mCherry10-expressing Mycobacterium tuberculosis (Mtb) was generated as described elsewhere [13]. Mtb H37Rv was obtained from American Type Culture Collection ((ATCC) 27294, Manassas, VA, USA). Strains were cultured as described in detail elsewhere [14]. In order to generate frozen stocks, mid-log-phase bacterial suspensions (OD₆₀₀, 0.2 to 0.4) were stored at −80°C. After four weeks of storage, the number of bacteria (colony forming units, CFU) was determined by plating serial tenfold dilutions on 7H10 plates (0.05% Tween 80 / 10% heat-inactivated bovine serum in phosphate buffered saline (PBS)) and counting colonies after 3–4 weeks of incubation at 37°C.

For spiking experiments, frozen stocks of Mtb H37Rv were thawed, centrifuged (4000xg), bacteria re-suspended in PBS and a defined amount of this solution added to buffer or saliva samples. To address the impact of proteinase K plus DTT treatment on bacterial viability, Mtb H37Rv spiked PBS samples were incubated in the presence or absence of Proteinase K (18U/ml, Carl Roth, Karlsruhe, Germany) and Dithiothreitol (DTT, 2mM, Sigma-Aldrich, St. Louis, USA) for 60 minutes at 37°C and subsequently subjected to CFU analysis as described.

5.6 x 10⁸ magnetic beads (Micromer-M, polystyrene body, surface: PEG-COOH, 5μm; in dH₂0) were incubated with 12.8 mg N-hydroxysuccinimide (NHS) and 6.4 mg 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDC) in 20% [v/v] 2-(N-morpholino)ethanesulfonic acid (MES) buffer (pH 6.3) and agitated for 1.5 hours. Activated beads were washed twice with PBS and then incubated with the biotinylated lipopeptide lipobiotin (PHCKKKKK(Aca-Aca-Biotin) x 3 TFA, N-Palmitoyl-S-(1,2-bishexadecyloxy-carbonyl) ethyl-[R]-cysteinyl-[S]-lysyl-[S]-lysyl-[S]-lysyl-[S]-lysyl-[S]-lysine(ɛ-aminocaproyl-ϵ-aminocaproyl-biotinyl) x 3 CF3COOH (EMC Microcollections, Tübingen, Germany) at 1.5 mg/ml in carbonate buffer (pH 9.2) at RT for 4.5 hours. Functionalized beads were washed twice with PBS and incubated in 10mM ethanolamine (in carbonate buffer; pH 9.2) to block unreacted ester groups (1h, RT). Lipobiotin-functionalized magnetic beads (LMB) were washed, re-suspended in PBS and kept at 4°C. Non-functionalized magnetic beads (NMB) that were equally washed and re-suspended in PBS served as a control.

PBS (0.5 ml) was spiked with Mtb H37Rv bacteria at the indicated concentrations. Subsequently, 4.6x10⁶ LMBs or NMBs were added and samples incubated at 37°C for 1 hour on a thermomixer (60 rpm; Eppendorf, Hamburg, Germany). After this step, beads were retained in the tube by use of a magnet (DynaMag-2 Magnet, Life technologies) and washed two times with PBS. Finally, beads were re-suspended in Tris and EDTA (TE) buffer (pH 10) before proceeding further.

Saliva samples were collected without stimulation from healthy donors or from patients with drug-sensitive, sputum smear positive pulmonary TB at different times of the day and were subsequently stored at -80°C. Samples were pseudonymized and processed in an observer-blinded fashion. The study was approved by the Ethics Committee of the University of Luebeck (ref. 14-079A).

The treatment with LMBs, also referred to as "LMB assay", for enrichment and quantification of mycobacterial DNA in saliva samples can be summarized as following. Saliva samples were thawed and either spiked with Mtb H37Rv or directly pre-treated with Proteinase K (18U/ml) and Dithiothreitol (DTT, 2mM) for 45 minutes at 37°C (Fig 6, step 1). Subsequently, samples were centrifuged (150xg), and supernatants transferred into fresh tubes. 4.6x10⁶ LMBs were added and samples incubated for 60 minutes at 37°C (Step 2). Samples were centrifuged (3629xg), and beads retained in the tube by use of a magnet (Step 3). Subsequently, beads were washed twice with PBS and re-suspended in TE buffer (pH 10), and heat-treated for the release of the DNA (95°C, 30 mins). After DNA purification and concentration by ethanol precipitation (step 4, for further details see below), samples were subjected to real-time quantitative PCR analysis (step 5, for details see below).

