Molecular Determinants of Substrate Selectivity of a Pneumococcal Rgg-Regulated Peptidase-Containing ABC Transporter.

Peptidase-containing ABC transporters (PCATs) are a widely distributed family of transporters which secrete double-glycine (GG) peptides. In the opportunistic pathogen Streptococcus pneumoniae (pneumococcus), the PCATs ComAB and BlpAB have been shown to secrete quorum-sensing pheromones and bacteriocins related to the competence and pneumocin pathways. Here, we describe another pneumococcal PCAT, RtgAB, encoded by the rtg locus and found intact in 17% of strains. The Rgg/SHP-like quorum-sensing system RtgR/S, which uses a peptide pheromone with a distinctive Trp-X-Trp motif, regulates expression of the rtg locus and provides a competitive fitness advantage in a mouse model of nasopharyngeal colonization. RtgAB secretes a set of coregulated rtg GG peptides. ComAB and BlpAB, which share a substrate pool, do not secrete the rtg GG peptides. Similarly, RtgAB does not efficiently secrete ComAB/BlpAB substrates. We examined the molecular determinants of substrate selectivity between ComAB, BlpAB, and RtgAB and found that the GG peptide signal sequences contain all the information necessary to direct secretion through specific transporters. Secretion through ComAB and BlpAB depends largely on the identity of four conserved hydrophobic signal sequence residues previously implicated in substrate recognition by PCATs. In contrast, a motif situated at the N-terminal end of the signal sequence, found only in rtg GG peptides, directs secretion through RtgAB. These findings illustrate the complexity in predicting substrate-PCAT pairings by demonstrating specificity that is not dictated solely by signal sequence residues previously implicated in substrate recognition.IMPORTANCE The export of peptides from the cell is a fundamental process carried out by all bacteria. One method of bacterial peptide export relies on a family of transporters called peptidase-containing ABC transporters (PCATs). PCATs export so-called GG peptides which carry out diverse functions, including cell-to-cell communication and interbacterial competition. In this work, we describe a PCAT-encoding genetic locus, rtg, in the pathogen Streptococcus pneumoniae (pneumococcus). The rtg locus is linked to increased competitive fitness advantage in a mouse model of nasopharyngeal colonization. We also describe how the rtg PCAT preferentially secretes a set of coregulated GG peptides but not GG peptides secreted by other pneumococcal PCATs. These findings illuminate a relatively understudied part of PCAT biology: how these transporters discriminate between different subsets of GG peptides. Ultimately, expanding our knowledge of PCATs will advance our understanding of the many microbial processes dependent on these transporters.

While substantial progress has been made in uncovering the mechanisms that allow PCATs to recognize GG peptides, comparatively little is known about how or if PCATs discriminate between different GG peptides. In addition to ComAB and BlpAB from pneumococcus, multiple PCATs have been shown to process and/or secrete multiple peptides with distinct signal sequences, sometimes even those from different strains or species (28,(32)(33)(34)(35). These data suggest that in general, PCATs are not particularly selective when it comes to choosing substrates.
In this work, we describe a previously uncharacterized locus in pneumococcus, rtg, which encodes the PCAT RtgAB and several GG peptides. This locus is regulated by the Rgg/SHP-like system RtgR/S, which provides a competitive fitness advantage during nasopharyngeal colonization. We demonstrate that RtgAB secretes the rtg GG peptides but not ComAB/BlpAB substrates and that ComAB or BlpAB cannot efficiently secrete the rtg GG peptides. Finally, we investigate the signal sequence determinants that selectively direct peptides toward either RtgAB or ComAB/BlpAB and show that a unique N-terminal motif is required for secretion by RtgAB. These findings shed light on how PCATs can use signal sequence motifs beyond the previously described conserved hydrophobic residues to distinguish different GG peptides.

