Detection of plasmids. B. aphidicola(T. caerulescens) has previously been shown to carry a small, cryptic plasmid (pBTc1, 1.74 kb) that contained therepA₁ replicon (related to IncFII plasmids) found in leucine plasmids (49). Restriction enzyme analysis of plasmid DNA preparations from this species indicated the presence of a second plasmid of about 3 kb. Similar observations were made with plasmid DNA isolated from B. aphidicola(P. spyrothecae) (48). Plasmid DNA preparations from both species were here further analyzed by Southern hybridization, cloning, and nucleotide sequencing. The two natural plasmids described below are designated pBTc2 and pBPs2, after their source B. aphidicola(T. caerulescens) and B. aphidicola(P. spyrothecae, respectively.

Characterization of pBTc2.An EcoRI shotgun cloning of B. aphidicola (T. caerulescens) plasmid DNA yielded four recombinants with an insert of 3.0 kb. Restriction fragment analysis showed these to be identical, and one clone was completely sequenced. A physical map of plasmid pBTc2 is presented in Fig. 1A. The G+C content of the 3.0-kb fragment is 20.74 mol%, which is similar to the 21.2 mol% of plasmid pBTc1 from this species (49) and is consistent with the low G+C content (28 mol%) of Buchnera genomic DNA (28). Two open reading frames (ORFs) were found, which correspond to trpE and trpG. The deduced amino acid sequences are 52 and 53% identical to TrpE and TrpG, respectively, from Buchnera of the phylogenetically treated aphid S. chinensis (Pemphigidae). Potential regulatory elements for gene expression (33) could not be identified.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Fig. 1.

Linearized physical maps of B. aphidicola trpEG plasmids. All genes are transcribed in the rightward direction. Open arrowheads indicate the approximate position and orientation of DnaA boxes. (A) Restriction site and genetic map of pBTc2 from B. aphidicola(T. caerulescens). The region denoted ori?contains the putative origin of replication. Restriction enzyme sites:X, XbaI; B, BglII;E, EcoRI. (B) Genetic maps of the repeated units of B. aphidicola (Aphididae) trpEGplasmids (adapted from reference 42 with permission from the publisher). B(sp.), species names (see below); trpEG/p, number of repeat units per plasmid (42); ori-3.6, putative origin of replication; Ψ, repeat units containing trpEG pseudogenes (32); striped box and ΨrepAC, location of a putative repAC pseudogene; open rectangle, 19-bp element similar to consensus sequence of RepA/C iterons. Species abbreviations: Dn, Diuraphis noxia; Ap, Acyrthosiphon piseum; Sg, Schizaphis graminum; Rp, Rhopalosiphum padi; Rm, Rhopalosiphum maidis.

Most trpEG-containing plasmids of Buchnera from aphids of the family Aphididae share a conserved 500- to 600-nucleotide (nt) sequence immediately upstream of trpE which has features characteristic of an origin of replication (29). The region has been designated ori-3.6 and contains three DnaA boxes (consensus 5′-TTATCCACA-3′) and regions of low G+C content (Fig. 1B) (43). A similar region, 717 nt in length, is present downstream of trpE on pBTc2 and may therefore contain the origin of replication of the plasmid. It has a very low G+C content (11 mol%) and contains five DnaA boxes in direct orientation, only one of which deviates by a single nucleotide from the consensus sequence. Beyond a high A+T content, the region has no further sequence similarity to ori-3.6.

