Surface display of a massively variable lipoprotein by a Legionella diversity-generating retroelement

Anatomy of a Legionella DGR.

The DGR in Fig. 1A is located within a 10-kb genomic island in the chromosome of L. pneumophila strain Corby. Flanking genes are predicted to encode the heavy metal transport system, cadmium translocating P-type ATPase (cadA–C), and a type IV secretion system with associated regulators, legionella vir region (lvrA–C, 10). The island shows signs of recent horizontal acquisition, as indicated by a G + C content (45%) that differs from the rest of the genome (38%), and the presence of a transposase and an unrelated RT (11). Interestingly, the 10-kb DGR-containing island is located within a larger (64-kb) integrative and conjugative element (ICE) that has recently been shown to be capable of horizontal transfer (10). The retroelement itself encodes a DGR-type RT (2), an Avd homolog, cognate TR, and VR sequences that differ at sites corresponding to adenines in TR, and tandem stem-loop structures downstream of VR (6, 7). The 148-bp TR contains 43 adenines that most often occupy the first two positions of AAC or AAT codons, allowing maximal amino acid diversity while excluding the possibility of nonsense mutations generated by adenine mutagenesis (Fig. 1B). Following mutagenic homing, the L. pneumophila TR can theoretically generate 4⁴³ (∼10²⁶) unique DNA sequences capable of encoding ∼10¹⁹ different polypeptides—a repertoire of massive proportions. VR-encoded sequences are located at the C terminus of LdtA, within a domain predicted to adopt a CLec fold similar to the distal end of the BPP-1 phage tail fiber (Fig. S1). Two protein-trafficking signals are predicted at the N terminus of LdtA: the first is a TAT motif that differs from the consensus in Escherichia coli but is characteristic of known and putative TAT substrates in L. pneumophila (12, 13); the second is a lipobox that is also noncanonical. The TAT pathway is an alternative secretion system found in plants and bacteria that can translocate folded proteins or complexes across membranes (12), and lipobox motifs mediate signal peptide cleavage, lipid modification, and anchoring to the inner or outer membrane (14). Although the ability of TAT and lipobox secretion motifs to function in concert has not been thoroughly characterized (15, 16), we hypothesized that the N terminus of LdtA mediates secretion, membrane localization, and, potentially, surface exposure.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

A Legionella pneumophila DGR. (A) A DGR in L. pneumophila Corby is found within a 10-kb genomic island (green dashed box) adjacent to heavy metal transport gene cluster (cadA–C) and type IV secretion system regulatory genes (lvrA–C). The DGR-containing genomic island has a higher G + C content (percentile score) than flanking regions, a non-DGR RT (dark blue), genes annotated as hypothetical proteins (gray), putative transposase (orange), and a gene annotated as a phage repressor (light green). DGR locus is expanded with the TP gene (ldtA); predicted secretion (TAT) and localization sequences (LPP); accessory protein gene (avd); and DGR-encoded reverse transcriptase (RT) identified. Additional DGR elements: DNA stem/loops, IMH, and IMH* are indicated. Black arrows represent mutagenic homing as transfer of nucleotide sequence information from TR to VR via an RNA intermediate and the predicted amino acid diversity. The LdtA signal peptide contains a polar N-region (residues 1–11), followed by a hydrophobic core (residues 12–15) and a nonpolar C-region (residues 17–23). A predicted noncanonical TAT motif, KKRHFFR, differs from the E. coli consensus [(S/T)RRxFLK]. The LdtA signal peptide contains a lipobox motif, FFSC, which is predicted to be lipid-modified (14). (B) Alignment of Corby DGR VR/TR shown in the ldtA reading frame. TR adenines (red) usually occupy the first two positions of AAC or AAT codons (shaded) and correspond to substitutions in VR. Nucleotide sequences downstream of the VR stop codon representing DNA stem/loop (underline) structures are shown.