RESULTS
Identification of an uncharacterized pneumococcal PCAT-encoding locus. As part of an effort to catalog the PCAT repertoire of S. pneumoniae, we searched pneumococcal genomes for putative PCAT genes that had not been previously described. One of the hits was CGSSp9BS68_07257 (henceforth 07257), a gene found in the clinical isolate Sp9-BS68 (36) (Fig. 1A). Upstream of 07257 is a gene oriented in the opposite direction predicted to encode an Rgg-family transcription regulator (12). We hypothesized that this regulator controls expression of 07257 and named the locus rtg (Rgg-regulated transporter of double-glycine peptides). We designated the transporter gene rtgA and the regulator gene rtgR. rtgR marks one end of the locus and is separated from a partially disrupted arginine biosynthesis cluster (37) by two transcription terminators. Downstream of rtgA are several genes arranged in a single operon. These include rtgB, which encodes a putative ComB/BlpB-like transport accessory protein, and the GG peptide genes rtgG, rtgT, rtgW1, and rtgW2. A transcription terminator separates the last gene, rtgD2, from a disrupted putative endoRNase gene and pspA. A different version of the rtg locus is found in the laboratory strain D39 (and its derivative, R6) but with a disrupted rtgA (Fig. 1A).
The Rgg/SHP-like pheromone pair RtgR/RtgS regulates rtg. We found that rtg expression is inhibited in mid-exponential phase during growth in the peptide-rich medium THY (Todd-Hewitt broth plus 0.5% yeast extract) but highly upregulated in the peptide-poor media CDMϩ (38) and RPMI (Fig. 1B). The start of rtg activation in both Sp9-BS8 and D39 during growth in CDMϩ occurs in early exponential phase at cell densities as low as an optical density at 620 nm (OD 620 ) of 0.01 (Fig. 1C). In contrast, rtg stays inactive in THY throughout the exponential and stationary phases. We concluded from these data that rtg is actively regulated, most likely by RtgR. Since Rgg regulators are often associated with peptide pheromones, we searched for and found an open reading frame (ORF) in Sp9-BS68 between rtgR and rtgA predicted to encode a SHP-like pheromone. D39 has two copies of the candidate pheromone: one located between rtgR and rtgA and the other downstream of rtgB. We named the only copy of the ORF in Sp9-BS68 and the first copy in D39 (between rtgR and rtgA) rtgS1 and the second copy in D39 rtgS2 (Fig. 1A).
Having identified a putative Rgg/SHP-like regulatory system, we sought to define the contributions of RtgR and RtgS to rtg regulation through deletional analysis. We monitored rtg activation in Sp9-BS68 ΔrtgS1, ΔrtgR, and ΔrtgS1 ΔrtgR strains during growth in CDMϩ and THY ( Fig. 2A). None of the mutants showed signs of rtg activation in either medium, indicating that both RtgR and RtgS promote and are required for rtg activation. In D39, rtgS1 encodes a peptide with a single amino acid change (S14L) compared to RtgS from Sp9-BS68, and rtgS2 encodes a peptide with a different single amino acid change (P27S) (Fig. 2B). We found that the D39 ΔrtgS1 and ΔrtgS1 ΔrtgS2 mutants failed to activate rtg in CDMϩ, while the ΔrtgS2 mutant was indistinguishable from the wild-type strain (Fig. 2C). This suggested that the P27S substitution in the rtgS2 product prevents it from activating rtg, while the S14L substitution in the rtgS1 product does not appreciably affect signaling. Therefore, we classified the rtgS1 product in both Sp9-BS68 and D39 as type A pheromone (RtgS A ) and the rtgS2 product in D39 as type B (RtgS B ).
To confirm that RtgS is the specific pheromone inducer of rtg, we performed dose-response assays using synthetic peptides corresponding to the C-terminal 8, 10, and 12 residues of RtgS A (RtgS A -C8, RtgS A -C10, and RtgS A -C12, respectively). All three synthetic peptides induced expression from the Sp9-BS68 rtgS1 promoter in both CDMϩ and THY, though the curves for the latter were shifted to the right in a manner consistent with pure competitive inhibition ( Fig. 2D; Table 1). We also confirmed that rtg induction by synthetic RtgS requires RtgR (Fig. 2E). In D39, RtgS A -C10 induces rtg similarly to Sp9-BS68, whereas RtgS B -C10 acts as a partial agonist with a 55-fold larger 50% effective concentration (EC 50 ) value than RtgS A -C10 ( Fig. 2F  Hairpins represent transcription terminators. GG peptide genes are marked with asterisks. (B) Activation of rtg in various growth media. A Sp9-BS68 P rtgA -luc reporter strain was grown in the indicated media, and luciferase activity was sampled when cells reached an OD 620 of 0.2. The plotted values were normalized against cell density and the luciferase activity from a strain harboring constitutively expressed luciferase grown in the same medium to account for signal difference due to medium alone. Plots show means Ϯ standard errors (SEs) from 3 independent experiments. ***, P Ͻ 0.001 by ANOVA with Tukey's honestly significant difference (HSD); AU, arbitrary units. (C) Timing of rtg activation in THY and CDMϩ. Sp9-BS68 (top) and D39 (bottom) P rtgS1 -luc reporter strains were grown in THY (red) and CDMϩ (black) and monitored for rtg activation (dark, left y axis) and cell density (light, right y axis). Plots show medians (lines) and 25% to 75% quantiles (shading) from 12 wells pooled from 3 independent experiments.
