Phycobiliprotein genes of the marine photosynthetic prokaryote Prochlorococcus: evidence for rapid evolution of genetic heterogeneity

The oceans are responsible for nearly half of the earths photosynthetically fixed carbon. In tropical and subtropical marine environments, the unicellular prokaryotes Prochlorococcus and Synechococcus are dominant members of the photosynthetic biomass (Partensky et al., 1999 ). Phylogenetic studies based on analyses of 16S rRNA and other (psbB, psbA, petB/D operon, rpoC1) genes indicate that Prochlorococcus is most closely related to phycobilisome-containing, marine Synechococcus species (marine cluster-A) (Palenik & Haselkorn, 1992 ; Urbach et al., 1992 , 1998 ; Hess et al., 1995 ). Branching orders between Prochlorococcus and marine Synechococcus are not well resolved and suggest a rapid and almost simultaneous diversification of both groups from a common ancestor (Urbach et al., 1998 ).

Although Prochlorococcus and Synechococcus share close phylogenetic ties and several ecological and physiological characteristics, a fundamental difference exists between these genera in their photosynthetic light-harvesting apparatus. A major antenna of the latter consists of phycobiliproteins that are covalently bound to linear tetrapyrrole prosthetic groups and are organized into phycobilisomes (Bryant, 1991 ). The phycobilisomes of marine cluster-A Synechococcus are phycoerythrin (PE)-rich, and two types of PE (I, II) have been identified that differ in their number of associated bilins (Ong & Glazer, 1991 ; Wilbanks et al., 1991 ). In contrast, the major antenna complex of Prochlorococcus is based on the pigments divinyl chlorophyll a (Chl a₂) and divinyl and/or monovinyl chlorophyll b (Chl b₂, Chl b) (Chisholm et al., 1992 ; Goericke & Repeta, 1992 ; LaRoche et al., 1996 ; Partensky et al., 1997 ). Its light-harvesting complex polypeptides (Pcb) are members of the Chl a-binding core complex family of antenna proteins, which includes CP43 and CP47 (LaRoche et al., 1996 ; Partensky et al., 1997 ).

Recently, genes encoding the α (cpeA) and ß (cpeB) subunits of PE were identified in Prochlorococcus sp. strain SS120 (CCMP1375) (Hess et al., 1996 ). In SS120, both cpeB and cpeA are found in a gene cluster containing genes that are homologous to those encoding other phycobiliproteins (MpeC, CpeZ) and bilin biosynthesis proteins (Hess et al., 1999 ). Although low levels of cpeB and cpeA expression have been detected in SS120 (Hess et al., 1996 , 1999 ), the functional significance of PE in SS120 is unknown. PE genes have since been found in Prochlorococcus isolates PAC1 and PAC2 (Penno et al., 2000 ), and we have also identified cpeB in the genome of Prochlorococcus MED4. This latter isolate is capable of growing at higher (>300 µmol photons m^-2 s^-1) irradiance levels than low-light-adapted Prochlorococcus isolates such as SS120, and is a member of the high-light-adapted clade (Urbach et al., 1998 ; Moore et al., 1998 ; Rocap, 2000 ).

We have in our collection two low-light-adapted Prochlorococcus isolates, MIT9303 and MIT9313, which are the most deeply branching in the Prochlorococcus lineage (Moore et al., 1998 ; Rocap, 2000 ). In this unique phylogenetic position, their 16S rDNA sequences differ by approximately 23% from other isolates of both Prochlorococcus sp. and marine Synechococcus sp. In order to obtain a more complete understanding of the evolution of phycobiliproteins in Prochlorococcus, we sequenced the cpeB and cpeA genes of both MIT9303 and MIT9313. Our analyses suggest that within the Prochlorococcus lineage the selective forces shaping the evolution of the PE gene set have not been uniform. Furthermore, the PE gene sequence heterogeneity we document between Prochlorococcus and Synechococcus is consistent with a model of elevated mutation rates, rather than relaxed selection. Elucidating the rates and pathways of genetic change in this important phototroph may help in assessing its response to rapid environmental change.

Culture conditions.
Prochlorococcus isolates MIT9303 and MIT9313, and marine Synechococcus sp. strain WH8103, were grown in batch cultures in Sargasso Sea water supplemented with Pro2 medium nutrients and trace metals (Moore & Chisholm, 1999 ). Cultures were maintained at 21±1 °C. Illumination was supplied during a 14 h light:10 h dark cycle by cool white fluorescent lights at 10 µmol photons m^-2 s^-1.

