Genomic Variation of Bartonella henselae Strains Detected in Lymph Nodes of Patients with Cat Scratch Disease

Abstract

Bartonella henselae is the primary agent of cat scratch disease (CSD). In order to study the genetic variation of B. henselae and the correlation of the various genotypes with epidemiological and clinical findings, two seminested, groEL- and pap31-based PCR assays were carried out with specimens from 273 patients. Amplicons were sequenced to determine the genotype of the causative Bartonella species. Compared to our reference intergenic spacer region-based PCR, the groEL- and pap31-based assays were 1.7 and 1.9 times more sensitive, respectively. All 107 positive patients were infected with B. henselae; neither Bartonella clarridgeiae nor other species were detected. Based on the groEL and pap31 sequences, B. henselae amplicons were classified into two genogroups, Marseille and Houston-1, and into four variants, Marseille, CAL-1, Houston-1, and a new variant, ZF-1. Patients infected with either one or the other genogroup did not exhibit different epidemiological or clinical characteristics. Our study highlights the genotypic heterogeneity of B. henselae in patients with CSD.

Cat scratch disease (CSD) is most frequently reported in children and young adults, in whom it usually presents as a benign, self-limited lymphadenopathy in lymph nodes draining the site of a cat scratch or bite (13). The first presumed agent of CSD was isolated in 1988 by English and colleagues and was named Afipia felis (11). In 1993, Dolan and colleagues isolated Bartonella henselae from the lymph nodes of patients with CSD for the first time (18). Since then, many data from epidemiological, serological, and molecular biology-based studies (2, 7, 17) have demonstrated the role of B. henselae rather than A. felis as the main causative agent of CSD. B. henselae exhibits an important heterogeneity; two serotypes, Houston-1 and Marseille, have been described (19), and these correspond to two genotypes based on 16S rRNA sequences, genotypes I and II, respectively (8). To date, the respective pathogenicity spectra of serotypes Marseille and Houston-1 have not been established. In addition, two other Bartonella species, Bartonella quintana and Bartonella clarridgeiae, have been suspected as alternative agents of CSD (4, 20, 34, 38, 47, 58).

Serological analysis by immunofluorescence or enzyme-linked immunosorbent assay is a useful tool for the diagnosis of B. henselae infections (5, 15, 46). However, the specificity of serological analysis has been questioned due to the cross-reactivity between B. henselae and other species including B. quintana, Coxiella burnetii (21, 36, 55), and Chlamydia species (41). Culture has not proved to be very useful for the diagnosis of CSD because of the fastidious nature of Bartonella species (37, 42, 43). PCR-based detection of Bartonella species coupled with nucleotide sequencing has also been used for the diagnosis of CSD. In our laboratory, we have been using primers derived from gltA (31), rpoB (50), and the 16S-23S intergenic spacer region (ITS) (51) for the detection of Bartonella DNA from human specimens. Unfortunately, none of these genes was adequate for the subtyping of B. henselae variants. Moreover, conventional PCR may be impaired by the inactivation of many DNA polymerases by immunoglobulins in patients' lymph nodes (1).

Recently, we demonstrated that sequences from the groEL and pap31 genes are suitable for the identification and subtyping of B. henselae genotypes, respectively (unpublished data).

The purpose of our work was to increase the sensitivity of the PCR-based detection of Bartonella DNA from lymph node biopsy specimens by using primers that allow the amplification of all known Bartonella species pathogenic for humans and, thus, study the involvement of Bartonella species in CSD. We designed two seminested PCR assays, one that amplifies the groEL gene, which results in specific sequence signatures for each of the seven currently known Bartonella species pathogenic for humans, and another that amplifies the pap31 gene, for which sequences present a signature region allowing the differentiation of B. henselae variants. We tested 289 lymph node biopsy specimens from 273 patients with a clinical diagnosis of CSD. We compared these results to those obtained by the ITS-based PCR. We also compared the epidemiological and clinical aspects of CSD according to the infecting B. henselae genogroup.

MATERIALS AND METHODS

Study design.

