Impact of Mycoplasma ovipneumoniae on juvenile bighorn sheep (Ovis canadensis) survival in the northern Basin and Range ecosystem

Study area

The populations of bighorn sheep we studied were located in southeastern Oregon and northern Nevada, between 41.3 and 42.8°N, and 117.0 and 118.2°W (Fig. 1). The entire study area fell within the North Basin and Range (Level III classification of ecoregions in Omernik & Griffith, 2014). Five populations in our study fell within or largely within the Dissected High Lava Plateau (Blue Mountain (BSP), Bowden Hills (BHP), Rattlesnake (RSP), Three Forks (TFK), Upper Owyhee (UOP), Fig. 1), although BSP and RSP partly occurred within the High Lava Plains ecoregion (Omernik & Griffith, 2014). The Ten Mile (TMP) population fell within the High Lava Plains, while High Lava Plains and Semiarid Uplands (Omernik & Griffith, 2014) dominate bighorn sheep habitat within the Trout Creek (Trout Creek—east (TCE),—south (TCS) and—west (TCW), Fig. 1) and Santa Rosa metapopulation (Calicos (CAL), Eight Mile (EML), Martin Creek (MCK), and Sawtooth (SAW), Fig. 1).

The Dissected High Lava Plateau and High Lava Plains ecoregions are both characterized by elevated plateaus, but the Dissected High Lava Plateau contains sheer-walled canyons as well as intermittent lakes, while the High Lava Plains contains isolated volcanic cones and buttes as well as intermittent lakes and ephemeral streams (Omernik & Griffith, 2014). Mountains of low to mid elevation, typically with steep slopes, and some ephemeral and perennial streams characterize the Semiarid Uplands (Omernik & Griffith, 2014). The most common geology types across all three ecotypes are basalt and rhyolite, interspersed with other rock types. The soils derived from these rock types are fairly shallow and poor (Omernik & Griffith, 2014). Mean precipitation across the study area is typically 22.5–35.0 cm per year, although some areas of the Trout Creeks and Santa Rosa Mountains, receive significantly more precipitation (Omernik & Griffith, 2014).

All three ecotypes contain sagebrush steppe. Big sagebrush (Artemisia tridentata) and low sagebrush (A. arbuscular) are the most common woody herbaceous species, while common indigenous herbaceous species being primarily made up of palatable perennial bunchgrasses such as Idaho fescue (Festuca idahoensis), bluebunch wheatgrass (Psudoroegneria spicata), bottlebrush squirreltail (Elymus elymoides), Thurber needlegrass (Achnatherum thurberianum) and the less palatable Sandberg bluegrass (Poa secunda). Western juniper (Juniperus occidentalis) is the most common woody plant across 3 ecotypes, typically found in rocky areas, while the Semiarid Uplands are distinguished by willows (Salix spp.) in riparian areas, and quaking aspen (Populus tremuloides) and mountain mahogany (Cercocarpus spp.) in snow pockets (Omernik & Griffith, 2014). Ungulate species occurring in the study area include elk (Cervus canadensis), mule deer (Odocoileus hemionus), and pronghorn (Antilocapra americana) (Omernik & Griffith, 2014). Potential bighorn sheep predators in the study area include cougar (Puma concolor), coyotes (Canis latrans), bobcat (Lynx rufus), and golden eagles (Aquila chrysaetos; Omernik & Griffith, 2014).

The most common land-use practices are cattle ranching and cultivation of grains and hay (Omernik & Griffith, 2014). Heavy grazing of these lands and suppression of natural fires has led to the spread of large, uncontrollable fires and encroachment by invasive annual plants, such as the cheatgrass (Bromus tectorum) and medusahead (Taeniatherium caput-medusae; Omernik & Griffith, 2014). These grasses outcompete indigenous vegetation post-fire leading to the domination of these grasses. Wildlife and cattle here rely on springs, wetlands, and artificial water sources (Omernik & Griffith, 2014).