Genomic DNA isolation.
Genomic DNA was prepared from Prochlorococcus according to standard methods (Ausubel et al., 1995 ) that were optimized for this bacterium (Rocap, 2000 ). Briefly, cells were lysed in the presence of lysozyme, and DNA was extracted with phenol/chloroform/isoamyl alcohol (25:24:1, by vol.), precipitated with 2-propanol, and resolubilized in TE buffer (pH 7·6).

PCR amplification, cloning and sequencing.
Genes encoding the ß- (cpeB) and α- (cpeA) subunits of PE were amplified from Prochlorococcus genomic DNA by PCR. The primers, cpeB1.2 (ATGCTTGATGCATTCTCAAG) and cpeA 2.2 (AAGGCATTAATAAGGTAATC), were designed as exact matches to the SS120 cpeB 5'-end and cpeA 3'-end (Hess et al., 1996 ). PCR samples contained 13 µg genomic DNA, 0·5 µM of each primer, 250 µM each dNTP, 03·5 mM MgCl₂, cloned Pfu buffer (1x) (Stratagene), and 2·5 units Pfu DNA polymerase (Stratagene). Cycle parameters were 94 °C for 1 min, 50 °C for 1 min and 72 °C for 2 min. Samples were processed for 30 or 40 cycles prior to a final extension at 72 °C for 10 min.

PCR products were separated on 1·5% agarose gels and visualized by UV fluorescence after staining with ethidium bromide. Selected PCR products of ∼1000 bp were purified from agarose gels using the QIAEXII Kit (QIAGEN), and then ligated to the pCR-Script Amp SK(+) Cloning Vector (PCR-Script Amp Cloning Kit, Stratagene). Vectors containing this insert were used to transform Escherichia coli (Epicurian Coli XL 10-Gold Kan ultracompetent cells, Stratagene). Size-selected clones were sequenced and final cpeB and cpeA sequences were obtained from three clones for MIT9303 and two clones for MIT9313. Plasmids were sequenced using M13 forward and reverse primers in our laboratory using SequiTherm LongRead Cycle Sequencing Kits on LI-COR model 4000L sequencers or by the MITCCR HHMI Biopolymers Laboratory.

Phylogenetic analysis.
Sequences used in comparisons were obtained from GenBank and from the complete MED4 genome (http://spider.jgi-psf.org/JGI_microbial/html/). All analyses utilized cpeB sequences minus the first 20 nucleotides (5' primer region) and cpeA minus the last 24 nucleotides (3' primer region). Data and alignments for 16S rDNA analyses were downloaded directly from the Ribosomal Database Project (RDP) (Maidak et al., 2000 ).

PE protein sequences were aligned using the CLUSTAL W program with the BLOSUM matrix (Thompson et al., 1994 ). Gaps were included to optimize the alignments. Only unambiguously aligned positions were used in the phylogenetic analyses (146 amino acids for α-PE, 159 amino acids for ß-PE). The percentage identities summarized in Table 1 were calculated from uncorrected distances. Total G+C base content ratios were calculated using GeneMark (Borodovsky & McIninch, 1993 ). MEGA version 2.0 (Kumar et al., 2001 ) was used to calculate the number of synonymous and nonsynonymous substitutions using both the original and modified NeiGojobori method (R=1·28 for cpeB, R=1·35 for cpeA) (Nei & Gojobori, 1986 ; Ina, 1995 ; Nei & Kumar, 2000 ; Kumar et al., 2001 ). The numbers of synonymous and nonsynonymous differences were compared using p-distances, where the number of synonymous differences was normalized to the number of synonymous sites, and the number of nonsynonymous differences was normalized to the number of nonsynonymous sites. Conclusions drawn from the original and modified NeiGojobori analyses were the same, and only the results of the latter analyses have been reported. Nonparametric relative rate tests were conducted in MEGA version 2.0 (Kumar et al., 2001 ) using Tajimas general method (Tajima, 1993 ), with one degree of freedom and a significance level of 5%.

