the first iteration followed by alignment of additional residues roughly superimposed by the first alignment. In those instances where the
CYTOCHROME SEQUENCES AND PHYLOGENY
27
FIG. 1. Evidence for the amino acid sequence of Rhodobacter sulfidophilus strain 2.8.1 cytochrome c 2. Lines above and below the sequence indicate the peptides used in the sequence determination. Double arrows are for residues established by manual Edman degradation. The symbol “T” is for tryptic peptides, “H” is for thermolysin peptides, and “C” is for chymotryptic peptides.
insertions may have caused localized rearrangements and then the elements of secondary structure on either side of the insertion were aligned in a third iteration to pinpoint the exact location in each instance. Such rearrangements are minimal in cytochrome c 2, and the sizes and locations of insertions and deletions have been accurately determined. N- and C-terminal insertions and deletions of any length were ignored as insignificant because they can easily be formed through frameshift mutations.
RESULTS AND DISCUSSION
We examined four strains of Rhodobacter, which appeared to be related based upon their microbiological properties and cytochrome compositions (19). We began with strains ATH 2.8.1 and 2.10.1 from the Van Niel collection, which were marine bacteria isolated by Dr. Sidney Elsden in Britain before either the Rb. sulfidophilus genus or species were described. These strains were never typed sufficiently to distinguish them from other Rhodobacter strains and species. The evidence for the sequence of the 126-residue strain 2.8.1 cytochrome c 2 is shown in Fig. 1. We could not detect any differences in the peptides of the cytochrome c 2 from strain 2.10.1 (which was not completely sequenced) in comparison with the reference strain 2.8.1 and conclude that the sequences are either identical or involve only one or a few compensating substitutions such as Gly-Ala to Ala-Gly. Subsequently, we examined the cytochrome c 2 from Rb. sulfidophilus strain W12, isolated from the marine environment in the Netherlands (12) and we could detect only 6 differences (at positions 65, 79, 80, 83, 90, and 94) through comparison of peptides with the reference strain 2.8.1 (evidence for the sequence not shown, but see Fig. 2 for
the actual sequence). Rhodobacter strain BSW8 was isolated from the beach at Scripps Institution of Oceanography in La Jolla California by Dr. Paul Weaver and the sequence of the cytochrome c 2 proved sufficiently different from the reference strain to warrant a more complete study (sequence shown in Fig. 2). This sequence is 13% different (17 substitutions) from that of strain 2.8.1. We compared the Rb. sulfidophilus cytochromes c 2 with other species as shown in the alignment of Fig. 2. Rb. sulfidophilus cytochrome c 2 averages about 60% difference from other species and is not significantly closer to any one of them by this criterion. This value is similar to the average for all cytochromes c 2 and does not provide any information on specific relationships because divergence is approximately balanced by convergence (7, 11). Therefore, Rb. sulfidophilus provides an ideal opportunity to evaluate alternative means of comparing sequences. Where there are large sequence differences among species, as in the present case, the numbers, sizes, and locations of insertions and deletions can be more informative than percentage identities (7, 8, 11). Because there are several three-dimensional structures for cytochromes c 2, the insertions and deletions have been accurately located as shown in Fig. 3. Rb. sulfidophilus cytochrome c 2 contains 3- and 8-residue insertions (at positions 69 –71 and 93–100 of the alignment) and a single-residue deletion (at position 109) relative to the smaller cytochromes c 2 and mitochondrial cytochromes c as highlighted in Figs. 2 and 3. Other species of cytochrome c 2 with this pattern of insertions and dele-
28
AMBLER ET AL.
FIG. 2. Alignment of cytochrome c 2 sequences based upon three-dimensional structures where available. (1) Rb. sulfidophilus strain 2.8.1 (this work), (2) Rb. sulfidophilus strain W12 (this work), (3) Rb. sulfidophilus strain BSW8 (this work), (4) Rb. sphaeroides strain 2.4.1 (10), (5) Rb. capsulatus strain SL (10), (6) P. denitrificans (15, 16), (7) Rb. capsulatus strain SB1003 cyt c y (soluble domain only) (42), (8) Rb. sphaeroides strain GA cyt c y (soluble domain only) (37), (9) Rb. sphaeroides strain CYCA65R7 iso-c2 (38), (10) Horse mitochondrial cytochrome c (43). Insertions and deletions relative to mitochondrial cytochrome c and most other cytochromes c 2, which are discussed in the text are boldfaced.
tions are Rhodobacter sphaeroides, Rb. capsulatus, Rhodospirillum rubrum, R. photometricum, Rhodopseudomonas palustris (10), Paracoccus denitrificans (15, 16), Thiobacillus versutus (17), Roseobacter denitrificans (26), and Rhodospirillum centenum (11). These three shared insertions and deletions in the cytochromes c 2 are highly significant when considered together and are not likely to have independently occurred more than once (11). In this sense, Rb. sulfidophilus cytochrome c 2 groups with the other Rhodobacter species as expected from microbiological properties (12) but apparently not from the 16S rRNA sequence (9). Rb. sulfidophilus cytochrome c 2 has a seven-residue insertion after the heme-binding site (at positions 29 –35 of the alignment) similar to that in Rb. spha-
eroides. However, in Rb. capsulatus, Ro. denitrificans, and P. denitrificans, there is a six-residue insertion at the same location as the seven-residue insertion and a two-residue deletion after the sixth heme ligand methionine (at positions 113 and 114 of the alignment). It should be emphasized that, in a genetic sense, large and small gaps are equally likely to occur, and occasionally small gaps result from protein and DNA sequencing errors (15, 16, 27) (these are excluded from the present analysis). However, among those gaps that are well-documented, small gaps are actually favored and large gaps are selected against, presumably due to structural constraints (28). Therefore, the six- to sevenresidue insertions are more significant than the tworesidue deletion due to their rarity (28) and distinguish the genus Rhodobacter from all other purple bacteria.
CYTOCHROME SEQUENCES AND PHYLOGENY
29
FIG. 3. First iteration superposition of the alpha carbons of the three-dimensional structures of Rb. sphaeroides cytochrome c 2 (red) (25) and horse mitochondrial cytochrome c (green) (22), as described under Materials and Methods in the standard front view towards the exposed heme edge. The locations of insertions and deletions were established from comparisons of all known structures and further iterations at each localized region. Insertions and deletions highlighted in Fig. 2 are colored cyan and are, clockwise from the upper left: the single-residue deletion (residue 109), the single-residue insertion (residue 16), the seven-residue insertion (residues 29 –35), the three-residue insertion (residues 69 –71), and the eight-residue insertion (residues 93–100).