History of bighorn sheep populations within the study area

Reestablishment of bighorn sheep in the study area started in 1978 with the translocation of bighorn sheep into the Eight Mile area of the Santa Rosas (Fig. 1). The three Trout Creek populations, TCE, TCS, and TCW, and RSP were established with single translocations each from Hart Mountain (Table S1) in 1987 and 1992. Hart Mountain’s bighorn sheep population was established in 1954 with a translocation of bighorn sheep from Williams Lake, British Columbia. Although bighorn sheep in TMP and UOP were translocated from various other populations in Oregon, all those populations were ultimately derived from the population established at Hart Mountain in 1954. Bighorn sheep in CAL, although translocated from the Pine Forest Range in Nevada, are also derived from Hart Mountain. The remaining bighorn sheep populations in the study were established from either a different original source, e.g., MCK and SAW, which had stock from Kamloops, British Columbia, and Penticton, British Columbia, respectively, or more than one source, e.g., EML. Two populations were established by dispersal into unoccupied habitat: BSP was established with suspected dispersal events from the Trout Creek populations in the mid to late 1990s (pers. comm. S. Torland, ODFW), and BHP is thought to have been established by dispersing bighorn sheep from RSP. Other movements observed include the dispersal of collared adult males between the Santa Rosa Mountain Range and neighboring mountain ranges in southeastern Oregon, observed in 2009 and 2010 (NDOW, unpublished data).

Capture and sampling

All capture, handling, and disease testing were conducted by Oregon Department of Fisheries and Wildlife (ODFW) and Nevada Department of Wildlife (NDOW). Capture methodology followed the recommendations of Foster (2004) and the American Society of Mammalogists (Sikes & the Animal Care and Use Committee of the American Society of Mammalogists, 2016). ODFW and NDOW captured, collared, and sampled adult female bighorn sheep across 13 populations in southeastern Oregon and northern Nevada between January 2016 and February 2018 (Fig. 1). Captures were conducted using a net gun fired from a helicopter, with individual bighorn sheep blindfolded and hobbled once captured (Krausman, Hervert & Ordway, 1985). Bighorn were brought to a centralized area at the base of their range to be fitted with a telemetry collar and to collect biological samples, except where capture location was too far from basecamp to transport them quickly, in which case they were field processed, at the capture location.

Each adult female was fitted with a Vertex Survey Globalstar collar (Vectronic Aerospace, Berlin, Germany). These collars provide a GPS location every 13 h as well as VHF signal and were set to report a mortality if stationary for 12 h. Each collar had its own unique VHF frequency, with the occasional duplicate placed on individuals in different populations between which dispersal was deemed unlikely. The collars were also fitted with colored tags with unique numbers, allowing for identification of individuals observed in the field.

The age of each adult female was estimated from horn growth rings (Geist, 1966; Hoefs & Konig, 1984). Blood was obtained via jugular venipuncture to determine pregnancy status of adult females, obtain DNA, and to screen for disease. We determined pregnancy status of adult females using a serum pregnancy-specific protein B (PSPB) assay (Drew et al., 2001). Pregnancy testing of adult females was only conducted in the year in which they were captured; samples were sent to Sage Laboratories (Emmett, ID) to conduct testing.

Diagnostics

Presence of M. ovipneumoniae was detected with polymerase chain reaction (PCR) tests using nasal, bronchial, and tympanic bullae swabs from each captured female bighorn sheep (Manlove et al., 2019). Previous exposure to M. ovipneumoniae was determined using a competitive enzyme-linked immunosorbent assay (cELISA) to detect antibodies in serum (Ziegler et al., 2014). All tests for M. ovipneumoniae were performed at Washington Animal Disease Diagnostic Laboratory (WADDL).

