The Prevailing O Serogroups Among the Serologically Differentiated Clinical Proteus Spp. Strains in Poland

In the years 2006-2011, 617 Proteus spp. strains isolated mostly from urine and wounds or other clinical sources from infected individuals were collected in Łódź, Poland. P. mirabilis species was dominating (86.9%), followed by P. genomospecies, P. vulgaris , and P. penneri . Ninety four per cent strains were recognized as S (smooth) forms. P. mirabilis exhibited the most intensive swarming growth. Serological studies (involving ELISA – enzyme-linked immunosorbent assay and Western blotting using native and adsorbed rabbit antisera) enabled classication of 80% S isolates into respective Proteus O serogroups among the 83 ones, described so far. The remaining strains seemed to be serologically unique. Detailed serological analysis conrmed that the recently described in Poland O78 serogroup is the most widespread. Also, 14 other serogroups have been found to predominate, being represented by ten or more strains. No unique structural feature of the prevalent O serotypes has been indicated. The observed big serological variety may suggest a low rate of Proteus spp. strains’ transmission between patients or rapid changes in their lipopolysaccharides. However, the prevalence of some O serogroups indicates that particular serotypes may be in some ways benecial to the strains producing these kinds of O antigen.


Introduction
Proteus spp. bacteria are peritrichously agellated members of Morganellaceae family distinguished from the other representatives of the order Enterobacterales by intensive and spectacular swarming growth on solid media as a result of the multicellular differentiation phenomenon [1,2]. These bacilli are detected in natural environments (water or soil) and in many species of wild and domestic animals. They have also been found to be a component of natural faecal micro ora in a part of human population. On the other hand, Proteus spp. are opportunistic pathogens, which may affect mainly immunocompromised individuals and cause infections of the urinary tract and wounds, bacteraemia, abscesses in many organs or other infections [3]. Among the pathogenic species, P. mirabilis is most common followed by less frequently isolated P. vulgaris, P. penneri, P. hauseri, and three genomospecies without names (P. genomospecies 4, 5, and 6). Lack of biochemical reactions which would reliably distinguish between different P. genomospecies resulted in limited knowledge on their characteristics and isolation frequency [4,5], when compared to well-described P. mirabilis. Urinary tract infections (UTIs) of P. mirabilis aetiology are frequently catheter-associated -CAUTIs, recurrent and dangerous due to the frequent and severe complications like pyelonephritis and urolithiasis [ 2].
Among numerous virulence factors of Proteus bacteria, the lipopolisaccharide (LPS) should be emphasized as an endotoxin and an important antigen determining the serological speci city of Proteus spp. strains. The LPS core region in the genus Proteus is structurally and serologically diverse, which has led to the formation of the R typing Proteus scheme containing 11 R serotypes [6]. However, mainly the highly immunogenic O-polysaccharide region determines the serological speci city of LPS S forms, which has become a basis for the classi cation of Proteus strains into 83 O serogroups [7,8]. The rst O type scheme for P. mirabilis and P. vulgaris strains including 49 O serogroups was founded by Kauffmann [9]. Later, Larsson [10] indicated among them O3, O6, O10, O11, O13, O23, O24, O26, O27, O28, O29, and O30 as the most prevalent serogroups in many countries. The Kauffmann's scheme was further extended with other O serogroups containing all Proteus species [7] but only a few P. penneri and P. mirabilis strains come from Poland [11,12]. Gathering a new wide collection of more than 600 Proteus spp. clinical isolates from the Łódź region (central Poland) enabled us to study their serological differentiation. Up to date, serologically unique strains from the collection have formed six new serogroups O77-O82 [13][14][15][16][17][18] and two new subgroups in O8 and O11 serogroups [19,20]. In this work the serological variety of the whole collection of Proteus spp. strains is shown focusing on Proteus O serogroups most widespread in central Poland.

