INTRODUCTION

Chronic wasting disease (CWD) is an insidious and invariably fatal neurodegenerative disease of cervids (Williams and Young 1980) that is caused by an infectious misfolded prion protein (Prusiner 1982). The prions of CWD maintain infectivity for extremely long periods of time and have remarkably low infectious doses (Denkers et al. 2020; Kuznetsova et al. 2020). Prions are putatively transmissible via oronasal exposure to environmental vectors such as soil (Johnson et al. 2006, 2007), plants (Pritzkow et al. 2015; Carlson et al. 2023), deer scrapes (Huang et al. 2024), and arthropods (Inzalaco et al. 2023; Soto et al. 2024). As of 1 August 2024, CWD has been detected in free-ranging cervids in 35 US states, four Canadian provinces, Finland (Sun et al. 2023), Norway (Pirisinu et al. 2018), and Sweden (SVA 2024). It has also been detected in captive cervid facilities in 19 states, three Canadian provinces (USGS 2024) and South Korea (Kim et al. 2005; Lee et al. 2013). The spread of CWD across North America is expected to have significant social and economic implications. Furthermore, CWD is a One Health issue, as it has implications for animal health and the environment in addition to its potential impact on human health (Gilch 2022). This disease presents many challenges for wildlife management agencies (Opsahl 2003; Chiavacci 2022). The worst-case scenario might involve the death of many cervids, resulting in substantial environmental contamination with infectious prions. Social implications include the loss of an important food source and cultural focus for many people, particularly indigenous and rural communities. Although there are no known instances of human CWD zoonosis (Tranulis and Tryland 2023) and the human species barrier appears to be robust (Groveman et al. 2024), the precedent for prion zoonoses exists (Will et al. 1996).

Key tools in the management of any disease are easy-to-use antemortem tests with high diagnostic sensitivity, specificity, and predictive value. Antemortem testing for CWD is a challenging endeavor because of the exceedingly slow progression of prion diseases generally. Body-wide dissemination of infectious prions accelerates in later stages of the disease but is slow initially (Henderson et al. 2020). Furthermore, the relative sensitivities of antibody-based methods (ELISA, western blot) are far lower than would be necessary for antemortem testing of readily accessible tissues (McNulty et al. 2019; Schwabenlander et al. 2022). Antemortem testing for CWD has been met with some success using tonsillar biopsies or rectoanal mucosa–associated lymphoid tissue (RAMALT) by either ELISA or immunohistochemistry (IHC; Wild et al. 2002; Keane et al. 2009; Spraker et al. 2009). However, sampling RAMALT or tonsils antemortem can be highly invasive for the cervid and can be technically challenging for the collector, requiring training as well as veterinary care.

The cervid ear represents an attractive biopsy source, as it is relatively easy to collect. Two publications have evaluated the diagnostic sensitivity and specificity of ear punches using real-time quaking-induced conversion (RT-QuIC; Ferreira et al. 2021; Burgener et al. 2022). RT-QuIC is an extremely sensitive seeded amplification assay that uses fluorescence intensity as a quantitative proxy for the presence of prions in samples of unknown status (Ferreira et al. 2021). Ferreira et al. (2021) reported that mule deer (Odocoileus hemionus) and white-tailed deer (Odocoileus virginianus) ear punch samples tested with RT-QuIC after iron oxide magnetic bead extraction showed 81% diagnostic sensitivity and 91% diagnostic specificity relative to retropharyngeal lymph node (RPLN) ELISA. Burgener et al. (2022) reported 95% diagnostic sensitivity and 100% diagnostic specificity using white-tailed deer ear punches and belly skin respectively, using traditional RT-QuIC. Taken together, these studies suggest that RT-QuIC testing of ear punches holds promise as an antemortem CWD diagnostic tool for mule and white-tailed deer. To date, no diagnostic evaluation of elk (Cervus canadensis) ear tissue has been conducted. Importantly, differences in physiology and disease progression between Odocoileus spp. and elk may lead to differences in the relative abundance and subsequent detectability of prions in ear tissue. Earlier work (Race et al. 2007) suggests that prion dissemination in elk is lower than in the Odocoileus spp. To address this knowledge gap, we evaluated the diagnostic concordance of RT-QuIC using elk ear punches relative to ELISA-tested obex and/or RPLN from matched animals.