Patients were clinically suspected of having CSD on the basis of findings such as the presence of chronic lymphadenopathy without a specific diagnosis and contacts with cats (22). Bartonella infection was diagnosed on the basis of the results of direct identification of a Bartonella sp. by PCR from lymph node tissue. Samples were collected from January 1996 to January 2001. Two hundred eighty-nine lymph node biopsy specimens or aspirates from 273 patients clinically suspected of having CSD were sent to the Unité des Rickettsies to be tested for the presence of Bartonella spp. A standardized questionnaire was completed for each CSD patient. Items completed on the questionnaire included (22) whether the patient had had contact with cats or cat fleas; the presence of fever, cat scratches or bites, or fleabites; the presence of a cutaneous lesion at the inoculation site; the previous administration of antibiotic therapy; and the outcome within 1 month following the diagnosis of CSD.

Molecular biology-based methods. (i) DNA extraction and PCR amplification.

Total genomic DNA was extracted from samples with a QIAamp Tissue kit (QIAGEN, Hilden, Germany), as described by the manufacturer. A total of 10 to 25 mg of tissue or 200 μl of aspirate was used. Samples were handled under sterile conditions to avoid the risk of cross-contamination. A total of 125 μl of elution buffer was used to resuspend the DNA. Genomic DNAs were stored at 4°C until their use as templates in PCR assays.

The primers used for amplification and sequencing are presented in Table 1. Primers were selected by using Primer3 software (52) and were purchased from Eurobio (Les Ulis, France). Primers positions were numbered relative to the ITS of B. henselae strain Houston-1 (GenBank accession number L35101), the sequence of the groEL gene of Bartonella bacilliformis (GenBank accession number Z15160), and the sequence of the pap31 gene of B. henselae strain Houston-1 (GenBank accession number AF001275). groEL-derived PCR primers were chosen to amplify a variable fragment that allowed the identification of the seven currently recognized pathogenic Bartonella species, B. henselae strains Houston-1 and Marseille, B. quintana, Bartonella elizabethae, B. clarridgeiae, Bartonella grahamii, B. bacilliformis, and Bartonella vinsonii subsp. berkhoffii, according to the sequences at many different positions, especially the specific nucleotide signatures at positions 1279 to 1280 and 1281 to 1286 (green in Fig. 1a). pap31-derived primers allowed both the detection and the subtyping of B. henselae variants (Fig. 1b). Many specific nucleotide signatures were found in this region (colored positions in Fig. 1b).

TABLE 1.

Primers used for PCR and sequencing

Primera	Primer sequence	Position (direction)b	Target gene
PAPn1 (ap, s)	5′-TTCTAGGAGTTGAAACCGAT-3′	438-457 (→)	pap31
PAPn2 (ap, s)	5′-GAAACACCACCAGCAACATA-3′	695-714 (←)	pap31
PAPns2 (ap, s)	5′-GCACCAGACCATTTTTCCTT-3′	629-648 (←)	pap31
PAPns1 (s)	5′-CAGAGAAGACGCAAAAACCT-3′	472-491 (→)	pap31
HSPps1 (ap, s)	5′-CAGAAGTTGAAGTGAAAGAAAA-3′	1154-1173 (→)	groEL
HSPps2 (ap, s)	5′-GCNGCTTCTTCACCNGCATT-3′	1366-1385 (←)	groEL
HSPps4 (ap, s)	5′-GCTGGNGGTGTTGCNGTTA-3′	1117-1135 (→)	groEL
HSPps3 (s)	5′-GCTGTNGAAGANGGNATTGT-3′	1216-1235 (→)	groEL

^aap, amplification primer; s, sequencing primer.

^bPrimer positions are numbered relative to the sequence of the groEL gene of B. bacilliformis (accession number Z15160) and to the pap31 gene of B. henselae strain Houston-1 (accession number AF001274). Arrows indicate the directions of the primers (→, forward; ←, reverse).

FIG. 1.

Multiple-sequence alignment of partial groEL sequences of B. henselae Houston-1 (BhHo), B. henselae Marseille (BhMa), B. quintana (Bqui), B. vinsonii subsp. berkhoffii (Bvbe), B. elizabethae (Beli), B. clarridgeiae (bcla), B. grahamii (bgra), and B. bacilliformis (bbac) (a) and partial pap31 sequences of seven B. henselae variants (b).

