Utility of circulating cell-free Mycobacterium tuberculosis DNA for the improved diagnosis of abdominal tuberculosis

Abdominal tuberculosis (ATB) continues to pose a major diagnostic challenge for clinicians due to its nonspecific clinical presentation, variable anatomical location and lack of sensitive diagnostic tools. In spite of the development of several assays till date; no single test has proved to be adequate for ATB diagnosis. In this study, we for the first time report the detection of circulating cell-free Mycobacterium tuberculosis (M. tuberculosis) DNA (cfMTB-DNA) in ascitic fluid (AF) samples and its utility in ATB diagnosis. Sixty-five AF samples were included in the study and processed for liquid culture, cytological, biochemical and molecular assays. A composite reference standard (CRS) was formulated to categorize the patients into ‘Definite ATB’ (M. tuberculosis culture positive, n = 2), ‘Probable ATB’ (n = 16), ‘Possible ATB’ (n = 13) and ‘Non-TB’ category (n = 34). Two molecular assays were performed, namely, the novel cfMTB-DNA qPCR assay targeting M. tuberculosis devR gene and Xpert MTB/RIF assay (Xpert), and their diagnostic accuracy was assessed using CRS as reference standard. Clinical features such as fever, loss of weight, abdominal distension and positive Mantoux were found to be strongly associated with ATB disease (p<0.05). cfMTB-DNA qPCR had a sensitivity of 66.7% (95% CI:40.9,86.7) with 97.1% specificity (95% CI:84.7,99.9) in ‘Definite ATB’ and ‘Probable ATB’ group collectively. The sensitivity increased to 70.9% (95% CI:51.9,85.8) in the combined ‘Definite’, ‘Probable’ and ‘Possible’ ATB group with similar specificity. The cfMTB-DNA qPCR assay performed significantly better than the Xpert assay which demonstrated a poor sensitivity of ≤16.7% with 100% (95% CI:89.7,100) specificity (p<0.001). We conclude that cfMTB-DNA qPCR assay is an accurate molecular test that can provide direct evidence of M. tuberculosis etiology and has promise to pave the way for improving ATB diagnosis.


