Upper respiratory tract disease (URTD), caused by Mycoplasma agassizii, has been deemed a threat to populations of Mojave Desert Tortoises, Gopherus agassizii. Previous work on URTD has focused on serology and visual health examinations to determine the extent of this disease in some natural tortoise populations. Here, we present the first range-wide study of the presence of the pathogen, M. agassizii, in Mojave Desert Tortoises. We detected M. agassizii in tortoise populations throughout the Mojave Desert, with notable differences in prevalence of M. agassizii among sampling sites within tortoise genotypes and sampling years. Analyses of three genetic markers in the M. agassizii genome indicated very low nucleotide diversity and no relevant spatial structuring of Mycoplasma haplotypes. We use published lines of evidence to discuss the roles of rare transmission events and long-term mycoplasmal persistence in individual hosts on tortoise URTD dynamics.
Diseases can have important effects on wildlife populations. When a pathogen is invasive and infects naïve host species and populations, such as the spreading of the fungus that causes chytridiomycosis (Batrachochytrium dendrobatidis), it can cause severe population declines in its amphibian hosts (Skerratt et al. 2007). Some host species can act as carriers, not directly affected by disease, but able to spread the pathogen to additional hosts (Eskew et al. 2015). Even within a single host species, a disease could affect some individuals, whereas others are carriers, such as in the typhoid disease system in humans (reviewed in Merrell and Falkow 2004).
An upper respiratory tract disease (URTD) caused by a species of Mycoplasma was first detected in Mojave Desert Tortoises, Gopherus agassizii (Cooper 1861), in California after a population crash in the late 1980s (Jacobson et al. 1991; reviewed in Sandmeier et al. 2009). This population crash was posited to be caused by URTD, and it spurred an abundance of research regarding the pathology (including mortality and morbidity) induced by the putative causative pathogen (Jacobson et al. 1991; Brown et al. 1999). The population crash, as well as other population declines, led to the listing of Mojave Desert Tortoises as threatened under the Endangered Species Act in 1990 (U.S. Fish and Wildlife Service 1994). Brown et al. (1994) verified that the pathogen causing URTD was the newly named Mycoplasma agassizii (Brown et al. 1994, 2001). In tortoises, M. agassizii forms a close association with the nasal epithelium (Jacobson et al. 1991), and URTD can cause clinical signs of nasal exudate, palpebral edema, and wheezing, with severe cases resulting in lethargy and potentially mortality (Brown et al. 1994). Subsequent research continued to investigate the extent to which the disease threatens tortoise populations (Schumacher et al. 1997; Dickinson et al. 2005; Sandmeier et al. 2013) and resulted in the development of serological tools (ELISA and Western blot) that can be used to detect Mycoplasma antibodies in tortoises (Schumacher et al. 1993; Brown et al. 2002; Hunter et al. 2008). Recently, a probe-based quantitative polymerase chain reaction (qPCR) for M. agassizii was developed and is used to detect the presence of the pathogen itself, to supplement the use of serological tools (Braun et al. 2014).
The ecological importance of URTD for tortoise populations continues to be controversial, because published reports have noted that this disease correlates with a high risk of morbidity, mortality, and population crashes (reviewed in Sandmeier et al. 2009). It is currently thought that M. agassizii is present in tortoises throughout the Mojave Desert and prevalence, measured by serology, differs among geographic areas (Sandmeier et al. 2013). Additional data indicate that multiple strains of M. agassizii are present in Mojave Desert Tortoises (Brown et al. 2001), but the extent to which different strains are found in tortoise populations remains unknown. Until now, no study has comprehensively investigated the presence of M. agassizii in Mojave Desert Tortoises across their geographic range, from southern California to southwest Utah in the Mojave and Colorado Deserts. Furthermore, genetic variation within and among Mycoplasma populations has not been assessed.
Here, we use samples of microbes obtained from nasal passages of Desert Tortoises to address the following questions:
What is the relative efficacy of conventional PCR and DNA sequencing compared with qPCR in detecting M. agassizii?
