Direct PCR on Tissue Samples To Detect Mycobacterium tuberculosis Complex: an Alternative to the Bacteriological Culture

Bovine tuberculosis (bTB) is an ongoing issue in several countries within the European Union. Microbiological culture is the official confirmation technique for the presence of Mycobacterium tuberculosis complex (MTBC) members in bovine tissues, but several methodological issues, such as moderate sensitivity and long incubation times, require the development of more sensitive and rapid techniques.

ABSTRACT

Bovine tuberculosis (bTB) is an ongoing issue in several countries within the European Union. Microbiological culture is the official confirmation technique for the presence of Mycobacterium tuberculosis complex (MTBC) members in bovine tissues, but several methodological issues, such as moderate sensitivity and long incubation times, require the development of more sensitive and rapid techniques. This study evaluates the analytical and diagnostic performance, comparative to culture, of a real-time PCR targeting the MTBC-specific IS6110 transposon using a panel of bovine tissue samples sourced from the Spanish bTB eradication campaign. Robustness and repeatability were evaluated in an interlaboratory trial between European Union National Reference Laboratories. The limit of detection with 95% confidence was established at 65 fg/reaction of purified genomic equivalents. Diagnostic sensitivity (Se) and specificity (Sp) were, respectively, 96.45% and 93.66%, and the overall agreement (κ) was 0.88. Cross-reactivity was detected against two mycobacterial isolates identified as Mycobacterium marinum and “Mycobacterium avium subsp. hominissuis,” and whole-genome sequencing (WGS) analysis of the latter isolate revealed an IS6110-like sequence with 83% identity. An identical IS-like element was found in other Mycobacterium avium complex species in the NCBI nucleotide and WGS databases. Despite this finding, this methodology is considered a valuable alternative to culture, and the strategy of use should be defined depending on the control or eradication programs.

INTRODUCTION

Bovine tuberculosis (bTB) is a chronic granulomatous infectious disease that mainly affects lymph nodes and lung tissue in cattle but can extend to other organs. This infection is mainly caused by Mycobacterium bovis and, to a lesser extent, by Mycobacterium caprae, but can also be produced by other members of the Mycobacterium tuberculosis complex (MTBC).

As bTB is a communicable disease, bTB surveillance or eradication campaigns are ongoing within the European Union (1, 2). Although 21 member states had officially tuberculosis-free (OTF) regions in 2018, bTB is still an ongoing problem in many parts of the European Union, with 11 non-OTF countries (3).

The difficulty in eradicating the disease could partially be explained by the complex interactions between the pathogen and different domestic and wildlife reservoirs (4). However, the success of the control measures used for the eradication of infectious diseases is also largely dependent on the performance of official diagnostic methods.

According to the European Union Council Directive 64/432/EEC (https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:01964L0432-20150527) on animal health problems affecting intracommunity trade in bovine animals, detection of bTB is based on immunological techniques that detect the cellular response of the host (single or comparative intradermal tuberculin test [SIT or SCIT] and supplementary testing as the gamma interferon assay) and on the identification of the agent. The presence of the members of the MTBC is confirmed by culture of the microorganism and is carried out in either liquid or solid media platforms. The fastidious growth of this microorganism can result in very long incubation times of up to 12 weeks (5–7), making this diagnostic method extremely slow. Furthermore, microbiological culture requires the use of a decontamination step, which negatively affects bacterial viability and can reduce recovery rates (5, 8). Although the diagnostic specificity (Sp) of microbiological culture is high, its diagnostic sensitivity (Se) is considered moderate (9). Consequently, the development of more sensitive and rapid tests for the confirmation of bTB could greatly improve the control of the disease.

Real-time PCR has been evaluated as a potential first-line technique for the detection of MTBC species in animal tissues across the globe (10). It is a simple, rapid, and robust technique with a performance comparable to that shown by microbiological culture (9, 11, 12). In addition to the reduced analysis times required (a few days versus several weeks in microbiological culture), it decreases the exposure to high bacterial loads and avoids contact with the hazardous chemicals used during decontamination. In comparison to microbiological culture, PCR can be affected by variations in the pathological status of the analyzed tissues, the reagents used, the genetic target, and, most importantly, the DNA extraction method (13).

The diagnostic performance of PCR has mostly been evaluated using microbiological culture as a reference. Despite the differences between studies in those, including animal tissues with visible lesions (VLs) and nonvisible lesions (NVLs), diagnostic Se of real-time PCR appears to be slightly better than that described for conventional PCR (74% to 100% versus 63% to 97%, respectively). On the other hand, diagnostic Sp is considerably higher in real-time PCR than conventional PCR (96% to 100% versus 55% to 100%, respectively) (9, 12, 14–19). The use of latent class analysis (LCA) for the evaluation of real-time PCR showed that diagnostic Se was higher than that of microbiological culture (87.7% versus 78.1%, respectively), and Sp was similar (97% versus 99.1%, respectively) (9).

This study describes the optimization and validation of a real-time PCR based on the IS6110 element, considered specific to the MTBC species. The performance of the IS6110-specific PCR is compared to that of microbiological culture in tissue samples obtained from the Spanish bTB eradication campaign. Additionally, this study describes the identification of an IS6110-like sequence in a “Mycobacterium avium subsp. hominissuis” strain and its conservation in different M. avium complex (MAC) species.

MATERIALS AND METHODS

PCR optimization.

The following oligonucleotides targeting 68 bp of the transposon IS6110, specific for MTBC species, were selected from the literature (20, 21): 5′-GGTAGCAGACCTCACCTATGTGT-3′ (F6110), 5′-AGGCGTCGGTGACAAAGG-3′ (R6110), and 5′-FAM-CACGTAGGCGAACCC-MGB-NFQ-3′ (S6110).

