The Xpert MTB/RIF Ultra assay detects Mycobacterium tuberculosis complex DNA in white rhinoceros (Ceratotherium simum) and African elephants (Loxodonta africana)

The study describes the novel use of the Xpert MTB/RIF Ultra assay for detection of Mycobacterium tuberculosis complex (MTBC) DNA in samples from white rhinoceros (Ceratotherium simum) and African elephants (Loxodonta africana). Culture negative respiratory sample matrices were spiked to determine if the Ultra could detect MTBC DNA in rhinoceros and elephant samples. Rhinoceros bronchial alveolar lavage fluid (BALF) was found to have an inhibitory effect on the Ultra. In this study, the limit of detection (LOD) of M. tuberculosis H37Rv in all spiked animal samples were 2 CFU/ml compared to 15.6 CFU/ml for humans, while the LOD for M. bovis SB0121 was 30 CFU/ml compared to 143.4 CFU/ml for M. bovis BCG in humans. Screening was performed on stored tissue and respiratory samples from known MTBC-infected animals and MTBC DNA was detected in 92% of samples collected from six rhinoceros and two elephants. Conversely, 83% of culture-negative tissue and respiratory samples from uninfected animals tested negative on the Ultra. In conclusion, the Ultra assay appears to be a sensitive and rapid diagnostic test for the detection of MTBC DNA from tissue and respiratory samples collected from African elephants and rhinoceros. Furthermore, the Ultra assay could provide a new tool for the detection of MTBC in various sample types from other wildlife species.

www.nature.com/scientificreports/ Due to the paucity of diagnostic tests and disease surveillance programs for wildlife, MTBC infections can remain undetected for years, resulting in uncontrolled spread 8 . In livestock, M. bovis-exposure within a population is often indirectly detected through the antemortem measurement of the host's antigen-specific cell-mediated immune (CMI) responses, rather than directly detecting the presence of the pathogen to determine the presence of an active infection and shedding by the host. Using a CMI-based test is a valuable tool for surveillance to determine the presence of exposure before decisions on mitigation are undertaken. However, in vitro diagnostic assays for detection of host CMI responses to MTBC are either very limited, as in the case of African rhinoceros 9 , or do not yet exist for African elephants. Therefore, antemortem TB diagnosis in these species still relies on mycobacterial culture of respiratory tract samples such as bronchoalveolar lavage fluid (BALF) from rhinoceros or BALF and trunk washes (TW) from African elephants.
Mycobacterial culture for wildlife testing is regarded as an imperfect standard 10,11 , perhaps due to aggressive decontamination methods applied to polymicrobial field samples, and results take weeks to months to be delivered. Therefore, detection of mycobacterial DNA in animal samples has been used as an alternative or ancillary diagnostic method 12 . In humans, the Xpert MTB/RIF assay (Cepheid) was initially developed to improve TB and rifampicin resistance (RIF-R) detection using human sputum 13 . To improve performance in paucibacillary samples, a next generation Xpert semi-quantitative assay, the Xpert MTB/RIF Ultra test (Ultra), was developed by improving several features including the first ever combined detection of two genetic insertion elements IS6110 and IS1081 for MTBC detection. Notably, field isolates from the MTBC contain varying copies of IS6110 and IS1081 between specific strains that may have an influence on a test's limit of detection (LOD). Varying copy numbers between strains within isolates from the MTBC largely depends on the geographical location of the infected hosts 14,15 . However, the additional detection of IS1081 by the Ultra, has truly increased its diagnostic appeal for use on specimens collected from animals suspected of having MTBC infections 16 . Initial analytical studies in human sputum have shown the LOD of the Ultra to be 15.6 CFU/ml for M. tuberculosis H37Rv and 143.4 CFU/ ml for M. bovis BCG with a sensitivity of 89% and specificity 98% for M. tuberculosis H37Rv detection 16 .