Despite their rarity, there is a remote possibility that Rb. sulfidophilus and Rb. sphaeroides may have experienced a seven-residue insertion independently of the six-residue insertion in Rb. capsulatus and P. denitrificans. That this is unlikely is shown by the similarity between the six- and seven-residue insertions, notably a Gly at position 29, Thr at 30, and an Ile at 31. Alternatively, we view the six- to seven-residue insertion as more likely to have occurred as two consecutive events, a six-residue insertion followed by a singleresidue insertion or a seven-residue insertion followed by a single-residue deletion. The difference in size of this insertion is not very significant because, as explained above, single-residue insertions and deletions are more common than larger ones (28) and therefore more likely to have occurred more than once, such as the deletion at position 16. Thus, Rb. sulfidophilus and Rb. sphaeroides are typical Rhodobacter species, but differ by two shared small insertions or deletions from Rb. capsulatus and close relatives. This observation indicates that the separation of Rb. sulfidophilus from Rb. sphaeroides and Rb. capsulatus and creation of a new genus on the basis of 16S rRNA (9) cannot be confirmed. That is, Rb. sulfidophilus and Rb. sphaeroides cytochromes c 2 are actually closer to one another than either is to Rb. capsulatus. It is thus likely that the 16S rRNA data, which led to the creation of a new genus (9), may have been overinterpreted. Support for this possibility comes from the fact that no two published RNA trees containing these species have exactly the same branching order (1, 9, 29 –36). To determine whether or not the percentage similarity in RNA might support the cytochrome c 2 results if reinterpreted more conservatively, we analyzed the 16S rRNA data for nonsulfur purple phototrophic bacteria in the same way as for cytochrome c 2 (11). In Fig.
4, the percentage identity in binary comparisons of 16S rRNA of purple bacteria shows a bimodal distribution with means of 87.3% identity (and standard deviation of 1.5) for data between 82.9 and 91.2% identity and 93.6% (SD 1.3) for data between 91.7 and 95.8% identity. The first distribution shows the approximate limit to change for purple bacterial RNA (13%) as a whole and the second distribution primarily reflects the relationships among the Rhodobacter species and close
FIG. 4. Distribution of the percentage identities in binary comparisons of 16S rRNA from representative purple phototrophic bacteria (9, 29, 30, 33–36). Variations in the data, presumably due to differing alignments, were averaged for individual species comparisons prior to plotting, resulting in an average error of 0.4%. The bimodal data were fit with two Gaussian curves shown as solid lines.
30
AMBLER ET AL.
relatives, Erythrobacter, Roseobacter, and Paracoccus (the mean of the distribution for these genera alone is 93.8%). There are more than four standard deviations between the means of the two distributions, it is thus clear that Rhodobacter is distinct from the other genera and species of purple bacteria by this criterion. Rb. sulfidophilus shows 93.8% average identity to the other Rhodobacter species, the same as for the genus as a whole, which shows that it is a genuine Rhodobacter and should not be split off in a separate genus. Rb. sulfidophilus 16S rRNA shows greatest apparent similarity to that of Rb. adriaticus (95.3% identity) and Rb. adriaticus is most similar to Rb. sulfidophilus as reported in (9). However, this value is only one standard deviation from the mean of the second distribution, which is not very significant and also indicates that these two species belong to Rhodobacter. Purple bacterial species occasionally produce more than one cytochrome c 2. There are two cytochromes c 2 in Rb. capsulatus and three in Rb. sphaeroides. The c 2 isozyme in Rb. capsulatus, known as c y, is membranebound and is also present in Rb. sphaeroides (37). It is able to substitute for the soluble c 2 in photosynthetic electron transfer when the soluble c 2 gene is inactivated in Rb. capsulatus but not in Rb. sphaeroides (37). The cytochrome c y is more closely related to the small c 2s and mitochondrial c that do not have 3- and 8-residue insertions and a single-residue deletion, as shown in Fig. 2. The second isozyme in Rb. sphaeroides is also grouped with the “small” c 2s as shown in Fig. 2 and is capable of substituting for the soluble c 2 following mutation of a regulatory gene (38). Thus, there is no possibility that the isozymes might be confused with the soluble cytochrome and that paralogous rather than orthologous genes might have been compared. We have not observed soluble c 2 isozymes in Rb. sulfidophilus and it is unknown whether or not c y may be present in the membrane. If present, it is expected to show the same relationship to Rb. sphaeroides as found for the soluble c 2. One- and three-residue insertions/ deletions separate the two c ys, thus it should be nearly as good a molecular marker as the soluble c 2. To summarize to this point, large insertions and deletions, such as the 6-, 7-, and 8-residue insertions, and the simultaneous occurrence of two or more smaller insertions in cytochrome c 2 is highly significant. They establish Rhodobacter as a genus and separate Rb. sulfidophilus and Rb. sphaeroides on the one hand from Rb. capsulatus and Paracoccus on the other (although still within the same genus). When percentage identity in 16S rRNA is analyzed conservatively, it supports the conclusion based upon insertions and deletions in cytochrome c 2, that sulfidophilus belongs to the genus Rhodobacter. One potential problem with analyzing insertions and deletions is the small sample size, they are decidedly
less common than amino acid substitutions. Thus, we examined a second type of cytochrome from Rb. sulfidophilus to determine if we would obtain a similar result as with cytochrome c 2. The sequence of Rb. sulfidophilus strain 2.8.1 cytochrome c556 was completely determined as shown in Fig. 5. We have not been able to detect any differences in the peptides from strains 2.10.1 or W12 in comparison with the reference strain 2.8.1, suggesting identical or closely related sequences in agreement with the results for cytochrome c 2. The sequence of strain BSW8 c556 has 20 differences out of 124 or 16% divergence as shown in Fig. 6. Cytochrome c556 is a class II cytochrome, homologous to the cytochromes c⬘ with Met 15 (position 17 of the alignment) likely to be the sixth ligand to the heme. Rb. sulfidophilus Strain W12 was the only strain which yielded sufficient pure cytochrome c⬘ for analysis. We determined about 90% of the sequence (except for a missing peptide), which is also shown in Fig. 6. Low-spin homologs of cytochrome c⬘ have been sequenced from Rps. palustris (14), Agrobacterium tumefaciens (39), Bradyrhizobium japonicum (40), and Rb. sphaeroides (41). The high-spin examples, known as cytochromes c⬘, are more often observed in the purple bacteria than are the low-spin examples and are present in higher concentrations (19). Rb. sulfidophilus cytochrome c556 is clearly more abundant than is cytochrome c⬘, thus an exception to the rule. Rb. sulfidophilus cytochrome c556 averages about 75% difference to all other class II cytochromes (which is near the mean for all pairwise comparisons) and is not significantly closer to any other species on a percentage basis. Furthermore, Rb. sulfidophilus cytochrome c⬘ is no closer to the c556 than it is to other species of c⬘ by this measure. This is consistent with the observation with cytochrome c 2 that percentage identities do not provide much information on specific relationships where there is a large amount of change relative to the average for that gene or protein. Insertions and deletions are present in the class II cytochromes such as c556, but they are usually small (i.e., relatively less significant than those in cytochrome c 2). Nevertheless, analysis of insertions and deletions indicate that the low-spin isozymes of cytochrome c⬘ probably arose independently in Rb. sulfidophilus, Rb. sphaeroides, and Rps. palustris from high-spin precursors. CONCLUSIONS
Protein or nucleic acid sequences may be used at several levels of analysis. They can establish homology between two genes or proteins and they can also be used to determine whether newly isolated strains are the same or different from previously described organisms. In the present study, we have used the sequences of the photosynthetic cytochromes in both
CYTOCHROME SEQUENCES AND PHYLOGENY
31
FIG. 5. Evidence for the amino acid sequence of Rb. sulfidophilus strain 2.8.1 cytochrome c556. Lines above and below the sequence represent peptides used in the sequence determination. Single and double arrows represent sequences determined by automated and manual Edman degradation. Peptide nomenclature is as follows: “T” is for trypsin, “C” is for chymotrypsin, “D” is for Pseudomonas Asp-N protease, “nx” is for cyanogen bromide, “H” is for thermolysin, and “F” is for Staphylococcus Glu-C protease.