the Pro-to-Ser substitution in RtgS B interferes with signaling at a step following pheromone secretion. Although partial agonists can act as competitive antagonists of full agonists, we did not observe an inhibitory phenotype associated with RtgS B during natural rtg activation (Fig. 2C), and competitive dose-response assays showed that RtgS B -C10 only antagonizes RtgS A -C10 at likely supraphysiological concentrations (Ն256 nM) (Fig. 2G). Next, we determined that, consistent with previously described Rgg/SHP systems, rtg activation requires both the Ami importer and the PptAB transporter (see Fig. S1A in the supplemental material). We also showed that response to exogenous RtgS treatment requires Ami but not PptAB (Fig. S1B), consistent with their respective roles as pheromone importer and exporter. Sp9-BS68 P rtgS1 -luc reporters were grown in CDMϩ or THY and monitored for rtg activation (dark, left y axis) and cell density (light, right y axis). Plots show medians (lines) and 25% to 75% quantiles (shading) from 12 wells pooled from 3 independent experiments. (B) Translated rtgS gene products from Sp9-BS68 and D39. The type-defining residues are highlighted in red. (C) rtgS1 but not rtgS2 is required for rtg activation in D39. D39 P rtgS1 -luc reporters were grown in CDMϩ and monitored for rtg activation (dark, left y axis) and cell density (light, right y axis). Plots show medians (lines) and 25% to 75% quantiles (shading) from 30 wells pooled from 3 independent experiments. C-terminal fragments of RtgS A induce rtg in an RtgR-dependent manner. Sp9-BS68 P rtgS1 -luc reporters were grown in CDMϩ and THY (D) or CDMϩ only (E) to an OD 620 of 0.02 and treated with synthetic RtgS A fragments. Response was defined as the maximum observed P rtgS1 activity within 60 min of treatment. Data in panel D were fitted to the four-parameter Hill model (curves). Plotted data points represent means Ϯ SEs from 3 independent experiments. (F) RtgS B is a partial agonist at the rtg locus. A D39 ΔrtgS1 ΔrtgS2 P rtgS1 -luc reporter was grown in CDMϩ to an OD 620 of 0.02 and treated with synthetic RtgS fragments. Response was defined as the maximum observed P rtgS1 activity within 60 min of treatment. Data were fitted to the four-parameter Hill model (curves). Plotted data points represent means Ϯ SEs from 3 independent experiments. (G) High concentrations of RtgS B -C10 antagonize RtgS A -C10. A D39 ΔrtgS1 ΔrtgS2 P rtgS1 -luc reporter was grown in CDMϩ to an OD 620 of 0.02 and treated simultaneously with 2 nM RtgS A -C10 and various concentrations of RtgS B -C10. Response was defined as the maximum observed P rtgS1 activity within 120 min of treatment. Plotted data points represent means Ϯ SEs from 3 independent experiments. **, P Ͻ 0.01; ***, P Ͻ 0.001 versus 0 nM RtgS B -C10 by ANOVA with Dunnett's correction for multiple comparisons.
Determinants of Pneumococcal PCAT Substrate Selection ® Finally, mutagenesis of the Sp9-BS68 rtgS1 promoter region revealed the presence of two nearly identical promoters, each contributing partially to rtgS1 expression (see Fig. S2A and B). We also identified an inverted repeat found in both promoters which is required for RtgS-induced expression and likely represents the RtgR binding site ( Fig. S2A and C).
