Rocky Mountain spotted fever (RMSF), caused by the bacterium Rickettsia rickettsii, was recognized as endemic in Arizona, US after a 2002 outbreak and has since been a public health concern. The brown dog tick (Rhipicephalus sanguineus sensu lato) is the principal vector of this pathogen in Arizona. Domesticated dogs (Canis lupus familiaris) are the tick's main host, so free-roaming dogs in peridomestic areas have been named the primary risk factor for human cases of RMSF. However, the sudden emergence and long-distance dispersal of the pathogen have not been adequately explained, and one possible mechanism could include wildlife. Coyotes (Canis latrans) are wide ranging in Arizona and closely related to dogs, so it is possible that brown dog ticks parasitize coyotes and infect them. Although R. rickettsii is the most severe spotted fever group (SFG) rickettsial pathogen in humans, others occur in Arizona, and antibodies raised against them are cross-reactive, so we more-broadly hypothesized that coyotes in Arizona are exposed to SFG rickettsiae. We collected coyote tissues in spring 2016 and 2017. We tested sera for antibodies to R. rickettsii and found 9% (8/94) of samples were antibody-positive with titers of ≥256. Subsequent quantitative PCR analyses of skin showed evidence for Rickettsia spp. in 2.9% (4/138) of samples. These data suggest that coyotes have a role in the maintenance of SFG rickettsiae in Arizona. Further investigation is warranted to reveal which specific pathogen-vector complexes act on coyotes in the region and whether they represent a risk to human health.
In 2002–04, Rocky Mountain spotted fever (RMSF) was diagnosed in 16 humans in two neighboring Native American communities in Arizona, US (Demma et al. 2005). Before 2002, RMSF had been considered of minor importance in the state because it was relatively rare. Only two cases were reported between 1981 and 1992 (Dalton et al. 1995), and six cases were reported between 1993 and 2002 (Treadwell et al. 2000); most of these cases were presumed to be associated with out-of-state travel (ADHS 2015a). Since the 2002–04 outbreak, RMSF has persisted and has become a greater public health concern within the state.
Caused by Rickettsia rickettsii, RMSF is a disease of the Americas that affects humans and is the most severe rickettsiosis in the world (Parola et al. 2013). With prompt diagnosis and treatment, the case fatality rate can range from 3% to 10% (Openshaw et al. 2010; Parola et al. 2013). Untreated, the case fatality rate is estimated at 20% (Nicholson et al. 2010) but has been reported as high as 38% in Mexico (Parola et al. 2013), 55% in Brazil (de Oliveira et al. 2016), and 100% in an outbreak in Panama (Parola et al. 2013).
Rickettsia is a genus of tick-borne bacteria that exhibit various levels of pathogenicity to vertebrates. Regardless of pathogenicity, members of Rickettsia are commonly placed into one of four phylogenetically distinct, antigenic-determinant groups: spotted fever group (SFG; e.g., R. rickettsii), typhus group (TG; e.g., Rickettsia typhi), transitional group (e.g., Rickettsia felis), and ancestral group (Rickettsia bellii; Kantsø et al. 2009; Parola et al. 2013; Brown and Macaluso 2016). Because R. rickettsii is the Rickettsia species of primary public health concern in Arizona and because we did not find reports of other groups in Arizona, except for the mention of R. felis in a booklouse (Liposcelis bostrychophila; Behar et al. 2010), we placed our emphasis on SFG Rickettsia.
Several other species within SFG rickettsia cause RMSF-like diseases that are collectively reported as SFG rickettsioses (Comer 1991; Parola et al. 2013). The US Centers for Disease Control and Prevention report these as a single category because antigenic characters between the bacterial species are difficult to differentiate for clinical diagnoses (Parola et al. 2013; Biggs et al. 2016). These pathogens are transmitted by all life stages of several species of hard-bodied ticks (Comer 1991; Biggs et al. 2016). In Arizona, the brown dog tick (Rhipicephalus sanguineus) is the principal vector of R. rickettsii (Demma et al. 2005) and also transmits Rickettsia massiliae (Parola et al. 2013). Rocky Mountain spotted fever is the most common rickettsiosis in Arizona (ADHS 2015b), but other SFG rickettsiae occur within the state, including Rickettsia parkeri and R. massiliae (Eremeeva et al. 2006; Allerdice et al. 2017), the latter having unknown pathogenicity in the US. The Gulf Coast tick (Amblyomma maculatum) transmits R. parkeri (Parola et al. 2013; Allerdice et al. 2017). However, dispersal across Arizona is not fully understood for these pathogens.
