Liver Mineral Levels Associated with Hoof Disease Occurrence and Severity in Roosevelt Elk (Cervus canadensis) in California, USA Open Access

Trace minerals are essential to animal health with important roles in nutrition, development, reproduction, and immune function. Host susceptibility and response to infection can be influenced by the availability of trace minerals, including copper (Cu), selenium (Se), and zinc (Zn). Adequate levels of these minerals help maintain the integrity of innate barriers, promote differentiation and production of immune cells, and regulate inflammation (Beisel 1976; Gombart et al. 2020). During infection, host modulation of mineral concentrations can limit mineral availability to bacteria and reduce pathogen survival (Monteith and Skaar 2021). In addition, Cu, Se, and Zn serve direct functions in skeletal formation and keratinization processes critical for hoof development and maintenance in ungulates (Tomlinson et al. 2004; Johnson et al. 2007; Langova et al. 2020). Inadequate levels of these minerals may predispose animals to infection and hoof-related diseases. Livestock with low concentrations of these three minerals have exhibited lameness, heel cracks, sole abscesses, and secondary bacterial infections (Tomlinson et al. 2004; Langova et al. 2020), while an apparent excess of Se has been associated with dysfunction of keratinization and resultant hoof deformities and lameness (James and Shupe 1984). An excess of some minerals may also inhibit the uptake of other minerals, leading to secondary deficiencies (e.g., excess Zn reduces Cu absorption; López-Alonso 2012). Because of these complex and interdependent effects, mineral supplementation has often been used to promote health and reduce the occurrence of hoof diseases in domestic livestock (Gomez et al. 2014; Faulkner et al. 2017; Gelasakis et al. 2019).

Treponeme-associated hoof disease (TAHD) is an infectious disease of free-ranging Roosevelt and Rocky Mountain ecotypes of elk (Cervus canadensis) first identified in Washington State, US, and later reported in Oregon, Idaho, and California, US (Han and Mansfield 2014; Wild et al. 2022; Robinson et al. 2023). Treponeme-associated hoof disease in elk is characterized by ulcerative lesions in the interdigital space and on the soles of the hooves, progressing to deformed, overgrown, and sloughed hoof capsules associated with lameness and debilitation (Han et al. 2019; Robinson et al. 2023). Detection of Treponema spp. is a hallmark finding in TAHD lesions; however, consistent detection of other bacterial families within hoof lesions suggests that TAHD may be a polybacterial and multifactorial disease (Han et al. 2019; Wild et al. 2022).

Note that the contribution of trace minerals to the occurrence and progression of TAHD is unclear. Preliminary investigations of liver samples from a small number of elk within TAHD-affected areas of southwestern Washington have indicated that both apparently unaffected elk and elk with hoof abnormalities have low concentrations of Cu and Se, but normal concentrations of Zn, relative to values from livestock and other elk studies (Han and Mansfield 2014). Although liver is the most biologically relevant sample for determining mineral status (Kincaid 1999), trace mineral analysis of elk hair has shown that relatively low Se concentrations are associated with higher odds of TAHD occurrence (Winter et al. 2022). This early evidence suggested that lower concentrations of trace minerals may influence the occurrence and severity of TAHD; however, balanced sampling of TAHD-affected and unaffected elk from a local area is needed to better characterize these relationships.

Our aims were to 1) establish reference ranges for hepatic trace mineral (Cu, iron [Fe], manganese [Mn], molybdenum [Mo], Se, and Zn) and heavy metal (arsenic [As], cadmium [Cd], lead [Pb], and mercury [Hg]) concentrations in Roosevelt elk from northern California, US, and 2) evaluate associations of liver concentrations of Cu, Se, and Zn with the occurrence and severity of TAHD lesions. Our hypotheses were that lower levels of these three minerals would be associated with higher occurrence and severity of TAHD lesions and that county, sex, and age also might affect occurrence and severity of TAHD lesions.