Table 1. Identity matrix for cpeB (bold type) and ß-PE (above the diagonal) and cpeA (bold type) and α-PE (below the diagonal) of Prochlorococcus, and other cyanophytes and rhodophytes

Protein phylogenetic analyses were done with PUZZLE 4.0 (Strimmer & von Haeseler, 1996 ) for maximum-likelihood, PHYLIP 3.5c (Felsenstein, 1989 ) for distance, and PHYLIP 3.5c (Felsenstein, 1989 ) for parsimony algorithms. Distances were calculated in PROTDIST using the DayhoffPAM substitution matrix and trees were inferred using the FitchMargoliash weighted least-squares criterion in FITCH. Distance and parsimony bootstrap analyses were performed with 1000 pseudosampled datasets generated by SEQBOOT. In the protein maximum-likelihood analyses, the JTT substitution model was used and the puzzling step was repeated 1000 times.

PAUP* version 4.0 beta 2a (Swofford, 1999 ) was used for all nucleotide analyses. Percentage identities reported in Table 6 were calculated from uncorrected distances. Distance trees were inferred using minimum evolution as the objective criterion and paralinear (logdet) distances, which are relatively insensitive to differences in G+C content (Lake, 1994 ). Nucleotide maximum-likelihood analyses were done via the HKY85 model of nucleotide substitution with rate heterogeneity and empirical nucleotide frequencies. Bootstrap analyses (100 resamplings for maximum-likelihood, 1000 for distance and maximum-parsimony) were performed with heuristic searches utilizing random addition and tree-bisection reconnection branch-swapping methods. Phylogenetic trees were displayed using TREEVIEW (Page, 1996 ).

Table 6. 16S rDNA sequence identity matrix for Prochlorococcus (Pro) and Synechococcus (Syn)

Accession numbers.
Sequences reported in this paper were deposited in GenBank under the following accession numbers: AF304132 (MIT9303 cpeB), AF304135 (MIT9303 cpeA), AF304133 (MIT9313 cpeB), AF304134 (MIT9313 cpeA). Database accession numbers of the PE sequences reported for other organisms, including Prochlorococcus (Pro) and Synechococcus (Syn), are as follows: Pro SS120 Z68890; Pro PAC1 AJ272069; Pro PAC2 AJ237612, AJ272030; Syn WH7803 X72961; Syn WH8103 M91809; Syn WH8020 M95288; Synechocystis PCC6701 M33812; Calothrix PCC7601 X04592; Pseudanabaena PCC7409 X63073, S96581; Porphyra purpurea U38804; Aglaothamnion neglectum Z11907; Cryptomonas sp. AFMDB; Chroomonas sp. S21775. Accession numbers for 16S rDNA sequences are as follows: Pro MIT9303 AF053397; Pro MIT9313 AF053399; Pro NATL2A AF001467; Pro PAC1 AF001471; Pro MED4 AF001466; Pro SS120 X63140; Syn WH8101 AF001480; Syn WH8103 AF001479; Syn WH7805 AF001478; Syn WH7803 AF081834; Syn PCC6307 AF001477; Syn PCC7001 AB015058; Syn PCC6301 X03538; Syn PCC7942 D88288; Synechocystis PCC6803 D64000; Microcystis holsatica U40336; Microcystis elabens U40335; Arthrospira PCC8005 X70769; Calothrix D253 X99213; Pseudanabaena biceps PCC7367 AF091108; Plectonema F3 AF091110; Nostoc punctiforme PCC73102 AF027655; Pleurocapsa PCC7516 X78681; Cyanidium caldarium chloroplast X52985; Antithamnion chloroplast X54299; Chondrus crispus chloroplast Z29521; Palmaria palmata Z18289; Cryptomonas phi chloroplast S73904; Cyanophora paradoxa cyanelle U30821. Amplification, cloning and sequencing of MIT9303 and MIT9313 PE genes
Using primers that were exact matches to the SS120 cpeB (5'-end) and cpeA (3'-end) genes, and Synechococcus sp. strain WH8103 as the control, we optimized the PCR conditions in the presence of sequence mismatches between the primers and the target cpeB and cpeA genes. Although the PE genes of Synechococcus sp. strain WH8103 possessed three mismatches to the cpeB primer (5'- end) and six mismatches to the cpeA primer (3'-end), we amplified a single DNA fragment of the approximate size expected for cpeB and cpeA (∼1000 bp) (Fig. 1, lane 2). Using the same PCR conditions, we then sought to amplify the genes encoding the α- and ß-subunits of PE from the genomic DNA of MIT9303 and MIT9313. Separation of the MIT9303 PCR products on an agarose gel revealed a single major DNA fragment of ∼1000 bp (Fig. 1, lane 3). Although this was also the case for MIT9313, additional minor DNA fragments were observed for this isolate (Fig. 1, lane 4). Subsequent cloning and sequencing of the ∼1000 bp DNA fragments from the two isolates identified them as putative cpeA and cpeB genes. Using these gene and derived amino acid sequences, as well as those from SS120 (Hess et al., 1996 ), we performed a BLAST search of the MED4 genome and were able to identify the MED4 cpeB gene. However, we found that MED4 lacks a cpeA homologue.