Unconfined, peridomestic dogs (Canis lupus familiaris) are the likely candidates for the dispersal of R. rickettsii–infected ticks within communities. Indeed, dogs are widely considered to be sentinels for RMSF (Gordon et al. 1983; Paddock et al. 2002), and experimental research has found that dogs are effective amplifier hosts of R. rickettsii for Rhipicephalus sanguineus ticks (Piranda et al. 2011). Although sentinel surveillance of dogs is useful in modeling the risk of human infections within communities, it reveals little about the origin and long-distance dispersal of a pathogen among communities. The relocation of infected or tick-infested domestic dogs could explain the sudden emergence of RMSF in Arizona and its translocation among distant communities (Demma et al. 2006; McQuiston et al. 2011), but this hypothesis has not been rigorously tested. Long-distance dispersal of brown dog ticks and SFG rickettsia could also be explained by the ticks parasitizing wide-ranging wildlife that might associate with peridomestic dogs (Demma et al. 2006). A prime candidate for that scenario is the coyote (Canis latrans). Antibodies for SFG rickettsia have previously been found in coyotes in Texas, Oklahoma, and Nebraska (Bischof and Rogers 2005; Starkey et al. 2013) but not in Arizona.
We tested whether coyotes act as hosts for SFG rickettsia in Arizona. If coyotes are exposed to SFG rickettsiae, then the nature of rickettsial pathology must be assessed to determine whether coyotes may amplify or dilute the pathogen's prevalence in nature and whether the coyotes could mediate the long-distance dispersal of infected ticks. Exposure to SFG rickettsiae could also make urbanized coyotes useful in sentinel surveillance for SFG rickettsioses.
MATERIALS AND METHODS
To test the hypothesis that coyotes in Arizona are exposed to SFG rickettsiae, we undertook an indirect immunofluorescence antibody (IFA) assay on coyote serum samples and genetic screening of DNA from skin samples using quantitative PCR (qPCR). We also examined coyote carcasses for ticks but found no ticks in 2 yr of sampling.
In April and May of 2016 and 2017, we collected cardiac blood and skin samples opportunistically from 138 coyotes harvested for predation management throughout Arizona. Collection locality (GPS coordinates or county), date of collection, sex, and age class (juvenile <1 yr old; adult ≥1 yr old) were recorded for each sample. In 2016, samples came from eastern Arizona (Apache, Cochise, and Graham counties). In 2017, samples came from northern, central, and eastern Arizona (Cochise, Coconino, Graham, Mohave, Navajo, and Yavapai counties). A total of 138 skin samples (66 in 2016 [48%]; 72 in 2017 [52%]) were cut from the bottom or side of the ear and placed into 1.5-mL microcentrifuge tubes. We collected 125 blood samples (53 in 2016 [42%]; 72 in 2017 [58%]) via cardiac puncture with disposable syringes and placed the blood in Corvac integrated serum separator tubes (Medtronic, Minneapolis, Minnesota, USA) containing a coagulation factor and acrylic gel barrier. All samples were collected postmortem and stored at 4 C in the field. In the laboratory, we separated sera from cells via centrifugation (15 min at 1,342 × G) and placed the sera into microcentrifuge tubes. Centrifugation yielded 94 usable serum samples (53 in 2016 [56%], 41 in 2017 [44%]); the remaining 31 samples (27%) were not used for IFA assays because sera did not separate from the cellular components of the blood. All tissue specimens were frozen at –20 C until assays or DNA extraction could be performed.
Rickettsia spp. IFA
We screened serum for antibodies raised against Rickettsia spp. with commercial RMSF IFA-substrate microscope slides designed for veterinary screening (Veterinary Medical Research and Development, Pullman, Washington, USA) per the manufacturer's technical instructions. An affinity-purified, polyclonal, anticanine immunoglobulin G (IgG) of rabbit origin was used as the secondary antibody in this assay. The secondary antibody is specific for canine IgG and is conjugated to a fluorescein isothiocyanate fluorescent label. To ensure that all reagents were working properly, we included positive and negative controls on each slide. The positive control consisted of canine serum that was positive to RMSF diluted in phosphate-buffered saline, 10% bovine serum, and 10 ppm ProClin 300 (Veterinary Medical Research and Development) as a reagent preservative. The negative control was canine serum diluted in phosphate-buffered saline, 10% bovine serum, and 0.09% sodium azide. Per the manufacturer's instructions, the controls were used undiluted. All serum samples were initially screened after 1:64 dilution with serum-diluting buffer. We performed serial twofold dilutions on specimens that were positive at 1:64 to determine the endpoint titer, henceforth, expressed as the reciprocals of the dilutions.
