Laboratory Diagnosis of 37 Cases of Bartonella Endocarditis Based on Enzyme Immunoassay and Real-Time PCR

ABSTRACT

Bartonella spp., mostly Bartonella quintana and B. henselae, are a common cause of culture-negative endocarditis. Serology using immunofluorescence assay (IFA) and PCR performed on cardiac tissues are the mainstays of diagnosis. We developed an enzyme immunoassay (EIA) and a novel multiplex real-time PCR assay, utilizing Bartonella genus-specific, B. henselae-specific, and B. quintana-specific SimpleProbe probes, for diagnosis of Bartonella endocarditis. We aimed to evaluate the performance of these assays. Thirty-seven patients with definite endocarditis, 18 with B. henselae, 18 with B. quintana, and 1 with B. koehlerae, were studied. Diagnosis was confirmed by conventional PCR and DNA sequencing of surgical cardiac specimens. Similar to the case with IFA, anti-Bartonella IgG titers of ≥1:800 were found in 94% of patients by EIA; cross-reactivity between B. henselae and B. quintana precluded species-specific serodiagnosis, and frequent (41%) but low-titer cross-reactivity between Coxiella burnetii antibodies and B. henselae antigen was found in patients with Q fever endocarditis. Low-titer (1:100) cross-reactivity was uncommonly found also in patients with brucellosis and culture-positive endocarditis, particularly Enterococcus faecalis endocarditis. Real-time PCR performed on explanted heart valves/vegetations was in complete agreement with results of sequence-based diagnosis with characteristic melting curves. The genus-specific probe identified five additional endocarditis-associated Bartonella spp. at the genus level. In conclusion, EIA coupled with a novel real-time PCR assay can play an important role in Bartonella endocarditis diagnosis and expand the diagnostic arsenal at the disposal of the clinical microbiologist. Since serology remains a major diagnostic tool, recognizing its pitfalls is essential to avoid incorrect diagnosis.

INTRODUCTION

Following Q fever, Bartonella spp. are the second most common cause of culture-negative endocarditis, accounting for up to 28% of cases (1–3). Most reported cases have been attributed to Bartonella quintana, followed by B. henselae, while other Bartonella spp. have been implicated in single cases of endocarditis, including B. elizabethae (4), Bartonella vinsonii subsp. berkhoffii (1, 5, 6), Bartonella vinsonii subsp. arupensis (7), B. koehlerae (8), B. alsatica (1, 9, 10), and “Candidatus Bartonella mayotimonensis” (11). Bartonella spp. are fastidious bacteria with notoriously poor sensitivity of axenic cultures performed on clinical specimens of human blood or tissue. Laboratory diagnosis of Bartonella endocarditis is therefore based mainly on serological assays and PCR performed on infected tissues. Microscopic immunofluorescence assay (IFA) is the most commonly used serological method utilized for diagnosis of Bartonella infections, although other methods, such as enzyme immunoassay (EIA), are used as well (12–16). Compared with EIA, IFA is laborious and time-consuming, and interpretation is subjective. A limitation of all serological assays for the diagnosis of Bartonella infection is cross-reactivity between B. henselae and B. quintana. Bartonella IFA has also been reported to have cross-reactivity with Coxiella burnetii and Chlamydia pneumoniae (17–19). The Bernard Pridan Laboratory for Molecular Biology of Infectious Disease at Tel Aviv Sourasky Medical Center serves as the national reference laboratory for human Bartonella infections in Israel and has developed over the past 30 years serological and molecular tools for the diagnosis of Bartonella infection, including an EIA and various conventional and real-time PCR assays. Although we have occasionally used these assays for non-henselae Bartonella-associated infections, they have been developed and utilized primarily for the diagnosis of cat scratch disease, which is caused predominantly by B. henselae (12, 14, 15, 20–24). A systematic evaluation of the performance of our assays for the diagnosis of Bartonella endocarditis has not been conducted thus far. The aim of the current study was to evaluate an EIA and molecular assays, including a novel multiplex real-time PCR assay, for the diagnosis of Bartonella endocarditis.

MATERIALS AND METHODS

Patients and clinical samples.

