The Bacteroidetes Q-rule and glutaminyl cyclase activity increase the stability of extracytoplasmic proteins

Porphyromonas gingivalis W83 cells were grown anaerobically in enriched trypticase soy broth (eTSB, 30 g/L) supplemented with yeast extract (5 g/L), l-cysteine (0.5 g/L), menadione (2 mg/L), and hemin (5 mg/L). We supplemented eTSB agar plates with 5% defibrinated sheep blood. Escherichia coli strain DH5α and E. coli S17-1 were grown in lysogeny broth (LB) (20 g/L). All bacterial strains used in this study are listed in Table S1.

We created P. gingivalis mutants by homologous recombination (15). For each mutant, a separate plasmid suitable for genomic integration was prepared in-house or synthesized commercially. For some experiments, the PgQC gene was expressed from vectors based on pTIO2-tet, which originates from pTIO-1 (16), using the RagAB promoter. Details of vector construction are provided in Supplemental Methods S1. The plasmids and primers used in this study are listed in Tables S2 and S3, respectively. All vectors were sequenced to confirm their integrity and introduced into P. gingivalis by electroporation or conjugation with E. coli S17-1. Transformants and transconjugants were propagated under antibiotic selection, and integration was confirmed by sequencing across the targeted genomic region.

The Asp126 residue in PgQC for D126A mutagenesis to generate a P. gingivalis strain expressing an inactive version of the enzyme was selected based on PgQC alignment with a Zn-dependent peptidase from Bacteroides vulgatus ATCC 8482 whose crystal structure at 1.8 A resolution (3GUX) is known. The motif AHWD¹²⁶ containing His and Asp residues essential for peptidase activity is conserved in PgQC, and as expected, the recombinant enzyme PgQ^D126A expressed in E. coli had no activity on H-Gln-7-amino-4-methylcoumarin (H-Gln-AMC).

Bacteria grown to the early stationary phase were harvested by centrifugation and washed once in cold phosphate-buffered saline (PBS). The cells were sonicated in a handheld UP200Ht homogenizer (Hielscher, Germany), and the protein concentration was determined using Pierce BCA Protein Assay Kit (Thermo Fisher Scientific, USA). QC activity was measured as previously described (4) using the fluorogenic substrate H-Gln-AMC (5 mM) pre-incubated with 10 µL of recombinant bacterial pyroglutamate aminopeptidase (12.5 U/mL, Unizyme Laboratories, Denmark) in 40 mM Tris-HCl and 400 mM KCl (pH 8.0) for 10 min at 30°C. The increase in fluorescence (λ_ex = 380 nm, λ_em = 460 nm) was recorded for 40 min at 30°C in a Flex Station 3 Multimode Microplate Reader (Molecular Devices, USA). The activity was normalized to the protein concentration in each sample.

Stationary cell cultures were adjusted to OD₆₀₀ = 1.4 with fresh eTSB containing 2 mM 2,2′-dithiothreitol (DTT; Sigma-Aldrich, USA) and were incubated for 15 min at room temperature to prepare the whole culture (WC) fraction. The cells were then centrifuged (5,000 × g, 20 min, 4°C), and the supernatant was clarified further (16,000 × g, 10 min, 4°C) to prepare the medium (M) fraction. The pellet was washed and resuspended in ice-cold PBS supplemented with 0.1 mM N_α-tosyl-l-lysine chloromethyl ketone hydrochloride (TLCK; Sigma-Aldrich, USA) and cOmplete EDTA-free inhibitor mix (Roche, Switzerland). The cells were disrupted using a BT40/TS2/AA Constant System cell disruptor (Thermo Fisher Scientific, USA) at 30 kPa. The lysate was digested with 0.02 mg/mL DNase I (Roche) for 30 min to prepare the whole-cell extract (WCE) fraction. The lysate was centrifuged (150,000 × g, 1 h, 4°C) to separate the periplasmatic/cytoplasmatic fraction from the insoluble cell envelope including IM and OM, known as the cell membrane fraction. The cell envelope pellet was treated with 200 mM MgCl₂ supplemented with 10% Triton X-100 for 30 min on ice to solubilize the IM and centrifuged as above. The collected supernatant was the IM fraction. The insoluble material was resuspended by sonication in PBS as the OM fraction. The obtained fractions were mixed with 0.1 mM TLCK, 1 mM DTT, 1 mM ethylenediaminetetraacetic acid (EDTA), and cOmplete EDTA-free inhibitor mix.

