A conserved rhizobial peptidase that interacts with host-derived symbiotic peptides

In the Medicago truncatula-Sinorhizobium meliloti symbiosis, chemical signaling initiates rhizobial infection of root nodule tissue, where a large portion of the bacteria are endocytosed into root nodule cells to function in nitrogen-fixing organelles. These intracellular bacteria are subjected to an arsenal of plant-derived nodule-specific cysteine-rich (NCR) peptides, which induce the physiological changes that accompany nitrogen fixation. NCR peptides drive these intracellular bacteria toward terminal differentiation. The bacterial peptidase HrrP was previously shown to degrade host-derived NCR peptides and give the bacterial symbionts greater fitness at the expense of host fitness. The hrrP gene is found in roughly 10% of Sinorhizobium isolates, as it is carried on an accessory plasmid. The objective of the present study is to identify peptidase genes in the core genome of S. meliloti that modulate symbiotic outcome in a manner similar to the accessory hrrP gene. In an overexpression screen of annotated peptidase genes, we identified one such symbiosis-associated peptidase (sap) gene, sapA (SMc00451). When overexpressed, sapA leads to a significant decrease in plant fitness. Its promoter is active in root nodules, with only weak expression evident under free-living conditions. The SapA enzyme can degrade a broad range of NCR peptides in vitro.

The symbiosis between legumes and rhizobia is initiated by the action of a variety of plant and bacterially-derived molecules including flavonoids 1-3 , lipochitooligosaccharide nodulation (Nod) factors 4,5 , and exopolysaccharides (EPS) 6,7 . Flavonoids secreted from plant roots stimulate the production of rhizobial Nod factor, which initiates the symbiotic developmental process by causing root hairs to curl around the bacteria [8][9][10] . In the Sinorhizobium meliloti -Medicago truncatula symbiosis, production of the bacterial EPS succinoglycan enables the development of infection threads within root hairs, allowing the rhizobia to colonize plant tissue 11,12 . A portion of bacteria within the infection threads are endocytosed into specialized nodule cells 13,14 , forming organelle-like assemblies called symbiosomes where the intracellular rhizobia (termed bacteroids) fix atmospheric nitrogen for the plant. While interkingdom signaling molecules play critical roles during bacterial entry and early nodule development, it has also become clear that symbiotic communication is occurring around the time that nitrogen fixation commences. This is evidenced by studies in which random pairings of symbiotically competent Medicago hosts and Sinorhizobium strains often give rise to infected nodules that do not fix nitrogen [15][16][17][18] . Additionally, several Fix − M. truncatula mutants have been isolated that allow bacterial entry into nodule cells, but nitrogen fixation is abolished 19,20 .
The first evidence of a new class of late-stage symbiotic signals emerged from a transcriptomic analysis in M. truncatula, where a large assortment of hundreds of host-derived nodule-specific cysteine-rich (NCR) peptides was observed 21 . Subsequent studies of NCR peptides revealed that they possess structural and bactericidal properties similar to the defensin class of antimicrobial peptides 22,23 . At non-lethal doses, NCR peptides have several effects on rhizobial cells: they permeabilize membranes, promote genome endoreduplication, and drive cell-enlargement and branching, all of which presumably facilitate effective nitrogen fixation, nutrient exchange, and suppression of rhizobial proliferation [24][25][26][27][28][29] . Genetic defects in a Medicago symbiosis-specific protein secretion pathway have been shown to block NCR peptide delivery to symbiosomes and prevent nitrogen fixation 26,30 . More recently, genetic disruption of specific NCR peptide-encoding genes has been linked to failure to fix nitrogen 31,32 . Taken together, legume-derived NCR peptides clearly play major roles in the later stages of symbiotic development, leading to the terminal differentiation of rhizobia within nodule cells and driving the nitrogen fixation and nutrient exchange that is characteristic of this symbiotic interaction.
The influence of host-derived NCR peptides on nodule-bound bacteria has given rise to a model in which corresponding bacterial peptidases may modulate this influence. It was previously demonstrated that a rhizobial metallopeptidase, HrrP, strongly suppresses nitrogen fixation in specific Medicago-Sinorhizobium combinations 33 . It was further shown that HrrP degrades NCR peptides in vitro, pointing to a direct host-strain interaction at the level of host NCR peptides and a microbial peptidase 33 . The hrrP gene, however, is found on a relatively rare accessory plasmid present in about 10% of Sinorhizobium isolates 34 . We have more recently turned attention to the question of whether NCR peptides are influenced, perhaps in more subtle ways, by peptidases that are more generally conserved in the core genome of S. meliloti. Considering that an ensemble of cooperating peptidases may make identification difficult by loss-of-function genetic analysis, we employed a plasmid-based overexpression screen focused on gene candidates that most likely encode peptide-hydrolyzing enzymes.

