Structural and functional probing of PorZ, an essential bacterial surface component of the type-IX secretion system of human oral-microbiomic Porphyromonas gingivalis.

Porphyromonas gingivalis is a member of the human oral microbiome abundant in dysbiosis and implicated in the pathogenesis of periodontal (gum) disease. It employs a newly described type-IX secretion system (T9SS) for secretion of virulence factors. Cargo proteins destined for secretion through T9SS carry a recognition signal in the conserved C-terminal domain (CTD), which is removed by sortase PorU during translocation. Here, we identified a novel component of T9SS, PorZ, which is essential for surface exposure of PorU and posttranslational modification of T9SS cargo proteins. These include maturation of enzyme precursors, CTD removal and attachment of anionic lipopolysaccharide for anchorage in the outer membrane. The crystal structure of PorZ revealed two β-propeller domains and a C-terminal β-sandwich domain, which conforms to the canonical CTD architecture. We further documented that PorZ is itself transported to the cell surface via T9SS as a full-length protein with its CTD intact, independently of the presence or activity of PorU. Taken together, our results shed light on the architecture and possible function of a novel component of the T9SS. Knowledge of how T9SS operates will contribute to our understanding of protein secretion as part of host-microbiome interactions by dysbiotic members of the human oral cavity.


Results and Discussion
PorZ is an essential component of T9SS. An isogenic deletion mutant of the porZ gene, Δ PorZ, was created by homologous recombination to assess its effect on T9SS cargo secretion and posttranslational processing. Deletion had a negligible effect on the P. gingivalis growth rate in complex media ( Supplementary Fig. S1). However, on blood agar, the mutant could not accumulate heme on the cell surface and therefore yielded non-pigmented colonies (Fig. 1a). This is attributable to gingipains, which are secreted T9SS cargos that are essential for hemoglobin degradation and heme recruitment [29][30][31] . Therefore, lack of pigmentation in Δ PorZ suggested failure of functional gingipain secretion and activation. Indeed, we found that Δ PorZ was deficient in extracellular Kgp and Rgp gingipain activities when compared to the wild-type P. gingivalis strain W83 (hereafter, "wild type"; Fig. 1b,c). In contrast, dipeptidyl peptidase IV and prolyl tripeptidyl peptidase A, which are surface enzymes but not secreted by T9SS, were produced and transported to the bacterial surface in significantly higher amounts than in the wild type (Fig. 1d). A similar response had been previously observed in an inactivation mutant of an essential T9SS component, PorT 32 . This presumably reflects general upregulation of peptidolytic enzymes as a response to the absence of functional gingipains, which account for 85% of the extracellular proteolytic activity of P. gingivalis 33 . Reconstitution in trans of the porZ gene in the Δ PorZ mutant-yielding PorZ + -restored both pigmentation and proteolytic activity to wild-type levels ( Fig. 1a-d).
To further investigate the fate of non-secreted T9SS cargos in the absence of PorZ, we performed Western blot analysis of distinct subcellular fractions to detect gingipains (Fig. 2a,b), PPAD (Fig. 2c), and the biotin-containing 15-kDa biotin carboxyl carrier protein (AccB alias MmdC or PG1609) as an IM marker ( Fig. 2d; see also ref. 32). The latter analysis revealed that the OM fractions obtained from the wild type and the mutant were contaminated with the IM. This is in contrast to undetectable contamination of the IM fraction with OM components, as indicated by the absence of gingipains and PPAD in the IM fraction. In the wild type, gingipains and PPAD were secreted onto the cell surface with CTD removal and proteolytic maturation of their precursors, which led to detectable activity in intact cells 34 . In Δ PorZ, they were not processed to the mature forms but rather accumulated as precursors in the periplasmic fraction and in clarified culture media (Fig. 2a-c). Additionally, partially processed gingipain precursors were found with anti-gingipain antibodies. Auto-processing of gingipains has been described in heterologously-overexpressed recombinant proteinases 35 . Moreover, the presence of considerable amounts of PPAD and gingipain precursors in the concentrated growth medium suggested that Δ PorZ had a "leaky" OM architecture. This contention was supported by peptide mass fingerprinting of proteins from the growth medium resolved on SDS-PAGE ( Supplementary Fig. S2). Although we found in the medium several proteins normally located in the periplasm, including prolyl oligopeptidase family proteins (PG0727 and PG1004), a MEROPS-M16-family peptidase (PG0196), thioredoxin (PG0275), HtrA protease/chaperone (PG0449) and TPR-domain protein (PG0449), CTD-bearing proteins were predominant. Indeed, of the 32 known T9SS cargos of P. gingivalis 28 , 12 were found in high abundance and apparently with intact CTDs in the growth medium of Δ PorZ, as indicated by high Mascot scores (Supplementary Table S1). These proteins included PorU (alias PG0026), carboxypeptidase D (alias Cpg70 or PG0232), PPAD (PG1424), internalin-like protein PG0350, putative hemagglutinin PG0411, immunoreactive 47-kDa antigen PG97 (PG1374), immunoreactive 46-kDa antigen PG99 (PG1798), heme-binding protein 30 (PG0616), and proteins PG0495, PG0654, PG1030, and PG2216. In Cultures were adjusted to OD 600 = 1.0 prior to testing and processing, and results shown correspond to triplicate experiments. Significant differences between the wild type and mutants are indicated by *P < 0.05 and ****P < 0.0001. addition, five other potential T9SS cargos were found in the medium, but none of the detected peptides corresponded to their CTDs. When the same proteins were detectable in the growth medium of the wild-type strain, they had much lower Mascot scores and no peptides corresponding to their respective CTDs. This "leaky" OM phenotype, which leads to release of non-cleaved CTDs from CTD-cargo proteins, is similar to that reported by Taguchi et al. 36 . These authors reported that the chaperone Skp-like protein (PGN_0300) is required for OM insertion of PorU sortase, which in turn is necessary for CTD cleavage from CTD-cargo proteins. Consistently, deletion of PGN_0300 resulted in failure of PorU insertion and, thus, T9SS function 36 . PorZ is located on the cell surface of P. gingivalis. Previous proteomics studies identified PorZ in the OM and OM vesicles (OMV) of P. gingivalis strain W50 27,28 . To determine the location of the protein more precisely, we performed Western blot analysis on wild-type cultures, quantitatively separated into whole cells and sub-cellular fractions, which included growth medium, periplasm, cytoplasm, and the cell envelope (IM+ OM). The latter was additionally fractionated with detergent into OM and IM fractions. Fraction purity was verified by Western blot using Rgp/Kgp and MmdC as markers for the OM and IM, respectively. Again, while the IM fraction was only very slightly contaminated with the OM, the latter fractions contained a notable amount of the IM, as indicated by the presence of MmdC in these fractions (Fig. 3a). In agreement with the predicted localization, PorZ was mostly found associated with the cell envelope and OM fractions (Fig. 3a), as confirmed by immunogold-staining electron microscopy (Fig. 3b). In addition, trace amounts of the PorZ protein were  detected in concentrated growth medium together with RgpA cat and Kgp cat , suggesting that, similarly to gingipain catalytic domains, also PorZ can be shed in low amounts from the bacterial surface into the medium.
