Bartonella spp. Infections Identified by Molecular Methods, United States

Molecular diagnostic testing can identify these pathogens, including uncommon and previously undescribed species.

specific results, differences in use and interpretation of serologic assays among laboratories, and crossreactivity with other pathogens, including Chlamydia spp. and Coxiella burnettii (9)(10)(11)(12)(13)(14)(15), which can lead to misdiagnosis and insufficient treatment. Bacterial cultures of blood or tissue can establish the diagnosis, but sensitivity of cultures is low because of the fastidious nature of Bartonella spp. and might result in underdiagnosis of infection. Combining enrichment culture techniques with molecular methods might increase detection of Bartonella spp. in blood or other clinical specimens (16).
Molecular detection of bacterial pathogens has emerged as an important clinical tool that can increase diagnostic yield compared with culture alone, particularly for detection of fastidious organisms (17). Bartonella spp. have been detected by using broadrange PCR-based assays that target conserved binding sites flanking regions of the rRNA gene and have species-specific variations (18) or by using organismspecific gene targets, such as gltA (19) and ribC (20). We describe the demographic, clinical, and microbiologic characteristics of patients with bartonellosis diagnosed by both broad-range and organism-specific molecular assays at a large clinical reference laboratory in Washington, USA.

Population
We included information for all patients who had an acceptable clinical specimen submitted to the University of Washington (Seattle, Washington, USA) Molecular Microbiology clinical diagnostic reference laboratory during 2003-2021. Acceptable specimen types were fresh frozen tissue, formalin-fixed paraffin embedded (FFPE) tissues, and body fluids other than blood. To identify Bartonella spp., we performed 16S rRNA sequencing by using broad-range PCR primers (PCR-16S) or next-generation sequencing (NGS-16S) or performed a Bartonella henselae and B. quintana bispecies-targeted PCR (BT-PCR) requested by the ordering clinician.
We evaluated patient information if the specimen had Bartonella spp. as the final species assignment, had sufficient material for testing, and was an acceptable specimen type for the assay performed and if the presence of PCR inhibitors was excluded (inhibitors precluded ruling out Bartonella spp.). If a patient had multiple specimens submitted >30 days apart, we included only the information provided with the first positive specimen submission in the analysis. From the specimen submission form, we obtained the patient's sex and age, US state and healthcare facility where the specimen was submitted, specimen collection date, and specimen description (e.g., anatomic location).

Ethics Approval
The study was approved by the University of Washington Institutional Review Board (approval no. STUDY00013877). This study was reviewed in accordance with policies and procedures of the Centers for Disease Control and Prevention and was determined to be exempt from Institutional Review Board requirements.

Molecular Methods for Bartonella spp. Identification
The University of Washington Molecular Microbiology laboratory performed all clinical testing pursuant to its high-complexity Clinical Laboratory Improvements Amendments license and College of American Pathologists accreditation. Laboratory-developed processes were validated in accordance with standards set by the Clinical Laboratory Standards Institute and monitored for quality through regular proficiency testing, biannual external inspections by the College of American Pathologists, and incorporation of control reactions as prescribed by the Clinical Laboratory Improvements Amendments. We performed DNA extraction as previously described for both fresh and FFPE tissue (21), except that we also validated and used multiple DNA extraction kits (QIAGEN) in 2021 because of COVID-19-related supply chain disruptions. We performed broad-range PCR amplification of the V1-V2 hypervariable region of the bacterial 16S rRNA gene (PCR-16S) as previously described (18). When sequencing results of 16S PCR products suggested mixed DNA templates, we performed amplicon-based NGS of the V1-V2 16S rRNA locus (NGS-16S) reflexively by using an Illumina Miseq instrument and 250 bp pairedend reads (22,23). Targeted detection of Bartonella spp. by BT-PCR incorporated broad-range 16S primers and 2 primer sets that amplified ribC alleles of B. quintana and B. henselae. The 16S primers also detect other Bartonella spp. but at a higher limit of detection: 100 templates per reaction for other Bartonella spp. compared with 5-25 for the Bartonella speciesspecific primers. We sequenced PCR products by using the Sanger method and assigned taxonomic classification after BLAST analysis (24) by using both the National Center for Biotechnology Information public databases and a curated database containing type strains and RefSeq (https://www. ncbi.nlm.nih.gov/refseq) records (21,22). Each case was reviewed by 2 independent certified medical laboratory scientists and a board-certified pathologist (21).