Analytical specificity of the oligonucleotides, as well as the IS6110 element (GenBank accession no. X17348.1), was evaluated in silico using the Basic Local Alignment Tool (BLAST) from the NCBI. Oligonucleotides were purchased from Eurofins Genomics (Ebersberg, Germany), and real-time PCRs were carried out using the Quantifast pathogen PCR +IC kit (Qiagen, Hilden, Germany). A heterologous exogenous internal amplification control (IAC), labeled with the MAX NHS Ester reporter dye, is included in the kit and was used to detect partial or complete inhibition in this study. According to the manufacturer, IAC cycle threshold (C_T) values should range between 30 ± 3. Therefore, when an amplification curve was detected and IAC C_T values were ≥33, samples were considered partially inhibited. Absence of amplification for the IAC and sample was considered complete inhibition. When inhibition was detected, samples were diluted 1:10 in ultrapure distilled water (Sigma-Aldrich, St. Louis, MO, USA), and the PCR was repeated.

M. bovis BCG strain Danish (CCUG 27863) genomic DNA was used to generate a standard curve. Genomic DNA was extracted from a pure culture grown in Middlebrook 7H9 medium supplemented with oleic albumin dextrose catalase (OADC) using phenol-chloroform-isoamyl alcohol (PCI). Briefly, 500 μl of culture was centrifuged, and the pellet was homogenized in ultrapure distilled water. The cell mixture was then transferred to a bead beater tube containing 500 μl of PCI 25:24:1 (Sigma-Aldrich) and incubated for 30 min. The tube was beaten in a Precellys Evolution homogenizer (Bertin Instruments, Montigny-le-Bretonneux, France) two times at 9,000 rpm for 40 s with 10-s intervals, and the genomic DNA was purified using ethanol as previously described (22). The obtained DNA pellet was dried at room temperature and resuspended in 50 to 100 μl of AE buffer (Qiagen). The original DNA stock concentration was measured, and quality was assessed using a NanoDrop spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). For standard curve generation, DNA concentration was set to 1.32 ng/μl using a Qubit 4 fluorometer (Thermo Fisher Scientific) and was serially diluted 1:10 from 1.32 ng/μl to 0.13 fg/μl. The standard curve was then used to evaluate the PCR efficiency as well as the linear range of the reaction in triplicates, which was defined as the range of dilutions where the reaction efficiency was 100% (±10%) and the coefficient of linearity was >0.997.

Following optimization, the setup for 25 μl of reaction volume was 7.5 μl of ultrapure distilled water, 5 μl of sample, 5 μl of Quantifast pathogen master mix, 2.5 μl of internal control assay (ICA), 2.5 μl of internal control DNA, 1 μl of F6110 (10 pmol/μl), 1 μl of R6110 (10 pmol/μl), and 0.5 μl of S6110 (10 pmol/μl). The run protocol was set as follows: an initial activation step of 95°C for 5 min followed by a denaturation step of 95°C for 15 s and an annealing/extension step of 60°C for 1 min. The denaturation and annealing/extension steps were repeated 45 times. Data acquisition was done at the annealing/extension step.

Analytical specificity, inclusivity, and limit of detection.

Inclusivity was tested in vitro using DNA from 6 MTBC species, including M. tuberculosis, M. africanum, M. bovis (including M. bovis strain BCG), M. caprae, M. microti, and M. pinnipedii. Analytical Sp was tested in vitro using 102 DNA preparations obtained from 24 nontuberculous mycobacteria (NTM) species, including M. avium subsp. avium (n = 12), M. avium subsp. hominissuis (n = 6), M. avium subsp. paratuberculosis (n = 3), M. avium complex (nonspecific strains) (n = 12), M. chitae (n = 2), M. elephantis (n = 1), M. europaeum (n = 2), M. flavescens (n = 1), M. fortuitum (n = 8), M. gordonae (n = 4), M. holsaticum (n = 1), M. intermedium (n = 1), M. intracellulare (n = 5), M. kansasii (n = 3), M. marinum (n = 2), M. neoaurum (n = 1), M. nonchromogenicum (n = 10), M. palustre (n = 1), M. parascrofulaceum (n = 1), M. peregrinum (n = 1), M. phlei (n = 5), M. seoulense (n = 1), M. shimodei (n = 1), M. smegmatis (n = 12), M. terrae (n = 2), M. thermoresistibile (n = 3), and M. vaccae (n = 1). Additionally, specificity was tested against DNA isolates of Corynebacterium spp. (n = 5), Streptomyces spp. (n = 5), and Lactobacillus brevis (n = 1). All DNA isolates were obtained from the strain repository at VISAVET Health Surveillance Center.

The analytical Se, or limit of detection (LOD), was defined as the lowest DNA concentration in which at least 95% of the replicates were positive. Twenty replicates of the lowest concentration in the standard curve in which all three replicates showed amplification, as well as 1:5 and 1:10 dilutions thereof, were evaluated.

Diagnostic validation of the IS6110-specific real-time PCR.

Diagnostic Sp of the direct real-time PCR was assessed in a panel of 100 tissue samples obtained from 78 bTB-free herds (SIT negative and originating from a region with a bTB prevalence of <1%). Diagnostic performance was further evaluated on 985 tissue samples obtained within the scope of the Spanish national bTB eradication campaign during the period from 2013 to 2017. The sampling population included positive (n = 608) and negative (n = 272) animals to the skin test, according to Spain’s Royal Decree 2611/1996, as well as carcasses with bTB-compatible lesions at the abattoir (n = 105). In all cases, different lymph nodes (retropharyngeal, mandibular, mediastinal, bronchial, prescapular, mesenteric, hepatic, and/or supramammary), and, if affected by VL, parenchymatous organs were sent for processing. Comparative diagnostic Se and Sp of the direct real-time PCR were evaluated against microbiological culture.