In this study, we describe the first use of the Ultra for the detection of M. tuberculosis and M. bovis from experimentally spiked animal samples and stored clinical tissue and respiratory tract samples collected from white rhinoceros and African elephants with known MTBC-infection status.

Results
Detection range of the Ultra assay in animals. The Ultra assay results for serial dilutions of M. tuberculosis and M. bovis, spiked into different matrices, are shown in Fig. 1. Briefly, the presence of DNA and RIF susceptibility were detected for all M. tuberculosis concentrations (30, 15.6, 5 and 2 CFU/ml), spiked into sets of elephant and rhinoceros respiratory tract samples. The lowest M. tuberculosis concentration of 2 CFU/ml was detected in this study (Fig. 1a). In contrast, M. bovis spiked respiratory tract samples at similar concentrations were undetectable at concentrations of 15.6 CFU/ml and below (Fig. 1b). However, M. bovis spiked samples, at concentrations of 30 CFU/ml and above 1 × 10 2 to 1 × 10 6 CFU/ml, were all positively detected by the Ultra assay (Fig. 1b). Notably, no amplification was observed for any of the negative controls (respiratory specimens left unspiked). With a Ct value ≤ 37 considered as positive, Ct-value detection ranges of the Ultra were between 17 and 26 for both elephant BALF and TW and between 22 -30 for M. tuberculosis DNA detection in rhinoceros BALF (Fig. 1a). For M. bovis detection the calculated Ct-value detection ranges were between 16 and 27 for both elephant BALF and TW and 16-32 for rhinoceros BALF (Fig. 1b).
Analytical specificity. The specificity of the primers and probes for MTBC used was demonstrated by spiking respiratory tract samples with a high concentration (1 × 10 6 CFU/ml) of M. avium. No signals were generated from the two M. tuberculosis-specific detection probes targeting IS6110 and IS1081 genes, with a result "MTB not detected" for all replicates tested (Fig. 1).
Identification MTBC-infected animals from MTBC endemic areas by mycobacterial culture. Six of the 36 white rhinoceros tested were identified as M. bovis-infected, based on mycobacterial growth in culture and PCR (Supplemental Table). Positive M. bovis cultures were obtained from 9/23 lymph nodes, 1/1 BALF sample (performed postmortem) and 5/6 lung tissue samples from the six individuals. None of the 31 antemortem BALF samples (animal #32 had a duplicate) collected from the 30 free-ranging rhinoceros were mycobacterial culture positive (Table 1 and Supplemental Table).
Two of the 22 African elephant samples were infected with MTBC; one with M. tuberculosis T2 family and another with M. bovis SB0121 (Supplemental Table). The captive elephant infected with M. tuberculosis had positive cultures from antemortem samples including one trunk wash and two BALF, and a postmortem lung tissue sample ( Table 1). The free-ranging M. bovis-infected elephant was only sampled postmortem and M. bovis isolated from four lymph node and one lung samples (Table 1 and Supplemental Table). The remaining 20 freeranging elephants were considered uninfected based on negative mycobacterial cultures from the antemortem collection of BALFs and TWs from 19 individuals, respectively and the postmortem collection of one lung and one lymph node sample at necropsy from one additional individual animal found dead in the field (less than 24 h) ( Table 1 and Supplemental Table).