Description of Detected Bartonella spp.
When available, results from BT-PCR provided specific taxonomic classification of Bartonella spp. If BT-PCR results were unavailable, we used results from PCR-16S or NGS-16S. Some 16S rRNA gene V1-V2 sequences lacked sufficient specificity to provide a species-rank classification, either because of close homology to multiple species or lack of homology to any established species. Therefore, we retrospectively analyzed those sequences reported as Bartonella spp. by using BLAST to attempt a species-rank classification (defined as 99.7% sequence homology of the V1-V2 sequence). For diagnostic reporting purposes, species with validly published names according to the taxonomic code were identified at the species level in the clinical diagnostic report; otherwise, we identified the genus in accordance with the laboratory's standard operating procedures (Appendix, https://wwwnc.cdc. gov/EID/article/29/3/22-1223-App1.pdf) (22,23).

Statistical Methods
We used χ2 tests to compare categorical variables and Mann-Whitney U tests to compare age distributions. For comparisons between patients who had B. quintana infections and those who had B. henselae infections, we excluded information from patients who had specimen results indicating another Bartonella sp. We performed all statistical analyses by using R software version 4.0.3 (25).
To examine the proportion of specimens infected with Bartonella spp. by year and US state, we only included information for patients who had specimens submitted for BT-PCR. We excluded specimens submitted for PCR-16S that was used to evaluate a diverse array of bacterial pathogens. To calculate percent positivity of BT-PCR results, we collected data on ordering location (US state) and year of specimen submission for all BT-PCR assays, including those in which Bartonella spp. was not detected. Similar denominator data for PCR-16S and NGS-16S testing were not available.
Among 397 patients who had Bartonella spp. detected in >1 specimen and available data regarding their sex, 245 (62%) were male ( Specimen descriptions were provided for all specimens. The most common specimens submitted were