As part of a surveillance study of Bartonella infections ongoing in Israel since 1991, all clinical specimens from patients with suspected Bartonella endocarditis, including serum samples and tissue for PCR assays and cultures, are referred to a single laboratory at Tel Aviv Sourasky Medical Center, thus allowing identification of essentially all laboratory-confirmed cases of Bartonella endocarditis in Israel. The referring hospital-based physicians are requested to fill out a questionnaire providing data on patients’ demographic and clinical features and are contacted to obtain additional information. For the purpose of the current study, we included patients with culture-negative endocarditis who met the following two criteria: (i) they have undergone valve surgery for endocarditis-associated valve dysfunction complicated by heart failure and/or the presence of intracardiac abscess, and (ii) Bartonella was identified in the valvular/paravalvular tissue or vegetation by PCR and DNA sequencing. The study included 37 patients with Bartonella endocarditis diagnosed in hospitals throughout Israel in 2001 to 2020: 18 patients with B. henselae, 18 with B. quintana, and 1 with B. koehlerae. All patients had definite endocarditis based on at least two major criteria of the Duke score: (i) echocardiogram positive for endocarditis as defined in the modified Duke criteria (25) and (ii) Bartonella-positive PCR from infected cardiac tissue corroborated by DNA sequencing, that is, if one accepts Bartonella-positive PCR as a finding with equal significance to a major criterion of the modified Duke criteria (25, 26). Patients with Bartonella endocarditis without PCR-based diagnosis for species identification were excluded from the study. The study was approved by the institutional review board (Helsinki Committee) of the Tel Aviv Sourasky Medical Center.

Bacterial strains and cultures.

Bacterial strains used in this study are described in Table 1, including 7 Bartonella spp. that have been implicated in human endocarditis and non-Bartonella bacterial species that have been associated with culture-negative endocarditis (27). All bacterial strains were obtained from known reference collections (American Type Culture Collection [ATCC], USA, National Collection of Type Cultures [NCTC], United Kingdom, Collection Nationale de Cultures de Microorganismes [CNCM], Institute Pasteur, France, and Collection of the Institute Pasteur, France [CIP]). Exceptions included Nine Mile phase I Coxiella burnetii DNA (Vircell, Spain), B. henselae isolate BhTA-4, cultured in our laboratory from lymph node biopsy specimen of a patient with typical cat scratch disease, and B. koehlerae isolate C-508, cultured in our laboratory from a bacteremic cat and found to have ribC and gltA DNA sequences identical to those of B. koehlerae amplified from a valve of an endocarditis patient reported by our group (8). Bartonella spp. and cardiac tissue specimens, whenever sufficient material was available, were cultured on chocolate agar plates as previously described (12). Bartonella spp. were identified by PCR targeting various genes with amplicon sequencing (Table 2).

TABLE 1.

Results of the real-time PCR assay performed on Bartonella spp. and other bacterial species associated with culture-negative endocarditis

Bacterial species/subsp.	Strain designation(s)b	Probe Tma (°C) or status
		Bartonella genus	Bartonella henselae	Bartonella quintana
B. henselae	BhTA-4, ATCC 49793 (87-66)	58–61	47–51	Negative
B. quintana	CIP 103739 (Toulouse), CIP 107027 (Fuller)	57–61	Negative	68–71
B. koehlerae	ATCC 700693	57–59	Negative	Negative
B. vinsonii subsp. berkhoffii	ATCC 51672	57	Negative	Negative
B. vinsonii subsp. arupensis	ATCC BAA-2250	58	Negative	Negative
B. elizabethae	ATCC 49927	43	Negative	Negative
B. washoensis	ATCC BAA-2254	50	Negative	Negative
Coxiella burnetii	ATCC VR-615 (Nine Mile phase I)	Negative	Negative	Negative
Tropheryma whipplei	CNCM I-2202T (Twist-Marseille)	Negative	Negative	Negative
Legionella pneumophila subsp. pneumophila	ATCC 33152	Negative	Negative	Negative
Aggregatibacter aphrophilus	ATCC 33389	Negative	Negative	Negative
Aggregatibacter paraphrophilus	NCTC 10557	Negative	Negative	Negative
Aggregatibacter actinomycetemcomitans	ATCC 700685	Negative	Negative	Negative
Aggregatibacter segnis	NCTC 10977	Negative	Negative	Negative
Cardiobacterium hominis	NCTC 10426	Negative	Negative	Negative
Eikenella corrodens	ATCC 23834	Negative	Negative	Negative
Kingella kingae	NCTC 10529	Negative	Negative	Negative
Abiotrophia defectiva	ATCC 49176	Negative	Negative	Negative
Granulicatella adiacens	ATCC 49175	Negative	Negative	Negative
Granulicatella elegans	ATCC 700633	Negative	Negative	Negative
Gemella bergeriae	ATCC 700627	Negative	Negative	Negative
Gemella haemolysans	ATCC 10379	Negative	Negative	Negative
Gemella morbillorum	NCTC 11323	Negative	Negative	Negative
Gemella sanguinis	ATCC 700632	Negative	Negative	Negative

^aT_m, melting temperature.

^bATCC, American Type Culture Collection, USA; CIP, Collection of the Institute Pasteur, France; CNCM, Collection Nationale de Cultures de Microorganismes, Institut Pasteur, France; NCTC, National Collection of Type Cultures, Central Public Health Laboratory, United Kingdom.

TABLE 2.