Introduction
Tuberculosis (TB) remains a significant global health challenge with an estimate of 10 million people developing the disease globally in 2018 [1].India alone accounts for 26.9% of this vast global TB burden [1].Extra-pulmonary TB (EPTB) constitutes ~15-20% of all TB cases in India [2].Abdominal TB (ATB) is one of the manifestations of EPTB that contributes to 3% of all EPTB cases in India and may involve peritoneal, intestinal, upper gastrointestinal (oesophageal and gastro-duodenal), hepato-biliary, pancreatic and peri-anal sites [3].ATB remains a major diagnostic challenge owing to the lack of specific clinical features and limited yield of the commonly used diagnostic tests [4].ATB diagnosis is usually made through a combination of clinical, radiological, microbiological, histological, biochemical and molecular techniques [5].Conventional microbiological techniques like AFB smear and culture suffer either from low sensitivity (0%-40%) or high turnaround time (culture) which renders them inadequate for appropriate management of the disease [4,6].Histopathology demonstrates tuberculous granulomas; however, it alone may not be pathognomonic and requires culture confirmation [7].Biochemical parameters including ascitic fluid (AF) total protein (>2.5 g/dL), serum-ascites albumin gradient (SAAG) (<1.1 g/dL) and elevated adenosine deaminase (ADA) values are routinely used for the clinical diagnosis of ATB [4].However, these parameters have a reduced sensitivity in confounding cases of ATB with co-morbidities like liver cirrhosis and chronic liver disease [4,[8][9][10][11].Laparoscopy is the diagnostic tool of choice for ATB and allows inspection of the peritoneum along with providing an option for obtaining biopsies for histology.However, laparoscopic features of ATB may mimic other diseases like sarcoidosis, peritoneal carcinomatosis and Crohn's disease that necessitates the need of taking biopsies during laparoscopy [4].Moreover, the procedure (for biopsy) is invasive, expensive and requires competent surgeons and hospitalization [4].Other tests for ATB diagnosis include immune response-based assays such as the Mantoux test and the IFN-γ release assays; but these tests are usually non-specific and are unable to discriminate between active and latent TB infection [7].Also, existing commercial serological assays are inconsistent, imprecise and have received a strong negative recommendation by WHO [12].
Various nucleic acid amplification tests (NAATs) including in-house PCR assays have been developed for ATB diagnosis.These assays have targeted different regions of M. tuberculosis in multiplex (IS6110, devR and 16S rRNA) or single-plex formats (IS6110, 38-kDa protein gene) with a variable sensitivity range of 35%-93% [13][14][15][16].This inconsistent performance of inhouse NAATs limits their usefulness for ATB diagnosis.Other NAATs include the Xpert MTB/RIF assay (Xpert) and its advanced version Xpert Ultra, which are WHO-endorsed, automated commercial NAATs, and have proved to be critical tools for the diagnosis of pulmonary TB, extra-pulmonary TB and for children with TB [17,18].A small number of studies have reported variable sensitivity (range: 18.3%-100%) for Xpert in diagnosing ATB, with an overall high specificity (range-90%-100%), limiting its role as a rule-in test [19][20][21][22].
Circulating cell-free DNA (cfDNA) is defined as extracellular DNA fragments circulating mainly in various body fluids such as blood, saliva, urine, stool, sputum, various body fluids (cerebrospinal fluid, peritoneal fluid, synovial fluid, lymph and amniotic fluid) and tissues [23].cfDNA-based tests have been successfully used in detecting cancer, parasitic infections, pathogens like fungi, bacteria, Epstein-Barr virus for nasopharyngeal carcinoma screening and pre-natal diagnosis [23][24][25][26][27][28].The presence of cfDNA in peritoneal fluid has been used in diagnosis and prognosis of therapy-related peritoneal degeneration [29].Detection of circulating cell-free M. tuberculosis DNA fragments in body fluids has also recently emerged as a novel tool for the diagnosis of TB disease [30][31][32][33][34][35][36].Circulating cell-free M. tuberculosis DNA (referred as cfMTB-DNA from here onwards) refers to the fragments of cell-free M. tuberculosis DNA that may have been released from dying human cells and mycobacteria in biological fluids [33].cfMTB-DNA detection has shown a sensitivity ranging from 50%-79.5% by PCR for pleural TB and 56.5-100% for tuberculous meningitis diagnosis with � 92% specificity in various studies [30][31][32][34][35][36][37].To the best of our belief, this is the first study to evaluate the utility of cfMTB-DNA detection in ascitic fluid samples for ATB diagnosis.The objectives of the present study were i) to develop a circulating cell-free M. tuberculosis DNA probe-based quantitative real-time PCR (qPCR) assay targeting devR gene of M. tuberculosis (cfMTB-DNA qPCR assay) and ii) to compare the performance of the developed assay with the Xpert test for ATB diagnosis.

Ethics statement
The ethical approval for this study was obtained from the Institute Ethical Committee of Translational Health Science and Technology Institute (THSTI), Faridabad [THS 1.

Study participants
Patients presenting with ascites (n = 95) visiting the Out-patient Department (OPD) of RMLH were enrolled in this study in a prospective manner.Ascitic fluid samples were collected only after taking written informed consent from patients.No minor cases were included in the study.The samples were collected over a period of 3 years in two phases; March 2014-August 2015 and February 2016-November 2017.Detailed clinical history, radiological findings, biochemical and haematological laboratory investigation results were recorded in a case record form as approved by the Ethics Committee (S1 Appendix).Patients with a past history of tuberculosis (pulmonary TB or extrapulmonary TB) in preceding 2 years or on anti-tubercular therapy (ATT) for > 7 days were excluded.Finally, ascitic fluid samples from 65 patients (age range of 15 to 70 years, 66.2% males and 33.8% females) were included in the study (Fig 1).Patient information was maintained in a classified manner.
The rate of isolation of M. tuberculosis is very low from ascitic fluid [3], thus a composite reference standard (CRS) was used in the study.The CRS was broadly based on published guidelines and previous categorizations for ATB and comprised of clinical, radiological, microbiological and biochemical findings associated with ATB patients and response to ATT (Fig 2) [3,22,38].Based on the CRS, the patients were categorised into 'Definite' ATB, 'Probable' ATB, 'Possible' ATB and 'Non-TB' groups (Fig 2).The 'Definite' ATB category included patients who were M. tuberculosis culture-positive (n = 2) and 'Probable' ATB category (n = 16) included patients who were clinically diagnosed as ATB and responded to anti tubercular therapy [3,38] (Fig 2).'Possible' ATB category (n = 13) included patients who presented with chronic symptoms of ATB with either radiology suggestive of ATB (palpable abdominal mass or omental thickening) or an abnormal chest X-ray (consolidation, calcification or pleural effusion) or presence of other features as described in Fig 2. [3,4,22,38].However, in this group, the information about response to ATT was not available because either the patient got lost to follow up (n = 11) or left against medical advice (n = 2).Since, the 'Possible' ATB category did not fulfil all pre-requisites for ATB classification, we performed a stratified analysis: i) one, including ascitic fluid samples from both 'Definite' ATB and 'Probable' ATB groups (referred as 'Definite and Probable' ATB group) and ii) two, including all three categories collectively i.e., 'Definite' ATB, 'Probable' ATB and 'Possible' ATB' groups (referred as 'ATB group').