MATERIALS AND METHODS

Sample preparation

We collected 6-mm ear punches using Integra Miltex disposable biopsy punches (Fisher Scientific, Hampton, New Hampshire, USA) from five locations on each intact ear. The samples generally weighed 100±20 mg. We designated those locations ventral, medial, dorsal, lateral, and central (Fig. 1). These designations correspond roughly to locations 7, 2, 1, 6, and 5 respectively in Burgener et al. (2022). They correspond to a site intermediate of 2 and 3, 4, 6, 5, and 1 respectively in Ferreira et al (2021). All samples were digested using the protocol described in Burgener et al. Ear punches (∼100 mg) were homogenized in 900 µL of a solution of 0.25% w/v collagenase A (Sigma Aldrich, St. Louis, Missouri, USA), 2 mM CaCl₂ in 1× phosphate-buffered saline (PBS) using a Beadblaster 24 microtube homogenizer (Neta Scientific, Marlton, New Jersey, USA) with 1.5-mm zirconia beads (Millipore Sigma, Burlington, Massachusetts, USA) for 1 min at 4 m/s. This yielded ∼10% (or 10^–1) homogenates that were heated and mixed using a ThermoMixer (Eppendorf, Hamburg, Germany) for 24–27 h at 45 C at 700 RPM. Supernatants were collected and frozen at –80 C after the homogenates were centrifuged for 3 min at 3,000 × G. Gross anatomical dissection of the remaining pinnae was performed in a biosafety cabinet using a scalpel and hemostat forceps (Fisher Scientific).

Figure1

(A) Rostral (concave) view of elk (Cervus canadensis) ear, left side, and (B) caudal (convex) view of elk ear, left side. Five 6-mm ear punches were made in each ear. Ear punches were designated dorsal (D), lateral (L), ventral (V), medial (M), and central (C).

Statistical assessment of binomial distribution using receiver operating characteristic analysis

For the second statistical diagnostic assessment, we conducted a semiparametric receiver operating characteristic (ROC) curve analysis (Pepe 2000) that established an AFR threshold for RT-QuIC positivity. To conduct the ROC analysis, we first normalized AFR values by plate, so the data were in the range (0, 1). We used the following formula to normalize the AFR values by plate:

where Xscaled is the normalized AFR value, X is the AFR value for each technical replicate of each elk ear biopsy on a given plate and xmin and xmax are the minimum and maximum AFR values respectively of all technical replicates of elk ear biopsies on a given plate.

To obtain fitted values for the ROC analysis, we fit the median of the normalized AFR values into a generalized linear model with a binomial distribution and logit link function, where l represents the log odds of the ELISA test having a positive test result (coded as 0, negative, or 1, positive), β₀ is the intercept of the model, and β₁ is the coefficient associated with the predictor variable, x₁, which is the median of the normalized AFR values. We determined the optimal threshold for AFR associated with an optimal combination of sensitivity and specificity from area under the curve (AUC). We used the pROC package (Robin et al. 2011) in R statistical software (R Core Team 2024) to plot ROC and determine AUC.

RESULTS

Specific details about ELISA (TeSeE Kit, Bio-Rad, Hercules, California, USA) testing and diagnosis are indicated in   Supplementary Material Table S1 (JWD-D-24-00071R2_Lichtenberg_SUPPL_Final.docx). The ELISA data from the three diagnostic laboratories indicated that 18/38 (47%) of the elk tested were CWD positive. Of these, 15/18 (83%) exhibited strong (RFU>1,000) RT-QuIC seeding activity (  Supplementary Material Table S1 (JWD-D-24-00071R2_Lichtenberg_SUPPL_Final.docx)). Of the 20 ELISA-diagnosed CWD-negative elk, one biological replicate (5%) exhibited strong RT-QuIC seeding activity. All 10 of the animals designated Wind Cave National Park Surveillance (  Supplementary Material Table S1 (JWD-D-24-00071R2_Lichtenberg_SUPPL_Final.docx)) were culled and exhibited suspect (RFU<1,000) seeding activity. Two of the four animals designated “research” had been found dead: One of these was a coyote predation victim and the other was of unknown status. Four of the five (75%) animals designated “sick/surveillance” were ELISA-diagnosed positive and RT-QuIC positive; the fifth was ELISA-diagnosed negative and RT-QuIC negative. No further details were available to us for sick/surveillance animals. The remaining elk were collected via hunter harvest.

