An Investigation of Bartonella spp.,Rickettsia typhi,and Seoul Hantavirus in Rats ( Rattus spp.) from an Inner-City Neighborhood of Vancouver,Canada: Is Pathogen Presence a Reflection of Global and Local Rat Population Structure?

Introduction

Urban Norway and black rats (Rattus norvegicus and Rattus rattus) are known to be reservoirs for a number of different zoonotic pathogens responsible for significant human morbidity and mortality (Meerburg et al. 2009, Himsworth et al. 2013a). The majority of rat-associated zoonoses (RAZ) are thought to have a worldwide distribution and to be endemic wherever Norway and black rats are found (Meerburg et al. 2009, Himsworth et al. 2013a).

Rickettsia typhi, Bartonella spp., and Seoul hantavirus (SEOV) are three important RAZ (Himsworth et al. 2013a). Although these pathogens have been identified in number of different geographic locations (Arikawa et al. 1994, Gundi et al. 2004, Easterbrook et al. 2007a, Heyman et al. 2009, Hsieh et al. 2010, Psaroulaki et al. 2010, Abramowicz et al. 2011), studies that specifically seek to quantify their prevalence in urban rat populations have occasionally found one or more of them to be absent (Yanagihara 1990, Ellis et al. 1999, Cueto et al. 2008, Inoue et al. 2008, Jiang et al. 2008, Eremeeva et al. 2012). In many cases, it is unclear if the pathogen is truly absent, or if absence is a result of undersampling (Yanagihara. 1990, Ellis et al. 1999, Inoue et al. 2008, Jiang et al. 2008, Eremeeva et al. 2012) and/or variability in disease detection methods (Klein et al. 2002, Jiang et al. 2008). When this uncertainty is coupled with the relative paucity of research on urban rats, the true distribution of rat-associated zoonoses becomes unclear.

From a public health perspective, it is important to know which rat-associated zoonoses are present in a specific geographic area, because this information is essential to developing appropriate surveillance and intervention strategies. However, it is equally important to know which pathogens are absent to avoid expending resources on programs targeting pathogens that are not a health risk in that particular jurisdiction. From a research perspective, accurate data on the distribution of RAZ is crucial for understanding the ecology of these pathogens. Finally, it is important to continuously update data on pathogen distribution because it may be subject to significant change over time, particularly given the fact that rats are prone to long-distance migrations in association with human transport (Wu et al. 2007). The objective of this study was to determine if rat populations in an inner-city neighborhood of Vancouver, Canada, carry R. typhi, SEOV, and Bartonella spp.

Materials and Methods

Rats were trapped in 43 contiguous city blocks and one location within the adjacent shipping port over the course of 1 year, as previously described (Himsworth et al. 2014). Briefly, trapping locations were randomized to a 3-week study period over the course of 1 year (September, 2011, through August, 2012). Within each city block, approximately 20 Tomahawk Rigid Traps for rats (Tomahawk Live Trap, USA) were set out along each side of the back alley that bisected the block, and at the port, traps were placed in areas where port staff had observed rats. Traps were prebaited for 1 week, followed by 2 weeks of active trapping. The majority of rats (n=693) were live-trapped, and blood was collected via intracardiac puncture under isoflurane anesthesia prior to euthanasia with human pentobarbital (Himsworth et al. 2014). Rats trapped at the port by a collaborating pest control professional (n=32) using snap-type kill traps were also collected. All rats underwent a complete necropsy with aseptic collection of kidney, lung, and spleen samples. The following measurements were collected from each rat: Date and location of trapping, species, mass (grams), body length (nose to anus, cm), sex, sexual maturity (animals were considered sexually mature if they had scrotal [vs. inguinal] testes or a perforate [vs. imperforate] vagina), presence and number of bite wounds in the skin, and volume of body fat (Himsworth et al. 2014).

For tissue-based PCR screening, 630 samples were considered to be the optimal sample size to calculate pathogen prevalence accurately in each city block (Himsworth et al. 2013b), although additional samples were included where resources allowed. The specific samples to be tested were selected randomly. Only 553 serum samples and 404 blood clot samples were considered to be of appropriate volume and quality for serology and Bartonella spp. culture, respectively. For this reason all serum and blood clots suitable for testing were included in subsequent analyses.

Serum samples were tested for the presence of immunoglobulin G (IgG) against R. typhi using an in-house immunofluorescence assay (Prabhu et al. 2011) at the National Microbiology Laboratory (NML), Winnipeg, Manitoba. Samples were applied to slides precoated with R. typhi antigen (CDC Viral and Rickettsial Zoonoses Branch) and incubated for 1 h at 37°C. The slides were then washed for 15 min in FTA buffer (BD Biosciences, Mississauga, Canada) and dried. Goat anti-rat IgG–fluorescein isothiocyanate (FITC) conjugate (Sigma Aldrich, Oakville, Canada) was applied to the slides, which were then incubated for 1 h at 37°C, washed, and dried. A reciprocal titer of ≥64 was used as the cutoff value suggestive of seroreactivity against typhus group Rickettsia (TGR) (which includes R. typhi and R. prowazekii, the latter of which is not known to infect Rattus spp.).