Xpert MTB/RIF assay
The Xpert assay was carried out according to manufacturer's instructions.Briefly, 0.8 ml of ascitic fluid was mixed with sample reagent provided in the assay kit in a ratio of 1:2 and incubated for 15 minutes. 2 ml of the mixture was transferred into the cartridge and inserted in the Xpert instrument for analysis.

Statistical analysis
The results of all the tests were compiled and analysed after completion of the study in a blinded manner.The performance of cfMTB-DNA qPCR assay and Xpert was evaluated using CRS as the reference standard.DNA amounts in 'Definite and Probable' ATB group (true positives) and 'Non-TB' group (true negatives) were used to generate receiver operating characteristics (ROC) curves using GraphPad Prism version 5.00 for Windows (GraphPad Software, La Jolla California USA, www.graphpad.com).The cut-off points were derived to achieve test specificity (~97%) consistent with the 'target product profile' (TPP) defined for newer diagnostic tests for extrapulmonary tuberculosis [39].The difference between cfMTB-DNA quantified in the 'Possible' ATB group versus other groups ('Definite' ATB and 'Probable' ATB) was calculated by unpaired t-test (2-tailed) using GraphPad Software.The diagnostic performance parameters for qPCR and Xpert were calculated using web resources (https://www.medcalc.org/calc/diagnostic_test.php).The difference in sensitivity of qPCR and Xpert was determined by McNemar's test (http://epitools.ausvet.com.au/comparetwotests).The statistical significance of the qualitative and quantitative variables between the 'ATB' group and 'Non-TB' group was performed by Fisher's exact test and Mann Whitney tests respectively (the significance was set at a p-value of < 0.05) using GraphPad Software.

Comparative evaluation of Xpert vs. cfMTB-DNA qPCR assay
The sensitivity of detection of circulating cell-free M. tuberculosis DNA by cfMTB-DNA qPCR assay was 66.7% as compared to 16.7% by Xpert in the 'Definite and Probable' ATB combined group (Table 3).The difference was further enhanced when the 'Possible' ATB group was included in the analysis (70.9% vs. 9.7%).The sensitivity increment of cfMTB-DNA qPCR over Xpert was statistically significant (p<0.001).Combining the results of Xpert and cfMTB-DNA qPCR yielded a sensitivity of 72.2% (95% CI: 46.5,90.3) in 'Definite and Probable' ATB category with 97.1% (95% CI: 84.7,99.9)specificity.The sensitivity was further enhanced to 74.2% (95% CI: 55.4,88.1) in the combined 'ATB' group with similar specificity (Table 3).