All fresh bovine tissue samples were sent to VISAVET Health Surveillance Center and processed within its biosafety level 3 (BSL3) facilities. Samples were collected from 12 autonomic regions in Spain, including Andalusia (n = 15), Aragon (n = 82), Canary Islands (n = 18), Castile-La Mancha (n = 160), Castile and Leon (n = 8), Basque Country (n = 1), Extremadura (n = 53), Balearic Islands (n = 57), Madrid (n = 465), Murcia (n = 44), La Rioja (n = 23), and Valencia (n = 59).

All samples were inspected for macroscopic lesions before processing. Four hundred and eight samples showed VL, while 577 showed NVL. Sample processing and culture were carried out as previously described (23) with minor modifications. After decontamination with hexadecylpyridinium chloride (HPC; 0.37% final concentration), centrifugation was carried out for 30 min at 2,500 × g. Löwenstein-Jensen medium with sodium pyruvate and Coletsos media (Difco, Madrid, Spain) were incubated at 37°C for a maximum of 3 months and were considered positive when growth compatible with MTBC was detected and confirmed using conventional PCR (24) and/or direct variable repeat (DVR) spoligotyping (25). For tissue DNA extraction, 1/10 of the volume (1 ml) of tissue homogenate used for culture was processed as previously described (11). Briefly, samples were lysed mechanically and chemically, and DNA was purified using a modified protocol from the DNeasy blood and tissue kit (Qiagen).

Culture-negative and direct real-time PCR-positive samples were further analyzed by the mpb70 real-time PCR as previously described (11), and an IS1245-specific PCR was used to confirm the absence of M. avium subspecies DNA in these samples (26). Positive samples by the mpb70 PCR were used as an indicator of presence of MTBC-specific DNA and, together with the culture results, were incorporated as adjusted results and compared with the IS6110 real-time PCR as a nonreference standard. In order to validate the agreement between the IS6110 and the mpb70 PCRs, a random selection of 200 samples were analyzed using both methods. In all reactions, double-distilled water and heat-inactivated M. bovis BCG isolate were used as negative and positive controls, respectively.

Statistical analyses.

The C_T cutoff used to determine if a tissue sample was positive or negative for MTBC DNA was established with a receiver operating characteristic (ROC) curve analysis using the ROCR package in R. Diagnostic Se, Sp, and predictive values, as well as their corresponding 95% confidence intervals (CI), were calculated using MedCalc statistical software 19.2.0 (MedCalc Software bv, Ostend, Belgium). Positive and negative percent agreements (PPA and NPA, respectively), as well as predictive values, were calculated using the same software for the IS6110 and mpb70 comparison and the adjusted results. Overall agreement between tests was evaluated using Cohen’s unweighted kappa in WinEpi 2.0 (27). Mann-Whitney two-tailed tests were used to assess the differences in the distributions of C_T values or DNA concentrations in different groups (negative versus positive culture, samples with VLs versus with NVLs, and inhibited versus noninhibited samples).

Interlaboratory performance evaluation.

An interlaboratory comparison trial was set up with four collaborating national and regional reference laboratories (RLs) for bTB. Laboratories A, B, C, and D tested 192, 93, 80, and 501 samples, respectively (n = 866). Each laboratory processed samples obtained from animals that had reacted positively to the official diagnostic techniques (skin test and/or interferon gamma [IFN-γ]) or had VLs compatible with bTB at the slaughterhouse inspection. The tissue samples were lymph nodes and parenchymatous organs with VLs or NVLs, except for laboratory C, which tested only MTBC culture-positive samples obtained from tissues with VL. The ratios of samples with VL and NVLs in the other laboratories were 84:108 (laboratory A), 53:40 (laboratory B), and 80:421 (laboratory D).

Microbiological culture was carried out in the different RLs according to their own validated procedures based on the OIE Manual for Terrestrial Animals with minor modifications. All laboratories used the same procedure for tissue DNA extraction, as previously described (11), and IS6110 real-time PCR protocol. Due to technical and biosafety reasons, two laboratories heat inactivated the samples prior to the DNA extraction protocol. Laboratory B incubated the samples at 95°C for 20 min, and laboratory C heat inactivated the samples at 80°C for 60 min. Additionally, laboratory B included the IAC DNA during the extraction protocol, whereas the rest added the IAC DNA directly to the master mix. Laboratories A and B carried out the real-time PCRs in CFX96 thermal cyclers (Bio-Rad, Hercules, CA, USA), whereas laboratory C and D used an Aria MX (Agilent, Santa Clara, CA, USA) and a StepOne Plus thermal cycler (Applied Biosystems, Foster City, CA, USA), respectively.

Whole-genome and Sanger sequencing.

Genomic DNA from an M. avium (CECT 7407) isolate that cross-reacted with the IS6110-PCR during specificity testing was sent to FISABIO (Valencia, Spain) for whole-genome sequencing. Libraries were generated using Nextera XT library prep kit (Illumina, San Diego, CA, USA) and sequenced using a Miseq reagent kit v3 (600 cycles) in a Miseq sequencer. Raw reads were adapter and quality trimmed using Trimmomatic v0.39 (slidingwindow,15:30; minlen, 36), clumped using clumpify v38.49 from BBTools, and deduplicated using PRINSEQ-lite v0.20 (28). Processed reads were aligned against the phiX-174 reference genome (GenBank accession no. NC_001422.1) using BWA (29). Unmapped reads were assembled de novo with SPAdes v3.13 (30), and assembly quality was assessed using QUAST (31) with default parameters and Mycobacterium avium subsp. hominissuis strain MAC109 (GenBank accession no. CP029332.1) as a reference.