Serum samples were also tested at the NML for the presence of IgG against SEOV using the hantavirus IgG DxSelect Kit (Focus Diagnostics, Cypress, USA). The human conjugate provided in the kit was substituted with peroxidase-conjugated Protein G (Calbiochem, Darmstadt, Germany) to detect antibodies produced in rats. The conjugate was diluted 1:3000 in sample buffer, and all other steps were performed as per manufacturer's instructions. Index values for the rat sera were calculated relative to the kit cutoff calibrator. Enzyme-linked immunosorbent assay (ELISA) index values of 0.9–1.0 were considered “borderline” and values ≥1.1 were considered “reactive.”

To investigate SEOV seroreactive samples further, RNA was extracted from the lung tissue of seroreactive rats and random selection of seronegative rats (n=22) using the RNeasy Mini Kit (Qiagen, Toronto, Canada) and tested using qRT-PCR for rodent-borne hantaviruses (including SEOV) (Johansson et al. 2010). At the Animal Health Centre, British Columbia Ministry of Agriculture, Abbotsford, British Columbia, RNA was extracted from kidney samples (n=633) using the RNeasy Mini Kit (Qiagen, Toronto, Canada) and screened for hantaviruses using the same qRT-PCR method (Johansson et al. 2010). SEOV RNA and nuclease-free water were used as positive and negative controls, respectively.

At the NML, genomic DNA was extracted from spleen samples (n=678) using the NucleoSpin 96 Tissue Kit (Machery-Nagel, Bethlehem, USA) and the QIAxtractor (Qiagen, Toronto, Canada) and screened using a multiplexed qPCR assay for the detection of R. typhi and R. felis (Karpathy et al. 2009). R. typhi, R. felis, and rat glial cells were used as positive controls. To confirm that the extractions were successful and that there was no PCR inhibition, 10% of the samples were subjected to qPCR for rat β-actin.

Blood clots (obtained after serum separation) were sent to the Bartonella & Rodent-Borne Diseases Laboratory, Centers for Disease Control and Prevention, Fort Collins, CO, for Bartonella culture. Briefly, clots were resuspended 1:4 in brain heart infusion broth supplemented with 5% amphotericin B, plated on heart infusion agar containing 10% rabbit blood, and incubated in an aerobic atmosphere with 5% CO₂ at 35°C for up to 4 weeks. Bacterial colonies were presumptively identified as Bartonella spp. on the basis of colony morphology, and definitively identified by PCR amplification of the citrate synthase gene (gltA) (Ying et al. 2002, Bai et al. 2007). Sequencing of gltA was performed for Bartonella spp. identification (Ying et al. 2002, Bai et al. 2007).

Bivariate and multivariate generalized linear mixed models (GLMMs) controlling for clustering by city block of origin were used to identify factors associated with pathogen carriage. The unit of analysis was the individual rat and the outcome variable was pathogen carriage (positive vs. negative). Explanatory variables considered included sex, sexual maturity, weight, length, body fat, bite wounds, species, and season (Himsworth et al. 2014). For all models, individuals with missing values for any of the variables under consideration were excluded. The final multivariate model was arrived at by manual stepwise regression using Akaike information criteria (AIC) to compare candidate models (the final model was that with the lowest AIC) (Dohoo et al. 2007). Highly correlated variables (ρ>0.8) were modeled separately (Dohoo et al. 2007). The proportion of the variation in the outcome associated with the random effect of block was obtained from the GLMM, as previously described (Dohoo et al. 2007). All statistical analyses were conducted using R software (R Development Core Team, Vienna, Austria), and an alpha level of 0.05 was used to determine statistical significance.

Spatial analyses were used to visualize the distribution of seropositive and pathogen-positive rats among the city blocks and across the study area as a whole. Note that the port site was excluded from these analyses due to privacy concerns. The location of each trap and the number of rats caught in each trap positive and negative on each test (i.e., serology, culture, PCR) were mapped using ArcGIS 10.0 (ESRI, Redlands, CA). This information was imported into SaTScan (SaTScan, Boston, MA) for cluster analysis using a purely spatial Bernoulli model and scanning for areas with high and low rates of pathogen prevalence using a circular window with a maximum spatial cluster size of 50% of the population at risk. It should be noted that scanning spatial analyses of this nature are useful for detecting “high-level” clustering (i.e., areas of high and low prevalence across the entire study area, irrespective of city block). They may not have the resolution to detect block-level clustering, particularly if blocks of high and low prevalence are somewhat evenly dispersed. Block-level clustering is better detected through calculation of block-level prevalence and through the GLMM.

