Rampant Horizontal Transfer and Duplication of Rubisco ... - LaCIM

1 downloads 0 Views 1MB Size Report
Charles F. Delwiche' and Jeflrey D. Palmer. Department of ... Reith 1995; Palenik and Swift 1996). In contrast ... and reprints: Charles Delwiche, De- partment of ...
Rampant Horizontal Transfer and Duplication Eubacteria and Plastids Charles F. Delwiche’ Department

of Biology,

of Rubisco Genes in

and Jeflrey D. Palmer

Indiana University,

Bloomington

Previous work has shown that molecular phylogenies of plastids, cyanobacteria, and proteobacteria based on the rubisco (ribulose-l$bisphosphate carboxylase/oxygenase) genes rbcL and rbcS are incongruent with molecular phylogenies based on other genes and are also incompatible with structural and biochemical information. Although it has been much speculated that this is the consequence of a single horizontal gene transfer (of a proteobacterial or mitochondrial rubisco operon into plastids of rhodophytic and chromophytic algae), neither this hypothesis nor the alternative hypothesis of ancient gene duplication have been examined in detail. We have conducted phylogenetic analyses of all available bacterial rbcL sequences, and representative plastid sequences, in order to explore these alternative hypotheses and fully examine the complexity of rubisco gene evolution. The rbcL phylogeny reveals a surprising number of gene relationships that are fundamentally incongruent with organismal relationships as inferred from multiple lines of other molecular evidence. On the order of six horizontal gene transfers are implied by the form I (L&J rbcL phylogeny, two between cyanobacteria and proteobacteria, one between proteobacteria and plastids, and three within proteobacteria. Alternatively, a single ancient duplication of the form I rubisco operon, followed by repeated and pervasive differential loss of one operon or the other, would account for much of this inconaruitv. _ _ In all probability, the rubisco operon has undergone multiple events of both horizontal gene transfer and gene duplication in different lineages.

Introduction Plastid bisphosphate

phylogenies based on rubisco (ribulose- 1,5carboxylase/oxygenase) conflict with those

based on other genes (Boczar, Delaney, and Cattolico 1989; Valentin and Zetsche 1989; Martin, Somerville, and Loiseaux-de Go& 1992; Douglas 1994; Delwiche, Kuhsel, and Palmer 1995). Analyses based on 16s and 23s rRNA, t&A, atpB, rpoC1, and HSP60 each indicate strongly that all plastids are derived from cyanobacterial ancestors and, possibly, from a single common plastid ancestor (fig. 1; see also Bhattacharya and Medlin 1995; Reith 1995; Palenik and Swift 1996). In contrast, analyses based on rbcL (the gene for the large subunit of rubisco) divide plastids into two groups of widely disparate bacterial affinity. As expected, the rbcL genes of two of the three primary plastid lineages (chlorophytes, which include green algae and land plants, and glaucophytes, as exemplified by Cyanophora paradoxa, whose plastids retain a peptidoglycan cell wall) group with those of cyanobacteria. However, the rbcL genes from red algae and related secondary plastid lineages (chromophytes, cryptomonads, etc.) are much more closely related to those of certain proteobacteria (purple bacteria) than to cyanobacteria. The gene for the small subunit of rubisco, rbcS, shows the same pattern, although it has not been as widely sampled as rbcL and is too short and variable to provide enough conserved characters to allow detailed phylogenetic reconstruction (Assali et al. ‘Present address: Department of Plant Biology, Maryland

at College

University

of

Park.

Key words: gene transfer, gene duplication, ribulose- 1,Sbisphosphate carboxylase, rub&o, &CL, mosaic evolution, chloroplast phylogeny, molecular phylogeny, cyanobacteria, proteobacteria. Address for correspondence and reprints: Charles Delwiche, Department of Plant Biology, H. J. Patterson Hall, University of Maryland. College Park, Maryland 70742-S 15. E-mail: delwiche@bio. indiana.edu. Mr~l. Biol. EL’/.13(6):X73-882. 0 1996 by the Soaety

1996 for Molecular Biology

and Evolution.

ISSN: 0737.4038

1991; Morden et al. 1992). No genes other than the rubisco genes have been identified that share this phylogenetic pattern. The discrepancy between phylogenies based on rbcL and those based on other genes has been noted and discussed by several authors in the context of plastid phylogeny (e.g., Boczar, Delaney, and Cattolico 1989; Valentin and Zetsche 1989; Assali et al. I99 1; Douglas and Turner 1991; Martin, Somerville, and Loiseaux-de Gob 1992; Morden et al. 1992; Delwiche, Kuhsel, and Palmer 1995). Although proposed mechanisms have

varied considerably in their details, most authors have favored a scenario involving a single horizontal gene transfer of a rubisco operon from a proteobacterium to a cyanobacterium or early plastid. An alternative explanation for the discrepancy is that ancestral proteobacteria and cyanobacteria possessed two rubisco operons (Martin, Somerville, and Loiseaux-de Go& 1992). The different rbcL lineages present in plastids would then reflect retention of different copies of the gene in the two main plastid lineages present in the rbcL tree. To distinguish between the alternate possibilities of gene transfer and gene duplication, we have analyzed all available eubacterial rbcL sequences and representative plastid rbcL sequences. We show that the evolutionary history of rubisco is far more complicated than previously envisioned; to reconcile it with organismal phylogeny requires many more events of lateral gene transfer and/or gene duplication and differential loss than previously suspected. Materials

and Methods

Forty-eight rbcL amino acid sequences (table 1) were aligned with the GCG (Genetics Computer Group 1991) utility PILEUP The alignment was checked manually and reconciled with alignments based on secondary structural data (Schneider et al. 1990). The total aligned length was 532 amino acids, of which 497 were 873

874

Delwiche

and Palmer

FIG. I.-Bootstrap support for relevant plastid and eubacterial groups from representative molecular phylogenetic studies. Included are those genes for which both red-like and green-like plastid and representative eubacterial sequences have been examined. For each gene, values in the left column indicate bootstrap support for monophyly of the group in question, while values in the right column indicate the number of taxa represented in that group. The total number of taxa included in each study is indicated at the head of the column. Cases that have not been tested (NT) or where bootstrap values are below 507~ are indicated by shading. (a) Bootstrap support for this clade was above 50% with only one analytical method, maximum likelihood. (b) The proteobacteria were not monophyletic, but bootstrap support for the observed topology was low (40%). (c) One expected taxon, Nrisseria gonorrhoeae, did not fall among the B-proteobacteria. The studies used to assemble this table were: 16s rRNA, *Bhattacharya and Medlin (1995), :Eisen (1995); 23s rRNA, De Rijk et al. (1995); tufA, Delwiche, Kuhsel, and Palmer (1995); at@, Douglas and Murphy (1994);HSP60, Viale et al. (1994); recA. Eisen (1995). Molecular phylogenetic studies with generally congruent conclusions include: 16s rRNA, Douglas and Turner (1991). DeLong, Frankel, and Bazylinksi (I 993). Ludwig et al. (1993) Distel and Cavanaugh (1994) Giovannoni et al. (1988), Olsen, Woese, and Overbeek (1994), Van de Peer et al. (1994); 23s rRNA. Somerville et al. (1993): r&A. Morden et al. (1992); atpB, Morden et al. (I 992). Leitsch and Kowallik (1992) Martin, Jouannic, and Loiseaux-de Goer (1993); HSP60, Viale and Arakaki (1994); rrcA, Lloyd and Sharp (1993).

