Lipoarabinomannan antigenic epitope differences in tuberculosis disease subtypes

An accurate urine test for diverse populations with active tuberculosis could be transformative for preventing TB deaths. Urinary liporabinomannan (LAM) testing has been previously restricted to HIV co-infected TB patients. In this study we evaluate urinary LAM in HIV negative, pediatric and adult, pulmonary and extrapulmonary tuberculosis patients. We measured 430 microbiologically confirmed pretreatment tuberculosis patients and controls from Peru, Guinea Bissau, Venezuela, Uganda and the United States using three monoclonal antibodies, MoAb1, CS35, and A194, which recognize distinct LAM epitopes, a one-sided immunoassay, and blinded cohorts. We evaluated sources of assay variability and comorbidities (HIV and diabetes). All antibodies successfully discriminated TB positive from TB negative patients. ROAUC from the average of three antibodies’ responses was 0.90; 95% CI 0.87–0.93, 90% sensitivity, 73.5% specificity (80 pg/mL). MoAb1, recognizing the 5-methylthio-d-xylofuranose(MTX)-mannose(Man) cap epitope, performed the best, was less influenced by glycosuria and identified culture positive pediatric (N = 19) and extrapulmonary (N = 24) patients with high accuracy (ROAUC 0.87, 95% CI 0.77–0.98, 0.90 sensitivity 0.80 specificity at 80 pg/mL; ROAUC = 0.96, 95% CI 0.92–0.99, 96% sensitivity, 80% specificity at 82 pg/mL, respectively). The MoAb1 antibody, recognizing the MTX-Man cap epitope, is a novel analyte for active TB detection in pediatric and extrapulmonary disease.

Tuberculosis (TB) accounts for more than 1.7 million deaths per year 1 . While most deaths could be prevented by early diagnosis and initiation of treatment, conventional diagnostic methods requiring sputum (Xpert Ultra real-time PCR, sputum microscopy) need specialized training and costly equipment, and are currently inaccessible to many underserved populations in countries where TB is endemic [1][2][3] . Non-invasive urine immunoassay screening tests for mycobacterial antigens are an attractive alternative to sputum testing, because they can be deployed in low resource settings 4 . Measurement of the urinary TB carbohydrate antigen lipoarabinomannan has proved valuable for HIV positive patients with active pulmonary TB who have low CD4 counts 5 .
Nonetheless the vast majority of patients (85%) with active pulmonary TB are HIV negative. In a previous study we used a nanotechnology-enhanced assay to document LAM in the urine of HIV negative TB positive patients; LAM concentration ranged between 5 and 5,000 picograms/mL 6 and positively correlated with disease severity. While LAM detection in HIV negative TB patients was initially controversial, several groups have now proven that LAM can be detected if the immunoassay has sufficient sensitivity.
In order to promote LAM urinary tests for widespread population screening for active TB, a number of important challenges need to be addressed. Firstly, LAM is a complex molecule and differences in its antigenic properties in the body fluids of infected patients, and animal models, compared to LAM produced by mycobacteria grown in vitro have been observed 7 . It is not clear which of its epitopes would be suitable for a highly sensitive assay for both HIV positive and HIV negative patients across geographic regions that differ in the prevalence of genetic strains of mycobacterium. We compared three different high quality and well-characterized antibodies 7 , recognizing known LAM epitopes. Our cohort included TB positive patients, healthy and diseased controls, across five different countries: Peru, Guinea Bissau, Uganda, Venezuela, and the United States.
A second unresolved challenge is determining the absolute concentration range of LAM in urine for HIV negative and HIV positive active TB. This is mandatory to establish the required sensitivity and quantitative range for ongoing and future development of point of care testing methods 5,8 . In the present study we calibrated the immunoassay to deterimine accurate concentration levels of LAM in microbiologically confirmed positive patients, for three different categories of anti-LAM antibodies.
A third issue is the interference, or influence, of comorbidities other than HIV to alter the urinary LAM concentration. Quality control (QC) of urine has not been emphasized in past studies. Specific gravity, urinary tract infections, ketonuria, hematuria and glycosuria can all influence a urine test 9 . For our study set of 430 patients we performed a full urine QC analysis and evaluated the percent of urine samples that have abnormal urinalysis values. Of particular interest was glycosuria, because diabetes is growing in prevalence among TB patients.