Results of the real-time PCR assay (expressed as melting temperatures) and the enzyme immunoassay (expressed as the reciprocal titers) among 37 patients with Bartonella endocarditis

Patient no.	Age (yrs)/sexa	Molecular diagnosisb	Conventional PCR target gene(s)	Real-time PCR Tm (°C)c			Maximal reciprocal titer (anti-Bartonella IgG antibodies)
				Tm-1	Tm-2	Tm-3	B. henselae	B. quintana	B. koehlerae
1	60/M	B. koehlerae	gltA, ribC	59	400	800	1,600
2	63/M	B. henselae	gltA	49	59	1,600	400	800
3	71/M	B. henselae	gltA	49	59	3,200	1,600	3,200
4	37/M	B. henselae	gltA	49	59	3,200	1,600	1,600
5	58/M	B. henselae	ribC, pap31	NAd	NA	NA	12,800	6,400	NDe
6	63/M	B. henselae	gltA	49	60	400	NA	ND
7	22/F	B. henselae	gltA	49	59	3,200	1,600	ND
8	71/M	B. henselae	pap31	NA	NA	NA	3,200	3,200	ND
9	74/F	B. henselae	ribC	49	59	3,200	800	ND
10	75/F	B. henselae	ribC	49	59	12,800	12,800	ND
11	73/M	B. henselae	ribC	49	59	1,600	800	ND
12	43/M	B. henselae	pap31, 16S rRNA	51	61	NA	NA	ND
13	56/M	B. henselae	gltA	49	60	1,600	ND	ND
14	71/F	B. henselae	gltA, 16S rRNA	49	59	6,400	6,400	ND
15	20/M	B. henselae	ribC	51	60	6,400	3,200	ND
16	56/F	B. henselae	16S rRNA	49	59	12,800	12,800	ND
17	50/M	B. henselae	gltA	49	59	3,200	1,600	ND
18	42/F	B. henselae	16S rRNA	50	60	12,800	6,400	ND
19	56/M	B. henselae	gltA	49	59	1,600	800	ND
20	32/M	B. quintana	gltA, groEL, rpoB, ftsZ	59	70	400	1,600	400
21	20/M	B. quintana	gltA, groEL, rpoB, ftsZ	59	70	400	1,600	200
22	56/M	B. quintana	gltA, groEL, rpoB, ftsZ	59	69	1,600	6,400	800
23	43/M	B. quintana	16S rRNA, gltA, pap31	NA	NA	NA	800	1,600	ND
24	72/M	B. quintana	gltA	NA	NA	NA	3,200	3,200	1,600
25	56/M	B. quintana	ribC	59	70	200	1,600	ND
26	48/F	B. quintana	gltA, groEL, rpoB, ftsZ	59	70	1,600	6,400	ND
27	12/F	B. quintana	gltA, groEL, rpoB, ftsZ	58	69	100	1,600	ND
28	7/F	B. quintana	gltA, groEL, rpoB, ftsZ	61	71	1,600	NA	ND
29	28/M	B. quintana	gltA, groEL, rpoB, ftsZ	61	71	3,200	6,400	ND
30	52/M	B. quintana	16S rRNA	58	69	12,800	25,600	ND
31	16/F	B. quintana	gltA, groEL, rpoB, ftsZ	58	68	12,800	6,400	ND
32	9/F	B. quintana	gltA, groEL, rpoB, ftsZ	59	70	3,200	6,400	ND
33	75/M	B. quintana	gltA, groEL, rpoB, ftsZ	58	69	25,600	25,600	ND
34	65/M	B. quintana	gltA, groEL, rpoB, ftsZ	57	68	3,200	6,400	ND
35	20/F	B. quintana	ribC	59	70	200	200	ND
36	11/M	B. quintana	gltA, groEL, rpoB, ftsZ	59	70	12,800	25,600	ND
37	75/F	B. quintana	16S rRNA	58	69	3,200	6,400	ND

^aM, male; F, female.

^bMolecular diagnosis was determined by sequencing of the conventional PCR products and by real-time PCR melting-curve analysis. All B. henselae strains were Huston-1 type as determined by DNA sequencing.

^cT_m-1, melting temperature generated by the B. henselae-specific probe; T_m-2, melting temperature generated by the Bartonella genus probe; T_m-3, melting temperature generated by the B. quintana-specific probe.

^dNA, not available. Specimen was not available for further testing.

^eND, not done.

Serology.