Samples of DNA extracted from ear biopsies were screened with a hybridization probe-based qPCR (TaqMan, Applied Biosystems, Life Technologies, Carlsbad, California, USA). After cutting up the skin samples with a scalpel to increase surface area, we extracted DNA from skin cells with DNeasy extraction kits (Qiagen, Inc., Valencia, California, USA), with minor modifications, including extending the initial incubation period from 10 min to overnight. In addition, during the elution step, 50 µL of sterile deionized water replaced the AE buffer (Qiagen, Inc.), samples were incubated with the elution solution for 5 min at 56 C before being centrifuged, the flow-through was re-eluted on the filter, incubated, and centrifuged again for optimal DNA yield. Samples from 2017 were extracted with a slightly different protocol: they were disrupted with glass beads during vigorous shaking in a Geno/Grinder® (Thomas Scientific, Swedesboro, New Jersey, USA) before the overnight incubation.
To screen total DNA for Rickettsia spp., we used a stepwise approach with two previously designed, real-time TaqMan assays. The first used primers and probes to target the 23S rRNA gene to detect the genus Rickettsia. The second targeted the A1G_04230 gene, which is specific to R. rickettsii (Kato et al. 2013). Each 20-µL reaction consisted of 3.6 µL of template DNA and 10 µL of master mix (SsoAdvanced Universal Probes Supermix, Bio-Rad Laboratories, Inc., Hercules, California, USA), with primers at a final concentration of 300 nM and probes at 250 nM (Applied Biosystems, Life Technologies). We included at least two negative controls (deionized H2O) on each plate. We then followed the two-step, thermal-cycling protocol recommended by the manufacturer with the following steps: 3 min at 95 C, followed by 40 cycles of 95 C for 10 s and 55 C for 30 s. All reactions were run with a CFX96 Touch real-time system (Bio-Rad). Samples that had a quantification threshold (Ct) value <40 and plots that demonstrated a logarithmic curve were considered positive. To confirm positive samples, each positive was rerun in duplicate.
Any samples that were qPCR-positive for SFG rickettsia, but not for R. rickettsii, were selected for further testing. For identification of the bacterial species, we performed nested PCR tests to amplify the 17-kDa and outer membrane protein A (ompA) genes of the SFG rickettsia genome (Kato et al. 2013). Successful amplification of the respective nested products, as determined by 208- and 540-base pair fragments in agarose gel electrophoresis (Paddock et al. 2004), was followed by Sanger sequencing of the PCR product. This was conducted on an ABI 3730 DNA Analyzer (Thermo Fisher Scientific, Waltham, Massachusetts, USA) at the Environmental Genetics and Genomics core facility (Northern Arizona University, Flagstaff, Arizona, USA), following the laboratory's standard protocol (Northern Arizona University 2017).
To characterize the distribution of positive titers in the IFA, we ran initial analyses with the manufacturer's diagnostic cutoff titer of 64. However, that is a clinical recommendation based on the comparison of titers from acutely ill patients (at presentation of clinical signs) and convalescent serum samples (collected 2–4 wk later), in which a primary infection is confirmed by a fourfold increase in endpoint titer. Because all our samples were collected postmortem, that was not possible. This limitation could have resulted in more false positives, so we used a conservative cutoff titer of 256, based on the natural decline in positives by titer (Figure 1).
We conducted statistical analyses based on both cutoff titers. First, we calculated summary statistics for apparent prevalence, range of positive titers, geometric mean titer (x̄G; Petrie and Watson 2013), and a geometric SD factor (SG; Taylor 1983). Next, we compared titer distribu-tions between years with a two-tailed Fisher's exact test using only those samples with a titer of 256 or greater. To compare positive and negative samples against sampling year, age class, and sex of the donor animal, we used chi-square contingency tables for the 1:64 dilution cutoff. We also used a Fisher's exact test on the 1:256 cutoff to analyze titer distributions based on year, age class, and sex. Fisher's exact test was used instead of the chi-square test whenever the expected frequency for any cell in a contingency table was less than five (Petrie and Watson 2013). All data analyses were conducted in R version 3.2.3 (R Foundation for Statistical Computing, Vienna, Austria), within RStudio version 1.0.136 (RStudio, Inc., Boston, Massachusetts, USA).