included in the analysis. All sequences were full length (ca. 470 amino acids) except for those from Prochloron and Prochlorococcus (228 amino acids). Percent amino acid identities of the genes rbcL and rbcS were calculated, excluding gaps, for selected taxa with the GCG utility DISTANCES. The alignment is available at http: //www.bio.indiana.edu/-jpalmer/rubisco-evolution/. Phylogenetic analyses were performed with PAUP v. 3.1.1 (Swofford 1993) and PAUP* v.4.0d31 (Swofford, personal communication). The most parsimonious trees were determined with 100 heuristic searches with random addition sequences. Bootstrapping was performed with 100 replicates, evaluating each replicate with 10 heuristic searches with random addition sequences. Neighbor-joining trees were determined and bootstrap analyses were performed using mean character distances with PAUP* v.4.0d3 1. Results and Discussion Phylogenetic Analysis Maximum-parsimony phylogenetic analysis of rbcL amino acid sequences found 24 equally parsimonious trees (fig. 2). Six form II rbcL sequences were included as outgroups to root the tree (fig. 2). This form of rubisco, composed of two large subunits but no small subunits, is restricted to certain proteobacteria and dinoflagellates (Gibson and Tabita 1977; Roy and Nier-

zwicki-Bauer 1991; Morse et al. 1995), whereas form I rubisco (composed of eight large and eight small subunits)-the subject of this study-is more widespread, present in all photosynthetic plastids (except certain dinoflagellates) and cyanobacteria, and widely in proteobacteria. The large subunits of form I and II rubisco, while readily alignable and clearly homologous (Schneider et al. 1990; Gibson, Falcone, and Tabita 1991), are so divergent (fig. 3) that one type is customarily used as outgroup for the other type (e.g., Martin, Somerville, and Loiseaux-de Go& 1992). The conclusions of the study are not, however, dependent upon rooting of the tree; in analyses with slightly fewer taxa in which the form II sequences were excluded, the unrooted tree was essentially identical to the form I ingroup in the rooted tree. The strict consensus of the 24 trees (indicated by bullets in fig. 2) is consistent with previous analyses of bacterial and algal form I rbcL sequences, dividing taxa into two major groups, termed here “green-like” and “red-like.” The green-like group includes two distinct sets of rbcL sequences, those from cyanobacteria and the plastids of Cyanophoru, green algae, plants, and euglenophytes (these correspond to the “type IB” genes in the classification of Tabita [ 1995]), and those from all examined y-proteobacteria and one (Y- and one B-proteobacterium (type IA). The red-like group also includes two distinct sets of sequences, those from red algal, brown algal (sensu lato), and cryptomonad plastids (type ID), and those from three (Y- and three B-proteobacterial sequences (type 1C). The green-like and red-like rbcL groups are quite distinct as measured both by branch lengths on phylogenetic trees (fig. 2) and by direct comparisons of amino acid sequences (fig. 3). Within-group amino acid sequence identities among the divergent bacterial and eukaryotic taxa examined here are 69%-92%, whereas between-group identities are much lower, in the range of 53%-60% (fig. 3, above diagonal). The same general pattern of sequence relationships also holds for the small subunit of rubisco; amino acid identities here are consistently higher within the red (52%-70%) and green (33%-70%) groups than between the two groups (24%39%; fig. 3, below diagonal). The major features of the rbcL tree (fig. 2) are strongly supported, and differences among the 24 shortest trees involve rearrangements within a few terminal groups. The critical branches defining red-like and green-like rubisco groups have 100% bootstrap values, as does the branch separating form I and II rubisco. There is also strong support for the monophyly of both green-like and red-like plastids, green-like proteobacterial sequences, and, except for the uncertain position of the c-Y-proteobacterial “Mn-oxidizing bacterium,” the Tred-like proteobacterial sequences (fig. 2). Comparable trees (except for the position of the dinoflagellate Gonyaulux; see below) and bootstrap support were found in neighbor-joining analyses using mean character distances (data not shown).

Lateral Transfer

Table 1 Taxa Used in Phylogenetic

Analysis,

Genbank

Accession

and Duplication

Numbers for &CL, and Classification

Taxon and Strain

of Rubisco

Genes

875

Based on rRNA

rbcL No.

Group

Ml744 U20584 U20585 D90204 M26396 M34536 D28135 D4362 1 D43622 L32182 L22885 D30764 U23 145 Wagner M64624 X00286 L37437 L42940 MS5061 x70355 Xl7252

Beta” Beta” Beta” Gammah Gammab Uncertain’ Gammad Gammad Gammad Alphat Alpha’ Uncertain’ Alpha” Alpha” Alpha” Alpha” Betag Betag Gamma” Gamma” Alpha’

I Proteobacteria Alcaligenes eutrophus ATCC 17707 (chromosomal) Alcaligenes eutrophus ATCC 17699 (chromosomal) Alcaligenes eutrophus ATCC 17699 (plasmid). Chromatium vinosum ATCC 17899 (rbcL). Chromatium vinosum ATCC 17899 (rbcA). Endosymbiont of Alvinoconcha (Thiobacillus sp.?) Hydrogenovibrio marinus MH-110 (form II) Hydrogenovibrio murinus MH-110 (form I rbcLI) Hydrogenovibrio marks MH- 1 IO (form I rbcL2) Mn-oxidizing bacterium SI85-9al Nitrobacter vulgaris T3 Pseudomonas hydrogenothermophila str.TH- 1 Rhodobacter capsulatus ATCC 11166 (form II) Rhodobacter sphaeroides ATCC 17023 (form II) Rhodobacter sphaeroides ATCC 17023 (form I) Rhodospirillum rubrum (form II). Thiobacillus denitrijicans ATCC 25259 (form II) Thiobacillus denitrificans ATCC 25259 (form I). Thiobucillus ferrooxidans Fe1 Thiobacillus ferrooxidans ATCC 19859 Xanthobacter jluvus H4- 14 II. Cyanobacteria Anabuenu 7120 ATCC 27893 Anacystis nidulans PCC 6301 Prochlorococcus mnrinus GP2. Prochloron didrmni (not cultivated) Prochlorothrix hollandicd Swechococcus sp. A 1 Synrchocystis sp. ATCC 27 184

501540 X03220 D21833 D21834 x57359 D13539 X65960

...... ...... ......