B. aphidicola (Aphididae) trpEGplasmids are typically composed of tandemly repeated, identical 3.2- to 3.6-kb units, each carrying a copy of trpEG andori-3.6 (42, 43). To determine the composition of pBTc2, a Southern hybridization analysis was performed of variousEcoRI-digested and undigested B. aphidicola (T. caerulescens) DNA preparations, using cloned pBTc2 as a probe. For each preparation, two strongly hybridizing bands were detected in lanes with undigested DNA and a single strong band was detected in lanes with EcoRI-digested DNA (Fig. 2). These three bands correspond to, respectively, the open-circular, covalently closed supercoiled, and linear forms of a plasmid of 3.0 kb. In addition, three larger, more weakly hybridizing bands were detected in undigested DNAs. Their sizes correspond to the different forms of a plasmid of 6.0 kb. EcoRI digestion converted all the different forms into a single linear fragment of 3.0 kb. These findings suggest that pBTc2 consists mainly of one copy of the sequenced 3.0-kb unit but that a fraction of the plasmid population is present as a dimer. In one preparation, molecules of an even higher molecular weight were detected (Fig. 2, right lane), indicating that further multimerization of the plasmid occurs.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Fig. 2.

Southern blot analysis of EcoRI-digested (E) and undigested (u) B. aphidicola (T. caerulescens) DNA preparations. Cloned pBTc2 was used as a hybridization probe. The different forms of the putative plasmids are marked: open circular (oc), linear (L), and covalently closed circular (ccc).

Identification of pBPs2.Restriction enzyme analysis of a plasmid DNA preparation from B. aphidicola(P. spyrothecae) indicated the presence of at least two plasmids. One of these (pBPs2) was larger than 12 kb as estimated from the migration of bands of undigested plasmid DNA in ethidium bromide-stained agarose gels. Analogous to trpEG plasmids ofB. aphidicola (Aphididae), it seemed to be composed of repeated elements, because summation of three restriction fragments generated with XbaI (2.1, 1.8, and 1.5 kb) added up to only 5.4 kb. Moreover, the 1.8-kb band stained more intensely than the other two fragments. Southern hybridization with atrpG-containing probe revealed that the plasmid carried sequences homologous to trpG and reinforced the assumption that it was composed of repeated elements (Fig.3). AccI and PvuII each yielded a single, slightly blurred band of 13 to 15 kb, whileEcoRI, PstI, ScaI, and XhoI yielded two bands, one of which also migrated at about 13 kb and the second of which migrated at more than 23 kb. Digests withEcoRV, ClaI, and XbaI yielded either one or two bands migrating at 1.5 to 2.1 kb. These hybridization patterns were at first explained in terms of the 13- to 15-kbAccI and PvuII bands representing the linear form of the putative plasmid. The fact that the plasmid could be linearized, in contrast to most previously characterizedtrpEG plasmids (42), implied that it contained a single-copy region in addition to repeated elements. Recognition sites for EcoRI, PstI,ScaI, and XhoI are absent from the plasmid, and hence the two detected bands correspond to the open-circular (at 23 kb) and the comigrating linear and closed-circular forms of a 13- to 15-kb plasmid. Finally, recognition sites for EcoRV,ClaI, and XbaI are present in the repeated part of the plasmid and hence convert it into the constituent repeat units.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Fig. 3.

Southern blot analysis of B. aphidicola (P. spyrothecae) plasmid DNA. Hybridizations were performed with a probe complementary totrpG. Restriction enzymes: A, AccI; P,PvuII; B, BamHI; H, HindIII; V,EcoRV; C, ClaI; X, XbaI; E,EcoRI; T, PstI; K, ScaI; O,XhoI. oc, open circular; L, linear; ccc, covalently closed circular.

Cloning and sequencing. B. aphidicola(P. spyrothecae) plasmid DNA was shotgun cloned withXbaI. Screening of 108 recombinant colonies showed that 10% of plasmids had an insert of 2.1 kb and 20% had an insert of 1.8 kb, corresponding in size to the two XbaI fragments detected in the Southern hybridization (Fig. 3). The 1.5-kb XbaI fragment, initially detected in agarose gels, was not found. Restriction enzyme analysis and end sequencing of the recombinant plasmids suggested that all clones of the same size were identical and one of each group was completely sequenced (Fig.4B).

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Fig. 4.

