Trypanosoma (Herpetosoma) diversity in rodents and lagomorphs of New Mexico with a focus on epizootological aspects of infection in Southern Plains woodrats (Neotoma micropus)

Protozoan parasites of the genus Trypanosoma infect a broad diversity of vertebrates and several species cause significant illness in humans. However, understanding of the phylogenetic diversity, host associations, and infection dynamics of Trypanosoma species in naturally infected animals is incomplete. This study investigated the presence of Trypanosoma spp. in wild rodents and lagomorphs in northern New Mexico, United States, as well as phylogenetic relationships among these parasites. A total of 458 samples from 13 rodent and one lagomorph species collected between November 2002 and July 2004 were tested by nested PCR targeting the 18S ribosomal RNA gene (18S rRNA). Trypanosoma DNA was detected in 25.1% of all samples, with the highest rates of 50% in Sylvilagus audubonii, 33.1% in Neotoma micropus, and 32% in Peromyscus leucopus. Phylogenetic analysis of Trypanosoma sequences revealed five haplotypes within the subgenus Herpetosoma (T. lewisi clade). Focused analysis on the large number of samples from N. micropus showed that Trypanosoma infection varied by age class and that the same Trypanosoma haplotype could be detected in recaptured individuals over multiple months. This is the first report of Trypanosoma infections in Dipodomys ordii and Otospermophilus variegatus, and the first detection of a haplotype phylogenetically related to T. nabiasi in North America in S. audubonii. This study lends important new insight into the diversity of Trypanosoma species, their geographic ranges and host associations, and the dynamics of infection in natural populations.


Introduction
Trypanosoma (class Kinetoplastida) is a diverse genus of flagellate blood parasites found in all groups of vertebrates [1]. Approximately 500 Trypanosoma species in dozens of lineages have been characterized and named based on morphological features, host associations, and DNA sequencing [2][3][4][5][6][7]. Trypanosomes are transmitted between vertebrate hosts by blood-feeding ectoparasites and micropredators. Trypanosomes of mammals fall into two biological groups based on the development of parasites in the vector and the mode of transmission from vector to host. The salivarian trypanosomes, including the causative agent of African sleeping sickness in humans, T. brucei, develop in the midgut of an insect vector, migrate to the salivary glands or proboscis, and pass to the host in the saliva. The stercorarian trypanosomes, including the agent of Chagas disease in the Americas, T. cruzi, develop in the hind gut of the vector and are transmitted through ingestion of the vector by the host or contamination of bite wounds with vector feces [1,8]. Trypanosomes are also grouped based on their pathogenicity in their hosts, with some species causing significant disease in humans and domestic animals (e.g., T. brucei, T. congolense, T. brucei) and others appearing to cause no disease in their hosts [9,10]. Finally, trypanosomes are classified into subgenera and other clades based on phylogenetic analysis of molecular sequences [2][3][4][5][6][7]. The stercorarian trypanosomes in the subgenus Herpetosoma (T. lewisi clade) are found in numerous rodent and lagomorph species globally [11]. At least 30 species from the clade have been named based on assumed host-specificity, but many species names require validation with DNA sequencing because they are morphologically similar [12][13][14]. While this group is often considered to be non-pathogenic [1], T. lewisi and T. lewisi-like parasites have been linked to illness in some rodent hosts [9,15] or in humans that become infected through contact with hosts or vectors [16][17][18]. Given the zoonotic potential of these parasites, there is a need to characterize the diversity of Trypanosoma species in vertebrate hosts and understand the ecology of these parasites within natural systems where humans could be exposed [19].
Within the United States, there is an added concern about the circulation of T. cruzi in natural hosts and the potential for zoonotic transmission. Cases of Chagas disease linked to autochthonous vector-borne transmission have been reported in humans in several western and southern states (California, Arizona, Texas, Louisiana, Mississippi, Tennessee), but there is limited information on the circulation of this species or other trypanosomes in potential reservoir hosts, especially rodents and lagomorphs [20,21]. T. cruzi has been reported in Neotoma spp. woodrats in California, Arizona, New Mexico, and Louisiana [22 -28], but it is unclear whether the methods used in these studies could have detected other Trypanosoma species. T. lewisi-like trypanosomes T. neotomae and T. kansasensis have also been observed in blood from woodrats in California and Kansas [29,30], but these parasites were only characterized morphologically and sequence information is therefore lacking. Surveys in Alaska have detected DNA sequences similar to T. microti in voles and lemmings in Alaska [31]. Finally, T. nabiasi infects European rabbits (Oryctolagus cuniculus) in Portugal, Spain, France, the UK, and Italy, and was introduced to Australia through the importation of O. cuniculus or its fleas [1,[32][33][34]. However, surveys of trypanosomes in rabbits in the US using molecular methods are yet to be published. Therefore, more surveys of trypanosomes in wild rodent and lagomorphs are necessary for surveillance of T. cruzi and to understand the prevalence and phylogenetic diversity of Trypanosoma parasites in this region.
As part of a surveillance study on vector-borne pathogens in New Mexico, the objectives of the study were to use molecular methods to estimate the prevalence of Trypanosoma infection in rodents and lagomorphs and evaluate the genetic diversity and host associations of detected Trypanosoma haplotypes. We hypothesized that multiple rodent species would be infected with T. cruzi or other trypanosomes within the T. lewisi clade and that haplotypes would exhibit a host-specific pattern, infecting either a single host or multiple related hosts. Previous studies on Bartonella infection in this same community showed high prevalence of infection in multiple species [35,36], so we hypothesized that immune suppression associated with Bartonella infection might lead to a higher likelihood of Bartonella-infected animals carrying trypanosomes. Therefore, we analyzed coinfection rates across species. Finally, due to the large number of samples collected from one species, Neotoma micropus, we examined whether characteristics such as age, sex, flea burden, or Bartonella infection could explain variation in the probability of Trypanosoma infection across individuals. This study reveals previously unknown diversity of Trypanosoma parasites in New Mexico and provides new information on the infection dynamics of trypanosomes in N. micropus.  [35]. The trapping sites were in pinyon-juniper woodlands in a suburban area nearby to homes and a recreational hiking area. For the purpose of broad surveillance of pathogens in rodents and lagomorphs, convenience sampling was implemented using a variety of traps: (1) small Sherman traps (2"x2"x6.5") for mouse-sized rodents (HB Sherman Trap Company, Tallahassee, FL), (2) large Sherman traps (3"x3.5"x9") for rat-sized rodents, and (3) Tomahawk traps (4"x4"x10") for squirrel-sized rodents and rabbits (Tomahawk Live Trap Company, Hazelhurst, WI). Trapping locations were chosen based on prior identification of woodrat nests. Traps were baited with oats and peanut butter in the afternoon and checked the following morning, and any captured animals were then processed at a centralized location.