III. Plastids Ahnfeltia fastigiatd Antithamnion sp>. Btypsis maximd. Chlumydomonas morwusiiJ. Chlorelln ellipsoidea C-27 Coleochuete orbicularis var. pondspride Cryptomonas + Cynidium caldarium RK-I Cyanophora paradoxa 2980 IPP Goettingen Cylindrotheca sp. N 1 Ectocarpus siliculosud Euglena gracilis Pringsheim Z. Gonyuulux p&edru 70 (form II). Marchantiu polymorphd. Nicotiuna ucuminatclJ. Olisthodiscus luteus’ Op:u sntivd Porphyridium aerugineumj Pseudotsugu menzie.siiJ Pyramimonns tetrarhyncus K-0002 A. e~rtrophu\ ATTCC17697

* SSU

rRNA

sequence

for

h SSU

rRNA

sequence

available

L Affiliation (’ rRNA

mformation

c Based on analysis

F Signature h Probable

in ribosomal

database

project

tree (RDP;

Redh Red Green Green Green Green Cryptomonad Red Glaucophyte Diatom Brown Euglenophyte Dinoflagellate Plant Plant Diatom Plant Red Plant Green

http://rdp.life.uiuc.edu).

in RDI?

unknown.

sequence

’ Probable

UO4167 X54532 X55877 Ml5842 D 10997 L13477 X62349 X55524 x53045 M59080 X52503 Ml2109 L41063 X04465 Ml6896 X61918 DO0207 x17597 X52937 L34833

phylogenetic analysis

phylogenetic

’ Phylogenetic

Y. Igarashi,

16s

placement

personal

rRNA sequence (R.

based on Serwaldt

commumcatmn. Caspi,

personal

et al. (1982)

communication).

and E. Bock.

personal

communicatmn.

of Woese (1987).

’ Probable phylogenetic 1Strain not specified. groups

tom~Bacillariophyceae;

from

of complete

placement

bahed on Kusano

placement

based on Wiegel

indicated

for plastids

Brown-Phaeophyta;

et al. (1991). (1992).

are: Red-Rhodophyta; Dinoflagellate-Pyrrhophyta;

Green-Chlorophyta; Plant-Embryophyta.

Cryptomonad-Cryptophyta;

Glaucophyte-Glaucocystophyta;

Dia-

876

Delwiche

and Palmer

1a

3 63

I

Rhodobacter sphaeroides II +--,

Proteobacteria

Thiobacillus denitrificans /I +-------------/. Hydrogenovibrio II +------------1.

1

Dinoflagellate plastid

3 p Proteobacteria 3 y Proteobacteria

3

a Proteobacteria

1p

Form II Rubisco

-

Proteobacteria

=

100

1

a Proteobacteria

Red type Form I Rubisco

I Red and Brown Plastids

1

1y 3

Cyanobacteria

-

Proteobacteria

-

:

100

q-

50 steps

i / / 3 y Proteobacteria

Hydrogenovibrio Ll +--_j-J “-wdomonas’ i//us ferrooxidans fe 1’ Thiobacillus ferr. 19859 denitrificans I

Ghr~ytyp?;

nechococcus labaena - Anacystis .rochlorothrix - Synechocystis Prochloron 7 Cyanophora

1

3

3 3

n Proteobacteria y Proteobacteria 8 Proteobacteria y Proteobacteria

1

Cyanobacteria

r

1

+I~

/! 3

1 3

L PyFamimonas 721 L_Chlamydomonas h/ore//a Bryopsis sochaete darchantia

Green type Form I Rubisco

Glaucophyte Plastid

Green Plastids

FIG. 2.-Maximum parsimony tree of rhcL amino acids from all available bacteria and representative plastids. The tree shown is one of 24 shortest trees of length 2.422, selected arbitrarily. Classification of taxa based on 16s rRNA and other evidence is indicated to the right. Bootstrap values above 60% are indicated above the branch, and branches which were not present in the strict consensus of the 24 shortest trees are indicated by a bullet (0) below the branch. Branch lengths correspond to the inferred number of character state changes on each branch (see scale bar). Dashed, arrow-headed lines connect multiple rbcL sequences determined from the same taxon. *Ribosomal RNA sequences are not available for this strain, and proteobacterial subgroup may be suspect. **Bootstrap analysis placed Chronzcztium L in a clade with Hwk~~mor~ihrio L2 and Prochlorococcus with 60% bootstrap support, but this topology did not occur among the shortest trees.

Incongruence Phylogeny

with Plastid and Proteobacterial

The division of form I sequences into red-like and green-like groups is strongly in conflict with all other studies of chloroplast phylogeny; there is little doubt that all plastids are ultimately cyanobacterial in origin, and reasonable evidence that they are monophyletic in origin (fig. 1). Furthermore, the division of proteobacteria into two widely disparate groups is similarly incongruent with current understanding of bacterial evolution. There is strong evidence from 16s and 23s rRNA, which is generally supported by data from protein genes (fig. l), that (Y-, p-, and y-proteobacteria together form a monophyletic group, with strongly supported subgroups that include the (Y-, p-, and p-/y-proteobacteria (the y-proteobacteria are split into a paraphyletic group by the P-proteobacteria). Several of these monophyletic groups are divided by the rbcL tree between the red-like and green-like sequence groups. For example, the rbcL tree places the a-proteobacterium Nifrobacter vulgaris (Strecker et al. 1994) among the green-type sequences but Rhodobacter, Xmthobacter,

and an unnamed Mn-oxidizing bacterium, also a-proteobacteria, among the red-types. Similarly, rbcL puts the P-proteobacterium Alcaligerzes in the red-like group, but Thionbacillus denitrijcans (Hernandez et al. 1996) in the green-like group. Aspects of the rbcL phylogeny are also incongruent with eubacterial gene phylogenies within the red-like and green-like groups. In particular, the c1- and p-proteobacterial sequences are intermixed in a manner that is inconsistent with the monophyly of the a-proteobacteria, and the cyanobacterium Proch1orncoccu.s falls solidly among the y-proteobacteria, contrary to other phylogenetic information. The conflict between rbcL and other sources of phylogenetic information is unlikely to reflect methodological error. Although poorly resolved data and weak analytical methods can create a false appearance of conflict by failing to find the correct tree, and sampling error (Cao et al. 1994; Cummings, Otto, and Wakeley 1995). long branch attraction (Felsenstein 1978), and convergent evolution can cause phylogenetic analyses to be positively misleading, the robust phylogenetic tree and

Lateral Transfer

RhodobacterForm

and Duplication

of Rubisco

Genes

877

II

Hydrogenovibrio

7

Rhodobacter Form 70

Xanthobacter

69

71

Alcaligenes Cryptomonas Porphyridium Ectocarpus Olisthodiscus Chromatium A/B Chromatium L/S Mtrobacter Prochlorothrci

71

I

Anabaena

33

Cyanophora

32

Euglena Chlamydomonas

! 36 1 35 I ’ 32 ! 35 / ’

-31131 34 34

39 _::_._. 38 ; 35 38 38 j 32

37

39 34

39

1 38

37

a-_ ..A 36 42

I\!

85

55



84

1 84 i

86

83

50

-

83

57 \



30_, 30 _-L_ 28 -_s

35

;

27

28

28 30

26.