A fourth unresolved question we addressed is the utility of urinary LAM testing in pediatric TB and extrapulmonary TB. The utility of urinary LAM has been previoulsy validated in EPTB patient living with HIV 5 ; here we investigate the urinary LAM concentration range in EPTB patients who are HIV negative. The rate of progression and the mortality rate for children and adults with extrapulmonary disseminated disease are higher than patients with isolated pulmonary TB 10 . These populations cannot produce sputum for conventional diagnostic modalities, thus an accurate urine screening test for active TB in these patients is critically needed 2-4,10-12 . Previous studies of urinary LAM diagnostics conducted on pediatric TB patients yielded low sensitivity, but nevertheless showed the promise of extending urinary LAM testing from adults to children 13,14 . In this study, we investigated new LAM epitopes and we identified one epitope that translates well to children.

Results
Three anti-LAM antibodies MoAb1 IgG, CS35 IgG, A194 IgG were selected from a panel of 12 (Supplementary  Table S1, Supplementary Methods) based on sensitivity, specificity and diversity of antigenic LAM epitopes using a previously published quantitative immunoassay described in Paris et al. 6 (Fig. 1, Supplementary Fig. S1). Analytical measurement range in non-concentrated undiluted urine was 0.6-5 ng/mL. Nanocage pre-analytical concentration increased the sensitivity 50-fold. Limit of detection was 12 pg/mL pg/mL for MoAb1 and CS35, and 13 pg/mL for A194; limit of quantification was 43 pg/mL for MoAb1 and CS35, and 46 pg/mL for A194.

Scientific Reports
| (2020) 10:13944 | https://doi.org/10.1038/s41598-020-70669-9 www.nature.com/scientificreports/  We observed within-patient variability of antibody reactivity using titers normalized with individual dose response calibration to the BEI Resources LAM standard (Fig. 1c). 57.5% of TB patients were positive for all three antibodies using a cut-off of 80 pg/mL, and 5.5% of TB patients were negative for all three antibodies. Additionally, 6.8% of TB patients were positive for CS-35 and A194 but not for MoAb1, 6.8% were positive for MoAb1 and CS-35 but not for A194, and 6.8% were positive for MoAb1 and A194 but not for CS-35. 10%, 2.7%, and 3.7% of TB positive patients were positive solely for MoAb1, CS-35 and A194, respectively (Fig. 2).

Discussion
We assessed whether a urinary LAM test for active TB could be extended beyond patients who are HIV coinfected. This study addressed several unresolved issues including the choice of antibodies, sources of assay variability, interfering comorbidities, and utility for childhood and extrapulmonary TB. We employed our previously described laboratory assay 6  It has recently been proposed that antigenically modified, or masked forms of the heterogeneous LAM molecule may be present in urine of infected patients. We cannot exclude that the LAM molecule might be altered depending on the stage of the disease, the host comorbidities, or the organ distribution of the infection. Since variability in the branched LAM structure has been previously described 7 we addressed the performance of CS35, MoAb1 and A194, three different purified previously characterized 7 anti-LAM mAbs, which recognize different sets of LAM epitopes 7 (Fig. 1, Supplementary Methods) using our previously described quantitative dot blot, which concentrates and immobilizes all forms of urinary LAM onto a solid phase for probing with a single species of anti-LAM antibody. www.nature.com/scientificreports/ www.nature.com/scientificreports/ Urinary LAM measurements successfully classified cases and controls with high significance (p = 8E−37, p = 5E−27, p = 6E−34, for MoAb1, CS35, A194, respectively) with a cut point of 80 pg/mL using CS35, MoAb1 and A194 for both HIV negative and HIV positive patients. Median LAM concentration values were 0.51 ng/mL, 0.46 ng/mL and 0.20 ng/mL for MoAb1, CS35, and A194 respectively. ROAUC was 0.85, 0.80, 0.84, respectively and ROAUC from the combination of the response from the three antibodies was 0.90, reaching 90% sensitivity at 73.5% specificity, confirming the utility of urinary LAM screening for active TB beyond patients with HIV co-infection.
Geographic location can introduce variability for TB screening because of heterogeneity in TB strain or clade prevalence, climate and state of hydration, disease stage at urine collection, comorbidity rate, and additional unknown factors.  www.nature.com/scientificreports/ sets. A194 yielded the best discriminatory power in Ugandan, and CS35 in Venezuelan patients. This intercountry difference was not entirely explained by difference in the frequency of HIV co-morbidity. For example, Venezuelan samples, which were all HIV negative, had the highest median value for CS35 and MoAb1 (Fig. 3).
