Development and validation of a multiplex real-time qPCR assay using GMP-grade reagents for leprosy diagnosis

Leprosy is a chronic dermato-neurological disease caused by Mycobacterium leprae, an obligate intracellular bacterium. Timely detection is a challenge in leprosy diagnosis, relying on clinical examination and trained health professionals. Furthermore, adequate care and transmission control depend on early and reliable pathogen detection. Here, we describe a qPCR test for routine diagnosis of leprosy-suspected patients. The reaction simultaneously amplifies two specific Mycobacterium leprae targets (16S rRNA and RLEP), and the human 18S rRNA gene as internal control. The limit of detection was estimated to be 2.29 copies of the M. leprae genome. Analytical specificity was evaluated using a panel of 20 other skin pathogenic microorganisms and Mycobacteria, showing no cross-reactivity. Intra- and inter-operator Cp variation was evaluated using dilution curves of M. leprae DNA or a synthetic gene, and no significant difference was observed between three operators in two different laboratories. The multiplex assay was evaluated using 97 patient samples with clinical and histopathological leprosy confirmation, displaying high diagnostic sensitivity (91%) and specificity (100%). Validation tests in an independent panel of 50 samples confirmed sensitivity and specificity of 97% and 98%, respectively. Importantly, assay performance remained stable for at least five months. Our results show that the newly developed multiplex qPCR effectively and specifically detects M. leprae DNA in skin samples, contributing to an efficient diagnosis that expedites the appropriate treatment.


Introduction
Leprosy is a neglected infectious disease that still represents a public health issue [1] with more than 200,000 cases every year worldwide. Diagnosis is generally late and, although a specific and effective treatment is available, it is likely that transmission occurs before the patient is diagnosed and adequately treated, thus contributing to sustained transmission. The high number of young patients (under 15 years old) and patients with disabilities due to the advanced stage of the disease, confirms this hypothesis [1]. Furthermore, clinical forms vary to a great extent, from localized (tuberculoid) to disseminated (lepromatous) forms, making diagnosis difficult. Evidence suggests that early diagnosis could prevent transmission and help epidemiological control [2].
Methods such as bacterial index detection by microscopy and histopathological examination have been the main complementary tools for the diagnosis of leprosy [2][3][4]. Classical bacteriological methods cannot confirm leprosy since M. leprae does not grow in vitro. In addition, there is no reliable marker to estimate the risk of disease progression [5,6]. In this regard, the sequencing of M. leprae genome [7] was a milestone towards the improvement of direct M. leprae detection, leading not only to better characterization of genomic targets unique to M. leprae strains but also to an extensive comparison of different mycobacteria.
At the time when the first sequences became available, the polymerase chain reaction (PCR) technique was laborious and very expensive, averting its universal application. However, as PCR was further developed, it became more affordable, versatile and reliable, with fully automated systems becoming commercially available from different companies [8][9][10]. For tuberculosis, routine tests using PCR reduced the turnaround time, allowing same day treatment initialization, which might impact resistance prevalence [11][12][13]. Cost-effective nucleic acid detection assays are relatively widespread, but assays for some neglected diseases are still missing. In leprosy, the situation is even more difficult due to reduced and late investments directed to diagnostic tests [14].
In the last few years, many studies have been carried out using the PCR technique to detect M. leprae DNA in clinical specimens. PCR has been used especially under challenging diagnoses such as equivocal paucibacillary [4,[15][16][17][18] or monitoring household contacts [19,20]. In this context, several different targets have been described in an attempt to establish the most sensitive and specific assay [16,[20][21][22][23][24][25][26][27][28]. However, most of the PCR protocols were developed, evaluated, and validated using reagents or tests produced without good manufacturing practices (GMP). Also, most of the studies enroll only leprosy patients and do not recruit patients with other common dermatological diseases that are differential diagnosis to leprosy. Thus, the development and validation of an assay over different laboratories has become a necessity.
Here, we present the development and validation of a multiplex real-time PCR assay aiming to standardize the leprosy molecular diagnostic assay. The protocol was designed to simultaneously detect two M. leprae targets (16S rRNA and RLEP genes), previously used in several studies [4,16,19,26,29], and one mammalian target (18S rRNA gene), that serves as reaction control [30]. Cross-reactivity was evaluated using DNA from 20 related mycobacterial and other skin pathogenic species, and no match was found. The new assay was validated using 97 skin biopsies and an independent panel enrolling 50 samples retrieved from patients previously characterized by clinical examination and histopathology, showed high sensitivity and specificity. The new multiplex PCR was also assessed for quality control standards and the data indicate that the assay is stable and reproducible. The results presented here are the basis of a novel and robust tool with potential to increase the accuracy of leprosy diagnosis in routine or reference laboratories.