Contigs were deposited in a local BLAST database using makeblastdb from the BLAST+ package from the NCBI. The IS6110 sequence from M. tuberculosis H37Rv (GenBank accession no. NC_000962.3; positions 1541952 to 1543306) was used as a query for a local BLASTn against the contig database. The aligned sequence was extracted from the contig using bedtools (32) for further analysis. Plasmid sequences were identified using BLASTn.

Further confirmation of the IS6110-like element was achieved through Sanger sequencing (Stab Vida, Caparica, Portugal) and the use of primers 5′-TCAGGTGGTGCGTCGAAGA-3′ (ISlike-F) and 5′-AATCCAGCACCCCTTCTTGT-3′ (ISlike-R).

Data availability.

This whole-genome shotgun project has been deposited at DDBJ/ENA/GenBank under the Whole Genome Shotgun accession no. JAAILH000000000. The version described in this paper is version JAAILH000000000.1. The nucleotide sequence of the IS6110-like element obtained through Sanger sequencing was deposited at the NCBI GenBank under accession no. MT818214.

RESULTS

Optimization and LOD.

A stock of 13.21 ng/μl (standard deviation [SD], 0.24) of genomic DNA, obtained from a pure culture of M. bovis BCG Danish, was serially diluted 1:10 from an initial concentration of 1.32 ng/μl to 0.13 fg/μl. After optimization, the range of C_T values spanned from an average of 18.83 cycles (SD, 0.09) at 6.5 ng/reaction to 39.18 cycles (SD, 0.27) at 6.5 fg/reaction (Fig. 1a). An overall efficiency of 100% was obtained from 6.5 ng/reaction to 650 fg/reaction, with an R² of 0.999 (Fig. 1b).

FIG 1.

(a) IS6110 real-time PCR amplification range and efficiency. The standard curve ranged from 6.5 ng/reaction to 6.5 fg/reaction. (b) Dynamic range of the reaction that shows best efficiency and coefficient of linearity (R²).

The LOD was established at 65 fg/reaction (average C_T value, 35.44, and coefficient of variation, 1.68%) with the 95% confidence limit, and all replicates (n = 20) were positive.

Analytical specificity and inclusivity.

The in silico specificity analysis using the 68-bp amplicon obtained from the IS6110 sequence (33) yielded no significant alignments. All tested MTBC species were positive for this real-time PCR. Five NTM isolates (2 M. fortuitum, 1 M. kansasii, 1 M. intermedium, and 1 M. nonchromogenicum) produced a positive amplification curve with a C_T value >38. Analysis of these isolates indicated a much higher concentration of mycobacterial DNA than that expected in tissue extracts (>400 ng/μl), and dilutions to expected working concentrations (<1 ng/μl) of mycobacterial DNA in tissue extracts produced negative results. Two additional NTM isolates (M. marinum and M. avium CECT 7407) produced amplification curves at C_T values of ≤38 and remained positive at lower concentrations of DNA.

Diagnostic validation.

None of the tissue samples obtained from 78 bTB-free herds (n = 100) produced an amplification curve.

In order to determine the diagnostic performance of the real-time PCR relative to microbiological culture, 985 bovine tissue samples were analyzed using both methods. The ROC analysis established that a cutoff of 38.70 produced the highest diagnostic Se and Sp relative to culture, with an area under the curve (AUC) of 96% (95% CI, 94.6% to 97.3%) (Fig. S1 in the supplemental material). As a result, 326 samples were MTBC positive by both microbiological culture and real-time PCR, and 12 were culture-MTBC positive and real-time PCR negative (Table 1). When other microorganisms were cultured from 63 samples (NTM, 23; Actinomycetales, 10; other microorganisms, 30), they were considered MTBC negative. There were 606 samples negative by both microbiological culture and real-time PCR, and 41 were positive only by the real-time PCR. Diagnostic Se and diagnostic Sp relative to culture were 96.45% (95% CI, 93.90% to 98.15%) and 93.66% (CI, 91.51% to 95.29%), respectively. Positive and negative predictive values (PPV, 88.83%; NPV, 98.06%) as well as agreement (κ = 0.88) between tests were good (Table 1).

TABLE 1.

Comparison of diagnostic performance between IS6110 real-time PCR and microbiological culture in 985 bovine samples

PCR result	Microbiological culture result			Diagnostic performance
	No. positive	No. negative	Total no.	Sensitivity (% [95% CI])	Specificity (% [95% CI])	Predictive value (% [95% CI])		Agreement (95% CI)
						Positive	Negative
No. positive	326	41	367	96.45 (93.90–98.15)	93.66 (91.51–95.29)	88.83 (85.19–91.66)	98.06 (96.64–98.89)	0.88 (0.85–0.91)
No. negative	12	606	618
Total no.	338	647	985

Cycle threshold distribution was analyzed in MTBC culture-positive and PCR-positive samples and MTBC-culture negative and PCR-positive samples. C_T values were statistically higher (Mann-Whitney two-tailed test; P < 0.01) in the culture-negative group than the culture-positive group (Fig. 2a).

FIG 2.

Cycle threshold comparison between culture-positive and culture-negative samples (a) and samples with VLs and NVLs (b). Statistical significance was carried out using a Mann-Whitney two-tailed test.