Results

The trapped population (n=725) consisted of 685 (95.3%) Norway rats and 40 (5.5%) black rats (Himsworth et al. 2014). Among the Norway rats, 56.3% were male, 63.9% were sexually mature, and the median body mass was 142.2 grams (Himsworth et al. 2014). Among the black rats, 47.5% were male, 61.8% were sexually mature, and the median body mass was 76.4 grams (Himsworth et al. 2014).

Two of 553 (0.36%) serum samples (located in blocks 6 and 8; Fig. 1) were reactive against TGR at a reciprocal titer of 64 (the cutoff value for reactivity). These samples originated from two Norway rats, one of which was a 159.6-gram sexually mature male, whereas the other was a 128.0-gram immature female. For SEOV, 8/553 (1.45%) serum samples (three and two in blocks 6 and 19, respectively, and one in each of blocks 8, 17, and 33) were borderline (n=2) or reactive (n=6) against SEOV at ELISA index values ranging from 0.92 to 8.51 (median=1.99). These rats had a median weight of 277.9 grams; 5/8 (62.5%) were male and 6/8 (75.0%) were sexually mature. There was no significant geographic clustering of either R. typhi or SEOV seroreactive samples. All PCR testing for R. typhi/R. felis and SEOV was negative.

OPEN IN VIEWER

FIG. 1.

Distribution of Bartonella spp.-positive rats (Rattus spp.) and clusters of high and low Bartonella spp. prevalence. The small circles indicate the location of traps in which at least one rat was caught, as well as the number of rats caught in each trap, and the number that were positive for Bartonella spp. The large circles indicate the clusters of high (red circle) and low (blue circle) Bartonella spp. prevalence across the study area as a whole. Observed versus expected number of Bartonella spp.-positive rats with relative risk and p values are noted for each cluster.

Bartonella spp. were isolated from 102/404 (25.2%) blood clots tested, although the prevalence varied significantly by city block, from 0% (95% confidence interval [CI] 0.0–9.4%) to 60.5% (95% CI 48.6–71.3%), and geographic clusters of higher and lower than expected prevalence were observed (Fig. 1). All Bartonella spp. isolated were identified as B. tribocorum and had identical gltA sequences. This sequence was identical to the predominant genotype found in urban rats from Los Angeles (JF429450) (Gundi et al. 2012).

In the final GLMM, the odds of being Bartonella spp. positive were associated with sexual maturity (odds ratio [OR]=5.0, 95% CI 1.36–18.06) and season (OR=0.16, 95% CI 0.05–0.51; OR=0.15, 95% CI 0.02–0.85; OR=0.24, 95% CI 0.09–0.69 for spring, summer, and winter vs. fall, respectively), and the proportion of variance associated with block of origin was 0.18. Body length was significantly associated with Bartonella spp. carriage on bivariate (OR=1.13, 95% CI 1.03–1.25) and multivariate (OR=1.13, 95% CI 1.01–1.26) analyses (in conjunction with season). However, length was highly correlated with maturity (ρ=0.81); therefore, these two variables were modeled separately. Maturity made a greater contribution to the final model (based on AIC) than body length. No other factors were significant in either bivariate or multivariate models.

Discussion

Bartonella spp., but not R. typhi or SEOV, were definitively identified in the populations of rats under study. Although spleen has been successfully used for detecting R. typhi infection in rats (Abramowicz et al. 2011), the period of active rickettsemia detectable by PCR can be transient (i.e., <1 month) (Azad 1990), and the bacterium can be maintained in fleas (the vectors for transmission) (Azad 1990, Azad and Beard 1998). For these reasons, negative PCR results alone cannot completely rule out the presence of circulating R. typhi. However, R. typhi infection appears to consistently elicit a long-term IgG immune response (for up to 40 weeks) (Azad 1990), suggesting that seronegative rats are not likely to be infected. In contrast, no vectors are involved in the ecology of SEOV, which is transmitted among rats via urine and other body secretions (Kariwa et al. 1998, Hinson et al. 2004). Rats often demonstrate persistent SEOV infection, particularly in the lungs and kidneys (Klein et al. 2002), despite the development of neutralizing antibody (Arikawa et al. 1994, Kariwa et al. 1996, Easterbrook et al. 2007b). However, hantavirus-seropositive rats, particularly those with low antibody titers, are occasionally found to be PCR negative (Johansson et al. 2010).