Sample collection and DNA extraction
Animals were anaesthetized with isoflurane and a small sample of blood was collected (100-200 μl) by retro-orbital bleed with heparinized microhematocrit capillary tubes (MWI Veterinary Supply Company, Denver, CO) and stored in EDTA-treated tubes. Animal species were distinguished morphologically using available field guides [37] and keys developed by the University of New Mexico Museum of Southwestern Biology for the study area. Additionally, data on sex, weight, and age class were recorded for each animal. Captured animals were marked with ear tags and/or subcutaneous transponders (AVID, Folsom, LA) and then released at the location of capture. All animal handling procedures were approved by the Centers for Disease and Control and Prevention Division of Vector-Borne Diseases (CDC DVBD) Institutional Animal Care and Use Committee, protocol number 06-008. Blood samples were placed on dry ice in the field and then kept at -80˚C in the laboratory until processing. Genomic DNA was extracted from 100 μl of blood using the KingFisher Flex Purification System and the associated MagMax Pathogen RNA/DNA Kit (ThermoFisher, Waltham, MA) according to the manufacturer's protocol. The DNA was eluted in 90 μl of elution buffer and stored at 4˚C while samples were being tested. The remaining blood samples have been archived at the CDC DVBD in Fort Collins, CO.

Trypanosoma spp. detection, sequencing, and phylogenetic analysis
Two PCR protocols were used to test for the presence of Trypanosoma DNA in samples. We first tested samples for T. cruzi DNA with species-specific qPCR targeting the kinetoplast minicircle [38]. The protocol for qPCR used the forward primer Cruzi 1: ASTCGGCTGATCG TTTTCGA, the reverse primer Cruzi 2: AATTCCTCCAAGCAGCGGATA, and the probe Cruzi 3: CACACACTGGACACCAA. Positive (T. cruzi) and negative (deionized water) controls were included in all reactions to ensure that PCR worked properly. A second test to detect Trypanosoma spp. DNA targeting a fragment (about 530 bp) of the small subunit ribosomal RNA gene (18S rRNA) used external primers TRY927F: CAGAAACGAAACACGGGAG and TRY927R: CCTACTGGGCAGCTTGGA and internal primers SSU561F: TGGGATAACAAAGGAGCA and SSU561R: CTGAGACTGTAACCTCAAAGC [39]. The 18S rRNA gene was chosen because previous analyses have demonstrated that this marker is adequate for phylogenetic differentiation of Trypanosoma species [40]. Multiple studies have used these primers to amplify Trypanosoma DNA from clades across the Trypanosoma phylogeny associated with mammals, sauropsids, amphibians, and fish [39,[41][42][43][44]. A study by Hodo et al. [45] was also able to amplify DNA from the related trypanosomatid genus Blastocrithidia in the blood of bats with these 18S rRNA primers. Therefore, we assumed that the nested primers were sufficiently sensitive to amplify a broad diversity of trypanosomes that might be present in our samples. T. lewisi DNA was used for the positive control for the 18S rRNA nested PCR and deionized water was the negative control. No internal control targeting host DNA was used to identify false negatives in PCR tests. Extractions and PCR were performed in separate laboratories to minimize contamination.
Amplified fragments of the 18S rRNA gene were separated in 1.5% agarose gels stained with Biotium GelGreen (Biotium, Hayward, CA). Positive amplicons were purified using the QIAquick PCR Purification Kit (QIAGEN, Valencia, CA) according to the manufacturer's protocol and sequenced in both directions with the internal primers (561F, 561R) on an Applied Biosystems Model 3130 Genetic Analyzer (Applied Biosystems, Foster City, CA). We assembled forward and reverse sequences using the SeqMan Pro program in Lasergene v12 (DNASTAR, Madison, WI). BLAST search was used to check whether the obtained sequences belonged to trypanosomes and to collect similar sequences from the GenBank database for the phylogenetic analysis. Amplified sequences from samples and reference sequences were aligned with MAFFT v7 [46] using the iterative local alignment method with generalized affine gap costs (E-INS-i) with default settings (https://mafft.cbrc.jp/alignment/server/). The alignment was manually inspected and adjusted, trimming the ends of sequences to match the newly obtained 18S rRNA sequences.
Novel Trypanosoma haplotypes were identified based on comparison with other sequences from this study and existing reference Trypanosoma sequences via maximum likelihood (ML) phylogenetic analysis using MEGA v7.0.26 [47]. An initial ML tree was estimated using the GTR+I+G model with four gamma categories. Identical sequences were labeled as distinct haplotypes and the sequence alignment was reduced to a single sequence per haplotype for final phylogenetic analysis. Phylogenetic model selection and phylogenetic reconstruction were performed with IQ-Tree v2.1.1 using the final alignment [48,49]. The top-ranking model based on the Akaike information criterion with a correction for finite sample sizes (AICc) was chosen for ML phylogenetic analysis [50,51]. The top model for the 18S rRNA alignment was a transition model with unequal base frequencies and two substitution rate categories (TIM2 +F+R2). Branch support across the phylogeny was estimated with 1000 ultrafast bootstrap replicates [52]. The consensus tree was viewed and annotated using the ape, phytools, and ggtree packages in R v4.0.3 [53][54][55][56].