Physical map of cloned and sequenced portion of pBPs2 from B. aphidicola (P. spyrothecae). (A) Restriction site and clone map. Double-headed arrows, clonedXbaI restriction fragments; arrows, oligonucleotide primers used for PCR; dotted line, cloned PCR product. Restriction enzyme sites: A, AccI; P, PvuII;S, SacI; C, ClaI;V, EcoRV. (B) Genetic map. All genes are transcribed in the rightward direction. The region denotedori? contains the putative origin of replication. Arrows, 19-bp iteron unique to pBPs2; arrows between brackets illustrate the copy number of the same iteron in the 1.70-kb PCR fragment; open rectangle, 19-bp element differing in one nucleotide from consensus sequence of RepA/C iterons; triple strip, 129-bp sequence identical to the 3′ end of trpE.

The 2.1-kb fragment contained three ORFs, the first two of which corresponded to the 3′ half of trpE and a complete copy oftrpG, with the stop and start codons of the two genes overlapping. BLASTP homology searches with the putative product of the third ORF yielded a high probability score (i.e., 893) with the N-terminal part of replication initiation protein RepA encoded byE. coli plasmid RA1, which belongs to the broad-host-range IncA/C incompatibility group (34). The amino acid sequences were 69% identical over the first 243 positions. The B. aphidicola (P. spyrothecae) gene is designatedrepAC₁ . The 1.8-kb fragment also contained three ORFs, the first of which was similar to the gene encoding the C-terminal part of the same RepA protein that was lacking from the 2.1-kb fragment. Because the coding sequences are in frame, the two fragments are most probably contiguous on the plasmid. The second ORF corresponded to a complete copy of the trpG gene, and the third ORF, again, corresponds to the gene encoding the N-terminal part of RepA. Both sequences were completely identical to the homologous sequences in the 2.1-kb fragment.

The 2.1-kb XbaI fragment contained single AccI and SacI restriction sites (Fig. 4A). Since the former enzyme was at first assumed to linearize the plasmid, this implied that the fragment contained part of a single-copy region of the plasmid. However, neither of the two sequenced fragments contained a restriction site for PvuII, suggesting that this site had to be present in the uncloned 1.5-kb XbaI fragment and, hence, that the 1.8-kb fragment represented the repeated unit of the plasmid. In addition, the uncloned 1.5-kb XbaI fragment was expected to contain the missing 5′ end of the trpE gene and possibly the 3′ end of a repAC gene. Oligonucleotide primers complementary to trpE and repAC₁ that would allow the amplification of this fragment were designed (Fig. 4A).

PCR yielded two products, which were 1.53 and 1.70 kb in size. After cloning, one recombinant plasmid from each size group was completely sequenced on both strands (Fig. 4B). Both fragments contained the 3′ end of a repAC gene and the missing 5′ part oftrpE, as well as a single PvuII site. However, the 3′ end of the repAC sequences differed slightly from the homologous sequence of the 1.8-kb XbaI fragment. The latter contained only one copy of a 19-bp element (5′-CAACAAACTATAAAAAAAA-3′), located 30 nt upstream of the stop codon TAA (Fig. 4B). In the smaller of the two PCR products, this element occurred at exactly the same position within the coding sequence but was directly followed by three additional copies, the second of which contained the stop codon (TAA) of repAC. The 1.7-kb PCR product contained 12 copies of the element. Since the third copy again included the stop codon, the repAC coding sequences of the two PCR products were identical, and this variant of the gene is designated repAC₂ . The presence of a repeated 19-bp element resembles the iterons found in the origin of replication of the RepA/C replicon of plasmid RA1 (34). The difference of eight copies of this element in the region downstream ofrepAC₂ explains the size difference of about 160 bp between the two PCR products. This length variation most probably reflects a heterogeneous plasmid population within our DNA preparation (see below).