A PCR protocol for amplification of the Bartonella ITS DNA fragment was carried out with these specimens as described previously (51). Seminested PCR protocols for amplification of the Bartonella groEL and pap31 genes with three primers at the same time (primers HSPps1, HSPps2, and HSPps4 and primers PAPn1, PAPn2 and PAPns2, respectively) were applied to the clinical samples. In each case the PCR was carried out with PTC-200 automated thermocyclers (MJ Research, Waltham, Mass.) with a Taq DNA Polymerase kit (Gibco-BRL, Cergy Pontoise, France). The 25-μl reaction mixture consisted of primers (12.5 pmol each), MgCl₂ (final concentration, 1.6 mM), deoxynucleoside triphosphates (dATP, dCTP, dGTP, and dTTP; concentrations, 2 mM each), 10× buffer, Taq DNA polymerase enzyme (0.03 U), sterile water, and 10 μl of a DNA sample. PCR amplification was performed under the following conditions: an initial 3 min of denaturation at 94°C was followed by 44 cycles of denaturation for 30 s at 94°C, annealing for 30 s at 58°C for both genes, and extension for 45 s at 72°C. The amplification was completed by holding the reaction mixture at 72°C for 7 min to allow complete extension of the PCR products. The PCR products were separated by electrophoresis on 1.5% agarose gels and visualized by staining with ethidium bromide and were then purified with the QIAquick PCR purification kit (QIAGEN), as proposed by the manufacturer. The products were stored at 4°C until they were further processed. Measures were taken to prevent PCR carryover contamination. Each PCR step was performed in a different room. All reactions included negative controls (Mycobacterium tuberculosis-infected lymph nodes from patients) in a proportion of one negative control for every seven samples.

(ii) Sequencing reactions.

PCR products were sequenced in both directions with a d-Rhodamine Terminator Cycle Sequencing Ready Reaction kit (Perkin-Elmer, Coignieres, France), as described by the manufacturer. Sequencing products were resolved with an ABI 377 or an ABI 310 automated sequencer (Perkin-Elmer).

(iii) Analysis of sequences and construction of phylogenetic trees.

Sequence analysis was performed with the ABI Prism DNA Sequencing Analysis Software package (version 3.0; Perkin-Elmer) and CLUSTAL W software (version 1.8) (60). The sequences were aligned with the groEL and pap31 sequences available in GenBank for other Bartonella species and subspecies. The partial pap31 sequences of five B. henselae isolates (Table 2) were previously determined in our laboratory (unpublished data) and allowed the classification of these isolates into two groups: the Houston-1 group (red in Fig. 1b), also referred to as genotype I (8), and the Marseille group (green in Fig. 1b), also referred to as genotype II.

TABLE 2.

Bacterial strains and sequences used in this study^a

Species (strain)	Source		pap31		groEL
	Collection no.	Reference	GenBank accession no.	Reference	GenBank accession no.	Reference
Bartonella henselae (Houston-1T)	ATCC 49882	48	AF001274	9	AF014829	39
Bartonella henselae Marseille (URLLY 8T)	CIP 104756	19	AF308169	UD	AF304019	UD
Bartonella henselae (CAL-1)	CDC, septicemia		AF308166	UD	AF304020	UD
Bartonella henselae (Fizz)		19	AF308167	UD	AF304022	UD
Bartonella henselae (90-615)	CDC	63	AF306165	UD	AF304023	UD
Bartonella henselae (SA2)	CDC	18	AF308168	UD	AF304021	UD
Bartonella henselae (ZF-1)			AF321116	This study
Bartonella bacilliformisT					ZI5160	UD
Bartonella clarridgeiaeT					AF014831	39
Bartonella grahamiiT					AF014833	39
Bartonella elizabethaeT					AF014834	39
Bartonella quintanaT					AF014830	39
Bartonella vinsonii subsp. berkhoffiiT					AF014836	39

^aAbbreviations: ATCC, American Type Culture Collection, Manassas, Va.; CIP, Collection de l'Institut Pasteur; CDC, Centers for Disease Control and Prevention, Atlanta, Ga.; UD, unpublished data.