The California Department of Fish and Wildlife (CDFW) collected feet and livers from Roosevelt elk via hunter harvest, opportunistic sampling of carcasses, and targeted surveillance from 2020 to 2021 for TAHD surveillance in Del Norte and Humboldt counties, California, US (Fig. 1; Munk et al. 2020). Samples were collected as part of management activities by CDFW and were exempt from Institutional Animal Care and Use Committee review. Data were used under the auspices of Data Use Agreement 145050-001 between Washington State University and the CDFW. All samples were frozen at −20 C until processed. As part of surveillance efforts, the CDFW recorded the location of collection, elk sex and age, and TAHD diagnostic results. Elk age was estimated by trained CDFW staff based on tooth wear and eruption patterns (C.H. pers. comm.).

Trained CDFW personnel visually inspected elk feet and classified gross lesion severity as 0–IV following the grading scheme in Han et al. (2019). Samples from hooves were collected and processed for histologic examination at the California Animal Health and Food Safety Laboratory (CAHFS; Davis, California, USA) or the Washington Animal Disease Diagnostic Laboratory (Pullman, Washington, USA) as described (Han et al. 2019). Immunohistochemistry (IHC) for Treponema denticola (Han et al. 2019) was performed at CAHFS as a supplementary diagnostic test. Due to logistical constraints, only some cases (n=4) were examined histologically with both H&E and silver stain (modified Steiner or Warthin-Starry) to confirm TAHD. For other cases, elk were considered TAHD-positive if they had graded hoof lesions accompanied by suppurative inflammation on histologic examination with H&E stain (Han et al. 2019) or, in the absence of H&E examination, were positive by IHC. Cases failing to meet these criteria were classified as TAHD negative. We omitted cases of apparently abnormal hooves (e.g., thickening or pitting of the interdigital space) that did not meet the criteria for a graded TAHD lesion. To examine trace mineral associations with TAHD lesion severity, we categorized severity as minor or major according to the most advanced gross lesion observed in the set of four feet collected from each individual. For the subset of TAHD-positive elk, gross lesion grades of I or II were categorized as minor, while grades III or IV were categorized as major. These minor and major cases were compared against TAHD-negative cases (n=32) that had no lesions. Therefore, we omitted the subset of TAHD-negative elk with gross lesions (n=8) in analyses of the effects of minerals on TAHD lesion severity, to avoid pooling apparently unaffected elk with elk that had hoof lesions unrelated to TAHD. We did not include cases for which fewer than four feet per individual were assessed for gross lesions.

Liver samples were analyzed for a panel of six trace minerals (Cu, Fe, Mn, Mo, Se, and Zn) and four heavy metals regarded as toxicants (As, Cd, Pb, and Hg; Table 1) using inductively coupled plasma optical emission spectroscopy by the CAHFS Laboratory. We calculated summaries of mineral concentrations for sampled elk (n=81) and reported quantifiable limits of detection in parts per million (ppm) to supplement information from previously published trace mineral reference ranges (Puls 1994).

We merged metadata from TAHD surveillance with mineral concentrations obtained from 81 elk (41 TAHD positive, 40 TAHD negative). Elk were grouped into three age classes of yearling, young adult, or older adult if they were aged 1–2, 2–4.5, or ≥4.5 yr, respectively. Age estimates were not available for 21 elk (26%). We examined differences in mineral concentrations between counties, sexes, and age classes by calculating means and 95% confidence intervals (CIs). Furthermore, we examined pairwise Spearman rank-order correlations between minerals and metadata variables; we omitted one of the pair of predictors in statistical models if there was evidence of high correlation (Spearman P>|0.5|; Dormann et al. 2013).

We focused on minerals that are known to directly influence hoof health and immune function (i.e., Cu, Se, and Zn) to address multiple a priori hypotheses. Hypotheses were broadly represented and categorized in two separate candidate model sets: One examined associations between mineral concentrations and TAHD occurrence, and the other examined associations between mineral concentrations and TAHD lesion severity (see Supplementary Material Table S1). We included county, sex, or age class as covariates in some models because we suspected demographic and geographic sources of variation could influence associations between mineral concentrations and TAHD; additive combinations of all of these covariates were also considered, given that pairwise correlations among them were low (range of Spearman P=0.0–0.1). Although Mn may also be important in hoof-related diseases (Anklam et al. 2022), we omitted it from consideration in these models because of its indirect role in hoof health (Tomlinson et al. 2004) and high correlation with Cu (Spearman P=0.5).