ways. We have shown that cytochrome c556 is a class II cytochrome, a result which is not obvious from spectral and redox properties. We have shown that the Elsden strains of purple bacteria in the Van Niel collection belong to the species Rb. sulfidophilus. Sequences can also be used to quantify the degree of similarity among isolates, but only where there is a small amount of change relative to the whole. Strain BSW8 is a good example in that the cytochrome c 2 is 13% different from the nearest homolog (where the average for all c 2 s is about 60% difference) and cytochrome c556 is 16% different from the same reference strain (where the average for all class II cytochromes is about 75% difference). Thus, these two proteins show good agreement in identifying strain BSW8 as a separate but closely related species to Rb. sulfidophilus. That is, only 17% of the residues have been substituted, which on average are able to change (13 divided by 60% of 126 ⫽ 17% and 16 divided by 75% of 124 ⫽ 17%). It is likely that any
other gene that has not been acquired through recent gene transfer, including that for rRNA, will show the same result. That is, a 17% substitution when the limiting variation is used as a reference. Thus, the 16S rRNA for strain BSW8 should be no more than 2% different from that of other sulfidophilus strains. Only insertions and deletions, being less common than amino acid substitutions, can be used to show that Rb. sulfidophilus is actually closer to Rb. sphaeroides than to Rb. capsulatus and should not be excluded from the genus Rhodobacter. To be even more precise about the relationship of Rb. sulfidophilus to other species will require analysis of additional genes and preferably, complete genome sequences. We can safely predict from what we have learned about cytochromes c 2 and c556, that Rb. sulfidophilus will have more genes in common with Rb. sphaeroides than either will have in common with Rb. capsulatus. Until then, name changes are premature.
32
AMBLER ET AL.
FIG. 6. Alignment of class II cytochromes. (1) Rb. sulfidophilus strain 2.8.1 cyt c556 (this work), (2) Rb. sulfidophilus strain BSW8 cyt c556 (this work), (3) Rb. sulfidophilus strain W12 cyt c⬘ (this work), (4) Rb. sphaeroides strain 2.4.1 cyt c⬘ (14), (5) Rb. sphaeroides strain 2.4.1 cyt c554 (41), (6) Rb. capsulatus strain SP7 cyt c⬘ (14), (7) Rb. capsulatus strain SB1003 orf 1662 [http://Rhodol.uchicago.edu/ capsulapedia], (8) Rps. palustris strain 2.1.37 cyt c556 (14), (9) Rps. palustris strain 2.1.37 cyt c⬘ (14). The sixth heme ligand Met is boldfaced.
ACKNOWLEDGMENT This work was supported in part by a grant from the National Institutes of Health (GM21277).
REFERENCES 1. Woese, C. R. (1987) Microbiol. Rev. 51, 221–271. 2. Koonin, E. V., Mushegian, A. R., Galperin, M. Y., and Walker, D. R. (1997) Mol. Microbiol. 25, 619 – 637. 3. Gupta, R. S. (1998) Microbiol. Mol. Biol. Rev. 62, 1435–1491. 4. Gupta, R. S. (1998) Mol. Microbiol. 29, 695–707. 5. Pennisi, E. (1998) Science 280, 672– 674. 6. Pennisi, E. (1999) Science 284, 1305–1307. 7. Meyer, T. E., Cusanovich, M. A., and Kamen, M. D. (1986) Proc. Natl. Acad. Sci. USA 83, 217–220. 8. Meyer, T. E., Van Beeumen, J. J., Ambler, R. P., and Cusanovich, M. A. (1996) in Origin and Evolution of Biological Energy Conversion (Baltscheffsky, H., Ed.), pp. 71–108, VCH, NY.
9. Hiraishi, A., and Ueda, Y. (1994) Int. J. Syst. Bacteriol. 44, 15–23. 10. Ambler, R. P., Daniel, M., Hermoso, J., Meyer, T. E., Bartsch, R. G., and Kamen, M. D. (1979) Nature 278, 659 – 660. 11. Samyn, B., Fitch, J., Meyer, T. E., Cusanovich, M. A., and Van Beeumen, J. J. (1998) Biochim. Biophys. Acta 1384, 345– 355. 12. Hansen, T. A., and Veldkamp, H. (1973) Arch. Microbiol. 92, 45–58. 13. Imhoff, J. F., Truper, H. G., and Pfennig, N. (1984) Int. J. Syst. Bacteriol. 34, 340 –343. 14. Ambler, R. P., Bartsch, R. G., Daniel, M., Kamen, M. D., McLellan, L., Meyer, T. E., and Van Beeumen, J. J. (1981) Proc. Natl. Acad. Sci. USA 78, 6854 – 6857. 15. Ambler, R. P., Meyer, T. E., Kamen, M. D., Schichman, S. A., and Sawyer, L. (1981) J. Mol. Biol. 147, 351–356. 16. Van Spanning, R. J. M., Wansell, C., Harms, N., Oltmann, L. F., and Stouthamer, A. H. (1990) J. Bacteriol. 172, 986 –996. 17. Ubbink, M., Van Beeumen, J., and Canters, G. W. (1992) J. Bacteriol. 174, 3707–3714.
CYTOCHROME SEQUENCES AND PHYLOGENY 18. Katayama, Y., Hiraishi, A., and Kuraishi, H. (1995) Microbiol. 141, 1469 –1477. 19. Bartsch, R. G. (1991) Biochim. Biophys. Acta 1058, 28 –30. 20. Ambler, R. P., Daniel, M., McLellan, L., Meyer, T. E., Cusanovich, M. A., and Kamen, M. D. (1987) Biochem. J. 248, 365–371. 21. Ambler, R. P., Daniel, M., Melis, K., and Stout, C. D. (1984) Biochem. J. 222, 217–227. 22. Bushnell, G. W., Louie, G. V., and Brayer, G. D. (1990) J. Mol. Biol. 214, 585–595. 23. Benning, M. M., Wesenberg, G., Caffrey, M. S., Bartsch, R. G., Meyer, T. E., Cusanovich, M. A., Rayment, I., and Holden, H. M. (1991) J. Mol. Biol. 220, 673– 685. 24. Benning, M. M., Meyer, T. E., and Holden, H. M. (1994) Arch. Biochem. Biophys. 310, 460 – 466. 25. Axelrod, H. L., Feher, G., Allen, J. F., Chirino, A. J., Day, M. W., Hsu, B. T., and Rees, D. C. (1994) Acta Crystall. D50, 596 – 602. 26. Okamura, K., Miyata, T., Iwanaga, S., Takamiya, K., and Nishimura, M. (1987) J. Biochem. 101, 957–966. 27. Ridout, C. J., James, R., and Greenwood, C. (1995) FEBS Letts. 365, 152–154. 28. Pascarella, S., and Argos, P. (1992) J. Mol. Biol. 224, 461– 471. 29. Kawasaki, H., Hoshino, Y., Hirata, A., and Yamasato, K. (1993) Arch. Microbiol. 160, 358 –362. 30. Kawasaki, H., Hoshino, Y., Kuraishi, H., and Yamasato, K. (1992) J. Gen. Microbiol. 38, 541–551. 31. Kawasaki, H., Hoshino, Y., and Yamasato, K. (1993) FEMS Microbiol. Letts. 112, 61– 66.