Ultra's performance on clinical specimens. Of the six M. bovis culture-confirmed white rhinoceros, the Ultra assay successfully identified all six infected animals from one BALF, eight lymph node, and four lung specimens (Table 1). Thirteen out of 30 samples obtained from these six M. bovis-infected rhinoceros had testpositive mycobacterium culture and Ultra assay results, and 10/30 samples were negative in both assays (Table 1). However, there were some discordant outcomes. Only for rhinoceros tissue samples, two were mycobacterial Scientific RepoRtS | (2020) 10:14482 | https://doi.org/10.1038/s41598-020-71568-9 www.nature.com/scientificreports/  www.nature.com/scientificreports/ culture-positive but Ultra-negative, and five were culture-negative but Ultra-positive (Table 1). Of the 30 culturenegative rhinoceros, antemortem BALF samples from seven animals tested positive in the Ultra assay (Table 1). Using samples from the M. tuberculosis or M. bovis culture-confirmed African elephants, the Ultra assay successfully identified both infected animals (Table 1 and Supplemental Table). Unlike the rhinoceros' specimens, all samples tested (four lymph nodes, two lung tissues, two BALs and one TW) from these two elephants were concordantly positive in the Ultra assay and mycobacterial culture (Table 1). Of the samples tested from the culture-negative elephants, 34/36 samples were also negative on the Ultra assay, with the additional detection of two test-positive samples (one lung and one TW sample) from this culture-negative cohort (Table 1).
Agreement between the Ultra and culture for all specimens was calculated as κ = 0.63, 95%CI 0.48 to 0.79 and a standard error (SE) of 0.08. According to Landis and Kock, this agreement can be considered as "substantial agreement" on their agreement scale 17 .

Discussion
The original Xpert MTB/RIF assay was one of the most widely used tests for TB and drug-resistance in humans 18 . However, since its endorsement by the World Health Organization (WHO) in 2010, the manufacturers have continued to enhance its diagnostic performance 13,16 , while unknowingly also adapting it into a candidate M. bovis diagnostic tool. Improved features are the inclusion of two new PCR targets within the system that target different multicopy genes (IS6110 and IS1081), conversion of assays into fully nested PCRs, and incorporation of a larger PCR reaction tube that doubles the amount of input DNA. These changes subsequently increased the analytical sensitivity of the assay for the detection of M. tuberculosis H37Rv from 81.0% to 87.5%, leaving the specificity unchanged at 98.7% 16 . In addition, the increased sensitivity of the Ultra was also observed through the enhanced LOD for other strains like M. bovis BCG. The Xpert MTB/RIF assay previously had a LOD for M. bovis BCG of 344.1 CFU/ml (95% CI 297.5-434.0) with the Ultra now having a LOD of 143.4 CFU/ml (95% CI 106.2-243.7). These findings suggest an increased sensitivity can be expected for different strains within the MTBC complex 16,19 . Findings from this study support those previously reported by Chakravorty et al., 2017 by demonstrating that different isolates such as M. bovis SB0121, M. tuberculosis H37Rv and M. tuberculosis T2 family can be detected using the Ultra assay. Furthermore, our preliminary findings suggest that (1) an overall relative sensitivity of 92.0% may be expected when using Ultra on wildlife tissue and respiratory tract samples and (2) lower LODs for M. tuberculosis H37Rv and M. bovis SB0121 isolates may also be observed for animal respiratory tract samples compared to that of human sputum. However, it must be noted that the latter observation may also be due to varying copy numbers of IS6110 and IS1081 between strain isolates from the MTBC and cannot be solely attributed to the performance of the Ultra in various specimen matrices 15 . Furthermore, interpretations of results from the unvalidated use of the Ultra on wildlife specimens must all be done cautiously since such specimens may interfere with results (pers. comm. Cepheid).
The overall observed "substantial agreement" between Ultra and culture on clinical specimens is supportive of Ultra's diagnostic potential to successfully differentiate between infected and presumed uninfected animals. However, the agreement between these two platforms was clearly affected by discordant results caused by the enhanced overall sensitivity of PCR technology compared to the known sub-optimal performance of mycobacterial culture techniques 10,11 . With regards to described discordant results, we believe Ultra's failure to detect two M. bovis culture positive rhinoceros tissue samples may have been due to a variety of reasons. These could include (1) pathogen DNA degradation as a result of freeze-thaw cycles during sample storage, transport or inappropriate sample collection 20 , (2) low levels of bacterial DNA below the limit of detection of the Ultra in tissue homogenates, especially since no prior DNA extractions on specimens were performed 13,16 , and (3) interference associated with the composition of the tissue specimens i.e. fat composition or presence of DNA degrading enzymes 21 . However, the Ultra platform does include an in-cartridge sample probe check that will identify the presence of any qPCR inhibition prior to commencing testing 18 .