We first built a set of 34 logistic regression models to ask whether the binomial response variable of TAHD status (TAHD negative or TAHD positive) was associated with mineral concentrations with and without the influence of the covariates. In practice, we evaluated two different model sets to address this question, because some cases had ambiguous or no age information. We evaluated the full model set (i.e., 34 models; see Supplementary Material Table S1) with a subset of the data with known age information (n=60), while we evaluated the model set that did not include age effects (18 models) with the full dataset (n=81). For most cases, we considered the influence of individual minerals on disease status with additive model terms. For a subset of models, we included an interaction term between Cu and Zn, given possible physiologic competition between them (O’Dell 1989; Bremner and Beattie 1995). The model set included models with all possible additive combinations of the three covariates paired with minerals (Supplementary Material Table S1) plus an intercept-only model with no effects to compare with the rest of the set with effects. We constructed univariate and multivariate logistic regression models using the stats package in R (version 4.3.2; R Core Team 2023).

Next, because mineral availability can change during the course of infection (Monteith and Skaar 2021) and TAHD lesion severity appears to be related to chronicity (Robinson et al. 2023), we considered whether mineral concentrations for Cu, Se, and Zn were associated with lesion severity (minor or major) in TAHD-positive elk (n=41). We used ordinal logistic regression models to address this question, because lesion severity has been shown to advance with disease progression (Robinson et al. 2023), enabling the inclusion of an “ordered” response variable (Harrell 2015). Ordinal regression models had the same combinations of sex and age covariates as the logistic regression models above; however, we could not include county as a covariate because the dataset did not contain any elk with major lesions in Humboldt County. As stated earlier, we included an intercept-only model without the effects of covariates. This resulted in our evaluation of a full model set with 17 models, including the effect of age (Supplementary Material Table S1) using the subset of the data with age and TAHD lesion severity information (n=55); nine models without age effects were evaluated with the full dataset containing cases with lesion severity information (n=73). We built univariate and multivariate ordinal regression models with a logistic link function using the ordinal package in R (Christensen 2023) after verifying assumptions of proportional odds (Brant 1990).

We used an information-theoretic framework for model selection and inference (Burnham and Anderson 2002). We used Akaike information criterion with a correction for small sample sizes (AICc) to rank models within each model set; the model with the lowest AICc value was considered the best model in the set given the data. For each model, using the R package AICcmodavg (Mazerolle 2023), we also calculated the differences in AICc with respect to the top-supported model (ΔAICc) and Akaike weights (w), which is the probability that a given model is the best given the data and the model set (Burnham and Anderson 2002; Anderson 2008). We examined model likelihoods with respect to AICc values, the number of parameters, and model coefficients to determine the possible presence of uninformative (pretending) variables (Arnold 2010). Finally, we demonstrated the relative strength of evidence for our hypotheses given the full dataset by calculating evidence ratios from model Akaike weights (Anderson 2008; Burnham et al. 2011).

We visualized the effects of minerals on response variables with marginal effects plots (Lüdecke 2018). We used model averaging to obtain parameter estimates from model sets that were developed a priori for basing inferences (Anderson 2008) about the effects of minerals from all additive models, including the effect of interest. We report the model-averaged estimates of the model coefficients (β_i) and the 95% CIs for Cu, Se, and Zn for both model sets.