33
32. Hiraishi, A., and Ueda, Y. (1995) Int. J. Syst. Bacteriol. 45, 319 –326. 33. Yurkov, V., Stackebrandt, E., Holmes, A., Fuerst, J. A., Hugenholtz, P., Golecki, J., Gadon, N.,, Gorlenko, V. M., Kompantseva, E. I., and Drews, G. (1994) Int. J. Syst. Bacteriol. 44, 427– 434. 34. Hiraishi, A., and Ueda, Y. (1994) Int. J. Syst. Bacteriol. 44, 665– 673. 35. Hiraishi, A., Urata, K., and Satoh, T. (1995) Int. J. Syst. Bacteriol. 45, 226 –234. 36. Pfennig, N., Luensdorf, H., Sueling, J., and Imhoff, J. F. (1997) Arch. Microbiol. 168, 39 – 45. 37. Myllykallio, H., Zannoni, D., and Daldal, F. (1999) Proc. Natl. Acad. Sci. USA 96, 4348 – 4353. 38. Rott, M. A., Witthuhn, V. C., Schilke, B. A., Soranno, M., Ali, A., and Donohue, T. J. (1993) J. Bacteriol. 175, 358 –366. 39. Van Beeumen, J., Tempst, P., Stevens, P., Bral, D., Van Damme, J., and De Ley, J. (1980) Protides Biol. Fluids 28, 69 –74. 40. Tully, R. E., Sadowsky, M. J., and Keister, D. L. (1991) J. Bacteriol. 173, 7887–7895. 41. Bartsch, R. G., Ambler, R. P., Meyer, T. E., and Cusanovich, M. A. (1989) Arch. Biochem. Biophys. 271, 433– 440. 42. Daldal, F., Cheng, S., Applebaum, J., Davidson, E., and Prince, R. C. (1986) Proc. Natl. Acad. Sci. USA 83, 2012–2016. 43. Margoliash, E., Smith, E. L., Kreil, G., and Tuppy, H. (1961) Nature 192, 1125–1127.
An Alternative to the Accepted Phylogeny of Purple Bacteria Based on 16S rRNA: Analyses of the Amino Acid Sequences of Cytochromes C2 and C556 from Rhodobacter (Rhodovulum) sulfidophilus Richard P. Ambler,* ,1 Terry E. Meyer,† Robert G. Bartsch,‡ and Michael A. Cusanovich† *Institute of Cell and Molecular Biology, University of Edinburgh, Edinburgh, EH9 3JR, Scotland; †Department of Biochemistry, University of Arizona, Tucson, Arizona 85721; and ‡Department of Biochemistry 0506, University of California at San Diego, La Jolla, California 92093
Received June 14, 2000, and in revised form November 20, 2000; published online March 9, 2001
It is becoming increasingly apparent from complete genome sequences that 16S rRNA data, as currently interpreted, does not provide an unambiguous picture of bacterial phylogeny. In contrast, we have found that analysis of insertions and deletions in the amino acid sequences of cytochrome c 2 has some advantages in establishing relationships and that this approach may have broad utility in acquiring a better understanding of bacterial relationships. The amino acid sequences of cytochromes c 2 and c556 have been determined in whole or in part from four strains of Rhodobacter sulfidophilus. The cytochrome c 2 contains three- and eight-residue insertions as well as a single-residue deletion in common with the large cytochromes c 2 but in contrast to the small cytochromes c2 and mitochondrial cytochromes. In addition, the Rb. sulfidophilus protein shares a rare six- to seven-residue insertion with other Rhodobacter cytochromes c 2. The cytochrome c556 is a low-spin class II cytochrome c homologous to the greater family of cytochromes cⴕ, which are usually high-spin. The similarity of cytochrome c556 to other species of class II cytochromes is consistent with the relationships deduced from comparisons of cytochromes c 2. Thus, our results do not support placement of Rb. sulfidophilus in a separate genus, Rhodovulum, which was proposed primarily on the basis of 16S rRNA sequences. Instead, the Rhodobacter cytochromes c 2 are distinct from those of other genera and species of purple bacteria and show a different pattern of relationships among species than reported for 16S rRNA. © 2001 Academic Press
1 To whom correspondence and reprint requests should be addressed. Fax: 44-131-668-3870. E-mail: [email protected].
0003-9861/01 $35.00 Copyright © 2001 by Academic Press All rights of reproduction in any form reserved.
Key Words: cytochrome; sequence; Rhodobacter sulfidophilus; RNA; phylogeny; purple bacteria; evolution.
Molecular sequence comparisons provide a powerful means for quantifying relationships among species. As we enter the new era of whole genome sequence analysis, these comparisons take on new meaning as thousands of genes may now be separately analyzed and compared. Nevertheless, there is no agreement on the best way to compare species. The most popular gene for evolutionary studies has been 16S rRNA (1), although it is now becoming apparent that there is only partial agreement with proteins or whole genome analysis (2– 6). The lack of complete agreement is likely to have little to do with the actual genes that are being compared but has more to do with the methodology (7). Thus, both phylogenetics and genomics appear to require development of new tools for quantifying relationships among genes and species which we propose may be provided in part by analysis of insertions and deletions. There is a strong case to be made for the analysis of insertions and deletions in cytochrome c 2 as well as other proteins for which an extensive structural database is available (8) although c-type cytochromes are more characteristic of the gram-negative bacteria and eucaryotes than of gram-positive species and therefore somewhat restricted in their utility. In a general sense, any protein, for which there are insertions and deletions and a number of three-dimensional structures, may be used for such analysis. At present, this ex25
26
AMBLER ET AL.
cludes the majority of genes within a given genome because of insufficient structural information. We have chosen Rb. sulfidophilus for analysis of insertions and deletions because it was reported that the 16S rRNA was sufficiently different from that of other Rhodobacter species to warrant creation of a new genus (9). Thus, one would expect that these differences would also be apparent with other molecules in general and with cytochrome c 2 in particular because we have previously found large sequence differences for this protein among species and genera of purple bacteria (10, 11). With virtually no comparable sequence data currently available for Rb. sulfidophilus, cytochrome c 2 may provide a test of how well the two approaches might agree. As will be shown, the published conclusions of the 16S rRNA analyses do not agree with those based upon insertions and deletions in the cytochrome c 2 from Rhodobacter sulfidophilus. Rb. sulfidophilus was first described by Hansen and Veldkamp (12) as a marine bacterium, related to the fresh-water bacteria, Rb. sphaeroides and Rb. capsulatus, but which would tolerate and utilize high concentrations of hydrogen sulfide similar to that of the purple sulfur bacteria. Imhoff et al. (13) transferred these species from Rhodopseudomonas and placed them in a separate genus, Rhodobacter, along with Rb. adriaticus. More recently, Rb. sulfidophilus was renamed Rhodovulum sulfidophilum on the basis of 16S rRNA sequences and marine habitat (9). We had previously determined the amino acid sequences of the cytochromes c 2 and c⬘ from Rb. sphaeroides and Rb. capsulatus, which showed that they were significantly more closely related to one another than to other species consistent with their inclusion in the same genus, Rhodobacter (10, 14). The cytochrome c 2 from the nonphototrophic Paracoccus denitrificans also belongs to this group and is closest to Rb. capsulatus (15, 16). Moreover, the cytochromes c 2 from Thiobacillus versutus and Thiosphaera pantotropha are 95% similar to that of Paracoccus denitrificans (17), which suggests that they are actually Paracoccus species. In fact, they have recently been transferred to that genus on the basis of 16S rRNA sequence data (18) (an instance where cytochrome c 2 and rRNA agree, presumably due to the small amount of change observed in these molecules). We have isolated the soluble cytochromes from four strains of Rb. sulfidophilus and have found that the principal soluble cytochrome is a c556 and there are lesser amounts of c 2 and c⬘ (19). All cytochromes from the Rb. sulfidophilus strains studied are acidic, which is characteristic of marine and halophilic species (20). We have now determined the amino acid sequences of the Rb. sulfidophilus cytochromes and find that insertions and deletions establish that the cytochromes c 2 are most similar to those of Rb. sphaeroides.