EIA for the detection of anti-Bartonella antibodies was performed essentially as reported previously, with some modifications (12, 14, 23). The EIA antigen was prepared as sarcosyl-insoluble, presumably outer membrane protein extracts of agar-derived B. henselae strain 87-66 (ATCC 49793), B. quintana Fuller strain (CIP 107027), or B. koehlerae strain C-508 (8), as needed. Sera were initially screened at a 1:100 dilution, and positive sera at this dilution, as determined previously (12), were tested at incremental dilutions to determine their endpoints. To assess assay specificity, serum specimens from the following control groups were subjected to Bartonella EIA. (i) The first group consisted of patients with blood culture-positive endocarditis (n = 62), including 31 cases of Enterococcus faecalis endocarditis and 31 cases with other pathogens, including Streptococcus viridans, Streptococcus sanguinis, Streptococcus mutans, Streptococcus mitis, Streptococcus infantarius, Streptococcus gallolyticus, Streptococcus pneumoniae, Aggregatibacter aphrophilus, Aggregatibacter paraphrophilus, Aggregatibacter actinomycetemcomitans, Abiotrophia defectiva, Gemella, Staphylococcus aureus, coagulase-negative staphylococci, and Aspergillus. The reason for enriching this control group with E. faecalis is a recent publication suggesting a cross-seroreactivity, based on immunoblotting, between E. faecalis and Bartonella (28). (ii) The second group consisted of patients suspected to have chronic Q fever (n = 34), with anti-phase I IgG titers of 1:1,600 to 1:51,200, determined using IFA performed at the national reference laboratory for Q fever, the Israel Institute for Biological Research, Ness-Ziona, Israel. (iii) The third group consisted of patients seropositive for Chlamydia pneumoniae (either IgG ≥ 1:64, IgA ≥ 1:32, or both) (n = 20) and seronegative for Chlamydia trachomatis and Chlamydia psittaci, using IFA performed at the National Chlamydia and Mycoplasma Center, Rabin Medical Center, Petah Tikva, Israel. (iv) The last group consisted of patients with brucellosis and serum agglutination test value of ≥1:160 (n = 10).

DNA extraction and PCR.

Genomic DNA was extracted from tissue specimens and bacteria using the QIAamp DNA minikit (Qiagen, Valencia, CA) or E.Z.N.A. bacterial DNA kit (Omega Bio-Tek) and plasmid DNA was extracted using the QIAprep spin miniprep kit (Qiagen), all according to the manufacturers’ instructions, and kept at −20°C until further analysis. Conventional PCR was performed using primers targeting various genes (Tables 2 and 3), followed by DNA sequencing of 1 to 4 PCR products per specimen (DNA Sequencing Unit at the G. S. Wise Faculty of Life Sciences, Tel Aviv University, and Hy Laboratories Ltd., Rehovot, Israel) and analysis using Chromas (version 2.6; Technelysium, South Brisbane, Australia) and BLAST search in the GenBank database (https://www.ncbi.nlm.nih.gov/GenBank/).

TABLE 3.

Sequences of oligonucleotides relevant to this study

Oligonucleotide use and name	Oligonucleotide sequence (5′→3′)	Target organism	Target gene	Product size (bp)	Reference
Conventional PCR
BhCS.781p	GGGGACCAGCTCATGGTGG	Bartonella sp.	gltA	379	37
BhCS.1137	AATGCAAAAAGAACAGTAAACA	Bartonella sp.
gltA-F	GCTATGTCTGCATTCTATCA	Bartonella sp.	gltA	790	38
gltA-R	GATCYTCAATCATTTCTTTCCA	Bartonella sp.
Barton-1	TAACCGATATTGGTTGTGTTGAAG	Bartonella sp.	ribC	588	39
Barton-2	TAAAGCTAGAAAGTCTGGCAACATAACG	Bartonella sp.
PAPn1	TTCTAGGAGTTGAAACCGAT	B. henselae	pap31	277	40
PAPn2	GAAACACCACCAGCAACATA	B. henselae
ftsZ-F	ATTAATCTGCAYCGGCCAGA	Bartonella sp.	ftsZ	935	38
ftsZ-R	ACVGADACACGAATAACACC	Bartonella sp.
groEL-F	GAACTNGAAGATAAGTTNGAA	Bartonella sp.	groEL	1,459	38
groEL-R	AATCCATTCCGCCCATTC	Bartonella sp.
rpoB-F	CGCATTGGCTTACTTCGTATG	Bartonella sp.	rpoB	893	38
rpoB-R	GTAGACTGATTAGAACGCTG	Bartonella sp.
16S-27f	AGAGTTTGATCMTGGCTCAG	Eubacteria	16S rRNA	862-3	41
16S-907r	CCGTCAATTCMTTTRAGTTT	Eubacteria	42
Real-time PCR
ribC-bart-F1	GGTTTGGCTGAGATYATTGA	Bartonella sp.	ribC	185	Present study
ribC-bart-R1	TCAAGTGTATGRCGAATAAT	Bartonella sp.	Present study
Bartonella genus SimpleProbe	GAAGGTGAXITGCAGTTCGTT-Ph	Bartonella sp.	ribC	Present study
B. henselae SimpleProbe	TTTCCTXIACAAGTACCAAA-Ph	B. henselae	ribC	Present study
B. quintana SimpleProbe	TCCTGCAAGTXITCCAACAAAATTCACGCCCTT-Ph	B. quintana	ribC	Present study
β-globin-F	TACACATATTGACCAAATCAGGGTA	Human	β-globin	165	Present study
β-globin-R	TTTAGAATGGTGCAAAGAGGC	Human	β-globin	Present study
β-globin-probe	Cy5-TACTTTCCCTAATCTCTTTCTTTCAGGGCAAT-BBQ	Human	β-globin	Present study