(93K): Fig. 1. PCR amplification of Prochlorococcus and marine Synechococcus DNA using a set of exact match primers to the 5'-end of SS120 cpeB and to the 3'-end of SS120 cpeA. The first lane shows a DNA ladder. Lane 1, SS120 (control). Note the single PCR product of ∼1000 bp. Lane 2, Synechococcus sp. strain WH8103. Lane 3, MIT9303. Lane 4, MIT9313.

Nucleotide sequence comparisons
The cpeB gene sequences of MIT9303 and MIT9313 exhibited 97% identity and were more similar to each other than they were to other Prochlorococcus or marine Synechococcus cpeB genes (Table 1). Most notably, of the 14 base-pair changes between MIT9303 and MIT9313, all but one of these changes occurred in the third codon position. Comparable patterns were observed for the MIT9303 and MIT9313 cpeA genes (Table 1). Pairwise cpeB sequence comparisons among the low-light-adapted isolates (MIT9303, MIT9313, SS120, PAC1; Table 1) indicated that sequence identities ranged from 65% (MIT9303/MIT9313, PAC1) to 97% (MIT9303, MIT9313). However, cpeB of MED4 (high-light-adapted clade) exhibited 42% or less nucleotide identity with cpeB of other Prochlorococcus isolates (Table 1).

The number of synonymous (K_s) and nonsynonymous (K_a) differences for cpeB and cpeA of Prochlorococcus as well as other cyanobacteria, and the K_a/K_s ratios, are reported in Tables 2 and 3, respectively. For both cpeB and cpeA, K_a/K_s values were not elevated within the Prochlorococcus genus relative to marine Synechococcus (Table 3). Comparisons between the different genera, as well as between specific strains, indicated that the range of K_a/K_s ratios was comparable (Table 3). Interestingly, while K_a/K_s was consistently greater than one for cpeB, K_a/K_s was less than one for cpeA, suggesting that the functional constraints on the evolution of these genes are different.

Table 2. Matrix of the number of synonymous (Ks) and nonsynonymous (Ka) differences for the cpeB (above the diagonal) and cpeA (below the diagonal) genes of different cyanobacterial genera

Table 3. Ratios of the number of nonsynonymous to synonymous differences (Ka/Ks) in the cpeB (above the diagonal) and cpeA (below the diagonal) sequences of different cyanobacteria

G+C base ratios
Both the total and third codon G+C base content of cpeB and cpeA were higher in MIT9303 and MIT9313 than in other Prochlorococcus isolates (Table 4). Among the six isolates that have been characterized thus far, the total G+C contents of cpeB and cpeA in MIT9303 and MIT9313 were closest to those of marine Synechococcus sp. For these two genes, the codon usage of MIT9303 and MIT9313 was shifted towards T at the third base position (T>C>G or A). This nucleotide frequency bias was similar to that of other Prochlorococcus isolates, whose cpeB and cpeA genes have third base positions favouring T or A (T>A>G C). Similar nucleotide frequency biases towards T (T >A>C or G) have been reported previously for psbB and petB/D genes of several Prochlorococcus isolates, including SS120 and NATL2A (Urbach et al., 1998 ; Partensky et al., 1999 ), and mutational biases may be a cause of these observed patterns.

Table 4. G+C base ratios of cpeB and cpeA of Prochlorococcus (Pro) and marine Synechococcus (Syn)

Derived ß-PE and α-PE amino acid sequences
As indicated in Table 1, Prochlorococcus ß-PE sequences (excluding MED4) were consistently more similar to each other (7799% identity) than they were to the ß-PEs of marine Synechococcus strains (6067% identity), or other cyanophytes and rhodophytes. Pairwise comparisons between MIT9303/MIT9313 and SS120 revealed that this was most likely because the majority of differences were in the third codon position and many were synonymous. The derived ß-PE sequence of MED4 exhibited the lowest amount (41%) of sequence identity with the ß-PE sequences of the other isolates (Table 1). Similar patterns were observed for the α-PE sequences of Prochlorococcus (Table 1).