RtgAB secretes rtg-encoded GG peptides. Sp9-BS68 and D39 both harbor four putative GG peptides at the rtg locus (Fig. 1A). To determine whether RtgAB secretes the rtg GG peptides, we employed a HiBiT tag-based peptide secretion assay (25). We constructed autoinducing-deficient ΔrtgS1 ΔrtgS2 R6 strains which harbor HiBiT-tagged rtgC (from D39) or rtgG (from Sp9-BS68) placed downstream of rtgB in RtgAB Ϫ (native R6 rtgA pseudogene) and RtgAB ϩ (pseudogene-repaired) backgrounds. These strains were also ComAB Ϫ and BlpAB Ϫ in order to remove the possibility of peptide secretion through these other PCATs. Upon RtgS-C10 treatment in CDMϩ, levels of extracellular RtgC-HiBiT and RtgG-HiBiT in the RtgAB ϩ cultures increased 26-and 376-fold, respectively, compared to their levels in the RtgAB Ϫ cultures (Fig. 3). From these data, we conclude that RtgAB secretes the rtg GG peptides.  The rtg locus exhibits extensive variation across different pneumococcal strains. To catalog the diversity found at rtg, we conducted a survey of the locus in a collection of pneumococcal clinical isolates from Massachusetts, USA (39). After removing genomes in which rtg spans multiple contigs or when conserved genes flanking rtg could not be found, we were left with 318 of 616 strains, all of which encoded at least one rtg gene. We analyzed the rtg loci from these 318 strains and clustered them into 23 groups based on overall architecture (see Fig. S3A). Across all 318 loci, we found 24 unique rtg genes (Table 2), including eight putative GG peptides which share highly conserved signal sequences (Fig. S3B). Searching for these genes in the full collection revealed that 615 of 616 strains had some version of the rtg locus. Next, we analyzed the variation in RtgS among the 318 filtered strains. We found at least one copy of rtgS in each strain. Because duplication of rtgS is common, we assigned the name rtgS1 to any copy located next to rtgR and the name rtgS2 to any copy located next to rtgR=. Based on the identity of the penultimate residue in the translated peptide, which we have shown is important for signaling activity, we catalogued a total of three pheromone types: types A (Pro), B (Ser), and C (Leu). Only two other positions in the last 12 residues of RtgS showed variation: Ala/Val at position Ϫ10 from the C terminus and Ile/Val at position Ϫ8. The functional significance of these other polymorphisms is unknown. Finally, we analyzed each strain's RtgAB status and found 5% of strains carry unambiguously intact rtgA and rtgB. Another 12% carry intact rtgB and a version of rtgA with a start codon mutation (ATGϾATT) but that is otherwise intact. We determined Determinants of Pneumococcal PCAT Substrate Selection ® that a strain with the ATGϾATT mutation still produces functional RtgAB, likely by using an alternative start site, and suffers only a minor reduction in secretion capacity compared to that of a strain with fully intact rtgA (see Fig. S4). Therefore, 17% of strains encode a functional RtgAB. Active RtgR/S confers a competitive fitness advantage during nasopharyngeal colonization. To determine the biological role of rtg, we tested the effect of a regulatory deletion on colonization of the nasopharynx, the natural niche of pneumococcus. Despite similar levels of colonization between the wild-type and ΔrtgR ΔrtgS1 strains in singly inoculated mice at 3 days postinoculation (Fig. 4A), the wild-type strain outcompeted the mutant in coinoculated mice (Fig. 4B). These data suggest that RtgR/S is active during nasopharyngeal colonization and show that active RtgR/S provides a fitness advantage over RtgR/S-inactive strains during cocolonization.
RtgAB and ComAB/BlpAB preferentially secrete different sets of peptides. The pneumococcal PCATs ComAB and BlpAB secrete the same diverse set of GG peptides (23)(24)(25). Therefore, we wondered if ComAB and BlpAB could also secrete the rtg GG peptides and if RtgAB is similarly promiscuous and could secrete ComAB/BlpAB substrates. We repeated the RtgC-HiBiT and RtgG-HiBiT secretion assays with ComAB ϩ and BlpAB ϩ strains, using treatment with the com and blp pheromones competencestimulating peptide (CSP) and BlpC, respectively, to induce their expression. ComAB and BlpAB secrete markedly reduced amounts of RtgC-HiBiT and RtgG-HiBiT compared to the amounts secreted by RtgAB ( Fig. 5A and B). To determine if RtgAB could secrete a ComAB/BlpAB substrate, we assayed secretion of a HiBiT-tagged version of the BlpI bacteriocin driven by its native promoter. RtgAB secretes roughly 10-fold less BlpI-HiBiT than BlpAB (Fig. 5C)  differences in com or blp activation compared to that of RtgAB Ϫ strains during growth in CDMϩ (see Fig. S5). Under these conditions, rtg is expected to turn on before com or blp in every strain background. Thus, even when RtgAB is highly expressed, it secretes too little CSP and BlpC to affect the timing of com and blp activation.