When considering the presence or absence of VLs in the analyzed tissues, 303 of the 326 culture-positive and PCR-positive samples were obtained from tissues with VL (Table 2). On the other hand, 28 out of 41 culture-negative and PCR-positive samples had VL. Diagnostic Se and Sp were, respectively, 97.74% (95% CI, 95.40% to 99.09%) and 71.43% (95% CI, 61.42% to 80.10%) in tissues with VLs and 82.14% (95% CI, 63.11% to 93.94%) and 97.63% (95% CI, 95.98% to 98.73%) in tissues with NVLs. Overall agreement was moderate (0.70 to 0.75) in both groups, whereas predictive values were generally high (90.91 to 99.8%), with the exception of the positive predictive value in NVLs (63.89%) (Table 2). Cycle thresholds were significantly lower (P < 0.01) in samples with VLs than samples with NVLs (Fig. 2b).

TABLE 2.

Comparison between the IS6110 real-time PCR and microbiological culture in relation to the presence or absence of anatomic lesions compatible with bTB

PCR resulta	Culture result			Diagnostic performance
	No. positive	No. negative	Total no.	Sensitivity (% [95% CI])	Specificity (% [95% CI])	Predictive value (% [95% CI])		Agreement (95% CI)
						Positive	Negative
VL				97.74 (95.40–99.09)	71.43 (61.42–80.10)	91.54 (88.78–93.67)	90.91 (82.63–95.46)	0.75 (0.67–0.82)
No. positive	303	28	331
No. negative	7	70	77
Total no.	310	98	408
NVL				82.14 (63.11–93.94)	97.63 (95.98–98.73)	63.89 (50.16–75.67)	99.08 (97.98–99.58)	0.70 (0.58–0.83)
No. positive	23	13	36
No. negative	5	536	541
Total no.	28	549	577

^aVL, visible lesions; NVL, nonvisible lesions.

From the total of analyzed samples, two showed complete inhibition, whereas 44 were partially inhibited. One of the two inhibited samples gave a positive reaction after a 1:10 dilution. Fifteen out of 44 partially inhibited samples resulted in target amplification despite a partial inhibition of the IAC. One of the 29 remaining partially inhibited samples turned IS6110 positive after dilution. These positive samples were included among the real-time PCR-positive samples.

The analysis of 200 randomly selected samples through the IS6110 and mpb70 PCRs showed a very high overall agreement (κ = 0.91 [95% CI, 0.85 to 0.97]) (Table S1). More specifically, positive and negative agreement (PPA and NPA) values were also very high (93.10% [95% CI, 85.59% to 97.43%] and 97.35% [95% CI, 92.44% to 99.45%], respectively). Therefore, discordant results between microbiological culture and IS6110 real-time PCR were further investigated using the mpb70 PCR. This mpb70 PCR confirmed the presence of MTBC DNA in 35 out of 41 culture-negative and IS6110 PCR-positive samples (Table S2). Of these, 27 proceeded from samples with VLs (Table S3). After detection of MTBC-specific DNA in culture-negative and real-time PCR-positive samples, PPA and NPAs with respect to the adjusted results were 96.78% (95% CI, 94.46% to 98.15%) and 99.02% (95% CI, 97.88% to 99.55%), respectively (Table S2). The overall agreement and PPV of the IS6110 PCR increased to 0.96 and 98.37%, respectively. Additionally, the presence of MTBC DNA was detected in 27/28 (VL) and 8/13 (NVL) positive samples (Table S3), and agreement between real-time PCR and the adjusted results was high when analyzing samples with VLs (κ = 0.93 [95% CI, 0.89 to 0.98]) and NVLs (κ = 0.85 [95% CI, 0.76 to 0.94]).

After the adjustment of discordant results, the C_T cutoff recommended through the ROC analysis was very similar (C_T, 39), and, when used, diagnostic performance did not vary significantly (data not shown).

Interlaboratory performance evaluation.

Diagnostic performance of the real-time PCR relative to microbiological culture was assessed in four RLs as part of interlaboratory testing. The same C_T cutoff as specified previously was used for the analysis. In laboratory A, 78 samples were positive by both the PCR and microbiological culture, whereas 2 were only positive to culture (Table 3). Forty out of 112 culture-negative samples were positive by the PCR. Diagnostic Se, Sp, and agreement between tests was, respectively, 97.50% (95% CI, 91.26% to 99.70%), 64.29% (95% CI, 54.68% to 73.12%), and 0.58% (95% CI, 0.85% to 0.91%) (Table 3). In laboratory B, 36 out of 54 culture-positive samples were also positive by the real-time PCR, whereas 4 out of 39 culture-negative samples were positive by the real-time PCR. Diagnostic Se, Sp, and agreement between tests were 66.67% (95% CI, 53.36% to 77.76%), 89.74% (95% CI, 76.42% to 95.94%), and 0.54% (95% CI, 0.39% to 0.69%), respectively. In laboratory C, 79 out of 80 culture-positive samples were also positive by PCR, and diagnostic Se was 98.75% (95% CI, 93.25% to 99.94%). Finally, 81 samples were culture and real-time PCR positive in laboratory D, whereas 416 were negative by both techniques. Four samples were positive only by culture, and no samples were positive by the real-time PCR alone. Diagnostic Se, Sp, and agreement between tests were, respectively, 100.00% (95% CI, 95.55% to 100.00%), 99.05% (95% CI, 97.58% to 99.74%), and 0.97% (95% CI, 0.94% to 1.00%). These results were translated into low to very good PPVs (66.10% to 95.29%) and NPVs (66.04% to 100.00%).