The additional MTBC DNA detection of five culture negative rhinoceros lymph node samples, seven rhinoceros culture negative BALFs, one elephant culture negative lung specimen and one elephant culture negative TW specimen by the Ultra may have been due to false negative culture results 10 , or the detection of MTBC complex DNA from dead non-viable mycobacteria in the sample, as frequently observed with specimens from human TB patients 22,23 . However, detection in tissue and respiratory samples may also represent early MTBC infection considering that the animals included in this study were all from known MTBC-endemic areas with varying MTBC prevalence's in various wildlife species 5,7,24,25 . It is also possible that there was recognition by the Ultra of environmental nontuberculous mycobacteria (NTMs). Further comparisons to culture-based methods may help to elucidate this.
Limitations of this study include limited sample size, multiple freeze-thawing of samples prior to testing, a possible underestimated sensitivity and overestimated specificity for the assay when using animal samples. Further investigations should include the thorough evaluation of the sensitivity and specificity of the Ultra assay in animal samples. This can be achieved by testing fresh samples from more animals sampled from both MTBC endemic and negative wildlife parks with the Xpert machine as close to the animal as possible.
In conclusion, the Ultra assay appears to be a rapid diagnostic test for the detection of MTBC DNA from tissue and respiratory tract samples collected from African elephants and rhinoceros. Furthermore, the Ultra assay could provide a new tool for the detection of MTBC in various sample types from other wildlife species.

Materials and methods
Animals. During  www.nature.com/scientificreports/ lected from 16 African elephants that included 15 free-ranging exposed animals in KNP and one zoo elephant from Pretoria (South Africa) suspected of being MTBC-infected. In addition, tissue samples were collected from a euthanized zoo elephant and from two elephant carcasses found in KNP. Specimen collection and mycobacterial culture. All animal carcasses included in this study were examined for gross bTB lesions. Tissue samples with bTB-consistent lesions were collected and stored at − 20 °C. For animals with no visible lesions, samples of head and thoracic lymph nodes were pooled by anatomical site. For antemortem samples, endoscopic BALF samples were collected endoscopically as previously described 26,27 . Additionally, TW samples were obtained from immobilized elephants by flushing 250 ml of sterile saline into each nostril, separately, elevating and lowering the trunk for approximately 30-40 s and then aspirating the fluid into a sterile 500 ml collection chamber. All respiratory samples (BALF and TW) collected in 50 ml sterile falcon tubes were concentrated by centrifugation at 2,000 ×g for 30 min and decanting 46 ml of the supernatant. Pellets were resuspended and the remaining 4 ml were stored at − 20 °C. All tissue and respiratory samples were processed for mycobacterial culture using Mycobacteria Growth Indicator Tubes (MGIT) and the BACTEC MGIT 960 Mycobacterial detection system (Becton Dickinson, Franklin Lakes, NJ, USA), as previously described 28 .
All Ziehl-Neelsen positively stained bacterial cultures were genetically speciated by polymerase chain reaction (PCR) as previously described 29 . An additional 150 ml of BALF and TW samples were collected and frozen for all animals immobilized. Any rhinoceros and elephant with a mycobacterial culture negative/Ultra assay negative BALF and TW sample was selected for downstream spiking experiments. Animals with any cultureconfirmed tissue or respiratory tract sample were defined as MTBC-infected. Prior to use for spiking, all stocks were agitated in tubes containing sterile zirconium silica beads (Merck, USA) to remove clumps of bacteria. Thereafter, the supernatant was transferred to sterile microcentrifuge tubes and centrifuged at 3,000 ×g for 20 min. The supernatant was discarded, and the pellet resuspended in sterile Tris-EDTA buffer (pH 7).