RtgAB and ComAB/BlpAB recognize their substrates through different signal sequence motifs. Given that we had found RtgAB and ComAB/BlpAB do not share the same substrate pool, we explored how the transporters discriminate between substrate and nonsubstrate GG peptides. We showed that the BlpI signal sequence (SS BlpI ) prevents secretion of the RtgG cargo peptide through RtgAB (Fig. 6A). However, it did not promote secretion of RtgG through ComAB/BlpAB, suggesting an incompatibility between the cargo peptide and these two transporters. On the other hand, the RtgG signal sequence (SS RtgG ) promotes secretion of the BlpI cargo peptide through RtgAB while preventing its secretion through ComAB and BlpAB (Fig. 6B).
To rule out the possibility of differences in peptide expression being solely responsible for the secretion differences, we also measured the amount of intracellular peptide in each assay ( Fig. 6A and B; right-hand graphs). The signal sequence swaps affected intracellular peptide levels. However, these intracellular differences cannot account for the observed changes in secretion; higher intracellular levels did not correlate with more secretion, and while intracellular levels of the same peptide were relatively consistent across different strains (RtgAB ϩ versus ComAB ϩ versus BlpAB ϩ ), secretion was not. Thus, the observed changes in secretion between the different peptides most likely reflect differences in peptide-transporter interactions. Determinants of Pneumococcal PCAT Substrate Selection ® In conclusion, while cargo peptide can dictate transporter compatibility in some cases, the signal sequences of GG peptides still contain all the necessary information to direct secretion of their own cargo peptides through the proper transporters. For all future assays, we used BlpI as the cargo peptide, since it can be secreted by all three transporters given the correct signal sequence.
Next, we searched for the specific signal sequence residues involved in transport selectivity. We found that secretion of peptide through ComAB/BlpAB depends on the identities of the residues at the conserved signal sequence positions Ϫ15, Ϫ12, Ϫ7, and Ϫ4. These positions were previously implicated in substrate recognition by PCATs (27,28,30). The combination of the four residues at these positions from SS BlpI (F/M/L/V) introduced into SS RtgG promote secretion through ComAB/BlpAB, although they were not strictly required for secretion in the context of SS BlpI (Fig. 7A). The complementary association did not hold for RtgAB-mediated secretion, in that the four residues from SS RtgG (Y/L/M/L) were neither necessary nor sufficient for secretion through RtgAB (Fig. 7A). Additionally, alanine substitutions at all four positions in SS RtgG only partially impeded secretion through RtgAB, but the same substitutions in SS BlpI prevented secretion through ComAB and BlpAB almost entirely (Fig. 7B).
A specific motif at the N-terminal ends of rtg GG peptide signal sequences promotes secretion through RtgAB. To identify the signal sequence residues that promote secretion through RtgAB, we turned our attention to the N-terminal ends of the signal sequences, which are conserved in rtg GG peptides but not in ComAB/BlpAB substrates. Residue swaps at positions Ϫ22 to Ϫ18 in SS RtgG and SS BlpI demonstrated that secretion through RtgAB, but not ComAB or BlpAB, depends on specific signal sequence residues in this region (Fig. 8A). The P(Ϫ18)M substitution in SS RtgG modestly decreased secretion through RtgAB, and removal of all residues on the N-terminal side of this substitution further decreased secretion (Fig. 8B). Meanwhile, removal of the residues at the same positions from SS BlpI did not change secretion through RtgAB (Fig. 8C). These data indicate that the residues in this region in SS RtgG were selected to interact with RtgAB rather than to avoid steric clash. Alanine scanning mutagenesis of the Ϫ22 to Ϫ19 region of SS RtgG revealed that secretion through RtgAB was not sensitive to mutation at any single site (see Fig. S6). These data can be explained by multiple redundant residues mediating the interactions in this region or the interactions being tolerant to alanine substitution. We conclude that RtgAB recognizes rtg GG peptides through interactions involving the signal sequence residues in the Ϫ22 to Ϫ18 region. At the same time, RtgAB's substrate recognition mechanism has evolved to be less reliant than that of ComAB or BlpAB on interactions with the hydrophobic signal sequence residues at positions Ϫ15, Ϫ12, Ϫ7, and Ϫ4.