Western blotting was performed on overnight whole bacteria cultures normalized according to OD₆₀₀ measurements or on isolated cellular factions. Proteins were resolved by SDS-PAGE, and blotting was carried out using a Mini Trans-Blot device according to the manufacturer’s instructions (Bio-Rad Laboratories, USA). The blots were probed with anti-QC (4), anti-Strep (Novus Biologicals, USA), anti-His₆ tag (GeneScript, USA), or in-house antibodies specific for PorQ, PorV, Kgp, RgpA, and RgpB. The secondary antibody was HRP-conjugated anti-rabbit or anti-mouse IgG, as appropriate (Sigma-Aldrich, USA). Signals were developed using Pierce ECL Western Blotting Substrate (Thermo Fisher Scientific, USA). The signals on photographed X-ray films (Agfa, Germany) were quantified by densitometry using Image Lab v6.1 (Bio-Rad Laboratories, USA).

Bacteria cell cultures at the early stationary phase were collected, OD₆₀₀ adjusted to 1, and gingipain amidolytic activity was determined in WC and cell-free medium (M) (after cells were removed by centrifugation) using N-benzoyl-DL-arginine p-nitroanilide (BApNA) and acetyl-Lys-pNA, respectively, for RgpA/B and Kgp. Briefly, samples were added to the assay buffer (200 mM Tris-HCl, 100 mM NaCl, 5 mM CaCl₂, 10 mM L-cysteine, pH 7.6) in microplates to make a total volume of 190 µL and pre-incubated for 5 min at 37°C. Then, 10 µL of a substrate (final concentration, 1 mM) was added, and its enzymatic hydrolysis was recorded as the increase of OD₄₀₅ per minute using Flex Station 3 Multimode Microplate Reader (Molecular Devices). Results are presented as mOD/min per 1 µL of a sample (mOD/min/µL).

Total RNA was isolated from bacteria in the late exponential phase (OD₆₀₀ = 1.0) using the Total RNA Mini kit (A&A Biotechnology, Poland). Residual genomic DNA was removed using DNase I (A&A Biotechnology), and the RNA was purified again using the kit. We used 800 ng of re-purified RNA for cDNA synthesis with the High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific). Quantitative RT-PCR was carried out in triplicate using SYBR Green JumpStart Taq ReadyMix (Sigma-Aldrich) and 333 nM of each primer (Table S3) on a CFX96 Touch Real-Time PCR Detection System (Bio-Rad Laboratories). Relative expression levels were calculated by normalizing each cycle threshold (Ct) value to the reference gene rpoB using the ΔΔCt method.

Membrane samples were analyzed by liquid chromatography-tandem mass spectrometry (LC-MS/MS) to find evidence of pGlu modifications. The samples were denatured, reduced, and alkylated in 20 mM Tris-HCl (pH 8) containing 8 M urea and 5 mM DTT, followed by the addition of 15 mM iodoacetamide. The reduced and alkylated samples were diluted and digested with trypsin at 37°C for 16 h. The tryptic peptides were micro-purified (17) before LC-MS/MS analysis on an eksigent nanoLC 415 system connected to a TripleTOF 6600 mass spectrometer (both from Sciex, Canada). The samples were trapped on a pre-column and then separated on a 15-cm analytical column, both packed with ReproSil-Pur C18-AQ 3 µm resin (Dr. Maisch, Germany). Peptides were eluted at a flow rate of 250 nL/min in a gradient of 5–35% solution B (0.1% formic acid in acetonitrile) lasting 30 min. Mass spectra were searched against P. gingivalis entries in UniProt or NCBI using an in-house Mascot search engine (Matrix Science, UK). Search parameters allowed semi-trypsin cleavage and carbamidomethylation as a fixed modification, with the peptide tolerance and MS/MS tolerance set to 10 ppm and 0.1 Da, respectively. Gln→pGlu (N-terminal Qln), Glu→pGlu (N-terminal Glu), and methionine oxidation were set as variable modifications.