Discussion
We report 420 cases of bartonellosis in the United States identified by uniform, clinical molecular diagnostic methods at a single reference laboratory during 2003-2021, representing 36 US states and including 140 persons who had endocarditis because of Bartonella spp. infections. Those cases highlight the broad clinical spectrum of disease caused by Bartonella spp., provide evidence that multiple species can cause bartonellosis, and demonstrate the utility of clinical molecular testing for pathogen identification.
B. henselae was the most frequently identified species causing bartonellosis. Although B. henselae was most often detected in lymph node specimens, consistent with cat-scratch disease (3), this species was also identified in cardiac specimens, underscoring its potential to cause endocarditis and endovascular disease (26). B. henselae was also detected in a diverse range of clinical specimens, including liver and bone, indicating the breadth of atypical B. henselae infections (4,5). Although atypical bartonellosis manifestations are rare, ≈25% of pediatric hospitalizations associated with cat-scratch disease are caused by atypical B. henselae infections (27). A previous study of US insurance claims data found that atypical infections with B. henselae accounted for 1.5% of cases; ocular and hepatic lesions were the most common clinical manifestations of atypical B. henselae infection (28). Overall, Bartonella spp. were more frequently found in cardiac specimens in our study, which might reflect the preponderance of certain specimen types submitted for molecular diagnostic tests.
B. quintana was the second most frequently identified species overall and in cardiac specimens, which is consistent with B. quintana as a causative agent of subacute endocarditis (11,26,29). California and Washington had the highest proportion of specimens infected with B. quintana; a high prevalence of antibodies against B. quintana has been described among PEH living in both states (6,7). Eight states reported >10 cases of bartonellosis during the study period, and Texas and Washington reported the highest number of cases. The higher proportion of B. quintana infections observed in California and Washington might be from epidemiologic clustering, because B. quintana is transmitted by body lice, which can be spread through shared clothing and bedding material, and clusters of B. quintana infections have been reported among PEH (6,8,30). The higher number of B. quintana infections in California and Washington might reflect geographic differences in the number or type of specimens submitted for molecular testing. Although other studies have reported that B. henselae infection is more common in the southeastern region of the United States (3,28), we did not identify a clear predominance of infections in this region, which might reflect specimen submission patterns. Despite a potential bias from specimen submission patterns, molecular testing methods might identify spatiotemporal clusters of B. quintana infections. *Abscess (n = 38), tissue (n = 32), neck/face/arm (n = 25), liver (n = 15), leg/hip (n = 14), breast/chest wall (n = 6), skin (n = 5), lung (n = 4), aorta (n = 2), vascular (n = 2), body fluid NOS (n = 4), bone (n = 4), spleen (n = 4), synovial fluid (n = 2), and dura mater (n = 1). †By χ 2 test.
Persons who had B. henselae infections were younger, and a higher proportion were female compared with those who had B. quintana infections. This finding likely reflects the higher incidence of B. henselae infection in children, possibly because children might spend more time in close proximity with domesticated cats (3). Persons <18 years of age commonly had lymph node or abscess involvement, typical of cat-scratch disease (5).
Nearly all patients with endocarditis had leftsided disease, which is consistent with other reports of endocarditis caused by Bartonella spp. (11,29,31). We found that B. henselae was a more frequent cause of endocarditis in our study than in prior studies that used molecular methods (26,29). In a study of 685 patients in the United Kingdom who had endocarditis, B. quintana was identified in 12/13 cases that had available PCR diagnostic results; B. henselae was identified in only 1/13 cases (29). A retrospective review of cases using PCR testing on heart valves identified B. quintana in 26/45 patients, compared with 19/45 patients who had B. henselae infections (26). However, the differences observed in our study might be because of differences in sample submission practices; we did not have information from the referring hospitals on the total number of patients with endocarditis caused by Bartonella spp. Although previous reviews have recommended the use of serologic testing to diagnose endocarditis caused by Bartonella spp. (31), our findings, in combination with those of other large studies (26,29), underscore the improved accuracy of molecular testing methods in obtaining a definitive diagnosis.
We identified 9 patients who had infections from novel or newly emerging Bartonella spp., including 2 patients with endocarditis caused by B. vinsonii and 4 patients with endocarditis caused by B. clarridgieae. Both B. vinsonii and B. clarridgieae have been reported to cause endocarditis (20,26,32), although only 1 case of endocarditis caused by B. clarridgieae infection has been previously reported (32); that infection was also confirmed by using molecular methods. We detected Candidatus B. mayotimonensis in 1 case of endocarditis, similar to findings for a single case reported previously (19). We also detected a case of B. washoensis lymphadenitis; this bacterium has been detected in ground squirrels and their fleas and was reported to be the etiologic agent in 2 human cases of myocarditis and meningitis in the United States (33) and prosthetic valve endocarditis in a patient in Germany (34). We detected a novel 16S rRNA sequence variant of Bartonella, most likely representing a previously undescribed species.
Although serology is an important diagnostic method for bartonellosis, case reports have described cross-reactivity of antibodies against B. clarridgieae, B. henselae, and B. quintana (32,35,36) and cross-reactivity between antibodies against B. vinsonii and Coxiella burnetii (20), highlighting the utility of molecular methods for species-specific diagnosis of those pathogens. Because B. quintana, B. henselae, and other non-Bartonella spp. pathogens are often treated with different antimicrobial drugs and for different durations, identification of infecting species can frequently have major effects on treatment, in addition to the epidemiologic value of recognizing novel organisms.
Despite increased specificity of molecular methods compared with serology and potential to detect novel organisms, disparate PCR positivity rates for fresh (16%) and FFPE (8%) tissue specimens highlight how preanalytical factors, such as formalin fixation, can limit assay yield. Other potential factors that can limit sensitivity of molecular methods include specimen selection, organism prevalence, and tissue volume. Formalin fixation damages DNA and can reduce assay sensitivity but has not always reduced positivity rates, perhaps because histopathologic evaluation permits selection of tissue most likely to contain microbial DNA (21). In this study, lymph nodes constituted the highest proportion of FFPE specimens. The observed discordant PCR positivity between specimen types might reflect anatomic site of disease, volume of tissue available for PCR, or tissue-specific variations in organism prevalence.
The first limitation of our study is that the analyzed specimens were not representative of all infections caused by Bartonella spp. because they represented more severe illness, and submitted specimens were primarily from tissues known to be infected by those pathogens. Second, the specimens did not represent a random sample of persons with bartonellosis; therefore, we could not estimate incidence or perform statistical inference testing. Third, not all geographic regions or states were equally represented, limiting our ability to compare results between different geographic areas. Fourth, we had little information regarding clinical features of the patients from whom specimens were collected, limiting our ability to determine potential risk factors on the basis of medical or social history. Fifth, preanalytical specimen handling can affect diagnostic yield; we could not assess or control variations in storage or transport conditions or tissue processing, such as formalin pH or acid-based decalcification of bone. Sixth, clinical testing might have missed Bartonella spp. other than B. henselae or B. quintana that were present at levels below the limit of detection for the 16S primers, because the ribC primers were designed to increase specificity and sensitivity for detecting B. henselae and B. quintana. Finally, some specimens might have been misclassified, leading to a higher proportion of specimens categorized as abscesses or tissue if more specific information on anatomic site was not provided on the specimen submission form.
In conclusion, molecular methods provide a powerful diagnostic tool to detect infections caused by Bartonella spp. Those methods should be considered for patients who have culture-negative endocarditis or lymphadenitis of unclear etiology, particularly in persons with established risk factors, including exposure to cats or prior homelessness. Broader use of molecular methods in suspected cases of bartonellosis will help elucidate the full clinical spectrum of Bartonella infections and increase awareness of this underrecognized pathogen.