MATERIALS AND METHODS Rb. sulfidophilus strains 2.8.1 and 2.10.1 were grown photosynthetically on a complex medium including yeast extract, peptone, malate, glutamate, acetate, and succinate plus mineral salts. Strains W12 and BSW8 were grown on a simpler medium composed of yeast extract, malate, and mineral salts. Accumulated frozen cells of strain 2.8.1 (the preparation was similar for strains 2.10.1 and W12) were suspended in 100 mM Tris–Cl, pH 8, they were broken in an automated French press, membranes were separated by ultracentrifugation, the clarified extract was desalted on Sephadex G-25, and then adsorbed to DEAE-cellulose from 20 mM Tris–Cl, pH 8. After the column was washed with 20 mM Tris–Cl plus 80 mM NaCl, a mixture of cytochromes c 2 and c⬘ was eluted in steps at 120 to 160 mM NaCl in buffer, followed closely by cytochrome c556 at 200 –240 mM NaCl. Pooled protein fractions were concentrated by pressure dialysis and fractionated by ammonium sulfate. Cytochrome c556 precipitated primarily at 50 – 80% saturation, cytochrome c 2 at 70 – 90% saturation, and cytochrome c⬘ at 50 –70% saturation ammonium sulfate. The cytochrome fractions were further purified by Sephadex G-75 and chromatographed on DEAE-cellulose using a linear gradient. For cytochrome c 2, the gradient was from 20 mM Tris–Cl to 20 mM Tris–Cl plus 100 mM NaCl. The absorption spectra were similar to those of other species. The purest fraction had a ratio of 280 nm to oxidized Soret absorbance of 0.18. For cytochrome c556, the gradient was 100 to 200 mM NaCl in 20 mM buffer. The oxidized absorption spectrum displayed a prominent peak at 630 nm, indicating that there was a significant amount of high-spin protein present in addition to the predominant low-spin form. The purest protein fraction had a ratio of 280 nm to oxidized Soret absorbance of 0.14. For cytochrome c⬘, the gradient was from 0 to 100 mM NaCl in buffer, resulting in elution at slightly lower concentration than for cytochrome c 2 . The absorption spectra were typical of the cytochromes c⬘. In strain BSW8, the cytochrome c 2 was slightly more acidic than the cytochrome c556, but the two were eluted from the initial DEAE-cellulose column in a single band at about 300 mM NaCl. There was much less cytochrome c⬘ than the other two proteins and further purification of this protein was not attempted. The cytochrome c556 was poorly resolved from cytochrome c 2 using a linear gradient from 160 to 320 mM NaCl, but further purification of leading and trailing edges of the elution band was similar to that for the other strains. The complete amino acid sequences of strain 2.8.1 cytochromes c 2 and c556 were determined by a combination of standard manual and automated Edman degradation plus mass spectroscopy (20, 21). Prior to sequence determination, the heme was removed by overnight treatment with acidified mercuric chloride. Details of enzyme digestions, peptide purification, amino acid analyses, and sequence determination may be obtained from the corresponding author. Peptides were isolated from the cytochromes c 2 and c556 of strains 2.10.1 and W12 and their mobilities, compositions, and N-terminal sequences were determined to detect differences from strain 2.8.1. The cytochromes c 2 and c556 from strain BSW8 were sequenced nearly as thoroughly as from strain 2.8.1 and the cytochrome c⬘ from strain W12 was completely sequenced except for a missing peptide in the center. Internal insertions and deletions were initially located by sequence alignment and their precise sizes and locations were determined by superposition of the three-dimensional structures of horse cytochrome c (22), Rb. capsulatus cytochrome c 2 (23), Paracoccus denitrificans cytochrome c 2 (24), and Rb. sphaeroides cytochrome c 2 (25). Usually, automatic structural alignment is not possible where the proteins do not have the same numbers of amino acid residues. Insertions and deletions normally occur outside regions of secondary structure. Therefore, regions of conserved secondary structure such as N-terminal and C-terminal helices were manually aligned in the first iteration followed by alignment of additional residues roughly superimposed by the first alignment. In those instances where the
CYTOCHROME SEQUENCES AND PHYLOGENY
27
FIG. 1. Evidence for the amino acid sequence of Rhodobacter sulfidophilus strain 2.8.1 cytochrome c 2. Lines above and below the sequence indicate the peptides used in the sequence determination. Double arrows are for residues established by manual Edman degradation. The symbol “T” is for tryptic peptides, “H” is for thermolysin peptides, and “C” is for chymotryptic peptides.
insertions may have caused localized rearrangements and then the elements of secondary structure on either side of the insertion were aligned in a third iteration to pinpoint the exact location in each instance. Such rearrangements are minimal in cytochrome c 2, and the sizes and locations of insertions and deletions have been accurately determined. N- and C-terminal insertions and deletions of any length were ignored as insignificant because they can easily be formed through frameshift mutations.
RESULTS AND DISCUSSION
We examined four strains of Rhodobacter, which appeared to be related based upon their microbiological properties and cytochrome compositions (19). We began with strains ATH 2.8.1 and 2.10.1 from the Van Niel collection, which were marine bacteria isolated by Dr. Sidney Elsden in Britain before either the Rb. sulfidophilus genus or species were described. These strains were never typed sufficiently to distinguish them from other Rhodobacter strains and species. The evidence for the sequence of the 126-residue strain 2.8.1 cytochrome c 2 is shown in Fig. 1. We could not detect any differences in the peptides of the cytochrome c 2 from strain 2.10.1 (which was not completely sequenced) in comparison with the reference strain 2.8.1 and conclude that the sequences are either identical or involve only one or a few compensating substitutions such as Gly-Ala to Ala-Gly. Subsequently, we examined the cytochrome c 2 from Rb. sulfidophilus strain W12, isolated from the marine environment in the Netherlands (12) and we could detect only 6 differences (at positions 65, 79, 80, 83, 90, and 94) through comparison of peptides with the reference strain 2.8.1 (evidence for the sequence not shown, but see Fig. 2 for
the actual sequence). Rhodobacter strain BSW8 was isolated from the beach at Scripps Institution of Oceanography in La Jolla California by Dr. Paul Weaver and the sequence of the cytochrome c 2 proved sufficiently different from the reference strain to warrant a more complete study (sequence shown in Fig. 2). This sequence is 13% different (17 substitutions) from that of strain 2.8.1. We compared the Rb. sulfidophilus cytochromes c 2 with other species as shown in the alignment of Fig. 2. Rb. sulfidophilus cytochrome c 2 averages about 60% difference from other species and is not significantly closer to any one of them by this criterion. This value is similar to the average for all cytochromes c 2 and does not provide any information on specific relationships because divergence is approximately balanced by convergence (7, 11). Therefore, Rb. sulfidophilus provides an ideal opportunity to evaluate alternative means of comparing sequences. Where there are large sequence differences among species, as in the present case, the numbers, sizes, and locations of insertions and deletions can be more informative than percentage identities (7, 8, 11). Because there are several three-dimensional structures for cytochromes c 2, the insertions and deletions have been accurately located as shown in Fig. 3. Rb. sulfidophilus cytochrome c 2 contains 3- and 8-residue insertions (at positions 69 –71 and 93–100 of the alignment) and a single-residue deletion (at position 109) relative to the smaller cytochromes c 2 and mitochondrial cytochromes c as highlighted in Figs. 2 and 3. Other species of cytochrome c 2 with this pattern of insertions and dele-
28
AMBLER ET AL.