One liter of P. gingivalis strains PG_0449^his and PG_0449^Q22N,his was grown up to the early stationary phase in eTSB media at 37°C and in anaerobic conditions. Bacteria were harvested by centrifugation at 10,000 × g for 25 min, washed two times in PBS, and resuspended in PBS supplemented with Roche Proteases Inhibitors Cocktail and TLCK (10 µM). Bacteria were then lysed by sonication (3 cycles of 10  ×  5  s pulses at 17 W), and the obtained lysate was clarified by ultracentrifugation at 150,000 × g for 1  h. Soluble proteins were then dialyzed two times against 4 L of Ni-Sepharose binding buffer (20  mM sodium phosphate buffer pH 7.4, 500  mM NaCl, 20  mM imidazole, and 0.02% NaN₃). Proteins of interest were purified by affinity chromatography on Ni-Sepharose High Performance matrix (GE Healthcare, Pittsburgh, PA, USA) and eluted in the same buffer supplemented with 500 mM imidazole. The protein concentration of the purified samples was determined by BCA Assay (Sigma-Aldrich), and its purity was determined by SDS-PAGE and Coomassie Blue staining.

For N-terminal sequencing, proteins separated by SDS-PAGE were electrotransferred onto a polyvinylidene difluoride membrane. Protein bands visualized by staining with Coomassie Brilliant Blue R-250 were excised and subjected to automated Edman degradation using a Procise 494 HT amino acid sequenator (Applied Biosystems, Carlsbad, CA, USA).

Data were analyzed using Student’s t-test or one-way analysis of variance (ANOVA) with Bonferroni’s correction (comparison with control group) or Tukey’s correction (comparison across all groups) in GraphPad Prism v8 (GraphPad Software, USA). Statistical significance was assumed at P < 0.05.

In agreement with transposon mutagenesis studies deeming PgQC essential (8), we were unable to create a viable PgQC deletion mutant (4). However, close inspection of the genomic locus revealed a complex arrangement of three open reading frames (PG_2157, PG_2158, and PG_2159), indicating that the PgQC gene (PG_2157) is the first of a three-gene operon (Fig. 1A). The accompanying genes are not well characterized but could be sensitive to polar effects when PgQC is mutated. PG_2158 encodes a protein with an Fe-S center whose function is unknown, whereas PG_2159 (hemG) encodes a protoporphyrin oxygenase involved in iron acquisition (18). To determine whether these genes are essential, we constructed two additional deletion mutants (ΔPG_2158 and ΔPG_2159) and the control strains QC+ and PG_2159+ (Fig. 1A). In all these strains, the colony morphology and growth rate were indistinguishable from the wild-type strain, and gingipain activity was normal, with some increase for soluble Kgp activity (Fig. 1B and C; Fig. S1A and B). These results show that PG_2158 and PG_2159 are not essential for P. gingivalis at least in vitro, supporting our hypothesis that the pyroglutamyl formation of secreted proteins is essential for P. gingivalis growth.

To provide more evidence that PgQC is essential, we used a system in which gene significance can be inferred from the stability of a plasmid carrying a candidate essential gene in the absence of antibiotic selection pressure. For this purpose, we engineered a QC^m strain (m = merodiploid, reflecting the presence of two copies of the PgQC gene, one in the genome and the other on the plasmid). Next, we replaced the genomic PgQC coding sequence with the ermF cassette, resulting in a QC^p strain (p = plasmid, now the only copy of PgQC). Initial characterization of the QC^m and QC^p strains showed that the production of PgQC protein from the plasmid was enhanced compared to the control containing the empty plasmid (pTIO2-tet) (Fig. 1D). We did also observe some increase in gingipain activity (Fig. S1D and E), and as expected, there was no impairment in pigmentation phenotype (Fig. 1E), which is dependent on gingipain activity. The QC^p strain grew more slowly, and the generation time (g) was longer than the other strains (wild type, g = 2.42 h; pTIO2-tet, g = 3.19 h; QC^m, g = 2.76 h; and QC^p, g = 4.83 h). The stationary phase biomass of the QC^p strain was ~40% lower than that of the QC^m strain (Fig. 1F).

In the plasmid stability experiment, the QC^m and QC^p strains were maintained for eight generations in the absence of antibiotics. The number of cells carrying the expTIO_QC_tet plasmid was determined on days 0 and 8 by comparing the number of colonies on tetracycline vs control plates without antibiotics (Fig. 1G). We found that the QC^m strain was prone to curing, resulting in a lower number of antibiotic-resistant colonies on day 8 compared to day 0. In contrast, the QC^p strain maintained the plasmid even in the absence of selection, and the number of antibiotic-resistant colonies was similar on day 8 compared to day 0 (Fig. 1H). Taken together, these data confirm that PgQC expression is indispensable for P. gingivalis viability.