Saliva samples were mixed with proteinase K (18U/ml) and DTT (2mM) or with the same volume of dH₂O as control and incubated (45 minutes, 37°C, 60rpm) on a roller mixer. The rheological properties of the saliva were measured using a Bohlin Rheometer (Bohlin CVO 120 High Resolution, Malvern Instruments Ltd., UK) and the 40mm diameter plate-plate apparatus. The slit width was set to 150μm and the temperature was constant at 36°C. The change in viscosity of 1 ml of each sample was recorded with a shear profile of 1 s^-1 to 100 s^-1. Measurements were repeated three times.

In order to assess the binding of Mtb bacteria to beads, mCherry10-expressing Mtb (at the concentration indicated) in PBS were incubated with 1.57x10⁶ LMBs or NMBs for 60 mins at 37°C and beads were separated by use of a magnet (see above). Subsequently, beads were washed twice with PBS and subjected to flow cytometric analysis (BD FACSCanto™ II (B&D Biosciences, Franklin Lakes, USA)). Data was analyzed by use of FCS Express 7.10.0007 (DeNovo Software, Los Angeles, USA) and earlier versions.

Samples containing Mtb were heat-treated on a thermomixer (95°C, 30 minutes, 1400 rpm). Subsequently, in order to concentrate and purify the extracted DNA, samples were vortexed in the presence of ethanol (81% [v/v], VWR International, Radnor, USA) ammonium acetate (9% [v/v], Thermo Fisher Scientific, Waltham, USA) and 60 μg glycogen (Roche, Basel, Switzerland) and incubated overnight at -20°C. Following centrifugation (30 minutes, 16.200xg), the supernatant was discarded. This step was repeated. Subsequently, the pellet was washed with ethanol (75% [v/v] and sample centrifuged (10 minutes, 16.200xg). Finally, the pellet was air dried for 15 minutes and dissolved in 5μl H₂O before proceeding further.

For detection of Mtb DNA, real-time quantitative PCR (qPCR) was performed based on a previously published protocol that detects a region in Rv2819 of Mtb [15]. In a final volume of 10 μl TaqMan Universal PCR Master Mix (Roche), forward and reverse primer (nBjF(5’-aagcattcccttgacagtcgaa); nBjR(5’-ggcgcatgactcgaaagaag) 500nM each), fluorescein-(FAM)-BlackBerry-(BBQ)-non-Beijing-Probe-(5’-6FAM-tcatcaaagaccctcttggaaggccc-BBQ; 200nM), double distilled water and the isolated Mtb DNA were mixed in a 96 well plate. The plate was sealed and centrifuged (2 minutes, 180xg, 4°C) before analysis with the LightCycler 480 Instrument II (Roche; Preincubation, 95°C, 10 min; Denaturation, 92°C, 15s; Extension, 60°C, 60s; Cooling, 40°C, 10s for 50 cycles).

Statistical evaluation was performed by use of R 3.5.0 and the R package nparLD (Open Source; [16]). The data was analysed by use of non-parametric analyses of variance for repeated measurements [16] and the obtained p-values were adjusted to the total number of tests performed in the study. For each dataset, data from all bacterial doses were included in the statistical evaluation.

