Extract from used Xpert MTB/RIF Ultra cartridges is useful for accurate second-line drug-resistant tuberculosis diagnosis with minimal rpoB-amplicon cross-contamination risk

Xpert MTB/RIF Ultra (Ultra) detects Mycobacterium tuberculosis and rifampicin resistance. Follow-on drug susceptibility testing (DST) requires additional sputum. Extract from the diamond-shaped chamber of the cartridge (dCE) of Ultra’s predecessor, Xpert MTB/RIF (Xpert), is useful for MTBDRsl-based DST but this is unexplored with Ultra. Furthermore, whether CE from non-diamond compartments is useful, the performance of FluoroType MTBDR (FT) on CE, and rpoB cross-contamination risk associated with the extraction procedure are unknown. We tested MTBDRsl, MTBDRplus, and FT on CEs from chambers from cartridges (Ultra, Xpert) tested on bacilli dilution series. MTBDRsl on Ultra dCE on TB-positive sputa (n = 40) was also evaluated and, separately, rpoB amplicon cross-contamination risk . MTBDRsl on Ultra dCE from dilutions ≥103 CFU/ml (CTmin <25, >“low semi-quantitation”) detected fluoroquinolone (FQ) and second-line injectable (SLID) susceptibility and resistance correctly (some SLIDs-indeterminate). At the same threshold (at which ~85% of Ultra-positives in our setting would be eligible), 35/35 (100%) FQ and 34/35 (97%) SLID results from Ultra dCE were concordant with sputa results. Tests on other chambers were unfeasible. No tubes open during 20 batched extractions had FT-detected rpoB cross-contamination. False-positive Ultra rpoB results was observed when dCE dilutions ≤10−3 were re-tested. MTBDRsl on Ultra dCE is concordant with isolate results. rpoB amplicon cross-contamination is unlikely. These data mitigate additional specimen collection for second-line DST and cross-contamination concerns.

spoligotyping 6 , a method useful for monitoring the molecular epidemiology of TB outbreaks. This additional testing does not require extra specimen collection nor additional downstream DNA extraction, both of which can exacerbate patient loss within the diagnostic care cascade.
As Xpert is a real-time PCR that generates quantitative information, a cycle threshold value (C T <24) was identified at which downstream dCE testing using MTBDRsl was successful and fully concordant with MTBDRsl results on matching isolates 7 . However, Xpert dCE was not useful for first-line DST using the WHO-endorsed MTBDRplus assay, likely due to interference from large numbers of Xpert rpoB amplicons. In addition to the dCE approach, others 8,9 have shown it is possible to test leftover specimen-sample reagent mix remaining after Xpert, however, remnant volume is not always present and DNA extraction and downstream clean-up might still be needed.
Xpert MTB/RIF Ultra (Ultra) recently superseded Xpert as WHO-endorsed frontline molecular test-of-choice for TB and rifampicin resistance 10 . Compared to Xpert, Ultra has higher sensitivity in paucibacillary samples, however, specificity is overall lower [11][12][13] . Ultra is a different assay compared to Xpert and it is not necessarily given that the extract approach would be feasible on Ultra dCE. We aimed to confirm that Ultra dCE would be useful for second-line DST. Furthermore, we asked if extract from other chambers within the cartridge other than the diamond (i.e., chambers that are likely rpoB amplicon-free), may contain DNA. We quantified this DNA using a Mycobacterium tuberculosis complex 16S rRNA real time qPCR and evaluated whether this DNA was useful for first-line DST using the FluoroType MTBDR (FT) (Hain Lifescience, Germany) assay 14,15 . A test such as FT could, for example, be used to check for isoniazid mono-resistance or confirm Ultra rifampicin-resistance results.