PgQC may be essential to P. gingivalis because its enzymatic activity is indispensable, or it may fulfill a non-catalytic role such as scaffolding or interacting with other proteins. Indeed, the spatial and functional coexistence of PgQC and the Sec translocon in the IM suggests that protein–protein interactions may contribute to the stabilization and/or function of Sec components (4). We therefore attempted to generate a P. gingivalis strain expressing catalytically inactive PgQC (mutation D126A). However, these attempts were unsuccessful, suggesting that PgQC enzymatic activity is necessary for P. gingivalis growth. This does not exclude the possibility that PgQC protein–protein interactions are also important, so we tested a possible dominant negative effect with an overexpression of the inactive enzyme (QC^D126A) in wild-type P. gingivalis to see if it would inhibit growth. We introduced the plasmid overexpressing PgQC^D126A (expTIO-QC-tet-D126A) and found that the resulting QC^mD126A (mD126A = merodiploid overexpressing the inactive enzyme) mutant strain was fully viable. Like the QC^m strain, also, the QC^mD126 strain produced much larger amounts of PgQC protein than the control strain with the empty plasmid (pTIO2-tet) (Fig. 2A), but, nevertheless, its enzymatic activity was similar to that of the control (Fig. 2B). Notably, the QC^mD126A mutant strain showed no difference in pigmentation phenotype (Fig. 2C), growth kinetics (Fig. 2D), or gingipain enzymatic activity (Fig. 2E and F) compared to the control strain. Moreover, proteomic analysis of pyroglutamyl formation revealed no significant differences between the wild-type and QC^mD126A strains (Table S4). Collectively, these results confirm that the importance of PgQC for P. gingivalis depends solely on the pyroglutamyl formation of proteins translocated across the IM by the Sec system.

Although these results suggest that physical interactions with Sec components are not an essential function of PgQC, its localization on the periplasmic side of the IM may be necessary for the modification of emerging N-terminal Gln residues as proteins are exported into the periplasm. Therefore, we replaced the Cys residue in the lipobox with Gln to produce mutant strain PgQC^C20Q. This was phenotypically indistinguishable from the parental strain (Fig. S2A through C), and the QC activity was also at the same level (Fig. S2D). However, the protein was distributed between the IM and periplasm (Fig. S2E) in the form of full-length QC and a truncated version, probably with a cleaved signal peptide. This makes it impossible to determine whether the association between QC and the IM is important for its function.

Given the large proportion of P. gingivalis secreted proteins that follow the Q-rule, we investigated the essential function of PgQC in more detail by replacing it with orthologs from other Bacteroidetes species with different proportions of Q-rule proteins: Porphyromonas macacae (PmQC), Porphyromonas somerae (PsQC), T. forsythia (TfQC), Prevotella intermedia (PiQC), Barnesiella intestinihominis (BiQC), and Pedobacter ginsenosidimutan (PedgQC). In addition to these animal-type QCs, we also replaced PgQC with the plant-type QCs expressed by Nonlabens sediminis (NsQC) and Alistipes indistinctus (AiQC). This allowed us to test QCs with a high level of sequence and structural divergence (Fig. 3A; Fig. S3A). The mutant strains were constructed by in-phase substitution of the PgQC coding sequence downstream of the lipobox with the coding sequence of heterologous QCs (without the signal peptide, but including a C-terminal Strep-tag).

Most of the substitutions yielded viable P. gingivalis mutant strains with colony morphology, generation time, stationary phase biomass, and gingipain activity indistinguishable from the parental strain (Fig. 3B and C; Fig. S3B and C). However, mutant strains were not recovered for AiQC or BiQC. The presence of PsQC^strep, PiQC^strep, and PedgQC^strep could not be detected on western blots probed with anti-Strep or anti-QC antibodies (Fig. 3D), but all mutant strains produced measurable QC activity that differed significantly from strain to strain, with the lowest values in the strains expressing NsQC and PsQC (Fig. 3E). Remarkably, although the QC activity was 44-fold lower in NsQC and 22-fold lower in PsQC compared to wild-type P. gingivalis, this was still sufficient for the pyroglutamyl formation of Sec-exported proteins at a level that supported normal bacterial growth (Table S4). Although the QC activity of P. gingivalis mutant strains expressing TfQC and PiQC correspond well to the activity of native T. forsythia and P. intermedia strains (19), we cannot exclude the possibility that NsQC and PsQC expressed by P. gingivalis are fully active on physiological substrates but exert a minute activity against the reference synthetic substrate.