What is the prevalence of M. agassizii across the geographic distribution of Mojave Desert Tortoises?
Can different strains of M. agassizii be detected in our samples, and are certain strains more associated with clinical signs of disease in the tortoises than other strains?
Is prevalence of M. agassizii or disease correlated with tortoise genotype?
We predict that M. agassizii is present in tortoise populations across the Mojave Desert, in accordance with previously measured seroprevalence (Sandmeier et al. 2013). We also predict that multiple strains of M. agassizii will be found, and that the presence of some strains in tortoises will be more closely associated with clinical signs of URTD than others.
The present study adds PCR and DNA sequencing to the diagnostic methods recently compared for their efficacy (Sandmeier et al. 2017). Additionally, we use the diagnostic methods presented here to answer broader questions regarding ecological patterns among the component parts of this disease system.
Materials and Methods
Field Sites and Sample Collection
Samples of upper respiratory bacteria were taken from wild-caught Mojave Desert Tortoises in the southwestern United States during the active season (April–October) between 2010 and 2012. Each tortoise was sampled only once. Our goal was to sample from sites relatively evenly spaced across the tortoise distribution (Fig. 1), and sites were generally delineated by mountain ranges, which confine tortoises to valley floors. Sites were selected to represent three previously determined tortoise genotypes: California (CA), Las Vegas (LV), and Northeast Mojave (NEM; Table 1; Hagerty and Tracy 2010). At most sites, tortoises are sparsely distributed (with home ranges ranging from 1 to 89 ha; Berish and Medica 2014); to increase the probability that our sampling was representative for each population, we found most tortoises along haphazardly selected 10-km transects. Approximately 1 tortoise encountered per 10 km transect can be expected based upon U.S. Fish and Wildlife surveys (available at https://psw.databasin.org/galleries/af8e55a0197a4c95a3120b278075a2b1), and our goal was to sample 20 tortoises per site. Although tortoises can move long distances, they are not likely to move among local populations, particularly not far enough to occupy a new range that falls within a different genotype group. Thus, we used the assumed genotypes relevant for each local population as determined by Hagerty and Tracy (2010), and we did not verify genotype group per tortoise. Genotype was used to detect general patterns across the tortoise distribution. We additionally sought to mitigate biases in sampling among years by sampling sites during multiple years, when possible (Table 1).
Visible presence of clinical signs of URTD, including nasal mucus or signs that the tortoise had recently exuded nasal discharge (occluded or asymmetrical nares, or eroded scales around nares), was evaluated for each tortoise. Tortoises with any of these signs of URTD were considered positive for clinical signs. Palpebral edema or any other ocular syndrome was not included in the list of relevant clinical signs of disease as they have not been included in the disease results from Koch's postulates regarding M. agassizii and URTD (Brown et al. 1994). To determine whether Mycoplasma DNA was present in the upper respiratory tract, a nasal lavage sample was collected by flushing 3 mL of sterile saline solution (0.9% NaCl) through the tortoise nares (Brown et al. 2002). Nasal lavage samples were immediately preserved in RNAlater Stabilization Solution (Ambion Inc.) at a ratio of 1 : 5 preservative to sample volume. Preserved samples were placed on ice in the field and frozen within 12 h of collection. In the laboratory, DNA from 500 μL of preserved nasal lavage samples was extracted with the use of the Qiagen DNeasy Blood and Tissue Kit (Qiagen Inc.) protocol for gram-negative bacteria.
Presence of M. agassizii was detected by qPCR under the protocol described in Braun et al. (2014). This protocol can detect 50 copies of the 16S ribosomal RNA (rRNA) gene in M. agassizii with a Cq value of 38.5, maximum intra-assay variation of 0.97, and a detection limit of a Cq of 40 (Braun et al. 2014). In qPCR, Cq value is a calculation of the number of cycles necessary for amplification of the desired region above a given threshold. Thus, a lower Cq corresponds to more of the desired template DNA in the sample. All qPCRs were conducted in triplicate with TaqMan Environmental Master Mix 2.0 with the use of a Step One Plus Real-Time PCR System with 40 thermal cycles and Step One Software v2.3 (Applied Biosystems, Foster City, CA). In accordance with the recommendation of Braun et al. (2014), samples were considered to include the DNA of M. agassizii (qPCR-positive), if two or three of the triplicates had a Cq value below 40.
