The prevailing O serogroups among the serologically differentiated clinical Proteus spp. strains in central Poland

In the years 2006–2011, 617 Proteus spp. strains isolated mostly from urine and wounds or other clinical sources were collected in Łódź, Poland, to determine the offensive O serotypes frequently occurring among patients. P. mirabilis exhibited the most intensive swarming growth and was dominating species (86.9%), followed by P. genomospecies, P. vulgaris, and P. penneri. Ninety four per cent strains were recognized as S (smooth) forms. Serological studies (involving ELISA—enzyme-linked immunosorbent assay and Western blotting using native and adsorbed rabbit antisera) enabled classification of 80% S isolates into respective Proteus O serogroups among the 83 ones, described so far. The remaining strains seemed to be serologically unique. Despite the observed big serological variety of Proteus spp. isolates, we found the O78 serogroup recently described in Poland as dominating and identified other widespread serotypes: O3, O6, O10, O11, O27, O28, and O30 reported earlier as predominating also in other countries; O77 and O79 detected lately in Poland; O16, O18, O20, and O50. No unique structural feature of the prevalent O serotypes has been indicated. However, the prevalence of some O serogroups indicates that particular serotypes may be in some ways beneficial to the strains producing these kinds of O antigen.

Serological investigation. We applied ELISA (enzyme-linked immunosorbent assay) and Western-blotting methods using bacterial biomasses and LPS preparations (extracted from selected studied strains or coming from our collection) as a source of O antigens as well as polyclonal rabbit sera specific to all the O serotypes reported in the genus Proteus (from our collection), as described earlier 15,25 .
Additionally, to confirm the classification of studied strains to 14 most numerous serogroups, we used the respective 14 antisera adsorbed by wet bacterial mass of cross-reacting strains, previously washed in PBS. The biomass was suspended in a serum diluted (1:50) in PBS in a proportion of 1:10 and incubated for 30 min. on ice. The cells were removed from the serum by centrifugation and the process was repeated twice 13 . Then, in ELISA, a lack of the reaction of the adsorbed antiserum with the homologous LPS confirmed the identity of the cross-reacting O antigen with the reference LPS due to removal of all the specific antibodies during adsorption. The stronger reaction observed in ELISA or Western blotting after adsorption, the lower similarity was proved.
In ELISA, LPSs (50 ng per well) or the bacterial biomasses (5 µg per well) were applied for coating the F96 Maxisorp Nunc-Immuno plates (12 h, + 4 °C). After washing the plates twice with distilled water to remove the residual antigens, the uncoated sites in the wells were blocked with casein buffer (2.5% casein, 240 mM of NaOH, 1.5 mM of KH 2 PO 4 , 8 mM of Na 2 HPO 4 , pH 7.2) for 1 h at 37 °C, to avoid subsequent unspecific adsorption of antibodies. The plates were doubly washed with phosphate buffer saline (PBS, 136.8 mM of NaCl, 2.7 mM of KCl, 8 mM of Na 2 HPO 4 , pH 7.2) and incubated with the appropriate polyclonal rabbit antiserum (twofold serially diluted in 50 µl of casein buffer) for 1 h at 37 °C. After triple washing the plates with PBS (to remove the unbound immunoglobulins), the rabbit-IgG specific goat antibodies labelled with horseradish peroxidase (Jackson Immu-noResearch) diluted 1:5.000 in casein buffer were added (50 µl/well) and the plates were incubated as described above. Triple washing with PBS and a single one with substrate buffer (0,1 M of C 6 H 5 Na 3 O 7 , pH 4.5) were applied and the solution of 2,2'-azino-bis(3-ethylbenzothiazoline-6-sulfonic acid) diammonium salt (ABTS) in substrate buffer with the addition of 3% H 2 O 2 (1:25) was used (50 µl/well) as a substrate for peroxidase. After 30 min. of incubation at 37 °C, the colour reaction was stopped by the use of oxalic acid (222 mM). Antibody titers were  Figure 1. The sources of isolation of the whole collection of 617 Proteus spp. strains and separately: P. mirabilis isolates (a), P. penneri isolates (b), P. genomospecies isolates (c), and P. vulgaris isolates (d) [this work and [13][14][15][16][17][18][19][20] ].  (25 mM Tris/HCl, 192 mM of glycine, 20% methanol) on 100 V for 1 h. The membrane was fixed for 5 min. with the solution of 10% isopropanol and 10% acetic acid in water and blocked with 10% skimmed milk in dot-blot buffer (50 mM of Tris/HCl pH 7,4, 200 mM of NaCl) for 1 h. Then, the nitrocellulose was incubated with appropriate rabbit antiserum diluted 1:500 in dot-blot-10% skimmed milk buffer (2 h) and, after washing with the dot-blot buffer, with the rabbit-IgG specific goat antibodies conjugated with alkaline phosphatase (Jackson Immu-noResearch) diluted 1:5000 in the dot-blot-10% skimmed milk buffer (2 h). AP Conjugate Substrate Kit (Bio-Rad) was applied as a substrate of alkaline phosphatase to obtain a colour reaction, which was stopped by washing the membrane in distilled water. Western blotting reactions were scanned by means of Canon Toolbox 4.9.

