Somatic Serogroups, Capsular Types, and Species of Fecal Klebsiella in Patients with Ankylosing Spondylitis

Abstract

The purpose of the present study was to find out whether patients with ankylosing spondylitis (AS) carry fecal Klebsiella strains that belong to serotypes or species specific for AS. Somatic serotypes (O groups), capsular (K) serotypes, and biochemically identified species were determined for fecal klebsiellae isolated from 187 AS patients and 195 control patients. The controls were patients with fibromyalgia or rheumatoid arthritis. The 638 isolates of Klebsiella that were obtained represented 161 strains; 81 from AS patients and 80 from the controls. The average number of Klebsiella strains per patient was 1.7 for the AS group and 1.5 for the control group. The most common O group was O1, which was observed for isolates from 23 of 187 AS patients and 24 of 195 control patients. Next in frequency was group O2, which was observed for isolates from 17 AS patients and 15 control patients. Regarding the K serotypes, 59 different types were identified, revealing a heterogeneous representation of Klebsiella strains, without a predominance of any serotype. By biochemical identification, Klebsiella pneumoniae was the most frequently occurring species, being found in 45 AS patients and 45 control patients. Next in the frequency was K. oxytoca, which was observed in 26 AS patients and in 29 control patients. K. planticola and K. terrigena occurred in only a minority of patients. Altogether, when analyzed either separately or simultaneously according to O groups, K serotypes, and biochemically identified species, no evidence of the existence of AS-specific Klebsiella strains was obtained. These findings do not indicate participation of Klebsiella in the etiopathogenesis of AS.

The question of whether Klebsiella contributes to the etiopathogenesis of ankylosing spondylitis (AS) has remained unsolved. Pros and cons against the role of Klebsiella are based on numerous studies of fecal carriage, antibody response, and molecular mimicry, and over the years they have been extensively reviewed and debated (2, 6, 11, 20, 21). One must conclude that, so far, no conclusive, indisputable evidence for participation of Klebsiella in the pathogenesis of AS exists. However, an aspect which has not received any attention is the detailed identification, including the serotypes, of the Klebsiella strains isolated from patients with AS. The only related approach has been a study of serum antibodies against Klebsiella capsular (K) antigens that suggested a predominance of serotypes K26, K36, and K50 in patients with AS (21, 22). The significance of bacterial serotypes is evident in infections due to Escherichia coli, but corresponding information is also available for an HLA B27-associated disease, reactive arthritis, which is sometimes even known to evolve into AS (14). For instance, Yersinia enterocolitica serotypes O3 and O9 are causes of reactive arthritis in humans, whereas serotype O8 is not (13). Likewise, dysentery due to Shigella flexneri is often followed by reactive arthritis in HLA B27-positive individuals, whereas this has been reported only rarely for Shigella sonnei (3).

The genus Klebsiella can be divided into five species: K. pneumoniae, K. oxytoca, K. planticola, K. terrigena, and K. ornithinolytica. They all typically express on the cell surface a lipopolysaccharide (LPS; O antigen) and a capsular polysaccharide (K antigen), both of which contribute to pathogenicity. K. pneumoniae and K. oxytoca frequently cause infections, whereas K. planticola, K. terrigena, and K. ornithinolytica are usually nonpathogenic. The somatic (O) antigens of Klebsiella have recently been recognized to divide into nine groups (O1, O2, O2ac, O3, O4, O5, O7, O8, and O12), most of which contain several serotypes (10). O typing of Klebsiella strains has rarely been applied to clinical isolates; during the past 40 years only five studies of O typing of Klebsiella strains have been published (7, 10, 15, 25, 26). In contrast to the small number of O groups, 77 K serotypes are recognized, and their distributions in clinical samples have been widely studied (5, 9).

The present work was undertaken to clarify the potential significance of serotypes and species of Klebsiella in patients with AS. For this purpose, we have analyzed the O groups and K serotypes of different species of fecal klebsiellae isolated from patients with AS and compared the results to those obtained for patients with fibromyalgia (FM) or rheumatoid arthritis (RA).

MATERIALS AND METHODS

Patients.