Trypanosoma-Bartonella coinfection
To assess the occurrence of coinfection by Trypanosoma parasites and Bartonella bacteria in captured animals, we included information on the presence of Bartonella DNA in collected samples from our previous study [36]. In brief, Bartonella DNA was detected using a genusspecific qPCR targeting the Bartonella transfer-messenger RNA gene (ssrA) [57] and conventional PCR targeting the 16S-23S rRNA gene intergenic transcribed spacer (ITS) [58]. Samples that were positive by ssrA or ITS were tested by nested PCR for the citrate synthase gene (gltA) [59,60], followed by sequencing and phylogenetic analysis of amplicons. Samples were considered Bartonella positive if they tested positive for at least two out of three targets (ssrA, ITA, and gltA) and were successfully sequenced.

Statistical analysis
The prevalence of Trypanosoma and Bartonella DNA in tested samples was estimated for each species across all months using the number of infected animals out of the total sampled between November 2002 and July 2004. Our convenience sampling resulted in numerous recaptures of individual animals over multiple months. We did not remove or attempt to randomize the selection of samples collected from recaptured animals when estimating Trypanosoma or Bartonella DNA prevalence by host species. Confidence intervals (95% CI) for Trypanosoma prevalence in host species over each month when a species was captured were calculated with Wilson score intervals [61]. Host specificity of Trypanosoma haplotypes identified in positive animals was examined by plotting the relative abundance of each haplotype (i.e., the percent of all infections attributed to a haplotype) by host species. Differences in the Trypanosoma prevalence between species were analyzed using a chi-square test of proportions in R [55]. For all statistical tests, α < 0.05 was used as the cutoff for statistical significance.
We investigated the effects of temporal and ecological predictors on Trypanosoma prevalence in N. micropus because it was captured in all months and resulted in a large sample size for this species; other species had insufficient samples for such analysis. We performed model selection to choose the best fitting set of covariates that explained variation in prevalence. Within this model selection procedure, we also performed stratified sampling by individual ID number (without replacement) to choose only one sample from recaptured animals. The joint procedure involved stratified sampling of N. micropus individuals followed by model selection based on AICc [51] using the R package MuMIn [62] and was repeated 1000 times to account for individual random effects. Model selection was performed on fitted generalized linear models (GLM) using covariates of infection at the time of capture including date of capture, sex, age class (adult, subadult, or juvenile), body weight, the presence/absence of fleas, flea count, and Bartonella infection status. We included these predictors based on previous studies of rodent-borne trypanosomes that found relationships between Trypanosoma infection and age, weight, and flea burden [19,[63][64][65][66][67]. Our analysis of Bartonella infection in this same population of N. micropus found a significant interaction effect of sex and weight [36]. Therefore, we added this interaction as a potential covariate. Information on all individual covariates can be found in S1 Dataset. The mean AICc values were then calculated over all 1000 iterations of resampling for each model including different combinations of main effects and interaction effects. The model with the lowest mean AICc was chosen as the top model for explaining variation in prevalence.
Using the top GLM, we then estimated the probability of infection for N. micropus individuals based on the covariates included in the model. For this procedure, we again performed stratified sampling by individual tag number before fitting the GLM and estimated 95% CI for fixed effects in the model. The fit of the model was assessed using the coefficient of determination (R 2 ) and the area under the receiver operating characteristic curve (AUC) using the R package pROC [68]. The process of stratified sampling and model fitting was repeated 1000 times and mean values were calculated for the probability of infection, upper and lower 95% CI, R 2 , and AUC.
The prevalence of coinfection with Trypanosoma and Bartonella was analyzed using a binomial test comparing the observed coinfection rate and an expected rate based on the product of the Trypanosoma and Bartonella infection rates. This test was performed on all species across all months and on separate species across all months. We also ran tests on those species with adequate sampling across dates (greater than one sample in a month and infection with Trypanosoma, Bartonella, or both).