25

I:

35

41 36

43 38

45 46

:

45 44

35

33

45

i

45;41-53

Form

:

45 SO

I

FIG. 3.-Percent amino acid identity of rubisco large (above diagonal) and small (below diagonal) subunit sequences from self xted taxa. Comparisons between the red-like and green-like form I sequence groups are shaded. rbcS is encoded in the nuclear genome of green algae and plants, but in the chloroplast genome of red-like plastids and Cyanophora.

dramatic differences (figs. 2 and 3) between green-like and red-like rubiscos make it quite unlikely that these phenomena explain the rbcL phylogeny. In some cases, apparently aberrant phylogenies result from inappropriate rooting. But even if one assumes that form II rubisco is an inappropriate outgroup, and rearranges the tree such that it and all form I rubiscos from proteobacteria (and red-like plastids and Prochlorococcus) form a monophyletic group (i.e., by rooting on cyanobacteria), only one of the phylogenetic discrepancies is removed, and this at the expense of introducing considerable rate heterogeneity throughout the tree. The assignment to proteobacterial subgroups in figure 2 of some strains, particularly those for which 16s rRNA sequences have not been determined, could be in error. However, while a few of the incongruencies noted here may be attributable to erroneous classification of a particular strain, there are rbcL sequences from unambiguously identified bacteria in both the red-like and green-like sequence groups. Thus, much or all of the conflict between the rbcL phylogeny and other information (fig. 1) must reflect a discrepancy between gene and organismal phylogeny. The most likely processes underlying the problematic rbcL phylogeny are horizontal gene transfer and ancient gene duplication coupled with differential gene loss. We will examine these two possibilities in turn, in each case first focusing on the division of rbcL sequences into redlike and green-like groups, and then on other features of the rbcL phylogeny. Horizontal Gene Transfer Between Red-like Form I Rubiscos

Green-

and

At least four independent horizontal gene transfers (fig. 4A) are required to explain the division of plastids

and proteobacteria into the green-like and red-like groups evident in figure 2. To explain plastid phylogeny, we postulate transfer of a red-like rubisco operon from a proteobacterium to a common ancestor of red and brown plastids. Given that mitochondria probably arose from within ol-proteobacteria and that this occurred prior to the origin of plastids (Gray 1992), the donor bacterium should be assignable to a modern proteobacterial group. Although the donor proteobacterium is shown in figure 4A as an o-proteobacterium, the intermixing of (Y- and P-proteobacteria in the rubisco tree (fig. 2) and the small number of characterized bacterial red-like genes make it difficult to infer the identity of the donor with confidence. Within the proteobacteria, at least three horizontal gene transfers are implied by the distribution of red- and green-like genes. A relatively simple scenario (fig. 4A) involves the transfer of a (cyanobacterial) green-like rbcL to an ancestor of y-proteobacteria early in their evolution (but after the emergence of the P-proteobacteria from within the y-proteobacteria), with subsequent transfers from this green-like y-proteobacterial lineage to ancestors of the a-proteobacterium Nitrobacter vulgaris and the P-proteobacterium Thiobacillus denitrificans (fig. 4A). A slightly different sequence of transfers would be postulated if the transfer of a green-type rubisco into the y-proteobacteria preceded the emergence of the P-proteobacteria, or if some of the taxa in figure 2 were not identified correctly. One potential source of direct evidence for horizontal gene transfer would be differences in codon usage and base composition between the putatively transferred gene and other host genes (Groisman, Saier, and Ochman 1992; Aoyama, Haase, and Reeves 1994; Haraya-

878

Delwiche

and Palmer Rubisco type Green Red

1

Proteobacteria

Proteobacteria

Proteobacteria

Cyanobacteria L

Cyanophora Red and Brown Algae Er;rnt21gae

FIG. 4.-Comparison of gene transfer and gene duplication hypotheses for red VS. green conflicts only. The topology of the two mirrorimage, wide, gray trees shows a simplified “organismal” phylogeny as inferred from phylogenetic analyses of 16s rRNA and other genes (fig. 1; see text). The black and white trees shown within the organismal phylogeny represent form I red-like and green-like rubisco, and indicate the events that would be implied if the observed distribution of these two kinds of rubisco reflects a history of horizontal gene transfer (A), or ancient gene duplication and differential loss (B). The type of rubisco present in each group is indicated in the right-center column. The figure assumes that the rbcL phylogeny and classifications shown in figure 2 are correct, Conflicts occurring entirely within green-like (i.e., Prochlorococcus) or red-like (i.e., intermixing of o- and P-proteobacterial sequences) groups are not shown. The diagrams are highly schematic, and are not intended to indicate the precise timing of the events, Many minor variations of these schemes would also be consistent with the data (see text).

ma 1994). However, in comparisons of codon usage in strains for which multiple genes are available, no cases were found where rbcL codon usage differed substantially from that of other genes (data not shown). Although codon usage does not provide any evidence in support of the horizontal transfer hypothesis, this does not rule out the possibility of gene transfer of sufficient age that the transferred gene would have fully adjusted to the recipient cell’s compositional environment (highly expressed genes such as rbcLS undergo relatively rapid amelioration; Sharp and Matassi 1994; Ochman and Lawrence 1995), or of transfer between cells of similar codon usage. Unfortunately, too few gene sequences for reliable comparisons of codon usage are available from the two taxa-Nitrobacter vulgaris (two other genes) and Thiobacillus denitri$cans (no other genes)-that seem to have undergone relatively recent gene transfer events. Other Horizontal

Gene Transfers

At least three other cases of horizontal gene transfer must be postulated to account for incongruities in the rbcL phylogeny besides those that involve the redgreen-like split. One of these seems ironclad. Recent work has shown that, unlike all other algal groups examined, the plastids of several peridinin-containing dinoflagellates (such as Gonyaulax) utilize a nucleat-encoded form II rubisco (Morse et al. 1995; Whitney, Shaw, and Yellowlees 1995; Rowan et al. 1996). Given the clearly cyanobacterial nature of the dinoflagellate

chloroplast, it is unlikely that the Gonyaulax rubisco is of plastid origin; form II rubisco is unknown among cyanobacteria. Morse et al. (1995) propose that the Gonyaular rubisco is derived from the mitochondrial genome, having first undergone gene transfer from the ancestral mitochondrion to the nuclear genome. Subsequent gene substitution presumably resulted in the expression of this nuclear-encoded gene in the chloroplast in lieu of the putative native form I rubisco. Consistent with this hypothesis, the Gonyaulax form II sequence groups within the a-proteobacteria in this, the first phylogenetic analysis performed with this sequence (fig. 2). However, neighbor-joining analysis places the Gonyaulax sequence outside all five diverse proteobacterial form II sequences with moderately high (78%) bootstrap support (data not shown). Regardless of whether the sequence is mitochondrial in origin or the product of transfer from a free-living proteobacterium, the Gonyaulax rubisco provides a clear example of successful transfer and substitution of rubisco genes in nature. A second very likely horizontal gene transfer involves the rbcL gene of the cyanobacteria Prochlorococcus and Synechococcus. The Prochlorococcus gene is part of a strongly supported (94%) clade that otherwise consists entirely of green-like proteobacterial sequences, and is distantly related to the other cyanobacterial sequences (fig. 2), while recent data from G. E Watson and E R. Tabita (personal communication) reveal that several marine species of Synechococcus con-