Results
Physiological features. Among the 617 studied strains, urine isolates and wound isolates (including bacteria from pus, bedsores, fistulas, skin and wound swabs, and abscesses) were dominating. Other frequent sources of isolation were connected with the respiratory tract (bronchoaspirates, sputum, and the pharyngonasal cavity), vagina, and faeces, while strains from other sources were isolated rarely ( Fig. 1). It was also shown that the eleven Proteus spp. strains (nine P. mirabilis, one P. vulgaris, and one P. genomospecies 5/6 including four strains reported earlier 13,14,20 ) were inhabiting the intestines of 5.8% from 189 individuals (carriers).
The strains species were identified on the basis of the metabolic features of the isolates (Table 1). P. mirabilis species was definitely predominant, while the contribution of the other detected Proteus species was much lower (Table 1). Totally, 41 isolates (6.6%) were found to belong to P. genomospecies 4 or 5/6, the two last are still impossible to distinguish on the grounds of their metabolic properties 4,5 , while no isolate was classified to P. hauseri species. Thus, it can be assumed that P. genomospecies are isolated from clinical sources with a frequency comparable to that observed for P. vulgaris and all the isolates belonging to the so called P. vulgaris group together account for 11.7% of the whole collection, where P. genomospecies 4 is the rarest. Contrary to P. mirabilis and P. penneri (Fig. 1a,b), among P. vulgaris and P. genomospecies the isolates from wounds were found to dominate, while the isolates from urine were less common (Fig. 1c,d).
To select the isolates which can be typed to O serogroups (S forms), we conducted S-R tests. Thirty seven studied strains (6% of the collection) which were totally or partly non-thermal stable and agglutinated in 0.85% NaCl were recovered as R or S/R forms possessing LPS deficient in O polysaccharide or LPS with only one O-repeating unit, respectively. R forms constituted only 4.9% of P. mirabilis strains (26 out of all collected P. mirabilis isolates), while in P. vulgaris it was 9.7% (three out of 31 P. vulgaris isolates), in P. genomospecies 4 and 5/6-14.6% (six out of 41 isolates), and in P. penneri-22.2% (two out of nine isolates). As the shortened LPS without O-polysaccharide part does not guarantee a proper and stable truncation of the flagella, the swarming growth of R forms is usually decreased or stopped. As expected, the majority of R strains were not able to swarm on the surface of solid Luria Broth medium. However, among the smooth strains, the non-swarming ones were also detected. Most of P. mirabilis isolates (more than 90%) were found to be able to swarm and 70% exhibited intensive swarming growth (on the distance > 30 mm within 24 h). In the case of P. penneri, non-swarming strains constituted 22% (R forms only) while the other isolates were able to swarm effectively. The weakest swarming

Serological features.
A total of the 580 S strains (possessing an O antigen in the LPS molecules) were serologically investigated to classify them into proper O serogroups in the genus Proteus. The results for 24 of these strains classified to O2, O8, O11, O77-O82 serogroups, respectively, had been reported previously [13][14][15][16][17][18][19][20] . Biomasses of the S strains were studied in ELISA with rabbit polyclonal antisera specific to all the Proteus O serogroups described so far. Strong cross reactions comparable to the ones in the homologous systems (mostly to the titre 1:128000 or more) were the first indication on the serological type of the studied strains. But the obtained results were not always so clear.
Several biomasses were not recognized by any antisera (the O antigens present on the cells were possible to be masked by other cell wall components) and some cross reacted with more than one antiserum achieving various titres (other surface antigens may have also been recognized by the antibodies present in the used rabbit antisera). In these cases, LPS preparations were extracted from the biomasses by the phenol-water procedure 24 . Purified LPS antigens coated on the ELISA plates might be better exposed and better accessible to specific antibodies. Indeed, in consequence the O serotype of several such strains was identified.