DISCUSSION
In this work, we have characterized the PCAT-encoding locus rtg and shown it is regulated by the RtgR/S system. RtgR/RtgS belongs to a family of regulatory systems found in streptococci that includes the Rgg/SHP and ComR/S systems (12). Rgg/SHP and ComR/S circuits can act as either cell density-dependent quorum-sensing systems (12) or timing devices (40). Our data suggest RtgR/S behaves like the former (see Fig. S7 in the supplemental material). A purely intracellular signaling pathway has been reported for XIP in Streptococcus mutans (14,41). Such a pathway is unlikely to exist for RtgR/S, since rtg autoinduction requires both PptAB and Ami (Fig. S1A). While the RtgS pheromone is similar to the previously described SHP and ComS/XIP pheromones, it also differs from these other pheromone classes in important ways. RtgS lacks the conserved aspartate or glutamate residue characteristic of SHPs and is divergently transcribed from its regulator unlike ComS (12). However, RtgS does contain a Trp-Gly-Trp motif near the C terminus which bears resemblance to the Trp-Trp motif found in some XIPs (12,20). RtgR is phylogenetically closer to the ComRs than SHP-associated Rgg regulators but does not cluster with either group (12). Using a published list of Rgg regulators (12), we found two RtgR-like regulators associated with Trp-X-Trp (WxW)  Determinants of Pneumococcal PCAT Substrate Selection ® motif-containing pheromones: SPD_1518 (Rgg1518) from S. pneumoniae D39 and SSA_2251 from Streptococcus sanguinis SK36 (predicted unprocessed pheromone sequences, MGFKKYLKNLPKNSGFLIWSWIQLIWFETWFWG and MKKIVYNLILLAVTSIVTTSVFP WWWLWW, respectively). Expression analysis of the pheromone operon associated with rgg1518 using PneumoExpress (42) revealed that the pheromone and genes SPD_1513 to SPD_1517 are specifically upregulated under the same conditions that result in upregulation of rtg. Therefore, the Rgg1518 system is likely functional. We propose that RtgR/S and other Rgg/WxW pheromone pairs constitute a distinct group of Rgg regulatory systems. We leave the work of characterizing the members of this group and the pathways they regulate to future studies. We showed that in the RtgAB ϩ strain Sp9-BS68, the ability to activate the RtgR/S system confers a fitness advantage during competitive colonization of the nasopharynx. While 78% of strains are predicted to harbor a functional RtgR and therefore can respond to pheromone, only 17% of strains are RtgAB ϩ . Most RtgAB Ϫ strains still harbor at least one rtg GG peptide but have no obvious means with which to secrete them, since they are not secreted by the other two PCATs commonly found in pneumococcus, ComAB and BlpAB. We have been unable to determine the function of the rtg GG peptides, but we speculate that they are bacteriocins. The reasons for this are that bacteriocin secretion is the most common function of PCATs and that five of the seven rtg GG peptide genes are always associated with downstream genes encoding hypothetical proteins that resemble bacteriocin immunity proteins (43). The fact that the Sp9-BS68 strain with a functional rtg locus demonstrated a competitive advantage over the ΔrtgR ΔrtgS1 strain during dual infection is consistent with the bacteriocin hypothesis, as the regulator mutant would be unable to upregulate immunity, although we cannot exclude that other rtgR-regulated factors play a role in this fitness advantage. Regardless of the specific function of the rtg GG peptides, the fact that most RtgAB Ϫ strains are still RtgR ϩ suggests that rtg retains a useful function that does not require secretion of these peptides. Further studies will be needed to determine the mechanism responsible for the RtgR/S-dependent competitive fitness advantage seen in colonization studies, the function of the rtg GG peptides, and the biological significance of active rtg loci with nonfunctional RtgAB.