Patterns of Trypanosoma DNA prevalence
Trypanosoma DNA was detected in 115 (25.1%) of the 458 samples tested over the whole study period, including samples from individuals recaptured in multiple months (Table 1). Trypanosoma DNA prevalence was lower than Bartonella DNA prevalence (66.4%) across all species and dates. Positive species included N. micropus, Peromyscus leucopus, P. maniculatus, D.  (Table 1). Considering all sampled species, there was a significant difference in Trypanosoma DNA prevalence among species over the whole study period (χ 2 = 42.8, df = 13, p-value < 0.001). Within positive species, detection of Trypanosoma DNA was not consistent over time due to variation in the number of samples collected in a month (Fig 1). Detection was more consistent for N. micropus because of the larger number of samples collected from this species across each month, although there was more variation after October 2003 when fewer samples were collected (Fig 1). Given the large number of samples collected and tested from N. micropus, we examined available covariates that might explain variation in the probability of Trypanosoma infection across captured individuals over the whole study period. We used a combination of stratified sampling by individual tag number and model selection by AICc to determine the best fitting combination of covariates in the model. The model with the lowest mean AICc after 1000 iterations of stratified sampling and model selection contained only age class as a covariate (S1 Table), indicating that subadults were more likely to be infected than juveniles or adults (Fig 2; S2 Table). However, the amount of variation in the probability of Trypanosoma infection among N. micropus explained by the model with age class was low, with a mean R 2 over the

PLOS ONE
1000 iterations of only 7.4%. The predictive power of the model was also low, with a mean AUC of 0.592, meaning that the ability of age class to predict the presence of Trypanosoma DNA was only marginally better than random.
The effect of higher prevalence in subadults was also observed to a small degree in the variation in prevalence in the whole N. micropus population over the study period. In three months when higher prevalence was observed (December 2002, June 2003, and September 2003; Fig  1), more than 14% of the captured individuals were subadults. However, the correlation between the prevalence and proportion of subadults across all months was positive but not statistically significant (Pearson's R = 0.19, p-value = 0.44).

Trypanosoma-Bartonella coinfection
Considering all host species together, Bartonella and Trypanosoma infections were not correlated. The observed coinfection prevalence was 18.1%, which is not statistically different (χ 2 = 0.7, df = 1, p-value = 0.4) from the expected 16.7% based on the prevalence of Bartonella DNA (66.4%) and Trypanosoma DNA (25.1%). For each of the host species with more than one individual infected with Trypanosoma over the whole study period, we saw no significant association based on binomial tests of proportions (p-value > 0.05) between Bartonella and

PLOS ONE
Trypanosoma infection (S3 Table). We also saw no significant association between infections in species across months of the study period when more than one individual was captured and tested positive for either organism (S4 Table).

Trypanosoma species diversity
Phylogenetic analysis of positive 18S rRNA sequences recovered five Trypanosoma haplotypes that were distinct from one another and from reference Trypanosoma species on GenBank (Fig 3). All identified haplotypes clustered within the Herpetosoma subgenus (T. lewisi clade) along with species T. nabiasi, T. microti, T. evotomys, T. otospermophili, and T. kuseli. Haplotype 1 sequences were found in two species in the family Cricetidae, N. micropus (85) and P. leucopus (3). Haplotype 1 was most closely related to haplotype 2 with 98.5% sequence identity (450/457 bp) although their phylogenetic relationship was not resolved. Haplotype 2 was found in two Peromyscus species, P. leucopus (5) and P. maniculatus (1). Haplotype 3 was strictly associated with S. audubonii and was most closely related to T. nabiasi previously detected in European rabbits (Oryctolagus cuniculus) and associated ectoparasites in Spain and the UK with 99.6% sequence identity (451/453 bp) and 99% bootstrap support (Fig 3). Haplotype 4 was found predominantly in D. ordii (7) with one sequence from N. micropus. This