To determine whether PorZ is located on the periplasmic or the extracellular side of the OM, we performed dot-blot analysis of intact and sonicated wild-type cells plus Δ PorZ as a negative control. While no signal was detected for the latter, equivalent staining was observed for both intact and sonicated cells employing two different anti-PorZ antibodies (Fig. 3c). Of note, the integrity of the P. gingivalis cell envelope was confirmed by dot-blot analysis with streptavidin-HRP showing the reaction only after cell disruption by sonication. These findings support cell-surface localization of PorZ (Fig. 3c). This location was further assessed by flow cytometry, with surface-exposed gingipain RgpB and intracellular MmdC as positive and negative controls, respectively. Indeed, when anti-PorZ and anti-RgpB antibodies were used, both proteins were identified on the cell surface (Fig. 3d,e), while no staining was detected with streptavidin-Alexa Fluor 488 conjugate (Fig. 3f,g). Specificity of the flow cytometry analysis was checked with Δ PorZ, which showed negligible staining, while a strong signal was found in the in trans porZ-complemented strain, PorZ + (Fig. 3h,i). Interestingly, FACS assays with P. gingivalis strain HG66, which cannot attach A-LPS to T9SS cargos so that these are subsequently released into the extracellular milieu, revealed that PorZ was still on the cell surface ( Supplementary Fig. S3). Finally, when intact P. gingivalis cells were gently mixed with pure distilled water or subcritical micellar concentrations of detergents, which are unable to disrupt OM integrity, a sizable fraction of PorZ was released into the liquid phase (Fig. 3j). Significantly, only trace amounts of gingipains were washed out of the bacterial surface under the conditions tested ( Supplementary Fig. S3).

PorZ affects expression of T9SS components and cargos.
To determine whether the absence of gingipain activity in the Δ PorZ mutant was also due to their reduced expression, quantitative RT-PCR was used to investigate expression of CTD-cargo proteins and other T9SS components in the Δ PorZ mutant. Interestingly, rather than suppression of CTD-cargo protein expression, the mRNA levels of CTD-cargo peptidases RgpB, Kgp and CPG70 in Δ PorZ were found to double or triple those of the wild type (Fig. 4). Thus, lack of activity was due to their failure to be secreted and maturated in Δ PorZ rather than to reduced expression. Further, deletion of PorZ induced significant upregulation of T9SS-components porT, lptO, and porN when compared to the wild type. In particular, the latter two triplicated the expression levels of the wild type (Fig. 4). Other components such as porO, porW, sov and porU did not show significant change. These trends were consistent regardless of whether r16s (Fig. 4) or gyrA (data not shown) was used as a reference gene. Overall, the significance of this variable effect on the expression of T9SS components in Δ PorZ is currently unknown.
Structural analysis of PorZ. We produced PorZ without its predicted signal peptide (residues Q 26 -R 776 ) by recombinant overexpression in Escherichia coli, and succeeded in crystallizing and solving its structure by  (a) Ribbon-type plot in cross-eye stereo of the crystal structure to 2.9 Å resolution of PorZ depicting domains β D1, β D2 and CTD, and the three domain-connecting linkers (white ribbons; labeled Lβ D2-β D1, Lβ D1-β D2, and Lβ D2-CTD). Each of the seven blades of propellers β D1 and β D2 (labeled counter-clockwise I to VII) is colored in yellow, orange, red, magenta, blue, turquoise and green, respectively; the CTD is in pink. A structural calcium-binding site (green sphere) is found within β D1-blade IV, and a tetraethylene glycol (TG) and a diethylene glycol (DG) were tentaively assigned on the protein surface (brown stick-models). Other (functionally probably irrelevant) ions and ligands were omitted for clarity. The central shafts of β D1 and β D2 are pinpointed on the entry and exit sides of the propellers by red and purple arrows, respectively. For labels and extension of regular secondary structure elements, see (b). (b) Topology scheme of PorZ, with β -strands as arrows and helices as cylinders, colored as in (a). The polypeptide chain spans residues G 29 -R 776 and the three constituting domains plus the linkers (in grey) are indicated with the residues delimiting each structural element (strands, bulges, helices, β -ribbons, blades and domains). The nomenclature adopted in the text for structure elements is "domain-blade-structural element", e.g. β D1-VI-β 3 or β D2-IV-β -ribbon. single-wavelength anomalous diffraction with a selenomethionine derivative. The structure was refined with data to 2.9 Å resolution and consists of three domains. The first two are consecutive N-terminal seven-stranded β -propeller or circular-leaflet moieties (β D1: residues K 39 -M 322 ; and β D2: G 29 -L 34 + Y 335 -T 679 37-39 ; PorZ residue numbering in superscript notation according to UP Q9S3Q8), each featuring a shallow cylinder or thick disk with an "entry side" and an "exit side" 38 . These domains are succeeded by a C-terminal domain (CTD, V 692 -R 776 ). The domains are connected by linkers (L): Lβ D2-β D1 (L 35 -H 38 ), Lβ D1-β D2 (P 323 -F 334 ), and Lβ D2-CTD (G 680 -G 691 ). The two propellers are offset from one another by a ~90° rotation about the intersection axis of the propellers' planes. This causes the overall molecular structure to be reminiscent of an easy chair of approx. maximal dimensions 95 × 80 × 55 Å, with β D1 as the seat, β D2 the backrest, and CTD the backrest support (Fig. 5a). The two entry-side surfaces of the PorZ propellers mimic, respectively, the seating and reclining surfaces of the chair.