L. pneumophila DGR Is Functional.

As a first step in characterizing the L. pneumophila DGR, we determined if it is capable of mutagenic homing. A PCR assay was used to detect RT-dependent transfer of an invariant sequence tag from TR to VR, and adenine mutagenesis was evaluated by sequencing amplified retrohoming products (Fig. 2). The avd, RT, and a modified TR containing a 20-bp PCR tag consisting of G + C residues (TR-GC) were expressed in trans to the wild-type DGR on a plasmid vector (pATR; Fig. 2B). Negative controls included constructs expressing wild-type TR (no tag) or a catalytically inactive RT. Following induction, DNA was extracted and homing products were amplified using primer sets that annealed to the GC tag and sequences upstream (P1/P3) or downstream (P2/P4) of VR. Homing products were readily detected and identified by transfer of the tag from TR-GC to VR (Fig. 2C), and sequences of GC-tagged VRs revealed adenine-specific mutagenesis (Fig. S2A). These results show that the L. pneumophila Corby DGR encodes functional components that are capable of catalyzing adenine-specific mutagenic homing to the VR region of ldtA.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

The L. pneumophila DGR is capable of mutagenic homing. (A) L. pneumophila Corby DGR cassette with components indicated as in Fig. 1A. (B) Experimental design. pATR encodes avd, TR with 20-bp GC PCR tag (red box), and RT expressed from Ptac. DGR-mediated transfer incorporates the GC tag into the chromosomal VR. Primer binding sites (P1–P5) for PCR-based mutagenic homing assays are shown, and deletions removing IMH, or disruption of the first (MSt1) or second (MSt2) stem/loop structure are indicated (black brackets). (C) DGR homing assays. Donor plasmids (pATR and derivatives) were transformed into wild-type L. pneumophila Corby, Ptac expression induced, and genomic DNA extracted and used for PCR with primers shown in B. Donor pATR derivatives encoded wild-type (wt) or mutant (RT−) RT alleles, with (+) or without (−) TR tags. Equivalent amounts of template DNA were used for PCR assays (P1 + P4). Nonspecific PCR amplification of donor plasmid indicated (red asterisk). (D) pATR constructs used in C and additional derivatives with deletions removing avd (avd−) or IMH (IMH−) were transformed into L. pneumophila Corby deleted for sequences from avd to RT (Δavd-RT), with IMH present (wt) or deleted (IMH−). DGR homing assays were conducted as in C with equal amounts of template DNA (P1 + P5). Nonspecific PCR amplification of donor plasmid indicated (red asterisk). (E) pATR donor plasmids with wild-type (wt) or mutant RT alleles (RT−) were transformed into L. pneumophila Corby Δavd-RT with wt or disrupted stem/loop structures (MSt1, MSt2). DGR homing assays were conducted as in C with equal amounts of template DNA (P1 + P5).

In the experiment in Fig. 2C, efficient transfer of the GC tag was detected following expression of avd, TR-GC, and RT in trans to the chromosomal DGR. To measure activity of the native element, we inserted the 20-bp GC tag into the chromosomal TR by allelic exchange, and assayed mutagenic homing by PCR. Transfer of the GC tag and adenine mutagenesis was both observed (Fig. S2 B and C), but detection required an increase in the number of PCR amplification cycles compared with the experiment in Fig. 2B (25 vs. 35 cycles). The low level of activity of the native chromosomal element was not surprising. DGR-mediated mutagenesis is stochastic, and the vast majority of diversified target genes are likely to encode inactive products. Mutagenic homing is expected to be tightly controlled to prevent a loss of fitness due to overdiversification (Discussion).

Cis- and Trans-Acting Factors Required for Mutagenic Homing.

Our understanding of mutagenic homing derives almost exclusively from studies of the phage BPP-1 DGR (2, 7). We were curious to determine if observations with BPP-1 are applicable to other DGRs, especially those encoded on bacterial chromosomes. To explore requirements for mutagenic homing by the L. pneumophila DGR, avd, TR, and RT were deleted en bloc from the L. pneumophila genome, and pATR (Fig. 2B) or derivatives with mutations in trans-acting factors were tested in our PCR-based assay. The ability to detect homing products required the TR tag and expression of an active RT, and homing was also dependent on expression of avd (Fig. 2D). The avd gene is predicted to encode a small, basic protein (14.1 kDa, pI 9.3) with homologs in other DGRs, including BPP-1, where it serves an essential function and has been shown to form a positively charged pentameric barrel that interacts with RT (2⇓–4). IMH*, located at the 3′ terminus of TR, is a 32-bp sequence with two mismatches to IMH. In BPP-1, IMH and IMH* are essential components that are predicted to facilitate assembly of a complex between VR, TR-RNA, and RT during priming and reverse transcription (6, 7). Deletion of IMH* on the pATR donor plasmid, or IMH on the recipient chromosome, abrogated activity of the L. pneumophila DGR (Fig. 2D).