Results
Identification of core genome-encoded peptidase candidates. Using the online S. meliloti 1021 genome and MEROPS peptidase databases, we identified 131 putative peptidase candidates that could be screened for effects on symbiosis. An OrthoMCL analysis of the 23 fully sequenced S. meliloti strains in the NCBI database showed there are 70 orthologs from our set of 131 that are present in all strains 35 (see Supplementary Table S1). The majority of the 131 putative peptidases (68%) are encoded on the main chromosome with the remaining 32% divided evenly between megaplasmids pSymA and pSymB. A similar distribution of peptidase genes is seen on the three primary replicons of the other fully sequenced strains of S. meliloti (data not shown). The 4 catalytic mechanisms for proteolysis represented in this set are metallopeptidases (37%), serine (28%), cysteine (7%), and aspartic (5%) proteases, with 23% having unknown or unannotated mechanisms. To predict the subcellular localization of each peptidase we used PSORTb 3.0 36 and it was determined that nearly half (49%) are predicted to localize to the cytoplasm. Another 16% are predicted to localize to the cytoplasmic membrane, 5% to the periplasm, 1% to the outer membrane, with 3% secreted extracellularly, while 26% have no predicted localization.
Effects of peptidase overexpression on plant fitness. Genes corresponding to 28 candidate bacterial peptidases were amplified from the S. meliloti 1021 genome and cloned in the overexpression vector pPG013 ( Supplementary Fig. S1), which is replicated at ~ 7 copies/cell. Expression of a candidate peptidase was driven by the hrrP promoter (PhrrP) rather than its native promoter because PhrrP has been shown to be highly active in nodules 33 . After cloning, individual overexpression plasmids were mated into S. meliloti C307 for inoculating M. truncatula A20 plants. This particular strain-host pair was chosen because of the previous observation that expression of hrrP in S. meliloti C307 leads to a significant fitness reduction of A20 plants but not plants from the A17 accession 33 . Thus, the S. meliloti C307 and M. truncatula A20 strain-host combination functions as a control while screening through candidate peptidases.
To test a given peptidase overexpression strain, 12 M. truncatula A20 plants were inoculated and grown for 28 days before measuring plant fitness. Empty-vector, hrrP-overexpression, and no-inoculum controls were included in each experiment. The 28 candidate peptidases screened in this study (Table 1) represent a variety of proteolytic mechanisms and cellular localizations. Almost all peptidases tested did not have a reproducibly noticeable impact on plant fitness compared to controls; however, overexpression of SMc00451 (referred to hereafter as sapA) led to a considerable decrease in shoot dry weight and nodule size in multiple independent experiments (Fig. 1).
Homologs of sapA are found in all 23 fully sequenced S. meliloti genomes in the NCBI database. It encodes an M16 family zinc metallopeptidase and is, based on an NCBI protein BLAST search, the closest homolog encoded in the core S. meliloti genome to HrrP. SapA is predicted to form a homodimeric complex that takes on a clamshell structure, characteristic of the M16B subfamily, whereas HrrP is a monomeric fusion of two domains connected by a linker, characteristic of the M16A and M16C subfamilies 37 . SapA has similarity to both halves of HrrP: a pairwise amino acid sequence alignment of SapA with the N-terminal half of HrrP shows 30% identity with 47% coverage, and alignment with the C-terminal half shows 23% identity and 62% coverage. Despite the difference in polypeptide length, their relatedness can be seen in a phylogenetic tree containing four SapA and four HrrP homologs, with a more distantly related sinorhizobial M16 peptidase outgroup ( Supplementary Fig.  S2). Predicted protein structures of HrrP and SapA underscores their similarity (Fig. 2).

Degradation of NCR peptides by SapA in vitro.