Ethics statement
The Ethics Committee of the Oswaldo Cruz Foundation approved this study (CAAE: 38053314.2.0000.5248, number: 976.330-10/03/2015). Written informed consent was obtained from all patients. If participant was a minor, written formal consent was obtained from the parent/guardian.

Clinical samples
Leprosy patients were enrolled at the Leprosy clinic from the Oswaldo Cruz Foundation in the city of Rio de Janeiro, Brazil. Skin biopsies were collected using a 6 mm punch and stored in 70% ethanol at -20˚C until processing. The samples were included according to patient enrollment consecutively and are representative of the period between 2010 and 2018.
Ninety-seven samples (53 skin biopsies from leprosy patients and 44 skin biopsies from patients with other skin diseases) were used for qPCR tests. Clinical and demographic characteristics of all patients are shown in Table 1.
Leprosy patients were defined according to the clinical, bacteriological, and histopathological Ridley-Jopling (R&J) classification and the operational classification in multibacillary (MB) or paucibacillary (PB) forms according to the WHO [31]. Leprosy or other dermatological diseases (ODD) patients were treated according to their respective condition. Leprosy paucibacillary (PB) or multibacillary (MB) patients were treated according to the Ministry of Health recommendations, while ODD patients were treated accordingly for each specific disease.

Replication Study
To validate the conditions and analysis parameters established with the clinical samples from Oswaldo Cruz's Leprosy Clinic, we tested a distinct collection of 50 skin biopsy samples that were also obtained at the Leprosy Clinic. The second set of samples was sent to the Global Health Institute, École Polytechnique Fédérale de Lausanne, Switzerland, where DNA samples were extracted and characterized by conventional PCR according to a previously published protocol [32]. Then, purified DNA was sent back to the Leprosy Clinic at Oswaldo Cruz Foundation, where it was blindly analyzed with the qPCR developed in the present study. After analysis, blinding was removed and the results were compared. Of these 50 samples, fifteen were from patients with other skin diseases, 27 patients had MB leprosy and eight from PB leprosy. The group presented a 1.27:1 ratio of males to females. The mean age was 44.8 (± 17.72 SD), and the range was 8-77. Details on the clinical characteristics are shown in S1 Table. Mycobacterial isolates samples

Synthetic DNA
The synthetic DNA (gBlock) was purchased from Integrated DNA Technologies (IDT) and consists of a double-stranded DNA containing the sequences of the three genomic targets

DNA extraction
DNA extraction from the biopsies was carried out using the DNeasy Blood and Tissue extraction kit (Qiagen, Germany). The total extracted DNA was quantified with NanoDrop (Thermo-Fisher Scientific, Waltham, MA, USA) and stored at -20˚C. M. leprae DNA from nude mice footpad was purified using TRIzol reagent (Life Technologies, Carlsbad, California) following the manufacturers' instructions, as previously described [3]. DNA used in the replication study were extracted using QIAmp UCP Pathogen Mini kit (Qiagen GmbH, Hilden, Germany).

Standard curve and 95% limit of detection (LoD 95% ) assessment
The standard curve was used for determination of the limit of detection and assay stability. A series of 10-fold dilutions was prepared from either M. leprae or synthetic DNA, using DNA purified from human blood obtained from healthy donors as a sample matrix. The dilution series used for the standard curve and the LoD 95% determination spans concentrations from 0.5 fg/reaction to 5 ng/reaction of purified M. leprae DNA (approximately 1.4 × 10 -1 to 1.4 × 10 6 genome-equivalents/reaction, considering the M. leprae genome size to be 3.3 Mbp), and 1.83 × 10 0 to 1.83 × 10 7 copies/reaction (equivalent to 0.5 ag/reaction and 5 pg/reaction, respectively) of synthetic DNA.