Lateral Transfer

tain rbcL genes that are closely related (>90% amino acid identity) to the Prochlorococcus gene. Shimada, Kanai, and Maruyama (1995) noted the similarity of the Prochlorococcus sequence to that of the Chromatium “cryptic” gene rbcL (Hydrogenovibrio was not included in their analysis) and from this inferred a phylogenetic relationship between Prochlorococcus and Chromatium. However, analyses of 16s rRNA (Urbach, Robertson, and Chisholm 1992) and rpoC1 (Palenik and Haselkorn 1992; Palenik and Swift 1996) indicate that other strains of Prochlorococcus are cyanobacterial, and that several cyanobacteria have independently acquired (or retained) the chlorophyll a/b pigment system characteristic of “prochlorophytes.” There is no evidence linking Prochlorococcus or Synechococcus with the proteobacteria other than their rbcL sequences. Therefore, these cyanobacteria probably acquired their rubisco genes by lateral transfer from the green-like proteobacterial group. Finally, as previously noted, the three a-proteobacterial sequences within the red-like clade (fig. 2) do not form the expected (fig. 1) monophyletic group. Assuming correct identification of bacterial strains, the simplest interpretation of this incongruity is that one of the three groups (RhodobacterlXanthobacter, Alcaligenes, or the Mn-oxidizing bacterium) acquired its form I rubisco genes by lateral transfer. Neither Rhodobacter nor Alcaligenes have codon usage patterns that suggest recent gene transfer (data not shown), indicating that, if transferred, these genes either have already adapted to codon usage in the host, or were acquired from a bacterium with similar codon usage. Too few genes are available from Xanthobacter and the Mn-oxidizing bacterium for reliable analysis. Gene Duplication Form I Rubiscos

and Loss for Green-

and Red-like

If a single duplication of the rubisco operon occurred prior to the divergence between cyanobacteria and proteobacteria, and if one or the other copy of the gene were lost (or simply remained undetected) in each of the lineages that have been examined, then the division of the rbcL phylogeny into green-like and red-like groups could be fully explained (fig. 4B). Under this interpretation the green-like and red-like kinds of rbcL would be paralogs rather than orthologs (and xenologs), and the distribution of taxa within the two groups would reflect which copy had been retained. The illustrative diagram of figure 4B clearly oversimplifies the complexity of the duplication/loss hypothesis. For example, if one takes the phylogeny of figure 2 at face value, then four separate gene losses within the cyanobacteria are required; red-like rubisco must have been lost independently from each lineage that branched prior to the origin of plastids. Although rRNAbased phylogenies of cyanobacteria are poorly resolved, even that study (Nelissen et al. 1995) which postulates the earliest origin of plastids within the group would require three independent losses of red-like rubisco among cyanobacteria rather than the one shown in figure 4B. Similarly, the numbers of losses among proteobacteria (fig. 4B) are probably also underestimates and are

and Duplication

of Rubisco

Genes

879

expected to rise as additional proteobacterial rbcL sequences are determined. The duplication hypothesis also implies a long coexistance of green- and red-like rubiscos in several lineages, as indicated by dual black and white horizontal lines in figure 4B. The duplicate genes in the cyanobacterial lineages would necessarily have persisted jointly for l-2 billion years, at least until after the origin of plastids and the separation between the red-like and green-like lineages (Knoll and Golubic 1992; Schopf 1993). If this is correct, the long persistence of both redgroup and green-group rubisco genes during eubacterial evolution implies that some extant cyanobacteria and proteobacteria might still contain both types of rubisco genes. To date, no organism has been identified that has both red-like and green-like rubisco genes, and indeed, no cyanobacteria have yet been found to have multiple rubisco genes of any type. However, the methods that have been used to isolate most rubisco genes could have failed to detect the presence of duplicate genes. Sequence divergence between green-like and red-like rubisco genes is high enough (fig. 3) that heterologous probes would probably detect only one of the two types under typical hybridization conditions, and amplification primers would be expected to be selective for one of the two types. Do any plastids contain both green- and red-like rubisco? The large subunit of rubisco is thought to be encoded by a single-copy plastid gene in all plants and algae (except certain dinoflagellates), but the possibility that a second, divergent rbcL gene resides in the nucleus is intrinsically difficult to rule out. An additional implication of the duplication hypothesis is that the long persistence of these duplicate genes in eubacterial evolution would suggest a functional distinction between green-like and red-like rubiscos. Duplicate genes of identical function are expected to persist only if there is a selective advantage to having two copies of the gene. Although there may be some advantage to having more than one identical copy of a highly expressed gene such as rbcL, the sequence differences between red-like and green-like rubisco are great enough that it is unlikely that both of these types would persist jointly without each having a functional role. Substantial kinetic differences between the various forms of rubisco have been documented (Tabita 1995), and Rhodobacter is known to control the form of rubisco synthesized (form I vs. form II) depending upon the availability of CO, (Gibson 1995). A physiological distinction between green-like and red-like rubiscos, comparable to that seen between the two rubiscos in Rhodobacter sphaeroides and in Chromatium (see below), would be expected if both types are found in a living cyanobacterium. By contrast, horizontal gene transfer postulates that the differences between green-like and red-like rubiscos are essentially neutral differences built up over long independent evolutionary histories. Although many differences between the individual genes certainly reflect adaptation to the form of photosynthesis carried out by the cell, the differences between green-like and red-like

880

Delwiche

and Palmer

rubisco developed independently, need not reflect distinct functional

and the two roles.

types

Other Gene Duplications The best evidence for duplication of rubisco genes involves situations other than the red- green-like split, where a given organism actually contains more than one rubisco gene. One case involves the coexistence of form I and II rubiscos in several diverse proteobacteria (fig. 2; Akazawa, Takabe, and Kobayashi 1984), which is thought to reflect a primordial duplication predating diversification within the extant form I and II types. Because of their simple, homodimeric (L2) form and low substrate specificity, form II rubiscos have been widely regarded as an ancestral form of rubisco (McFadden et al. 1986; Roy and Nierzwicki-Bauer 1991). However, a genuinely ancient form of rubisco would be expected to have a wide phylogenetic distribution, whereas form II rubisco (known only from proteobacteria and certain dinoflagellates) is less widespread than form I rubisco (present in proteobacteria, cyanobacteria, and all other algal groups). For this reason, and because form II rubiscos appear to be changing faster than form I enzymes (fig. 3; Rowan et al. 1996), one should not dismiss the possibility that form II rubisco arose by gene duplication during proteobacterial evolution, followed by loss of the small subunit and acceleration of sequence evolution. A likely example of more recent gene duplication is the presence of the two rubisco-encoding operons (rbcLS and rbcAB) in the y-proteobacterium Chromatium vinosum (fig. 2) and comparable dual operons (rbcLS1 and rbcLS2) in Hydrogenovibrio (which also contains a third, form II gene; Yaguchi et al. 1994). The Hydrogenovibrio rbcL1 and Chromatium rbcA group together, to the exclusion of the Hydrogenovibrio rbcL2 and Chromatium rbcL (fig. 2). Although the shortest trees found by parsimony analysis did not place the latter two genes in a monophyletic group (fig. 2), parsimony bootstrap analysis found 60% support for such a group (including Prochlorococcus), and neighbor-joining analysis found a similar clade with 59% support. Recent gene transfer is unlikely because codon usage is uniform among both Chromatium genes (Kobayashi et al. 1991) and all three Hydrogenovibrio genes (data not shown) but is distinct between the two taxa. Consequently, we believe that a duplication early in y-proteobacterial evolution, in a common ancestor of Chromatium and Hydrogenovibrio, is the best explanation for these data. The putatively long period of coexistence of these duplicates has led to functional differentiation in Chromatium: under normal culture conditions the predominant rubisco is from the rbcAIB operon, and the rbcWS genes are expressed only at very low levels (if at all), but both genes encode fully active enzymes with different rate constants and CO, specificities (Kobayashi et al. 1991). Finally, Alcaligenes eutrophus H16 is known to have two rubisco operons, one located on the bacterial chromosome, and the other on a plasmid. A similar situation is known to occur in A. eutrophus ATCC17707 (Anderson and Caton 1987), but, interestingly, the chromosomal and plasmid rbcL genes from A. eutrophus