The subcellular location of heterologous QCs was determined by western blot, revealing that most were anchored into the IM, with only NsQC^strep found exclusively in the OM (Fig. 4F). Interestingly, the IM retention signal of E. coli lipoproteins is defined by Asp (+2 residue) following lipidated Cys (+1 residue), whereas glycine (Gly +2 or Gly +3) may fulfill the same function in P. gingivalis (20). Our experimental data provide supporting evidence because only NsQC was anchored into the OM and only this enzyme lacks a Gly residue in the Cys(+1)-Lys(+2)-Thr(+3) motif (Fig. S3A). It is remarkable that P. gingivalis, which is clearly dependent on PgQC, can survive despite the very low activity of heterologous enzymes, regardless of their type and subcellular location.

The unequivocally verified dependence of P. gingivalis on PgQC activity suggests that (i) there are one or more essential secreted proteins that require N-terminal modification for their stability or activity or (ii) the overall lack of pyroglutamyl formation may exert a cumulative detrimental effect. We, therefore, inspected all proteins with a signal peptide in the new reference genome NZ_CP025932.1. Where possible, we assigned a cellular localization, protein structure, molecular function, and biological process based on published research, databases (UniProt, KEGG, and AlphaFold), and BLAST homology searches. Among 210 secreted proteins, 164 (77.14%) appeared to follow the Q-rule and are, therefore, potential substrates of PgQC (Table 1; for details, see Table S5). These included 41 (20%) annotated as “conserved hypothetical proteins,” indicating that their function cannot be predicted because, at most, they are only distantly related to characterized proteins.

The very high prevalence of Gln following the signal peptide (Q-value) occurs among proteins predicted to be functionally and structurally associated with the OM (88.4%), T9SS components and cargo proteins (83.3%), remnants of the type 4 secretion system (T4SS) (100%), periplasmic chaperones (85.7%), and, surprisingly, the products of genes that were deemed essential by transposon mutagenesis (84.6%) (8). The Q-value of proteins involved in peptide catabolism, peptidoglycan biosynthesis, and cell response/regulation was 61.5–75%, but it was much lower for proteins involved in carbohydrate catabolism (27.3%) and lipopolysaccharide synthesis (25.0%). These potential PgQC substrates were found in all extracytoplasmic compartments. We also looked for conserved secondary structures at the N-terminus using Alpha-fold, but few crystallographic structures are available. The analysis revealed a tendency for β-sheet proteins, but this structure was predicted unambiguously only for a subgroup of OM β-barrel proteins.

We, therefore, tested the effect of replacing the key Gln with Ala or Asn in four of the OM β-barrel proteins (PorG, PorQ, PorT, and PorV) and two predicted periplasmic proteins (PG_0320 and PG_1788). Homologous P. gingivalis strains expressing each mutated protein were generated, and protein levels were determined by probing western blots with antibodies specific for the target proteins (PorQ and PorV) or antibodies specific for the C-terminal His₆ tag (all other proteins). In all cases, the mutated protein accumulated to a significantly lower level than the normal protein in the control strain (Fig. 4A through K, left panels). This deficit was not due to transcriptional repression because, generally, the mutants showed elevated gene expression compared to the control (Fig. 4, right panels). The mutants were generally indistinguishable from the parental strains in terms of growth kinetics, pigmentation phenotype, and gingipain activity/distribution (Fig. S4 and S5). However, the porT ^Q30N,his strain was an exception, with slower growth than the control strains (Fig. S4C). Cumulatively, these results argue that pyroglutamyl formation is important for the stability of proteins exported to the periplasm and OM, apparently protecting them from proteolytic degradation and truncation, clearly visible in the case of PG_1788^Q21N,his (Fig. 4F). This was confirmed by the analysis of gingipains, which in the absence of a functional T9SS accumulate in the periplasm as multidomain single-chain proenzymes at a significantly lower level compared to control if the N-terminal Gln is replaced with Asn (Fig. 4G through I).