Bacterial strains, LPSs, physiological tests
The studied clinical Proteus spp. strains from different sources ( Fig. 1) were kindly provided in 2006-2011 by four big medical laboratories in Łódź (Poland): in Barlicki Hospital, in Biegański Hospital, "Synevo", "Diagnostyka". Some strains were isolated from faeces of 189 volunteers using McConkey and Nogrady media plates and recognized as forming the urease-positive and lactose-negative colonies, as described [20]. All isolates were identi ed by the cultivation on the media proposed by Senior [21], modi ed as described [13], additional tests were applied for the strains recognized as P. genomospecies, as described [4,15] (Table 1). The strains were retained in stocks (Luria Broth cultures with 25% glycerol) at -80°C.
S-R tests included both the thermal stability in the boiled broth medium test and the pseudoagglutination test [22]. Additionally, the swarming growth ability on Luria Broth medium with 1.5% agar was studied, according to Wilkerson and Niederhoffer [23].
Biomasses of the studied strains were obtained by 18-h cultivation on a nutrient broth medium supplemented with 0.2% glucose, with aeration, at 37°C. The strains were killed by 1% phenol, centrifuged, washed with distilled water and stored in +4°C (wet biomass) or lyophilized (dry biomass).
LPSs were extracted from strains' dry biomass using the classic phenol-water method, published by Westphal and Jann [24].

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 speci c to all the O serotypes reported in the genus Proteus (from our collection), as described earlier [15,25]. In ELISA 50 ng of LPSs or 5 µg of the bacterial biomasses per well were applied for coating the F96 Maxisorp Nunc-Immuno plates. 2,2'-Azinobis(3-ethylbenzothiazoline-6-sulfonic acid) diammonium salt (ABTS) was used as a substrate for peroxidase conjugated with the rabbit-IgG speci c goat antibodies (Jackson ImmunoResearch). Antibody titers were determined by measuring the absorbance (A 405 ) using Multiskan Go microplate reader (Thermo Scienti c). Additionally, we used some antisera adsorbed by wet bacterial masses, according to the method by Drzewiecka et al. [13]. In Western blotting, antigens (bacterial biomass or LPS, 5-8 µg per lane) were separated during sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS PAGE).
Then, the samples were transferred to nitrocellulose (Whatman Schleicher & Schuel). The sera were diluted 1:500 in dot-blot-10% skimmed milk buffer. AP Conjugate Substrate Kit (Bio-Rad) was applied as a substrate of alkaline phosphatase conjugated with the rabbit-IgG speci c goat antibodies (Jackson ImmunoResearch).