To assess whether lipobiotin-functionalized magnetic beads (LMBs) are of potential use to enrich Mtb bacteria from patient material such as saliva, we assessed whether Mtb bacteria bind to LMBs and compared the binding to non-functionalized magnetic beads (NMBs). Thus, PBS-samples (0.5ml) were spiked with various numbers of fluorescent, mCherry-expressing Mtb bacteria (10⁶, 10⁷ or 10⁸ bacteria per 0.5ml; equivalent of 2x10⁶ - 2x10⁸ bacteria per ml) and incubated in the presence of LMBs or NMBs. Subsequently, magnetic beads were isolated, washed and subjected to flow cytometric analyses. As shown in Fig 1, the median fluorescence intensity increased in a dose-dependent manner with increasing numbers of added bacteria when samples were incubated with LMBs. In contrast, there was no substantial increase in median fluorescence intensity for samples incubated with NMBs, irrespectively of the amount of bacteria added. Considering data from all bacterial doses, a highly significant increase in fluorescence intensity was observed when LMBs were compared to NMBs (p = 3.729186 x 10⁻²⁶), strongly suggesting that Mtb bacteria bind to lipobiotin-functionalized beads, whereas only minimal or no binding occurs with non-functionalized beads.

qPCR is utilized as a molecular diagnostic test for tuberculosis by detecting mycobacterial DNA in samples. Thus, we assessed whether utilization of LMBs could improve qPCR-based detection of Mtb. In a first set of experiments, we spiked various amounts of Mtb H37Rv (10²,10³ and 10⁴ bacteria per 0.5ml, equivalent of 2x10² - 2x10⁴ bacteria per ml) to PBS samples (0.5 ml), which were then left untreated or underwent pre-treatment with LMBs. The magnetic beads were separated from the samples by use of a magnetic stand and–after two washing steps–re-suspended in PBS. The same volume of untreated or LMB pre-treated sample was subjected to heat treatment (95°C, 30 min), DNA purification and qPCR analysis detecting a well-established target in Mtb [15]. As depicted in Fig 2, with 10⁴ bacteria, Mtb was not detectable in 1 out of 6 (17.3%) of the cases when samples were untreated, while all samples were tested positive when samples were pre-treated with LMBs (LMB assay). Notably, treatment reduced the average ct-values from 37.6 to 33.9 indicative of an increased concentration of Mtb DNA in samples treated with the LMB assay. After determining the reaction efficiency of the conducted qPCR (E = 1.77, compare S1 Fig) we calculated the relative DNA concentration of the samples. For 10⁴ bacteria, this analysis revealed an ~8.7-fold enrichment of Mtb DNA in LMB-assay treated samples when compared to untreated samples. In line with this, with 10³ bacteria, Mtb was not detectable in 4 out of 6 (66.6%) of the cases when samples were untreated, while all samples were tested positive when samples were pre-treated with the LMB assay. Similarly, with 10² bacteria, Mtb DNA was not detected when samples were left untreated, but in 50% of the LMB-assay treated samples. In summary, LMB pre-treatment substantially reduced the mean ct-values and the rate of false negative PCR results. Results from this first set of experiments analysing aqueous, Mtb-spiked samples suggest that LMB pre-treatment has the potential to improve qPCR-based detection of Mtb also in other biological fluids.

In contrast to aqueous solution, saliva has a higher viscosity, which may interfere with the binding of Mtb to the LMBs. To circumvent this issue, we aimed to perform a pre-treatment employing the protease proteinase K and the reducing agent dithiothreitol (DTT). This treatment, however, could affect viability of Mtb. Thus, we first incubated Mtb for 60 minutes at 37°C with Proteinase K (18U/ml) and DTT (2mM) and plated the bacteria on solid 7H10 agar, revealing similar CFU counts between untreated and Proteinase K/DTT- treated samples (S2 Fig, p = 0.3753). Second, to test the anti-viscosity effect of these compounds, we incubated saliva samples for 45 minutes at 37°C in the absence and presence of Proteinase K (18U/ml) plus DTT (2mM). Rheological analyses revealed that the viscosity of the saliva samples varied considerably between donors (Table 1). Importantly, treatment with Proteinase K plus DTT for already 45 minutes reduced the viscosity of saliva samples in all donors compared to the respective untreated controls. In summary, our data show that Proteinase K / DTT treatment reduces viscosity of saliva samples, while not affecting viability of Mtb.

