Introduction

Reproduction is one of the most physiologically taxing processes of amniote life, particularly among capital breeders1. Trade-offs between finite resources invested toward survival and reproduction are thus ubiquitous and should be especially pronounced in females, which bear the brunt of reproductive expenditure2,3. Capital breeders are especially vulnerable to population declines when female survival or reproductive success is compromised by resource limitations or acute selection pressures, such as emergent disease3,4,5. Thus, understanding how reproductive costs and ecological stressors inform female survival is essential for identifying and responding to threats to population stability6.

Many macrostomate snakes are capital breeders3, which accumulate requisite energetic reserves prior to reproducing. During reproduction, somatic stores (i.e., muscle and fat) are often substantially depleted by the effects of increased metabolic rate7,8,9, autophagy associated with anorexia/hypophagia10, formation of ova11, matrotrophy (to varying degrees12), and obligate water provisioning13,14. Such reproductive costs reduce resources available for growth and maintenance, prompting delayed maturity at larger body sizes15. Larger females may more readily meet the physiological requirements of reproduction16,17,18,19 and better withstand reproductive costs, gaining the advantage of surviving a greater number of pregnancies throughout their lifetime20,21. However, ecological pressures and senescent processes can shift directional selection for body size, optimizing fitness advantages to intermediate sizes22,23. In iteroparous species, high reproductive costs are also associated with long birthing intervals required to restore capital3, hinging fitness on survival24,25.

When the timely recovery of maternal investments (e.g., lipids, muscle) and/or adaptive responses to stressors (e.g., disease) fail, reproduction is expected to reduce survival26,27,28,29. Ovoviviparous snakes, which express a transitional reproductive mode and an impressive capacity for reproductive plasticity, are ideal organisms for studying trade-offs between survival and reproduction. Survival may be negatively affected months or years following parturition30. Yet, few studies26,31,32,33 have accounted for temporal lags in mortality costs among live-bearers. Parturition occurs shortly before winter, increasing the potential for brumation to delay starvation-associated mortality in temperate species26,30. However, in southerly (subtropical) populations, dormant season survival could be complicated by sporadic winter activity34. Lack of information on free-ranging species reflects the limited availability of telemetry data that spans timeframes (i.e., months or years) necessary to assess delayed survival effects.

Reproduction has suppressive effects on immunocompetency, reflecting support of pregnancy and conservation of physiologic processes during periods of constrained resources35,36,37. In females, immune costs can be fecundity-dependent or independent, varying with adaptive or innate immune responses and biochemical components involved38. Pregnancy can also exacerbate trade-offs related to immunocompetency and other forms of maintenance (e.g., digestion39). Immune suppression could continue well beyond the reproductive cycle in slow-lived capital breeders, given evidence that body condition is positively associated with immune function40 and postpartum females are often emaciated41. In temperate latitude species, females of suboptimal body condition may be particularly vulnerable to disease in the winter as ambient temperatures are probably insufficient for mounting immune responses42.

We used an information theoretic approach to assess survival of mature female Crotalus adamanteus (eastern diamondback rattlesnakes; EDBs) using radio-telemetry data collected in the twelve months (October 2022–September 2023) following the parturition season (August–September 2022) on the Marine Corps Recruit Depot at Parris Island, South Carolina (MCRDPI). We evaluated six survival covariates: 1) reproductive status (reproductive or nonreproductive), 2) relative fecundity, 3) body size, 4) behavioral season, 5) body condition, and 6) severity of clinical signs of disease consistent with ophidiomycosis (also known as snake fungal disease), caused by the pathogen Ophidiomyces ophidiicola (Oo). We expected reproduction would be negatively associated with survival, but that survival would be optimized when reproductive investment (i.e., relative fecundity) was intermediate. We further hypothesized that intermediate body sizes would be favored, but poor health (i.e., poor body condition and greater disease severity) and dormancy (i.e., October–January) would negatively affect survival.

Methods

Data collection and radio-telemetry

From 2021–2023, we conducted visual encounter surveys for mature females (≥ 100 cm snout-vent-length34,43; SVL). We measured body mass (± 1 g), SVL (nearest cm), and determined sex using cloacal probes. We tagged unmarked snakes using subcutaneous unique Passive Integrated Transponder (PIT) tags, injected approximately twelve ventral scales cranial to the anal plate. We used ventral scale cauterization44 as a secondary mark but withheld cautery in some instances to minimize stress or dermal injury (e.g., during late gestation or when snakes appeared substantially afflicted with dermatological disease, evidenced by lesions or growths). We recorded SVL only at the time of first capture in 2022 to minimize subsequent handling times. Growth rates decrease when female EDBs reach maturity34, and are likely negligible during gestation19,32. Thus, we assumed growth rates were too slow to detect short-interval changes in body size. We used bleach or alcohol to sanitize snake handling and processing equipment between uses following Lind et al.45 or Rzadkowska et al.46.