PLOS ONE
Trypanosoma (Herpetosoma) diversity in rodents and lagomorphs of New Mexico haplotype shared 99.8% sequence identity (452/453 bp) with haplotype 2 and T. microti detected in Microtus miurus in Alaska, USA (GenBank accession AY586623) and 99.6% sequence identity (452/454 bp) with T. microtus, T. evotomys, and other Trypanosoma haplotypes detected in voles and lemmings (Myodes, Microtus, and Lemmus spp.) from Japan, the UK, and Alaska, USA (AB242276, AB242275, AY043356, AY586622, AY586621, AJ009158, AJ009158). However, the phylogenetic relationship between haplotype 4 and its close relatives is uncertain, demonstrated by the polytomy in this region of the tree (Fig 3). Haplotype 5 was predominantly associated with O. variegatus (7) with one sequence from N. micropus. This haplotype shared 99.3% sequence identity (453/456 bp) with T. otospermophili detected in ground squirrels (Spermophilus and Urocitellus spp.) imported into Japan from the USA (AB175625, AB190228) and 98.7% sequence identity (450/456 bp) with T. kuseli detected in the flying squirrel Pteromys volans imported into Japan from China (AB175626). However, haplotype 5 is a distinct lineage from T. otospermophili and T. kuseli in the phylogenetic tree (Fig 3). The sequences of haplotype 4 and 5 from N. micropus were detected from animals captured in May 2003 and the sequences of haplotype 1 from P. leucopus were amplified from animals from April 2003 (S1 Dataset). No sequences of T. cruzi were amplified with the nested 18S rRNA primers and none of the samples were positive for the qPCR assay specific to T. cruzi minicircle DNA. Representative sequences amplified from each host species and used to define new haplotypes were uploaded to GenBank (S5 Table). Information on Trypanosoma species used as references in the phylogenetic analysis can be found in S6 Table. The trimmed sequences from all new haplotypes submitted to GenBank and Trypanosoma references and used for phylogenetic analysis are included in S5 and S6 Tables.

Trypanosoma in recaptured Neotoma woodrats
A total of 31 Neotoma woodrats (26 N. micropus, 5 N. albigula) were recaptured three or more times over the course of the study. Recaptured animals showed a sparse pattern of Trypanosoma infection, with 20 individuals showing no infection over the study period and the remaining 11 showing a repeating pattern of infection with the same Trypanosoma haplotype over multiple months of sampling (Fig 4). Positive N. micropus individuals were predominantly infected with haplotype 1, occasionally interrupted by a negative test or, in one case (adult female N. micropus H-037), detection of haplotype 4 in May 2003.

Discussion
The aim of the present study was to investigate the prevalence and genetic diversity of Trypanosoma parasites in a community of rodents and lagomorphs in New Mexico, document patterns of coinfection in hosts with trypanosomes and Bartonella bacteria, and determine factors that predict infection risk in an abundant species in the community, N. micropus. We found that that the estimated prevalence of Trypanosoma DNA in species ranged from 7.1% in P. leucopus to 50% in S. audubonii, with consistent prevalence around 30% for N. micropus throughout the study period. All Trypanosoma haplotypes sequenced in this study clustered in the Herpetosoma subgenus (T. lewisi clade) along with T. kuseli, T. evotomys, T. otospermophili, T. microti, and T. nabiasi and were largely specific to one host or two phylogenetically related hosts (Fig 3). We also found that the probability of Trypanosoma infection in N. micropus varied by age class, with a higher prevalence in subadults compared to juveniles and adults. We provide evidence that N. micropus individuals can be persistently infected over multiple months with the same Trypanosoma haplotype. Finally, coinfection patterns with Bartonella bacteria did not indicate that Trypanosoma infections are facilitated by infection with Bartonella in N. micropus or across sampled host species.
The phylogenetic position of the Trypanosoma haplotypes enables prediction of their natural transmission cycle [69]. Trypanosomes in the T. lewisi clade are believed to be transmitted by fleas, with supporting experimental evidence for T. grosi, T. lewisi, T. microti, T. nabiasi, and T. evotomys [12,69,70]. Animals contract infection through wounds contaminated with infected flea feces or by ingesting infected fleas and flea feces during grooming [1,71]. Fleas become infected upon ingesting a blood meal from an infected animal [71]. As all obtained sequences cluster within the T. lewisi clade, their potential vectors could be among 107 local flea species [72]. However, experimental manipulation of flea prevalence demonstrated vectorindependent transmission of T. microti in field voles (Microtus agrestis) [73], suggesting that direct transmission between hosts may also be important.

PLOS ONE
Due to the high variation in Trypanosoma DNA prevalence among host species and the observed pattern of host-specificity, comparing prevalence rates is more feasible at the host species or genus level, rather than in the whole community of rodents and lagomorphs. We discuss our results in the context of previous surveys of trypanosomes in each species or related species in the United States.