M N T K M M S Q A S V A D N E E A E I A S G G
The case of RtgAB and ComAB/BlpAB allowed us to study how two sets of PCATs which coexist in the same strain preferentially secrete different sets of peptides through slight differences in substrate recognition. Unlike ComAB and BlpAB, RtgAB recognizes its substrates partially using a motif located at the N-terminal ends of their signal sequences. This motif is located 18 residues away from the signal sequence cleavage site and is exclusively found in rtg GG peptides. Where data are available, previous studies of PCAT substrates have found that positions at the N terminus located farther than 18 residues from the cleavage site are either dispensable for recognition by PCATs (28,30) or can be missing entirely (33,44,45). As far as we are aware, RtgAB is unique among PCATs in recognizing a signal sequence motif located so distantly from the cleavage site. Future efforts will be directed toward identifying the specific nature of the interaction between the N-terminal motif and RtgAB and the exact signal sequence residues involved.
The insights into the sequence determinants of PCAT substrate selectivity gained here illuminate a relatively understudied aspect of this class of transporters. They will also be useful in guiding future efforts to predict substrates for ComAB, BlpAB, RtgAB, and other PCATs. Some GG peptides are found without a closely associated or coregulated PCAT (18). In these cases, it would be helpful to have sequence-based approaches for assigning potential transporters to these "orphan" GG peptides. Moreover, for strains that harbor multiple PCATs, predicting if GG peptides can be secreted by PCATs that are not necessarily closely associated can guide mechanistic studies that lead to new insights into function and regulation, such as with ComAB and BlpAB substrates in pneumococcus. Our work lays the groundwork for identifying signal sequence motifs of GG peptides that are important for transporter selectivity. The next step will be to study the corresponding sequence and structural motifs in PCATs that contribute to this selectivity. In addition to bacteriocins and quorum sensing (3), GG peptides have now been linked to biofilm formation, colonization of host niches, and dissemination during infection (18,46). Ultimately, the ability to predict and rationalize PCAT-GG peptide pairings will advance our understanding of a broad range of biologically significant microbial processes.

MATERIALS AND METHODS
Strains and growth conditions. All strains were derived from Sp9-BS68 (36), D39, or the R6 strain P654 (referred to as PSD100 in reference 47) (see Table S1 and methods in Text S1 in the supplemental material for details). The modified R6 strain was used for some in vitro assays because previous work demonstrating the blp-com connection was performed in this strain background. Pneumococcus was grown in either filter-sterilized THY (Todd Hewitt broth plus 0.5% yeast extract) or CDMϩ (see methods in Text S1) (38) at 37°C. All media contained 5 g/ml catalase. All CDMϩ was supplemented with 0.5% (vol/vol) THY. Except where noted otherwise, pneumococcal cultures used for experiments were inoculated to an OD 620 of 0.0015 from starter cultures grown in THY (pH 7.4) to an OD 620 of 0.275 and frozen at Ϫ80°C in 13% glycerol. Starter cultures were pelleted at 6,000 ϫ g for 5 min at room temperature and resuspended in the appropriate growth medium for the experiment before being used for inoculation. Antibiotics were used at the following concentrations: chloramphenicol, 2 g/ml; gentamicin, 200 g/ml; kanamycin, 500 g/ml; spectinomycin, 200 g/ml; streptomycin, 100 g/ml.
Transformations. Transformation protocols were adapted from those described in reference 48. See methods in Text S1 and Table S1 for details and primers used for constructing transforming DNA products. Unmarked chromosomal mutations were created via Janus (49), Sweet Janus (50), or Janus2 (Text S1) exchange. Transformants were verified by Sanger sequencing.