Monitoring of juvenile bighorn sheep

From 2016 to 2018, we conducted semi-monthly observations of all collared adult females between April 1 and August 31. Juvenile identification was determined via observation of physical contact between adult females and juveniles, such as nursing or bedding down together. Juveniles are weaned at approximately 4-months of age (Festa-Bianchet, 1988); thus, our observation period was intended to cover birth through weaning. We located adult females for observation using an R-1000 telemetry receiver fitted with an RA-23K VHF directional antenna (Telonics, Inc., Mesa, AZ). We conducted observations with Kowa TSN-601 spotting scopes fitted with a 20–60x magnification mounted on tripods. Once an adult female was confirmed to not or no longer have a juvenile, i.e., two consecutive observations where the adult female was observed without a lamb, we stopped tracking that individual adult female (e.g., Cassirer & Sinclair, 2007).

We opportunistically located dead juveniles and collected samples, including either the entire corpse, the pluck (heart, liver, and lungs), the head, nasal and ear swabs, and/or tissue samples depending on the state of decomposition. Swabs were inserted into vials dry or containing tryptic soy broth, and other samples put in a cooler until returning to our field base. We then stored both the samples and swabs at −20 °C until laboratory submission. WADDL conducted gross- and histo-pathology on lung tissue samples and PCR tests on swabs. Additionally, M. ovipneumoniae strain-typing using multi-locus sequence typing (MLST) tests with four locus sequences, 16-23S intergenic spacer regions, the small ribosomal subunit, the genes encoding RNA polymerase B, and gyrase B, was conducted on lung tissue and swab samples from two of the juvenile mortalities recovered (Cassirer et al., 2017). We then compared these strain-types to strain-typed samples from the Santa Rosa metapopulation (n = 6) and from the Rattlesnakes (n = 2). WADDL conducted the M. ovipneumoniae strain-typing.

Genetic sampling

We used a combination of both blood samples (n = 125) and feces (n = 66) as sources for DNA samples. Whole blood (3 mL) provided by ODFW and NDOW from captured bighorn sheep was collected in EDTA tubes and spun at 4,000 × g for 10 min to separate the buffy coat. We extracted DNA from this material using a Qiagen DNeasy Blood and Tissue Kit (Qiagen Inc., Valencia, CA, USA). Fecal samples were collected opportunistically while conducting observations of bighorn sheep across the different populations, and were generally a week or less in age, as estimated from pellet color, odor, and surface condition. Fecal samples that were still moist after deposition were dried and then stored at room temperature. Fecal pellets were scraped to target dried epithelial cells on the surface of the pellet (Wehausen, Ramey & Epps, 2004), and we extracted DNA from the scraped material using a modified version of the Aquagenomic Stool and Soil protocol (Multitarget Pharmaceuticals LLC, Colorado Springs, CO; see details in Appendix S4).

Genotyping, markers, individual identification, and marker

We used a suite of 16 microsatellite markers in three panels (Table S2) that had previously been used to investigate population connectivity and genetic variability in bighorn sheep (Creech et al., 2020; Creech et al., 2020; Epps, Crowhurst & Nickerson, 2018). Genotyping followed protocols outlined in Epps, Crowhurst & Nickerson (2018). Briefly, all samples were run in at least two (for blood) or three (for feces) independent PCR reactions to generate a consensus genotype for each individual at each locus. For blood samples, any discrepancy between the two replicates resulted in the sample being rerun at that panel, although consistency across replicates was very high. Because allelic dropout can be higher in fecal samples, for those samples a homozygous genotype was considered verified if the single allele was seen in all three replicates. A heterozygous genotype was considered to be verified if each allele was seen in at least two of the three replicates; any other discrepancies resulted in reruns. Other studies on bighorn (e.g., Epps, Crowhurst & Nickerson, 2018) reported screening for recaptures using as few as 6 loci to achieve a desired probability of identity (PID; Waits, Luikart & Taberlet, 2001) of <0.001 and a probability of identity for full siblings of <0.05. However, our initial genotyping demonstrated that we needed to genotype at all 16 loci to achieve those thresholds. We identified recaptured individuals using cervus (Kalinowski, Taper & Marshall, 2007) by screening for individuals that matched at all 16 loci and removing them from the data set. We repeated this analysis using successively reduced numbers of matching loci and the presence of one to two mismatches (to account for missing data and genotyping error, respectively), until the matches that the program returned seemed unlikely due to geographic location and/or the mismatches were not explainable by simple allelic dropout. Finally, we used gimlet (Valière, 2002) to calculate two types of error rates in our genotypes: allelic dropout, and the presence of false alleles.