Positive results were categorized as being either strongly positive (with a mean Cq < 38) or weakly positive (with 38 < mean Cq < 40). Although this qPCR assay can be used to detect the presence of M. agassizii in a sample up to a Cq value of 40, we divided qPCR results into three data sets in order to detect differences in patterns associated with strong versus weak infections: Data Set 1 (DS1) = both strongly positive and weakly positive samples were categorized as positive for M. agassizii, all other samples were categorized as negative; Data Set 2 (DS2) = only strongly positive samples were categorized as positive for M. agassizii, all other samples were categorized as negative; and Data Set 3 (DS3) = results from weakly positive samples were excluded from analyses, with only strongly positive samples categorized as positive and only negative samples categorized as negative.
Samples were screened for three M. agassizii genetic markers with the use of PCR: 16S rRNA, 16-23S intergenic spacer region (IGS), and ribosomal polymerase β subunit (rpoB; see Table S1 in the Supplemental Materials available online). PCRs were conducted with Qiagen Multiplex PCR mix (Qiagen Inc.) with a 15-min hot start at 95°C and a final extension at 72°C. Nested and pseudo-nested PCR products of appropriate length were extracted from agarose gel with the use of a Qiagen QIAquick Gel Extraction Kit. Gel-extracted products were sequenced at the Nevada Genomics Center on an ABI3730 DNA Analyzer (Applied Biosystems). Sequences were verified as M. agassizii with the use of NCBI's BLAST search (available at http://blast.ncbi.nlm.nih.gov). Samples were considered positive for M. agassizii if at least one PCR amplified the appropriate marker.
qPCR and PCR comparison
Correspondence of M. agassizii results from qPCR and PCR assays per sample was assessed by calculating unweighted kappa statistics with the use of the kappa2 function in the irr package (Gamer et al. 2010) in the programming language R (v3.2.1; R Core Team 2015). Cohen's kappa (κ) ranges from −1.00 to 1.00, where values <0.00 = poor agreement, 0.00–0.20 = slight agreement, 0.21–0.40 = fair agreement, 0.41–0.60 = moderate agreement, 0.61–0.80 = substantial agreement, and 0.81–1.00 = almost perfect agreement (Landis and Koch 1977). PCR results were compared with the three qPCR data sets described above.
Mycoplasmal Genetic Diversity and Geographic Structure
DNA sequences were aligned in ClustalX v2.1 (Larkin et al. 2007) and trimmed in MEGA v6.06 (Tamura et al. 2013). Following Clement et al. (2000), TCS haplotype network analyses were conducted in PopART (Leigh and Bryant 2015). Haplotype networks included sequences from GenBank (available at http://www.ncbi.nlm.nih.gov) for M. agassizii (Accession Nos. U09786.1, NR_114450.1, AY780802.1, AY780801.1, and EU925153.1) to determine similarity between mycoplasmas in our samples and known strains, such as PS6 (Brown et al. 2001). Spatial analyses of molecular variance (SAMOVAs) were conducted separately for each marker in SAMOVA v1.0 (Dupanloup et al. 2002) to detect spatial patterns of haplotypes. Sampling localities were input as the mean latitude and longitude coordinates of all tortoises sampled per site. SAMOVA was used to produce F statistics by grouping sampling sites based upon their spatial coordinates and associated haplotypes. We ran SAMOVA with groups K = 2 to 16 (when appropriate) and 100 simulated annealing processes (Dupanloup et al. 2002). The output having the highest Fct value (i.e., the number of groups that maximized among-group variance relative to the total variance) defined the best number of groups. Sequences shorter than 75% of the average length for each genetic marker were excluded from haplotype analyses (excludes three samples for IGS and four samples for rpoB).