A Kruskal-Wallis H test of ELISA-diagnosed CWD-positive (only) ear punch AFR data indicated a statistically significant effect of spatial location (H=14.95, P=0.0048). Post hoc Dunn tests indicated a statistically significant reduction in the ventral punch AFR relative to medial (mean rank difference=–96.49, P=0.0067), dorsal (mean rank difference= –81.75, P=0.0396), and lateral (mean rank difference=81.71, P=0.0398) ear punches (Fig. 2). Based on these data, and because few people would simultaneously collect five ear punches per elk in practice, we performed a nonparametric analysis and semiparametric analysis on the medial ear punches only Nonparametric statistical analysis of all of our data showed a statistically significant reduction (P=0.0067, 0.0396, 0.0398 respectively) in the AFR of samples collected from the ventral area of the elk ear, relative to medial, dorsal and lateral areas. A statistically nonsignificant reduction in AFR (P=0.9288) was also observed in the central punch.

Figure2

Amyloid formation rate (entire real-time quaking-induced conversion run) of elk ear punches sorted by spatial location and ELISA status. Five spatially distinct samples were collected from each of 38 biological replicates. Eight technical replicates were run for each sample. ELISA-diagnosed chronic wasting disease (CWD)–positive (n=18 biological replicates) and CWD-negative ear punches (n=20 biological replicates) are indicated by + and –, respectively. ELISA-diagnosed CWD-positive (n=12) and CWD-negative (n=14) elk retropharyngeal lymph nodes (RPLNs) are also plotted for comparison. The median and 95% confidence interval are indicated by horizontal bars. Deer CWD-positive and CWD-negative plate controls are subsampled from 1 biological replicate each. A Kruskal-Wallis H test of ELISA-diagnosed CWD-positive ear punches was statistically significant (H=14.95, P<0.0048). Post hoc Dunn tests indicated a statistically significant reduction in the ventral punch AFR relative to medial (**mean rank difference =–96.49, P=0.0067), dorsal (*mean rank difference=–81.75, P=0.0396) and lateral (*mean rank difference=81.71, P=0.0398) ear punches. All other post hoc comparisons of ELISA-diagnosed, CWD-positive samples were statistically insignificant.

  Supplementary Material Table 2 (JWD-D-24-00071R2_Lichtenberg_SUPPL_Final.docx) shows the sensitivity, specificity, PPV, and NPV of RT-QuIC relative to ELISA, using the 50% of 40 technical replicates threshold; i.e., across all technical replicates and spatial locations per elk ear). All other spatial comparisons were statistically insignificant. We found that 15/38 elk were CWD positive based on seeding in 50% of the eight replicates of the medial ear punch location. A total of 12/38 elk were classified as CWD positive based on the median AFR threshold for detection established by the ROC analysis. Using the seeding in 50% of the eight technical replicates approach, RT-QuIC had a sensitivity of 67% (95% CI, 41.0–86.7%), a specificity of 85% (95% CI, 62.1–96.8%), a PPV of 80% (95% CI, 51.9–95.7%), and an NPV of 74% (95% CI, 51.6–89.8%; Table 1). The threshold of the median normalized AFR was 0.2198, and AUC was 78.3%. Using ROC analysis, RT-QuIC had a sensitivity of 67% (95% CI, 41.0–86.7%), a specificity of 100% (95% CI, 83.2–100%), a PPV of 100% (95% CI, 73.5–100.0%), and an NPV of 77% (95% CI, 56.3–91.0%; Table 2).

Table 1

Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of chronic wasting disease (CWD) diagnosis using real-time quaking-induced conversion (RT-QuIC) on elk (Cervus canadensis) ear biopsies, with 50% of 8 medial biopsy technical replicates (any seeding activity in 4/8 technical replicates) established as the threshold for RT-QuIC diagnosis of CWD positivity. Sensitivity, specificity, and predictive value were subsequently calculated relative to the ELISA data as a reference standard.

Table 2

Sensitivity, specificity, positive predictive value, and negative predictive value of chronic wasting disease (CWD) diagnosis using real-time quaking-induced conversion (RT-QuIC) on elk (Cervus canadensis) ear biopsies, with a semiparametric receiver operating characteristic (ROC) analysis used to establish the threshold for RT-QuIC diagnosis of CWD positivity. Sensitivity, specificity, and predictive value were subsequently calculated relative to the ELISA data as a reference standard.

When comparing detection of CWD prions between ELISA and the RT-QuIC method based on seeding in 50% of the eight technical replicates of the medial ear punch location, McNemar test was not significant (χ²=1.00; P=0.317), and KAPPA was 0.52 (0.25–0.79), indicating moderate agreement between the two methods. When comparing detection of CWD prions between ELISA and ROC, McNemar test was significant (χ²= 6.00; P=0.014), and PABAK was 0.68 (0.37–0.88), indicating substantial agreement between the two methods.