Lastly, as the cartridge extraction (CE) procedure involves aspirating fluid rich in rpoB amplicons, it may represent a source of cross-contamination. We sought to evaluate this risk, both under a prolonged exposure scenario (where collection tubes were purposely exposed during extended batch extractions) and an absolute worst-case scenario (directly adding dCE to a sample later tested by Ultra). Showing that the extracted cartridge approach in Ultra is compatible with MTBDRsl and represents minimal rpoB amplicon cross-contamination risk would increase the likelihood of implementation, especially as Xpert is in the process of being phased out in lieu of Ultra. In turn, this could reduce both sputum collection requirements for complete DST and time-to-effective-treatment initiation.

Methods
Ethics statement. Methods and protocols were carried out in accordance with relevant guidelines and regulations. The study was approved by the Health Research Ethics Committee of Stellenbosch University (N09/11/296) and the City of Cape Town (10570). Permission was granted to use anonymised residual specimens collected during routine diagnostic practice and thus patient informed consent was waived.
Ultra and Xpert on dilution series of Mycobacterium tuberculosis strains. Culturing of genotypically-confirmed drug-susceptible (DS-TB) and extensively-drug resistant (XDR) M. tuberculosis isolates were done in a Biosafety Level (BSL) 3 laboratory to an OD 600 of 0.6-0.8 (Fig. 1A). A triplicate tenfold dilution series from three separate cultures [10 0 -10 4 colony forming units (CFU)/ml] was prepared in phosphate buffer (33 mM Na 2 HPO 4 , 33 mM KH 2 PO 4 ; pH 6.8) with 0.025% Tween80 (Sigma-Aldrich, United States). Colony counts were done on 7H11 Middlebrook agar (BD Biosciences, United States). A total of 52 dilutions [four dilutions, 10 1 -10 4 CFU/ml in triplicate for both strains plus a negative control for each strain; (4 × 3 × 2 + 2) × 2] were made up to 1 ml and tested by Ultra (n = 26) or Xpert (n = 26) per the manufacturer's instructions 16,17 . Used positive cartridges were stored prior to extraction at 4 °C for ≤3 days. Crude DNA (heat inactivated for www.nature.com/scientificreports www.nature.com/scientificreports/ 2 hours at 100 °C) from the same strains served as positive controls for downstream tests (16S rRNA gene qPCR, MTBDRplus, MTBDRsl, FT). Ultra on sputum from TB patients. Forty used positive Ultra cartridges done on NALC-NaOH decontaminated sputa from pre-treatment TB patients with known drug resistance [5 rifampicin-mono-resistant, 15 MDR, 10 pre-XDR (resistance to rifampicin, isoniazid and either a fluoroquinolones or a second-line injectable), 10 XDR] were collected from November 2015 to September 2017 and dCEs were extracted as described previously 6 (Fig. 1B). To confirm MTBDRsl results from dCEs, MTBDRsl was done per the manufacturer's instructions directly on corresponding decontaminated sputa 18,19 . Ultra cartridges were processed in a manner blinded to MTBDRsl results.
Recovery of mycobacterial genomic DNA from used Ultra and Xpert cartridges. Preparation of work space. BSL2 hood surfaces were sterilised [1% NaOCl (bleach), 70% EtOH, 5 min UV irradiation] before and after each batched extraction. Each cartridge was wiped with 1% bleach and 70% EtOH before and after each extraction.
Description of cartridge design. To investigate the feasibility of testing extract from Ultra and Xpert cartridge chambers, an understanding of their design and inner processes is required. As described previously, each cartridge has a similar design consisting of a foot, valve, body, reaction tube and lid 20,21 . The five internal chambers hold buffers and lyophilised PCR reagents used for sputum homogenisation, washing away debris, and DNA extraction, purification, and amplification 22 . The Xpert and Ultra procedures, including the processes inside the cartridges and the contents of each chamber are described in the supplement. After assay conclusion, the volumes typically remaining in each chamber are ~500 µl for Chamber 1 (C1), ~3 ml for Chamber 2 (C2), ~5 ml for Chamber 3 (C3) and ~500 µl for Chamber 4 (C4) [Chamber 5 (C5) had no volume remaining after test completion].