The study was carried out in two parts. Altogether, 187 patients with AS and 195 control patients were enrolled (Table 1). In part I, 72 patients with AS admitted to the Heinola Rheumatism Foundation Hospital during the period from August to November 1993 were included. The controls were 83 patients with FM enrolled during the same period. In Part II 115 patients with AS admitted to Turku University Central Hospital during the period from November 1995 to March 1997 were included. The controls were 112 patients with RA enrolled during the same period. Each AS patient or control was included only once. Patients with FM or RA were chosen as controls since these diseases are similar to AS regarding the need for hospitalization and the use of anti-inflammatory agents, both of which might affect the intestinal flora. The diseases were diagnosed according to the generally accepted criteria. The HLA B27 status was determined for 137 AS patients, with 113 being HLA B27 positive. Patients who were vegetarians or who had received antibiotics during the preceding 2 months were excluded from the study. Also excluded were those with any intestinal disorders (Crohn’s disease, ulcerative colitis etc.), celiac disease, lactose intolerance, or diabetes mellitus.

TABLE 1.

Patients and controls

Study part and patient group	No. of patients		Mean disease duration (yr)	Age (yr [mean ± SD])
	Total	Females/males
Part I
AS	72	21/51	9.9	42.6 ± 11.2
Controls (FM)	83	72/11	7.7a	46.9 ± 8.3
Part II
AS	115	47/68	14.1	44.1 ± 10.6
Controls (RA)	112	72/40	10.7	54.4 ± 11.2
Both parts combined
AS	187	68/119	12.7	43.5 ± 8.3
Controls	195	144/51	10.6	51.2 ± 5.4

^aReliable information was not available for all patients. 

Isolation of Klebsiella.

Stool samples were collected at the time of hospital admission. For part I of the study they were stored for transportation at −20°C and were thawed later, immediately before culture. For part II of the study the samples were cultured within 2 to 6 h after collection. In both parts of the study the initial cultures were done on MacConkey inositol-carbenicillin agar that was less than 72 h old (4). The medium contains inositol as the selective substrate for the growth of klebsiellae and carbenicillin (10 μg/ml) to prevent the growth of other enterobacteria. Within 20 h of incubation at 37°C, klebsiellae appear as red or pink colonies on the agar surface, indicating fermentation of inositol. After incubation, all (≤10 for each patient) differently looking red or pink colonies were subcultured on lactose-containing agar plates. Identification of the strains was carried out with the API 20 E system (bioMérieux, Marcy-L’Etoile, France). All Klebsiella strains were preserved in Protect tubes (STC, Heywood, England) at −70°C until serological and biochemical characterization. The final biochemical identification was carried out as described in Bergey’s Manual of Determinative Bacteriology (12). A utilization test was done only with histamine, as described by Monnet and Freney (16).

Serotyping.

For each patient with fecal Klebsiella, one strain from those identified as Klebsiella was initially K serotyped. If the following (up to nine) strains from the same patient reacted with the same K antiserum, they were considered to be identical. Those that did not react with the same K antiserum were typed further by the same strategy.

K serotyping was carried out by countercurrent immunoelectrophoresis (CCIE) by using a modification proposed by Palfreyman (19). All 77 known serotypes (K1 through K72, K74, and K79 through K82) were looked for. An extract described by Oerskov and Oerskov (18) instead of a whole-cell suspension was used as the antigen. The extract was heated only once for 1 h at 100°C before centrifugation. All strains with no, weak, or unclear reactions by CCIE were investigated by the classical quellung reaction. O serotyping was done by an inhibition enzyme-linked immunosorbent assay (ELISA) as described by Hansen et al. (10). All currently recognized O groups (O1, O2, O2ac, O3, O4, O5, O7, O8, and O12) were looked for. Strains that did not react in any of the ELISA systems were divided by sodium dodecyl sulfate-polyacrylamide gel electrophoresis into nontypeable strains with smooth LPS and rough (O−) strains without LPS.

RESULTS

In part I of the study fecal Klebsiella strains were isolated from 12 of 72 (17%) AS patients and from 9 of 83 (11%) control patients (Table 2). Among the Klebsiella strains isolated from AS patients, 9 of 18 strains were of group O1; in the controls the corresponding ratio was 3 of 16 (P > 0.05). All other differences, including K serotypes, between strains from AS patients and the controls were even less. Seventeen percent of AS patients and 11% of the controls were carriers of fecal Klebsiella. Due to the low frequency of carriage and to the finding that group O1 strains tended to be more frequent among AS patients, we decided to expand the study to include more AS patients and controls.

TABLE 2.

Rates of Klebsiella isolation and identification

Study part and patient group	No. of patients with fecal Klebsiella/total no. of patients (%)	No. of Klebsiella isolates studied	No. of Klebsiella strains identified
			Total	Per patient
Part I
AS	12/72 (17)	54	18	1.5a
Controls (FM)	9/83 (11)	52	16	1.8
Part II
AS	36/115 (31)	251	63	1.8
Controls (RA)	46/112 (41)	281	64	1.4
Both parts combined
AS	48/187 (26)	305	81	1.7
Controls	55/195 (28)	333	80	1.5

^aMean; only data for patients with fecal Klebsiella are included. 