In this study, the R. typhi- and SEOV-seroreactive rats are somewhat difficult to explain given the very low prevalence of seropositivity, the lack of spatial clustering of seroreactive samples, and the absence of PCR evidence for pathogen presence. They could represent cross-reaction with related organisms or antigens, particularly in the case of R. typhi seroreactive samples, which were rare, weakly positive, and originated from small/young rats. The SEOV seroreactive samples, however, originated primarily from large, mature male rats, which would be the demographic group most likely to be infected (Hinson et al. 2004). Although we suspect that our intensive and systematic trapping protocol allowed us to collect a representative sample of Norway rats within the study area (Himsworth et al. 2014), a very low prevalence of SEOV infection cannot be ruled out. It should be noted that there were relatively few black rats in the study population; however, it is unclear whether this was a result of trapping-related bias or if it is a true reflection of the species composition in this area. Infection in adjacent rat populations with spillover into the study population is possible. However, few seroreactive samples were located at the periphery of the study area. Additionally, studies have shown that the home range of urban rats is often limited to a city block (Feng and Himsworth 2013), suggesting that mingling of rat populations in adjacent blocks is unlikely and/or infrequent. Spillover from other animal reservoirs is also unlikely. Seoul hantavirus has only been found in Rattus spp., and although Rickettsia typhi can be maintained in opossums (Civen and Ngo. 2008), opossums are not found within the city of Vancouver.

If it is the case that the rat populations included in this study are not endemically infected with R. typhi and SEOV, it could be that the physical and/or climactic environment within the study area is not conducive to the maintenance of these pathogens, or that appropriate vectors (in the case of R. typhi) are not present. However, it might also be the case that R. typhi and SEOV were never introduced into these rat populations in the first place.

It has been suggested that the geographic distribution of RAZ may be dependent on the dispersion and establishment of infected rats (Reynes et al. 2003, Bai et al. 2007). Given that both Norway and black rats, which originated in Asia, reached their current global distribution through numerous introduction events over time (Feng and Himsworth 2013), and given that rats are not prone to long-distance migrations in the absence human transport (Feng and Himsworth 2013), it could be the case that the presence of a zoonotic pathogen in a particular location is largely dependent on the characteristics of the founding or subsequently introduced rat populations. This theory is supported by the Bartonella spp. genetic homogeneity observed in this study. All isolates were found to be B. tribocorum and had identical gltA sequences. This is consistent with what has been found in other North American cities with invasive Rattus spp. (Billeter et al. 2011) and in contrast to the high Bartonella spp. diversity observed among Rattus spp. in Asia (Bai et al. 2007, Bai et al. 2009). It is interesting to note that the gltA sequence of the B. tribocorum from these rats was identical to that of the predominant Bartonella spp. circulating in rats from Los Angeles, another city on the west coast of North America (Gundi et al. 2012). That being said, more than one Bartonella sp. was identified in that study (Gundi et al. 2012), and R. typhi (Abramowicz et al. 2011) has been found in rats from Los Angeles. Genetic characterization of global rat population structures may help to elucidate the distribution of RAZ among different cities.

Similarly, spatial clustering of Bartonella spp. within the study could reflect local population genetic structures. Specifically, there was both significant variation in Bartonella spp. prevalence by city block (even after controlling for season and rat demographic/morphometric variables), and clusters of high and low prevalence were observed across the study area as a whole (irrespective of block). Additionally, the built environment within the study area is relatively homogeneous (high-density, mixed-use urban) and does not clearly account for observed heterogeneity in pathogen distribution. It is interesting to note that similar clustering was also observed for Leptospira interrogans carriage in these rat populations and likewise could not easily be explained by environmental factors or rat phenotypic characteristics (Himsworth et al. 2013b).

Given the paucity of data on the ecology of Bartonella spp. in rats, the results of statistical modeling are difficult to clearly explain. A positive correlation between mass and Bartonella spp. infection has been observed previously (Costa et al. 2014) and suggests that the probability of infection increases with age (likely due to cumulative exposure). In our study, positive associations with maturity and length might reflect a similar relationship. Another study also observed variations in Bartonella spp. prevalence among seasons, and attributed that finding to seasonal variations in the proportion of juvenile versus adult animals (Liu et al. 2010). In our study, however, both sexual maturity and season were independently associated with Bartonella spp. infection, suggesting that these relationships be more complex than previously appreciated. It is thought that Bartonella spp. is transmitted among rats by ectoparasites (Himsworth et al. 2013a); therefore, intra-annual variations in infection prevalence could be a result of seasonal differences in vector abundance.

Conclusion

Overall, these results suggest that both global and local rat population structures influence the ecology of RAZ, including Bartonella spp., R. typhi, and SEOV. There is very little information regarding the genetic relationships between Norway and black populations within and among different urban centers, and further study on the subject might help to elucidate patterns of disease distribution in these species. Additionally, the potential for future long-distance migrations of rats and their pathogens in association with human transport (Gage and Kosoy 2005) suggests that ongoing surveillance of urban rats is necessary to maintain an accurate and contemporary understanding of rat-associated health risks.