Linkage disequilibrium, Hardy-Weinberg tests, and genetic diversity

We used genepop Version 4.2 on the Web (Rousset, 2008) to conduct the probability test for Hardy-Weinberg equilibrium (HWE) for each population by locus and then for each locus by population, as well as across populations for each locus and across loci for each population (Fisher’s method, Fisher 1948). We then used genepop 4.0 Desktop to test for linkage disequilibrium across each pair of loci within each population, and each pair of loci across all populations, applying a sequential Bonferroni correction in both cases across all loci. We used the R package diveRsity (Keenan et al., 2013) to calculate population genetic diversity metrics including expected heterozygosity (H_E), observed heterozygosity (H_O), and allelic richness (A_R). We accounted for imbalanced sample sizes amongst populations using rarefaction.

NDVI data

We used 14-day composite, 250 m resolution NDVI data from the Moderate Resolution Imaging Spectroradiometer (eMODIS). We utilized pre-processed data from 2016-2018 obtained from Earth Explorer (https://earthexplorer.usgs.gov/), which is managed by the United States Geological Survey’s Earth Resources Observation Center (Jenkerson, Maiersperger & Schmidt, 2010).

We used GPS data from all collared adult females in each population to generate a single utilization distribution per population for each year from 2016–2018, using the R package adehabitatHR (Calenge, 2006). We estimated 95% utilization distributions using the kernel method with the default smoothing parameter (h_ref). We did not use least square cross validation (h_lscv) due to repeated use of locations, which may cause convergence issues (Kie et al., 2010).

We then extracted NDVI data from each population polygon for each 14-day composite image and generated a 90th percentile statistic using program R (R Core Team, 2019). Bighorn sheep are selective feeders; as such, we assume that the 90th percentile NDVI statistic represents a high-quality choice of forage, while accounting for the fact that maximum forage is not always attainable (Creech et al., 2016). Additionally, selecting the 90th percentile excludes the selection of false maximum values caused by measurement error. Finally, we averaged NDVI values across each 14-day composite for the 3-months before and after the first juvenile observed in each population and used this variable as a measure for forage quality for each population pre- and post-parturition.

Drivers of juvenile survival

We analyzed our known-fate data in Program Mark (White & Burnham, 1999), to estimate juvenile survival (S) using a Kaplan–Meier estimator (Kaplan & Meier, 1958) with staggered entry (Pollock et al., 1989). First, we considered three population-level measures of exposure to M. ovipneumoniae in our models of juvenile survival (Table 1). Primarily, we considered the presence of M. ovipneumoniae-infected juveniles in each population, as the limited numbers of adults captured and tested in each population and the lack of annual testing precluded clear estimates of infection rates among adults. However, we also considered whether presence of infected adults (PCR) or exposed adults (cELISA) in each population influenced juvenile survival. Second, we considered two population-level measures of genetic diversity, expected heterozygosity (H_E) and allelic richness (A_R) (Table 1). We used univariate models of the three M. ovipneumoniae measures and the two genetic diversity measures as initial screening methods, using Akaike’s information criterion, corrected for small sample sizes (AIC_c) to determine which M. ovipneumoniae and genetic diversity metric most strongly linked to juvenile survival before proceeding with subsequent analyses.