FIG. 2. Alignment of cytochrome c 2 sequences based upon three-dimensional structures where available. (1) Rb. sulfidophilus strain 2.8.1 (this work), (2) Rb. sulfidophilus strain W12 (this work), (3) Rb. sulfidophilus strain BSW8 (this work), (4) Rb. sphaeroides strain 2.4.1 (10), (5) Rb. capsulatus strain SL (10), (6) P. denitrificans (15, 16), (7) Rb. capsulatus strain SB1003 cyt c y (soluble domain only) (42), (8) Rb. sphaeroides strain GA cyt c y (soluble domain only) (37), (9) Rb. sphaeroides strain CYCA65R7 iso-c2 (38), (10) Horse mitochondrial cytochrome c (43). Insertions and deletions relative to mitochondrial cytochrome c and most other cytochromes c 2, which are discussed in the text are boldfaced.
tions are Rhodobacter sphaeroides, Rb. capsulatus, Rhodospirillum rubrum, R. photometricum, Rhodopseudomonas palustris (10), Paracoccus denitrificans (15, 16), Thiobacillus versutus (17), Roseobacter denitrificans (26), and Rhodospirillum centenum (11). These three shared insertions and deletions in the cytochromes c 2 are highly significant when considered together and are not likely to have independently occurred more than once (11). In this sense, Rb. sulfidophilus cytochrome c 2 groups with the other Rhodobacter species as expected from microbiological properties (12) but apparently not from the 16S rRNA sequence (9). Rb. sulfidophilus cytochrome c 2 has a seven-residue insertion after the heme-binding site (at positions 29 –35 of the alignment) similar to that in Rb. spha-
eroides. However, in Rb. capsulatus, Ro. denitrificans, and P. denitrificans, there is a six-residue insertion at the same location as the seven-residue insertion and a two-residue deletion after the sixth heme ligand methionine (at positions 113 and 114 of the alignment). It should be emphasized that, in a genetic sense, large and small gaps are equally likely to occur, and occasionally small gaps result from protein and DNA sequencing errors (15, 16, 27) (these are excluded from the present analysis). However, among those gaps that are well-documented, small gaps are actually favored and large gaps are selected against, presumably due to structural constraints (28). Therefore, the six- to sevenresidue insertions are more significant than the tworesidue deletion due to their rarity (28) and distinguish the genus Rhodobacter from all other purple bacteria.
CYTOCHROME SEQUENCES AND PHYLOGENY
29
FIG. 3. First iteration superposition of the alpha carbons of the three-dimensional structures of Rb. sphaeroides cytochrome c 2 (red) (25) and horse mitochondrial cytochrome c (green) (22), as described under Materials and Methods in the standard front view towards the exposed heme edge. The locations of insertions and deletions were established from comparisons of all known structures and further iterations at each localized region. Insertions and deletions highlighted in Fig. 2 are colored cyan and are, clockwise from the upper left: the single-residue deletion (residue 109), the single-residue insertion (residue 16), the seven-residue insertion (residues 29 –35), the three-residue insertion (residues 69 –71), and the eight-residue insertion (residues 93–100).
Despite their rarity, there is a remote possibility that Rb. sulfidophilus and Rb. sphaeroides may have experienced a seven-residue insertion independently of the six-residue insertion in Rb. capsulatus and P. denitrificans. That this is unlikely is shown by the similarity between the six- and seven-residue insertions, notably a Gly at position 29, Thr at 30, and an Ile at 31. Alternatively, we view the six- to seven-residue insertion as more likely to have occurred as two consecutive events, a six-residue insertion followed by a singleresidue insertion or a seven-residue insertion followed by a single-residue deletion. The difference in size of this insertion is not very significant because, as explained above, single-residue insertions and deletions are more common than larger ones (28) and therefore more likely to have occurred more than once, such as the deletion at position 16. Thus, Rb. sulfidophilus and Rb. sphaeroides are typical Rhodobacter species, but differ by two shared small insertions or deletions from Rb. capsulatus and close relatives. This observation indicates that the separation of Rb. sulfidophilus from Rb. sphaeroides and Rb. capsulatus and creation of a new genus on the basis of 16S rRNA (9) cannot be confirmed. That is, Rb. sulfidophilus and Rb. sphaeroides cytochromes c 2 are actually closer to one another than either is to Rb. capsulatus. It is thus likely that the 16S rRNA data, which led to the creation of a new genus (9), may have been overinterpreted. Support for this possibility comes from the fact that no two published RNA trees containing these species have exactly the same branching order (1, 9, 29 –36). To determine whether or not the percentage similarity in RNA might support the cytochrome c 2 results if reinterpreted more conservatively, we analyzed the 16S rRNA data for nonsulfur purple phototrophic bacteria in the same way as for cytochrome c 2 (11). In Fig.
4, the percentage identity in binary comparisons of 16S rRNA of purple bacteria shows a bimodal distribution with means of 87.3% identity (and standard deviation of 1.5) for data between 82.9 and 91.2% identity and 93.6% (SD 1.3) for data between 91.7 and 95.8% identity. The first distribution shows the approximate limit to change for purple bacterial RNA (13%) as a whole and the second distribution primarily reflects the relationships among the Rhodobacter species and close
FIG. 4. Distribution of the percentage identities in binary comparisons of 16S rRNA from representative purple phototrophic bacteria (9, 29, 30, 33–36). Variations in the data, presumably due to differing alignments, were averaged for individual species comparisons prior to plotting, resulting in an average error of 0.4%. The bimodal data were fit with two Gaussian curves shown as solid lines.