We equipped snakes with very-high-frequency external (ATS, R1640, 2–2.5 g or HOLOHIL, BD-2, 3 g) or internal (HOLOHIL, SI-2, 13–13.5 g) radio-transmitters following methods adapted from Jungen et al.47 or Reinert & Cundall48, respectively (Supplementary Table S1). All transmitters weighed < 2% of snake mass, and we did not implant snakes judged to be ill or emaciated (based subjectively on tail corpulence and spinal protuberance). We conducted transmitter implantation surgeries under sterile conditions within a veterinary hospital, March–August 2022. We radio-located snakes 1–3 times weekly and maintained 2–5 m from snakes to minimize disturbance. We captured telemetered snakes during the dormant season (October–January) to measure body mass; not all snakes were available for capture in winter, affecting the sample size of summary statistics.

We conducted all research procedures under approved Marshall University Institutional Animal Care and Use Committee protocol (IACUC #759) and South Carolina Department of Natural Resources scientific collection permits (#SC-20-2021, #SC-20-2022, and #SC-20-2023); all methods followed relevant guidelines and regulations. Our reporting follows ARRIVE guidelines (https://arriveguidelines.org).

Survival

We condensed weekly telemetry observations into 12 monthly intervals (corresponding to October 2022–September 2023) to assess true survival using a known fate framework49. We right-censored encounter histories after snakes died or were lost (e.g., transmitters failed or were dropped) before the study conclusion. We used two datasets to examine covariate effects on adult female (AF) survival and predictors of post-reproductive female (PF) survival in the year following parturition. The AF dataset included encounter histories of all adult females, regardless of their reproductive status (i.e., females were reproductive or nonreproductive in 2022). We used this combined dataset to examine survival as a function of five predictors: (1) body size, (2) reproductive status, (3) dormant season body condition, (4) season, and (5) severity of lesions consistent with ophidiomycosis. The PF subset was limited to encounter histories for post-reproductive females (i.e., those that gave birth in August or September 2022), which allowed us to examine the effects of relative fecundity on survival in addition to body size, body condition, and season.

We quantified body size (z-standardized SVL) and body condition covariates independently for each dataset (AF and PF). We used standardized residuals50 from ordinary least squares regression between log(SVL) and log(mass) to describe dormant season body condition (dBCI). We used the most recent record of snake mass during October–January to calculate dBCI. For example, if a snake was weighed in October but thereafter died in January, we used the postmortem mass if we detected (visually/olfactorily) no signs of advanced decomposition or scavenging affecting the carcass. We categorized reproductive status as a binary factor (0 = nonreproductive, 1 = reproductive), which we diagnosed via ultrasonography (MediSono P3V, 2–10 MHz). We determined fit litter size based on follicular echogenicity51,52. To quantify relative fecundity (FLS), we extracted residuals from an ordinary least squares regression between fit litter size and maternal SVL. We held SVL constant across calculations of dBCI, FLS, and z-standardized SVL.

We examined seasonal effects on survival by stratifying monthly encounters into periods of late fall/winter (dormant season; October–January, intervals 1–4) or spring/early fall (active season; February–September, intervals 5–12), which represented behaviorally-based seasons for EDBs at the study site. The dormant season encompassed months when we encountered telemetry-equipped females at hibernaculum, while the active season included egress, foraging, and breeding seasons, based on our EDB telemetry observations.

We used ordinal scores of disease severity as a survival covariate for the AF dataset. It was not the original aim of our study to assess the effects of disease on the MCRDPI population, but we observed a dramatic increase in the incidence and severity of dermal lesions on EDBs during dormant season body condition assessments, suggesting an emergent conservation concern. All lesions appeared consistent with the presentation of ophidiomycosis53,54 (e.g., Supplementary Figure S1a). We categorized disease severity using lesion scoring criteria adapted from McCoy et al.42: 0 = absent, 1 = minimal [< 5], 2 = numerous [> 5], 3 = numerous and/or present on face, cloaca, or internal organs. We assigned scores post facto using notes or media recorded at or nearest to the time of dormant season condition assessments that described lesion number and placement. When quantitative data were unavailable, we relied on qualitative descriptions to assign scores, e.g., “some lesions” = 1, “a significant number of lesions” = 2, “lesions on ventral caudal half of snake” = 3. All scoring was conducted by E.R.G. We collected intact EDB carcasses with unknown cause of death for gross necropsy or necropsy with histopathology, respectively, which were performed by the Avian and Exotic Animal Hospital of Georgia or the Athens Veterinary Diagnostic Laboratory at the University of Georgia. Necropsy findings usurped severity classifications based on gross presentation in the field, i.e., females with disseminated fungal granulomatous disease (Supplementary Figure S1b) were assigned the highest severity score (‘3’).