The seven blades of the propellers consist of a four-stranded (β 1-β 4) antiparallel β -sheet of simple up-and-down "W" connectivity or β -leaflet topology 40 . Two β 1 strands and one β 4 strand are interrupted by bulges (strands β D2-III-β 1, β D2-IV-β 1, and β D1-III-β 4 Fig. 5b; for structural-element notation, see the legend to Fig. 5b). The blades are radially arranged around a central propeller shaft, which is lined by the respective first strands of each blade (β 1) originating on the entry sides, and the strands of each leaflet are connected by short loops. Exceptions are those connecting β D2-I-β 2 with β D2-I-β 3, and β D2-IV-β 2 with β D2-IV-β 3, which span twelve and eight residues respectively, and protrude from the entry side of β D2 (Fig. 5a). Within each sheet and ongoing from β 1 to β 4, the strands accumulate a twist of ~45°, which is right-handed as usual for such structures 41 . Uniquely, blade IV of β D2 contains a metal, which was assigned to a calcium based on the octahedral ligand sphere, chemical nature of the six (oxygen) ligands, and binding distances of the ligands (~2.4 Å; see http:// tanna.bch.ed.ac.uk/newtargs_06.html and ref. 42). The ion is bound by three main-chain oxygens, two aspartate side-chain oxygens, and a threonine side-chain oxygen (Fig. 5c). Each blade further consists of a C-terminal linker, which connects respective strand β 4 with strand β 1 of the downstream blade. Linkers vary in length between four and sixteen residues, and may contain extra regular secondary-structure elements such as β -ribbons (blades IV and V of β D2) and α -helices (blades I of β D2 and II of β D1; see Fig. 5b). The shafts of the PorZ propellers have an internal diameter spanning ~5-10 Å, and, in contrast to other propellers such as the four-fold hemopexin domains 38 , they do not show evidence of tight ion or ligand binding. We only found some loosely bound chemicals from the crystallization and vitrification conditions (data not shown), which could indicate potential binding sites of functional relevance. Further research will be required to verify this assertion, though.
In general, β -propeller symmetry ranges from four-fold to eight-fold, the most populated group being the seven-fold propellers [37][38][39] . This may result from the packing of the blades, which is considered more stable the larger the number of blades 39 . Thus, to circumvent low stability, four-fold and five-fold propellers incorporate additional elements to tether the circular arrangement, such as disulphide bonds between the N-and C-terminal blades 38,43 . Other mechanisms entail that the first blade is made up by the N-terminal part of the polypeptide chain for some of its strands and by the C-terminal part after completion of the entire propeller moiety for the remaining strands, thus featuring a kind of "velcro" mechanism 38 . In the case of PorZ, which consists of two consecutive seven-fold propellers, the N-terminal segment of the polypeptide chain follows such a mechanism and forms strand β D2-VII-β 4, while the remaining three strands of this blade are provided by the polypeptide chain after forming both propellers (see Fig. 5b). After strand β D2-VII-β 4, the polypeptide enters short, four-residue Lβ D2-β D1 and then forms β D1 starting with blade I. After β D1, which ends with strand β D1-VII-β 4, Lβ D1-β D2 leads to strand β D2-I-β 1. Altogether, β D1 and β D2 tightly approach each other through respective blades I and VII and linkers Lβ D2-β D1 and Lβ D1-β D2 (Fig. 5a). This may explain the stable, nearly perpendicular relative arrangement between propellers, which is reminiscent of that of the two four-fold propellers found in rabbit hemopexin 38,44 .
After β D2, Lβ D2-CTD leads to CTD (Figs 5a,b and 6a), which spans 85 residues and is organised as a β -sandwich consisting of a three-fold antiparallel β -sheet (strands CTD-β 1, -β 2, and -β 5; CTD-β 1 is interrupted by a bulge, see Fig. 5b) and a four-fold antiparallel β -sheet (CTD-β 4, -β 3, -β 6, and -β 7). Despite some general differences, in particular in loops, the architecture and strand connectivity is equivalent to that of the CTD of RgpB, which is the structural paradigm of a functional CTD required for T9SS secretion 12 . Both structures were superposed for 61 topologically equivalent residues, giving rise to an overall core rmsd of 1.8 Å and a sequence identity of 20% (see Fig. 6a-c). Although the latter value is rather low 45 , both structural and sequence similarity are particularly significant for the last ~25 residues of either molecule (Fig. 6b), a segment that encompasses the apparent signal recognized by T9SS for translocation 12,15 . A common pattern (G-V-Y-V/A-V-X-I/V) arises when the two sequences are aligned based on structural criteria (Fig. 6c) and the sequence of the T9SS cargo HBP35 15 is further included (Fig. 6b). Thus, PorZ possesses a potentially functional CTD for T9SS secretion.
Possible function of the PorZ propeller domains. The modular architecture of PorZ is reminiscent of that of the periplasmic sensor-domain moiety of protein BT4663 from Bacteroides thetaiotaomicron, which belongs to the human microbiome and is the most prevalent gut colonizer 10 . BT4663 is the transmembrane histidine kinase of a two-component signal transduction system engaged in detection and degradation of complex carbohydrates. BT4663 transits between distinct unbound and bound conformations, and in this way activates the intracellular kinase domain 46 . The general architecture of BT4663 is also found in related potential histidine kinases BT4673 and BT3049 from the same organism 47 . Similarly to PorZ, the three proteins likewise comprise two N-terminal seven-fold β -propeller domains and a C-terminal all-β domain, termed the Y_Y_Y domain for BT4663 46 .
However, while BT4663 and related proteins are dimeric 46,47 , PorZ was verified in size-exclusion chromatography to be a monomer in solution (data not shown). In addition, the relative arrangement of the three domains in either the bound or unbound forms, differs from that of PorZ. Differences also arise between the Y_Y_Y domain and PorZ CTD: although they share topology and strand connectivity, the former is much larger (120 vs. Scientific RepoRts | 6:37708 | DOI: 10.1038/srep37708 C-terminal strands β 6 and β 7, which contain the reported molecular determinants for T9SS secretion 12,15 , are framed and labeled in the right panel. (b) Sequence alignment of the 22 C-terminal residues of HBP35 (UniProt Q8G962) and RgpB (UniProt P95493), and the 25 final residues of PorZ after the structural alignment of the structures of the latter two proteins (see also [c]). Identical residues are in red, similar ones in green. G-X-Y sequences are also found in PKD proteins within strands equivalent to CTD-β 6. These are also seven-stranded immunoglobulin-like all-β domains 84 , although the function of the G-X-Y motif therein is unknown. In addition, the Y_Y_Y domain of BT4663 protein contains this signature 46 . (c) Detail in cross-eye stereo showing the strands framed in (a) as full-atom models, i.e. segments S 752 -R 776 of PorZ (with pink carbons and magenta labels) and N 486 -K 507 of RgpB (mature protein numbering as subscripts, add 229 for full-length protein numbering; cyan carbons and labels).
Scientific RepoRts | 6:37708 | DOI: 10.1038/srep37708 85 residues) and has an extra strand. Moreover, while the Y_Y_Y domain functionally acts as a spacer from the cytoplasmic membrane surface further engaged in dimerization, PorZ CTD is a potential signalling domain (see below).