These results allowed recognizing the similarity of most of the isolates (447/556 S forms studied in this work) to particular O serotypes among 83 Proteus O serogroups recognized so far, while the remaining strains seemed to be serologically unique and were considered to represent new O serogroups in the genus Proteus. One hundred and fifty six isolates had been preliminarily classified into 47 O serogroups (data not shown) formed by fewer than 10 studied strains. More than a half (291) of the studied here 556 S Proteus spp. clinical strains were initially considered as belonging to 14 predominating O serogroups consisting of 10 or more strains: O3, O6, O10, O11, O16, O18, O20, O27, O28, O30, O50, O60, O77, and O78.
To confirm that consideration, the sera specific to the 14 most prevalent O serotypes were adsorbed by the cross-reacting biomasses and their reactions with the homologous LPSs were analyzed in ELISA. In most cases, the reactions of the antisera in the homologous systems were totally abolished due to the removal of all LPSspecific antibodies by the cross-reacting biomasses, indicating their antigenic identity and confirming the initial serological classification (Table 2). However, in some cases the adsorption was not complete and some remaining fractions of antibodies still reacted with homologous LPSs, although the reactions were obviously weaker than the ones of the native antisera (Table 2). In these 34 cases, some differences in the strains serospecificity were responsible for not full adsorption of the antisera. The antibodies which could not be removed from the adsorbed antisera should be specific to some epitopes, which are present in the reference homologous LPSs (core region or O antigen) but absent in the cross-reacting studied strains. Thus, such reactions indicated more or less antigenic similarity but not the identity of the studied strains and the reference LPSs.
The nature of the cross reactions, observed in ELISA, was visualized in the Western-blotting method using LPS preparations obtained by proteinase-K treated biomasses as antigens (Figs. 2, 3, 4). The reactions are similar and comparable to the ones observed for the LPS preparations obtained using phenol-water extraction, which is demonstrated on the example of strains classified to O77 serogroup (Fig. 2a,b). By the degradation of proteins, www.nature.com/scientificreports/ we could remove possible non-specific cross reactions caused by protein antigens rather than by O antigens 26 . Indeed, the initially observed in ELISA reactions of several biomasses not fully adsorbing the respective antisera, could have been caused by protein antigens, as they were not confirmed in Western blots. That was the case of four strains (Fig. 3b, paths 2-5) preliminarily classified into O10 serogroup, two strains (Fig. 3c, paths 8, 9) classified into O28 serogroup, and one strain (Fig. 3d, path 8) classified into O50 serogroup, which displayed completely no reaction after proteinase-K treatment in Western blots. The Western-blotting technique also allowed localizing LPS epitopes responsible for the strong cross reactions initially observed in ELISA. The epitopes may have been present in the O-specific part and/or in the core region of the LPSs exposed on the biomasses of the studied strains. Indeed, among the strains preliminarily classified into O10 serogroup on the basis of strong reactions of their biomasses in ELISA, one isolate showed in Western blot reactions only in the fast-migrating fractions corresponding to the lipid A-core region, while there were no reactions observed for the slowly migrating LPS bands additionally possessing O-polysaccharide chains (Fig. 3b,  path 8). Similar results were obtained in the case of as many as four strains (Fig. 3a, paths 2, 6, 7, and 9) initially classified into O3 serogroup. The O3 and O10 antisera are visibly rich in antibodies recognizing core-localized common epitopes and responsible for strong cross reactions observed both in ELISA and in Western blots. Obviously, these five strains cannot be included in O3 or O10 serogroups, although they may form R serogroups.
Noteworthy, the 15 strains initially classified into the most numerous O78, O16, and O11 serogroups but not fully adsorbing the respective O78, O16, or O11 antisera ( Table 2) displayed in Western blots visible strong reactions of the slowly-migrating LPS molecules possessing O polysaccharides (O-PS), which is showed on the example of the strains (2-7) classified to O78 serogroup (Fig. 4a, paths 2-7). The use of the O78, O16, or O11 antisera adsorbed by the respective biomasses in the Western-blotting technique proved that the observed serological differences refer mainly to the O-specific parts of these strains' LPS. As the antisera contain small amounts of anti-core antibodies, the possible differences in core antigens were hard to be detected. The adsorbed O78, O16, and O11 antisera did not react with the respective adsorbing biomasses, which confirmed the properly conducted adsorption processes (control), but still in these three antisera there were remaining some fractions www.nature.com/scientificreports/ of antibodies recognizing the epitopes located in the homologous O antigens (slowly-migrating LPS bands) but absent in the studied strains. The Western-blotting results for these 15 strains were similar and they are shown on the example of O78 antiserum adsorbed with strain 2 (Fig. 4b). Thus, the 15 strains were recognized as representing different subgroups within O78, O16, or O11 serogroups, respectively.