We fit 15 candidate models using the AF dataset to evaluate survival as a function of season, dBCI, reproductive status, additive and interactive effects of SVL (both linear and second-order relationships), and as a constant. Due to convergence issues, we fit a model examining the interaction between reproductive status and disease severity using simulated annealing optimization. We fit 17 candidate models to assess annual survival of post-reproductive females (PF subset), which examined survival as a constant, by season and dBCI, and as linear and second-order polynomial functions of SVL and relative fecundity. In this analysis, we were unable to include disease severity as a covariate due to high uniformity; all but four PF females had scores of ‘3’. See Supplementary Table S2 for all candidate model structures and expectations of covariate effects.

We retained all outliers and assessed covariate collinearity using R Statistical Software55 (package: stats, v.4.4.2 or ltm, v.1.2-056). We fixed four dBCI and three fecundity values to “0,” representing respective means when true values were not known. This approach allowed us to retain these individuals within analyses without influencing the magnitude or directionality of relationships examined. Using program MARK49 (v. 10.1), we ranked models using a modified Akaike’s Information Criterion (AICc) and evaluated covariate effects using 95% confidence intervals of model-specific and model-averaged coefficients49,57. For supported models, we used weighted averaging (based on AICc weights) to estimate monthly survival probability58. We estimated annual survival by averaging derived estimates across supported models. We report estimates with unconditional standard errors, which account for model selection uncertainty in parameter estimation57,58.

Results

Summary of samples

The AF dataset included 23 females (15 reproductive, 8 nonreproductive) and the PF subset included 14 females; one female was eggbound in 2022, which was unknown at the study start, and we thus considered her reproductive but did not include her in the PF subset. Females averaged 121 cm (SD = 6 cm, n = 23) SVL and 1293 g (SD = 304 g, n = 19). Reproductive (postpartum or eggbound) females were 122 cm (SD = 6 cm, n = 15) SVL and 1248 g (SD = 354 g, n = 13) and nonreproductive females were 118 cm (SD = 6 cm, n = 8) SVL and 1390 g (SD = 121 g, n = 6), on average. We monitored five of 23 females with internal transmitters, three with external transmitters, and 15 with a combination thereof (Table S1). Nineteen (83%) of 23 females showed clinical signs of disease consistent with ophidiomycosis, 11 (58%) of which were severe. Two (13%) reproductive females (n = 15) had no clinical signs of disease, three (20%) had minor or moderate signs, and ten (67%) had severe signs of disease. Three (38%) nonreproductive females (n = 8) lacked clinical signs, four (50%) had mild or moderate signs, and only one (13%) had severe signs. Dermal lesions occurred primarily on the cloaca and ventral scutes along the caudal third of the body. Two of three females with clinical signs of ophidiomycosis were positive for Oo based on commercial testing; both Oo-positive females were tested antemortem, while the Oo-negative female was tested postmortem (Supplementary Table S3). Of the seven snakes that underwent necropsy with histopathology examination, fungal granulomas were found on the liver (n = 5, 71%), trachea or lungs (n = 5, 71%), pancreas (n = 1, 14%), kidney(s) (n = 2, 29%), and GI tract (n = 1, 14%), though autolysis occasionally prevented assessment of some organs. All snakes with systemic granulomas had concomitant dermal involvement, though extent was variable. Litter size of reproductive females averaged 12 neonates (SD = 3, range: 7–18, n = 11).