Transmission of M. agassizii in adults is likely associated with social behaviors, and little is known about transmission or disease progression in juveniles. Thus, we chose to exclude possible confounding effects from juveniles by excluding four samples from tortoises smaller than 100-mm midline carapace length (MCL) from the range-wide analyses. One sample from the large-scale translocation site in Clark County, Nevada, was also excluded from further analyses because of ambiguities surrounding the historical origin of the tortoise. Analyses regarding genotypes included all remaining samples, whereas those regarding site-level patterns in prevalence exclude sites with fewer than 10 samples assayed.
Presence of Mycoplasma in nasal lavage samples (via qPCR) and clinical signs of URTD were analyzed separately with the use of generalized linear mixed effects models (GLMM) with the use of the glmer function in the lme4 and lmertest packages in R (Bates et al. 2015; Kuznetsova et al. 2016). Predictor variables included tortoise genotype, sampling year, sex, and MCL, with a random variable of site nested within genotype. Analyses of clinical signs of URTD also included Mycoplasma presence or Cq as a predictor variable. We used Tukey's post hoc tests to conduct pairwise comparisons to compare factors within the genotype, year, and sex variables by implementing the glht program in the R package multcomp (Hothorn et al. 2008). We analyzed Cq values with the use of a linear mixed-effects model with the lmer program in the lme4 package using the same predictor and random variables as for the analysis of Mycoplasma presence and Tukey's post hoc tests as above. To detect whether some sites differ in Mycoplasma presence from the rest of the genotype (Table 1), we analyzed Mycoplasma as predicted by site for each genotype with logistic regressions with the use of the glm function. We compared average Cq values with Mycoplasma prevalence by site using linear regressions with the lm function in R to determine if higher average Mycoplasma loads tended to correspond with higher prevalence of Mycoplasma among sampling sites.
Sample Collection and Mycoplasma Detection
Nasal lavage samples were collected from 402 wild Mojave Desert Tortoises from 2010 to 2012. At least 10 tortoises were sampled from each of 21 sampling sites, and fewer samples were collected from six additional sites (Table 1, Fig. 1, and see Table S2 in Supplemental Materials available online). Out of the 402 samples, PCR detected M. agassizii in 64 samples, and 198 samples were qPCR-positive with Cq < 40.
Detection of M. agassizii via PCR produced similar results with slight agreement to two qPCR data sets (DS2, n = 402, κ = 0.159, P < 0.001; DS3, n = 315, κ = 0.138, P < 0.01; Table S3) in which the weakly positive samples were considered to be negative or excluded. The κ statistic between PCR and Data Set 1 was not significant (DS1, n = 402, κ = 0.045, P = 0.23).
Mycoplasmal Genetic Diversity and Geographic Structure
The PCR reactions for 16S rRNA, IGS, and rpoB detected M. agassizii in 49, 51, and 26 samples, respectively. Nucleotide diversity of aligned, trimmed sequences was two to three times higher in IGS (47 samples, 352 bp, π = 0.0016) than rpoB (23 samples, 548 bp, π = 0.0005) and 16S (47 samples, 453 bp, π = 0.0007). TCS network analyses for each genetic marker detected one principal haplotype, with additional haplotypes differentiated by 1–3 single-nucleotide polymorphisms, represented by a single sample (Fig. 2). A second haplotype was detected in five samples in the IGS marker (Fig. 2b). Despite this shared secondary haplotype, SAMOVAs did not detect any meaningful spatial distribution of haplotypes for any of the three genetic markers, with or without the inclusion of indels in the analyses (Table S5; Fig. S1). Because each genetic marker resulted in one common haplotype, we could not determine whether uncommon haplotypes were more or less associated with clinical signs of disease.