Diamond chamber extract. dCEs were extracted from all positive cartridges by puncturing the rear chamber with a sterile 29 G × 1/2′′ 1 ml insulin syringe (Avacare, South Africa) ( Fig. 2A,B) as described previously 6 . The full volume was extracted (~15 µl for Xpert; ~35 µl for Ultra). CEs were stored in microcentrifuge tubes at −20 °C prior to analysis.
Other chambers. Five cartridge chambers (C1, C2, C3, C4, C5) were accessed by inserting a 22 G spinal needle (Becton Dickinson, United States) fixed a 5 ml syringe ( Fig. 2A; a pipette may also be used for C1) and the entire volume withdrawn ( Fig. 2A,B). C5 had no remaining volume left after Xpert or Ultra test completion. No DNA extraction or purification steps were done for downstream assays.
16s rRNA gene quantitative PCR (qPCR) on cartridge extract. CEs from C1-4 and dCE from Ultra and Xpert done on the serial dilutions were tested (heat extracted crude DNA from matching isolates was used as positive control). For each qPCR, 5 µl iTaq Universal SYBR Green Supermix (Bio-Rad), 0.3 µl (300 nM) of M. tuberculosis specific forward (V4 515F) primers, 0.3 µl (300 nM) of M. tuberculosis specific reverse (V4 806R) primers (Table S1) and 1.4 µl nuclease-free water was used 23 . 3 µl CE was added and amplification occurred using a Bio-Rad CFX-96. The threshold used to determine if a reaction was excluded from subsequent analyses was defined as a C q value greater than the average of the triplicate negative controls for that run. Chambers with a C q less than that average value were considered positive for M. tuberculosis complex (MTBC) DNA and used for MTBDRplus, MTBDRsl and FT. www.nature.com/scientificreports www.nature.com/scientificreports/ MtBDRplus and MtBDRsl line probe assays on cartridge extract. Diamond chamber extract. MTBDRplus and MTBDRsl (both version 2.0) were performed on dCEs from Ultra and Xpert done on the in vitro dilution series. For Ultra done on sputa from patients, only MTBDRsl was done. 5 µl dCE was used for MTBDRplus and MTBDRsl each. MTBDRplus and MTBDRsl results were reported as described 24 : either actionable [TUB-band positive and determinate (gene-specific locus bands present)] or non-actionable [TUB-band negative or TUB-band positive but indeterminate (gene-specific locus band absent)]. Susceptibility calls were made for all actionable results. Banding patterns were read by two experienced independent readers blinded to each other's calls, the Ultra and Xpert results, and, for the dilution series experiement, the strain antibiograms (if there was a discrepancy between readers, a third experienced reader reviewed results and did the final classification).
Other chambers. MTBDRplus and MTBDRsl were done on C2 and C4 CEs from both Ultra and Xpert done on the dilution series. C1, C3, and C5 were not tested with LPAs as their CEs were 16S rRNA qPCR-negative or there was no volume remaining to test after the Ultra or Xpert test had completed (C5).
FluoroType MTBDR on cartridge extract. Diamond chamber. dCEs from Ultra and Xpert cartridges done on the in vitro dilution series were tested by FT using the manufacturer's instructions 25 . A total of 26 tubes for each test (Ultra, Xpert) were tested [four dilutions from 10 1 -10 4 CFU/ml in triplicate for both strains plus a negative control for each strain, (4 × 3 × 2 + 2)]. As Xpert dCE had a volume of ~15 µl, after MTBDRplus (5 µl), MTBDRsl (5 µl), and the 16S rRNA qPCR (3 µl) were all done on the same Xpert dCE, the remaining volumes (5-14 µl) were made up to 20 µl with dH 2 O for FT (the recommended input volume) 25 . All Ultra dCEs (~35 µl originally) had 20 µl remaining and the full 20 µl was used for FT. FT results were classified in a manner similar to that for the line probe assays: actionable (MTBC detected; rifampicin and isoniazid susceptible or resistant) or non-actionable (no MTBC detected, MTBC indeterminate or MTBC detected but rifampicin or isoniazid indeterminate).