It is important to note that even if the median LAM concentration varied among countries, the discriminatory power for TB cases versus controls was maintained (Peru: ROAUC = 0.88; Guinea Bissau: ROAUC = 0.89; Uganda: ROAUC = 0.90; Venezuela: ROAUC = 0.90). This is an encouraging finding supporting expanded studies for worldwide screening. Urinary sample quality and concentration state can have a profound effect on a screening assay. In the present study, approximately 20% of cases and controls had abnormal urinalysis values. The most significant endogenous interfering substance was glucose. The prevalence of type 2 diabetes mellitus (DM) is steeply increasing worldwide, in particular in low income countries where TB is endemic and the effect of glycosuria and diabetes on urinary LAM have not previously studied. Urinary LAM concentration in TB positive patients with glycosuria (n = 18) was significantly lower than TB positive patients with normal urinary glucose levels for CS35 (p = 0.015) and A194 (p = 0.035), but not for MoAb1 (Fig. 3). This finding is in keeping with reports suggesting that the Mtb cell wall polysaccharide composition is altered in diabetic patients 19 . MoAb1, which recognizes the MTX-Man cap of LAM, successfully discriminated TB positive patients presenting with glycosuria from TB negative controls (AUC = 0.87).
As previously observed 20-24 , LAM concentration was higher in HIV+/TB+ patients (CS35 p = 5E−05, A194 p = 0.0021). Patients with HIV co-infection were collected in Uganda (N = 36), Peru (N = 13), and Guinea Bissau (N = 6). 4/55 HIV co-infected patients were children. These covariates can increase the heterogeneity of LAM concentration values. In our previous study by Paris et al. 6 , we demonstrated that it is possible to detect urinary LAM in HIV negative TB patients using an experimental workflow which included a pre-analytical concentration step. Other investigators confirmed these findings 7,8,17,18,25 . Recent reports of LAM immunoassays employing high affinity purified antibodies indicate that acceptable analytical sensitivity required for HIV negative TB positive patients can be achieved without the use of a pre-analytical concentration step. The present study combined with the body of published work 6,8,26 demonstrates that at least one of the tested antibodies has a sensitivity sufficient for a future field deployable test for HIV−/TB+ patients.
EPTB accounts for 15% of diagnosed TB cases 18 and in its disseminated forms, leads to higher mortality rates compared to pulmonary TB 18 . In face of non-specific clinical symptoms and imaging tests, a definitive diagnosis of EPTB requires invasive and often impractical surgical procedures. We analyzed N = 24 EPTB HIV negative cases, which included pleural effusion, lymph node, bone, laryngeal, and meningoencephalitis TB. Three out of 24 patients were children. MoAb1 distinguished EPTB cases versus controls with high accuracy (AUC = 0.96, sensitivity = 96%, specificity = 80%, cut-off = 82 pg/mL, Fig. 4). A higher LAM signal can be explained by higher burden of disease that results in lower treatment success rates, which generally characterizes EPTB and disseminated TB patients 4 . In a previous study 6 , we have demonstrated correlation between urinary LAM values and severity of TB symptoms. Thus, the MoAb1 based LAM assay is a promising new approach to non-invasive detection of extrapulmonary disease.
Tuberculosis is among the top 10 causes of childhood mortality worldwide with a global estimate of 130,000 pediatric deaths per year 27 . Lack of appropriate screening, diagnostic, and treatment protocols has a severe toll on the life of TB pediatric patients: it is estimated that 1 in 4 undiagnosed children will die of the disease 19 . Because of the hurdle to procure invasive samples such as induced sputum and gastric lavage to perform Mtb culture, the need to develop sensitive technologies that enable Mtb detection non-invasively is of high priority to address the pediatric TB epidemic. We analyzed 19 culture positive pediatric patients (age, years: 2-17). 17/19 were positive for MoAb1, 12/19 and 9/19 were positive for A194 and CS35, respectively using a cut off of 80 pg/mL. MoAb1 showed great promise for pediatric screening (AUC = 0.87, sensitivity = 90%, specificity = 80%, cut-off = 80 pg/mL).
The present study has several weaknesses. Our goal was to mimic real-world urine testing conditions. Consequently, all samples, including those with abnormal urinalysis (e.g. glycosuria, bilirubinuria, hematuria, leukocytes, proteinuria) were included. In our patient cohort TB negatives are either asymptomatic, or symptomatic for non-TB diseases, but are not matched quantitatively in each country. A further limitation is that we used a one-sided immunoassay as opposed to a sandwich assay. Thus, while it can achieve a calibrated accurate concentration value for LAM, each antibody has to be judged against its own dose response curve. The overall combined sensitivity and specificity we observed for these heterogeneous cases for all countries combined was somewhat lower than we observed in our previous study of only HIV negative, TB culture positive, pre-treatment patients, from one hospital in one country 28 . Higher sensitivity and specificity were achieved when we subdivided the data by country of origin (Supplementary Figure S5). These regional comparisons imply potential variability in stage of disease or time of day for urine collection. Overall, across the heterogeneity of geographic regions, sample collecting institutions, urinary LAM continued to show strong promise for screening for patients with active TB, for assays that are adequately sensitive (LOD 12 pg/mL).