Discussion
In the years 2006-2011 a total of 617 Proteus spp. strains were collected in Łódź, Poland ( Fig. 1) including the 24 clinical strains reported earlier [13][14][15][16][17][18][19][20] . What is worth noticing, no strains were collected from blood, although Proteus spp. bacteria may also cause severe bacteremia and sepsis 2,27 . Urine and wound isolates were prevailing (Fig. 1). Frequent isolation of P. mirabilis (Table 1) is not surprising, as the species is regarded as the most virulent and widespread species in the genus Proteus, especially in patients suffering from nosocomial urinary tract and wound infections and it is still the most frequently isolated species among Proteus spp. 2 . P. genomospecies 4, 5, 6, or P. hauseri are not usually distinguished from P. vulgaris in the routine laboratory work and so there are very few reports available on their occurrence in patients. Similarly to our results, O'Hara et al. 4 and Janda et al. 5 studying several P. hauseri, P. genomospecies 4, 5, and 6 strains, mostly from human sources, found P. genomospecies 5 and 6 to be the most numerous. Our results confirm the low frequency of P. penneri isolation from clinical sources 28,29 . Recently, Dai et al. 30 have proposed a new molecular MLSA (multilocus sequence analysis) technique for identification of Proteus species, suggesting a reclassification of P. cibarius and P. genomospecies 5 into the formerly described species P. terrae isolated from soil. Among 210 isolates from clinical specimens and food in China, the authors found P. terrae as being the second most abundant species after P. mirabilis, when P. vulgaris and P. penneri as the third and fourth ones, respectively. R forms which are regarded as less virulent, we found among the isolates belonging to all the identified Proteus species. However, in the most pathogenic P. mirabilis species, the smallest number of R forms was detected and the best swarming ability was noticed which is considered to be an important virulence factor. P. genomospecies and P. vulgaris isolates demonstrated weaker swarming growth in comparison to P. penneri strains, similarly to the results achieved by Kwil et al. 31 .
Serotyping is an important way of typing and classification of bacteria, especially in diagnostics and epidemiology of some pathogenic genera like Salmonella. In the case of this genus, besides the classical phenotypic www.nature.com/scientificreports/ methods new precise molecular tools are recommended, such as whole genome sequencing (WGS)-based serotype prediction and microarray-based method 32 . WGS was also applied in studies of a P. mirabilis strain virulence 33 . Many O-antigen gene clusters were analyzed and O serotype-specific suspension array for detecting selected Proteus serotypes has been proposed 34 . However, the prevalence and significance of particular serogroups in patients have still not been obvious. Serological studies of Polish strains confirmed that the studied strains possess various core types which is typical in the genus Proteus. Frequently, strains forming a common R serogroup do not belong to a common O serogroup and the strains classified into one O serogroup may differ in their core-region serospecificity 6,11 . Our studies showed a big serological diversity among Proteus spp. strains isolated recently from patients in central Poland. However, half of the isolates (299) including 18 strains reported before [13][14][15]19 belong to 15 most numerous O serogroups, comprising ten or more studied strains (Fig. 5). O78 serogroup is the biggest one, formed by 61 studied strains (10.5% of 580 collected S strains). This serogroup together with O77 and O79 serogroups were described as new ones not long ago by Drzewiecka et al. 13,14 and Arbatsky et al. 15 and they currently include only the isolates from Poland. O50 serogroup, in turn, was reported in 2003 35 and has so far been represented by P. mirabilis strain TG 332 described earlier as being serologically unique 36 . O60 serogroup was also formed in 2003 37 for a non-clinical P. myxofaciens isolate from a gypsy larva. Lately, the species has been proposed to be excluded from the genus Proteus 38 . However, our results clearly indicate that O60 serotype exists among P. mirabilis and P. genomospecies clinical strains. The other 10 most prevalent serogroups have been included in the first classic serological scheme proposed for P. mirabilis and P. vulgaris bacteria 9 . What is interesting, the majority of these predominant serogroups had been previously reported as frequently found among hundreds of clinical P. mirabilis and P. vulgaris isolates from urine, faeces, blood, or unknown sources. Summarizing these data, Larsson 10 indicated O3 serogroup as dominating in all reports and O10, O13, O26, O28, and O30 serogroups as the most prevalent ones. O6, O11, O23, O24, O27, and O29 serogroups were also widely distributed. Analyzing the serological properties of 99 Swedish and 24 Polish P. mirabilis strains from urinary tract infections, Kaca et al. 12 also found O10 and O30 to be the most numerous serogroups. However, the authors applied only 20 Proteus O-specific sera in their studies so the attachment of the strains to the other prevalent serogroups (e.g. O6, O26, O28, or O29) was not analyzed. In our studies, some of the prevalent serogroups turned out to be multi-species (Table 3), including not only the ubiquitous P. mirabilis but also the other species. Three P. penneri isolates were classified into O10 serogroup (so far found mostly in O17, O61, O64, and O65 serogroups 11,39 including also P. mirabilis and/or P. vulgaris strains 7 ), several P. genomospecies strains were classified into O11, O60, O78, and O79 serogroups and two P. vulgaris isolates into O27 and O78 serogroups. In all the prevalent serogroups, the isolates from urine and/or wounds were most numerous and we could not see any relations between serotypes and sources of isolation.