Discussion
ATB is a clinical challenge because of its non-specific clinical presentation and lack of accurate diagnostic tools [4,41].In spite of the availability of several diagnostic platforms; no test has proved to be solely adequate for diagnosing ATB [4,19,42].The detection of circulating cellfree M. tuberculosis DNA (cfMTB-DNA) has recently emerged as a tool for diagnosing paucibacillary forms of TB such as tuberculous meningitis and pleural TB [30][31][32][34][35][36][37].In the present study, to the best of our knowledge, we for the first time assessed the utility of cfMTB-DNA for ATB diagnosis.
The developed cfMTB-DNA devR-based qPCR assay demonstrated a sensitivity of 66.7% (95% CI:40.9,86.7)collectively in the 'Definite and Probable' ATB group and 70.9% (95% CI: 51.9,85.8) in the combined 'ATB' group ('Definite', 'Probable' and 'Possible' ATB).Nine (9/31) samples were false negative by cfMTB-DNA qPCR assay in the combined 'ATB group' which included one (1/2) from 'Definite' ATB group, five (5/16) from 'Probable' ATB group and three (3/13) from the 'Possible' ATB group.The false negative qPCR results might be attributed to the varying amounts of cfMTB-DNA available in ascitic fluid for amplification.Although, a few qPCR results were positive with very low volumes of ascitic fluid (~41% positive ascitic fluid samples had a volume of <300 μl), we believe that the false negative qPCR results might be reduced by using a larger volume of ascitic fluid for cfMTB-DNA extraction.The volume range of ascitic fluid used in this study was in the range of 200 μl to 1 ml (depending on sample volume availability), and we believe that an increase in sample volume can ensure optimum amount of cfDNA extraction which will lead to fewer false negative results [43].An improved sensitivity was obtained for urinary EGFR cfDNA detection in NSCLC patients when ~90-100 ml of urine was used for analysis [44].The specificity for the cfMTB-DNA qPCR assay was 97.1% (95% CI: 84.7,99.9) in both 'Definite and Probable' ATB group and combined 'ATB' group.One sample (1/34) was false positive in this study.The diagnosis of this patient was chronic liver disease (CLD) and it might be possible that the false positive qPCR result was due to an underlying ATB disease that was masked by CLD; as it has been reported that in countries with high TB prevalence there is a high possibility of developing ATB disease in CLD patients [45,46].
The origin of cfDNA in the circulation is believed to be either through apoptosis, necrosis, phagocytosis, active secretion, exocytosis or breakdown of pathogens and the increase in the amount of cfDNA released has been correlated with the increase in the dead and dying cells [47][48][49][50][51].This released cfDNA is further cleared via blood, body fluids, liver, spleen, kidney or lymph nodes [52][53][54] by immune system mainly by macrophages [55,56].In healthy individuals, the cfDNA and apoptotic cells are rapidly cleared; thus leading to a reduced level of cfDNA in circulation [47,52]; however, during chronic inflammation, excessive cell death and pathological conditions, there is an insufficient clearance of cfDNA leading to its accumulation in circulation [47,52].This insufficient clearance might explain the accumulation of cfDNA in pathological conditions (here abdominal TB) [47,52].The actual levels of cfDNA in circulation relies upon the balance between the DNA release and DNA clearance processes and this inter-play between them marks the actual levels of cfDNA in the circulation [52].It was observed in the present study that the mean DNA amount quantified by the cfMTB-DNA qPCR assay in ascitic fluid was higher in the 'Possible' ATB group (Mean 7492 fg/ml ± 4545) followed by 'Probable' ATB group (Mean 1754 fg/ml ± 1068) and 'Definite' ATB group (Mean 468.1 fg/ml ± 417.9).The difference between circulating cfMTB-DNA amount quantified in the 'Possible' ATB group versus combined group of 'Definite and Probable' ATB was highly significant (p < 0.0001).This observation proves a reverse relationship between the presence of number of intact M. tuberculosis bacilli and circulating cfMTB-DNA, i.e. greater is the number of intact bacilli (which is expected in 'Definite' group), lower is the amount of circulating cfMTB-DNA.This significant difference between the mean DNA amounts between different categories might be because of different cellular fate (death modality) of M. tuberculosis infected macrophages which influences the infection outcome [57].During apoptosis or necrosis of M. tuberculosis infected macrophages, the viability of M. tuberculosis bacilli is reduced whereas cfMTB-DNA release is increased in the circulation [49,57] which might be the reason for increased cfMTB-DNA in the 'Probable and Possible ATB' categories.However, in the 'Definite ATB' group the reduced amount of cfMTB-DNA might be due to the inhibition of apoptosis of M. tuberculosis infected macrophages as one of the virulence mechanisms against host defence which allows M. tuberculosis to replicate [57], thereby leading to more number of intact bacilli and not as much amount of cfMTB-DNA.
Xpert exhibited a modest sensitivity of 16.7% and 9.7% in the combined group of 'Definite and Probable' ATB and combined 'ATB' ('Definite', 'Probable' and 'Possible') group, respectively in the present study.One study reported a similar sensitivity of ~18.3% for Xpert against CRS [21]; however, a higher sensitivity of 32% and 50% has also been reported [22,58].In microbiologically confirmed ATB cases, a comparatively higher sensitivity has been reported for Xpert.A pooled sensitivity of 59% (Range:33%-100%) has been recently described by WHO which was similar to the sensitivity of Xpert in definite ATB cases in our study (50%), albeit we had only 2 cases [20].We believe that since Xpert is based on the principle of detecting 'genomic DNA' from intact M. tuberculosis bacilli (which is expected to be higher in microbiologically confirmed ATB cases), the sensitivity of Xpert is higher in 'Definite' ATB samples.In CRS confirmed cases, intact M. tuberculosis bacilli are expected to be fewer, leading to a higher sensitivity of cfMTB-DNA qPCR assay vs. Xpert.However, the excellent specificity (100%) of Xpert provided credence to a sample being not labelled as false positive and indicated its utility as a rule in test [59].
The major strength of our study was the development of a cfMTB-DNA based assay and establishing its utility in ATB diagnosis.cfDNA is an attractive biomarker for TB diagnosis that indicates the presence of pathogen in the system [33].This approach enabled utilization of both the sediment and supernatant fractions of ascitic fluid for M. tuberculosis culture (reference standard) and circulating cell-free DNA based tests, respectively.The limitations of this study were small sample size, limited sample volumes and identification of only two 'Definite' ATB patients.Overall, cfMTB-DNA qPCR assay showed a better diagnostic accuracy for ATB as compared to culture and Xpert, which indicates its diagnostic utility in ATB diagnosis.However, the assay missed one culture positive and one Xpert-positive patient which might be because of the inverse relationship between the disease category and the amount of genomic DNA.Therefore, we believe that tests based on both genomic DNA and circulating DNA should be performed on ATB samples, as the combinatorial outcome of both these formats will enable us to diagnose both Definite and Probable/Possible forms of ATB.By combining our novel cfMTB-DNA qPCR assay with Xpert, the overall diagnostic sensitivity in 'Definite and Probable' ATB group increased to 72.2%, with no decrease in the specificity.When test results of 'Possible' ATB category were also included, the combined sensitivity of qPCR and Xpert increased to 74.2%.The 'Possible' ATB category was formulated specifically for this study, we believe that the inclusion of this category and the detection of subjects belonging to it highlights the diagnostic challenges that plague ATB.These patients possibly had TB which was missed due to lack of optimal diagnostic tools [46].
The detection of cfMTB-DNA in the ascitic fluid of ATB patients has proved to be a promising diagnostic tool.We conclude that the developed cfMTB-DNA qPCR assay has an immense potential for the diagnosis of ATB and can also be adapted to other paucibacillary forms of EPTB.Nevertheless, for the application of this assay in the clinic, we will require its validation on a larger number of samples.
Ascitic fluid samples of all patients were subjected to biochemical and cytological investigations at RMLH, New Delhi.The left over ascitic fluid was transferred to the AIIMS laboratory and processed for liquid culture and Xpert on the same day.Briefly, liquid culture for M. tuberculosis was performed from the sediment obtained after centrifugation of ascitic fluid in Middlebrook 7H9 broth enriched with oleic acid, albumin, dextrose, catalase (OADC) and an antibiotic cocktail; Polymyxin B, Amphotericin B, Nalidixic acid, Trimethoprim and Azlocillin (PANTA) [Becton and Dickinson (BD), Franklin Lakes, New Jersey, United States] (Fig 1).The supernatant was stored at -80˚C and used later for cfMTB-DNA extraction (Fig 1).Quantification of cfMTB-DNA was performed targeting the devR region of M. tuberculosis by qPCR.