Both the derived ß-PE sequences and α-PE sequences exhibited 99% amino acid identity between MIT9303 and MIT9313 (Table 1). While the single amino acid residue difference between ß-PE of MIT9303 and MIT9313 involved the conserved substitution of Asp-144 (MIT9303) and Asn-144 (MIT9313), the single residue difference between α-PE involved the semi-conserved substitution of Ala-29 (MIT9303) and Val-29 (MIT9313).

Within the Prochlorococcus lineage, relative rate tests for ß- and α-PE revealed that mutations in these sequences appear to be accumulating at approximately the same rate for low-light-adapted strains such as MIT9303 and SS120 (χ² was not significant at the 5% level) (Table 5). However, the rate at which the ß-PE sequence is evolving in MED4 is not the same as that in MIT9303 or SS120. Instead, relative rate tests suggested an elevated rate of evolution for the high-light-adapted MED4 sequence (Table 5).

Table 5. Nonparametric relative rate test scores for the ß-PE and α-PE sequences of Prochlorococcus and other cyanobacteria

The results of the relative rate tests between Prochlorococcus and Synechococcus PE sequences depended on whether Synechococcus PE (I) or (II) was used in the actual test (Table 5). Differences in the rate at which mutations have accumulated in the PE sequences of Prochlorococcus strains MIT9303/SS120 and the PE (I) sequence of Synechococcus WH8020 were significant (Table 5). However, relative rate tests suggested that mutations have accumulated at approximately the same rate in the Prochlorococcus strain MIT9303/SS120 PE sequences as in the Synechococcus PE (II) sequences. In the case of MED4, all relative rate tests with Synechococcus suggested that the rate at which ß-PE is evolving is not equivalent, and that mutations have accumulated faster in MED4 (Table 5).

Identification of putative functional residues
Alignments of the derived amino acid sequences of the ß-PE subunit of MIT9303 and MIT9313 indicated that many residues known for their functional importance in the ß-PE subunits of other organisms (Apt et al., 1995 ) were conserved in all of the Prochlorococcus sequences, excluding MED4 (Fig. 2a, b). These included residues that serve as putative chromophore attachment sites (Cys-50, Cys-61, Cys-82, Cys-162), function in chromophore interaction and stabilization (Arg-77, Arg-78, Ala-80/81, Arg-84, Asp-85), and have a role in subunit interactions (Asp-13, Arg-91, Tyr-92, Tyr-95, Arg-108) (Apt et al., 1995 ). In contrast, the MED4 ß-PE sequence lacked several of these highly conserved residues, including Cys-50(43), Cys-162(155), Ala-80(73)/81(74) and Tyr-92(85) (Fig. 2a).

(121K): Fig. 2. Alignment of deduced amino acid sequences of ß-PE (a) and α-PE (b) of Prochlorococcus isolates MIT9303, MIT9313, SS120 and MED4 (ß-PE only), marine Synechococcus sp. strains WH7803, WH8103 and WH8020, and Calothrix sp. strain PCC7601 (formerly Fremyella diplosiphon). Identical residues are presented in white type on a black background. Black type on grey squares indicates highly conserved substitutions (within the groups STA NEQK NHQK NDEQ QHRK MILV MILF HY FYW). Putative chromophore-binding sites are indicated with an asterisk above the corresponding Cys residue.

Alignment of the derived α-PE sequences indicated that several residues important in chromophore attachment or protein interactions were also conserved in this subunit (Fig. 2b). Most notably, both MIT9303 and MIT9313 possessed a cysteine residue at position 130 that is highly conserved among other cyanobacteria and red algae and functions in chromophore binding (Apt et al., 1995 ). However, this Cys-130 residue is not conserved in the α-PE sequences of any other Prochlorococcus isolate (Fig. 2b). In SS120, this position contains an aspartic acid residue (Hess et al., 1996 ), which does not represent a conserved substitution.