The total length of the three assembled contiguous fragments, including the sequence derived from the smaller, 1.53-kb PCR product, is 5,477 bp (Fig. 4B). The sequence has a G+C content of 25.6 mol% and 85% is occupied by putative ORFs. In contrast to pBTc2, putative −35 (TTGAC) and −10 (AGTTAA) promoter elements could be identified, 125 bp upstream of trpE. The elements are separated by 16 bp and resemble previously identifiedBuchnera promoters (32). A second possible −10 box (TAACTA), which is identical to the E. coli trp −10 sequences (33), was found overlapping 3 bp with the first. The intergenic region betweenrepAC₁ and trpG contained a 129-nt sequence that was identical to the 3′ end of trpE. Similar “remnants” of trpE have been described for thetrpEG plasmid of B. aphidicola(Uroleucon sonchi) (4). The intergenic region between repAC₂ and trpE was highly AT rich (85 mol%) and contained one copy of a 19-bp element (5′-TATATGGGAATGTTGCACA-3′) that differs in only one position from the consensus sequence (5′–-AT-TGGG---G-TGCACG–3′) of a 13-copy iteron described for the origin of replication of E. coli plasmid RA1 (Fig. 4B) (34, 35). In addition,ori of this replicon contains one copy of a DnaA box. This element was absent from our sequence, but it may not be required for replication (40). The high A+T content and similarity with respect to the presence of iterons suggest that therepAC₂-trpE intergenic region contains the origin of replication of pBPs2.

Structure of pBPs2.To establish the overall structure of the plasmid, we performed the following experiments. First, quantitative hybridizzations were carried out to determine the relative abundance of the two XbaI fragments detected in the initial hybridizations (Fig. 3). Densitometric assays of Southern blots containing XbaI digested plasmid DNA indicated that the 1.8-kb fragment was 5.19 ± 0.97 times more abundant than the 2.1-kb fragment (data not shown). A copy number of 5 or 6 of the 1.8-kb fragment relative to one copy each of the 2.1- and 1.5-kb fragments adds up to a plasmid size of either 12.8 or 14.6 kb, which was in good agreement with the initial estimations. Second, partial XbaI digestions of plasmid DNA predigested with AccI plusSacI resulted in a ladder of bands in which each step consisted of two bands differing 200 bp in size (Fig.5A). The size difference between the double steps corresponded exactly to multiples of 1.8 kb plus 1.6 kb and multiples of 1.8 kb only. These findings were in agreement with the expectation based on the restriction site map (Fig. 4A), the 1.6-kb fragment resulting from the SacI site in the 2.1-kbXbaI fragment.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Fig. 5.

(A) Southern blot analysis of partial digestion ofAccI-SacI-digested B. aphidicola (P. spyrothecae) DNA by using serial dilutions of XbaI. Lanes: 1 to 8, digestions with 5.0, 2.5, 1.24, 0.63, 0.31, 0.16, 0.08, and 0.04 U of XbaI, respectively, for 15 min. X, A, S, control digestions performed to completion (X, XbaI; S, SacI; A, AccI. Arrows indicate the two principal forms of the plasmid. (B) Proposed structure of pBPs2 from B. aphidicola (P. spyrothecae) containing five tandem repeats of a 1.8-kb unit. Boxes with a similar pattern represent identical genes. All genes are transcribed in a clockwise direction. The region denotedori? contains the putative origin of replication. Arcs correspond to restriction fragments removed from molecules withAccI and SacI before partial digestion experiment with XbaI. The restriction enzyme sites shown are as in Fig. 4A.

Control digestions with AccI and SacI revealed that the plasmid population was in fact heterogeneous and was composed of molecules of different sizes. A single digestion withSacI, which linearizes the plasmid, yielded a short ladder of bands, differing in size by steps of 1.8 kb (Fig. 5A). The two most strongly hybridizing bands were 12.8 and 14.6 kb, corresponding to molecules containing either five or six copies of the 1.8-kbXbaI repeat, respectively. Four more weakly hybridizing bands measured approximately 18.2, 16.4, 11.0, and 9.2 kb, corresponding to molecules containing 8, 7, 4, and 3 copies of the same unit. A similar ladder of bands was observed for AccI, although all the bands were 1.3 kb smaller, which is consistent with the presence of two sites, 1.3 kb apart, in the single-copy region of the plasmid (Fig. 4A). Superimposed on the length variation due to differences in repeat number is the length variation accounted for by the difference in copy number of the 19-bp direct repeat downstream ofrepAC₂ . The inferred overall structure of a plasmid containing five tandem repeats of the 1.8-kb XbaI unit is presented in Fig. 5B.