In BPP-1, target-site recognition is also dependent on the presence of an inverted repeat that flanks IMH and forms a hairpin/cruciform structure in supercoiled DNA (7). These elements are highly conserved in phage DGRs, where they consist of 7- to 10-bp GC-rich stems and 4-nt loops. A divergent yet potentially analogous element, composed of two tandem repeats with GC-rich stems of different lengths and identical 3-nt loops, is present downstream of the L. pneumophila ldtA IMH (Fig. 2B). To examine their role in homing, we generated mutants in which sequences in the 3′ halves of either stem were replaced with complementary nucleotides to disrupt base pairing. As shown in Fig. 2E, disruption of stem-loop 1 (SL1) or stem-loop 2 (SL2) eliminated homing. Our results show that the L. pneumophila and BPP-1 DGRs operate in a fundamentally similar manner, using conserved cis- and trans-acting components for adenine-specific mutagenic homing.

LdtA Is Surface-Exposed.

L. pneumophila Corby cells expressing LdtA or control proteins with C-terminal HA epitope tags were lysed and separated into soluble and membrane fractions. LdtA partitioned to the membrane fraction along with macrophage infectivity potentiator (MIP) [outer membrane (OM) protein control] (17) and defective organelle trafficking A (DotA) [inner membrane (IM) protein control] (18), whereas recombinase A (RecA) appeared in the soluble fraction (Fig. 3A) (19). On further separation by isopycnic gradient ultracentrifugation, DotA was enriched in fractions containing IM proteins, whereas MIP and LdtA preferentially partitioned to the OM (Fig. 3B). To determine the orientation of LdtA, intact bacterial cells were treated with proteases under conditions that preferentially degrade surface-exposed proteins (Fig. 3C). Cells were induced to express HA-tagged LdtA, MIP, DotA, RecA, or intracellular multiplication X (IcmX) as a periplasmic control (20), and incubated with increasing concentrations of proteinase K (21). MIP, an integral OM protein, showed moderate protease sensitivity, whereas periplasmic (IM) and cytoplasmic control proteins were relatively unaffected. In contrast, LdtA was highly sensitive to protease treatment, indicative of surface localization.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

LdtA is a surface-exposed lipoprotein. Epitope (HA)-tagged LdtA or control proteins were used for subcellular localization experiments. (A) Cells were lysed by French press and cellular constituents separated by high-speed centrifugation (I). Proteins with known subcellular locations: MIP (OM), DotA (IM), and RecA (cytoplasmic) were compared with LdtA. RecA was enriched in the soluble fraction(S), whereas LdtA, MIP, and DotA were enriched in membrane fractions(M). Membrane proteins detected in soluble fractions are due to overexpression and increased pools of secretion intermediates. (B) High-speed isopycnic sucrose gradient fractionation was used to separate membrane proteins and fractions were probed by Western blotting with anti-HA antibodies. Fractions corresponding to IM or OM are shown. (C) L. pneumophila Corby cells expressing epitope-tagged proteins were treated with increasing concentrations of proteinase K (0–200 µg/mL) to digest surface-exposed proteins, and whole-cell lysates were probed, as above. HA-tagged IcmX, a periplasmic T4SS component, was included as a control. (D) (Top) Mutation of the essential Cys in the LdtA LPP processing signal to Ala (C20A) or Ser (C20S) protected LdtA from digestion by proteinase K. (Middle) An in-frame deletion in tatB (Corby∆tatB) eliminated LdtA digestion, and complementation by plasmid-expressed tatB (ptatB) restored protease sensitivity. (Bottom) L. pneumophila Corby cells expressing epitope-tagged IcmX, LdtA C20A, or CorbyΔtatB expressing LdtA were permeabilized with EDTA/lysozyme and treated with increasing concentrations of proteinase K. (E) Indirect immunofluorescence detection of epitope-tagged surface proteins in L. pneumophila Corby. Cells were treated with anti-HA or anti-L. pneumophila antibodies under permeabilizing (+) (EDTA and lysozyme) or nonpermeabilizing (−) conditions, as indicated. EV, empty vector control. (F) (Left) Indirect immunofluorescence detection of epitope-tagged LdtA or LdtA-C20S mutant derivative under permeabilizing (+) or nonpermeabilizing (−) conditions. (Right) Indirect immunofluorescence detection of epitope-tagged LdtA in L. pneumophila Corby∆tatB in the presence or absence of a complementing plasmid (ptatB).

