During June–August 2018, a total of 114 persons experiencing homelessness (101 men, 13 women; average age 42.5 ±13.4 years) from a day-shelter in the city of São Paulo signed written informed consent forms and participated in this study, which was approved by the National Ethics Committee in Human Research (protocol no. 80099017.3.0000.0102). Persons responded to a questionnaire that, combined with medical and demographic records (Appendix), we used to assess risk factors. We carefully examined personal clothing and found lice in 14.9% (17/114, 95% CI 6.9%–19.7%) of persons; the lice were taxonomically identified as P. humanus humanus (5).

We analyzed 109 serum samples from study participants by using IFA to detect IgG against Bartonella spp. and TGR. We used commercial slides for B. quintana (12-well IFA Substrate Slides; Fuller Laboratories, http://www.fullerlaboratories.com) and in-house slides for B. henselae sequence type 9, B. machadoae 56A, R. typhi Galveston, and R. prowazekii Breinl strains. We found 79/109 (72.5%, 95% CI 63.1%–80.1%) persons were seropositive for Bartonella spp. and 40/109 (36.7%, 95% CI 27.7%–46.5%) were seropositive for TGR (titers >64). All antibody titers were >128 (Appendix Table 1), except for 2 B. quintana–positive and 8 TGR-positive samples. An endpoint titer >4-fold higher for a particular Bartonella/Rickettsia spp. antigen than that observed for other Bartonella/Rickettsia spp. antigens was considered the possible antigen involved in a homologous reaction (PAIHR) (6). Thus, B. quintana was the PAIHR in 75/79 (95.0%, 95% CI 87.5%–98.6%) persons, R. typhi was the PAIHR in 13/40 (32.5%, 95% CI 18.6%–49.1%) persons, and R. prowazekii was the PAIHR in 3/40 (7.5%, 95% CI 1.6%–20.4%) persons (Appendix Table 1).

We extracted DNA by using the Blood/Tissue DNA Kit (MEBEP Bio Science, https://www.mebep.com) for 114 blood samples and guanidine isothiocyanate and phenol/chloroform technique (7) for 638 lice (82 pools). We confirmed successful extractions by PCR of glyceraldehyde-3-phosphate dehydrogenase (blood) and invertebrate mitochondrial cytochrome c oxidase subunit I (lice) genes (8,9). We screened DNA samples for Bartonella spp. by PCR of citrate synthase (gltA) and β subunit of RNA polymerase (rpoB) genes and for Rickettsia spp. by PCR of rickettsial 17-kDa antigen gene, as previously described (10–12). We used ultrapure water as a negative control and genomic DNA from B. henselae and R. sibirica as positive controls. A total of 14/82 (17.0%, 95% CI 9.7%–27.0%) louse pools and 1/114 (0.9%, 95% CI 0.02%–4.8%) blood samples were positive for gltA and rpoB but negative for Rickettsia spp. (Appendix Table 2).

Amplicons were purified and sequenced at the University of Texas Medical Branch (Galveston, TX, USA). The gltA and rpoB sequences showed 100% identity to B. quintana strain NCTC12899 (GenBank accession no. LS483373.1) by BLASTn analysis (https://blast.ncbi.nlm.nih.gov). B. quintana sequences generated in this study were deposited in GenBank (accession nos. ON808843 and ON808844). The person whose blood was PCR-positive for B. quintana was not infested with body lice but demonstrated high levels of IgG against B. quintana (titer >1,024) and TGR (titers were 1,024 for R. typhi and 512 for R. prowazekii).

We chose risk factor variables by using unconditional logistic regression models (p<0.25) and conditional logistic regression to determine relationships between putative risk factors and serologic status. We used Bayesian information criteria to assess the goodness-of-fit for the models. We used R software version 4.1.2 (The R Project for Statistical Computing, https://www.r-project.org) for all statistical analyses and summarized the final conditional logistic regression model (Table). Although the final model for Bartonella spp. revealed 3 variables, only 1 was statistically significant and showed an association between body louse infestation and higher risk for Bartonella spp. seropositivity (OR [odds ratio] 2.9, 90% CI 1.1–8.1). The final TGR model contained 5 variables of which 3 were associated with higher seropositivity risk, including self-identifying as white (OR 3.9, 90% CI 1.6–10.7), syphilis seropositivity (OR 3.6, 90% CI 1.5–9.4), and homelessness because of unemployment (OR 2.3, 90% CI 1.02–5.5). We detected 5 variables for combined Bartonella spp. and TGR of which 4 variables were associated with seropositivity, including self-identifying as white (OR 5.6, 90% CI 2.2–15.5), monthly change of clothes (OR 0.08, 90% CI 0.07–0.4), homelessness because of family conflicts (OR 0.4, 90% CI 0.2–0.8), and higher total plasma protein (OR 2.0, 90% CI 1.1–4.0).

Our study revealed Bartonella spp. and TGR exposure, associated risk factors related to serologic status, and B. quintana detection in lice and one blood sample among persons experiencing homelessness in São Paulo, Brazil. Seroprevalence of Bartonella spp. (72.5%) was higher in our study than previous reports for persons experiencing homelessness (1.8%–65%) (13), and B. quintana was the dominant antigen involved in homologous reactions. The highest B. quintana seroprevalence was previously found in France (65%, antibody titers >100) and Japan (57%, titers >128) (13); those titers were considered indicative of previous exposure. In our study, the antibody titer cutoff was >64, explaining our high seroprevalence results, although all but 2 titers were >128. TGR seropositivity in our study (36.7%) was within the range observed in the United States, Europe, and Colombia (0.54%–56.2%) (1,3).

Persons experiencing homelessness in São Paulo had P. humanus humanus body louse infestation and seropositivity for B. quintana and TGR similar to that reported previously (1). Body louse infestation (14.9%) was within the range of other reports (7%–22%) (14), highlighting global vulnerability to louse infestation and louseborne diseases in persons experiencing homelessness (1).

Through logistic regression, we showed seropositivity for Bartonella spp. was associated with louse infestation. Because the association of white ethnicity and TRG seropositivity (alone and in combination with Bartonella spp.) might be from a skewed population sampling, our findings should be further investigated. Nonetheless, higher TGR seropositivity was associated with homelessness because of unemployment, duration of homelessness, and syphilis seropositivity, which represent risk factors that reflect vulnerability and socioeconomic conditions. In addition, seropositivity for both Bartonella and TGR was associated with infrequent changes of clothing.

The first limitation of our study is that the small sample size and power for the examined variables might have weakened associations of Bartonella seroreactivity with other variables included in our questionnaire, such as alcoholism, tobacco or intravenous drug use, and homelessness as previously reported (15), and variables that were significant in univariate analysis. In addition, IgG seropositivity reflects past Bartonella and TGR infections (6,13). IFA cross-reactivity should be addressed with future studies by using cross-adsorption techniques.

Our results should alert public health professionals in the city of São Paulo to initiate preemptive measures and active vector control among persons experiencing homelessness and confirm circulation of Bartonella and TGR species. Clinicians should also consider these vectorborne agents as probable etiologic agents of febrile syndromes in this vulnerable population.