Tortoises with MCL ≥ 100 mm were sampled from California (n = 156), Las Vegas (n = 106), and Northeast Mojave (n = 135) genotypes (Table 1). The following results were calculated with the use of Data Set 1 to include all positive samples. In the California genotype, 57% of tortoises tested positive by qPCR for M. agassizii (Cq < 40), and Chuckwalla Bench had lower prevalence than other California sites (25%, t = −2.15, P = 0.03; Fig. 3). Fifty-two percent of tortoises from the Las Vegas genotype had a detectable amount of M. agassizii. Of the sites within this genotype, Pahrump (25%, t = −3.141, P = 0.002) and South Las Vegas (45%, t = −2.103, P = 0.04) samples had significantly lower levels of prevalence compared to other sites in the Las Vegas genotype, and prevalence in the Northwest Vegas site was marginally different (50%, t = −1.803, P = 0.07). In the Northeast Mojave (36% prevalence), Gold Butte, Mormon Mesa/Halfway, Red Cliffs, and Zion all stood out with significantly lower prevalence (P < 0.05).
The GLMM did not detect effects of genotype, sex (female, n = 168; male, n = 179; unknown sex, n = 47), or MCL on presence of M. agassizii (in all cases, P > 0.1). There was an effect of year on Mycoplasma presence, with 2011 having higher prevalence than 2010 (z = 2.354, P = 0.05), and marginally higher prevalence than 2012 (z = −2.173, P = 0.08).
Clinical signs of URTD (n = 91) were marginally correlated with presence of M. agassizii in nasal lavages (z = 1.879, P = 0.06), and were only predicted by sampling year, with 2010 having a lower occurrence than both other sampling years (2011, z = 2.502, P = 0.03; 2012, z = 2.860, P = 0.01). When Mycoplasma presence was replaced with Cq value in the GLMM, Cq predicted presence of clinical signs (z = −1.948, P = 0.05), with similar effects of sampling year. Thus, animals with more Mycoplasma per sample generally had an increased likelihood of exhibiting clinical signs of URTD.
By site, there was a trend for the prevalence of M. agassizii to be negatively correlated with higher Cq values (inferring decreased average Mycoplasma load; F1,18 = 3.77, R2 = 0.17, P = 0.07; Fig. 4). Importantly, the Cq variable does not represent a linear change in starting quantity of target DNA, but rather the number of PCR cycles necessary to reach a threshold. Thus, a decrease in Cq by 3.5 is associated with an approximately 10-fold increase in target DNA. Cq values from our sample DNA ranged from 28.98 in a sample from Ord Rodman to 39.93 in a sample from Gold Butte, representing a 1000-fold increase in mycoplasmal DNA from the strongest to weakest infection load detected.
From the kappa statistics, we found that PCR and qPCR provided similar diagnostic value on the scale of the individual host. The best agreement between these techniques occurred when weakly positive samples were considered to be negative, indicating that qPCR might be more sensitive than PCR. Further supporting this notion was the observation that most of the disagreement between the two techniques resulted from positive qPCR results corresponding with negative PCR results (20–40%, depending on the data set), whereas we detected up to 14.7% negative qPCR results corresponding with positive PCR results (Table S3). In addition to lower detection sensitivity, PCR and DNA sequencing is expensive and time consuming compared to qPCR. Thus, as a diagnostic technique to detect this pathogen in nasal lavage samples from tortoises, we recommend qPCR over PCR, interpreting Cq values under 40 as positive for M. agassizii, as suggested in Braun et al. (2014). Although our previous work detected higher intra-assay variability at Cq values of 38 and above (Sandmeier et al. 2017), the high chemical specificity of a probe-based qPCR indicates that Type II errors increase more than Type I errors at low DNA concentrations. Thus, even Data Set 1 could underestimate true prevalence of the microbe (Sandmeier et al. 2017), especially considering the imprecision of the nasal lavage sampling process.