Quantitative PCR (real-time PCR assays)
A multiplex real-time qPCR assay targeting simultaneously two M. leprae regions and an internal reference human sequence was developed. The primers and hydrolysis probes were designed to detect regions from RLEP and 16S rRNA genes [29] from M. leprae, and the human 18S rRNA [30] ( Table 2). According to BLAST and in silico PCRs, the RLEP primers are specific to M. leprae without any complete (whole oligonucleotide) identical matching to M. lepromatosis or any other mycobacteria. However, 16S rRNA primers completely match M. lepromatosis genome sequences, except for a single base mismatch in the probe sequence. Reactions were performed on an ABI7500 Standard instrument (Thermo-Fisher Scientific, Waltham, MA, USA), using Multiplex PCR Mastermix (IBMP/FIOCRUZ PR, Curitiba, Brazil). For each reaction, 5 μL of DNA solution was added for a 25 μL final volume. Reaction mixtures were prepared in triplicates and amplified at 95˚C for 10 min, and 45 cycles of 95˚C for 15 sec and 60˚C for 1 min. All reactions included a positive control (mouse foot-pad M.  PLOS NEGLECTED TROPICAL DISEASES leprae DNA and/or high-bacterial load lepromatous leprosy patient purified DNA), and water as a non-template control (NTC; PCR reaction without any template DNA).
All reagents (oligomix 25X and qPCR mix) were maintained in independent aliquots at -20˚C at the Leprosy Laboratory (FIOCRUZ-RJ). Tests with the dilution series described above were repeated weekly for the first month, and then once a month for five months.

Data analyses and statistics
Qualitative (diagnostic sensitivity and specificity, accuracy) and quantitative (intra-and interlaboratory repeatability and reproducibility, analytical sensitivity and specificity) validation tests were performed. The 95% limit of detection (LoD 95% ) was calculated by fitting a Probit model to the estimated detection probabilities. Samples that fall under the "equivocal" category were considered negative when calculating the diagnostic sensitivity, specificity and the predictive values. These parameters were calculated according to Altman and Bland [34]. Data were processed and analyzed using customized scripts for R version 3.5.1 (downloaded from http://www.Rproject.org/).

Analytical performance
Primers and hydrolysis probes designed to target 16S rRNA and RLEP sequences of M. leprae were tested in multiplexed reactions to concomitantly detect the human 18S rRNA sequence. Optimal fluorescence thresholds were chosen based on the common practice that it should be positioned on the lower half of the fluorescence accumulation curves plot from the 10-fold dilutions, crossing most if not all fluorescence signals on the exponential segment of the curve on a logarithmic scale (Fig 1). Therefore, after setting the baseline to the automatic function, fluorescence threshold values chosen for determining C p (Crossing point) values for each target were set to intercept the positive controls and avoid the negative ones, being established as follows: 0.2 for RLEP, 0.15 for 16S rRNA, and 0.16 for 18S rRNA.
The analytical 95% limit of detection (LoD 95% ) was determined from a series of tests in which DNA extracted from M. leprae was diluted from 5 ng to 100 ag/reaction. The developed multiplex reaction was evaluated against a collection of microorganisms to assess the specificity of the primers and probes under these conditions. The selection included several mycobacteria, as well as a few other pathogens associated with skin diseases such Leishmaniasis (Fig 3). We only considered any species as cross-reactive if all the technical replicates displayed amplification for at least one of the targets, which was not the case for any of the species tested. Most positive amplifications observed correspond to RLEP, which was detected in two out of three replicates in M. fortuitum and M. kyroniense. Even though some reactions presented 16S rRNA signals above the threshold, these amplifications are very uncharacteristic and are easily distinguishable from a proper amplification when compared with the positive control with 500 fg/reaction of M. leprae.