Table 1 . Characteristics and clinical features of the study participants # . Clinical features
[4,40] denoted as number of patients showing clinical signs in each category; values in brackets denotes the percentage of patients presenting the clinical features.ATB in other studies[4,40].Nineteen percent (6/31) of the ATB category patients had a co-existing pulmonary involvement of TB as evident by an abnormal X-ray finding [cavitation, mediastinal lymphadenopathy, pleural effusion and consolidation].Six percent (2/31) of the ATB patients had a past history of TB disease and ~29% (9/31) had a history of contact with a TB patient.In the ATB category, 26% (8/31) of the patients showed hepatomegaly, followed by 'Definite' ATB (n = 2) 'Probable' ATB (n = 16) 'Possible' ATB (n = 13) 'ATB' group $ (n = 31) 'Non-TB' group (n = 34) p-value � # $ ATB group includes 'Definite', 'Probable' and 'Possible ATB' categories collectively.^Mantoux test data was available for n = 24 patients in ATB group and n = 17 patients in non-TB group.� statistical significance calculated between 'ATB' group vs. Non-TB group by Fisher's test, significant values (p-value < 0.05) are in bold.https://doi.org/10.1371/journal.pone.0238119.t001with

outcome of cMTB-DNA devR-qPCR assay
Circulating cell-free M. tuberculosis DNA extracted from ascitic fluid samples were quantified using the devR-qPCR assay.Total DNA amounts in ascitic fluid from 'Definite and Probable ATB' group (true positives) and 'Non-TB' group (true negatives) were used to construct ROC curves and cut-off value was set at 575.7 fg/ml (equivalent to 115 M. tuberculosis bacterial genome equivalents) to attain a specificity of 97% (Fig3A).cfMTB-DNA devR-qPCR assay