Trypanosoma spp. in Neotoma woodrats
The observed Trypanosoma prevalence in our sampled population of N. micropus is comparable to estimates in Neotoma woodrats in other areas. The only prior study in New Mexico found one N. micropus and one N. albigula positive for T. cruzi [23]. However, it is unclear whether the methods used in their study (xenodiagnosis in triatomine bugs) could have detected other trypanosomes besides T. cruzi. In California, T. neotomae was described from 19.4% of N. fuscipes [29] and later found in N. lepida there as well [74]. In Texas, T. cruzi prevalence in N. micropus ranged from 21% to 46% [22,[24][25][26]75]. Studies in Louisiana and California revealed T. cruzi prevalence of 73.3% in N. floridana [27] and 14.3% in N. macrotis [28]. Again, these studies used laboratory methods that are specific for T. cruzi detection, so it is unknown whether other Trypanosoma species were truly absent in the tested woodrats. Studies in Texas also identified T. neotomae, T. neotomae-like, and T. lewisi-like parasites in N. micropus [1,22,76]. A recent molecular study in Texas found that 24% of N. micropus were infected with T. neotomae-like parasites [76]. Comparison of partial sequences of the 24Sα gene showed that the T. neotomae-like trypanosomes shared 99.1% sequence identity compared with T. kuseli and T. otospermophili [76]. In Kansas, 13% of N. floridana carried T. kansasensis [30]. An interesting outcome of our study was that all 47 tested N. albigula samples were negative for Trypanosoma species. However, T. cruzi was reported in N. albigula in Arizona and New Mexico and T. neotomae was detected in Mexico [23,77].
Considering that we used specific PCR primers that targeted T. cruzi and more general primers for detection of other Trypanosoma species, we believe that our results indicate that T. cruzi is likely absent or at very low prevalence in our studied rodent community, and that Trypanosoma sp. haplotype 1, belonging to the T. lewisi clade, is prevalent in N. micropus. Since there have been no other studies that have targeted the 18S rRNA gene in Neotoma woodrats, it is challenging to interpret the identity of Trypanosoma haplotype 1 in relation to other Trypanosoma species identified in Neotoma spp. While the prevalence values estimated in our study compare favorably with estimates of T. neotomae or T. neotomae-like parasites in California and Texas [29,76], we cannot determine whether the haplotype found in N. micropus in New Mexico is T. neotomae or another species without morphological examination of the trypanosomes. Such examination was not possible with the archived samples in our study, so additional studies using both morphological examination of trypanosomes in fresh blood smears and molecular sequencing will be necessary to conclusively identify these parasites in N. micropus. Furthermore, broader surveys using such methods in other Neotoma species may recover T. kansasensis or other Trypanosoma species occurring in the US.

Trypanosoma spp. in other rodents
This study also provided evidence of Trypanosoma infection in several other rodent species for the first time, including D. ordii and O. variegatus [23,76]. Previous studies have detected T. otospermophili, T. cruzi, and a T. lewisi-like species in O. beecheyi [78][79][80][81][82], a congener of O. variegatus. Trypanosoma haplotype 5 found in O. variegatus in our study is genetically similar to T. otospermophili and T. kuseli, both species associated with sciurid rodents, although additional sequencing and morphological examination will be necessary to ascertain whether haplotype 5 represents a new Trypanosoma species.
For other species that were negative for Trypanosoma DNA in our study, the limited number of samples from these species prevents us from adequately assessing their host status. More studies with increased sampling from O. leucogaster, P. truei, P. flavus, R. megalotis, D. spectabilis, M. musculus, and X. spilosoma will be necessary to determine if trypanosomes are truly absent from these species in New Mexico. T. cruzi has been reported from M. musculus [22,76] and O. leucogaster [25] in Texas, as well as P. truei and R. megalotis in California [83,84]. The single P. leucopus previously tested from Texas was negative for T. cruzi and other trypanosomes [76]. P. maniculatus was T. cruzi-negative in New Mexico and California [23, 78], but T. peromysci was found in this host in Canada [85] and Arizona [77]. Like the relationship of Trypanosoma haplotype 1 from N. micropus and T. neotomae or T. kansasensis, the question of whether haplotype 2 found in P. leucopus and P. maniculatus is T. peromysci will require additional sampling, morphological characterization of trypanosomes, and molecular sequencing.