We determined mineral concentrations for 81 liver samples (Table 1; Winter et al. 2024). Most samples originated from hunter harvest (n=65; 80%), followed by CDFW-targeted surveillance for TAHD (n=11; 14%), and opportunistic collections (n=5; 6%). More livers were collected from female elk (n=48) than from males (n=33), but ratios of TAHD-positive to TAHD-negative individuals between sexes were similar (1.09 females; 0.94 males). Levels of toxicant metals (i.e., As, Cd, Pb, and Hg) were below detectable limits in all sampled elk, so no further analysis was conducted. Mean values for most minerals were within reference ranges available from Puls (1994; Fig. 2 and Table 1); however, mean Cu and Se levels (Cu x¯_ppm=12.89; Se x¯_ppm=0.17) were lower than reference levels (Cu range_ppm=20–120; Se range_ppm=0.25–1.4; Puls 1994). When comparing mineral concentrations across covariates, we did not find clear differences in Cu or Zn by county, sex, or age class (see Supplementary Material Fig. S1). Elk from Humboldt County (n=14) had higher Se than elk from Del Norte County (n=68; Supplementary Material Fig. S1A), but Se did not differ by sex or age. We also found moderate correlations between Cu and Zn levels (Spearman P=−0.4) and between Se and Zn levels (Spearman P=0.4; Supplementary Material Table S2).

We did not identify a single, clear model that best explained variation in mineral concentrations with TAHD status. The model with the effect of Zn alone was top ranked and received the most support (w_i=0.27), while models containing Zn with county or sex covariates had competitive levels of support (ΔAICc<2; Table 2). Taken together, these three models accounted for over half of the cumulative Akaike weight in the model set (w_cumulative=0.54). The Zn-only model and Zn with county model had similar model probabilities (w_i=0.27 and w_i=0.17, respectively), while the Zn-only model was 3.96 times more likely than a model of no effect (w_i=0.07). Models containing Cu with or without an interaction with Zn had less support than the model of no effect using the full dataset (Table 2), and only minimal support in the smaller dataset with known ages (Supplementary Material Table S3). Models with Se had little support, regardless of the dataset. Further, sex most likely was an uninformative covariate, because there was a ΔAICc∼2 with the addition of one parameter for this covariate, negligible changes to model likelihoods, and parameter CIs that overlapped zero (Arnold 2010). This finding suggests that sex did not account for meaningful variation in mineral concentrations.

We observed limited associations between TAHD status and Cu or Se concentrations and a positive association between TAHD status and Zn concentrations in univariate models (Supplementary Material Fig. S2). Based on model averaging, Zn concentrations were positively associated with TAHD occurrence, but the effect size was small (β=0.03; CI 95%, 0.002–0.058; Fig. 3). Model-averaged estimates for Cu and Se did not show a clear directional effect, as 95% CIs overlapped zero (Fig. 3). Similarly, coefficients for the interaction of Cu with Zn, and additional covariates (county, sex, and age class) had CIs overlapping zero in other models.

The best-supported models contained the effects of Zn alone, while the interaction of Cu and Zn had competitive support (ΔAICc=0.32; Table 3), and probabilities for the two models were similar under the full dataset (w_i=0.41 and w_i=0.31, respectively). Further, the addition of sex or age class or both with the Zn-only and Cu-Zn interaction models had competitive levels of support (ΔAICc∼2–5; Table 3) and accounted for almost the entire Akaike weight within the model sets (w_sum >0.99). Although sex most likely was an uninformative covariate (Arnold 2010; Table 3; Supplementary Material Table S4), age class slightly improved model likelihood and resulted in ΔAICc∼2 from two parameters, rather than one (Supplementary Material Table S4). Thus, age class, but not sex, accounted for meaningful variation in mineral concentrations. All models containing Cu (without the Zn interaction), Se, or the model of no effect did not have support in either dataset (ΔAICc>10).

We observed a trend of higher Zn concentrations and generally lower Cu in elk with advancing TAHD lesion severity (Fig. 4), but no trends in Se (Supplementary Material Fig. S3). Congruently, model-averaged parameter coefficients (excluding interactive models) reflected a positive association with Zn, but the effect size was small (β=0.05, CI 95%: 0.03–0.08; Fig. 3). Finally, model-averaged coefficients for Cu and Se did not suggest a clear association with TAHD lesion severity (Cu β=−0.032; CI 95%, −0.07–0.002; Se β=−0.61; CI 95%, −4.67–3.45; Fig. 3).