O78, O16, and O11 serogroups seem to be heterogenous. Further chemical analysis may show the differences in the structures of O antigens, responsible for their slight serological variety observed within these serogroups and will allow the formation of new subserogroups or new cross-reacting O serogroups. www.nature.com/scientificreports/ It can be seen that for many years, Proteus spp. strains belonging to O3, O6, O10, O11, O27, O28, and O30 serogroups have been most frequently isolated in many different countries. What is worth noticing, the Proteus O60 antigen structure resembles that of the Proteus O30 antigen 7 so before the O60 serogroup was created in 2003 37 , O60 isolates could be classified to O30 serogroup due to their mutual similarity and strong cross reactivity revealed in our work (data not shown). It should be noted that at present, O78 serogroup obviously seems to be the most prevalent among Polish patients from the Łódź region accounting for 10.5% of all S isolates, although its predominance is not big. The prevalence of the serotypes dominating over others may be connected with some features of their O antigens. However, it is difficult to indicate common structural O-PS features (Table 3) distinguishing the biggest serogroups from the smaller ones, e.g. the compounds present exclusively in the frequent O antigens and, simultaneously, absent in the rarely found O serotypes 7 . Among the 15 most numerous serotypes, both branched and linear O-PS repeating units can be found, which are built of three-to-five sugars, including glucose, galactose, and their derivatives frequently occurring in Proteus O-PSs. For example, the repeating unit of the most ubiquitous O78 antigen is composed of five different sugars and sugar derivatives. Three of them constitute the main chain and two form side branches. This antigen is not characterized by any unusual or unique structural feature. It shares disaccharide fragments with several Proteus O antigens, but no cross reactions have been noticed 14 . It is possible that the most prevalent Proteus O serotypes may be especially connected with some virulence factors, like it is observed e.g. in Vibrio cholerae O1 40 or with drug resistance, which is the case of P. mirabilis O78 (unpublished data).

Conclusion
In general, clinical Proteus spp. isolates from central Poland are characterized by high serological diversity and include the representatives of almost all described O serogroups. These results may suggest that either Proteus spp. strains do not intensively transmit from one patient to another or they are able to rapidly change their antigenic specificity by avoiding the immune response from host organisms, increasing their own virulence and adapting to environmental conditions. Such examples of antigenic variation are known in other Gram-negative genera and may be caused either by mutations in O antigen genes or by inter/intra-species lateral transfer of O antigen genes by homologous recombination, IS elements, plasmids, or bacteriophages 41   www.nature.com/scientificreports/ the high serological diversity among Proteus bacilli, the results of the presented research confirm the persisting for many years predominance of some serogroups reported earlier 10 and reveal new prevalent O serotypes (especially O78). The presence of the representatives of different Proteus species among these serogroups suggests that the bacteria acquire these special O phenotypes as a result of adaptation in convergence processes. Thus, these O antigens may be in some ways attractive and beneficial to the expressing bacteria. It will be reasonable to investigate a clonal homology of one-species isolates belonging to the respective predominating O serogroups. Their diversity would confirm the special role of a particular O-antigen type in their survival and virulence. The appointment of the most virulent O serotypes among the Proteus spp. bacilli would allow developing methods, which might enable rapid detection and eradication of such strains and the prophylaxis of recurrent complicated UTIs and CAUTIs.