Urine lipoarabinomannan in HIV uninfected, smear negative, symptomatic TB patients: effective sample pretreatment for a sensitive immunoassay and mass spectrometry

Abstract

Our study sought to determine whether urine lipoarabinomannan (LAM) could be validated in a sample cohort that consisted mainly of HIV uninfected individuals that presented with tuberculosis symptoms. We evaluated two tests developed in our laboratory, and used them on clinical samples from Lima, Peru where incidence of HIV is low. ELISA analysis was performed on 160 samples (from 140 adult culture-confirmed TB cases and 20 symptomatic TB-negative child controls) using 100 μL of urine after pretreatment with Proteinase K. Two different mouse monoclonal antibodies-CS35 and CHCS9-08 were used individually for capture of urine LAM. Among cases, optical density (OD₄₅₀) values had a positive association with higher bacillary loads. The 20 controls had negative values (below the limit of detection). The assay correctly identified all samples (97-100% accuracy confidence interval). For an alternate validation of the ELISA results, we analyzed all 160 urine samples using an antibody independent chemoanalytical approach. Samples were called positive only when LAM surrogates-tuberculostearic acid (TBSA) and D-arabinose (D-ara)-were found to be present in similar amounts. All TB cases, including the 40 with a negative sputum smear had LAM in detectable quantities in urine. None of the controls had detectable amounts of LAM. Our study shows that urinary LAM detection is feasible in HIV uninfected, smear negative TB patients.

Figure 1

Inactivation of Proteinase K before ELISA, Urine from a healthy volunteer from TB non endemic region was spiked with Mtb (CDC1551) LAM (ranging from 12.5 to 0.02 ng/mL) and pretreated with Pro K (200 ug/mL) to release LAM. Before ELISA, Pro K was inactivated at different time points (0 min to 30 min) to ensure complete inactivation. At 0 min inactivation after pretreatment, the ELISA signal is very weak (brown line), however, with 10 (blue line), 20 (red line) and 30 (green line) min inactivation a proper standard curve is achieved with no difference between the inactivation times.

Figure 2

Capture ELISA showing LAM standard curve. Urine from two healthy volunteers (NEU#2, NEU#3) from a TB non endemic region and two clinical samples (UP148, UP158) that were collected from clinically TB negative patients in a TB endemic region were spiked with known amount of LAM and serially diluted two fold (ranging from 12.5 to 0.02 ng/mL) to derive a standard curve and pretreated with Pro K (200 ug/mL). Urine with no LAM but treated with Pro K were used as negative control for background signals. (A) Urine from healthy volunteers spiked with LAM show the LoD at ~ 0.05 ng/mL and (B) urine from TB negative patients spiked with LAM also show LoD at ~ 0.05 ng/mL. The background signal from urine from one of the healthy volunteers is lower than the signals from the urine from clinical samples with no LAM. This also shows that urine from different sources give varying background levels in an ELISA.

Figure 3

ELISA assay validation on the initial cohort of 60 urine samples. OD values are plotted along the x-axis; each dot represents a single sample. A vertical solid line at 0.252 represents the cutoff value that separates all Non-TB samples from all TB positive samples. A dashed line is placed at the maximum cutoff chosen by bootstrap. The ten samples that would be classified differently with the higher cutoff are outlined in green. Samples in orange dots are all considered positive.

Figure 4

Comparative ELISA on follow-up cohort of 100 urine samples using two sets of antibodies. OD values for 100 samples are listed in descending order, with cutoff values for prediction shown by the horizontal lines. Points below the line were identified as negative, and above the line as positive. Yellow triangles indicate culture positive samples, black dots indicate culture negative samples. 9/100 samples were misidentified perhaps due to experimental error which were later correctly identified (Supplemental File Fig. S1) after reanalysis.

Figure 5

Correlation between d-ara and TBSA by GC/MS analysis in each urine sample analyzed.

Figure 6

Correlation of smear− and smear+ category with LAM amounts in urine. For TBSA, the LoD is set to be 2 ng/mL below which the peaks are discernable and for d-ara 4 peaks must be present to conclude as positive. LoD for d-ara was found to be < 2.4 ng/mL. Both TBSA and d-ara must calculate out to be in similar amounts (± 5% error for manual integration). Overall, 138 of 140 TB-positive samples has LAM above LoD by TBSA (98.5%) and 127 has LAM by d-ara (90.7%). All 20 clinically TB negative samples had no LAM by either assay (Specificity > 99%).

Figure 7

TBSA detection of spiked (non-endemic urine; 1 mL) CDC1551 LAM on Varian CP3800 and TSQ 8000 mass spectrometer in 2015 and 2017 respectively. Out of 15 ng LAM spiked, 12.5 ng was detected on Varian and out of 10 ng LAM spiked, 9.0 ng was detected on TSQ 8000. These spiked samples were subjected to Octyl-Sepharose column chromatography and subsequent pentafluorobenzyl derivative synthesis before GC/MS analyses. D2-Palmitic acid (10 ng and 5 ng respectively) were added to the samples as internal standard before synthesis. Calculations are shown on the right.