TABLE 3.

Comparison between microbiological culture results and real-time PCR results in the interlaboratory diagnostic validation study^a

PCR result	Culture result		Diagnostic performance
	No. positive	No. negative	Sensitivity (% [95% CI])	Specificity (% [95% CI])	Predictive value (% [95% CI])		Agreement (95% CI)
					Positive	Negative
Laboratory A			97.50 (91.26–99.70)	64.29 (54.68–73.12)	66.10 (60.27–71.48)	97.30 (90.10–99.30)	0.58 (0.48–0.68)
No. positive	78	40
No. negative	2	72
Laboratory B			66.67 (52.53–78.91)	89.74 (75.78–97.13)	90.00 (76.95–96.04)	66.04 (52.59–77.31)	0.54 (0.39–0.69)
No. positive	36	4
No. negative	18	35
Laboratory C			98.75 (93.25–99.94)	NA	NA	NA	NA
No. positive	79	ND
No. negative	1	ND
Laboratory D			100.00 (95.55–100.00)	99.05 (97.58–99.74)	95.29 (88.42–98.17)	100	0.97 (0.94–1.00)
No. positive	81	4
No. negative	0	416

^aND, not determined; NA, not applicable.

Identification of an IS6110-like element in Mycobacterium avium subsp. hominissuis.

Due to the possibility of detecting M. avium from cattle, a more careful analysis of the M. avium isolate that cross-reacted during the analytical Sp validation was carried out using WGS.

After processing of the raw reads, 3,462,172 reads remained. The de novo assembly of M. avium CECT 7407 isolate (Whole Genome Shotgun project accession number: JAAILH000000000) yielded 133 contigs (>500 bp) and an N₅₀ of 108,694 bp. Compared to the reference using QUAST, the assembly represented 94.96% of the genome and had a duplication ratio of 1.00 and an NGA of 88,010 bp.

Alignment of the IS6110 sequence against these contigs using local BLASTn produced a hit with 83% similarity (Fig. S2), which corresponded to contig 83 (5,095 bp). Although a misalignment with the reference was detected in this contig (positions 1 to 2734 and positions 2735 to 5095), this did not include any part of the IS6110-like sequence (positions 3919 to 4934). Oligonucleotide sequence identity with the IS6110-like element was 86.40% (19/22) for F6110, 82.30% (14/17) for R6110, and 85.70% (12/14) for S6110.

An online search using BLASTn from the NCBI against the nucleotide and WGS databases was carried out to assess the existence of similar sequences. Apart from IS6110 elements from members of the MTBC, alignment against the nucleotide archive produced a hit with 100% query coverage and identity with Mycobacterium avium subsp. hominissuis MAC109 (GenBank accession no. CP029332.1; positions 89000 to 90959). Annotation of this region indicated the presence of two truncated sequences from the IS256 family flanking the IS6110-like element.

In the case of the WGS contig database, alignments with 99 to 100% coverage and identity were found in 31 Mycobacterium avium subsp. hominissuis strains, 6 M. avium strains, as well as in 1 Mycobacterium bouchedurhonense and 1 Mycobacterium timonense strain (GenBank accession nos. MVIL01000161.1 and MVHL01000039.1, respectively).

Further analysis of the IS6110-like element with ISFinder produced partial alignments with IS6110 in the IS3 family of insertion sequences. The alignment had an approximate identity of 83% and was characterized by an incomplete left inverted repeat (IRL), which contained a 5-bp-long insertion at the 5′ end (Fig. S2), as well as a truncated 3′ extreme missing the right inverted repeat (IRR).

DISCUSSION

Real-time PCRs have become an important technique in the rapid detection and identification of pathogens around the world. Although bTB detection by direct PCR has been evaluated previously with remarkable results (9, 12), microbiological culture is still considered the reference technique for the detection of tuberculosis in bovine tissues. This study described the optimization and validation of a real-time PCR targeting the IS6110 element, considered specific to members of the MTBC, and evaluated its diagnostic performance with tissue samples obtained from the Spanish bTB eradication campaign. An interlaboratory performance comparison was pursued through a collaboration with four reference laboratories from the European Union.

The IS6110-specific real-time PCR detected, during analytical validation, at least 13.6 genome equivalents/reaction (65 fg/reaction) with 95% confidence (34), taking into account that IS6110 is only present in one copy in the M. bovis BCG strain. In M. bovis strains and MTBC species with multiple copies of the IS6110, the LOD could be lower.

The IS6110-specific real-time PCR, in combination with the DNA extraction method, resulted in a rapid technique with good diagnostic performance (Se, 96.45%, and Sp, 93.66%) compared to culture. These results, including the high agreement between tests (κ = 0.88), highlight the potential of this real-time PCR as a first-line technique. Compared to other real-time PCR studies, the diagnostic Se for this PCR was similar (87.70 to 100%) in some cases (9, 11, 12) and higher in others (66.70 to 76.70%) (14–16, 35, 36). Specificity was, in general, slightly higher in these studies (96.03 to 100%). Nevertheless, the individual comparison between studies can be affected by the differences in molecular targets used (single copy versus variable copy), DNA extraction protocols, experimental design (type of culture isolation, number and type of samples, nested versus nonnested, etc.), and validation method (experimental versus Bayesian analysis) (9, 11, 12, 14–16, 34–36).

Although agreement between microbiological culture and the real-time PCR was very good (κ = 0.88), discordant results between the two methodologies were identified and further investigated.