Similarity between pBPs2 repAC and B. aphidicola (Rhopalosiphum) trpEGplasmids.The single-copy region betweenrepAC₂ and trpE of pBPs2 was found to share similarities with the origin of replication of plasmid RA1, but no significant similarities to the 500- to 700-nt putativeori of pBTc2 or the B. aphidicola(Aphididae) trpEG plasmids were observed. However, a comparison of the repAC sequence to the latter group of plasmids revealed a striking similarity (73% identical positions in a 955-bp overlap) to the region immediately upstream ofori-3.6 of the B. aphidicola(R. maidis) plasmid (Fig. 1B). Analysis of the recent update of this sequence (provided by P. Baumann) showed that this plasmid in fact carries a potentially intact repAC gene, whose deduced amino acid sequence was 71% identical to the two pBPs2 RepAC sequences and 64% identical to the plasmid RA1 homologue (Fig.6). In addition, the 19-bp element (5′-TATATGGGAATGTTGCACA-3′) present in a single copy in the putative ori of pBPs2 and in 13 copies in the oriof RA1 (34, 35) also occurs in the ori ofB. aphidicola (R. maidis) plasmid, with only a 1-bp difference from the pBPs2 sequence.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Fig. 6.

Multiple alignment of translations ofrepAC-like sequences. Black background indicates stop codons (x). The C-terminal residues differing between RepAC₁and RepAC₂ are shown in boldface type. Ps_RepAC,repAC of pBPs2; Rm_RepAC, repAC oftrpEG plasmid of B(R. maidis); RA1_RepAC, repAC of E. coliplasmid RA1.

A subsequent comparison of B. aphidicola(R. maidis) repAC to the region upstream ofori-3.6 from the closely related B. aphidicola (R. paid) trpEG plasmid yielded a similarity of 71% in an 843-bp overlap. The latter sequence, however, did not contain an uninterrupted reading frame. A putative remnant of the 19-bp element (5′-TATACGAGAATAGTGCACA-3′) was also found in exactly the same position as in the B. aphidicola (R. maidis) sequence. The similarity of repAC to the upstream region of ori intrpEG plasmids from other species was much lower (44 to 59%), and the 19-bp element or remnants thereof could not be identified. Nevertheless, the high sequence similarity in B. aphidicola (Rhopalosiphum) suggests that at least the B. aphidicola (R. padi) sequence contains an repAC pseudogene. This mode of gene silencing has a well-documented precedent in the trpEGplasmids of Buchnera from the aphids Diruaphis noxia (32) and Uroleucon sonchi(4). Both plasmids carry only one functional copy oftrpEG, in addition to a variable number of repeat units containing trpEG or trpE pseudogenes.

The multiple alignment further illustrates differences among the termini of the RepA sequences (Fig. 6). RepA from plasmid RA1 has an additional 33 residues at its N-terminal end, while the B. aphidicola (P. spyrothecae) sequences are longer at their C-terminal end. Remarkably, however, a high similarity to the RA1 sequence is retained at the nucleotide level after the stop codon, in the second reading frame of the RA1 sequence. The similarity drops immediately after the stop codon of the pBPs2 repACsequences. Because it is unlikely that Buchnera would have retained such a high similarity in a noncoding sequence, it is possible that the published RA1 sequence contains a sequencing error. A reestimation of the size of RepA in the sodium dodecyl sulfate-polyacrylamide gel electrophoresis analysis of Fig. 2 of Llanes et al. (35) yielded a molecular mass of 42 kDa, in contrast to the 35-kDa estimate reported by the authors. Our estimate corresponds exactly to a RepA protein of 366 residues (Fig. 6).