Phylogenetic trees were generated from multiple-sequence alignments by using computer algorithms implemented in the PHYLIP software package (25). Trees were obtained by the maximum parsimony method (DNAPARS software in the PHYLIP package) and distance methods (DNADIST [distance matrix with the Kimura two-parameter method] and NEIGHBOR [neighbor joining and the unweighted pair group method with arithmetic means in the PHYLIP software package]). Bootstrap replicates were performed to estimate the reliabilities of the nodes in the phylogenetic trees obtained by both methods (12). The bootstrap values were obtained from 100 trees (23) generated randomly with SEQBOOT and CONSENSE programs in the PHYLIP software package. Only values above 90 were considered significant. Trees were drawn with the TreeView program (version 1.5) (45).

Statistical analysis.

The Mantel-Haenszel method with Epi Info (version 6.0) was used to compare proportions, and Fisher's exact test was used to compare different groups of patients. Observed differences were considered significant when P was <0.05 for two-tailed tests.

The sensitivity was calculated for each of the three PCR assays, and the sensitivities were compared with each other.

Nucleotide sequence accession numbers.

The accession numbers for the sequences obtained in the present study are reported in Table 2.

RESULTS

Molecular biology-based assay results.

Of the 289 lymph node specimens tested, amplicons were obtained from 114 samples from 108 patients by combining the results from the three PCR assays.

ITS amplification and sequencing.

Amplicons were obtained from 57 specimens (19.7%) from 56 patients by the ITS-based PCR, 49 of which were also groEL positive and 56 of which were also pap31 positive (Fig. 2). One specimen was amplified only by the ITS-based PCR. None of the negative controls were positive. The sequences derived from these PCR products were 100% identical to the B. henselae sequence for all 57 specimens. Therefore, these sequences did not allow us to distinguish between B. henselae genotypes.

FIG. 2.

Comparison of results of the PCR assays carried out with ITS-, groEL-, and pap31-derived primers with lymph node specimens from 273 patients. The numbers represent the numbers of specimens.

groEL seminested PCR and sequence variation.

PCR amplification of part of the groEL gene resulted in two bands. The PCR products generated consisted of 231- and 269-bp fragments, corresponding to the amplifications obtained by PCR with the HSPps4-HSPps2 and HSPps1-HSPps2 primer pairs, respectively. None of the negative controls were positive. Amplification of the groEL sequence resulted in the identification of B. henselae from 104 (36%) specimens from 98 patients, all of which were also pap31 positive and 49 of which were also ITS positive (Fig. 2). Only B. henselae could be detected. Additionally, sequences from this amplified region also allowed us to subtype B. henselae amplicons into two clusters according to the sequence variations seen at position 1348 (red in Fig. 1a). Samples from two of the six patients were positive; the amplicons from both samples had the same sequence. Of the sequences from the 98 patients, 60 (61.2%) exhibited 100% sequence identity with the sequence of B. henselae genotype Marseille and 38 (38.8%) exhibited 100% sequence identity with the sequence of B. henselae genotype Houston-1.

pap31 seminested PCR, sequence variation, and phylogeny.

PCR amplification of part of the pap31 gene resulted in two bands. The PCR products generated consisted of 209- and 275-bp fragments, corresponding to the amplification products obtained by PCR with the PAPn1-PAPns2 and PAPn1-PAPn2 primer pairs, respectively. None of the negative controls were positive. The sequences deduced by the pap31-based PCR identified only B. henselae DNA in 113 specimens from 107 patients. Samples from two of the six patients were positive; the amplicons from both samples had the same sequence. Among the 107 patients, 64 (59.8%) were infected with the Marseille genogroup and 43 (48.2%) were infected with the Houston-1 genogroup. Within these two main genogroups, four different pap31 sequences were identified: the CAL-1 genotype in 63 patients (58.8%), the Marseille genotype in 1 patient (0.9%), the Houston-1 genotype in 3 patients (2.8%), and a previously undescribed pap31 genotype, which we called the ZF-1 genotype, in 40 patients (38.1%). The CAL-1, Marseille, Houston-1, and ZF-1 genotypes were differentiated on the basis of specific pap31 signatures at 22 positions (Fig. 1b). For one patient, only the ITS-based PCR was positive, but an amplicon could not be obtained by either the pap31-based PCR or the groEL-based PCR; therefore, the precise genotype could not be determined.

Phylogenetic analysis by the maximum parsimony and neighbor-joining methods provided trees with similar organizations when pap31 sequences were used (Fig. 3). B. henselae isolates clustered into two genogroups. The first, which we called Marseille, contained B. henselae strains Fizz, CAL-1, and Marseille. The second group, Houston-1, included B. henselae strains SA-2, 90-615, Houston-1, and ZF-1. The two clusters were statistically supported by elevated bootstrap values (100%) by both methods. Elevated bootstrap values were also obtained within both clusters.