The detection of MTBC DNA by mpb70 real-time PCR in 35 out of 41 culture-negative samples highlights possible limitations of culture, namely, lack of sensitivity, reduction of cell viability, and growth of other microorganisms, including NTM. The type of chemical decontamination and the time of exposure have an important effect on bacterial viability (8), which can have a significant effect in samples with already low bacterial loads. Indeed, real-time PCR C_T values revealed smaller amounts of MTBC DNA in culture-negative samples than in culture-positive ones. In addition, cellular viability could also have been affected by the pathological status of the granuloma (37). Although the majority (27/35) of these samples were obtained from tissues with VLs, no histopathological analysis was carried out on these tissues.

Diagnostic performance of the IS6110 real-time PCR compared to the adjusted results was very high irrespective of the presence or absence of lesions in the analyzed tissue (Table S3 in the supplemental material). Indeed, 27/28 (VL) and 8/13 (NVL) positive samples would have been missed if only culture had been carried out, in comparison to 7 (VL) and 5 (NVL) culture-positive and real-time PCR-negative samples. This suggests that real-time PCR could be an effective technique in the detection of bTB in tissues with NVLs and possible paucibacillary loads in comparison to microbiological culture.

Even though the DNA extraction protocol was carried out using 1/10 of the amount used for microbiological culture, only a small number (12/338) of culture-positive samples were negative by this PCR, indicating a good extraction efficiency. Repetition of the DNA extraction protocol in 8 out of 12 culture-positive and PCR-negative samples yielded two PCR positive results by the IS6110 and mpb70 PCRs (data not shown), indicating that the extraction procedure is critical for the performance of PCR and can be responsible for some of the false PCR-negative results.

The presence of inhibitors can also have an important role in the performance of PCR for the detection of pathogen DNA. Inhibitors generally include excess host DNA and organic compounds such as urea, collagen, or bile salts, which can already be present in the sample or carried over during DNA extraction. Only 0.2% (n = 2) and 2.94% (n = 29) of the samples were completely or partially inhibited with no IS6110 amplification, respectively. Eight of the partially inhibited samples were culture positive, highlighting the importance of including an inhibition control. DNA concentration in inhibited samples was higher (median [M], 353.65 ng/μl; interquartile range [IQR], 269.05 to 453.73 ng/μl) to that observed in noninhibited samples (M, 292.20 ng/μl; IQR, 201.7 to 391.8 ng/μl), indicating that excess host DNA could be responsible for the inhibition effect (P = 0.02). In order to overcome the inhibition, samples were diluted 1:10, which resulted in the amplification of MTBC DNA in 1 completely and 1 partially inhibited samples. However, in those samples with low bacterial loads, dilution can result in undetectable concentrations of target DNA. Certainly, the presence of MTBC DNA was confirmed in one of these samples using the mpb70 real-time PCR. Therefore, the DNA extraction should be repeated in all negative diluted samples, for which appropriate backup frozen material should be available.

Another reason behind discordance between culture-positive and PCR-negative samples could be the absence of the IS6110 transposon. The existence of isolates lacking the IS6110 element, although infrequent, has been described previously (38). When a single copy of the IS6110 element is present, it is usually located in spacer 24 in the direct repeat locus (25). An initial analysis was carried out in some isolates available in our laboratory characterized by spoligotyping as sp24-deleted spoligotypes (21 out of 198 sp24-deleted patterns on Mbovis.org). No IS6110 amplification was detected in 6 of them, which include SB1263, SB1561, SB1630, SB1678, SB1888, and SB1901 (data not shown). All these patterns lost at least 3 spacers around spacer 24. From the total of SB patterns detected in the validation of this real-time PCR (n = 30), only one sp24-deleted spoligotype (SB1263) was detected, which represented two of the culture-positive and PCR-negative samples. The presence of MTBC DNA was confirmed using the mpb70 real-time PCR in these two samples.

In order to evaluate the diagnostic performance of the real-time PCR in different scenarios, an interlaboratory trial was carried out in four RLs using samples obtained from the eradication campaigns of the participating countries. Samples were processed using the standard procedures and biosafety measures of each laboratory. The interlaboratory performance evaluation revealed an overall high diagnostic Se relative to microbiological culture. However, diagnostic Se was considerably lower in laboratory B than in the rest of the laboratories. An important difference with the established protocol in this laboratory was the addition of a heat inactivation step of 95°C for 20 min before the extraction protocol. The effect of high-temperature treatment on the tissue homogenate before extraction has not been evaluated during the validation of the protocol, and therefore, a negative effect cannot be ruled out. A pilot test in our laboratory evaluated the effect of high-temperature treatment (100°C for 15 min) on 43 randomly selected tissue homogenates before DNA extraction, and a reduction in PCR Se was reported (19%; data not shown). Interestingly, laboratory C also heat inactivated samples at 80°C and obtained higher diagnostic Se, but only samples with visible lesions were analyzed. Therefore, other factors can also be related to this reduction, such as a paucibacillary bacterial load in the animal tissues and/or the subsequent mechanical disruption steps. Diagnostic Sp with respect to culture was moderate to high in 3 laboratories (A, B, and D). Discrepancies in Sp could be explained by differences in isolation efficiency due to different decontamination procedures and sample processing before culture, which could have affected bacterial viability and microbiological yield. In conclusion, these results indicate that the performance of the real-time PCR can vary depending on the technical requirements of each facility and that these must be carefully evaluated before its implementation.