In part II, 31% of the AS patients and 41% of the controls harbored Klebsiella in the stool (Table 2). Otherwise, no significant differences between results from the two parts of the study were observed, and the results for both parts combined are presented. Likewise, the results were not affected by the sex of the patients, and the findings from the serological and biochemical identifications are presented together regarding both sexes.

With results from both parts of the study combined, 638 isolates of Klebsiella were obtained: 305 from patients with AS and 333 from those with FM or RA. The isolates turned out to represent 161 Klebsiella strains: 81 from AS patients and 80 from control patients. Forty-eight of 187 (26%) AS patients and 55 of 195 (28%) of the control patients harbored fecal Klebsiella. When analyzed according to sex, the corresponding figures were 20 of 68 (29%) for females with AS and 37 of 144 (26%) for female controls. Among the males, 28 of 118 (24%) with AS and 18 of 51 (35%) of the controls harbored fecal Klebsiella. The average number of Klebsiella strains for the patients with fecal Klebsiella was also quite equal in the two study groups, being 1.7 per patient for patients with AS and 1.5 for the controls (P > 0.05) (Table 2). Altogether, 87 patients had only one strain of fecal Klebsiella, 6 patients had two different strains, 4 patients had three different strains, 3 patients had four different strains, 2 patients had five different strains, and one patient had six different strains.

Regarding disease activity, 57 AS patients had an erythrocyte sedimentation rate of >30. In addition, 23 AS patients had clinically active disease, as determined by a physician’s general clinical evaluation. Among these 80 patients with active disease, 20 (25%) had fecal Klebsiella. The corresponding figure for the 107 other AS patients was 25 (23%), indicating that patients with active disease did not harbor fecal Klebsiella more often than those with inactive AS.

The O-group distribution among the fecal Klebsiella isolates has not previously been reported. Among our strains, O1 is the most frequently occurring O group of fecal klebsiellae; this was followed by O2. These findings are in accordance with studies of clinical Klebsiella isolates from other sources (7, 10, 26). Altogether, the distribution of O groups (Table 3) does not reveal any difference between strains from AS patients and the controls when the distribution was analyzed either according to the number of strains or according to the number of patients harboring fecal Klebsiella. Regarding the K serotypes, 59 different types were identified, revealing a heterogeneous representation of Klebsiella strains in the human stool (Table 4). Sixteen of the K serotypes occurred only in AS patients, and 15 occurred only in the controls. In other words, no meaningful difference between the K serotypes of strains from patients with AS and the controls could be observed.

TABLE 3.

Distribution of O groups

O group	No. of strains		No. of patients
	AS	Controls	AS	Controls
O1	33	29	23	24
O2	19	18	17	15
O3	10	12	7	8
O4	2	2	2	2
O5	11	8	9	4
O12	1	1	1	1
O-a	2	1	1	1
NTb	3	9	4	14
Total	81	80	NAc	NA

^aRough strains without LPS. 

^bNT, nontypeable. 

^cNA, nonapplicable, since 16 patients had more than one strain. 

TABLE 4.

Distribution of K serotypes

K serotype	No. of patients
	AS	Controls
1	0	1
2	1	0
3	3	1
6	1	3
8	0	3
8.52.59	0	1
9	1	0
9.81	1	0
12	1	2
14	3	1
15	0	3
16	1	0
17	1	0
18	6	1
19	1	0
20	1	1
21	2	1
22.37	1	1
24	0	1
26	1	0
27	1	2
28	1	1
30	2	2
31	6	6
33	1	1
34	1	0
35	1	2
37	0	1
38	1	0
39	0	1
41	5	0
41.61	0	1
42	0	1
43	0	1
44	2	1
45	0	1
46	0	1
47	3	4
48	0	1
51	1	0
53	1	0
54	1	1
55	1	2
56	1	1
57	0	1
58	2	0
60	2	1
61	1	1
62	1	1
65	1	2
66	1	0
67	0	1
68	1	2
70	2	1
71	1	0
74	2	1
79	2	1
80	2	0
81	0	1
NTa	7	14

^aNT, nontypeable. 

We identified four different Klebsiella species. The most common was K. pneumoniae (55.9%), followed by K. oxytoca (34.2%). These two species are known to be pathogens that cause infections, whereas the other two species observed as a minority of strains, K. planticola (8.7%) and K. terrigena (1.2%), usually do not cause infections. No difference in the distribution of biochemically identified Klebsiella species was observed between strains from AS patients and the controls. Likewise, when analyzed simultaneously according to O groups, K serotypes, and biochemical identification, no evidence for AS-specific Klebsiella strains emerged.