Repeatability and reproducibility
Three independent operators performed three replicate runs each, in consecutive days, and evaluated the repeatability and reproducibility of the multiplex reactions. For each replicate, a new dilution series for the synthetic gene was prepared from a concentrated aliquot to be used as a template. The data (S2 Table) shows that all intra-operator replicates were remarkably reproducible, with only one point (Op. 1, 16S rRNA 1.83 × 10 2 ) displaying a relative standard deviation (rRSD%) above 5%, but still well below 10%. The inter-operator variability was also very low, and the largest variation was observed for the 16S rRNA target. Nonetheless, the rRSD% was between 1.38 and 11.57 across the dilution range, which shows an excellent reproducibility for a quantitative test (see also S3 Table).
The accuracy of the determinations performed by the multiplex real-time qPCR assay was also estimated using the synthetic DNA. To evaluate the intra-and inter-repeatability (or intermediate precision) for operators, we calculated the arithmetic mean, standard deviation, and relative standard deviation percentage of three independent experiments. It is noteworthy that the detection of the human target 18S rRNA does follow the same dilution trend for the other targets because the synthetic template was not diluted in human DNA.
In summary, for both M. leprae targets we observed that all points showed excellent reproducibility and repeatability. As expected, detection of the human target 18S rRNA loses  Fig 2. Analytical 95% limit of detection (LoD 95% ) for 16S rRNA and RLEP in multiplexed qPCR. Mycobacterium leprae DNA was diluted in DNA extracted from whole blood from healthy donors and tested from 5 ng to 0.5 fg/ reaction. Probability of detection was calculated for 16S rRNA and RLEP (top and bottom panels, respectively) from nine independent experiments, and a Probit model was fit to the data (black lines). The gray ribbon around the model fit indicates the 95% CI on the predicted probability. Dotted lines indicate the interpolation to determine the concentration at a 95% probability. The calculated LoD 95% is displayed on each plot in femtograms of DNA/reaction. https://doi.org/10.1371/journal.pntd.0009850.g002

PLOS NEGLECTED TROPICAL DISEASES
reproducibility as it becomes scarce in the reaction due to the dilution factor. It is noteworthy that there is no variation in the detection of the human target 18S rRNA when M. leprae DNA was present in the synthetic control molecule, i.e., in a 1:1 ratio, supporting the notion that the multiplexed reactions do not interfere with each other.

Stability
Storage stability was assessed by performing monthly evaluations of reactions with different concentrations of the synthetic DNA molecule for 5 months. Most of the data points tested varied below the established limit of three standard deviations above the average of all time points. Fig 4 shows the C p obtained for the three evaluated targets (16S rRNA, RLEP, and 18S rRNA) in representative concentrations for brevity, over a 5-month period. The test remained reliable for the entire range of concentrations tested.

Diagnostic performance
The implemented setup involved the interrogation of two target sequences from M. leprae to classify clinical samples correctly while mitigating possible false positives. To evaluate the diagnostic performance, we first established optimal parameters for the analysis, considering possible cross-reactions that may occur in the laboratory routine. The C p cutoff for both targets were determined iteratively by analyzing the receiver operating characteristic (ROC) curves for each combination of cutoff cycles. S1  Table. Based on these results the best combination of cutoff values (35.5 for 16S rRNA and 34.5 for RLEP) showed a sensitivity of 91% and specificity of 100%, positive and negative predictive values (PPV and NPV) were 100% and 90%, respectively. These values were similar between MB and PB samples, as shown in the S5 Table. These parameters for analysis and summarized them in the decision algorithm presented in Table 3.
Next, the molecular diagnosis obtained using the new multiplex PCR, was compared to the clinical diagnosis of each sample (Fig 5 and S6 Table). Results show that the qPCR reaction and classification algorithm correctly characterized 48 of the 53 samples previously described as "Leprosy" by the clinical outcome. Of the 5 misclassified samples, one was classified as negative for M. leprae and four were in the "equivocal" quadrant. All the misclassified samples have a Bacterial Index of 0.
None of the 44 samples characterized as "Other skin diseases" were classified as M. lepraepositive by our reaction and decision algorithm. Thirty-eight of these samples were classified as "Negative" and 6 as "equivocal".