We sequenced three genetic markers of M. agassizii from Mojave Desert Tortoises and found very little genetic variation. The type strain of this pathogen (PS6) was cultured from a Mojave Desert Tortoise and had similar rpoB and IGS sequences to the mycoplasmas collected in our study, but its 16S rRNA sequence differed slightly from the main haplotype that we detected (Fig. 2). This type strain (PS6) is considered to be especially pathogenic (Brown et al. 1994, 2001), and the fact that it differs from the rest in 16S rRNA could indicate that the pathogen has evolved since its first discovery, or that we did not sample tortoises in areas where the type strain originally occurred (e.g., Desert Tortoise Natural Area, Kern County, CA). Importantly, this pathogenic strain was not detected in our survey.
The three genetic markers that we sequenced are in conserved regions of the genome or linked to conserved regions. In a previous report, multiple strains that differed in physical size and cell viability after storage were found in the host species (Brown et al. 2001). Whereas we did not find evidence of distinct, shared strains, the true determinant of genetic differentiation requires genome-level analyses of isolates from field samples. A better understanding of the M. agassizii genome could allow for identification of virulence factors, such as adhesins, which might be an important indicator of virulence at a population level of the host. If there are virulence factors associated with disease, it would be important to determine which local host populations seem most at risk based on the Mycoplasma strains present. With this kind of information, managers could better address the implications of tortoise translocation programs and the potential effects of habitat-altering stressors on disease outbreaks at a local population scale.
Mycoplasmal URTD in Mojave Desert Tortoises is categorized by slow transmission and slow disease progression (Maloney 2011; Aiello et al. 2016). Previous work by Jacobson et al. (1995) found that tortoises appearing to be healthy could in fact be infected, and it is possible that an infection is not detectable by field techniques that preserve the tortoise's life. Although histology is the best determination of URTD (Jacobson et al. 1995), it is not a practical method to determine population-level disease and infection patterns in a threatened species. Whereas we detected samples that were qPCR negative for M. agassizii, many more of our cultured field samples contained M. agassizii, even though its presence was not detected by pre-culture qPCR (C.L. Weitzman, personal observation). Thus, it is important to note that many more individuals likely harbor this pathogen than can be detected.
Although there is a downward trend in population sizes over most of the Mojave Desert Tortoise distribution (U.S. Fish and Wildlife Service 2015), since the crashes of 1989–1990 (Corn 1994; Peterson 1994), we know of only one crash in the mid-1990s, seemingly caused by drought (Longshore et al. 2003; reviewed in Sandmeier et al. 2009). Mycoplasma agassizii is widespread across its host's distribution, but to the best of our knowledge, no die-offs have been linked to disease in the last 25 yr, despite locations with high pathogen prevalence detected in our study (e.g., 84% in Fenner Valley). High prevalence paired with low morbidity and mortality rates supports the hypothesis that this pathogen maintains colonies in the tortoise host, becoming infective at higher loads that could be triggered by environmental stressors or interactions within the upper respiratory microbiome.
We found that the presence of M. agassizii was associated with clinical signs of disease in individual tortoises, supporting previous findings of a similar correlation at the population level (Sandmeier et al. 2017). This correlation was tighter when only strongly positive samples were included in the pool of those positive for M. agassizii (Data Sets 2 and 3), indicating that a higher pathogen load might be more likely to cause URTD. Low pathogen loads might be attributable to recent colonization; alternatively, pathogen load could oscillate between high and low levels, with similar cycles in morbidity.