Real-time PCR was developed in our research laboratory and has been used by the clinical microbiology laboratory since 2017. This assay was designed to specifically identify B. henselae and B. quintana, which account for >95% of all Bartonella endocarditis cases, and also identify other Bartonella spp. implicated in endocarditis at the genus level. Bartonella genus-specific degenerate primers (Sigma-Aldrich, Rehovot, Israel) were used to amplify a 185-bp fragment of the ribC gene. The resulting PCR product was detected postamplification using 3 SimpleProbe probes for the identification of Bartonella genus, B. henselae, and B. quintana DNA (TIB MOLBIOL, Berlin, Germany, and Dyn Diagnostics, Migdal HaEmeq, Israel) (Table 3). Once hybridized to its target sequence, the SimpleProbe probe emits more fluorescence than the free probe, which is self-quenched. Each probe has a distinctive melting temperature (T_m) at which the probe no longer binds the amplified target, resulting in an instantaneous drop in the fluorescent signal. As a result, changes in fluorescence are based solely on the hybridization status of the probe. Even minor changes in the target DNA sequence, such as single nucleotide polymorphisms, might result in a detectable shift of the melting curve, thus facilitating pathogen identification, without sequencing the amplicon (technical note no. LC 18/2004, Roche Applied Science). PCR conditions were as follows: 5 min at 95°C followed by 50 cycles of 10 s at 95°C, 20 s at 55°C, and 10 s at 72°C. The melting-curve analysis step was performed postamplification and consisted of 60 s at 95°C and 120 s at 37°C, with gradual increase to 80°C and cooling for 30 s. Real-time PCR and analysis were performed using the LightCycler 96 instrument and software version 1.1 (Roche Diagnostics, Basel, Switzerland). For the construction of positive-control plasmids, we cloned the ribC PCR amplicons of B. henselae and B. quintana into high-copy-number plasmids designated pMR-ribC185-Bh and pMR-ribC185-Bq, respectively (PCR cloning kit; Qiagen, Valencia, CA). Thirty, 300, and 3000 copies of both were used as positive controls along with the clinical specimen DNA in a multiplex real-time PCR assay. The analytical sensitivity was determined to be 3 to 30 plasmid copies per reaction (M. Giladi, unpublished data). To identify Bartonella, melting peaks of unknown specimens were compared to those of known standards (positive-control plasmids). Each run of clinical specimens also included a negative no-template control and an internal control gene (human β-globin) to verify template DNA integrity and absence of PCR inhibitors in the specimens. For the purpose of this study, all clinical specimens were subjected to both real-time PCR and conventional PCR except for four patients for whom tissue or DNA samples were not available for real-time PCR (Table 2). All PCR products of both real-time PCR and conventional PCR assays were sequenced to corroborate the identification of the Bartonella sp. To determine its specificity, the real-time PCR assay was performed on DNA extracted from non-Bartonella bacteria known to be associated with culture-negative endocarditis as presented in Table 1.

RESULTS

Serology.

Of 17 patients with B. henselae endocarditis who were tested with EIA using B. henselae-derived antigen, 16 (94%) had anti-B. henselae IgG titers of ≥1:800 and 1 patient had an IgG titer of 1:400 (range, 1:400 to 1:12,800). Of 17 patients with B. quintana endocarditis who were tested using B. quintana-derived antigen, 16 (94%) had anti-B. quintana IgG titers of >1:800 and 1 patient had an IgG titer of 1:200 (range, 1:200 to 1:26,600). One patient with B. koehlerae endocarditis had an anti-B. koehlerae IgG titer of 1:1,600 when tested using the corresponding antigen. Significant cross-reactivity between B. henselae, B. quintana, and B. koehlerae was found in all sera tested (Table 2). Two or more serum specimens taken from the same patient at least 1 month apart were available for 12 patients, 10 with B. henselae and 2 with B. quintana endocarditis (Fig. 1). The median duration to achieve IgG titer decrease of two dilutions, observed in 11 patients, was 5 months (range, 2 to 41 months).

FIG 1.

Kinetics of anti-Bartonella IgG antibodies expressed as reciprocal titers in a logarithmic scale in patients with Bartonella endocarditis (n = 12). Only results for patients with 2 or more serum specimens taken at least 1 month apart are presented. Each circle represents the mean enzyme immunoassay optical density of triplicate determinations for 1 patient. Each patient is marked with a different color.