Assay validation
Conditions established with the training cohort were tested on an independent set of samples, which were previously characterized using a distinct qPCR method described in Girma et al. [32]. The comparison between the original classification and the new results is shown in Fig 6 and S1 Table. We tested 50 samples, of which 34 were previously characterized as positive and 16 as negative. The 50 samples were classified according to our algorithm, resulting in 33 correctly classified as positive and 11 correctly classified as negative. Of the four samples classified as equivocal, two were negative for the reference method and one was positive. The sensitivity, specificity and accuracy calculated for this sample set were 97.1%, 100% and 98%, respectively.
Taken together, these results demonstrate that the multiplexed reaction for M. leprae RLEP and 16S rRNA is able to classify samples precisely, by combining the strengths of each molecular target and improving on their use in isolation. This setup is also able to flag samples with low bacilli burden for further monitoring, avoiding unnecessary treatment of uninfected patients and proper follow up of M. leprae carriers.

Discussion
Leprosy is a chronic infectious disease with a wide range of clinical forms, each distinguished by immunological and histopathological features. Leprosy can be tuberculoid, which is a localized form exhibiting few or no bacteria, or lepromatous, which is a systemic form with high loads of mycobacteria. Among the tuberculoid patients, there is a range of skin granulomatous diseases phenotypically comparable to leprosy [5].