FIG. 3.

Neighbor-joining unrooted tree (obtained by using the distance matrix with the Kimura two-parameter method and neighbor-joining analysis with the PHYLIP software package) generated with pap31-based sequencing data for seven B. henselae variants. The numbers at the nodes indicate the percentages of occurrence of the branching order in 100 bootstrapped trees for the neighbor-joining trees. Scale bar, 0.5% divergence.

Comparative sensitivities of the three PCR assays.

All three PCR assays were conducted under the same conditions and so were subjected to the same effects of inhibitory elements. Overall, 114 lymph node specimens were positive by at least one of the three tests (Fig. 2). When compared to each other, the sensitivity of the ITS assay was 0.50, that of the groEL assay was 0.91, and that of the pap31 assay was 0.99.

Patients' characteristics.

The mean ± standard deviation age of the 108 patients with proven B. henselae lymphadenopathy was 25.6 ± 22.7 years (range, 4 to 82 years). Forty-three patients (39.8%) were younger than 20 years, including 23 (21.3%) who were younger than 10 years. Sixty-eight patients (63.0%) were male. All patients had contact with cats, but only 82 (75.9%) reported scratches, 34 (31.5%) reported bites, and 20 (18.5%) reported fleabites. None was immunocompromised. Forty-nine (45.3%) patients developed a lesion at the inoculation site, and 35 (32.4%) had a temperature of >38.5°C. The infected lymph nodes were the axillary lymph nodes in 58 patients (53.7%), the inguinal lymph nodes in 33 patients (30.5%), the cervical lymph nodes in 13 patients (12.0%), and the popliteal lymph nodes in 3 patients (2.8%). None of the patients developed any visceral complication. Compared to the 64 patients with CSD caused by B. henselae genogroup Marseille, the 43 patients with CSD caused by B. henselae genogroup Houston-1 were not significantly different epidemiologically or clinically (Table 3). We also compared patients within each group infected with one or the other genotype but observed no significant differences.

TABLE 3.

Comparison of epidemiological and clinical characteristics of patients infected with B. henselae genogroup Marseille and genogroup Houston-1 and those infected with B. henselae genotype CAL-1, Marseille, ZF-1, and Houston-1

Variable	B. henselae		P value	B. henselae genogroup Marseille			B. henselae genogroup Houston-1
	Genogroup Marseille	Genogroup Houston-1		Genotype CAL-1	Genotype Marseille	P value	Genotype ZF-1	Genotype Houston-1	P value
No. of patients	64	43		63	1		40	3
Sex ratio (male/female)	1.78	1.52	0.70	1.74	1.0	0.64	1.5	2.0	0.66
Mean ± SD age	24.2 ± 23.4	25.9 ± 23.3	0.71	24.6 ± 23.7	11.0a	NAb	27.0 ± 23.9	15.0 ± 14.8	0.73
No. of patients with:
Cat scratches	52	30	0.17	51	1	0.70	28	2	0.67
Cat bites	21	13	0.78	21	0	0.67	11	2	0.21
Cat flea bites	13	7	0.60	13	0	0.79	6	1	0.42
Inoculation lesion	29	20	0.90	28	1	0.45	19	1	0.55
Fever	19	16	0.41	19	0	0.74	14	2	0.30
Usual affected lymph node location (axillary + inguinal)	55	36	0.75	54	1	0.86	33	3	0.58
Unusual affected lymph node location (cervical + popliteal)	9	7	0.75	9	0	0.86	7	0	0.58
Axillary lymph nodes affected	35	23	0.90	34	1	0.54	22	1	0.44
Inguinal lymph nodes affected	20	13	0.91	20	0	0.68	11	2	0.21
Cervical lymph nodes affected	8	5	0.89	8	0	0.87	5	0	0.68
Popliteal lymph nodes affected	1	2	0.35	1	0	0.98	2	0	0.86

^aStandard deviation not applicable.

^bNA, not applicable.