None of the mutant strains discussed above showed any loss of viability, but the target genes are known to be nonessential for in vitro growth on complex medium. We, therefore, repeated the same approach, but this time we mutated two genes (PG_0076 and PG_0449) that were previously shown to be essential by transposon mutagenesis (8). In both mutant strains, the N-terminal Gln was replaced with Asn. Both mutant strains were viable and were indistinguishable from the parental strains in terms of growth, pigmentation, and gingipain activity (Fig. S4J, K, S5J and K). This was the case, despite the much lower level of PG_0076 protein and the truncation of PG_0449 (Fig. 4J and K).

In order to determine the impact of pyroglutamyl formation on extracytoplasmic proteins of P. gingivalis, we purified the modified PG_0449 protein from PG_0449^his and PG_0449^Q22N,his strains and subjected to N-terminal sequencing. The results revealed 13 amino acid truncations of the N-terminus of PG_0449 in the latter strain indicating N-terminal proteolytic degradation. Nevertheless, for other substrates, we cannot exclude a direct structure/function stabilizing effect of pyroglutame formation. Generally, the mutated genes were expressed at a higher level than their wild-type counterparts, but the increase in mRNA abundance was inversely correlated with the amount of protein (Fig. 4B, D through H and K). Although a lower rate of translation or a difference in mRNA stability cannot be excluded, this result strongly suggests the presence of a feedback loop that enhanced transcription but was unable to compensate for the degradation of the modified protein. This seems to be a common phenomenon when Gln is replaced, resulting in low levels of mutated proteins compared to the parental strains (Fig. 4L).

Bacteria from the phylum Bacteroidetes are unique among procaryotes in that cleavage of signal peptides by signal peptidase I exposes a glutamine residue as the new N-terminus in the majority of substrates. In prior work, we had already shown that this N-terminal glutamine residue is indiscriminately cyclized to a pyroglutamate residue (4). Saturating transposon mutagenesis data for P. gingivalis further indicated that the single glutaminyl cyclase in this species, a type II (animal type) QC, was likely to be essential (8). However, direct evidence for essentiality remained elusive, and—assuming the enzyme was essential—it remained unclear whether the enzyme activity was essential or whether the protein was required because of some other non-catalytic function. Finally, the biological role of glutaminyl cyclization remained unresolved. Our initial hypothesis that pyroglutamate formation was a sorting signal distinguishing between periplasmic and extracellular targeting could be ruled out by the observation that glutaminyl cyclization in P. gingivalis and other representative Bacteroidetes species were pervasive and occurred for proteins irrespective of their final destination (4).

In this work, we provide definitive experimental evidence that the QC from P. gingivalis is essential, and we show that the activity of the protein is required, even under laboratory growth conditions in rich medium. Furthermore, we could demonstrate that orthologues of the P. gingivalis QC could substitute for the endogenous QC, even though their activity, at least as assayed using a fluorogenic reporter substrate, was drastically lower than the activity of the endogenous protein. The observation that a seemingly residual QC activity was sufficient to rescue P. gingivalis growth in culture conditions was surprising because a lead QC inhibitor compound had previously been shown to block the growth of P. gingivalis, T. forsythia, and P. intermedia in laboratory cultures (21). The data may be reconciled with our data assuming that the lead compound reduced QC levels to below the 2% of activity measured in our assays for some of the QC orthologues. Alternatively, it is possible that the activity levels measured with our short fluorophore reporter peptide may not be representative of protein substrates of the cyclization pathway.