This notwithstanding, BT4663 may yet provide a hint to the molecular function of PorZ, as most of the known propeller domains coordinate ligands or catalyse reactions at or close to the central shafts on their entrance side 37,48 . Among its widespread functions are binding of sugar moieties and related molecules as reported for the five-fold propellers β -fructosidase, α -L-arabinanase and levan-sucrase. This is also the function of BT4663, which binds disaccharide ligands on the respective entry surfaces of both propeller domains, central to the shafts 46 . Along this line, the crystal structure of PorZ revealed potential polyethylene glycol fragments bound at the hinge between β D1 and β D2 on their entry side, as well as on the exit surface of β D2 (see Fig. 5a). Other smaller molecules were found in or on the shaft entrances (see above). Moreover, the two extended intra-leaflet loops, which protrude from the entry side of β D2 (see above), could also be potentially engaged in ligand binding, thus providing a functional explanation for their exceptional length. In any case, further experiments will be required to verify a potential glycan-binding function of PorZ as part of, or independently from, T9SS secretion.
PorZ is itself secreted via T9SS with an intact CTD. The presence of a CTD reminiscent of that found in T9SS cargos and components at the C-terminus of PorZ suggests that the protein is itself secreted to the surface by the same secretion system. In this event, PorZ would be absent from the bacterial surface in P. gingivalis mutants with dysfunctional T9SS, as shown for other T9SS cargos. To validate this hypothesis, we investigated the surface location of PorZ by flow cytometry analyses in three secretion-deficient mutants: Δ PorN, Δ PorU, and a mutant expressing inactive PorU, in which the catalytic cysteine had been mutated to alanine (PorU C690A ). The defective secretion phenotype of the Δ PorN and Δ PorU mutants was confirmed by the lack of staining for RgpB (Fig. 7a). In contrast to the wild type, PorZ was absent from the cell surface of mutant Δ PorN but surprisingly not of mutants Δ PorU and PorU C690A (Fig. 7b), which evinced similar levels to the wild type (Fig. 7c). This strongly argues that PorZ is translocated across the OM and anchored at the bacterial surface in a PorU-independent manner.
Due to the phenotypic similarity of the Δ PorZ mutant with the Skp-like PGN_0300 mutant, where PorU failed to insert into the OM 36 , we proceeded to investigate the surface exposure of PorU in the Δ PorZ and Δ PorN mutants as compared to wild type. Flow cytometry using an anti-PorU antibody revealed that PorU was absent from the cell surface of both Δ PorZ and Δ PorN, in stark contrast to the wild type (Fig. 7d). It was not present in the Δ PorU mutant as expected but the complementation of the Δ PorZ mutant restored the surface exposure of both PorU and RgpB (Fig. 7e). This result suggests that the export of PorU to the cell surface may be a complex process involving both an intact T9SS pathway as well as the Skp-like chaperone PGN_0300. It further suggests that PorU may be the last component of T9SS to be secreted to the surface as it requires a functional PorN and PorZ but not vice-versa, as PorZ was exported to the cell surface in the absence of wild-type or inactive PorU (Fig. 7a).
To determine the fate of PorZ in Δ PorN, we performed Western blot analysis on cultures that were quantitatively separated into whole cells and subcellular fractions. When compared with the wild type, the partition of PorZ between the OM and periplasmic fractions was only slightly higher in the periplasm of Δ PorN (Fig. 7f). This is in stark contrast with Rgps and Kgp, which are fully processed and associated with the OM in the wild type (Fig. 2a,b) but found in unprocessed and partially processed forms in the periplasm of Δ PorN ( Supplementary Fig. S5). The same distribution of unprocessed gingipains in the periplasm is apparently a hallmark of all T9SS secretion mutants characterized to date, including Δ PorZ (Fig. 2), PorT 49 , PorU 17 , Sov 50 , and LptO 51 . Taken together, our results unambiguously show that PorZ is associated with the OM, and its surface exposure is dependent on PorN but not PorU (Fig. 7b,c). Conversely, PorU exposure on the cell surface requires the presence of PorZ (Fig. 7d,e).
As to which variant of PorZ is found on the cell surface, the relative molecular mass of a PorZ-immunoreactive band in SDS-PAGE was ~80 kDa, which suggests that the protein is full length, without the signal peptide (theoretic molecular mass: 81 kDa). To verify this contention, we constructed P. gingivalis mutant strain R776i8H, which expresses PorZ with an octahistidine at the C-terminus (see also the next section). This mutant possesses a secretory phenotype that is indistinguishable from the wild type, as determined by colony pigmentation (data not shown), cellular distribution (Fig. 7e) and gingipain activity (Fig. 8f). Western blot analysis with anti-His-tag antibodies revealed reactivity to a band of ~80 kDa, which confirmed the presence of intact CTD in the mature PorZ protein (Fig. 7g). This observation is consistent with proteomics data reporting that PorZ appears to retain its CTD and does not undergo A-LPS modification as seen in other T9SS cargos 27 .
An intact CTD is required for cargo translocation across the OM and post translational processing. It was previously described that an intact CTD was needed for proper secretion, because its removal or C-terminal truncation prevented OM translocation and resulted in accumulation of the cargo protein in the periplasm 32 . However, insertion of additional residues into the linker region between the CTD and the preceding immunoglobulin superfamily domain of RgpB resulted in cleavage of the CTD but prevented A-LPS attachment. In addition, the N-terminal pro-domain responsible for latency maintenance in pro-RgpB 52 was cleaved off, so soluble, fully active gingipain was secreted into the medium 53 .
The crystal structure of PorZ revealed that CTD is preceded by domain β D2, with the inter-domain linker Lβ D2-CTD spanning residues G 680 -G 691 (see above). To study the effect of oligohistidine insertions or replacements on PorZ expression, processing and translocation, we inserted hexa/octahistidines or replaced six consecutive residues with histidines within Lβ D2-CTD (six mutants) and at the C-terminus of the CTD (three mutants) (Fig. 8a). Western blot analysis with either anti-PorZ (Fig. 8b) or anti-His-tag antibodies (Fig. 8c) revealed Scientific RepoRts | 6:37708 | DOI: 10.1038/srep37708 81-kDa (full-length) and/or ~75-kDa (truncated) immunoreactive bands, which were found with the former antibody in all strains tested except Δ PorZ (Fig. 8b). Both forms were found in mutant I770i6H with the anti-PorZ antibody, and the shorter fragment was exclusive for the I770 > 6 H and L689 > 6 H mutants. In turn, results with anti-His-tag antibodies showed the oligohistidine motif in full-length and/or truncated PorZ in all mutants expressing the protein tagged in the Lβ D2-CTD segment (Fig. 8c, left panel). In the case of the C-terminal mutants, the tag was detected only in two mutants expressing, respectively, PorZ C-terminally extended with an octahistidine attached to C-terminal residue R 776 (mutant R776i8H) and when hexahistidine was inserted after I 770 (mutant I770i6H) (Fig. 8c, right panel). In contrast, no tag was detected in the ~75-kDa form of PorZ with hexahistidine substituting the six most C-terminal residues (I770 > 6 H) and when inserted in the middle of the last β -strand (CTD-β 7) of the protein (I770i6H). The lack of reactivity of the ~75-kDa band with anti-His-tag antibodies in this mutant clearly indicates that mutated PorZ was cleaved at the C-terminus, losing ~5-kDa.