Phylogenetic analyses
Phylogenetic trees constructed using protein sequences deduced from cpeB and cpeA genes (Fig. 3a, b) revealed that the position of the Prochlorococcus cluster relative to marine Synechococcus was not congruent with 16S rDNA based phylogenetic groupings (Fig. 4). Prochlorococcus ß-PE and α-PE sequences did not group with either type I or type II ß-PE and α-PE sequences from marine Synechococcus (Fig. 3a, b). However, within the Prochlorococcus limb, the clustering of isolates based on both ß- and α-PE sequences was consistent with their positions in the 16S rDNA-based tree. As in ribosomal trees, MIT9303 and MIT9313 are very closely related (Table 6, Fig. 4) and form the basal branch of the Prochlorococcus clade. Similar tree topologies and levels of support were obtained in trees rooted using phycocyanin sequences and in trees constructed using cpeB and cpeA nucleotide sequences (both all positions and only the first two codon positions) (data not shown). The MED4 ß-PE sequence was not included in the tree (Fig. 3a) because it is highly degenerate and shared low sequence identity (36%) with other cyanobacterial and red algal ß-PEs (Table 1). This resulted in its branch length being five times longer than any of the others in the tree. In phylogenetic trees, the placement of such long branches is often at odds with true evolutionary position (Swofford et al., 1996 ).

(26K): Fig. 3. Phylogenetic analysis of PE proteins from Prochlorococcus, other members of the cyanobacterial lineage, red algae and cryptomonads, using maximum-likelihood, distance and parsimony methods. Likelihood trees are shown for both ß-PE (a) and α-PE (b); identical tree topologies were obtained in distance and parsimony analyses. Support values for internal branches (percentage support from quartet puzzling for likelihood or percentage bootstrap values from 1000 resamplings for distance and parsimony analyses) are displayed at each node in the following order: likelihood/distance/parsimony.

(39K): Fig. 4. Phylogenetic relationships among Prochlorococcus and other organisms based on 16S rDNA sequences. The distance tree is shown with bootstrap values for all three methods displayed in the following order: likelihood/minimum evolution-distance/parsimony.

16S rDNA-based phylogenetic analyses suggest that Prochlorococcus and marine Synechococcus share a common, phycobiliprotein-containing ancestor, and may have both diversified at a similar point in time (Urbach et al., 1998 ). Although Prochlorococcus has evolved to use a chlorophyll a₂/b₍₂₎ light-harvesting complex as its major antenna, it still retains genes encoding the phycobiliprotein PE (Hess et al., 1996 ; LaRoche et al., 1996 ; Ting et al., 1999 ). The branching order we observed among Prochlorococcus PE genes (not including the degenerate MED4 cpeB) was congruent with previous 16S rDNA sequence analyses for this genus. Furthermore, both complete sequence and third codon G+C base frequencies of cpeB and cpeA were correlated with 16S rDNA phylogenetic tree branching patterns, with isolates such as MIT9303 and MIT9313 exhibiting the highest (0·480·53) G+C base content ratios for these particular genes. A similar correlation between third codon position G+C base frequencies and the branching patterns of 16S rRNA gene phylogenetic trees was demonstrated previously for several Prochlorococcus and Synechococcus genes (Urbach et al., 1998 ; Rocap, 2000 ). In particular, other G+C-rich genes identified in MIT9303 include psbB (0·52), petB (0·58) and petD (0·58) (Urbach et al., 1998 ). Although these data support the idea that the PE genes of Prochlorococcus were unlikely to have been acquired by lateral gene transfer (Hess et al., 1999 ), they do not completely rule out the possibility of recent horizontal transfer from a close relative, particularly in the case of MED4. Nevertheless, no MED4-like cpeB sequence has been previously described in any other photosynthetic organism. Thus it is most parsimonious to presume that the PE genes were retained from the phycobilisome-containing ancestor of Prochlorococcus and Synechococcus.