Spiking of matrices with
In order to determine the detection range of M. bovis and M. tuberculosis bacilli by the Ultra assay, respiratory tract samples from culture-negative elephants (TW and BALF) and rhinoceros (BALF) were used as the matrices. Specific working concentrations of M. tuberculosis and M. bovis were spiked into all matrices to assay concentrations and tested as a single event on the same day as previously described 16 . Briefly, concentrations ranging from 30-2 and 0 CFU/ml for M. tuberculosis and 1 × 10 6 -2 and 0 CFU/ml for M. bovis 16 were spiked into BALF and TW samples from 6 elephants and BALF samples from 3 rhinoceros. The assay was performed in duplicate for each dilution. The limit of detection (LOD) was defined as the lowest number of CFU/ml which, when spiked into 1 ml of matrix, would result in 100% detection of M. tuberculosis and M. bovis by the Ultra assay in this study. Furthermore, 1 × 10 6 CFU/ml. M. avium were spiked into the respiratory sample matrix, in duplicate, and tested to determine assay specificity (Fig. 1).
Specimen preparation and testing with Ultra. All rhinoceros and elephant clinical tissue and respiratory tract samples were processed for mycobacterial culture as previously described 28,30 . Following a second thaw of the samples, they were also processed for Ultra testing according to the Xpert MTB/RIF alternate sample testing guide. Clinical specimens were directly tested using the Ultra, sample treatment and cartridge loading were done as previously described 13 . First, each sample (TW, BALF, tissue homogenate or spiked BALF and spiked TW) was directly mixed at a ratio of 2:1 with the Xpert lysis buffer containing NaOH and isopropanol (Cepheid, Sunnyvale, CA). The mixture was incubated for 15 min with occasional shaking and then loaded into the sample chamber of the cartridge for automatic processing and analysis as previously described 16 . The presence of MTBC DNA targets was confirmed using real-time signal from probes detecting the multicopy IS6110 and IS1081 genes. If confirmed, the Ultra uses four rpoB sloppy molecular beacon (SMB) probes to determine whether mutations associated with RIF resistance is present 16 . Additionally, the "trace" category was designated to identify samples containing lower number of MTBC bacilli according to the presence of the IS6110 and/or IS1081 molecular signals (Ct ≤ 37) in the absence of a signal from at least 3 of the rpoB SMBs. Automated MTBC detection and RIF susceptibility calls were performed by modified Xpert Diagnostic software (Cepheid). Results were defined as follows: (1) "MTB not detected" (assay will not proceed to high-resolution melt); (2) "MTB trace detected" but "RIF resistance indeterminate" (assay will not proceed to melt); (3) "MTB detected high/medium/ Scientific RepoRtS | (2020) 10:14482 | https://doi.org/10.1038/s41598-020-71568-9 www.nature.com/scientificreports/ low/very low" and "RIF resistance detected/not detected, " respectively. The assay allows for the identification of the presence of MTBC DNA in samples within 60 min and RIF resistant calls in less than 90 min. Full description of assay parameters are reported in Chakravorty et al. 16 .
Statistical analysis. Mean bacterial concentrations detected by the Ultra were shown as a line-graphs, left y-axis as Ct values, x-axis as CFU/ml and interquartile ranges generated for all replicates (GraphPad Prism v7.0.4, GraphPad Software). All positive and negative Ultra results and mycobacterial culture results were reported as proportions of the total number of samples tested. For all animal specimens, the agreement between the Ultra and culture was calculated as Cohen's kappa coefficient (κ) using the online agreement calculator on the Graph-Pad software website (https ://graph pad.com/quick calcs /kappa 1/). All results for the Ultra assay and mycobacterial culture results for each sample tested from individual animals are provided in the Supplemental Table. Received: 3 October 2019; Accepted: 17 August 2020 Scientific RepoRtS | (2020) 10:14482 | https://doi.org/10.1038/s41598-020-71568-9 www.nature.com/scientificreports/