To verify on which side (N-or C-terminal) the protein was cleaved in Lβ D2-CTD linker mutants, we purified the truncated PorZ variant derived from the L689 > 6 H mutant by affinity chromatography, and subjected the protein to N-terminal sequencing analysis. The N-terminus was found to be intact, so cleavage must have occurred at the C-terminus, downstream of the inserted histidine-tag. The same can be assumed for the ~75-kDa PorZ variant derived from mutant Q678i6H.
Next, we determined the function of the PorZ variants in the processing and activation of gingipains. A membrane-bound form of RgpB (mt-RgpB) and the mature RgpA catalytic domain (RgpA cat ) were clearly visible in the wild type and in mutants F677i8H, Q678i6H, S683 > 6 H, A686 > 6 H, D690i6H, I770i6H and R776i8H (Fig. 8d), which all possessed full-length PorZ (Fig. 8b,c). Conversely, mutants L689 > 6 H and I770 > 6 H, which do not have this PorZ variant, entirely lacked both mt-RgpB and RgpA cat (Fig. 8d). Similarly, the mature catalytic domain of Kgp was found only in mutants possessing full-length PorZ, while only the Kgp precursor was present in mutants missing full-length PorZ (Fig. 8e). Lack of gingipain processing in these mutants correlated with substantially lower Rgp and Kgp activities in the culture than in the wild type (Fig. 8f). The only exception was mutant Q678i6H, which seemed to process gingipains, but with much lower yields than the wild type (Fig. 8f). Finally, flow cytometry studies (Fig. 8g) revealed that all mutants with full-length PorZ displayed surface-exposed PorZ at levels comparable to those of the wild type. Conversely, strains with truncated PorZ (L689 > H6 and I770 > 6 H) had negligible amounts of this protein on the cell surface, despite comparable expression levels of modified PorZ, as shown by Western blot (Fig. 8b,g).

Conclusions
Members of the dysbiotic oral microbiome, such as T. forsythia and P. gingivalis, and some environmental Gram-negative bacteria uniquely possess T9SS dedicated to the export of proteins. This multi-component machinery consists of a minimum of twelve Por proteins and is responsible for the secretion of at least 32 proteins in P. gingivalis. The export signal is not a flexible peptide but rather a full ~70-residue protein domain located at the C-terminus called the CTD, which is removed upon secretion. Here, we discovered a new component of T9SS, PorZ, and found it loosely associated with the P. gingivalis cell surface in a manner independent of A-LPS anchorage. In T9SS cargo enzymes such as PPAD and gingipains, its absence prevented proper secretion to the extracellular medium, cleavage of CTDs, activation from precursor forms, and anchoring in the OM. In this case, cargos accumulated in the periplasmic space and the culture medium, the latter owing to leaky OM architecture. Increased mRNA expression of CTD-cargo proteins in the Δ PorZ mutant suggested that absence of these functional proteins on the cell surface induced a feedback response, which augmented expression of these proteins and of some T9SS components, presumably in an effort to secrete more cargo proteins to the surface.
To gain insight into the molecular determinants of PorZ function, we solved its full-length crystal structure, which revealed two N-terminal seven-fold β -propeller domains in tandem. Such domains are widely used in macromolecular recognition and are engaged in protein-protein and protein-substrate interactions, as the architecture is generally versatile enough to enable binding of small molecules like sugars 43 . Based on structural similarities with sugar-binding proteins from B. thetaiotaomicron from the human gut microbiome, we hypothesized that PorZ may likewise have a potential glycan-binding function as part of, or independently from, T9SS secretion. Downstream of the β -propeller domains, PorZ further comprises a C-terminal seven-stranded β -sandwich, which conforms to the canonical CTDs of other T9SS-secreted proteins. This allowed us to further hypothesize that PorZ may itself be a T9SS cargo. Further functional studies showed that PorZ is actually transported to the cell surface via T9SS as a full-length protein with an intact CTD-a translocation that was independent of the presence or activity of sortase PorU. Consistently, PorZ was absent from the surface but apparently remained associated with the periplasmic side of the OM in a T9SS mutant lacking protein PorN. We further studied the effect of oligohistidine insertions or replacements on PorZ expression, processing and translocation to the bacterial surface, and inserted hexa-or octahistidines or replaced six consecutive residues with histidines within linker Lβ D2-CTD and at the C-terminus of CTD. Collectively, these mutations revealed that although CTD can be elongated at the C-terminus without affecting the secretory phenotype of the mutant, alteration of the native C-terminal sequence was absolutely prohibited 26,53 . Interestingly, in both cases truncation/ substitution of C-terminal residues resulted in partial cleavage of CTD and accumulation of modified protein in the periplasm. However, modifications of the loop preceding CTD in PorZ had a negligible effect on protein function, as long as they did not affect strand CTD-β 1. In T9SS cargos, such mutations decouple CTD removal by sortase PorU from the attachment of A-LPS, thus releasing fully-processed cargo proteins into the culture medium. This accounts for different functions of the linker domain preceding CTD, which is unstructured in cargos and provides a cleavage/A-LPS attachment site for sortase PorU 12 . In stark contrast, the intra-domain loop is well-structured in PorZ and resistant to proteolysis by PorU, thus suggesting an important role of CTD in PorZ function in T9SS.
To sum up, the reported full molecular and functional characterization of PorZ, a novel essential surface component of T9SS, contributes to our understanding of protein secretion as part of host-microbiome interactions by dysbiotic members of the human oral cavity.