Subsequently, we conducted two analyses of juvenile survival. The first included all study populations (n = 13) and the selected measures of M. ovipneumoniae presence in each population, population-level metrics of forage quality (3-month pre- and post-parturition NDVI values) and genetic diversity (expected heterozygosity, H_E) (Table 1), and both additive and multiplicative effects of time. For each model that included the multiplicative effect of time and M. ovipneumoniae, we fixed survival interval 1 for the M. ovipneumoniae group, as no mortalities occurred during this period. The second analysis included all of the same variables except M. ovipneumoniae presence and was restricted to populations where M. ovipneumoniae-infected juveniles were not detected (n = 11, see Results). That second analysis was undertaken to determine whether effects of M. ovipneumoniae obscured the effect of the other covariates of interest. For both analyses, we used AIC_c and AIC_c weights (w_i) to select the best-supported model. We included a null model in both model selection sets to evaluate model performance (Burnham & Anderson, 2002). We selected the model with the lowest AIC_c and highest w_i as our best-supported model. We used evidence ratios between the top model and competitive models (those within 2 AIC_c units, to evaluate each model relative to the top model (Burnham & Anderson, 2002).

Post-study observations

After our final planned field season in 2018, the single PCR+ adult female in the RSP (of 21 tested) died of suspected bluetongue (Orbivirus spp.). Because prevalence of PCR+ adult females in this population was low (4.76% of tested—see Results), and no adult males tested PCR+ throughout the study, we considered it possible that no additional PCR+ individuals remained at RSP, potentially removing the source of infection for new juveniles. Therefore, we decided to conduct a single observation of juveniles in the RSP and BHP, populations in early August of 2019; we included BHP given its proximity to RSP. Juveniles in both these populations were typically 4-months old at that time; thus, that observation aligned closely with the 4-month juvenile survival we estimated during the study.

Diagnostics

Between 2016 and 2018, 78 adult females were tested via cELISA to determine M. ovipneumoniae exposure, and 95 adult females were tested via PCR to determine active M. ovipneumoniae infections. The proportion of adult females PCR tested in each population varied from 0.10 to 0.43 ($\overline{x}=0.27$, Table S4). For the 10 adult females in the Santa Rosa metapopulation that were recaptured during our study and retested for infection via PCR, PCR status remained the same for the single positive individual in EML, and the rest were negative. None of the adult females in populations west of U.S. Route 95 (BSP, TCE, TCS, or TCW) showed evidence of M. ovipneumoniae exposure (Fig. 1, Table S3). However, all of the populations east of U.S. Route 95 (the four populations in the Santa Rosa Mountain’s metapopulation, as well as BHP, RSP, TMP, and UOP (Fig. 1)), except TFK, included adult females with evidence of exposure to M. ovipneumoniae (Table S3). We were unable to get a sample from the single individual adult female captured in TFK. The proportion of M. ovipneumoniae exposed adult females in those eight populations varied between 0.60 in CAL and EML to 1.00 in TMP ($\overline{x}=0.75$; Table S4). Only two adult females tested PCR+ to M. ovipneumoniae infection across all 13 populations (Table S3): one in EML on two occasions (2017 and 2018), and one in RSP in 2016. Additionally, one adult female in BHP yielded an indeterminate PCR test result, meaning M. ovipneumoniae detection could not be determined (Besser et al., 2019).