Indirect immunofluorescence was used as an independent approach to test surface exposure. Cells expressing full-length LdtA-HA were recalcitrant to visualization attempts, and we hypothesized this was due to sequestration of the epitope tag within a folded structure. Based on structural modeling predictions, we constructed a variant that expressed the first 370 amino acids of LdtA fused to a triple-HA C-terminal epitope tag (LdtA-370-3HA). Surface immunofluorescence was readily detected using intact cells, whereas visualization of the IcmX or DotA negative controls required OM permeabilization (Fig. 3E). In sum, membrane fractionation, protease sensitivity, and surface immunofluorescence support the conclusion that LdtA is an OM protein with a surface-exposed C terminus.

LdtA Is a TAT-Secreted Lipoprotein.

Translocation through the TAT pathway requires cytoplasmic recognition by the TatABC complex of signal sequences with the consensus motif SRRxFLK, which is conserved in plants and other bacteria but more variable in Legionella (12, 13). The predicted LdtA secretion motif retains requisite recognition components, including polar N-terminal residues, a hydrophobic core with the arginine required for TAT transport, and a nonhydrophobic C-terminal residue (Fig. 1A). To determine if TAT is required for LdtA secretion, we generated an in-frame deletion in L. pneumophila Corby tatB (22, 23). LdtA-HA was protease resistant in the ∆tatB strain, and protease sensitivity was restored by complementation with tatB (Fig. 3D). Confirmatory results were obtained by immunofluorescence (Fig. 3E).

Having shown that LdtA is secreted through the TAT pathway, we were curious to determine how it becomes localized in the OM. TAT motifs often contain a consensus sequence at their C terminus, Ala-Xaa-Ala, which mediates cleavage by signal peptidase I (12). In contrast, we identified a potential lipobox motif, FFSC, at the analogous position in LdtA (Fig. 1A); this suggested a hybrid signal that combines TAT translocation with signal peptide cleavage, lipid modification of the conserved cysteine (Cys-20), and trafficking to the OM by the localization of lipoprotein (LOL) system. To test this hypothesis, we first determined if Cys-20 is required for surface exposure. As shown in Fig. 3D, alanine (C20A) or serine (C20S) substitutions at this site rendered LdtA-HA protease resistant, with a corresponding loss of surface immunofluorescence (Fig. 3F). To determine the step at which trafficking is blocked, we compared the effects of OM permeabilization on phenotypes of Cys-20 substitutions in wt vs. ∆tatB mutants. Wild-type L. pneumophila expressing LdtA-C20A showed sensitivity to proteinase K under permeabilizing conditions but not in intact cells, as observed with the periplasmic control protein IcmX (Fig. 3 C–F). In contrast, LdtA-C20 was protease resistant in the ∆tatB strain, even after OM permeabilization. Immunofluorescence results paralleled protease sensitivity assays (Fig. 3 E and F) and showed that mutation of Cys-20 causes missorting to the periplasm, a phenotype that would be predicted to result from the lack of modification of an OM lipoprotein.

Bacteria often modify lipobox cysteines with palmitic acid by linkage to N-acyl-S-diacylglyceryl-Cys moieties (24, 25). Acylbiotin-exchange chemistry, which provides a sensitive alternative to labeling cells with radioactive fatty acids, was used to detect posttranslational modification of LdtA (Fig. S3 A and B) (25). Protein lysates from E. coli and wild-type or ∆tatB L. pneumophila cells expressing LdtA or LdtA-C20S were treated with hydroxylamine to remove acyl-linked lipids, which were then replaced with biotin. Biotinylated proteins were column purified, eluted, and analyzed by Western blotting. Biotin-labeled LdtA was found in wild-type L. pneumophila but not in ∆tatB mutants, and LdtA-C20S was refractory to biotinylation (Fig. S3B). Taken together, our results show that after secretion across the IM by TAT, LdtA is anchored in the outer surface of the OM by an acyl-linked lipid modification.

Distribution of DGRs in L. pneumophila Isolates.