Materials and Methods
Bacterial strains and general growth conditions. Porphyromonas gingivalis strain W83 (wild type and mutants, listed in Table 1) was grown in enriched tryptic soy broth (eTSB per liter: 30 g trypticase soy broth, 5 g yeast extract, 5 mg hemin, pH 7.5; further supplemented with 0.5 g L-cysteine and 2 mg menadione) or on eTSB  blood agar (eTSB medium plus 1.5% agar, further supplemented with 4% defibrinated sheep blood) at 37 °C in an anaerobic chamber (Don Whitley Scientific, UK) with an atmosphere of 90% nitrogen, 5% carbon dioxide and 5% hydrogen. Escherichia coli strains (listed in Table 2), used for all plasmid manipulations, were grown in Luria-Bertani (LB) medium and on 1.5% agar LB plates. For antibiotic selection in E. coli, ampicillin was used at 100 μ g/ml, kanamycin at 50 μ g/ml, spectinomycin at 50 μ g/ml and erythromycin at 250 μ g/ml. P. gingivalis mutants were grown in the presence of erythromycin at 5 μ g/ml and/or tetracycline at 1 μ g/ml.

Generation of P. gingivalis ΔPorZ, ΔPorU and ΔPorN deletion mutants.
All P. gingivalis deletion mutants were generated by homologous recombination, as previously described for other genes 54 . Three suicide plasmids were generated analogously. For deletion mutant Δ PorZ, briefly, two 1-kb flanking regions on either side of the porZ gene were amplified from genomic DNA by PCR with primer pairs PG1604FrANdeIF/ PG1604FrASmaR (for primers and sequences, see Supplementary Table S2) and PG1604FrBXbaIF/ PG1604FrBSalIR. The PCR product was cloned into plasmid pUC19 with an inserted erythromycin-resistance (erm) cassette (ermF-ermAM), amplified from plasmid pVA2198 55 with primers ermFAMSmaIF/ermFAMSalIR. Correct placement and orientation of the DNA segments in resulting plasmid p1604AeB-D was confirmed by sequencing.
A plasmid for mutant Δ PorN, designated p291AeB-C, was generated similarly using vector pUC19 and primer pairs PG291FrBXbaIF/PG291FrBPstIR and PG291FrANdeIF/PG291FrASmaIR, as well as primers ermFAMSmaIF/ermFAMXbaIR to introduce cassette ermF-ermAM between chromosomal DNA fragments.
In like manner, a plasmid for mutant Δ PorU was obtained using primer pairs PorUFrag5_F/PorUFrag5_R and PorUFrag3_F/PorUFrag3_R to amplify 1-kb upstream and downstream flanking regions of the porU gene respectively. Blunt-end PCR products were cloned into plasmid pCR-BluntII-TOPO using the Zero Blunt TOPO PCR cloning kit (Invitrogen) and then subcloned into a pUC19 plasmid containing a promoterless erm resistance gene 55 yielding plasmid pPorU/pUC19/Erm.
The resulting suicide plasmids for generating the Δ PorU, Δ PorZ and Δ PorN mutants were introduced by electroporation into electrocompetent wild-type P. gingivalis cells prepared as previously described 56 . Resultant clones were selected on erythromycin plates and double-crossover genomic recombination was confirmed by DNA sequencing of the manipulated region. Isogenic mutant Δ PorU was further verified for inoffensiveness on translation of the downstream gene encoding protein LptO (alias PorV and PG0027) through Western blot analysis using rabbit polyclonal anti-LptO antibodies.
Similarly to the reported generation of RgpB and Cpg mutants 53 , SLIM mutagenesis 57 was employed to replace the catalytic cysteine of PorU (C690) with alanine within the pPorU-E master plasmid to yield mutant PorU C690A .
PorZ mutants including insertion of an oligohistidine-tag or substitution of six consecutive residues with six histidines at various locations at the inter-domain linker between domains β D2 and the CTD (mutants PorZ F677i8H, Q678i6H, S683 > 6 H, A686 > 6 H, L689 > 6 H, and D690i6H) or at the C-terminus (mutants PorZ I770 > 6 H, I770i6H, and R776i8H) were generated by the SLIM method within the porZ master plasmid (see Supplementary Table S2). Mutated plasmids were verified by DNA sequencing of the pertinent region before being electroporated into wild-type P. gingivalis for homologous recombination 54 . Resistant clones were selected using erythromycin-selective media.

Construction of porZ complementation strain (PorZ + ). To construct a complementation plasmid
(pT-COW-porZ), a 3166-bp fragment containing the entire porZ coding sequence plus 544 bp upstream and 291 bp downstream was obtained by PCR from P. gingivalis genomic DNA with primers pT-COWporZForSalI and pT-COWporZRevNheI. Next, the fragment was cloned into SalI and NheI restriction sites of the pT-COW plasmid 58 by a standard procedure. After pT-COW-porZ sequencing, the plasmid was introduced into Δ PorZ strain by conjugation according to 59 . Culture partitioning and subcellular fractionation. Two-day-old cultures in the stationary phase of wild-type and mutant P. gingivalis were adjusted to OD 600 = 1.5 and designated "whole culture" (WC). Bacterial cells were collected from WC by centrifugation (8000× g, 15 min). The cell pellet was then washed and resuspended in PBS up to the initial volume of WC, and centrifuged. This fraction is referred to as "washed cells". The collected cell-free culture medium was ultracentrifuged (100,000× g, 1 h) to remove vesicles and the supernatant was concentrated 10-fold by ultrafiltration with filter devices of 10-kDa cut-off; this fraction was designated "medium" (Med). Subcellular fractionation was carried out as described 60 , with some modifications. Briefly, bacteria were cultured until OD 600 = 1.2 and fraction "cell extract" was obtained by centrifugation of 0.5 ml of WC and a single wash step with PBS. The rest of the culture (7.5 ml) was centrifuged to produce a supernatant and a pellet. The supernatant was filtered through a 0.22-μ M syringe filter and concentrated three times with centrifugal filter devices of 10-kDa cut-off to obtain the "medium" fraction. The pellet was rinsed with PBS and suspended in 2.5 ml buffer containing 0.25 M sucrose, 30 mM Tris·HCl, pH 7.6. After a 10-min incubation period, cells were pelleted (12,500× g, 15 min) and rapidly resuspended in 2.5 ml of cold distilled water to disrupt outer membranes.