The anti-Bartonella IgG titer distribution among patients with Bartonella endocarditis and patients in the control groups tested to evaluate specificity is presented in Fig. 2. Of 62 patients with blood culture-positive endocarditis, 58 (94%) were considered seronegative, with an anti-Bartonella IgG titer of <1:100, while 4 patients (6%), 3 of them with Enterococcus faecalis endocarditis, had an IgG titer of 1:100. Of 34 patients with serological profile consistent with chronic Q fever, 20 (59%) were seronegative, while 5 patients were seropositive for Bartonella henselae and/or B. quintana with IgG titers of 1:100, 4 patients with 1:200, 4 patients with 1:400, and 1 patient with an anti-Bartonella IgG titer of 1:800. Of note, among the 14 patients with chronic Q fever and cross-reactivity with Bartonella, anti-Bartonella IgG titers were 3 to 7 dilutions lower than the matching anti-Coxiella burnetii phase I IgG titers. Of the patients with Bartonella endocarditis, only one patient (3%), with an anti-Bartonella IgG titer of 1:3,200, tested positive for Q fever, with a phase I IgG titer of 1:200. Of 20 patients with Chlamydia pneumoniae infection, 18 (90%) were seronegative for Bartonella (IgG < 1:100) and 2 were seropositive, with IgG titers of 1:100 and 1:200. Of 10 patients with brucellosis, 9 (90%) were seronegative for Bartonella and 1 had an IgG titer of 1:100.

FIG 2.

Distribution of anti-Bartonella IgG reciprocal titers, expressed in a logarithmic scale, in patients with Bartonella endocarditis and 4 control groups: patients with culture-positive endocarditis, patients with anti-Coxiella burnetii IgG phase I antibody titers of ≥1:1,600, Chlamydia pneumoniae-seropositive patients, and patients with brucellosis and serum agglutination test values of ≥1:160. Each circle represents the mean enzyme immunoassay optical density of triplicate determinations for 1 patient.

Molecular diagnosis.

We first evaluated the performance of the Bartonella genus-specific probe to identify Bartonella spp. known to cause endocarditis. All species tested were identified, each with a single fluorescent peak, some with a unique T_m (e.g., B. elizabethae, T_m, 43°C, and B. washoensis, T_m, 50°C) and others with overlapping T_ms which do not allow interspecies separation (B. henselae, B. quintana, B. koehlerae, and B. vinsonii, T_m, 57 to 58°C) (Fig. 3 and Table 1). DNA sequencing confirmed the species identity.

FIG 3.

Multiplex real-time PCR performed on 7 Bartonella spp. reported to cause endocarditis in humans, utilizing a Bartonella genus-specific probe. The melting temperature (T_m) was calculated by plotting the negative derivative of fluorescence over temperature (−dF/dT). The T_m value was defined as the peak of the curve. Values were rounded to the nearest integer.

The real-time PCR assay performed on infected cardiac specimens was positive for all 33 tested specimens and was in complete agreement with results of the sequence-based diagnosis of the conventional PCR. The assay specifically identified all cases of B. henselae and B. quintana endocarditis based on the fluorescent pattern and T_m values. T_m-1 (range, 49°C to 51°C) and T_m-3 (68°C to 71°C) were generated by the specific B. henselae and B. quintana probes, respectively. T_m-2 (57°C to 61°C), representing the genus-specific probe, was common to B. henselae, B. quintana, and B. koehlerae (Table 2). Figures 4 and 5 present the identification of B. quintana DNA in infected native aortic valve and B. henselae DNA in infected native mitral valve, respectively, of two patients with culture-negative endocarditis. Specificity of the real-time PCR assay was established when the multiplex assay, including Bartonella genus, B. henselae, and B. quintana probes, was performed using DNA of non-Bartonella culture negative-associated bacteria and resulted in negative fluorescent signals (Table 1).

FIG 4.

Multiplex real-time PCR using 3 SimpleProbe probes: Bartonella genus-specific, B. henselae-specific, and B. quintana-specific probes. PCR was performed on valvular tissue of a 65-year-old patient with native aortic valve Bartonella quintana endocarditis (patient number 34 [Table 2]). Two fluorescent peaks corresponding to melting temperatures (T_ms) of 57.2°C and 67.6°C are the result of detachment of the genus-specific and B. quintana-specific probes from the PCR product, respectively. Plasmids pMR-ribC185-Bq and pMR-ribC185-Bh were used as positive controls for B. quintana and B. henselae, respectively.

FIG 5.

Multiplex real-time PCR using 2 SimpleProbe probes, B. henselae-specific and genus-specific probes. PCR was performed on valvular tissue of a 56-year-old patient with native mitral valve Bartonella henselae endocarditis (patient number 13 [Table 2]). Two fluorescent peaks corresponding to T_ms of 49.3°C and 59.8°C are the result of detachment of the B. henselae-specific and genus-specific probes from the PCR product, respectively. Plasmids pMR-ribC185-Bq and pMR-ribC185-Bh were used as positive controls for B. quintana and B. henselae, respectively. The peak corresponding to the B. quintana-specific probe is not seen since the B. quintana-specific probe was not used in this specific assay.