To test whether the LMB assay improves qPCR-based detection of Mtb in saliva, samples obtained from healthy donors (5 ml) were incubated with Proteinase K plus DTT, spiked with different amounts of Mtb (10²–10⁵ per 5ml; equivalent of 2x10¹ - 2x10⁴ bacteria per ml) and were either left untreated or LMB-assay treated prior to heat-treatment, DNA extraction and qPCR analysis. With 10⁵ bacteria, all samples (15 in each group) were tested positive, irrespective of whether the LMB-assay was used or not (Fig 3). Importantly, the mean ct values decreased from 36.99 to 34.5, which represents an enrichment of DNA concentration by performing the LMB assay of 4.2-fold, when considering the reaction efficiency of the qPCR (E = 1.77, see S1 Fig). With 10⁴ bacteria, Mtb was not detectable in 10 out of 15 (66.6%) of the untreated samples, whereas detection rate remained at 100% with prior LMB-Assay. Similarly, with 10³ bacteria, Mtb was not detectable in 9 out of 15 (60%) of the cases, while the LMB assay allowed Mtb detection in 93% of the samples. Strikingly, the LMB-Assay allowed the detection of 10² Mtb in 40% of the samples (6 out of 15), whereas Mtb DNA detection was not possible without prior LMB-Assay (0 out of 15). Importantly, when integrating data from all bacterial doses, ct-values were significantly reduced when statistically comparing LMB assay- treated with untreated samples (p = 6.038205 x10^-35). In summary, our data demonstrate that the LMB-Assay improves qPCR-based detection of Mtb in spiked saliva samples, suggesting it to be a valuable tool to enrich bacteria even from biological fluid with a high viscosity.

To assess whether the LMB assay can improve a saliva-sample based diagnosis of pulmonary tuberculosis, two cohorts were tested in an observer-blinded manner, each consisting of ten saliva samples from healthy subjects (as a control for specificity) and eight and nine saliva samples from sputum-smear positive patients, respectively (Figs 4 and 5). One fraction of each sample was left untreated (control), while another fraction was incubated with LMBs before undergoing heat treatment, DNA isolation, and qPCR analysis (LMB assay; workflow as shown in Fig 6). Notably, from each saliva sample, three technical replicates were conducted and detection of Mtb was considered positive if any values were recorded positive. According to this, in the first cohort (Fig 4A), Mtb was detected in 44,4% of the TB patients when samples were left untreated (Sample ID 1, 3, 6 and 8; 4 out of 9), while detection of Mtb improved to 66,67% when LMB assay was employed prior to the qPCR (Sample ID 1, 2, 3, 6, 8, and 9; 6 out of 9). However, the average ct-values in the four patients detected with both methods did not change, and there was no statistically significant difference between the groups (p = 0.7506). In addition, in one patient (sample ID 9), qPCR detection after performing the LMB assay appeared less robust, as positive results were obtained in only two out of three cases in the technical replicates. Finally, there were no differences in the healthy subject’s samples with 10% false positive results for control and LMB assay (Fig 4B).

In the second cohort (Fig 5), positive qPCR signals were detected in 50% and 75% of cases, respectively, when untreated and LMB assay-treated samples were compared (Fig 5A). In two patients, Mtb was detected only when the LMB assay was applied (Sample ID 2 and 4). In four patients, both methods lead to a positive detection (Sample ID 1, 6, 7, 8), but the LMB assay showed a more reliable positive readout of the replicates and lowered the mean ct-value in two patients substantially (Sample ID 6 and 7). Consistent with that, we found significantly reduced ct-values in LMB-assay treated saliva samples when compared to untreated samples (p = 0.0253), confirming that the LMB assay improves qPCR-based detection of mycobacterial DNA in this cohort. Surprisingly, the LMB assay led to three single positive PCR results (10% out of 30 tests performed) in saliva samples from healthy subjects compared to those that were left untreated (Fig 5, sample ID 15, 16 and 19). Consequently, we again collected saliva (2 of 3 donors were available), re-examined the samples and this time obtained negative qPCR results, regardless of whether the LMB test was performed or not (S3 Fig, sample ID 15 and 19).

In summary, our data from two small cohorts show that more TB patients can be correctly identified from saliva analysis when samples were subjected to LMB assay prior to analysis.