30
AMBLER ET AL.
relatives, Erythrobacter, Roseobacter, and Paracoccus (the mean of the distribution for these genera alone is 93.8%). There are more than four standard deviations between the means of the two distributions, it is thus clear that Rhodobacter is distinct from the other genera and species of purple bacteria by this criterion. Rb. sulfidophilus shows 93.8% average identity to the other Rhodobacter species, the same as for the genus as a whole, which shows that it is a genuine Rhodobacter and should not be split off in a separate genus. Rb. sulfidophilus 16S rRNA shows greatest apparent similarity to that of Rb. adriaticus (95.3% identity) and Rb. adriaticus is most similar to Rb. sulfidophilus as reported in (9). However, this value is only one standard deviation from the mean of the second distribution, which is not very significant and also indicates that these two species belong to Rhodobacter. Purple bacterial species occasionally produce more than one cytochrome c 2. There are two cytochromes c 2 in Rb. capsulatus and three in Rb. sphaeroides. The c 2 isozyme in Rb. capsulatus, known as c y, is membranebound and is also present in Rb. sphaeroides (37). It is able to substitute for the soluble c 2 in photosynthetic electron transfer when the soluble c 2 gene is inactivated in Rb. capsulatus but not in Rb. sphaeroides (37). The cytochrome c y is more closely related to the small c 2s and mitochondrial c that do not have 3- and 8-residue insertions and a single-residue deletion, as shown in Fig. 2. The second isozyme in Rb. sphaeroides is also grouped with the “small” c 2s as shown in Fig. 2 and is capable of substituting for the soluble c 2 following mutation of a regulatory gene (38). Thus, there is no possibility that the isozymes might be confused with the soluble cytochrome and that paralogous rather than orthologous genes might have been compared. We have not observed soluble c 2 isozymes in Rb. sulfidophilus and it is unknown whether or not c y may be present in the membrane. If present, it is expected to show the same relationship to Rb. sphaeroides as found for the soluble c 2. One- and three-residue insertions/ deletions separate the two c ys, thus it should be nearly as good a molecular marker as the soluble c 2. To summarize to this point, large insertions and deletions, such as the 6-, 7-, and 8-residue insertions, and the simultaneous occurrence of two or more smaller insertions in cytochrome c 2 is highly significant. They establish Rhodobacter as a genus and separate Rb. sulfidophilus and Rb. sphaeroides on the one hand from Rb. capsulatus and Paracoccus on the other (although still within the same genus). When percentage identity in 16S rRNA is analyzed conservatively, it supports the conclusion based upon insertions and deletions in cytochrome c 2, that sulfidophilus belongs to the genus Rhodobacter. One potential problem with analyzing insertions and deletions is the small sample size, they are decidedly
less common than amino acid substitutions. Thus, we examined a second type of cytochrome from Rb. sulfidophilus to determine if we would obtain a similar result as with cytochrome c 2. The sequence of Rb. sulfidophilus strain 2.8.1 cytochrome c556 was completely determined as shown in Fig. 5. We have not been able to detect any differences in the peptides from strains 2.10.1 or W12 in comparison with the reference strain 2.8.1, suggesting identical or closely related sequences in agreement with the results for cytochrome c 2. The sequence of strain BSW8 c556 has 20 differences out of 124 or 16% divergence as shown in Fig. 6. Cytochrome c556 is a class II cytochrome, homologous to the cytochromes c⬘ with Met 15 (position 17 of the alignment) likely to be the sixth ligand to the heme. Rb. sulfidophilus Strain W12 was the only strain which yielded sufficient pure cytochrome c⬘ for analysis. We determined about 90% of the sequence (except for a missing peptide), which is also shown in Fig. 6. Low-spin homologs of cytochrome c⬘ have been sequenced from Rps. palustris (14), Agrobacterium tumefaciens (39), Bradyrhizobium japonicum (40), and Rb. sphaeroides (41). The high-spin examples, known as cytochromes c⬘, are more often observed in the purple bacteria than are the low-spin examples and are present in higher concentrations (19). Rb. sulfidophilus cytochrome c556 is clearly more abundant than is cytochrome c⬘, thus an exception to the rule. Rb. sulfidophilus cytochrome c556 averages about 75% difference to all other class II cytochromes (which is near the mean for all pairwise comparisons) and is not significantly closer to any other species on a percentage basis. Furthermore, Rb. sulfidophilus cytochrome c⬘ is no closer to the c556 than it is to other species of c⬘ by this measure. This is consistent with the observation with cytochrome c 2 that percentage identities do not provide much information on specific relationships where there is a large amount of change relative to the average for that gene or protein. Insertions and deletions are present in the class II cytochromes such as c556, but they are usually small (i.e., relatively less significant than those in cytochrome c 2). Nevertheless, analysis of insertions and deletions indicate that the low-spin isozymes of cytochrome c⬘ probably arose independently in Rb. sulfidophilus, Rb. sphaeroides, and Rps. palustris from high-spin precursors. CONCLUSIONS
Protein or nucleic acid sequences may be used at several levels of analysis. They can establish homology between two genes or proteins and they can also be used to determine whether newly isolated strains are the same or different from previously described organisms. In the present study, we have used the sequences of the photosynthetic cytochromes in both
CYTOCHROME SEQUENCES AND PHYLOGENY
31
FIG. 5. Evidence for the amino acid sequence of Rb. sulfidophilus strain 2.8.1 cytochrome c556. Lines above and below the sequence represent peptides used in the sequence determination. Single and double arrows represent sequences determined by automated and manual Edman degradation. Peptide nomenclature is as follows: “T” is for trypsin, “C” is for chymotrypsin, “D” is for Pseudomonas Asp-N protease, “nx” is for cyanogen bromide, “H” is for thermolysin, and “F” is for Staphylococcus Glu-C protease.
ways. We have shown that cytochrome c556 is a class II cytochrome, a result which is not obvious from spectral and redox properties. We have shown that the Elsden strains of purple bacteria in the Van Niel collection belong to the species Rb. sulfidophilus. Sequences can also be used to quantify the degree of similarity among isolates, but only where there is a small amount of change relative to the whole. Strain BSW8 is a good example in that the cytochrome c 2 is 13% different from the nearest homolog (where the average for all c 2 s is about 60% difference) and cytochrome c556 is 16% different from the same reference strain (where the average for all class II cytochromes is about 75% difference). Thus, these two proteins show good agreement in identifying strain BSW8 as a separate but closely related species to Rb. sulfidophilus. That is, only 17% of the residues have been substituted, which on average are able to change (13 divided by 60% of 126 ⫽ 17% and 16 divided by 75% of 124 ⫽ 17%). It is likely that any
other gene that has not been acquired through recent gene transfer, including that for rRNA, will show the same result. That is, a 17% substitution when the limiting variation is used as a reference. Thus, the 16S rRNA for strain BSW8 should be no more than 2% different from that of other sulfidophilus strains. Only insertions and deletions, being less common than amino acid substitutions, can be used to show that Rb. sulfidophilus is actually closer to Rb. sphaeroides than to Rb. capsulatus and should not be excluded from the genus Rhodobacter. To be even more precise about the relationship of Rb. sulfidophilus to other species will require analysis of additional genes and preferably, complete genome sequences. We can safely predict from what we have learned about cytochromes c 2 and c556, that Rb. sulfidophilus will have more genes in common with Rb. sphaeroides than either will have in common with Rb. capsulatus. Until then, name changes are premature.
32
AMBLER ET AL.