This study demonstrates that LAM as recognized by MTX-Man cap specific antibodies is a previously unappreciated, promising marker for pediatric and extrapulmonary disease patients. These findings support assay validation in larger cohorts of extrapulmonary and pediatric patients.

Methods
Study design and participants. In this case-control study, we assessed urine samples collected from adults and children undergoing evaluation for pulmonary or extra-pulmonary tuberculosis in four countries (Uganda, Peru, Venezuela, and Guinea Bissau). The George Mason University Institutional Review Board approved the study described herein (IRBnet IDs 1317042, 1244866, and 869592). The study followed princi- Procedures. Nanocage fabrication, covalent incorporation of binding baits, and one sided immunoassay was performed as described in Paris et al. 6 and in the Supplementary Methods. Lipoarabinomannan preparation from Mycobacterium tuberculosis (Mtb) strain H37Rv (BEI Resources, catalog #NR-14848) was analyzed using the Anthrone method 15 (Supplementary Fig. S6) and used as calibrator for the immunoassay.
Anti-LAM monoclonal antibodies CS35, A194, and MoAb1 were produced and purified according to Choudhary et al. 7 . CS-35 monoclonal antibody (mAb) was obtained from hybridoma cells derived from mice immunized with cell wall extracts of M. leprae. MoAb1 mAb was identified by phage display of single-chain variable fragment libraries generated from rabbits (Otsuka Pharmaceutical). A194 mAb was generated using in vitro cultures of fractionated memory B cells from a patient diagnosed with pulmonary TB. FIND 28 monoclonal antibody was provided by FIND 7 . All antibodies were purified using protein A chromatography and biotinylated.
LAM immunoassay was performed (Supplementary Methods) by spotting 2 µL of each sample in duplicate on polyvinylidene fluoride membranes (PVDF, Biorad) previously activated with methanol and equilibrated in DI water. Spots were allowed to dry and membranes were incubated with biotinylated anti-LAM MoAb1, CS-35, A194 mAbs and streptavidin-HRP (Biorad), and imaged using enhanced chemiluminescence system (Super-Signal West Dura, Thermo Fisher Scientific) and an Azure c300 imager. Densitometry analysis of immunoassay signals was conducted using ImageQuant software. Negative urine samples (N = 20) collected from healthy volunteers were considered as blanks and were used to calculate the background signal. The limit of detection (LOD) was calculated as the average of the blanks plus 2 standard deviations and the limit of quantification (LOQ) was the blank mean plus 10 standard deviations 16 . Statistical analysis. Patients were grouped by binary levels of categorical variables representing TB status, HIV status, pulmonary or extrapulmonary disease, and age (pediatric under NIH criteria of 18 years and younger versus adults). Statistical differences in urinary LAM values between the two groups were assessed by the Kluskal-Wallis H test with an alpha level of 0.05. Significance of LAM concentration differences among countries was assessed using the median test followed by a pairwise comparison by the Kluskal-Wallis H test. The analysis was performed using SPSS version 26. Relationship between covariates and outcome was investigated performing logistic and linear regression in Python 3 using the Scikit-Learn library. Covariates of interest included urinalysis measurements of glucose, bilirubin, ketone, blood, protein, urobilinogen, nitrite and leukocyte esterase, auramine staining scores, Microscopic Observation Drug Susceptibility Assay (MODS) results, MODS time to positivity. LAM concentration was considered as linear outcome. For the logistic regression, LAM measurements obtained with different antibodies were considered as covariates against TB status as binary outcome. Statistical significance of regression coefficients was assessed by two-sided t tests and z tests with an alpha level of 0.05. Receiver operating characteristics (ROC) analysis was performed using Python Scikit-Learn library; confidence intervals were calculated with 1,000 stratified bootstrap replicates. The urinary LAM concentration cut off was defined using logistic model regression, ROC analysis and tenfold cross validation studies ( Supplementary Fig. S2). Visualization of box plot, ROC plots, and Venn Diagrams were obtained using Python 3 Matplotlib 3.1.1 and Seaborn 9.0 libraries.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.