Survival

Of the 23 females monitored, we recovered nine (39%) deceased (reproductive = 8, nonreproductive = 1). Two dropped their external transmitters and were lost. Eight of 14 females included in the PF subset died following parturition (mean = 7 months, range = 2–12); all but one were afflicted with systemic fungal granulomatous disease (ophidiomycosis clinically suspected or confirmed in all cases; Supplementary Table S3). Predictors were not correlated (\(\left|\text{r}\right|\)≤ 0.7). Disease severity and dBCI were important predictors of female survival (AF dataset; Supplementary Table S5). The top model, {S(DS)}, detected a negative effect of DS on survival (\(\hat {\beta}\) = − 0.9749, SE = 0.4759, 95% CI = − 1.9077 to − 0.0421; Supplementary Table S5, Fig. 1a), and the third-ranked model, {S(dBCI)}, revealed a positive effect of dBCI (\(\hat {\beta}\) = 1.0029, SE = 0.4460, CI = 0.1286–1.8772, n = 23, Supplementary Table S5, Fig. 1b). Reproductive status (STATUS) was not statistically supported. Further, parameters included in the model fit to examine effects of STATUS and DS, i.e., S(DS*STATUS)} were uninformative, evidenced by large SE relative to beta coefficients; we thus culled the model from our AIC selection table59. Model-averaged estimates of monthly and annual survival were 97% (SE = 2%, 95% CI = 92%–99%) and 68% (SE = 13%, 95% CI = 41%–87%; Supplementary Table S6), respectively. Three models examining post-reproductive female survival received support (PF subset; Supplementary Table S4). The top-ranked model, {S(dBCI)} suggested a positive effect of dBCI (\(\hat {\beta}\) = 0.9223, SE = 0.5516, 95% CI = − 0.1588–2.0035), but 95% CIs included zero; no covariates received statistical support. Model-averaged annual survival was 53% (SE = 14%, 95% CI = 27%–77%) and monthly survival was 95% (SE = 2%, 95% CI = 89%–98%) (Supplementary Table S6).

Fig. 1
figure 1

Monthly survival probability of female EDBs (AF dataset). Monthly survival probability of mature female eastern diamondback rattlesnakes (Crotalus adamanteus; n = 23) given (A) ophidiomycosis-consistent disease severity (0–3, in which 0 represented no clinical signs of disease, 1 represented minimal signs of disease, 2 represented moderate disease, and 3 represented highest observed disease severity); and (B) dormant season body condition (positive or negative signum represented above- or below-average body condition, respectively). Whiskers represent standard errors, gray shading represents 95% CIs.

Full size image

Discussion

While we failed to detect significant effects of reproductive covariates on female survival, this study was the first to document that severity of clinical signs of disease suggestive of ophidiomycosis is negatively associated with true survival of wild snakes. Annual survival (68%) was the lowest estimate on record for the focal species: 17% less than the estimate provided by Waldron et al.34 for inland South Carolina EDBs and 19% less than the estimate provided by Kelley et al.60 for resident MCRDPI EDBs, though it should be noted that prior studies included both sexes. Similarly, monthly survival probability of females with severe disease (i.e., 91%) was 8% lower than the estimate (i.e., 99%) provided by Kelley et al. Our results are particularly concerning given the focus on mature females, which drive recruitment and function as the key demographic for maintaining population viability via high survival34. Thus, our work poses concerning implications for the vulnerability of EDBs (and other snakes) across the Southeast if similar disease dynamics are currently present, or should begin, at other sites. Relatively recent reports from the Southeast indicate observations of ophidiomycosis-attributed mortalities among crotalids are not restricted to MCRDPI. For example, cases of ophidiomycosis in an opportunistically encountered adult female EDB from North Florida in 2020 (U.S. Geological Survey, National Wildlife Health Center, personal communication) and a radio-tracked adult male timber rattlesnake (C. horridus) from Central Georgia in 202261 share pathological similarities with those observed on the MCRDPI.

Although this study was not originally designed to test hypotheses related to effects of disease on survival and not all individuals were evaluated for criteria necessary to clinically diagnose ophidiomycosis62, our findings underline advancing concern for the long-term effects of Oo infection on snake biodiversity. Slow-lived snakes are especially vulnerable to increases in adult mortality63, and adults experience complex reproductive trade-offs (e.g., trade-offs between reproduction and immunocompetency38) that can make them especially vulnerable to acute selection from emergent disease. Further, Oo has potential to reduce recruitment by not only truncating female longevity but concomitantly increasing juvenile mortality64 and interfering with adult reproductive cycles as a sublethal effect65. Our work thus highlights the need for further research into (1) population-level effects of Oo, which may vary by species, reproductive mode, and local selection pressures (e.g., there is no support for a direct influence of ophidiomycosis on eastern foxsnakes [Pantherophis vulpinus] survival66), (2) variation in mortality and morbidity associated with Oo strain67, and (3) strategies for managing disease in vulnerable populations. It is unknown whether the apparent outbreak of ophidiomycosis-consistent disease at our study site resulted from the introduction of a locally novel fungal strain, underlying population health issues, environmental factors, or various interactive effects. Ophidiomyces was first tested for and detected at the MCRDPI in 2015 (T. Norton, unpublished data) and affects EDBs and colubrids68. Targeted research to better understand Oo and ophidiomycosis dynamics at the site began in 2023 and is ongoing.