TrpEG phylogenetic relationships.A phylogenetic analysis ofBuchnera TrpE and TrpG sequences was conducted to assess the relationships of the newly sequenced genes. In accordance with previous 16S rDNA phylogenies (39, 49), the TrpG sequences from theB. aphidicola (Pemphigidae) form a cluster (Fig. 7). In the TrpE analysis, the outgroup sequences (E. coli and Salmonella typhimurium) rooted the tree within the B. aphidicola (Pemphigidae). The branch lengths of the groupings around the root were relatively short, particularly in the TrpE phylogeny, while the longer branches leading to the B. aphidicola (T. caerulescens) sequences indicate that the substitution rate in this lineage has been higher than in any of the other lineages.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Fig. 7.

Phylogenetic trees constructed by the neighbor-joining method (23) from estimated distances between TrpE and TrpG sequences. Species designations: Ec, Escherichia coli; St,Salmonella typhimurium; Ps, B. aphidicola (Pemphigus spyrothecae); Tc,B. aphidicola (Tetraneura caerulescens; Sc, B. aphidicola(Schlechtendalia chinensis); Dn, B. aphidicola (Diuraphis noxia); Ap, B. aphidicola (Acyrthosiphon pisum); Sg,B. aphidicola (Shizaphis graminum); Rp, B. aphidicola (Rhopalosiphum padi); Rm, B. aphidicola(Rhopalosiphum maidis).

Properties of B. aphidicola (Pemphigidae)trpEG plasmids. B. aphidicola (T. caerulescens) carried a 3.0-kb plasmid (pBTc2) which resembled the trpEG plasmids found in Buchnera from the Aphididae. With the exception of the B. aphidicola (R. maidis) plasmid (see below), none of these plasmids encodes proteins potentially involved in replication. The distinctive feature of their putative origin of replication is the presence of multiple DnaA boxes.B. aphidicola (P. spyrothecae) contained trpEG-encoding plasmids (pBPs2) with a gene content and putative replicon different from those described above, including a putative replication gene that was highly similar to the replication initiation gene repA of the RepA/C replicon from E. coli plasmid RA1 (34) (Fig. 6). The origin of replication of the RepA/C replicon is located downstream of repA and contains iterons, which serve as binding sites for RepA and are required for replication. pBPs2 contains 4 or 12 copies of a 19-bp iteron that overlap the 3′ end of the repAC₂ gene and the downstream putative origin of replication (Fig. 4B). The sequence of this iteron was unique to pBPs2, but it also carries a single copy of the 19-bp element that constitutes the iteron of RepA/C.

Besides carrying multiple copies of this replication gene, a noticeable feature of pBPs2 is the amplification of trpG overtrpE. This only further occurs in thetrpEG plasmid of B. aphidicola(U. sonchi) (4). Similar to this plasmid,trpG in the repeat unit of pBPs2 is preceded by a remnant oftrpE (Fig. 4B). This suggests either that a region containing only the 3′ end of trpE was initially duplicated or that after the initial duplication of an entirerepAC-trpEG unit, trpE was silenced through deletion. In either case, we suspect that the amplification is merely a by-product of recombination. A functional explanation would have to invoke an involvement of TrpG in a pathway other than tryptophan biosynthesis. In some bacteria, includingPseudomonas acidovorans, Bacillus subtilis, andAcinetobacter calcoaceticus (13), TrpG participates in both anthranilate synthase andp-aminobenzoate synthase activities. The latter enzyme synthesizes p-aminobenzoate, which is a component of the vitamin folic acid (50). In most other bacteria, however, including the closest known relative of Buchnera,E. coli (47), this activity is provided by a different though significantly similar protein (PabA) (13).