The HrrP peptidase is known to degrade NCR peptides in vitro, with degradation occurring much more rapidly under reducing conditions where peptide disulfide linkages are prevented 38 . This is consistent with the predicted localization of HrrP in the bacterial cytoplasm (a reducing environment). In testing the ability of SapA to degrade NCR peptides, we began by addressing the question of whether the peptidase is likely to encounter peptides in the bacterial cytoplasm or extracytoplasmically. Using predictive tools to assess the subcellular localization of SapA, analysis was carried out with SignalP-5.0 39 , DeepSig 40 , and TMHMM 2.0 41 . This analysis indicated a lack of a secretion signal or transmembrane segment in the N-terminal region of SapA. Additionally, in the predicted structural model of SapA (Fig. 2), the N-terminus maps to a structurally conserved 6-stranded ß-sheet, strongly suggesting that it folds into the protein's globular structure and does not mediate trafficking to an extracytoplasmic location.
With SapA likely to encounter NCR peptides in the cytoplasm, we purified recombinant SapA and incubated it individually with several NCR peptides under reducing conditions. A portion of each reaction was stopped at 0-, 2-, and 4-h time points to observe degradation over time via gel electrophoresis (Fig. 3). Most peptides were at least partially degraded and 3 were fully degraded within 2 h. A variant form of SapA (E50A), which is presumably catalytically inactive, was purified and tested for activity against NCR peptides in the same manner. This mutant showed no evidence of peptide degradation. The NCR peptides used in this experiment span a range of amino acid lengths and isoelectric points, but all had four similarly spaced cysteine residues. While it is difficult www.nature.com/scientificreports/ to explain the peptide substrate specificity based on sequence or other chemical properties, it is notable that the same peptides degraded by SapA were also the most readily degraded by HrrP 33 . NCR peptide degradation is likely not a universal property of metallopeptidases, as evidenced by our observation that S. meliloti SMb20697, a putative M20 metallopeptidase, was purified and tested in the same conditions with no observable NCR peptide degradation (see Supplementary Fig. S4).
To test whether NCR peptide degradation by SapA is specified by peptide sequence, we incubated SapA with NCR035 (a readily degraded substrate) and three NCR035 variants with scrambled sequences. These sequences were designed to maintain overall amino acid composition. The results at 0-and 2-h time points indicate that SapA exhibits selectivity for NCR035 over the randomized sequence variants (see Supplementary Fig. S5).
Expression pattern of sapA. To test whether sapA is expressed in nodules, we fused the sapA promoter (PsapA) to a gus reporter gene and followed reporter gene activity in nodules and free-living cells. For assessment of nodule expression, M. truncatula A20 plants were infected with promoter-gus strains, nodules were harvested 7 days post-inoculation (dpi), and these were stained to visualize GUS activity (Fig. 4). PsapA, as well as both positive controls (Ptrp and PhrrP) displayed GUS activity in nodules, while no GUS activity was observed in the absence of a promoter. Similar expression patterns were seen in nodules harvested and stained at 10 and 14 dpi (not shown). To monitor PsapA activity in free-living conditions, we performed Miller assays on three biological replicates of each promoter-gus strain in both rich and minimal media. In contrast to nodules, PsapA displayed very low activity in either free-living condition, while positive controls (Ptrp and PhrrP) were both active in both media. These data indicate that sapA expression is relatively low under free-living conditions and at least moderately high in nodules.