Cultures.

Of the 28 clinical cardiac specimens available for culture, 1 (4%) was positive, growing B. henselae. Two blood cultures were positive, one growing B. henselae and the other B. quintana.

DISCUSSION

This report describes the modalities used in a national reference center for the diagnosis of Bartonella endocarditis. The fact that all Bartonella endocarditis cases in Israel are referred for diagnoses to a single laboratory permitted analysis of the performance of these methods for 37 patients diagnosed with Bartonella endocarditis during a 20-year period. To date, the mainstay of the literature on Bartonella endocarditis, and in particular description of diagnostic methods, stems from series originating in France and one series from the United Kingdom, with limited data reported elsewhere (1, 18, 29–31). In fact, a recent review published in 2019 identified only 13 cases of B. quintana endocarditis reported from North America, although Bartonella spp. are believed to be the most common cause of culture-negative endocarditis in the United States (32, 33).

IFA, either as a “homemade” or a commercial product, is the most commonly used serological method for the diagnosis of Bartonella infections, including Bartonella endocarditis, utilized among others by the national reference center for Q fever and Bartonella infection in France and the Centers for Disease Control and Prevention (CDC) in the United States (1, 34). In comparison with EIA, IFA has several limitations. Preparation of antigen requires cultivation of Bartonella spp. in Vero cells or other cell lines. Attempts to use agar-derived antigens have resulted in autoagglutination of the Bartonella bacilli, which interferes with the IFA (34). In addition, IFAs in general suffer from interobserver variation and are also not suitable for automation or for screening large numbers of specimens. We found that anti-Bartonella IgG titers of ≥1:800 were recorded for 94% of patients with Bartonella endocarditis, while the remaining patients, though PCR positive on cardiac infected tissue, demonstrated lower titers. A previous study from France utilizing an IFA on 106 patients with Bartonella endocarditis also demonstrated that IgG titers of ≥1:800 correlate with endocarditis in 94% of cases (1). Clinicians should be aware, however, that relatively low IgG titers, obtained by either IFA or EIA, may represent rare cases of B. henselae or B. quintana endocarditis but may also be attributable to non-quintana, non-henselae Bartonella spp., as was demonstrated in our B. koehlerae endocarditis patient who had IgG titer of 1:400 when tested using B. henselae antigen, compared with a titer of 1:1,600 when B. koehlerae antigen was incorporated in the EIA. Similarly, an IgG titer of <1:800 was reported for patients with B. alsatica endocarditis who were tested with IFA using either B. henselae or B. quintana antigens (1).

Although, similarly to IFA, our EIA cannot distinguish between endocarditis due to B. henselae and B. quintana as a result of cross-seroreactivity between these two species, the clinical significance of this finding is limited since the medical approaches to B. henselae and B. quintana endocarditis are quite similar (17, 31). More concerning is the cross-reactivity between Bartonella spp. and Coxiella burnetii, since they are important etiologic agents of culture-negative endocarditis with very similar clinical characteristics yet with dissimilar practices for antimicrobial treatment and serological follow-up (17, 18). Of 34 patients with serology consistent with chronic Q fever, we found that 13 patients (38%) showed cross-seroreactivity with Bartonella spp.; however, all had IgG titers of <1:800, except 1 patient (3%) with anti-Coxiella burnetii phase I IgG titer of 1:6,400 who tested positive for Bartonella with an IgG titer of 1:800. Cross-reaction between C. burnetii antibodies and Bartonella antigen have been reported before (17, 35). La Scola and Raoult (17) have found that using IFA, more than 50% of chronic Q fever patients had sera which reacted to a significant level with B. henselae and/or B. quintana. Although cross-adsorption studies and Western blot analysis have been shown to assist identification of the true culprit in cases with cross-reactivity while also increasing the diagnostic sensitivity (1), these assays are not currently available in most clinical microbiology laboratories (1, 17). However, Bartonella-Coxiella cross-seroreactivity infrequently poses a clinical quandary since in the majority of cases, IgG titers against the true pathogen are several dilutions higher than those of the pathogen causing the false-positive serology and thus allowing for the correct diagnosis. To further explore the specificity of our EIA, we tested sera from several control groups. We have also demonstrated low-level cross-reactivity in cases of Chlamydia pneumoniae infections similarly to a previous study using an IFA (19), in cases infected with Brucella melitensis, a pathogen phylogenetically close to Bartonella, and in cases with culture-positive bacterial endocarditis. Cross-reactivity, however, was uncommon in these cases and always in low titers, supporting an anti-Bartonella IgG titer of ≥1:800 as a cutoff for Bartonella endocarditis diagnosis even in the presence of cross-reactivity. Such low-level cross-reactivity does not necessarily indicate antigenic similarity but may also be the result of polyclonal B cell activation, known to occur in endocarditis (27, 36). Of particular interest is a subgroup of patients with E. faecalis endocarditis, since a recent publication demonstrated cross-seroreactivity with this pathogen in Bartonella-infected endocarditis patients tested by Western immunoblotting (28). We have tested serum samples from 31 patients with E. faecalis endocarditis, and although 10% were Bartonella seropositive, none demonstrated high-level seropositivity (>1:100) that might interfere with the interpretation of Bartonella EIA. Further studies are needed to evaluate the added value of Western blotting, which may increase diagnostic sensitivity (at the expense of lower specificity). The potential benefit, however, must be weighed against the increased burden placed on the microbiology laboratory that does not perform Western immunoblotting or cross-adsorption studies on a routine basis.