Currently, it is unknown how M. agassizii persists in hosts in a subclinical state. While M. agassizii is an extracellular pathogen that binds to the epithelium of its host (Jacobson et al. 1991), some predominantly extracellular mycoplasmas can also survive and reproduce within host cells, including M. pneumoniae, M. penetrans, M. genitalium, M. gallisepticum, and M. fermentans (Dallo and Baseman 2000; Rosengarten et al. 2000; Waites and Talkington 2004). It seems that this pattern of life-history traits might be common among Mycoplasma species. Hosts of M. pneumoniae might spread the pathogen without experiencing morbidity, and the presence of a carrier state might explain the cyclical epidemic trends in M. pneumoniae that occur every 3–5 yr (reviewed in Waites and Talkington 2004). Possibly, M. agassizii also has an intracellular stage or hides elsewhere in the respiratory tract of the host, where sampling cannot reach. Alternatively, the microbe could exist in such small population numbers that they are not easily represented in lavage samples. The presence of infection cycles in this host–pathogen system could explain the differences that have been observed in M. agassizii prevalence among years and sites. Sandmeier et al. (2013) found higher serological prevalence in the Northeast Mojave tortoise genotype than in the California genotype, but our sampling for Mycoplasma by qPCR years later did not reveal a similar pattern. The large time delay between serological and pathogen assays could have allowed for cycling of disease dynamics, explaining why patterns from the serological survey did not correspond with similar patterns in the present pathogen survey. Additionally, if serology and pathogen presence are not correlated on a local population scale (sampling locality) at a single time point (see Sandmeier et al. 2017), then cyclical patterns in URTD-associated variables might also explain these discrepancies. Alternatively, tortoise immune responses to similar loads of M. agassizii could vary among populations, possibly caused by differences in thermal environment or nutrition (Sandmeier et al. 2013; Drake et al. 2016).
Most disease models focus on parameters of pathogen transmission. We propose that this tortoise/Mycoplasma system might be unique in the interplay between persistence in a long-lived host and rare transmission events. There are currently no available data on the recovery of tortoises from URTD, as the progression of this disease is extremely slow. If an individual host cannot clear itself of the pathogen, then even a low transmission rate could allow M. agassizii to accumulate across entire populations of Desert Tortoises, as this host species exhibits a long life-span and low turnover of adults (2%; U.S. Fish and Wildlife Service 1994). If that is the case, then predictive models need to incorporate the long-term persistence in individuals as well as in populations.
Lastly, it is important to note that pathogens other than M. agassizii are also associated with respiratory disease in Gopherus tortoises. Mycoplasma testudineum is another causative agent of URTD in Desert Tortoises (Jacobson and Berry 2012). Pasteurella testudinis and a herpes virus are also associated with tortoise respiratory disease (Snipes et al. 1980; Snipes and Biberstein 1982; Jacobson et al. 2012). Mycoplasmas can be present in co-infections in tortoises and other vertebrate species (Salinas et al. 2011; Ley et al. 2012), yet we know neither the extent to which co-infections of URTD-associated pathogens occur in Gopherus tortoises, nor the importance of co-infections in the development from subclinical to clinical disease. Although M. agassizii can cause URTD (Brown et al. 1994), it is possible that in wild tortoise populations, an interaction among pathogens is required before clinical signs of disease are detected.
We thank S. Snyder, M. Gordon, J. Jahner, and two anonymous reviewers for their helpful comments on earlier drafts of this manuscript. Our work was partially supported by the National Science Foundation Graduate Research Fellowship under Grant No. DGE-1447692. Funding was also provided by the U.S. Fish and Wildlife Service (1320-114-23DZ) and the Desert Legacy Fund from the Community Foundation. Work was conducted under the following permits: U.S. Fish and Wildlife Service 10(a)1(A) Recovery Permit Nos. TE076710-7 and TE076710-8; California Department of Fish and Game Scientific Collecting Permit SC-007374; Nevada Department of Wildlife Scientific Collecting Permits S-33080 and S-35517; Utah Division of Wildlife Resources Permit 5COLL8886; National Parks Service Permits MOJA-2011-SCI-0022, MOJA-2012-SCI-0009, ZION-2012-SCI-0006, and JOTR-2011-SCI-0021; and University of Nevada, Reno Institutional Animal Care and Use Committee Protocols 00465 and 00555. We thank M. Tuma for field samples, M. Teglas and N. Nieto for laboratory assistance and advice, and multiple field and laboratory technicians.
Supplemental material associated with this article can be found online at http://dx.doi.org/10.1655/Herpetologica-D-16-00079.S1
Associate Editor: Pilar Santidrián Tomillo