On the caudal (i.e., convex) surface of the pinna there were two main cartilage ridges and three visible neurovascular bundles. Figure 3A provides the clearest visualization of the ear punch locations on the caudal surface of the pinna. At the base of the pinna, there were three to four locations where neurovascular bundles entered the auricular cartilage (Fig. 3B, C).

Figure3

Ear biopsy sampling locations on the auricular cartilage of a right elk (Cervus canadensis) pinna. (A) An image of the caudal surface (convex) of the auricular cartilage after removal of hair and skin showing the biopsy locations dorsal (D), lateral (L), ventral (V), medial (M), and central (C). Note the two prominent ridges and three depressions where neurovascular bundles were found. (B) Medial border of right ear and (C) ventromedial view at the base of auricular cartilage—both show neurovascular bundles (indicated by the hashtag # and asterisk *) entering the cartilage. The scutiform cartilage is also labeled.

DISCUSSION

Based on our data, we recommend the application of ROC curve analysis to AFR data in order to obtain the best achievable RT-QuIC diagnostic accuracy. This recommendation is based on our evaluation of the sensitivity, specificity, and predictive value of RT-QuIC testing of elk ear punches in the context of CWD diagnosis. Sensitivity and specificity are calculated to evaluate diagnostic performance, whereas predictive value is calculated to predict the likelihood that, given a certain test result, the subject does or does not have the disease (Monaghan et al. 2021). In contrast to sensitivity and specificity, which are a function of the test itself, predictive value varies as a function of disease prevalence, such that PPV increases and NPV decreases when disease prevalence increases.

In our study, ELISA of the obex and/or RPLN results were represented as “true disease” status. There is a possibility that an animal may be classified as negative because of a nondetect using ELISA with RPLN while still actually being CWD positive. This could arise because of the animal being in the earliest stages of disease progression, reduced prion trophism in elk relative to deer (Race et al. 2007), the somewhat low sensitivity of ELISA (at least compared to seeded amplification assays), or the lack of obex for testing. The temporal gap between detectability in RPLN vs. obex is reasonably small and is only present in about 12% of animals (Spraker et al. 2004). Keeping the aforementioned caveat in mind, we observed no false-positive data using the ROC-established threshold for CWD positivity: A positive test using ROC curve analysis to determine the binomial distribution (CWD positivity) correctly identified subjects with CWD. Our ROC PPV data indicated that subjects that returned a positive RT-QuIC ear punch were highly likely to have a positive obex and/or RPLN by ELISA. These data are consistent with prior publications that observed 91% and 100% specificity in RT-QuIC testing of ear punches from white-tailed and mule deer (Ferreira et al. 2021) and white-tailed deer (Burgener et al. 2022), respectively. Conversely, we did observe false-positive data in the nonstatistical 50% of technical replicates approach, indicating that this analysis is suboptimal for management decision-making: 3/ 20 ELISA-diagnosed negative animals were designated CWD positive by RT-QuIC using this approach.

The RT-QuIC testing of elk ear punch biopsies may be susceptible to false-negative data. We found that the statistical approach was a slight improvement over the nonstatistical approach. A 74–77% NPV indicates a 23–26% chance that a subject diagnosed CWD negative by ear punch biopsy using RT-QuIC may actually be CWD positive by ELISA of the obex and/or RPLN. Sensitivity using ear punch biopsy RT-QuIC (67%) was comparable to testing of RAMALT by RT-QuIC and IHC in elk (Haley et al. 2016). Under conditions of relatively low sensitivity, it may be worthwhile to employ various tactics such as more extensive sampling of the ear, enhanced (iron oxide magnetic extraction) RT-QuIC assays (Ferreira et al. 2021), and US Department of Agriculture (USDA)–approved tests such as ELISA or IHC using RPLN or obex. Along these lines, the qualitative descriptions “strong” and “suspect” were employed in the current study to evaluate data that incorrectly (i.e., fail to) meet the threshold for significance using statistical or nonstatistical approaches. For example, elk 38 was ELISA-diagnosed CWD positive but RT-QuIC negative by statistical and nonstatistical analysis. Looking at the “strong” (>1,000 RFU) seeding activity, an analyst might designate it RT-QuIC positive, assuming that it is early in the disease progression. In contrast, elk 42 evidenced “strong” seeding activity but was diagnosed negative by ELISA and RT-QuIC. Interpretation of RFU values generated by RT-QuIC in the absence of other metrics (AFR, maximum slope, maximum point ratio, etc.) is basically impossible. To date there has been no definitive analysis of what exactly is represented by maximum RFU values or if these values are representative of any biological phenomenon at all. In our laboratory, we generally see what we call “strong” RFU values associated with true-positive samples, but this is by no means universal across all samples, or even across replicates at times. This inconsistency renders interpretation of RFU values alone mostly moot. Despite having sufficient statistical power, our data highlight diagnostic uncertainty that remains in the face of the most sensitive assays.