Other chambers. FT was done on C2 and C4 (as for LPAs) from both Ultra and Xpert cartridges used for the dilution series. Spiking of amplicons. The same XDR-TB strain with known Xpert and Ultra rpoB resistance profiles was used in the dilution series (Fig. 1C). Ultra and Xpert were each done on 1 ml of a 10 4 CFU/ml concentration (in triplicate). dCEs were extracted and used for a dilution series (10 0 , 10 −3 , 10 −6 , and 10 −9 ; each 1 ml final volume). For all dilutions, 5 µl was added to 700 µl of the DS-TB strain (10 4 CFU/ml) and tested with Ultra [700 µl was used as, when combined with the recommended two-fold sample reagent volume, the 2 ml input volume is reached with minimal sample unused (~100 µl)].

Results
Mycobacterium tuberculosis complex genomic DNA detection in different chambers from cartridges done on dilution series. Though qPCR-positive results were obtained from C2, C4 and the dCE (Fig. S1), these results were highly variable even at high concentrations of bacilli (at least 10 4 CFU/ml), suggesting interference. As C2, C4 and dCE gave positive qPCR results on cartridges done on some dilutions, and C1 and C3 gave none, we only explored the utility of the former for downstream testing using MTBDRplus, MTBDRsl, and FT.
Resistance detection. MTBDRsl correctly identified FQ and SLID resistance on Ultra dCE done on all XDR strain aliquots ≥10 3 CFU/ml (Fig. 5). On the DS-TB strain, MTBDRsl identified FQ susceptibility in all three 10 4 CFU/ ml replicates and in 2/3 (67%) replicates for SLIDs (one indeterminate). At 10 3 CFU/ml for the DS-TB strain, 2/3 (67%) were correctly identified as FQ susceptible (one indeterminate) and all were SLID-indeterminate. The C Tmin threshold at which all MTBDRsl results was feasible on Ultra CE was <25, which was used for further experiments. Similar results were obtained for MTBDRsl on Xpert dCE.
MtBDRsl on extract from cartridges done on clinical specimens. TB  www.nature.com/scientificreports www.nature.com/scientificreports/ on dCE from Ultra done on clinical sputa had 37/40 (93%) actionable results (the rest were non-actionable). Non-actionable results corresponded to "trace" or "very low" semi-quantitative categories.
Resistance detection. Of the actionable results, 35/37 (95%) fell within the defined threshold (C Tmin <25) and of these all FQ results were concordant with MTBDRsl on sputum and all but one SLID result were concordant (false-susceptible). Though this percentage is slightly higher than the number of patients with C Tmin <25 in our setting, which was determined to be 86% (based on an evaluation of Ultra done in sympotmatic patients in primary care 26 ), which further show that this approach would benefit the majority of patients in our setting. Of the 2/37 (5%) results that were actionable but fell above the defined threshold, one was concordant with MTBDRsl on sputa and one was indeterminate for FQs and discordant for SLIDs (false-resistant).  All sixty tubes exposed were FT MTBC-negative and had no rpoB amplification.

Discussion
We have validated MTBDRsl on CEs from used Ultra cartridges for genotypic second-line DST. We show: (1) MTBDRsl on Ultra dCE when C Tmin <25 enabled DST concordant with sputum results, (2) risk of rpoB extract cross-contamination is unlikely if standard aseptic protocols are followed, (3) neither 16S rRNA qPCR, MTBDRplus, MTBDRsl nor FT are feasible on other cartridge chambers, nor was MTBDRplus or FT on Ultra and Xpert dCEs. These data support the use of Ultra extract for second-line genotypic DST.