Our sequence analyses suggest that within the Prochlorococcus lineage, the selective forces shaping the evolution of the PE gene set have not been uniform. Notably, this is apparent in the unique loss of the gene encoding the α-PE subunit in the MED4 strain compared to more deeply branching isolates such as MIT9303 and MIT9313. Loss of PE genes has been suggested to have occurred in other cyanobacterial lines (Apt et al., 1995 ). MIT9303 and MIT9313 have the highest degree of 16S rDNA sequence similarity to marine Synechococcus relative to other cultured Prochlorococcus isolates (Table 6), and only they have retained a highly conserved putative chromophore binding site (Cys-130) in their α-PE subunit. Furthermore, relative rate tests suggest that mutations have not accumulated at an equivalent rate in the ß-PE sequences of Prochlorococcus. In MED4, the presence of a degenerate form of ß-PE, characterized by several point mutations and deletions at highly conserved sites, is consistent with an elevated rate of evolution. In addition, the ratios of nonsynonymous (K_a) to synonymous (K_s) substitutions were not elevated in MED4 relative to other Prochlorococcus or Synechococcus strains. Thus, even though MED4 lacks the gene encoding the α-PE subunit, these results suggest that its cpeB gene is still under selection, although likely a different type of selection than in other Prochlorococcus strains. As a high-light-adapted strain of Prochlorococcus, MED4 is thought to have arisen more recently than its low-light-adapted counterparts (e.g. MIT9303, MIT9313, SS120, PAC1) (Urbach et al., 1998 ; Rocap, 2000 ), and thus could have been subjected to different selective pressures (i.e. light and/or nutrient availability), leading to dissimilar genetic recombination and/or deletion events.

Although cpeB and/or cpeA expression has been detected in SS120 (Hess et al., 1996 , 1999 ), and in MIT9303 and MIT9313 (C. Ting, unpublished results), the exact role of PE in the Prochlorococcus genus remains to be established. PE is clearly not the major constituent of the light-harvesting apparatus in any extant Prochlorococcus strain (Chisholm et al., 1992 ; Goericke & Repeta, 1992 ; LaRoche et al., 1996 ; Partensky et al., 1997 ), as it is in Synechococcus (Ong & Glazer, 1991 ; Wilbanks et al., 1991 ). In SS120, the PE genes are part of a larger cluster, within which other phycobiliprotein-related genes (cpeZ, cpeY, mpeX, ppeC) can be found (Hess et al., 1999 ). But there is no evidence that PE forms part of a cyanobacterial-like phycobilisome structure in Prochlorococcus (Chisholm et al., 1988 ; Fields et al., 1997 ; C. S. Ting and others, unpublished). It remains to be explored whether the phycobiliproteins of Prochlorococcus are located within the intracytoplasmic lamellar space, as in cryptophytes (Gantt et al., 1971 ).

Our phylogenetic analyses based on PE sequences show that the position of the Prochlorococcus cluster relative to Synechococcus was not congruent with 16S rDNA-based trees. Additional analyses indicated that the ratios of nonsynonymous (K_a) to synonymous (K_s) substitutions for the cpeB and cpeA genes were not elevated in Prochlorococcus relative to Synechococcus. Instead, K_a/K_s ratios were similar between these genera. This suggests that the sequence differences we observed may be due to elevated mutation rates rather than relaxed selection. This is consistent with the results of relative rate tests between Prochlorococcus ß/α-PE and Synechococcus ß/α-PE (I) sequences. Thus although PE has a different function in Prochlorococcus, these results suggest that the cpeB and cpeA genes are still under selection, albeit a different type of selection than in Synechococcus.

Future work on the role of PE in Prochlorococcus may help to identify whether this protein confers a biological advantage under specific environmental conditions. However, with the change in selection on the PE genes we have observed within the Prochlorococcus lineage, it is unlikely that the function of PE will be conserved among all strains. Comparisons between Prochlorococcus and Synechococcus provide a striking example of how the evolution of a key protein complex, the light-harvesting antenna system, has proceeded along very different paths in two globally important marine prokaryotes.

We would like to thank John Waterbury (Dept of Biology, Woods Hole Oceanographic Institution) for kindly providing us with Synechococcus sp. strain WH8103, Mitchell Sogin for the use of the LICOR model 4000L Sequencer at the Marine Biological Laboratory (Woods Hole, MA) and the DOE Joint Genome Institute for use of preliminary sequence data (http://www.jgi.doe.gov/JGI_microbial/html/). Special thanks are also extended to Thomas Owens and Wolfgang Hess for their critical reading of the manuscript. This study was supported by a NSF Postdoctoral Research Fellowship in Biosciences Related to the Environment (C.S.T.), by a NSF Graduate Student Fellowship (G.R.), by NSF OCE-9820035 (S.W.C.), and by NASA (NAG5-3727) (G.R., S.W.C.).