Structural variation of trpEG plasmids.The most conspicuous characteristic of Buchnera trpEGplasmids is their structural fluidity. B. aphidicola (Aphididae) contains plasmids of a concatenate structure, consisting of 4 to 10 tandem duplications of a basic, 3.6-kbtrpEG-containing unit (42). Exceptions areB. aphidicola (R. maidis), which contains a plasmid consisting of only one 3.6-kb unit (42), and B. aphidicola (D. noxia) and B. aphidicola (U. sonchi), which carry plasmids consisting of one 3.2-kb unit encoding a functional copy of TrpEG and a variable number of repeat units carrying trpEG pseudogenes (4, 32). The plasmid population in B. aphidicola (T. caerulescens) was composed mainly of molecules containing a single 3.0-kb trpEG unit, but Southern hybridization indicated the presence of multimers (Fig. 2).

The molecular basis of concatenation in these plasmids is unknown, but it is likely to be initiated by multimerization, which involves homologous recombination between plasmid monomers (25, 46). An indication that recombination (and mismatch repair) occurs in thetrpEG plasmids comes from the observation by Lai et al. of gene conversion among trpEG pseudogenes (32). Multimerization of low-copy-number plasmids increases the chance of their loss at cell division (46). Several maintenance systems are known to counteract this risk, including multimer resolution and partitioning systems (46, 51). There is no evidence for the presence of such systems on thetrpEG plasmids. However, stable inheritance may not be essential if selection for trpEG copy number does not act primarily on individual cells. In its intracellular environment, the slowing dividing Buchnera (3) may be viable with a reduced number of trpEG plasmids or even with none. This could shift the need for plasmid stability in individual cells to the need for partitioning bacteria with an adequate genetic composition to the next generation of aphids. We speculate that selection acting at this higher level underlies the observed structural variability of Buchnera trpEG plasmids.

Phylogenetic analysis of TrpEG.The fact that pBTc2 was more similar to the B. aphidicola (Aphididae)trpEG plasmids than to pBPs2 raised the possibility thatB. aphidicola (T. caerulescens) acquired its plasmid through horizontal transfer. Phylogenetic analysis, however, showed that the plasmid-borne sequences ofB. aphidicola (T. caerulescens) and B. aphidicola (P. spyrothecae) are most closely related to the chromosomal TrpEG sequence ofB. aphidicola (S. chinensis) (Fig.7). This refutes horizontal acquisition of pBTc2 by B. aphidicola (T. caerulescens). The slight incongruence at the root between the TrpE and TrpG topologies can be attributed to an accelerated substitution rate in B. aphidicola (T. caerulescens), causing its long branch to be “artifactually attracted” (24) by theB. aphidicola (Aphididae) sequences in the case of the TrpE phylogeny.

Replicons in Buchnera trpEG plasmids.Two replicons are possibly linked to the amplification oftrpEG: (i) a RepA/C-like replicon, in which replication is most probably initiated by plasmid-encoded RepAC, and (ii) anori-3.6-like replicon, which is characterized by the presence of multiple DnaA boxes in the ori and for which DnaA is thought to be involved in initiation of replication (30,42, 43). After the discovery of RepA/C in B. aphidicola (P. spyrothecae), computer analysis of previously characterized trpEG plasmids (42) revealed the presence of a putative intactrepAC gene in B. aphidicola(R. maidis) and a repAC pseudogene inB. aphidicola (R. padi) (Fig.1B and 6). In addition, ori-3.6 of both plasmids contained a single copy of the 19-bp iteron of the RepA/C replicon. None of these elements could be identified unequivocally in any of the otherB. aphidicola (Aphididae) plasmids or in pBTc2. The phylogenetic distribution of the two replicons is summarized in Fig. 8.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Fig. 8.

Phylogenetic distribution oftrpEG-associated replicons in Buchnera from aphids of the families Pemphigidae [B(P.)] nd Aphididae [B(A.)]. Species designations are as in Fig. 7. Location: p, plasmid; c, chromosomal. Replicon designation: R, RepA/C; ΨR, RepAC pseudogene; o; ori-3.6 like.