Our data from free-ranging Roosevelt elk in California supplements baseline information on the range of trace mineral concentrations in elk. Reference ranges (intervals) are useful for summarizing variation and evaluating trace mineral concentrations in animals with respect to health status. For domestic animals, trace mineral levels of individuals or herds are compared against values considered normal in apparently healthy individuals (Spears and Brandao 2022). Information on mineral levels in elk and other wildlife, however, is limited (Puls 1994). This limitation is partially because of logistical difficulties in collecting and storing samples of biologically representative organs (e.g., liver, kidneys) at a large scale for mineral analysis. Consequently, reference ranges may be extrapolated from domestic animals (e.g., Mo values for cattle used as a proxy for elk; Johnson et al. 2007; Han et al. 2019), although these values may not be representative due to species variability and conditions of domestication, such as nutritional supplementation. In addition, reference ranges reflect mineral levels in apparently healthy livestock, but free-ranging wildlife have unknown health histories, and the term healthy is difficult to define (Stephen 2022). Reference ranges derived from conspecific or closely related wildlife species in different locations may more accurately reflect species-specific values but are not without limitations. For example, many elk we sampled had Se and Cu levels lower than previously reported reference ranges for Cervus spp. (Puls 1994). It is unclear whether the observed low Se and Cu is inconsequential and reflects local adaptations of animals or whether these low levels represent mineral deficiencies. Copper values in elk we sampled were similar to those from another C. canadensis ecotype, Tule elk, experiencing brittle antlers in central California (Johnson et al. 2007). Likewise, low Se has been reported in nearby populations of black-tailed deer (Odocoileus hemionus columbianus), possibly contributing to declines in fawn survival (Flueck 1994). However, there have been no reports of these issues in elk from the study area (Nigon 2020). Continuing to explore trace mineral values in wildlife populations across the geographic ranges is necessary to develop informed inferences about the clinical significance of variation in mineral concentrations (Flueck et al. 2012; Poppenga et al. 2012).

We found limited to no support for our hypothesized associations of Se and Cu concentrations with TAHD occurrence and severity. These results were surprising, considering the physiologic basis linking Se and Cu with the occurrence of infectious hoof diseases in livestock (Hall et al. 2013; Langova et al. 2020); however, similar findings have been reported in investigations of TAHD-affected and unaffected elk in Washington (Han and Mansfield 2014; Han et al. 2019). Our results may suggest that Se and Cu contribute less to hoof disease susceptibility or development in wild elk than in livestock. Alternatively, Se and Cu levels in both groups of elk (i.e., TAHD affected and unaffected) in the sampled populations may be below optimal levels. When considered together with the current study, prior associations of low Se and Cu in the hair of elk with TAHD in the northwestern US (Winter et al. 2022) might imply that differences in mineral availability influence risk for TAHD at broader spatial scales than can be easily captured at the local level studied here. That is, local heterogeneity in TAHD occurrence may be driven by risk factors other than mineral concentrations. This hypothesis may partially explain why TAHD appears concentrated in some elk populations when Se in forage is low, such as southwestern Washington, and sporadic in regions with reportedly higher Se in forage, such as eastern Washington (Welch et al. 1991; Wild et al. 2022; Winter et al. 2022, 2023). However, a similar observation cannot be made for Cu because less information is available about Cu in forage from these regions (Welch et al. 1991). Adjusting for multiscale environmental risk factors may be needed to better understand the relative differences in TAHD risk from low Cu and Se levels.