H16 are more closely related to each other than either is to the chromosomal gene from A. eutrophus ATCC17707 (fig. 2; unfortunately, the plasmid-encoded rbcL sequence from this strain is not available). This situation could reflect either very recent gene duplication, perhaps complicated by concerted evolution, or lateral transfer between closely related bacteria. Conclusion The two hypotheses discussed here, horizontal gene transfer and gene duplication, need not be mutually exclusive. There is ample reason to believe that both of these processes have played a role in the evolution of rubisco: form I and form II rubiscos are probably the product of an ancient gene duplication, and duplication is probably also responsible for the dual form I genes in Alcaligenes H16, and in Chromatium and Hydrogenovibrio, while the Gonyaulax form II gene and Prochlorococcus form I gene probably result from gene transfer. Thus the best explanation for the overall rubisco phylogeny will certainly involve some combination of these mechanisms. On the basis of current knowledge, division of plastid and proteobacterial rubiscos into green-like and red-like groups could reasonably be attributed to either horizontal gene transfer or gene duplication, although on balance we tend to favor horizontal transfer. The phylogenetic distribution of green-like and red-like rubiscos is so complex that numerous independent events must be postulated in either case, but the number of independent losses necessary under the duplication hypothesis is potentially huge. Furthermore, if the duplication hypothesis is correct, it is rather surprising that no bacteria have yet been found that contain both green- and red-like rubisco, especially given the implied long periods of coexistence of the genes in various bacterial lineages. The chaotic nature of the rubisco phylogeny at the deep level explored in this study is ironic, for rbcL is by far the most widely sequenced gene for systematic purposes within algae and plants (e.g., Clegg 1993; Freshwater et al. 1994; Manhart 1994; Sytsma and Hahn 1994). Fortunately, at these lower taxonomic levels, the rubisco phylogeny is entirely consistent with other evidence, and phylogenetic analyses based on rbcL have been valuable in plant and algal taxonomy. Nor are the rubisco genes the only genes that show signs of phylogenetic discontinuities attributable either to gene transfer (xenology) or duplication and loss (paralogy). Among the proteins that seem to have undergone similar events in bacteria are GAPDH (Martin et al. 1993), HSP60 (Viale et al. 1994), and ATPase (Hilario and Gogarten 1993), and it is reasonable to believe that few, if any, genes will be entirely free of these complications. As a widely sampled and conservative gene, rbcL provides a case study of these phenomena, and it is to be hoped that an understanding of the evolution of rubisco will facilitate interpretation of similar patterns in other genes. Acknowledgments We acknowledge Ron Caspi, Jonathon Eisen, Yasuo Igarashi, David Morse, Brian Palenik, Rob Rowan,