Though the Legionella genus contains over 50 species, more than 84% of cases of legionellosis are caused by L. pneumophila serogroup I (Sg1) (26). The DGR in L. pneumophila Corby, a Sg1 clinical isolate, is absent in several other sequenced Sg1 strains, and conserved flanking sequences define the endpoints of the DGR genomic island (Figs. 1A and 4A). To further investigate the distribution of DGRs within L. pneumophila, we conducted a PCR screen of 12 additional Sg1 clinical isolates and found three that contained DGR-associated loci (Fig. S4); to determine their composition and architectures, multiplexed genomic libraries of three newly identified DGR-containing isolates were sequenced. Complete sets of conserved DGR components, including RT, avd, TR, VR, stem loops, IMH, IMH*, and TP loci were identified in strains D5572 and D5591. The third isolate, D5549, contained RT, TR, and avd genes without an identifiable linked TP. Given the ability of the diversifying machinery to act in trans (7), unlinked target proteins may exist in this isolate. In D5591, the DGR is located on a genomic island similar to the one in L. pneumophila Corby, with identical regulatory genes to the right (Fig. 4; Fig. S4D). A chromosomal rearrangement at the left boundary appears to have substituted different flanking loci and removed genes from the genomic island. For D5572 and D5549, DGRs are encoded on similar genomic islands but flanking sequences could not be identified, even at 50-fold coverage, because the high degree of plasticity in the L. pneumophila genomes precluded assembly of long stretches of contiguous sequence (Fig. S4C) (27).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

A subset of L. pneumophila strains contain DGRs. (A) The genomic island (dashed box) containing the L. pneumophila Corby DGR is absent in sequenced and assembled genomes of L. pneumophila strains 130b and Lens, and DGR and flanking genomic island sequences are present in Sg1 strains D5549, D5572, and D5591, which were partially sequenced in this study. (B) Sequence comparisons between DGRs. TP genes (ldtA–C), TAT/LPP secretion signals (green boxes), and Sec/LPP secretion signal (gray box) are shown. Other DGR components are indicated with percent identities between nucleotide (black) and amino acid sequences (red). Predictive programs (SI Material and Methods) identified similar stem loops (St1/2) in Corby and D5572, whereas D5591 contains a single, highly structured stem loop (St3). The predicted CLec folds in LdtA, B, and C are modeled in Fig. S1.

Fig. 4B shows details of the DGRs in three Sg1 isolates, and their comparison is quite interesting. RT and avd genes are predicted to encode nearly identical products, and the TR loci share 97% nucleotide identity. VR sequences are also similar, with differences at positions corresponding to adenine residues in their cognate TRs (Fig. S5). Protein threading predicts that the VR domains of LdtA, LdtB, and LdtC adopt similar CLec folds with binding pockets composed of diversified residues (Fig. S1). Though LdtA and LdtB are related throughout their entire length, including secretion and localization signals, sequences upstream of the LdtC VR have significantly diverged. Despite this, the extreme N terminus of LdtC is predicted to encode a Sec-dependent secretion signal followed by a lipobox motif, which could provide an alternative means for surface localization (28). These observations illustrate the modular nature of DGRs and is consistent with the notion that trans-acting factors and diversified scaffolds act in a generic manner to evolve ligand binding specificities, which can be adapted to different functions through connection with different N-terminal domains.

LdtA as a Model for Diversified Surface Lipoproteins.

The ability to introduce massive amounts of diversity in surface-anchored bacterial proteins seems likely to be an adaptive trait. In support of this idea, a partial survey of sequenced bacterial genomes revealed multiple DGRs that are predicted to diversify TPs with N-terminal lipoprotein sequences. As shown in Fig. 5, predicted DGR-diversified TPs in Treponema denticola, Bacteroides fragilis, Bacteroides thetaiotaomicron, Vibrio angustum, and Shewanella baltica contain N-terminal LPP motifs. Although LdtA is known, and LdtB is predicted, to be TAT secreted, the other TPs in Fig. 5 have Sec secretion signals. T. denticola encodes seven TPs that are particularly interesting; we predict they are diversified by a single DGR, and the majority includes highly predicted LPP processing signals. Spirochetes are known to make extensive use of the LPP secretion pathway to anchor proteins to their outer surface (21). Taken together, our results suggest that lipid modification and surface exposure are conserved features of proteins that are diversified by bacterial chromosomal DGRs.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Putative DGR target proteins in diverse bacteria are predicted lipoproteins. Species and protein ORFs are indicated. Analysis of putative TPs (28) identified N-terminal lipoboxes (red) and C-terminal VR sequences (colored arrows) found within predicted CLec domains (white box). The T. denticola 35405 genome encodes seven TPs (two are shown), which are predicted to be diversified in trans by a single DGR linked to TDE2269. All other TPs are encoded within their cognate DGRs. Numbers above arrows represent theoretical levels of nucleotide diversity generated by mutagenic homing.