PLOS NEGLECTED TROPICAL DISEASES
The use of PCR for leprosy diagnosis has been extensively tested [4,16,[35][36][37][38][39][40][41][42][43][44][45][46]. However, limitations towards the experimental designs for some published studies were identified. We observed that most studies: (i) test only samples from leprosy patients, creating difficulties in determining some diagnostic parameters such as specificity; (ii) were performed on small sample sizes; and (iii) do not have independent validation on the same assay or an evaluation of the same protocol in different centers. Furthermore, no studies have used reagents produced under good manufacturing practices (GMP), a set of guidelines that allow for traceability and batch-to-batch reproducibility of characteristics such as physical parameters and performance of the reagents [47].
In this study, we solved some of these issues by (i) developing and validating an assay based on the two most tested targets in the literature with better accuracy so far [7,8,43,48] (ii) following guidelines for validation of diagnostic tests [47,49,50], and (iii) using GMP grade reagents. We were also able to include a reaction for the detection of human 18S rRNA gene in the sample, to assess the quality of DNA extraction and reagent performance in the same reaction as the M. leprae determination occurs.
RLEP and 16S rRNA are the most frequent markers used in leprosy studies, displaying PCR sensitivity values up to 80% for each target. However, it is important to note that the sensitivity  of targets varied between sample types, clinical settings, and also between studies of the same authors [8,9]. Tatipally et al. [9] showed that using more than one marker in a multiplex format of conventional endpoint PCR yields significantly higher PCR positivity.
In the current study, a multiplex qPCR assay simultaneously amplifies two specific M. leprae targets (16S rRNA and RLEP), and the mammalian 18S rRNA gene as internal reaction control. The assay validation comprised analytical performance, diagnostic sensitivity and specificity, as well as reproducibility and repeatability. Development of multiplex qPCR assays provides a greater challenge than designing singleplex assays because it often requires extensive optimization as primer-dimers and non-specific interactions may interfere with amplification of the desired targets. Additionally, it is important that the amplification of two or more targets does not preferentially amplify one of the targets [51,52]. Combining multiple primers and probes did not affect the efficiency of the triplex qPCR in comparison to the corresponding singleplex reactions used in Martinez et al. [16]). They evaluated the independent detection of 16S rRNA and RLEP using the same primers and probes and obtained 0.51 and 0.91 for sensitivity and 0.73 and 1 for specificity, respectively. Barbieri et al. [4] also used the same 16S rRNA target to evaluate paucibacillary leprosy samples and obtained 0.57 for sensitivity and 0.91 for specificity. Here, we evaluated a panel with 53 leprosy and 44 non-leprosy patient samples, and later a different sample panel (50 patient samples) and achieved high sensitivity (> 90%) and specificity (100%) for both panels tested.
However, we understand that the small number of paucibacillary (PB) individuals in our study is a limitation. In fact, the greatest importance of using qPCR as a complementary diagnosis is precisely for PB samples. Generally, PB patients exhibit low (or zero) bacterial index and the histopathology examination does not distinguish them from other dermatoses. Therefore, these are the cases where clinical evaluation alone might not be able to determine the diagnosis, and where a qPCR confirmation becomes more important. However, due to the scarcity of bacterial DNA in these samples, it is known that the detection of M. leprae in PB patients by real-time PCR is difficult [4].
The reactions we developed in this study predict the equivocal classification of early-stage infections based on the finding from Martinez et al. [16], who showed that RLEP displays higher sensitivity than 16S rRNA whereas the ribosomal gene displays higher specificity. Thus, samples lacking 16S rRNA amplification but with RLEP amplification with a C p lower than the threshold are suggested to be re-analyzed.
In general, our data ( Fig 5) show a correlation between BI and C p values. Biopsies from patients with higher BI values were deemed positive for bacteria earlier in the amplification cycle, as seen by the lower Cp values and high copy numbers of bacilli.
The "analytical sensitivity" or "limit of detection" of an assay is defined as the ability of the assay to detect very low concentrations of a given substance in a biological specimen [47]. The result of the limit of detection (LoD 95% ) determination when tested on a purified M. leprae sample indicated a higher sensitivity for RLEP (4.6 fg of DNA/reaction, equivalent to approximately 1.3 M. leprae genomes) versus 16S rRNA (450 fg of DNA/reaction, approximately 126 M. leprae genomes). This difference in sensitivity was expected since the 16S rRNA is a single copy gene [29] and the RLEP presents an average of 36 copies per genome [26].
Applicability in a reference laboratory setting was also considered during the development of these reactions. Intra and inter-operator variability were low, ensuring consistent results in routine testing (S3 Table). Moreover, reagents remained stable for at least five months, allowing for adequate stock maintenance (Fig 4).
Leprosy is a silent disease with a very long incubation time. Currently, transmission can only be halted if patients obtain early diagnosis. High-risk individuals, which are the patients' close contacts, should be traced and treated whenever leprosy is detected. Recently, it has been suggested that novel policies towards this group of contacts such as immuno-and chemoprophylaxis are effective to help control the disease burden [15,53,54]. These approaches provide a screening of the high-risk population that, coupled with a pharmacological or immunological intervention, has been suggested to decrease disease incidence.
In some situations, clinical diagnosis needs the accuracy of a laboratory analysis, and qPCR is a reliable technique to enable diagnostic confirmation [10]. Indeed, we confirmed that the availability of molecular tests can be very helpful in diagnosing patients during contact monitoring [55]. When contacts present a leprosy-like lesion, a positive PCR has resulted in a leprosy diagnosis with 50% sensitivity and 94% specificity [55]. Other indirect methods based on simultaneous detection of host humoral as well as cellular immune responses directed against the bacteria are also promising new diagnostic tools. Recently, lateral flow assays (LFA), combining detection of mycobacterial components and host proteins, proved to be specific and sensitive [56][57][58][59][60][61][62]. The signature detected by this platform identified 86% of the leprosy patients, with a specificity of 90% (AUC: 0.93, p < 0.0001) [60]. Thus, a multicentric study comparing different available methods such as qPCR and LFA is still necessary. It is noteworthy that our data showed accuracy, sensitivity, and specificity values quite similar to LFA.
We believe that the diagnosis of tropical and neglected diseases needs molecular-based methods such as PCR, especially due to the robustness and capillarity of the technique in clinical analysis laboratories worldwide. Towards that future, we present a real-time quantitative PCR produced with GMP reagents that adheres to all quality control specifications, allowing batch-to-batch performance reproducibility and repeatability, and that can be used in research and clinical laboratories with reasonable infrastructure in endemic countries. Finally, we envision the multiplex qPCR assay developed adapted to more affordable, rapid, point-of-care tests to be used in low-resourced settings, enabling on-site early and specific diagnosis of leprosy, hopefully helping disease control.