Luciferase reporter time course assays. For com-blp activation assays only, starter cultures were grown in THY (pH 6.8) to an OD 620 of 0.075 to prevent com-blp activation. Cells were grown in THY or CDMϩ in a white, clear-bottom 96-well plate (655098; Greiner Bio-One), 200 l per well. For assays using For assays using NanoLuc luciferase, the following concentrations of Nano-Glo substrate (N1121; Promega) were added to the media: 1:5,000 (CDMϩ), 1:10,000 (THY). The plate was incubated in a Synergy HTX plate reader set to read absorbance at 620 nm and luminescence every 5 min. For single reporter assays, no filter was used for luminescence readings. For dual reporter assays, 450/50 band-pass and 610 long-pass filters were used to isolate NanoLuc and red firefly luciferase signals, respectively. For D39 strains only, the plate was shaken before readings were taken. Promoter activities were calculated from luminescence and absorbance readings as described in reference 25. For locus activation assays, timings of activation events were calculated as described in the methods in Text S1 and compared using survival analysis. Differences between groups were assessed by log-rank tests using the FHtest package (v1.4) in R, and when appropriate, the Holm correction was applied for multiple comparisons. RtgS dose-response assays. Cells expressing P rtgS1 -luc reporters were grown in THY or CDMϩ containing 330 M firefly luciferin. At an OD 620 of 0.02, cultures were aliquoted into a white, clear-bottom 96-well plate (655098; Greiner Bio-One), 100 l per well. Each well of the plate was prefilled with 100 l sterile medium containing 0.5% (vol/vol) dimethyl sulfoxide (DMSO), 330 M firefly luciferin, and appropriate concentrations of synthetic RtgS peptide (Genscript). The plate was then incubated in a Synergy HTX plate reader set to read absorbance at 620 nm and luminescence every 5 min. For D39 strains only, the plate was shaken before readings were taken. P rtgS1 activity was calculated, and the response was defined as the maximum observed P rtgS1 activity within 60 min (Sp9-BS68) or 120 min (D39) of treatment. When applicable, curves were fit to a Hill model using the nls() function in R 3.5.1.
Peptide secretion assays. Cells were inoculated from starter cultures to an OD 620 of 0.005 and grown in CDMϩ. At an OD 620 of 0.05, cells were treated with 200 ng/ml CSP1, 200 ng/ml BlpC R6 , and 20 nM RtgS A -C10. Samples were taken for HiBiT quantification at appropriate time points. For native BlpI-HiBiT assays only, clarified supernatants were obtained after centrifugation at 6,000 ϫ g for 5 min at 4°C. For all other assays, cells were retained in the samples. HiBiT signal was quantified by mixing samples with HiBiT extracellular detection reagent (N2421; Promega) at a 1:1 ratio and reading luminescence with a Synergy HTX plate reader. Samples were also taken for quantification of intracellular peptide; for endpoint assays, they were taken concurrently with the extracellular samples, and for time course assays, they were taken at the last time point. Extracellular peptide was removed from these samples by proteinase K digestion, and then the cells were lysed and HiBiT signal was quantified as described above. Standards consisting of synthetic L10-HiBiT peptide (25) mixed with samples of a non-HiBiT-expressing strain were used to generate standard curves to use for calculating HiBiT-tagged peptide concentrations in experimental samples. See methods in Text S1 for more details. Differences between groups were assessed by analysis of variance (ANOVA) using the emmeans package (v1.2.3) in R.
Genomic analysis of rtg. Analysis of rtg was performed using the assembled genomes of the Massachusetts isolate collection (BioProject accession PRJEB2632). See methods in Text S1 for details.
Mouse colonization assays. Mouse colonization was performed as described in reference 25. Briefly, dual or single-strain mixtures of Sp9-BS68 were inoculated into the nasopharynx of unanaesthetized 5to 7-week-old female BALB/c mice (Taconic)with 1.0 ϫ 10 6 to 3.0 ϫ 10 6 CFU/mouse in 10 l of sterile phosphate-buffered saline (PBS). For dual inoculated mice, the ratio of the kanamycin-resistant strain to the spectinomycin-resistant strain was between 0.25 and 0.6. Mice were euthanized with CO 2 overdose after 72 h, and nasopharyngeal colonization was sampled by nasal wash. See methods in Text S1 for IACUC approval and details on how colonization density and competitive indices were calculated. Differences in colonization densities and competitive indices between groups were evaluated by the Mann-Whitney (2 groups) and Kruskal-Wallis (Ͼ2 groups) tests using the wilcox.test() and kruskal.test() functions in R 3.5.1.
Data availability. Sequences of Janusϩ and Janus2 constructs were deposited in GenBank under accession numbers MN848328 and MN848329, respectively. The rtg locus from Sp9-BS68 including the new sequencing that allowed us to connect existing contigs and rtg gene designations established here was deposited as accession number MN848330.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. TEXT S1, PDF file, 0.3 MB.

ACKNOWLEDGMENTS
This work was supported by funding from the National Institutes of Health T32 AI007528 (to C.Y.W.), R01 AI101285 (to S.D.), and R56 AI101285 (to S.D.).