DISCUSSION

PCR-based detection methods have been widely used for the diagnosis of CSD (Table 4). We compared the DNA detection efficacy of one conventional, ITS-based PCR assay, previously considered in our laboratory as the best tool for the detection of Bartonella DNA from clinical specimens, and those of two seminested, groEL- and pap31-based PCR assays. All three tests were carried out with 289 lymph node samples from 273 patients with a clinical diagnosis of CSD, which represents the largest reported series of lymph node samples from patients with a diagnosis of CSD tested by PCR (Table 5).

TABLE 4.

Literature summary of the geographic distributions of Bartonella species and variants isolated from humans and cats by PCR-based methods

Geographic origin	Origin	No. (%) of specimens						Reference
		Totala	B. henselae	B. henselae Houston-1	B. henselae Marseille	Other B. henselae variants	B. clarridgeiae
North America	Cats	2	2	0	2 (100)	0	0	8
Switzerland	Patients	34	34	9 (26.5)	25 (73.5)	0	NDb	10
	Cats	19	19	0	19 (100)	0	0	10
Germany	Patients	39	39	23 (59)	9 (23)	7 (18)	ND	54
	Cats	17	17	1 (5.8)	16 (94.2)	0	ND	57
The Netherlands	Patients	41	41	32 (78)	7 (17)	2 (5)	ND	8
	Cats	25e	21	7 (28)	14 (56)	0	9 (36.0)	6
France	Patients	18	18	5 (27.8)	13 (72.2)	0	0	16
		98c	98	38 (38.8)	60 (61.2)	0	0	Present study
		107d	107	3 (2.8)	1 (0.9)	63 (58.9 %) for CAL-1, 40 (37.4 %) for ZF-1	0	Present study
	Cats	50	35	17 (34.0)	18 (36)	0	15 (30.0)	30
		7e	9	41	5	0	5 (71.0)	28
		23	16	11 (68.7)	4 (25)	1 (6.3)	7 (30.4)	La Scola et al.f

^aOnly PCR-positive amplicons are considered.

^bND, undetermined value because the primers used in the study did not allow the amplification of the bacterial target gene.

^cBy the groEL-based assay.

^dBy the pap31-based assay.

^eCoinfection of cats with many Bartonella species and variants.

^fB. La Scola et al., unpublished data.

TABLE 5.

Literature summary of the main PCR results obtained with different partial gene fragments for patients with suspected CSD

Target genea	Product size (bp)	Geographic origin	No. of patients	No. (%) of patients positiveb	Reference
16S rRNA	280	The Netherlands	41	32 (78)	8
16S rRNA	153	United States	42	27 (64)	59
16S rRNA	153	United States	6	2 (33)	35
htrA	414	United States	25	21 (84)	2
htrA	414	United States	13	7 (53)	44
16S rRNA	460	Israel	32	32 (100)	3, 26
htrA	414	Israel	32	22 (68)	3
gltA	379	Israel	32	30 (93)	3
16S rRNA	280	The Netherlands	226	167 (73)	7
16S rRNA	990	France	68	42 (61)	16
rpoB	825	France	94	21 (22)	50
htrA	414	Germany	60	26 (43)	54
16S rRNA	296	Germany	60	36 (60)	54
ITS	723	France	273	56 (20.5)	Present study
pap31	269			107 (39.2)
groEL	275			98 (35.9)

^ahtrA, 60-kDa heat shock protein gene; groEL, 60-kDa heat shock protein gene; gltA, citrate synthase gene; rpoB, RNA polymerase beta subunit gene; pap31, B. henselae phage 60457 Pap31 protein gene.

^bAll positive specimens were identified as B. henselae.