Scientific RepoRts | 6:37708 | DOI: 10.1038/srep37708 After a further 10-min incubation period, spheroplasts were obtained and separated by centrifugation (12,500× g, 15 min). The supernatant was collected and designated as the "periplasmic" fraction. Next, spheroplasts were washed once with PBS, suspended in 2.5 ml of fresh PBS, and disrupted by brief sonication. The resulting solution was ultracentrifuged (150,000× g, 1 h) and the soluble fraction was designated as the "cytoplasmic" fraction. The pellet was resuspended in 2.5 ml PBS and briefly sonicated, and was designated as the "cell envelope" fraction. Inner membranes were solubilized by incubation in PBS supplemented with 1% Triton X and 20 mM magnesium chloride (final concentration) for 20 min. Membranes were separated by ultracentrifugation (150,000× g, 1 h) and the resulting supernatant was dubbed the "inner membrane" fraction. The pellet was then resuspended in PBS and sonicated, which resulted in the "outer membrane" fraction. All fractions were supplemented with peptidase inhibitors (5 mM tosyl-L-lysyl-chloromethane hydrochloride [TLCK], 1 mM 2,2'-dithiodipyridine [DTDP], 1x EDTA-free protein inhibitor cocktail (all from Roche) before storage at − 20 °C.
Antibodies. Rabbit anti-RgpB antibodies, which also recognize the catalytic domain of RgpA 61 , were used as the primary antibody. Mouse monoclonal antibodies specific for the Kgp catalytic domain and a single epitope on RgpB (18E6), respectively, were developed at the University of Georgia (GA, USA) Monoclonal Antibody Facility using recombinant protein as the antigen 54,62 . A rabbit polyclonal antibody recognizing P. gingivalis peptidylarginine deiminase (PPAD) was generated by Cambridge Research Biochemicals, Billingham, UK 63 . Mouse monoclonal antibodies specific for PorU were developed using standard procedures in the Laboratory of Monoclonal Antibodies at the Malopolska Center of Biotechnology, Jagiellonian University (Krakow, Poland) with recombinant PorU as the immunizing antigen.
Mouse polyclonal antibodies against protein PorZ were obtained according to standard procedures. Briefly, six-week old Balb/C mice were injected intraperitoneally with 100 μ g of antigen diluted in PBS and mixed 1:1 with Complete Freund's Adjuvant (Sigma Aldrich). Subsequent immunizations were performed with 50 μ g of antigen mixed with Incomplete Freund's Adjuvant (Sigma Aldrich). The serum titer of polyclonal antibodies specific to the antigen was determined by ELISA according to standard procedures. After administration of an adequate anaesthetic (ketamine and xylazine), blood was collected from immunized animals by cardiac puncture. Polyclonal antibodies were purified from the serum by affinity chromatography with protein G, according to the manufacturer's instructions (Thermo Fisher Scientific), and then dialyzed against sterile PBS. Finally, the antibodies were purified by affinity chromatography employing recombinant PorZ, which was covalently bound to Pierce ™ NHS-Activated Agarose Spin Columns (Thermo Fisher Scientific), according to the manufacturer's instructions. This work was carried out in the Laboratory of Monoclonal Antibodies at the Małopolska Center of Biotechnology, Jagiellonian University (Krakow, Poland). Mouse monoclonal antibodies against PorZ were also generated commercially by Abmart (Shanghai, China) using a synthetic peptide of sequence E-K-G-R-K-T-T-Q-F-P. All experiments were performed in accordance with relevant guidelines and regulations.
Western blot analysis. Sample preparation, separation by SDS-PAGE, electro-blotting onto polyvinylidene difluoride or nitrocellulose membranes and blocking were performed as described previously 53 . RgpA and RgpB catalytic domains were detected with rabbit polyclonal anti-Rgp antibodies; Kgp with mouse monoclonal anti-Kgp antibodies; PPAD with rabbit polyclonal anti-PPAD antibodies; and PorZ with mouse polyclonal anti-PorZ antibodies (see also above). In addition, oligohistidine-tagged PorZ was detected using mouse monoclonal THE TM His-Tag antibodies (Genscript, USA) at 0.125 μ g/ml. Development with polyclonal anti-mouse or anti-rabbit horseradish peroxidase-conjugated secondary antibody (BD Pharmingen and Amersham Pharmacia) was carried out using the ECL Western Blotting substrate kit according to the manufacturer's instructions (Pierce, UK). Streptavidin conjugated to horseradish peroxidase was used to detect MmdC, a biotinylated IM-associated protein.
Dot blot analysis. P. gingivalis cells were harvested at OD 600 = 1.5, washed and resuspended in cold PBS to yield OD 600 = 1.0. Half of the suspension was sonicated to disrupt cell membranes, and 5 μ l of either intact or disrupted cell suspension were spotted onto 0.22-μ m nitrocellulose membranes and air dried. Subsequent steps were performed as described for Western blot analysis.
Flow cytometry analysis. Wild-type and mutant P. gingivalis strains were grown in eTSB until they reached the late exponential or early stationary growth phase (OD 600~1 .2-1.5). Bacterial cells were harvested by centrifugation, washed twice with PBS and adjusted to OD 600 = 1.0 with buffer (PBS supplemented with 1% bovine serum albumin). Then, 100 μ l of cell suspension was transferred to a 96-well plate and incubated for 30 min with the previous buffer. Cells were collected by centrifugation (5,000× g, 5 min) and the pellet was resuspended in the previous buffer containing mouse antiserum specific for PorZ, RgpB or PorU at a total protein concentration of 30 μ g/ml, and incubated for 30 min. Thereafter, cells were centrifuged (5,000× g, 5 min) and the newly obtained pellet was resuspended in the previous buffer containing goat anti-mouse antibody conjugated with fluorescein isothiocyanate (Abcam) at 1:200 dilution and incubated for 30 min. Cells were washed twice with PBS after each incubation with antibodies or streptavidin-Alexa Fluor 488 conjugate. The whole staining procedure was performed on ice. After staining, one-color flow cytometry analyses were performed using a FACSCalibur apparatus (BD Biosciences) operating with CellQuest software (BD Biosciences). Graphs were prepared using the FLOWJO v.10 program (Ashland, USA). Grids were then treated with 2% osmiate vapour for 5 min, stained first with 1% uranyl acetate and then with Reynolds' lead citrate 64 . Grids were visualized in a 100-kV FEI CM120 electron transmission microscope and digital images were taken with a SIS Morada CCD camera.
Quantitative RT-PCR analysis. Quantitative RT-PCR data were obtained from four separate experiments.
Briefly, wild-type and ∆ PorZ cells were cultured to the mid-exponential phase and collected at OD 600 = 0.8-0.9. Cells were stabilized with RNAprotect Bacteria Reagent (Qiagen, Germany) prior to mRNA extraction with the RNAqueous Total RNA Isolation Kit (Ambion, USA). Reverse transcription was carried out on 2 μ l of total RNA using the AffinityScript qPCR cDNA Synthesis Kit (Stratagene, Australia) in 10 μ l total volume, according to the manufacturer's instructions. Quantitative RT-PCR was carried out in 25-μ l singleplex reactions using 5 μ l of a 1:50 dilution of cDNA along with TaqMan probe and primers against genes related to T9SS components (porT, sov, porU, lptO, porN, porO and porW) or cargos (rgpB, kgp, and cpg70). A housekeeping DNA gyrase (gyrA) gene and ribosomal 16 s (r16s) were used as calibrator genes (see primers and probes in Supplementary Table S2).