Trypanosoma spp. in lagomorphs
Our study found that S. audubonii carried Trypanosoma haplotype 3, which was closely related in T. nabiasi. First discovered in the blood of European rabbits (Oryctolagus cuniculus) in France [86], T. nabiasi has been reported in wild O. cuniculus in Europe [34], domesticated O. cuniculus outside Europe [1], and introduced O. cuniculus in Australia [33]. In Spain, T. nabiasi prevalence in O. cuniculus was reported as high as 82.4% [32,87]. Trypanosomes morphologically similar to T. nabiasi have previously been detected in North American cottontails (Sylvilagus audubonii, S. nuttallii, S. floridanus) [85,88,89], but were unnamed and have not previously been sequenced. Despite the introduction of O. cuniculus to the USA and its current widespread status [90], the sequence distance between haplotype 3 and T. nabiasi (99.6%) suggests that these organisms may be distinct species. Sequences from trypanosomes in introduced O. cuniculus in Australia were identical to T. nabiasi in Europe, likely the result of a recent introduction of the parasites in rabbit fleas or infected rabbits [33]. Such a recent introduction does not appear to explain the origin of Trypanosoma haplotype 3. Instead, a more parsimonious explanation is that T. nabiasi and haplotype 3 diverged along with their hosts when the genera Oryctolagus and Sylvilagus separated between 7 and 16 million years ago [91][92][93]. More studies on trypanosomes in native and introduced lagomorphs in the United States will help to clarify the evolutionary history and host associations of T. nabiasi and related species.
Another argument against a recent introduction of T. nabiasi is that the rabbit flea Spilopsyllus cuniculi, the vector of T. nabiasi in Europe [33], does not occur in the United States [94]. However, a congener, S. inaequalis, was collected from wild lagomorphs in New Mexico [72]. In our study community, S. audubonii harbored the flea species Euhoplopsyllus glacialis, Cediopsylla inaequalis, anrd Orchopeas sexdentatus. E. glacialis fleas parasitize cottontails but can also invade buildings and bite humans [95]. Therefore, given the potential for atypical human infections by trypanosomes in the T. lewisi clade, more studies could determine the vectors of trypanosomes in S. audubonii and evaluate risks for human exposure.

Trypanosoma infection dynamics in Neotoma micropus
Our focused analysis on the large number of samples collected from N. micropus found that Trypanosoma prevalence varied with age class, with a higher prevalence in subadults than juveniles or adults. This pattern of infection may partially explain some months of higher prevalence in the N. micropus population when the proportion of subadults among the captured animals was higher, but this trend was weak over the whole study period. The relationship between age class and prevalence over time was likely weakened by variation in sampling intensity (especially after October 2003) and whether an animal with a persistent infection was captured in a given month.
Previous studies have observed relationships between Trypanosoma prevalence and rodent age or weight, although the shape of the relationship is not certain. T. musculi-like infection was found in immature rather than adult rodents [64] and T. evotomys infection was detected predominantly among older rodents [65]. Other authors found age played only a minor role [66,67]. The most similar results to our study were from T. microti in voles in the UK, showing a convex relationship where prevalence first increases with age/weight then later decreases [19]. A relationship of this shape suggests that susceptibility or exposure to trypanosomes varies over an animal's lifetime and that animals may develop some level of immune protection from reinfection later in life.
Evidence of a protective effect of past trypanosome infection has been documented for multiple Trypanosoma species in rodents. Specifically, the presence of persistent dividing forms of T. musculi in the kidneys of previously infected M. musculus was associated with high antibody titers and a lack of parasitemia following a second challenge with T. musculi [96]. These persisting forms of T. musculi in the kidney could be observed almost a year after recovery from parasitemia [97,98]. Persistent dividing forms of T. grosi in the kidneys of Mongolian jirds (Meriones unguiculatus) following the clearance of trypanosomes from the blood appear to have similar protective role. However, pregnancy may lead to reactivation of T. musculi persisting in the liver of M. musculus, producing short bouts of parasitemia [99].
Our data suggest Trypanosoma infections in N. micropus appear to be persistent in some individuals. Ten of the 26 N. micropus recaptured more than three times in our study were positive for the same Trypanosoma haplotype over multiple months (Fig 4). Unfortunately, due to our exclusive use of molecular detection of trypanosome DNA in blood, we cannot determine whether the persistent infections we observed are patent parasitemia or simply detection of DNA from trypanosomes persisting in the kidneys or another organ in these individuals.
For the one individual (H-037) that was positive for haplotype 4 in May 2003, it is possible that this animal was briefly infected with haplotype 4 circulating in D. ordii in this community, possibly via contact between N. micropus and D. ordii or sharing of ectoparasites. The animal may have been coinfected with haplotype 1 at this time, but only haplotype 4 was detected. Repeated PCR and sequencing from this sample confirmed detection of haplotype 4. A previous study of multiple rodent species showed that animals that were serially challenged with two different Trypanosoma species had lower levels of parasitemia and shorter patent periods than control animals challenged only once [100]. A cross-protective immune response activated by an initial infection might explain the short duration of haplotype 4 infection in individual H-037. However, we could not confirm the coinfection with a sequence of haplotype 1 and cannot rule out the possibility of sample contamination during DNA extraction.
Despite these uncertainties, our findings provide some important information on the dynamics of Trypanosoma infection in natural populations of N. micropus. The decline in prevalence among adults might be partly explained by some immune protection from reinfection after recovery from parasitemia earlier in life. Persistence of trypanosomes in the liver or other organs could be involved in the stimulation of a protective antibody response but may result in occasional relapse of parasitemia during pregnancy or other forms of stress. Obviously, additional longitudinal studies that monitor individuals for a much longer period would be able to track how the force of infection varies with age with more accuracy than the age classes we have used. Additionally, analysis of parasitemia and persistent infection in organs would require collection of fresh blood and tissue. Measures of antibody levels in previously infected individuals, especially during reproduction or periods of nutritional stress, might be able to explain the reactivation of persistent infections or reinfection in adult N. micropus.