Discussion
Prior to symbiotic nitrogen fixation, M. truncatula compels rhizobia to terminally differentiate, in part through the activity of secreted NCR peptides. It was previously shown that the plasmid-encoded peptidase, HrrP, enables endosymbionts to counteract this pressure, by selectively degrading a portion of these peptides 33 . The hrrP gene is only found in a small fraction of Sinorhizobium isolates 34 . The current study explores the possibility that HrrP involvement in modulating bacteroid differentiation reflects a broader phenomenon involving peptidases present in all or most strains of a rhizobial species. Peptidases encoded in the S. meliloti 1021 genome were

SMc02577
Probable heat shock protein Cytoplasmic +

SMc04091
Putative protease transmembrane protein; Peptidase M48 Cytoplasmic membrane +     37 . Based on data from a tissue-specific transcriptomic study, each of the 131 peptidases identified in this study could be detected in nodules, with varying patterns of expression across nodule zones 42 . More than half of these peptidases are expressed at their highest mRNA levels in the interzone (IZ) or zone III regions where bacteroid development and nitrogen fixation occur. Expression of sapA also falls in this category, and in agreement with the transcriptomic data, we show that sapA expression is readily detectable throughout the nodule. Similarly, the vast majority of NCR peptides, including the NCR peptides degraded by SapA in vitro, are expressed by the host plant at their highest levels where bacteroid development and nitrogen fixation occurs 42 . Two of the NCR  www.nature.com/scientificreports/ peptides degraded by SapA (NCR247 and NCR035) have been shown to significantly affect bacterial cell division, membrane permeability, and DNA replication in vitro 24,26,43 . That NCR peptides can have such profound effects on rhizobia, while being susceptible to degradation by conserved rhizobial peptidases, supports a model in which the precise cocktail of peptides and the precise level of peptidase activity are vital determinants of symbiotic compatibility. It should be considered that NCR peptides are exogenously introduced to developing rhizobia (through a host secretory apparatus), and the rhizobial peptidases that degrade them are likely to be localized in the bacterial cytoplasm. Thus, the peptides must enter bacterial cells in order to be destroyed. This may explain the seemingly counterintuitive observation that the conserved rhizobial peptide uptake transporter, BacA, imports NCR peptides and protects rhizobia from their antimicrobial effects 44,45 . Transporter-mediated uptake likely hastens the delivery of NCR peptides to cytoplasmic bacterial degradation machinery before they can interact with and cause damage to cell membranes. Consistent with this model, BacA expression is known to be upregulated while bacteroid maturation is underway in the nodule 46 . While it is possible for developing bacteroids to secrete peptidases to meet their NCR peptide substrates extracellularly, this strategy may be less effective due to the more structured state of NCR peptides in oxidizing environments. A study on the HrrP peptidase showed that NCR peptides are nearly impervious to degradation when in their disulfide-bonded (oxidized) state 38 . By delivering these peptides to the reducing environment of the cytoplasm, the disulfide-associated protection is likely lost, making this a more ideal location for degradation.
We present evidence, based on peptide motif-finding algorithms and structural modelling, that SapA (like HrrP) is a cytoplasmic enzyme and, as such, can potentially inactivate imported NCR peptides. From our analysis, it is clear that some peptides are more susceptible to degradation than others; though there is no obvious correlation amongst susceptible peptides based on general properties such as isoelectric point, length, or cysteine positioning. Comparing SapA-mediated degradation of NCR035 and several of its scrambled variants, we observe that the native sequence is more actively targeted than sequence variants that have the same overall amino acid composition. This suggests that rhizobial peptidases recognize substrate peptides with some degree of sequence specificity. Analysis of substrate specificity in an M16 peptidase isolated from Bacillus halodurans revealed a preference for aromatic residues upstream of the initial cleavage site 37,47 . A possible explanation for the differences in degradation of NCR035 and the scrambled variants, therefore, may be the varied positioning of aromatic residues in these peptides. More focused studies in which residues with specific properties are manipulated could shed more light on these kinds of questions.
The discovery of sapA came about through screening a small fraction of the total peptidases in S. meliloti strain 1021. Continued efforts to screen for sap genes will likely allow for the discovery of an ensemble of rhizobial peptidases that cooperate to modulate symbiotic transactions with the plant host. We likewise suspect that no single peptidase will prove to have a dominant influence on symbiosis based on loss-of-function phenotypes. Thus, continuing the analysis based on overexpression phenotypes will likely be most productive. Considering that sapA overexpression yields modest symbiotic defects, one can imagine using this as a sensitized background for carrying out further screening.

Materials and methods
Bacterial growth conditions. E. coli and S. meliloti strains are described in Supplementary Table S2 and were grown in LB (lysogeny broth) at 37 °C and 30 °C respectively and supplemented with the following antibiotics, as appropriate: streptomycin (Sm, 200 or 100 μg/ml), neomycin (Nm, 100 or 50 μg/ml), kanamycin (Km, 30 μg/ml or 15 μg/ml). Strains of E. coli used for purifying peptidases were grown at 30 °C. The defined medium used for growth of S. meliloti was composed of 0.5% sucrose, 50 mM NaCl, 10 mM KH 2 PO 4 (pH 7), 10 mM NH 4  Identification of putative core genome-encoded peptidases. The S. meliloti 1021 online genome database includes annotations from Interpro, PubMed, Swiss-Prot, and trEMBL as well as enzyme codes and functional classifications for most genes. Enzyme codes, and generic terms including "peptidase", "protease", "hydrolase" and others, as well as variations on those terms, were used to compile a large list of candidate peptidases. This list was refined and compared with all peptidases for S. meliloti 1021 found in the MEROPS database to obtain a final set of 131 peptidases. Any enzyme annotated as "acting on C-N (but not peptide) bonds" or classified with an Enzyme Commission number other than 3.4 were removed. Conserved hypothetical proteins with a generic hydrolase or hydrolase-like annotation were included in the set.