A biological role of glutamine cyclization was identified by our observation that glutaminyl cyclization appeared to stabilize selected Q-rule substrates. The protection of proteins by the pyroglutamyl formation of N-terminal Gln is illustrated here by the significantly less abundant mutant proteins in which the key Gln is replaced with Asn. These are 40%–80% less abundant than the normal proteins in wild-type P. gingivalis, even though the mutant genes show an increase in transcriptional activity (Fig. 4). It is likely that degradation would be even more profound if Gln was replaced with a hydrophobic residue to match the preference of the more potent peptidases DPP7 and DPP11, which cleave dipeptides from substrates with hydrophobic N-terminal residues (22). Remarkably, none of the strains expressing a Gln→Asn mutated protein, including those essential for P. gingivalis, showed a significant loss of vitality when growing in vitro in rich medium. This contrasts with our experimental confirmation that PgQC is essential for P. gingivalis growth. The discrepancy may reflect the residual amount of mutated Q-rule essential proteins as apparent in the case of PG_0078^Q25N, which is sufficient for normal P. gingivalis growth. In addition, among eight essential proteins following the Q-rule (Table 1), we tested only two: PG_0076 and PG_0449. It is possible that one or more of the remaining proteins need N-terminal pGlu for its function or stability or that the cumulative effect of several protein depletions is necessary to evoke the detrimental effect. Alternatively, the lethal effect of PgQC inactivation may be due to the accumulation of partially degraded, inactive proteins in the periplasm.

On a technical level, our data only show that pyroglutamate formation stabilizes proteins, without providing a mechanism. It is extremely likely, however, that the protein stabilization is based on protection against aminopeptidases. The asaccharolytic nature of P. gingivalis metabolism makes this bacterium dependent on peptides generated in the extracellular milieu by an array of cell-surface endopeptidases (23). These peptides are translocated across the OM into the periplasm by a RagAB system (24) where they act as substrates for dipeptidyl-peptidases (DPP4, DPP5, DPP7, and DPP11) that release dipeptides from any N-terminus with a free α-amino group (9). DPPs work in concert with an acylpeptidyl-oligopeptidase and prolyl-tripeptidyl-peptidase to fragment oligopeptides (regardless of their sequence) into dipeptides (25). The latter are transported into the cytoplasm, where they are used as a carbon and energy source (26, 27). Interestingly, whereas DPP4 is widely distributed among prokaryotes, the Bacteroidetes genera Tannerella, Prevotella, Porphyromonas, Capnocytophaga, and Bacteroides also carry genes encoding homologs of DPP5, DPP7, and DPP11. The activity of these enzymes has been detected in Porphyromonas endodontalis, T. forsythia, and P. intermedia (26), the pathobionts contributing to the development of periodontitis. Notably, the growth of T. forsythia and P. intermedia as well as P. gingivalis was suppressed by a QC inhibitor (21), supporting our hypothesis that QC is an essential enzyme in species that follow the Q-rule. Signal peptidase I substrates in the phylum Bacteroidetes appear to have evolved to expose N-terminal Gln for pyroglutamyl formation by QC once the signal peptide is cleaved off (the Q-rule) in order to protect proteins exported into the periplasm or OM and secreted via the T9SS. Interestingly, the Q-rule has also been adopted by the recently discovered P. gingivalis phages to secrete proteins of yet unknown function (28). The intense proteolytic pressure in the periplasm and extracellular milieu of P. gingivalis and related Bacteroidetes species may well explain the applicability of the Q-rule to Bacteroidetes, but not other bacterial phyla, which may experience less pressure to stabilize periplasmic and extracellular proteins against proteolytic attack.

To our knowledge, the observation that protein cyclization serves as a protein stabilization mechanism is novel for bacteria. By contrast, it is a well-known phenomenon that, in eukaryotes, the cyclization of N-terminal residues often prolongs the half-life of proteins and peptides by protecting them from ubiquitous aminopeptidases (19, 29), as shown in studies focusing on chemoattractant proteins 1 and 3 (MCP-1 and MCP-3) and fractalkine (CX3CL1) (19, 30). However, this reaction also has broader biological significance. For example, malaria parasites evade the mosquito immune system by the pyroglutamyl formation of sporozoite surface proteins (31). In mammals, pyroglutamyl formation influences protein–protein interactions, as exemplified by ligands and receptor binding (32, 33). Importantly, extensive cyclization is related to pathological processes. It enhances the stability, hydrophobicity, and aggregation potential of the Aβ peptide underlying β-amyloid formation in Alzheimer’s disease (34); it forms pGlu79-α-synuclein, which promotes oligomerization and synucleinopathies in Parkinson’s disease (35, 36); and the modification of CD47 directly influences interactions with SIRPα to modulate the immunological surveillance mechanisms essential for the removal of cancer and senescent cells (37, 38). Together with the eukaryotic data, our data for bacteria make the case for deep evolutionary conservation of protein stabilization by glutaminyl cyclization.