Lateral Transfer and Duplication of Rubisco Genes

88 1

DOUGLAS,S. E. 1994. Chloroplast origins and evolution. Pp. 91118 in D. A. BRYANT, ed. The molecular biology of cyanobacteria. Kluwer Academic Publishers, Amsterdam. DOUGLAS, S. E., and C. A. MURPHY. 1994. Structural, transcriptional, and phylogenetic analyses of the atpB gene cluster from the plastid of Cryptomonas $ (Cryptophyceae). J. Phycol. 30: 329-340. DOUGLAS, S. E., and S. TURNER. 1991. Molecular evidence for the origin of plastids from a cyanobacterium-like ancestor. J. Mol. Evol. 33:267-273. EISEN, J. A. 1995. The recA protein as a model molecule for molecular systematic studies of bacteria: comparison of trees of recAs and 16s rRNAs from the same species. J. Mol. Evol. 41:1105-l 123. FELSENSTEIN,J. 1978. Cases in which parsimony or compatibility LITERATURE CITED methods will be positively misleading. Syst. Zoo]. 27:40AKAZAWA, T, T TAKABE, and H. KOBAYASHI. 1984. Molecular 410. evolution of ribulose- 1,.%bisphosphate carboxylase/oxygenase FRESHWATER,D. W., S. FREDERICQ,B. S. BUTLER,M. HOMMER(RuBisCo). Trends Biochem. Sci. 9:380-383. SAND, and M. W. CHASE. 1994. A gene phylogeny of the red ANDERSON, K., and J. CATON. 1987. Sequence analysis of the algae (Rhodophyta) based on plastid rbcL. Proc. Natl. Acad. Alcaligenes eutrophus chromosomally encoded ribulose bisSci. USA 91:7281-7285. phosphate carboxylase large and small subunit genes and their GENETICSCOMPUTERGROUP. 1991. Sequence analysis software gene products. J. Bacterial. 169:4547-4558. package. Version 7.2. Genetics Computer Group, Madison, Wis. AOYAMA, K., A. M. HAASE, and l? R. REEVES. 1994. Evidence GIBSON, J. L. 1995. Genetic analysis of CO, fixation genes. Pp. for effect of random genetic drift on G+C content after lateral 1107-l 124 in R. E. BLANKENSHIP,M. T. MADIGAN,and C. E. transfer of fucose pathway genes to Escherichia coli K- 12. BAUER, eds. Anoxygenic photosynthetic bacteria. Kluwer AcMol. Biol. Evol. 11:829-838. ademic Press, Amsterdam. ASSALI, N.-E., W. MARTIN, C. SOMMERVILLE,and S. LOISE- GIBSON, J. L., D. L. FALCONE,and E R. TABITA. 1991. Nucleotide AUX-DE GoER. 1991. Evolution of the rubisco operon from sequence, transcriptional analysis, and expression of genes enprokaryotes to algae: structure and analysis of the rbcS gene coded within the form I CO, fixation operon of Rhodobacter of the brown alga Pylaiella littoralis. Plant. Mol. Biol. 17:853sphaeroides. J. Biol. Chem. 266: 1464614653. 863. GIBSON, J. L., and E R. TABITA. 1977. Different molecular forms BHAT~ACHARYA,D., T. HELMCHEN,C. BIBEAU, and M. MELKONof ribulose 1,5-bisphosphate carboxylase from RhodospeudoIAN. 1995. Comparisons of nuclear-encoded small-subunit rimonas sphaeroides. J. Biol. Chem. 252:943-949. bosomal RNAs reveal the evolutionary position of the GlauGIOVANNONI,S. J., S. TURNER, G. J. OLSEN, S. BARNS, D. J. cocystophyta. Mol. Biol. Evol. 12:415-420. LANE, and N. R. PACE. 1988. Evolutionary relationships BHATTACHARYA,D., and L. MEDLIN. 1995. The phylogeny of among cyanobacteria and green chloroplasts. J. Bacterial. 170: plastids: a review based on comparisons of small-subunit ri3584-3592. bosomal RNA coding regions. J. Phycol. 31:489-498. GRAY, M. W. 1992. The endosymbiont hypothesis revisited. Int. BOCZAR, B. A., T. I? DELANEY,and R. A. CATTOLICO.1989. The Rev. Cytol. 141:233-357. gene for the ribulose-1,5-bisphosphate carboxylase small subGROISMAN,E. A., M. H. SAIER JR., and H. O~HMAN. 1992. Horunit protein of the marine chromophyte Olisthodiscus luteus is izontal transfer of a phosphatase gene as evidence for mosaic similar to that of a chemoautotrophic bacterium. Proc. Natl. structure of the Salmonella genome. EMBO J. 11: 1309-I 3 16. Acad. Sci. USA 86:4996-4999. -YAMA, S. 1994. Codon usage patterns suggest independent CAO, Y., J. ADACHI, A. JANKE, S. P&~Bo, and M. HASEGAWA. evolution of two catabolic operons on toluene-degmdative 1994. Phylogenetic relationships among eutherian orders estiplasmid TOL ,,WWO of Pseudomonas putida. J. Mol. Evol. mated from inferred sequences of mitochondrial proteins: in38:328-335. stability of a tree based on a single gene. J. Mol. Evol. 39: HEFWANDEZ,J. M., S. H. BAKER, S. C. LORBACH,J. M. SHIVELY, 5 19-527. and E R. TABITA. 1996. Deduced amino acid sequence, funcCLEGG, M. T. 1993. Chloroplast gene sequences and the study of tional expression, and unique enzymatic properties of the form plant evolution. Proc. Natl. Acad. Sci. USA 90:363-367. I and form II ribulose bisphosphate carboxylase/oxygenase CUMMEVGS,M. F!, S. F! 07~0, and J. WAKELEY. 1995. Sampling from the chemoautotrophic bacterium Thiobucillus den&-$ properties of DNA sequence data in phylogenetic analysis. cans. J. Bacterial. 178:347-358. Mol. Biol. Evol. 12:814-822. HILARIO, E., and J. I! GOGARTEN. 1993. Horizontal transfer of DELONG, E. E, R. B. FRANKEL, and D. A. BAZYLINSKI.1993. ATPase genes-the tree of life becomes a net of life. BioSystems 31:11 l-l 19. Multiple evolutionary origins of magnetotaxis in bacteria. SciKNOLL, A. H., and S. G0LUBIC. 1992. Proterozoic and living cyence 259:803-806. anobacteria. Pp. 450-462 in M. Schidlowski et al., eds. Early DELWICHE,C. E, M. KUHSEL, and J. D. PALMER. 1995. Phyloorganic evolution: implications for mineral and energy regenetic analysis of tufA sequences indicates a cyanobacterial sources. Springer-Verlag, Berlin. origin of all plastids. Mol. Phylogenet. Evol. 4: 110-128. KOBAYASHI, H., A. M. VIALE, T. TAKABE, T. AKAZAWA, K. DE RLIK, F?, Y. VAN DE PEER, I. VAN DEN BROECK, and R. DE WADA, K. SHINOZAKI,K. KOBAYASHI,and M. SUIURA. 1991. WACHTER. 1995. Evolution according to large tibosomal subSequence and expression of genes encoding the large and unit RNA. J. Mol. Evol. 41:366-375. small subunits of ribulose 1,5-bisphosphate carboxylase/oxyDISTEL, D. L., and C. M. CAVANAUGH. 1994. Independent phygenase from Chromatium vinosum. Gene 97:55-62. logenetic origins of methanotrophic and chemoautotrophic Kusmo, T., T. TAKESHIMA,C. INOUE, AND K. SUGAWARA.1991. bacterial endosymbioses in marine bivalves. J. Bacterial. 176: Evidence for two sets of structural genes coding for ribulose 1932-1938.

Jess Shively, Bob Tabita, Gregory Watson, Spencer Whitney, and David Yellowlees for sharing unpublished data, and Eberhard Bock, Jess Shively, and Bob Tabita for useful discussions. We are particularly grateful to an anonymous reviewer for helpful suggestions based on a careful reading of the manuscript. Dave Swofford kindly provided a test version of PAUP* v.4.0 and allowed publication of analyses performed with it. We also thank Heather Diederick and Peter Kuhlman for critically reading the manuscript. This study was supported by NSF grant GM-35087.