The correlation that we found of higher Zn levels with TAHD may be indirect evidence of a nutritional immune response by the host to combat pathogens. We hypothesized a priori that elk diagnosed with TAHD and those with major TAHD lesions would have lower Zn concentrations relative to unaffected elk, given the contribution of Zn to keratinization, maintenance of hoof health, and general immune function (Tomlinson et al. 2004; Gammoh and Rink 2017; Langova et al. 2020). Contrary to our expectations, we found higher Zn levels were associated with TAHD occurrence and were highest in elk with major lesions. The reasons for elevated Zn in TAHD-affected elk are unclear; in other species, redistribution of minerals can occur through systemic and local changes to starve pathogens of essential minerals (Mir et al. 2020; Monteith and Skaar 2021). Such nutritional immune responses can result in higher liver Zn levels in animals with signs of infection and inflammation (Rahman and Karim 2018; Mir et al. 2020). These changes have been linked with the activity of acute phase proteins that influence Cu and Zn homeostasis and control oxidative stress from pathogens (Beisel 1976; Rahman and Karim 2018). Alternatively, considering that mineral levels in animals depend on mineral availability in vegetation and soil, elevated Zn in TAHD-positive elk may relate to forage or soil with high Zn content or exposure from unknown agricultural practices (e.g., Zn-containing fertilizers) near foci of infection. Relatively high soil Zn levels have been reported near the study area (White et al. 1997), but this would not fully explain why we observed differences among sympatric elk. Therefore, further characterizing elevated Zn levels in TAHD positive elk may benefit from examining fine-scale land use practices in areas where the disease occurs.

We hypothesized that the covariates of county, sex, or age might influence the effects of mineral concentrations and therefore, when accounted for, improve inferences on mineral associations with TAHD; however, including these variables seldom resulted in improved explanations for variation in TAHD occurrence or lesion severity. We did not find strong effects of county on either TAHD occurrence or lesion severity. Limitations of our data restricted the comparison of counties in lesion severity models. Considering that CDFW has detected elk from Humboldt County with major TAHD lesions (E.L.L. pers. comm.), it may be more likely that the absence of major lesions in animals from this county relates to the small sample size rather than higher Se concentrations also found in these elk, but this cannot be discounted. Similarly, sex was not an informative covariate in mediating mineral concentrations with TAHD, but more data could clarify any potential relationship. Finally, age is known to alter mineral requirements in livestock (Spears and Brandao 2022) and can be a risk factor for chronic infectious diseases, yet we did not identify clear modifying effects of age class on minerals with respect to TAHD. This may relate to missing information on elk age, uneven representation of age classes in our dataset, or inconsistent differences in mineral content with age (Cygan-Szczegielniak and Stasiak 2022).

Further investigation is needed to understand variation in mineral levels and the impact on wildlife health more fully. The contribution of Cu and Se to TAHD occurrence and severity remains unclear. The Cu and Se values that we found in both unaffected and diseased elk may suggest relatively lower values are normal for elk in this region. It could be that elk sampled in this study have suboptimal levels of Cu and Se, posing a potential risk factor for TAHD susceptibility. Regardless, artificial mineral supplementation should be approached cautiously due to potential unintended consequences. For example, mineral licks may increase host contact rates, potentiating pathogen transmission in TAHD or other diseases (Lavelle et al. 2014). In addition, if nutritional immune responses are occurring, supplying minerals could counteract any host response to sequester Zn from pathogens (Monteith and Skaar 2021; Maywald and Rink 2022). Given geographic and host variation in mineral levels, as well as challenges in collecting samples and monitoring health of free-ranging wildlife, controlled trials with trace mineral supplementation are probably needed to directly understand the role of these minerals on TAHD occurrence and severity and to clarify any benefit of supplementation.

We thank Kyle Taylor for examining hooves, as well as Charlie Park and biologists with the California Department of Fish and Wildlife, including Megan Moriarty, John Ly, and Nicholas Shirkey, for sample collection and processing. We are grateful to Tricia Talcott, Chelsea Sykes, and Glen Sargeant for thoughtful discussion and advice about earlier versions of the manuscript. Funding was provided by the state of Washington through accounts dedicated to research on elk hoof disease.

Supplementary material for this article is online at http://dx.doi.org/10.7589/JWD-D-24-00135.