Physiological features
Six hundred and seventeen strains were isolated in the Łódź city, central Poland ( Fig. 1), in that number 606 ones from infected individuals. Urine isolates and wound isolates (including bacteria from pus, bedsores, stulas, 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), faeces, and vagina, while strains from other sources were isolated rarely. Moreover, the bacteria ( ve P. mirabilis strains) were also detected in faecal samples taken from only ve out of 181 healthy volunteers (2.8%) and in faecal samples from as many as six out of eight infected individuals (four P. mirabilis, one P. vulgaris, and one P. genomospecies 5/6 including four strains reported earlier [13,14,20]). Thus, it was shown that the 11 Proteus spp. strains were inhabiting the intestines of 5.8% studied volunteers.
The strains species were identi ed on the basis of the metabolic features of the isolates (Table 1). P. mirabilis species was de nitely predominant, while the contribution of 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 classi ed to P. hauseri species. Thus, 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. 1 a,b), among P. vulgaris and P. genomospecies the isolates from wounds were found to dominate, while the isolates from urine were less common ( Fig. 1 c,d).  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 de cient 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 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, nonswarming strains constituted 22% (R forms only) while the other isolates were able to swarm effectively. The weakest swarming growth ability was observed among P. vulgaris and P. genomospecies isolates as 50 % of strains have not showed this ability or could swarm very weakly (<10 mm within 24 h).
Biomasses of the S strains were studied in ELISA with rabbit polyclonal antisera speci c 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 rst 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]. Puri ed LPS antigens coated on the ELISA plates might be better exposed and better accessible to speci c antibodies. Indeed, in consequence the O serotype of several such strains was identi ed.
These results allowed to recognize 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 fty six isolates have been preliminarily classi ed into 47 O serogroups (datat 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 con rm that consideration, the sera speci c 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 LPS-speci c antibodies by the cross-reacting biomasses, indicating their antigenic identity and con rming the initial serological classi cation (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 serospeci city were responsible for not full adsorption of the antisera. extraction, which is demonstrated on the example of strains classi ed to O77 serogroup (Fig. 2a,b). By the degradation of proteins, we could remove possible non-speci c 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 con rmed in Western blots. That was the case of four strains (Fig. 3b, paths 2 -5) preliminarily classi ed into O10 serogroup, two strains (Fig. 3c, paths 8, 9) classi ed into O28 serogroup, and one strain (Fig. 3d, path 8) classi ed 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-speci c part and/or in the core region of the LPSs exposed on the biomasses of the studied strains. Indeed, among the strains preliminarily classi ed 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 classi ed 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 ve strains cannot be included in O3 or O10 serogroups, although they may form R serogroups.
Noteworthy, the 15 strains initially classi ed 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) classi ed 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-speci c 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 con rmed the properly conducted adsorption processes (control), but still in these three antisera there were remaining some fractions 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 especially among the ubiquitous P. mirabilis. 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. In comparison, 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 result con rm the low frequency of P. penneri isolation from clinical sources [28,29]. R forms which are regarded as less virulent, were found among the isolates belonging to all the identi ed 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. The feature of swarming is considered to be an important virulence factor of P. mirabilis bacteria and may be a form of their adaptation to the host environment (especially urinary tract). Moreover, P. mirabilis swarming cells facilitate an entry into the urinary tract of other non-motile species, which often leads to following polymicrobial UTIs [30]. P. vulgaris isolates demonstrated weaker swarming growth in comparison to P. penneri strains, similarly to the results achieved by Kwil et al. [30].
Serological studies con rmed 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 classi ed into one O serogroup may differ in their core-region serospeci city [6,11]. Thus, the use of antisera which are rich in core-speci c antibodies is informative, but in the studies aimed at classifying Proteus spp. strains into proper O serogroups may be misleading. [19]. A good solution might be the use in the preliminary classi cation studies of the antisera deprived of core-speci c antibodies removed by adsorption, which is hard to obtain, or the application of further detailed tests allowing a precise serological classi cation of strains on the basis of their O antigens serospeci city, as presented in this work. Among the studied strains, 13 isolates preliminarily classi ed into the respective predominating O serogroups were further excluded on the basis of subsequent detailed studies.
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 [32] and has so far been represented by P. mirabilis strain TG 332 described earlier as being serologically unique [33]. O60 serogroup was also formed in 2003 [34] for a non-clinical P. myxofaciens isolate from a gypsy larva. Lately, the species has been proposed to be excluded from the genus Proteus [35]. 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 rst 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-speci c sera in their studies so the attachment of the strains to the other prevalent serogroups (e.g. O6, O26, O28, or O29) was not analysed. Research on P. penneri strains revealed that clinical strains and isolates from faeces of healthy people in the USA, Great Britain, Canada, France, Germany, and single strains from other countries belonged mostly to O17, O61, O64, and O65 serogroups [11,36] including also P. mirabilis and/or P. vulgaris strains [7]. In the presented studies three P. penneri isolates were classi ed into O10 serogroup. Several P. genomospecies strains were classi ed into O11, O60, O78, and O79 serogroups and two P. vulgaris isolates into O27 and O78 serogroups. Thus, 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.
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.
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 [34], O60 isolates could be classi ed 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.