Thereafter, thermal cycling with a Stratagene Mx3005 P Real-Time PCR System ® using the Brilliant ® II QPCR Master Mix (Stratagene, Australia) entailed activation for 15 min at 95 °C plus 40 cycles of annealing at 95 °C for 20 s and extension at 60 °C for 1 min. Fluorescence intensities were normalized against a passive fluorophore carboxy-X-rhodamine (ROX) present in the Master Mix and converted to absolute quantities using standard curves. The expression of target genes was normalized to the housekeeping gyrA and r16s genes independently by  expressing the data as the average threshold cycle value of the target gene divided by that of either housekeeping gene. The average ratio in the wild type was arbitrarily set to 1.0 for reference.

Large-scale expression, purification and oligomeric characterization of recombinant PorZ.
Genomic DNA was isolated from wild-type P. gingivalis strain W83 using the Genomic Mini System (A&A Biotechnology, Gdansk, Poland), according to the manufacturer's instructions. The gene encoding residues Q 26 -R 776 (PorZ residue numbers in superscript notation, see UniProt entry Q9S3Q8), i.e. without the predicted signal peptide (M 1 -A 25 ), was amplified by PCR, purified, and cloned into the pGEX-6P-1 expression vector using SmaI/NotI sites (for primer sequences, see Supplementary Table S2  Mass spectrometry of media samples. Proteins in concentrated (20x) particle-free growth medium were ultracentrifuged and resolved by SDS-PAGE (NuPAGE Novex Bis-Tris System). Gels were stained with SimplyBlue ™ SafeStain (Novex) and washed in distilled water. Gel lanes were cut into sections and subjected to mass spectroscopy and follow up analysis, as described earlier 65 .
Enzyme activity assays. The extracellular hydrolytic activity of P. gingivalis proteases was determined as described previously 66,67 . Briefly, the chromogenic p-nitroanilide (pNA) substrates benzoyl-Arg-pNA, acetyl-Lys-pNA, N-Gly-Pro-pNA and N-Ala-Phe-Pro-pNA (all from Bachem) were used to detect RgpA/B, Kgp, dipeptidyl peptidase IV, and prolyl tripeptidyl peptidase A, respectively. Samples were preincubated in buffer in 96-well plates prior to the addition of substrate to a total volume of 200 μ l. The final concentration of each substrate was 1 mM. The rate of substrate hydrolysis, monitored through accumulation of pNA, was followed at 405 nm and the activity of each enzyme was given as mOD/min/μ l or as a % of wild-type activity.
Crystallization and diffraction data collection. Cryosystems 700 series cryostream) native crystal at beam line ID23-2 of ESRF synchrotron (Grenoble, France) using a MAR225 CCD detector. A complete dataset from a selenomethionine-derivatised crystal was similarly collected at the absorption peak wavelength of selenium on a Pilatus 6 M pixel detector (from Dectris) at beam line XALOC of the ALBA synchrotron (Cerdanyola, Barcelona 68 ). Crystals were tetragonal and contained one protein molecule in the asymmetric unit (solvent content 57%, V M = 2.9Å 3 /Da 69 ). Diffraction data were processed with programs XDS 70 and XSCALE 71 , and transformed with XDSCONV to formats suitable for the PHENIX 72 and CCP4 suites 73 of programs. Table 3 provides essential data-processing statistics.
Structure solution and refinement. The structure of PorZ was solved by single-wavelength anomalous diffraction with data collected at the selenium absorption peak wavelength from the selenomethionine-derivatized crystal, applying the AUTOSOL protocol from the PHENIX package 74 . The program found 10 of the 12 theoretical selenium sites plus seven minor sites corresponding to surface ions, as determined a posteriori, which resulted from the crystallization conditions containing zinc and cacodylate (see the previous section). Phasing with all these sites identified P4 3 2 1 2 as the correct enantiomorphic space group and yielded an initial figure of merit of 0.26. Subsequent automatic density modification and model building produced a partial model of 373 (mostly alanine) residues in 72 chains, which showed an overall model-map correlation of 0.34 and an R free value of 0.46. The resulting Fourier map enabled us to perform manual model building with the COOT program 75 , which was assisted by a homology model of PorZ obtained by threading with the LOMETS server (http://zhanglab.ccmb. med.umich.edu/LOMETS 76 ). A more complete partial model was obtained and refined against the diffraction data of the selenomethionine derivative to 3.1 Å resolution with the PHENIX 77 and BUSTER/TNT 78 programs. The refined partial model was subsequently used to solve the native structure with data to 2.9 Å resolution by Fourier synthesis with PHENIX. Given the anisomorphism in cell axes a and b between the native and derivative data (see Table 3), these calculations included rigid-body refinement and simulated annealing, in addition to individual coordinate, translation/libration/screw-motion (TLS) and grouped thermal-displacement-parameter refinement. Thereafter, careful manual model building alternated with two cycles of automatic model building and refinement with the AUTOBUILD routine of PHENIX 79 . Finally, additional manual model building was alternated with regular crystallographic refinement using PHENIX and BUSTER/TNT, including TLS refinement, over several cycles until the final model of PorZ was obtained. This consisted of residues G 29 -R 776 plus a strongly bound calcium ion. In addition, eight zinc ions, four calcium ions, one chloride ion, one cacodylate ion, one tetraethylene glycol, one diethylene glycol and four glycerol molecules arising from the crystallization and cryo-protection conditions were tentatively assigned on the surface of the molecule in addition to 43 unambiguous solvent molecules. See Table 3 for the final refinement and model quality statistics.
Bioinformatics and statistical analyses. Structural similarity searches were performed with DALI 80 , and figures were prepared with the CHIMERA program 81 . Structure superpositions were performed with the SSM routine 82 within COOT. The final structure of PorZ was validated with MOLPROBITY 83 and deposited with the Protein Data Bank (PDB) at www.pdb.org (access code 5M11). Quantitative RT-PCR results were tested for normality distribution, and differences were analysed using Student's t-test with SPSS v.16 software. Differences in enzyme activities and PorZ surface exposure in mutants in comparison to the wild-type parental strain were analyzed with GraphPad Prism 7 software (GraphPad Software, CA, USA) using one-way ANOVA with Bonferroni's correction. P-values below 0.05 were considered significant.