Rickettsia spp. IFA
At the manufacturer's recommended cutoff titer of 64, 26% (14/53; 95% confidence interval [CI], 16–41%) of the samples from 2016 were antibody positive (x̄G=115.93; SG=2.35), and 44% (18/41; 95% CI, 29–60%) of the samples from 2017 were antibody positive (x̄G=161.27; SG=3.05). Overall, 34% (32/94; 95% CI, 25–45%) of the samples were antibody positive (titer range, 64–4,096; x̄G=139.59; SG=2.74). Chi-square tests showed no difference because of sampling year (P=0.120), sex (P=0.827), or age class (P=1.000). Fisher's exact test showed no significant differences in titer distributions between sampling years (P=0.265).
At the adjusted cutoff titer of 256, 9% (8/94; 95% CI, 4–17%) of the samples were antibody positive (titer range, 256–4,096; x̄G=608.87; SG=2.43). Fisher's exact test showed no significant difference in prevalence by year (P=0.290), sex (P=1.000), or age class (P=0.343). Similarly, the titer distributions were not significantly different by year (P=0.286), sex (P=0.257), or age class (P=1.000).
Meaningful spatial statistics were not feasible with the available data, but positive samples at the 1:64 cutoff dilution were found in all counties sampled, except for Mojave and Navajo counties. At the 1:256 cutoff, positive samples were found in Apache (2016), Coconino (2017), Cochise (both years), and Graham (2016) counties (Table 1).
Initial qPCR assays resulted in four samples that amplified in the pan-rickettsia assay, all with high Ct values (range, 34–38). No samples were positive for R. rickettsii. The repeated qPCR run of the pan-rickettsia–positive samples resulted in one triplicate and three duplicate positive samples (Ct range, 34–38.5). Three of these four samples (75%) were from animals on which serology was also performed; one sample had an antibody titer of 64, and two samples had antibody titers of 128. This suggested that some low-titer samples might have been from positive animals, even with titers that fell below our 1:256 cutoff. Nested PCR yielded additional nonspecific amplification products that interfered with sequencing of the four qPCR-positive samples. As such, Sanger sequencing did not yield any results that could be matched to entries in GenBank.
Of the four DNA samples that qualified as qPCR positive, one was collected in Cochise County, about 60 km south of an RMSF-affected community. The other three were collected in a cluster less than 10 km from each other, on the Coconino-Navajo county line, located roughly equidistantly between two affected communities, which are 41.6 km north and 49.5 km south of the cluster, respectively.
We hypothesized that coyotes are exposed to SFG rickettsia in Arizona. The serologic and genetic evidence presented here support our hypothesis. In addition, the similar level of antibody prevalence and titer between years and sexes suggests that SFG rickettsiae are endemic in Arizona's coyote population. However, our data only cover the spring seasons of two consecutive years. Confirmation of endemism would require more widespread sampling in multiple seasons over several years.
We are confident that our data provide evidence of SFG rickettsiae in coyotes. Although cross-reactivity in serology can sometimes occur between SFG and TG rickettsiae (Ormsbee et al. 1978), most sources treat cross-reactivity as only occurring within antigenic determinant groups (Parola et al. 2013). In addition, although the pan-rickettsia qPCR primer set is sensitive to all species in the genus (Kato et al. 2013), reports of TG rickettsiae in Arizona are rare. On the other hand, we acknowledge that although the three replicates helped us to rule out false positives in qPCR, false negatives might exist in the samples that were only run once.
Antibody-based detection is a powerful approach for detecting a pathogen in wildlife (La Scola and Raoult 1997). Immunofluorescence has been criticized for its lack of specificity in some cases (Parola and Raoult 2001); however, a review of guidelines for diagnosing rickettsioses in humans found that a cutoff titer of 64 for the R. rickettsii IFA resulted in a sensitivity and specificity of 84.6% and 100%, respectively, in clinically relevant samples (Brouqui et al. 2004). In addition, a comparative analysis of serologic tests has shown that increasing the cutoff titer for positive diagnoses improves specificity (Kantsø et al. 2009). A known prevalence value for Rickettsia spp. antibodies in coyotes is not available, so we were not able to calculate the sensitivity and specificity of our test. However, because specificity increases with higher titers, our adjusted cutoff titer of 256 should have reduced the number of false positives in our assay.