We also showed that the multiplex real-time PCR assay performed on cardiac tissue accurately identified B. quintana and B. henselae, which are responsible for almost all Bartonella endocarditis cases in humans, without the need for DNA sequencing. The unique melting curves generated by the interaction of the genus-specific and species-specific probes with the PCR product results in 2 distinct T_ms, 9 to 11°C apart, which are easily recognized. Melting peaks of clinical specimens were compared to those of known standards (plasmids pMR-ribC185-Bh and pMR-ribC185-Bq). This comparison to standards is essential since T_m is not a constant value and a significant T_m shift may occur due to multiple factors, such as oligonucleotide concentration, cation concentration, particularly free magnesium (which by itself is inversely related to the amount of deoxynucleoside triphosphates [dNTPs], number of probes and primers, or amount and length of the target DNA), and other components of the reaction. We have observed a T_m shift of 1 to 4°C (Tables 1 and 2); however, all probe-generated peaks of a tested specimen and positive controls are shifted in the same direction and magnitude, resulting in a fixed interval between peaks. We demonstrated that the genus-specific probe also identified 5 non-henselae, non-quintana Bartonella spp. known to cause endocarditis. B. alsatica and “Candidatus Bartonella mayotimonensis,” also reported to rarely cause endocarditis, were not tested with our real-time PCR assay; however, due to the use of degenerate primers designated to encompass a large number of Bartonella spp. and the characteristics of the genus-specific probe, we predict that they will be recognized by this probe. To further support this prediction, we tested 21 additional Bartonella spp. isolated from various hosts, including ruminants, rodents, felines, squirrels, a kangaroo, and a bat, from different continents, including North America, Africa, Europe, Asia, and Australia, and all were recognized by the genus-specific probe, resulting in a single peak with variable T_ms (Giladi, unpublished). Along these lines, we expect that if a new Bartonella sp. is implicated in endocarditis, it will be recognized by the real-time PCR assay. Contrary to the case with B. quintana and B. henselae, which are accurately identified by specific probes, the identity of other species recognized by the genus probe alone needs to be corroborated by either DNA sequencing or a second PCR to distinguish between members of the same melting group. As non-henselae, non-quintana Bartonella spp. are extremely rare causes of endocarditis, a decision to add additional specific probes for the identification of these species is unlikely to be supported by cost-benefit analysis. However, if future epidemiological studies demonstrate that the incidence of another endocarditis-associated Bartonella sp. is increased, new species-specific probes can be incorporated into this assay. The high specificity of the real-time PCR system is achieved by using Bartonella-specific degenerate primers to amplify part of the ribC gene and 3 Bartonella-specific probes. This was demonstrated by testing a large number of bacterial pathogens known to be associated with culture-negative endocarditis, including members of the HACEK group (a group of Gram-negative bacilli consisting of Haemophilus spp., A. actinomycetemcomitans, Cardiobacterium hominis, Eikenella corrodens, and Kingella spp.), nutritionally variant streptococci (e.g., Abiotrophia spp.), Coxiella burnetii, Legionella pneumophila, and Tropheryma whipplei, none of which produced false-positive results. The high positivity rate of PCR performed on infected cardiac tissue is in agreement with previous studies reporting 93 to 98% positivity rates of PCR performed on valvular specimens in spite of the fact that in one of these studies, 62% of the specimens were obtained from patients receiving antibiotic therapy. To summarize, the current study highlights the SimpleProbe real-time PCR as a sensitive and specific method which is also rapid and high throughput for the detection and differentiation of Bartonella spp. causing endocarditis.

In conclusion, we have demonstrated that EIA coupled with a novel real-time PCR assay can play an important role in the laboratory diagnosis of Bartonella endocarditis and expand the diagnostic arsenal at the disposal of the clinical microbiologist. Since serology remains the major diagnostic tool in the absence of tissue suitable for PCR, recognizing its many pitfalls is essential to avoid incorrect diagnosis.