As dermatological injury can affect Oo pathogenesis69, careful consideration of research risks/benefits with respect to goals, focal organism, and project timeframe is necessary for developing appropriate monitoring protocols70. For example, while there was no pathological evidence that transmitter implantation affected disease progression among snakes in the present study, 2024 pathological findings from a female of this cohort indicated that transmitter implantation (which occurred 22 months prior to death) can precipitate systemic ophidiomycosis (Southeastern Cooperative Wildlife Disease Study, unpublished data). Though we were unable to evaluate transmitter (external versus internal) effects, recent work indicates clinical signs of ophidiomycosis and Oo prevalence decreased with the cessation of internal transmitter use within a population of eastern massasauga rattlesnakes (Sistrurus catenatus)71. These observations suggest use of less invasive monitoring methods is a safer approach for monitoring vulnerable species/populations.

We acknowledge that noninvasive marking and monitoring techniques may not be feasible for all taxa or sufficient to meet certain project goals. Characterizing behavior, true survival, and other fitness metrics is only possible via individual-based monitoring72,73,74, and though we have detected clinical signs of advanced disease in untelemetered and unmarked EDBs, we would not have been able to assess disease effects on true annual survival without the use of internal transmitters, especially given how slowly disease can progress in reptiles. We thus advocate for improvements in transmitter implantation/removal protocols that can ameliorate increased fungal disease risks. Additionally, our findings emphasize the importance of advancing and prioritizing biosafety methods during ophiological research. Knowledge of the efficacy of isopropyl alcohol, glutaraldehyde, betadine, zephiran chloride, and hydrogen peroxide as Oo disinfecting agents would be particularly useful.

The significance of dBCI for female EDB survival is consistent with Waldron et al.34. In addition to the direct risk of starvation, poor nutritional status during the dormant season could prompt winter surface activity (e.g., foraging or basking), elevating risks of exposure or predation. One year of postparturient monitoring may have been insufficient for detecting a delayed effect of dBCI on survival in our small sample. Confounding effects of food items, waste, and follicular mass pose unique challenges to quantifying nutritional status in snakes, and the general resilience of snakes to starvation75 may require extended timeframes to confidently describe relationships between survival and energetic reserves. Further, snake body condition can rapidly shift in the wild, likely due in part to cryptic digestive cycles. We have recorded several dramatic mass changes unrelated to reproduction or conspicuous boluses: + 661 g [+ 48%] over 95 days, + 726 g [+ 57%] over 99 days, and -420 g [-18%] over 56 days. Such shifts could interfere with efforts to quantify body condition where one measurement is used to represent nutritional status over an extended period. Yet, while dBCI confidence intervals included zero in the PF subset analysis, it was the only parameter in the top-performing model, and thus a positive association between dBCI and PF female survival warrants future testing.

Annual survival of PF females (i.e., 53%) was lower than anticipated. However, low survival following reproduction in a capital breeding, live-bearing snake is not surprising30. For example, reproductive female European adders experienced 40% annual mortality31 with elevated mortality risks sustained for about nine months following parturition26. Poor post-reproductive survival may also help explain the importance of long birthing intervals26, which allow time for subadult females to reach maturity during lag periods, facilitating the replacement of reproductive females to a population. However, we failed to detect significant effects reproductive parameters (i.e., reproductive status [AF dataset] and FLS [PF dataset]) on survival. Only one (11%) deceased female was nonreproductive, yet over half (57%) of reproductive females died in the year following parturition. We thus suspect a small sample coupled with short (monthly) encounter intervals may have blurred our ability to detect differences in survival related to reproduction.

Our work indicates an urgent need for further research into how fungal diseases (such as ophidiomycosis) impact wild snakes in North America, a challenging feat given the temporal scale at which clinical signs can progress in individuals and changes in demographic processes can manifest within populations. Our work also lends itself to future efforts to forecast EDB status, but additional research is needed to understand threats, demography, life-history attributes, and local selection pressures across the species’ range. To our knowledge, our work provides the most compelling evidence to date that ophidiomycosis can affect population-level vital rates in wild snakes.