Screen for symbiotically relevant peptidases. Genes for candidate peptidases were cloned into the overexpression vector pPG013 (plasmids are listed in Supplementary Table S3) using primers listed in Supplementary Table S4. Clones were verified by Sanger sequencing and mated into S. meliloti C307 before inoculating plants. The pPG013 plasmid was constructed by assembling four fragments: the first fragment contained, pVS1 repA, staA, and oriV, elements; the second fragment contained, kanR and the p15A oriV; the third fragment contained, the RK2 mobilization element, oriT; finally, a multiple cloning site and the hrrP promoter were added. A map for this plasmid is provided in Supplementary Fig. S1  Phylogenetic comparison of SapA and HrrP homologs. Phylogenetic analysis was carried out using the Phylogeny.fr platform 48,49 and comprised the following steps: Sequences were aligned with ClustalW (v2.1) 50 . After alignment, ambiguous regions were removed with Gblocks (v0.91b) 51 using the following parameters: minimum length of a block after gap cleaning was 5, positions with a gap in less than 50% of the sequences were selected in the final alignment if they were within an appropriate block, all segments with contiguous nonconserved positions larger than 8 were rejected, and the minimum number of sequences for a flank position was 55%. The phylogenetic tree was reconstructed using the Bayesian inference method implemented in the MrBayes program (v3.2.6) 52 . The number of substitution types was fixed to 6. The Poisson model was used for amino acid substitution, while rate variation across sites was fixed to "invgamma". Four Markov Chain Monte Carlo (MCMC) chains were run for 100,000 generations, sampling every 10 generations, with the first 5000 sampled trees discarded as "burn-in". Finally, a 50% majority rule consensus tree was constructed. Graphical representation of the phylogenetic tree was rendered using TreeDyn (v198.3) 53  In vitro NCR peptide degradation assays. Peptides were synthesized commercially from GenScript and prepared at a 1 mg/ml concentration. Approximately 4 μg of purified SapA (or the E50A variant) was incubated with 4 μg of peptide in buffer composed of 10 mM DTT, 50 mM KH 2 PO 4 pH 7.2, and 1 μM ZnSO 4 • 7H 2 O. At 0-, 2-, and 4-h time points, a portion of the reaction was stopped by adding sample buffer and transferring to ice. Sample buffer contained 32% glycerol, 150 mM Tris pH 6.8, 20 mM EDTA, 40 mg/ml SDS, 15.5 mg/ml DTT, and 0.3% bromophenol blue. Following the addition of stop buffer, samples were moved to an 85 °C heat block for 3 min and then placed back on ice before visualization on a tricine gel. In the experiment where SapA and NCR035 (or scrambled variants of NCR035) were tested, less enzyme was used (approximately 2 μg of purified SapA and 4 μg of peptide), to make subtle differences in degradation easier to visualize.
Quantification of GUS activity. Promoter-gus fusion strains were made by cloning various inserts into the gus expression vector pPG178 (promoter-gus plasmids are listed in Supplementary Table S3) using primers listed in Supplementary Table S4. This plasmid has the same backbone as pPG013 but, unlike pPG013, various promoters were cloned upstream of the gus reporter gene. A map of the empty plasmid is available in Supplementary Fig. S1 followed by relevant insert sequences. Cultures used to measure GUS activity were grown in three biological replicates. To each reaction, 20 μl of cells in log phase (OD 600 0.6-1.1) were added to a reaction composed of 520 μl basal buffer (60 mM Na 2 HPO 4 , 40 mM NaH 2 PO 4 pH 7), 3 μl ß-mercaptoethanol, 3 μl 20% tween, 20 μl CHCl 3 , and 60 μl p-Nitrophenyl α-D-glucopyranoside (8 mg/ml). Reactions proceeded 15 min before stopping with 600 μl 1 M Na 2 CO 3 and taking OD 405 measurements. www.nature.com/scientificreports/ Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.