The most frequently occurring strain appeared to be K. pneumoniae of serotype O2:K31, which was harbored by 11 patients, 6 in the AS group and 5 in the control group, all with RA (Table 5). Only 1 of the 11 strains was from part I of the study, which probably explains why this serotype was not observed in FM patients (which were included only in part I). This serotype comprises 6.8% of all the Klebsiella strains identified. Next in frequency was a group of eight strains of K. oxytoca of the O1 group and with a nontypeable K antigen (two from AS patients and six from RA patients).

TABLE 5.

Five most common Klebsiella strains observed

Strain		No. of strains
Species	Serotype	Total	According to disease
			AS	FM	RA
K. pneumoniae	O2:K31	11	6		5
K. oxytoca	O1:K(NT)a	8	2		6
K. pneumoniae	O1:K(NT)a	5	1		4
K. pneumoniae	O1:K44	3	2		1
K. oxytoca	O1:K18	3	2	1

^aThese strains are necessarily not identical even within the species, since the nontypeable (NT) K antigen could represent several different K antigens. 

DISCUSSION

In the present study no serotype or biochemically identified species of Klebsiella specific for AS could be observed. Likewise, the results obtained do not reveal increased levels of excretion or increased rates of carriage of Klebsiella in the patients with AS. An important and new observation is the finding that the distribution of fecal Klebsiella strains seems to be extremely individual; no clear dominance of any type was observed, and almost each person seemed to have been infected with his or her own specific type of Klebsiella strains. This finding includes the fact that 16 K serotypes (22 strains) were observed only in AS patients and not in the controls (Table 4). It remains theoretically feasible that these 16 K serotypes would be AS-specific Klebsiella serotypes. However, 15 other K serotypes (18 strains) were observed only in the controls and not in AS patients. Therefore, no substantiated evidence of a claim that those 16 K serotypes would be AS specific remains, even though serotype-specific K polysaccharides may be important for Klebsiella-macrophage interaction (1). The scattered occurrence of different Klebsiella strains in our material is also compatible with the use of stool samples collected at the time of hospital admission to avoid nosocomially acquired strains. Of particular interest is the detail that of the three K serotypes (K26, K36, and K50) against which Sahly and colleagues (21, 22) found increased antibody levels in AS patients, only K26 was observed in our material, and this was observed only once.

AS is predominantly a disease that occurs in males, and our controls were patients with FM or RA, which more often affect females than males. However, when the results for females and males were analyzed separately, no differences were observed. In general, no evidence that indicates that sex would affect either fecal carriage or the distribution of bacterial types exists, and this was also the case in our study. In contrast to some previous studies and in accordance with some other studies (6, 20), we did not find increased rates of fecal carriage of Klebsiella in AS patients. However, the true nature of such a discrepancy has already been extensively debated and reviewed (6, 20). The possibility that the use of frozen samples in part I of the study would have affected the results of serological or biochemical typing can also be excluded, since, as a whole, no significant difference in this respect between the two parts of the study was evident. In part II of the study, in which freshly cultured stool samples were used, the frequency of Klebsiella carriers was the same as that reported in several other studies of patients with AS (6, 20) or healthy subjects (17). We may only conclude that if AS patients carried a disease-specific Klebsiella serotype or a biochemically identifiable strain, it would have become apparent in the present study.

Finally, even though the present results did not reveal any difference between AS patients and the controls, they do not totally exclude the potential role of Klebsiella in the etiopathogenesis of AS. This conclusion is based on the findings of studies of reactive arthritis following enteric infections caused by Campylobacter, Salmonella, Shigella, or Yersinia (24). At the time of development of arthritis, the diarrhea-causing bacterial triggers are only rarely culturable from stool samples, even though traces of them may be found within the joint tissue; diagnosis of the disease is mostly based on the patient’s history and antibody responses (8, 23). Parallel to this, a possibility exists that an initial process (e.g., enteric colonization) that leads to development of AS occurs and subsides before AS becomes manifest. In such a case, the specific bacterial trigger might already be absent from the feces at the time of fully developed disease. Nevertheless, we must conclude that the results of the present study do not indicate involvement of Klebsiella in the etiopathogenesis of AS. This is based both on the rates of carriage of Klebsiella as well as the prevalence of serotypes, including O groups and K serotypes, and the distribution of Klebsiella species in AS patients and the controls.