The importance of the groEL gene as a sensitive and valuable tool for the phylogenetic analysis (39, 61) and identification (27) of bacterial species has recently been highlighted. We selected primers that allowed the amplification and identification, following sequencing, of all seven Bartonella species currently recognized to be pathogenic for humans (Fig. 1a). We also used pap31-derived primers. The pap31 gene encodes a major protein associated with a phage isolated from B. henselae and is probably implicated in its pathogenesis (9). Previously, this gene allowed the classification of B. henselae strains into two clusters, B. henselae Houston-1 and B. henselae Marseille, on the basis of 22 nucleotide differences (Fig. 1b) (unpublished data). To the best of our knowledge, our study is the first to use a seminested PCR assay. Compared to the conventional, ITS-based PCR assay used in our laboratory for the diagnosis of CSD, our seminested assays with the groEL and the pap31 genes increased the numbers of positive samples by 82 and 98%, respectively. Therefore, while our new seminested detection method exhibits a specificity of 100%, it was nearly twice as sensitive as the conventional, ITS-based PCR for the detection of Bartonella DNA in specimens from patients with CSD. When the assays were compared to each other, the seminested, pap31-based assay was 8% more sensitive than the seminested, groEL-based assay. The sensitivity of the PCR assay may be influenced by the fact that the sequences of the pap31- and groEL-derived primers have been chosen from conserved regions of DNA and the fact that the sequences of the primers used for ITS amplification have been chosen from a variable region of DNA. However, one ITS-positive amplicon was not detected by the two seminested PCR assays; this may be due to either the presence of a different B. henselae genotype not amplified by any of two detection assays or DNA damage because of long-term storage of the sample. Compared to the largest (7) of the previously published studies of B. henselae DNA detection (Table 5), the proportions of positive samples by our PCR assays (i.e., 36.0 and 39.1% by the groEL- and pap31-based assays, respectively) may seem low, but in the previous study, PCR was performed with specimens from patients with a diagnosis of CSD that had already been confirmed by a skin test, whereas in our series, patients were only clinically suspected of having CSD.

In contrast to ITS sequences, which were discriminant only to the species level, the groEL sequences classified positive patients as being infected with B. henselae genogroup Houston-1 (38.8%) or genogroup Marseille (61.2%); and pap31 sequences allowed a similar but more precise subtyping of B. henselae variants, as four B. henselae genotypes were identified. When groEL sequences were used, 60 patients were found to be infected with B. henselae genogroup Marseille and 38 were found to be infected with genogroup Houston-1. When pap31 sequences were used, 1 patient exhibited a genotype Marseille infection, 63 patients exhibited genotype CAL-1 infections, 40 patients exhibited genotype ZF-1 (a previously undescribed genotype) infections, and 3 exhibited genotype Houston-1 infections (Fig. 1b). The great heterogeneity among B. henselae variants confirms the interest in the pap31 gene for the subtyping of this bacterium.

B. henselae has been described as the main agent of CSD, but two additional Bartonella species, B. quintana and B. clarridgeiae, have also been proposed as possible agents of the disease. The role of B. quintana as an agent of CSD appears to be limited, as it has been isolated only from lymph node biopsy specimens of two immunocompromised adults in contact with cats and/or cat fleas (20, 47) and was recently reported in a 10-month-old girl with a chronic lymphadenopathy (4). However, although both patients presented with B. quintana-induced lymphadenopathies, their epidemiological and clinical characteristics were not the same as those of patients with typical CSD. Recently, we isolated both B. quintana and M. tuberculosis from a human immunodeficiency virus-positive patient who had no contact with cats and who presented with lymphadenitis (unpublished data). Moreover, B. quintana is the agent of rare cases of chronic lymphadenopathy (47). However, as observed in patients with bacillary angiomatosis, B. quintana is less likely than B. henselae to cause adenopathies, and when B. quintana was present, the infection did not proceed to suppuration (33). Therefore, the role of B. quintana in CSD and other lymphadenopathies should be further investigated.

The role of B. clarridgeiae as an agent of CSD has been suspected on the basis of the fact that cats are the reservoir of both B. clarridgeiae and B. henselae (30, 32, 34, 40, 50) and that coinfection with both species has been reported (14, 28). As B. clarridgeiae represents up to 36% of Bartonella isolates from cats (6, 30), one would expect humans to be frequently exposed to this bacterium. The demonstration of the pathogenic role of B. clarridgeiae for humans is currently based on indirect evidence obtained either by PCR-restriction fragment length polymorphism analysis of an isolate from the cat of a patient with CSD (34) or by serology for a patient with CSD (38). Recently, Sander et al. (58) detected antiflagellin (anti-FlaA) antibodies to B. clarridgeiae in 3.9% of 724 patients suffering from lymphadenopathy. However, among the previously published studies, only two reported the use of PCR primers able to detect B. clarridgeiae (16, 51). In our study, and despite the use of groEL-derived primers that allowed the amplification of B. clarridgeiae, we did not find this bacterium among 98 positive patients. Moreover, in the study of Sander et al. (58), no bacterial agent was reported in cultures of samples from serologically positive patients, and the only PCR carried out with a sample from a B. clarridgeiae anti-FlaA-positive patient detected and identified B. henselae but not B. clarridgeiae. None of the other diagnostic studies conducted so far could detect B. clarridgeiae from patients with CSD either by culture or by PCR (2, 7, 8, 29, 49, 54, 57). Therefore, if B. clarridgeiae is pathogenic for humans, it is probably much less virulent than B. henselae.