Few diagnostic specificity discrepancies could be related to cross-reactivity of the IS6110 probes with other bacteria. Although none of the 100 tissue samples obtained from 78 bTB-free herds were positive to this real-time PCR, giving an absolute diagnostic Sp of 100%, two cross-reactivity issues were detected during the validation of this PCR. Cross-reactivity with M. marinum has limited significance in the diagnosis of bTB, as this mycobacterial species is not a common pathogen of cattle. On the other hand, cross-reactivity with M. avium CECT 7407, identified as Mycobacterium avium subsp. hominissuis by PCR and hsp65 sequencing (data not shown), required further analysis. Mycobacterium avium subsp. hominissuis is mostly isolated in pigs and represents between 6 and 14% of NTM isolates in cattle (21, 39–41), being present as an opportunistic microorganism in bTB diagnosis (40). However, the real prevalence of Mycobacterium avium subsp. hominissuis in cattle is probably underestimated, as this pathogen is not searched for actively. Most isolations have been recorded in mesenteric lymph nodes, in accordance with the suspected oral-fecal transmission of MAC species, but isolates have also been detected in cattle respiratory samples (21, 42).

Other non-MTBC microorganisms were isolated in nine IS6110 real-time PCR-positive samples. Seven of these samples were confirmed as MTBC infections with the mpb70 direct real-time PCR. The remaining isolates were confirmed as Mannheimia granulomatis and one NTM species using 16S sequencing and conventional PCR, respectively (24). Interestingly, the latter isolate was negative to the IS1245-specific PCR (data not shown), indicating that it probably did not contain MAC DNA but other NTM DNA.

The IS6110 has been generally considered specific for the MTBC complex, although cross-reactivity with certain primer pairs or probes has been identified previously (36, 43, 44). The WGS analysis of the cross-reacting M. avium strain resulted in the discovery of an IS6110-like element with a sequence similarity of 83% with IS6110. The presence of this element was further confirmed through PCR and Sanger sequencing (Fig. S2). The truncation of the DDE motif and its IRs deemed this element inactive (Fig. S2). The IS6110-like sequence identified in this study, along with the rest of the contig’s sequence, was identical to that of Mycobacterium avium subsp. hominissuis MAC109 (45), which had not been deposited in the NCBI at the time of the initial in silico specificity analysis. Another inactive IS6110-like element has been previously described in M. smegmatis strain MKD8 (46), although it only shares 50% and 70.79% coverage and identity with the IS6110-like element described in this study.

Interestingly, 100% identity and coverage were detected between the IS6110-like sequence and other contigs from Mycobacterium avium complex strains stored within the NCBI WGS database, including several Mycobacterium avium subsp. hominissuis strains, M. bouchedurhonense, and M. timonense (a selection is aligned in Fig. S2). However, the last two species have not been detected in cattle in any NTM surveillance study so far (21, 39–41, 47, 48). Genomic comparative studies of mycobacterial species showed that these two species cluster together with M. avium subspecies separately from the rest of the MAC (49, 50).

Despite the identification of an IS6110-like element, the 68-bp amplicon from CECT 7407 could not be sequenced, and therefore, a causal relationship between this element and the unspecific fluorescence signal could not be established (data not shown). Additionally, the absence of amplification in the rest of MAC isolates tested during the specificity study (n = 33), including 5 Mycobacterium avium subsp. hominissuis isolates, indicates that the cause of cross-reactivity may not be conserved in all MAC species or that the accumulation of mutations may hamper oligonucleotide binding. Furthermore, no cross-reactivity issues were detected against MAC isolates in a recent study published in France when the same primers and probe were used (21). A possible reason for this could be related to lineage-specific recombination and inversion events in M. avium genomes (51, 52). Further WGS comparisons between MAC species are required in order to confirm the presence of the IS6110-like sequence and understand the possible origin of this cross-reactivity.

Nevertheless, the appearance of an IS6110-like element in the genomes of several MAC species is an interesting finding and could be a result of horizontal gene transfer events (HGT) between a species of the MTBC and a MAC ancestor. Several HGT phenomena have been described in mycobacteria, such as the transfer and recombination of chromosomal DNA through distributive conjugal transfer (DCT) in M. smegmatis and M. canetti (53, 54). Although DCT has not been experimentally demonstrated in the MTBC (54), transitory conjugation has been observed in M. tuberculosis and M. bovis BCG (55), and recombination events have been identified in M. tuberculosis, indicating that DNA exchange and recombination is still possible even after the clonal expansion and genomic deletion events that led to MTBC speciation (56). Certainly, further research is required to understand the importance of HGT between MAC and MTBC species.

Based on the results, the application of the IS6110 real-time PCR on tissue samples is a very promising first-line molecular technique for the detection of MTBC species in bovine tissues. The rapidity of this technique, together with the high diagnostic sensitivity and specificity, makes this protocol an alternative to microbiological culture in some scenarios. Quick results would allow animal health authorities to make future decisions regarding bTB-infected herds in a shorter time than microbiological culture. However, the isolation of MTBC will still be required on most of the PCR-positive samples in order to carry out molecular epidemiological studies. Therefore, the strategic use of the direct PCR should be defined specifically considering the epidemiological situation, prevalence of bTB in the area of study, or the presence/absence of compatible TB lesions and could lead to cost-effective management of bTB diagnostic resources. Although the effect of the abovementioned specificity issues is limited, the possibility of cross-reactivity with other NTMs cannot be ruled out. Therefore, the additional use of an MTBC-specific PCR, such as the mpb70 PCR, and/or MAC-specific real-time PCR, as a confirmation of positive samples could further improve the effectivity of direct molecular detection of bTB in animal tissues. This would be especially relevant in those situations where speed, high specificity, and sensitivity are required (such as a low bTB prevalence area or a bovine OTF herd with a suspended status). In conclusion, the implementation of this protocol could aid in the eradication of this persevering infectious disease.