FIG. 6. Alignment of class II cytochromes. (1) Rb. sulfidophilus strain 2.8.1 cyt c556 (this work), (2) Rb. sulfidophilus strain BSW8 cyt c556 (this work), (3) Rb. sulfidophilus strain W12 cyt c⬘ (this work), (4) Rb. sphaeroides strain 2.4.1 cyt c⬘ (14), (5) Rb. sphaeroides strain 2.4.1 cyt c554 (41), (6) Rb. capsulatus strain SP7 cyt c⬘ (14), (7) Rb. capsulatus strain SB1003 orf 1662 [http://Rhodol.uchicago.edu/ capsulapedia], (8) Rps. palustris strain 2.1.37 cyt c556 (14), (9) Rps. palustris strain 2.1.37 cyt c⬘ (14). The sixth heme ligand Met is boldfaced.
ACKNOWLEDGMENT This work was supported in part by a grant from the National Institutes of Health (GM21277).
REFERENCES 1. Woese, C. R. (1987) Microbiol. Rev. 51, 221–271. 2. Koonin, E. V., Mushegian, A. R., Galperin, M. Y., and Walker, D. R. (1997) Mol. Microbiol. 25, 619 – 637. 3. Gupta, R. S. (1998) Microbiol. Mol. Biol. Rev. 62, 1435–1491. 4. Gupta, R. S. (1998) Mol. Microbiol. 29, 695–707. 5. Pennisi, E. (1998) Science 280, 672– 674. 6. Pennisi, E. (1999) Science 284, 1305–1307. 7. Meyer, T. E., Cusanovich, M. A., and Kamen, M. D. (1986) Proc. Natl. Acad. Sci. USA 83, 217–220. 8. Meyer, T. E., Van Beeumen, J. J., Ambler, R. P., and Cusanovich, M. A. (1996) in Origin and Evolution of Biological Energy Conversion (Baltscheffsky, H., Ed.), pp. 71–108, VCH, NY.
9. Hiraishi, A., and Ueda, Y. (1994) Int. J. Syst. Bacteriol. 44, 15–23. 10. Ambler, R. P., Daniel, M., Hermoso, J., Meyer, T. E., Bartsch, R. G., and Kamen, M. D. (1979) Nature 278, 659 – 660. 11. Samyn, B., Fitch, J., Meyer, T. E., Cusanovich, M. A., and Van Beeumen, J. J. (1998) Biochim. Biophys. Acta 1384, 345– 355. 12. Hansen, T. A., and Veldkamp, H. (1973) Arch. Microbiol. 92, 45–58. 13. Imhoff, J. F., Truper, H. G., and Pfennig, N. (1984) Int. J. Syst. Bacteriol. 34, 340 –343. 14. Ambler, R. P., Bartsch, R. G., Daniel, M., Kamen, M. D., McLellan, L., Meyer, T. E., and Van Beeumen, J. J. (1981) Proc. Natl. Acad. Sci. USA 78, 6854 – 6857. 15. Ambler, R. P., Meyer, T. E., Kamen, M. D., Schichman, S. A., and Sawyer, L. (1981) J. Mol. Biol. 147, 351–356. 16. Van Spanning, R. J. M., Wansell, C., Harms, N., Oltmann, L. F., and Stouthamer, A. H. (1990) J. Bacteriol. 172, 986 –996. 17. Ubbink, M., Van Beeumen, J., and Canters, G. W. (1992) J. Bacteriol. 174, 3707–3714.
CYTOCHROME SEQUENCES AND PHYLOGENY 18. Katayama, Y., Hiraishi, A., and Kuraishi, H. (1995) Microbiol. 141, 1469 –1477. 19. Bartsch, R. G. (1991) Biochim. Biophys. Acta 1058, 28 –30. 20. Ambler, R. P., Daniel, M., McLellan, L., Meyer, T. E., Cusanovich, M. A., and Kamen, M. D. (1987) Biochem. J. 248, 365–371. 21. Ambler, R. P., Daniel, M., Melis, K., and Stout, C. D. (1984) Biochem. J. 222, 217–227. 22. Bushnell, G. W., Louie, G. V., and Brayer, G. D. (1990) J. Mol. Biol. 214, 585–595. 23. Benning, M. M., Wesenberg, G., Caffrey, M. S., Bartsch, R. G., Meyer, T. E., Cusanovich, M. A., Rayment, I., and Holden, H. M. (1991) J. Mol. Biol. 220, 673– 685. 24. Benning, M. M., Meyer, T. E., and Holden, H. M. (1994) Arch. Biochem. Biophys. 310, 460 – 466. 25. Axelrod, H. L., Feher, G., Allen, J. F., Chirino, A. J., Day, M. W., Hsu, B. T., and Rees, D. C. (1994) Acta Crystall. D50, 596 – 602. 26. Okamura, K., Miyata, T., Iwanaga, S., Takamiya, K., and Nishimura, M. (1987) J. Biochem. 101, 957–966. 27. Ridout, C. J., James, R., and Greenwood, C. (1995) FEBS Letts. 365, 152–154. 28. Pascarella, S., and Argos, P. (1992) J. Mol. Biol. 224, 461– 471. 29. Kawasaki, H., Hoshino, Y., Hirata, A., and Yamasato, K. (1993) Arch. Microbiol. 160, 358 –362. 30. Kawasaki, H., Hoshino, Y., Kuraishi, H., and Yamasato, K. (1992) J. Gen. Microbiol. 38, 541–551. 31. Kawasaki, H., Hoshino, Y., and Yamasato, K. (1993) FEMS Microbiol. Letts. 112, 61– 66.
33
32. Hiraishi, A., and Ueda, Y. (1995) Int. J. Syst. Bacteriol. 45, 319 –326. 33. Yurkov, V., Stackebrandt, E., Holmes, A., Fuerst, J. A., Hugenholtz, P., Golecki, J., Gadon, N.,, Gorlenko, V. M., Kompantseva, E. I., and Drews, G. (1994) Int. J. Syst. Bacteriol. 44, 427– 434. 34. Hiraishi, A., and Ueda, Y. (1994) Int. J. Syst. Bacteriol. 44, 665– 673. 35. Hiraishi, A., Urata, K., and Satoh, T. (1995) Int. J. Syst. Bacteriol. 45, 226 –234. 36. Pfennig, N., Luensdorf, H., Sueling, J., and Imhoff, J. F. (1997) Arch. Microbiol. 168, 39 – 45. 37. Myllykallio, H., Zannoni, D., and Daldal, F. (1999) Proc. Natl. Acad. Sci. USA 96, 4348 – 4353. 38. Rott, M. A., Witthuhn, V. C., Schilke, B. A., Soranno, M., Ali, A., and Donohue, T. J. (1993) J. Bacteriol. 175, 358 –366. 39. Van Beeumen, J., Tempst, P., Stevens, P., Bral, D., Van Damme, J., and De Ley, J. (1980) Protides Biol. Fluids 28, 69 –74. 40. Tully, R. E., Sadowsky, M. J., and Keister, D. L. (1991) J. Bacteriol. 173, 7887–7895. 41. Bartsch, R. G., Ambler, R. P., Meyer, T. E., and Cusanovich, M. A. (1989) Arch. Biochem. Biophys. 271, 433– 440. 42. Daldal, F., Cheng, S., Applebaum, J., Davidson, E., and Prince, R. C. (1986) Proc. Natl. Acad. Sci. USA 83, 2012–2016. 43. Margoliash, E., Smith, E. L., Kreil, G., and Tuppy, H. (1961) Nature 192, 1125–1127.