Linkage disequilibrium, Hardy-Weinberg tests, and genetic diversity

After removing the fecal samples from recaptured individuals (n = 28), the genetic data set contained 191 individuals at 16 microsatellite loci, representing 10 to 26 ($\overline{x}=15.9$) individuals per population (Table S5). The mean rate of false allele occurrence per locus was 0.001 and the mean allelic dropout rate across all loci was 0.008 (range = 0.000–0.017). No locus was determined to be out of Hardy-Weinberg equilibrium by either test employed (Tables S6 & S7). Evaluating each locus pair by population showed no evidence of linkage disequilibrium (p_critical = 0.000035 for α = 0.05); evaluating linkage for each pair of loci across populations using Fisher’s test indicated that BL4 and HH62 were in disequilibrium (p = 0.00039; p_critical = 0.00042 for α = 0.05). However, those loci have not appeared to be in disequilibrium in other bighorn sheep studies (e.g., Epps, Crowhurst & Nickerson, 2018), suggesting that this relationship may have been an artifact. Thus, and because linkage disequilibrium is more likely to bias estimates of genetic structure rather than estimates of genetic diversity as employed in this study, we retained all 16 loci in our analyses. H_E across the 13 study populations varied from 0.26 in BSP to 0.48 in EML ($\overline{x}=0.37$; Table S5) and A_R varied from 1.82 in BSP to 2.88 in SAW ($\overline{x}=2.37$; Table S5).

NDVI data

Pre- and post-parturition NDVI varied spatially and temporally (Table S8). BSP had consistently low pre-parturition NDVI, whilst EML and TMP had consistently high pre-parturition NDVI (Table S8). Post-parturition NDVI was lowest in BHP in 2018, was consistently low in RSP but was consistently high in TCE and EML (Table S8). Pre- and post-parturition NDVI values were not correlated (Table S9).

Pregnancy rates and observation of juvenile bighorn sheep

Seventy-six of the 82 (93%), pregnancy tests conducted across all 13 populations between 2016 and 2018 were positive (Table S10). Six of the 82 tests conducted were on recaptured adult females from EML (n = 3) and SAW (n = 3), all of which were positive. All populations had 100% pregnancy rates, except for TCE (67%; n = 8∕12), TCS (50%; n = 1∕2), and TFK (0%; n = 0∕1) in 2016. The single collared adult female in TFK did not yield a positive pregnancy test result in January, but was later observed in early June with a juvenile, suggesting that the test was administered too early to detect a positive pregnancy result. Sixty-five of those 82 (79%) pregnant adult females survived to parturition; of those 59 (91%) were observed with juveniles. Population pregnancy rates were not correlated with genetic diversity (Pearson pairwise correlation; H_E, r = 0.24, p = 0.272; A_R, r = 0.14, p = 0.529).

We observed 121 juveniles with radio-collared adult females between 2016 and 2018; 78% of radio-collared adult females were observed with juveniles. Populations with the lowest proportion of radio-collared adult females with juveniles were BSP (2016—0.00; 2017—0.67; 2018—0.33), and TCE (2016—0.64; 2017—0.64; 2018—0.17), although sample sizes were small in BSP (n = 3). A set of twins was observed with a radio-collared adult female in both the BHP and MCK populations in 2018.

The observation rate of collared adult females with juveniles varied from 50 to 100%, with a mean observation rate of 94% across all semi-monthly sampling periods and study populations (Table S11). Throughout the study, we collected samples suitable for M. ovipneumoniae testing from 17 juvenile mortalities (Table S12). All juvenile mortalities (n = 15) recovered from the BHP (n = 1) and RSP (n = 14) populations tested positive for M. ovipneumoniae. None of the other juvenile mortalities from BSP (n = 1) or TCS (n = 1) tested positive for M. ovipneumoniae. The M. ovipneumoniae strain-type of the samples collected from the BHP and RSP juveniles matched the strain-type (NV_BHS_SantaRosas_2651_2014_4; Kamath et al., 2019) of all the other previously strain-typed bighorn sheep at the available sequences from the Santa Rosa meta-population and Rattlesnake population (Table S13).