Given the profound implications of CWD management, stakeholder confidence in CWD diagnostics is important. RT-QuIC has yet to be formally approved by the USDA as a test for diagnosing CWD, and its findings should be interpreted cautiously. Limitations to this assay include lack of consensus on the analysis and interpretation of its data (Rowden et al. 2023). Our study advances this conversation by documenting improved RT-QuIC diagnostic accuracy when ROC curve analysis is used to determine the binomial distribution of unknown samples. The value proposition of RT-QuIC is that it increases antemortem diagnostic access, providing stakeholders with more time to be careful and precise with decision-making. We are unaware of any published evidence of successful ELISA or IHC diagnosis of CWD using ear punches or skin. Current USDA approved ELISA and IHC tests require obex or RPLN for definitive CWD diagnosis (USDA, Animal and Plant Health Inspection Service, 2024). Although RAMALT and tonsil have been successfully tested antemortem (Wild et al. 2002; Keane et al. 2009; Spraker et al. 2009), they are technically challenging to collect and are highly invasive. Given the high specificity and PPV documented by this study and the nascent literature (Ferreira et al. 2021; Burgener et al. 2022), RT-QuIC analysis of punches from any ear location is an easy and useful first step for evaluating the CWD status in cervids.

A secondary aim of this study was to evaluate the extent to which anatomical features influenced results. Spatial differences in RT-QuIC seeding activity are likely explained by prion neurotropism and differences in nerve innervation patterns. There are few anatomical descriptions of the cervid ear in the literature (Rashid et al. 1987). The most relevant literature used was centered on bovine (Budras and Habel 2011) and canine ear anatomy (Evans and de Lahunta 2012). The pinna and any neurovascular bundles (i.e., veins, arteries, and nerves) entering or exiting the ear were the focus, as these structures follow each other as they branch to reach tissues. It is likely that the largest branches of nerves are found at the base of the pinna (most ventrally), and that branches become smaller as the nerves reach the medial and lateral edges and apex of the pinna. Previous studies have shown that the auricular cartilage has several neurovascular branches running within the cartilage and its ridges (Rashid et al. 1987). These are probably where the nerves of the rostral (concave) surface enter to provide sensory innervation of the pinna and the external ear canal. There are four main nerves related to sensory innervation of the pinna. These nerves include branches of three cranial nerves: the trigeminal (auriculotemporal with rostral auricular and external acoustic branches), facial (several internal auricular branches), and vagus (auricular branch), as well as the second cervical nerve (great auricular and greater or major occipital nerves; Budras and Habel 2011; Evans and de Lahunta 2012). The predicted rostral and caudal distribution of these nerves is shown in Figure 4A, B. An in-depth study of comparative ear anatomy across cervids should be a next step to determine if neurovascular patterns are consistent, which could increase sampling efficacy. Additionally, it is likely that auricular branches of the vagus nerve are also intermingled with the facial nerve branches that are found in the external ear canal (Budras and Habel 2011; Evans and de Lahunta 2012). This could be significant in that the vagus nerve, and potentially other cranial nerves (e.g., facial) are associated with the autonomic nervous system pathways, which play a large role in transport and neuroinvasion of prions in the body (Seelig et al. 2011).

Figure4

Diagram showing the predicted distribution of sensory nerves of a right elk (Cervus canadensis) ear. (A) Rostral (concave) view with four nerves from lateral (L) to medial (M): great auricular nerve (n.), internal auricular n., greater occipital n., auriculotemporal n. (B) Caudal (convex) view with three nerves from M to L: auriculotemporal n., greater occipital n., great auricular n.

Although Ferreira et al. (2021) reported higher seeding activity from punches around the central nerve of the mule/white-tailed deer (areas 1, 6, and 7), Burgener et al. (2022) did not observe an effect of ear punch location on seeding activity in white-tailed deer. Our finding of reduction in the AFR of samples collected from the ventral area of the elk ear is somewhat consistent with Ferreira et al. (2021), as our dorsal location is similar to location 6 in that publication. A statistically nonsignificant reduction in AFR was also observed in the central punch, which corresponds to position 1 or 7 in Ferreira et al. (2021). Collectively, these data have implications for elk tracking and herd management decisions. The opportunity cost of placing an ear tag where tissue is biopsied for CWD testing from the periphery of the elk ear may be reduced tracking efficacy, as tags may be more easily lost from the ear over the long term.