882

Delwiche and Palmer

bisphosphate carboxylase in Thinbucillus ferrooxidans. J. Bact. tal structures of L1 and L,S, rubisco suggests a functional role 173:7313-7323. for the small subunit. EMBO J. 9:204-2050. LE~TSCH,C. E. W., and K. V. KOWALLIK. 1992. Nucleotide seSCHOPF,J. W. 1993. Microfossils of the early Archean Apex chert: quence and phylogenetic implication of the ATPase subunits new evidence of the antiquity of life. Science 260640-646. SEEWALDT, E., K. H. SCHLEIFER, E. BOCK, and E. STACKEB and E encoded in the chloroplast genome of the brown alga BRANDT. 1982. The close phylogenetic relationship of NitroDicn/ota dichotoma. Plant Mol. Biol. 19:289-298. hatter and Rhodopseudornonas pczlustris. Acta Microbial. 131: LLOYD, A. T., and I? M. SHARP. 1993. Evolution of the recA gene 287-290. and the molecular phylogeny of bacteria. J. Mol. Evol. 37: SHARP, I? M., and G. MATASSI. 1994. Codon usage and genome 399-407. evolution. Cum Opin. Genet. Dev. 4:851-860. LUDWIG, W., J. NEUMAIER,N. KLUCBAUERet al. (12 co-authors). SHIMADA, A., S. KANAI, and T MARUYA~ZA.1995. Partial se1993. Phylogenetic relationships of bacteria based on comparquence of ribulose- 1$bisphosphate carboxylase/oxygenase ative sequence analysis of elongation factor Tu and ATP-synand the phylogeny of Prochloron and Prochlorococcus thase B-subunits. Antonie Van Leeuwenhoek 64:285-305. (Prochlorales). J. Mol. Evol. 40:67 i-677. MANHART,J. R. 1994. Phylogenetic analysis of green plant rbcL SOMERVILLE,C. C., S. JOUANNIC,W. E MARTIN, B. KLOAREG, sequences. Mol. Phylog. Evol. 3: 114-127. and S. LOISEAUX-DE Go~R. 1993. Secondary structure and MARTIN, W., S. JOUANNIC,and S. LOISEAUX-DEGo~R. 1993. Mophylogeny of the chloroplast 23s rRNA gene from the brown lecular phylogeny of the atpB and atpE genes of the brown alga P$zirl/a littorulis. Plant. Mol. Biol. 21:779-787. alga Pyluiella littorulis. Eur. J. Phycol. 28: 11 l-l 13. STRECKER,M., E. SICKINGER,R. S. ENGLISH,J. M. SHIVELY,and MARTIN, W., C. C. SOMERVILLE,and S. LOISEAUX-DEGo~R. E. BOCK. 1994. Calvin cycle genes in Nitrobucter vu/gari.s T3. 1992. Molecular phylogenies of plastid origins and algal evoFEMS Microbial. Lett. 120:45-50. lution. J. Mol. Evol. 35385-404. MARTIN, W. S., H. BRINKMANN, C. SAVONNA, and R. CERFF. SWOFFORD,D. L. 1993. PAUP: phylogenetic analysis using par simony. Version 3.1. Computer program distributed by the Il1993. Evidence for a chimeric nature of nuclear genomes: eulinois Natural History Survey, Champaign, 111. bacterial origin of eukaryotic glyceraldehyde-3-phosphate deSYTSMA, K. J., and W. J. HAHN. 1994. Molecular systematics: hydrogenase genes. Proc. Natl. Acad. Sci. USA 908692-8696. 1991-1993. Prog. Bot. 55:307-333. MCFADDEN,B. A., J. TORRES-RUIZ,H. DANIELL,and G. SAROJINI. TABITA, E R. 1995. The biochemistry and metabolic regulation 1986. Interaction, functional relations and evolution of large of carbon metabolism and CO, fixation in purple bacteria. Pp. and small subunits in rubisco from prokaryotd and eukaryota. 855-914 in R. E. BLANKENSHIP,M. T MADIGAN, and C. E. Philos. Trans. R. Sot. Lond. B 313:347-358. BAUER, eds. Anoxygenic photosynthetic bacteria. Kluwer, AmMORDEN,C. W., C. E DELWICHE,M. KUHSEL, and J. D. PAI.MER. sterdam. 1992. Gene phylogenies and the endosymbiotic origin of plasURBACH, E., D. L. ROBERTSON,and S. W. CHISHOLM.1992. Multids. BioSystems 28:75590. tiple evolutionary origins of prochlorophytes within the cyMORSE, D., I? SALOIS,I?MARKOVIC, and J. W. HASTINGS. 1995. anobacterial radiation. Nature 355267-269. Dinoflagellate rubisco is nuclear encoded and similar to the VALENTIN,K., and K. ZETSCHE. 1989. The genes of both subunits form II enzyme of c-u-proteobacteria. Science 268: 1622-1624. of ribulose- 1,5-bisphosphate carboxylase consitute an operon NELISSEN,B., Y. VAN DE PEER, A. WILMOTTE,and R. DE WACHon the plastome of a red alga. Cum Genet. 16:203-209. TER.1995. An early origin of plastids within the cyanobacterial VAN DE PEER, Y., J.-M. NEEFS, I? DE RIJK, P DE Vos, and R. divergence is suggested by evolutionary trees based on comDE WACHTER. 1994. About the order of divergence of the plete 16s rRNA sequences. Mol. Biol. Evol. 12:1166-l 173. major bacterial taxa during evolution. Syst. Appl. Microbial. OCHMAN, H., and J. G. LAWRENCE. 1995. Phylogenetics and the 17:32-38. amelioration of bacterial genomes. In E C. NEIDHARDTet al., VIALE, A. M., and A. K. ARAKAKI. 1994. The chaperone coneds. E. co/i and S. ~~~himuriurrz:molecular and cellular aspects. nection to the origins of the eukaryotic organelles. FEBS Lett. 2nd edition. ASM Press, Washington (in press). 341:146-151. OLSEN, G. J., C. R. WOESE, and R. OVERBEEK. 1994. The winds VIALE, A. M., A. K. ARAKAKI, E C. SONCINI.and R. G. FERREYRA. 1994. Evolutionary relationships among eubacterial groups of (evolutionary) change: breathing new life into microbiology. as inferred from GroEL (chaperonin) sequence comparisons. J. Bacterial. 176:265-267. Int. J. Syst. Bacterial. 44:527-533. PA~ENIK, B., and R. HASELKORN. 1992. Multiple evolutionary WEIGEL, J. 1992. The genus Xnnthobacter. Pp. 2365-2383 in A. origins of prochlorophytes, the chlorophyll b-containing proBALOWS, H. G. TR~PER, M. DWORKIN,W. HARDER, AND Kkaryotes. Nature 35.5265-267. H. SCHLEIFER,eds. The prokaryotes. second edition: a handPALENIK, B., and H. SWIM. 1996. Cyanobacterial evolution and book on the biology of bacteria: ecophysiology, isolation, idenprochlorophyte diversity as seen in DNA-dependent RNA tification, applications. Springer-Verlag, Berlin. polymerase gene sequences. J. Phycol. (in press). WHITNEY, S. M., D. C. SHAW, and D. YEI.LOWLEES.1995. EviREITH, M. 1995. Molecular biology of rhodophyte and chromodence that some dinollagellates con:,un a ribulose- 1,5-bisphosphyte plastids. Annu. Rev. Plant Physiol. Plant Mol. Biol. 46: phate carboxylase/oxygenase related to that of the o-proteo549-575. bacteria. Proc. R. Sot. Lond. B 259:271-275. ROWAN, R., S. M. WHITNEY, A. FOWLER, and D. YELLOWLEES. WOESE, C. 1987. Bacterial evolution. Microbial. Rev. 51:2211996. Rubisco in marine symbiotic dinoflagellates: form II en271. zymes in eukaryotic oxygenic phototrophs, encoded by a nuYAGUCHI,T, S. Y. CHUNG, Y. ICARASHI,and T KODAMA. 1994. clear multi-gene fdmily. Plant Cell (in press). Cloning and sequence of the L, form of rubisco from a marine ROY, H., and NIERZWICKI-BAUER.1991. RuBisCo: genes, strucobligately autotrophic hydrogen-oxidizing bacterium, H$roture, assembly, and evolution. Pp. 347-364 in L. BOGORAD genovibrio mczrinus strain MH- I 10. Biosci. Biotechnol. Bioand I. K. VASIL, eds. Cell Culture and Somatic Cell Genetics them. 58:1733-1737. of Plants, vol. 7B, the photosynthetic apparatus: molecular biology and operation. Academic Press, San Diego. THOMAS EICKBUSH, reviewing editor SCHNEIDER,G.. S. KNIGHT, I. ANDERSSON,C.-I. B~~ND~N, Y. LINDQVIST,and T LUNDQVIST.1990. Comparison of the crysAccepted April 1, 1996