Patterns of Trypanosoma-Bartonella coinfection
Our findings of coinfection by Bartonella and Trypanosoma in multiple species are in line with other studies of these pathogens in rodents. Prior studies detected Bartonella coinfections with T. evotomys, T. microti, and T. grosi in voles from Norway [101] and with T. musculi-like species in 26.7% of British rodents [64]. Studies in Poland revealed Trypanosoma-Bartonella coinfection rates from 3.1% to 17.3% in voles [66,67,102]. The lack of association between Trypanosoma and Bartonella infection may indicate that these two pathogens do not compete for host resources or facilitate infection through immune suppression, but additional work using experimental infections would be useful to understand how these parasites interact with one another and the host's immune system [103].

Limitations
One important limitation of our study is that we relied solely on molecular detection of Trypanosoma DNA to assess prevalence in host species over time, thus we could not determine whether a positive individual had living trypanosomes in their blood at the time of sampling. Living trypanosomes could have been observed in fresh blood smears prepared at the time of collection, but this was not done in our study and would not have been possible from the archived samples. As already mentioned, these smears would also have been useful in describing Trypanosoma haplotypes using morphological features, thereby allowing us to reconcile the observed haplotypes with Trypanosoma species previously described in some of our study species but not sequenced.
Another drawback of our methodology was the lack of internal controls to determine whether DNA extraction from blood was successful. Since the main focus of our study was estimation of Trypanosoma prevalence and diversity, the potential presence of false negatives was not important. However, internal controls would have been useful in the analysis of infections in recaptured Neotoma woodrats. For example, four of the ten positive recaptured N. micropus (H-027, H-037, H-060, H-149; Fig 4) showed a pattern of positive-negative-positive over multiple months. The negative tests might have been due to the absence of trypanosomes from blood (or at a level below the detection limit of PCR) or due to a failure in DNA extraction. An internal control PCR targeting host DNA present in blood might have been able to distinguish between the different causes of a negative test for Trypanosoma DNA. With the inclusion of internal controls and more consistent recapture of animals over a longer period of time, future longitudinal studies of Trypanosoma infection in natural host populations would be able to measure the duration of infection in individual animals.
Our study also used only one marker for DNA sequencing. While sequences of the 18S rRNA gene have frequently been used for the detection and identification of Trypanosoma species, the information contained in these short sequences is insufficient to reconstruct deeper evolutionary relationships among Trypanosoma species. Sequencing of additional markers (e.g., 70 kDA heat-shock protein, gGADPH) from the haplotypes identified in our study and from other species in the T. lewisi clade would help to clarify the evolutionary history of this group [7,104]. Deep sequencing methods could also be beneficial for detecting coinfections of different Trypanosoma haplotypes [4,105].
Finally, we recognize that differentiation between some rodent species based on morphological observations in the field is known to be challenging in this study area, particularly within the genera Neotoma and Peromyscus. We removed some Peromyscus mice from the study because they could not be identified to the species level, but it is possible that some other Neotoma woodrats or Peromyscus mice were misidentified. While these errors would have had minimal effect on our conclusions about Trypanosoma associations at the host family or genus level, uncovering more subtle patterns of host specificity among trypanosomes would require more robust methods of host identification. We recommend that future studies in this area use molecular methods (e.g., barcoding) to confirm identification of host species. These data, along with sequencing of additional markers from trypanosomes, may reveal additional host specificity between Trypanosoma haplotypes and host species, subspecies, or regional populations.

Conclusions
Rodents and lagomorphs of New Mexico host a variety of host-specific Trypanosoma haplotypes. Trypanosoma infection in N. micropus had consistent prevalence around 30% in the population over time, with varying prevalence across age groups, and some N. micropus appeared to have persistent infections with the same haplotype over multiple months. We report the first molecular detection of a Trypanosoma haplotype related to T. nabiasi in North America in the blood of S. audubonii and the first detection of Trypanosoma infection in D. ordii and O. variegatus. While the pathogenicity of the trypanosomes reported in this study is unknown in humans, the Trypanosoma species present in this rodent and lagomorph species should be monitored as potential agents of human trypanosomiasis.
Supporting information S1 Table. Generalized linear models ranked by AICc following 1000 iterations of stratified sampling and model selection. "FleaPos" represents the presence/absence fleas at the time of capture while "Fleas" is the number of fleas at capture. "Bartonella" is the presence of Bartonella DNA and "Sex � Weight" is an interaction term for N. micropus sex and weight. (XLSX) S2 Table. Model coefficients and 95% confidence intervals for top GLM from S1