Two main pathogenic B. henselae variants, Houston-1 and Marseille, have been identified in both animals and humans, and both of these variants are involved in CSD (8, 19, 24, 28, 53, 54, 56, 57). Sander et al. (54) reported on the detection of three B. henselae variants in human lymph node specimens by PCR: Houston-1 in 59% of patients, Marseille in 23% of patients, and a third variant in 18% of patients. In our study, we identified two B. henselae genogroups using the groEL sequences and four genotypes using the pap31 sequences; among the 107 patients infected with the Marseille genogroup, 0.9% were infected with the Marseille genotype and 58.9% were infected with the CAL-1 genotype, whereas among the patients infected with the Houston-1 genogroup, 2.8% of the patients were infected with the Houston-1 genotype and 37.4% of the patients were infected with the ZF-1 genotype. This is the first description of the ZF-1 genotype, which represented the major genotype among isolates within the Houston-1 genogroup in our study. Therefore, the genotypic variability of B. henselae is greater than that observed previously and may explain the absence of cross-protection between both variants in cats (28). Other investigators have studied the geographic distributions of variants Houston-1 and Marseille detected in human lymph node specimens (Table 4). In Germany and The Netherlands, the Houston-1 variant is more frequent than the Marseille variant (8, 54), whereas in France (16) and Switzerland (10), the Marseille variant is the most common. As cats are the source of both B. henselae variants, it may be speculated that the prevalence in cats reflects that in humans. However, when trying to correlate the distribution of B. henselae variants in humans and cats (Table 4), discrepancies were noted in the literature. In Switzerland, both cat and human B. henselae isolates are mainly the Marseille variant (10) (Table 4). In The Netherlands and Germany, the majority of B. henselae isolates from cats are the Marseille variant, whereas isolates from humans are mostly identified as B. henselae Houston-1 (6, 8, 54, 57) (Table 4). In France, the opposite results have been reported: humans are mainly infected with B. henselae Marseille, whereas most cats are infected with the Houston-1 variant (16, 28) (Table 4).

On the basis of the distributions of the B. henselae variants observed in humans and cats in Germany, Sander et al. (54) have proposed that B. henselae Houston-1 may be more pathogenic for humans than the Marseille variant. As described above, this assertion is not supported by the results of Swiss and French studies (Table 4). Moreover, in our study, when we compared the epidemiological and clinical characteristics of patients infected with either one or the other of the two main B. henselae genogroups, Marseille and Houston-1, we found that our data were in accordance with the findings usually observed for patients with CSD (62). However, we observed no statistically supported differences between the two genogroups and, within each genogroup, between genotypes; these results do not favor differences in the pathogenicities of the two genotypes. We suggest that further epidemiological studies with B. henselae variants should take the four genotypes described above into consideration and should investigate whether these various genotypes are correlated with specific serotypes, human pathologies, or distributions in cats.

In this study we have reported on a seminested PCR-based detection and identification method which allowed us to increase the sensitivity of detection and identification of causative agents of CSD directly with lymph node specimens by use of the groEL and pap31 genes. Although our seminested, pap31-based PCR was the most sensitive detection tool, it failed to detect B. henselae in a patient found to be positive by the ITS-based PCR. Therefore, we recommend the use of a combination of both the ITS- and the pap31-based PCR methods for the diagnosis of CSD from lymph node specimens. By the pap31 gene-based PCR, the B. henselae organisms that caused CSD were classified into two genogroups, Marseille and Houston-1, and four genotypes, mainly CAL-1 and ZF-1, but also Marseille and Houston-1. The pap31 gene should be considered a valuable tool for the subtyping of B. henselae.

Further studies with patients with CSD may clarify whether the various pathogenic genotypes of B. henselae have specific geographic distributions and/or virulences and may identify additional variants.