The presence of conserved elements of both replicons in theB. aphidicola (R. maidis) plasmid and a repAC pseudogene in its closest relativeB. aphidicola (R. padi) demonstrates that repAC can be silenced in the presence of an ori-3.6-like replicon. The silenced repAC gene of B. aphidicola (R. padi) apparently diverged rapidly from the B. aphidicola (R. maidis) sequence through the accumulation of substitutions (only 71% similarity at the nucleotide level). This observation, together with the fact that the noncoding regions between trpG and ori-3.6 are often unusually long [up to 1 kb in B. aphidicola (A. pisum) (Fig. 1)], raises the possibility that a RepA/C-like replicon was ancestral to allB. aphidicola (Aphididae) plasmids but that the respective repAC genes have been silenced in parallel lineages and have diverged beyond recognition.

Evolution of trp in Buchnera. Lai et al. (32) have previously proposed a scenario for the evolution of trp in Buchnera. In brief, they hypothesized that the ancestor of all present-day Buchnera species had the trp genes arranged in two linkage groups [trpEG andtrpDC(F)BA] on its chromosome, an organization still found in B. aphidicola(S. chinensis) (31). trpEG was relocated to a plasmid in the ancestor of the B. aphidicola (Aphididae), and the resulting enhanced capacity of tryptophan production promoted increased growth and reproductive rates of the insect host. This scenario must be amended in the light of two significant findings made in the present study. Firstly, trpEG plasmids are found in Buchneraspecies from the two most divergent lineages of aphids, i.e., the Aphididae and Pemphigidae. This suggests that selection for amplification of trpEG in Buchnera may be as ancient as the symbiosis itself. Second, similar replicons are found in both lineages (Fig. 8). On the basis of these findings, we hypothesized that the transfer of trpEG to a plasmid carrying a RepA/C-like replicon occurred in the ancestor of all present-dayBuchnera species. Like the related E. coliRA1 plasmid, the origin of replication may have contained a single DnaA box. An ori-3.6 like origin of replication evolved de novo on this plasmid through amplification of the resident DnaA box. (Note that, in contrast to the RepA/C-like replicon and therepA₁ replicon linked to leucine plasmids [49], there are no reports on bacterial replicons related to the ori-3.6-like replicon.) De novo evolution ofori-3.6 must have occurred independently in the lineages leading to B. aphidicola (T. caerulescens) and B. aphidicola(Aphididae) but not in B. aphidicola (P. spyrothecae). In the presence of ori-3.6 and its hypothesized involvement in a regulatory mechanism of initiation of chromosome replication (32), control over replication of thetrpEG plasmid was more advantageously exerted by the DnaA protein than by RepAC. This led to silencing of the ancestral RepA/C replicon in most lineages. Although it has yet to be established whether repAC is functional in B. aphidicola (R. maidis), it remains difficult to explain why the elements of RepA/C are much more highly conserved in this species than in any of the other B. aphidicola (Aphididae) strains.

Our scenario implies that trpEG in the lineage leading to B. aphidicola (S. chinensis) must have been transferred back into the chromosome (Fig. 8) rather than that its organization represents the ancestral state (31). In E. coli, the closest-known relative ofBuchnera (47), all the genes of thetrp pathway are organized into a single operon (13). If a similar organization was present in the presymbiotic ancestor of Buchnera, then the proposed reversal in B. aphidicola (S. chinensis) provides an explanation for the organization of itstrp genes into two separate linkage groups.

Finally, our scenario would predict that (i) trpEGplasmids, containing elements of either RepA/C or ori-3.6like replicons (or both), are likely to be found in Buchnerastrains from other families of aphids and (ii) in lineages wheretrpEG is chromosomally encoded, its location is expected to be different from the one found in B. aphidicola (S. chinensis).