The range of antibody-positive titers we found is comparable to those found in experimental R. rickettsii infections of dogs that developed clinical illness sometimes resulting in death (Levin et al. 2014). Levin et al. (2014) experimentally infected dogs with R. rickettsii and monitored IgG antibody titers daily. Detectable titers began 7–10 d postinfection, rose to a maximum of 2,048, and began falling 28–33 d postinfection (Levin et al. 2014). Although the relationship between the infection kinetics and antibody titer for coyotes is not known, our data are similar to what is seen in dogs. In the three qPCR-positive samples we were able to match with serology from the same animals, the titers were relatively low (range, 64–128). These titers were below our adjusted cutoff, but they are at levels we would expect from an early primary infection if rickettsial pathophysiology is similar in coyotes as it is in humans (Biggs et al. 2016) and the dogs reported by Levin et al. (2014). Although our samples were likely taken from acutely infected animals, it is also possible for SFG rickettisae to be detectable at low titers for several years after infection (Mansueto et al. 1985).
Our data suggest that coyotes are exposed to SFG rickettsiae in Arizona, but the specific ecologic relationships among coyotes, ticks, and rickettsial pathogens remain inadequately explored. This is a complex problem, given that Arizona is home to several tick species that may parasitize coyotes and transmit multiple SFG rickettsiae that have varying degrees of pathogenicity. For example, a small proportion of SFG rickettsiosis cases in Arizona residents are attributed to R. parkeri (Allerdice et al. 2017), which presents with similar symptoms to RMSF and is commonly associated with the Gulf Coast tick (Parola et al. 2013; Starkey et al. 2013; Chitwood et al. 2015). In the eastern US, there are multiple accounts of Gulf Coast ticks parasitizing coyotes. In Arizona, those ticks have recently been found infected with multiple SFG rickettsiae (Allerdice et al. 2017). Given that ticks of the A. maculatum group do not appear to have a host preference (Estrada-Peña et al. 2005), nondog vertebrate hosts seem likely in the wild. Further, experimental evidence has shown that brown dog ticks attached more readily to humans when exposed to high temperatures (Parola et al. 2008). We find it plausible that such a decrease in host specificity could also lead to infestations of brown dog ticks on coyotes because coyotes are more closely related to dogs than humans are. Finally, the immunologic cross-reactivity among rickettsial species further complicates our understanding of this system. For instance, several laboratory studies have shown that non- or mildly pathogenic species, such as Rickettsia montana, Rickettsia amblyommii, and Rickettsia rhipicephali can induce protective immunity against R. rickettsii in guinea pigs (Feng and Waner 1980; Gage and Jerrells 1992; Rivas et al. 2015).
In addition to the limitations discussed, our study could have been improved with more-rigorous sample collection and storage procedures. For instance, blood samples could not be immediately centrifuged in the field, so hemolysis occurred in some samples. We also collected specimens opportunistically; therefore, we did not have equal geographic representation of samples or any data directly from RMSF-affected communities.
Although we do not have sufficient evidence to classify coyotes as bridge hosts for SFG rickettsiae, there is a need for additional research on the role of wildlife in the maintenance of Rickettsia spp. in Arizona. Potential topics of interest for future studies include pathogen-vector-host dynamics, cross-reactive adaptive immune responses, and the potential for wildlife-mediated pathogen dispersal into peridomestic habitats or across political borders. We suggest that follow-up studies include more-robust laboratory methods, such as competitive IFA and the performance of a southern blot before Sanger sequencing. Washing skin samples before DNA extraction should reduce the probability of exogenous contamination by flea feces, which could produce false positives in qPCR because of the presence of R. felis. In addition, conducting specimen collection efforts in the months during which ticks are most active might yield a greater prevalence of infected tissues for analyses and reveal which tick species parasitize coyotes in the area.
We thank Joseph Bahe for generating hypotheses and inspiring this study; Aubrey Funke, assistant director of the Northern Arizona University Imaging and Histology Core Facility, for the use of the fluorescence microscope; and Lela Andrews, manager of the Northern Arizona University Environmental Genetics and Genomics Laboratory, for the use of DNA sequencing equipment. We thank Nashelly Meneses, Julie Wachara, Erik Settles, and Thomas Lowrey for their valuable advice and 33 private donors on our https://experiment.com crowd-funding campaign. Four anonymous reviewers greatly improved the quality of an earlier draft. This work is dedicated to the late Nathan C. Nieto.
6 Current address: University of Regina, Department of Biology, Laboratory Building LB109, 3737 Wascana Parkway, Regina, Saskatchewan S4S 0A2, Canada