Drivers of juvenile survival

Preliminary analyses with univariate models of the relationship between juvenile survival and different population-level measures of M. ovipneumoniae revealed that the presence of M. ovipneumoniae-infected juveniles in the population was a much stronger predictor of juvenile survival than the presence of infected (PCR+) or exposed (ELISA+) adults (Table S14). Preliminary analyses with univariate models of juvenile survival as a function of genetic diversity demonstrated that H_E was a stronger predictor of juvenile survival than A_R (Table S14). Therefore, in our subsequent multivariate models of juvenile survival, we used the presence of M. ovipneumoniae-infected juveniles in each population as our measure of M. ovipneumoniae presence, and H_E as our measure of genetic diversity. In those multivariate analyses, we identified two competing models predicting survival probability of juveniles across all populations (Table 2). Both models contained only two predictors: (1) whether M. ovipneumoniae-infected juveniles were detected, and (2) time. The model containing time as a multiplicative effect with M. ovipneumoniae had 2.36 times more support than the competing model that treated time as an additive effect (Table 2), meaning that the temporal pattern of juvenile mortality differed in populations where M. ovipneumoniae-infected juveniles were present. Neither forage quality (pre- and post-parturition NDVI), nor genetic diversity (H_E) predicted survival probability of juveniles (Table 2).

The odds of a juveniles surviving to 4-months of age in our study populations where M. ovipneumoniae-infected juveniles were detected were 8.00 (95% CI [−4.00–255.78]) times less likely than juveniles in populations where we did not detect M. ovipneumoniae-infected juveniles (Table 3). The derived probability of survival for the entire 4-month annual study period for juveniles in populations where M. ovipneumoniae-infected juveniles were detected was 0.02 (95% CI [0.00–0.13]) compared to 0.44 (95% CI [0.29–0.59]) for juveniles in other populations (Fig. 2B).

We observed no significant difference in semi-monthly juvenile survival probability between populations where M. ovipneumoniae-infected juveniles were not detected in observations at 0.5, 1, and 1.5-months post-parturition (Fig. 2A). However, semi-monthly survival probability in populations where M. ovipneumoniae-infected juveniles were present were significantly lower 2-months (S exposed = 0.42, 95% CI [0.24–0.62]; S unexposed = 0.88, 95% CI [0.75–0.95]) and 2.5-months (S exposed = 0.40, 95% CI [0.16–0.70]; S unexposed = 0.94, 95% CI [0.82–0.98]) post-parturition (Fig. 2A). At 3, 3.5, and 4-months post-parturition, estimates in semi-monthly survival probability again did not differ between populations where M. ovipneumoniae-infected juveniles were or were not detected (Fig. 2A). The increase in uncertainty in our estimates is tied to decreasing sample size as individuals died.

Although no direct effect of either population genetic diversity or post-parturition NDVI was supported based on our model selection approach, both variables were highly correlated with the presence of M. ovipneumoniae (H_E = 0.58; post-NDVI = −0.66, Table S9). In fact, the two populations where M. ovipneumoniae-infected juveniles were present, BHP and RSP, had relatively high genetic diversity (H_E: BHP = 0.43; RSP = 0.45) compared to other study populations (Table S5). Similarly, RSP had low post-parturition NDVI values relative to the other populations in 2017 and 2018, while the same was the case for BHP in 2018 (Table S8).

In the second analysis, which excluded the two populations where M. ovipneumoniae-infected juveniles were present, competing models included both indices of forage quality and population genetic diversity as predictors of juvenile survival (Table 2), as well as a time-varying effect. However, the 95% confidence intervals for the parameter estimates of the nutritional and population genetic diversity model both overlapped zero, and the null model with no predictors was 2nd ranked. The top competing model, a time-varying model, had 1.69 times more support than the next best model, the null model (Table 2).

Post-study observations

In early August of 2019, following the death of the single adult female in RSP that tested PCR+ for M. ovipneumoniae, we located the two remaining collared individuals in the BHP (linked to RSP by adult male and female movements, data not shown). Neither the collared adult females nor any of the eight other adult females we observed in this population had juveniles (0% juvenile survival). We then located 14 of the RSP collared adult females, of which 11 had juveniles (∼79% juvenile survival). We located approximately 35 adult females total in the RSP population and observed approximately 27 juveniles, which equates to an approximate 4-month survival rate of ∼77%.