We defined a threshold at which MTBDRsl is likely to work on Ultra dCE from the vast majority of Ultra-positive patients, thereby avoiding time and resources wasted on dCE unlikely to give a valid result. We are mindful that there were some indeterminate SLID results (in line with previous reports of higher MTBDRsl indeterminate result rates for SLIDs vs. FQs) [27][28][29] . However, all dCE SLID-indeterminate results from the dilution series  www.nature.com/scientificreports www.nature.com/scientificreports/ were from the DS-TB strain and there were no indeterminate SLID results on XDR-TB dCEs. On clinical sputum (and falling within our threshold), one MTBDRsl SLID susceptibility result was discordant with sputum (one false-negative). We thus suggest that MTBDRsl Ultra dCE results are interpreted in the same manner as recommended by the WHO for MTBDRsl on clinical specimens 30 . If, for example, MTBDRsl on dCE is non-actionable or susceptible, MTBDRsl on sputum or isolates should be done. If there is still no evidence of resistance in a high burden setting, phenotypic DST should still be done given the suboptimal rule-out accuracy of MTBDRsl 19,30 .
The possibility of contamination from rpoB amplicons during extractions has not been investigated. We implemented systematic testing for possible environmental contamination. No tubes exposed for each extraction batch were rpoB-positive when tested with FT. FT was used for testing for rpoB amplicons as it is more sensitive than MTBDRplus 14,15 .
We further tested a worst-case contamination scenario with dCEs from both Ultra and Xpert cartridges done on a XDR-TB strain, diluting these dCEs, and adding them to a DS-TB strain which was subsequently tested by Ultra. The undiluted and most concentrated dCE dilutions (10 0 , 10 −3 ) showed false rifampicin-resistance indicating that, although the GeneXpert platform does have proven ability to remove large numbers of amplicons 31 , it was not able to remove all amplicons during the pre-amplification wash steps, however, amplicons diluted beyond 10 −3 were successfully removed to the point of not being detected 22,32,33 . These results, together with those from the environmental samplings during extractions, shows that when standard aseptic techniques are used, amplicon cross-contamination is highly unlikely except in the artificial worst case scenarios. Finally, it should be noted that, in line with good practice in any molecular biology laboratory providing results for patient management, dCEs should not be collected in the same room where rpoB-based tests are done, and that the risk of cross-contamination from the dCE approach is only pertinent to tests for rifampicin resistance.
We suggest that diagnosticians considering implementing this approach use the cartridge itself as a transport vessel (upright and in sealed containers) to a central laboratory where dCE can be extracted appropriately (the diamond is a sealed chamber and should remain safe during transport). Most peripheral laboratories will be unable to do the dCE procedure safely and downstream molecular DST like MTBDRsl. This cartridge transport can interface with existing specimen referral networks. If dCE is planned purely for molecular epidemiology, we suggest that dCE be extracted and stored at −80 °C or alternatively the whole cartridge be stored at −20 °C until extractions can be done in a batched, centralised fashion. The long term stability of these approaches will require examination.
We further hypothesised that liquid from other cartridge chambers may avoid interference by rpoB amplicons. However, upon testing, this approach gave variable non-replicable results. This was true for qPCR, MTBDRplus, MTBDRsl and FT assays. This may also be due to very low concentrations of template in these chambers, for example C3 -which is the "wash chamber", and/or remnant PCR inhibitors (e.g., salts from the sample reagent). In light of this, we believe that the presence of these amplicons may prevent newer approaches, such as next generation sequencing methods, from performing well on dCE without to clean up steps. This warrants further investigation. CE from the diamond chamber hence remains the best option for downstream genotypic DST.
The results of this study should be interpreted within its limitations, namely aseptic techniques done in an assay-or procedure-specific biosafety cabinet are needed to minimise amplicon cross-contamination. However, this infrastructure should already be implemented per WHO guidelines 34 where LPAs are done routinely for patient care. Furthermore, per good laboratory practice, CEs should not be collected in the same room where rpoB-or IS6110/1081-based assays are done, nor should either procedure be done by the same personnel on a daily basis. Lastly, further investigation into cross-contamination risk should be done in a routine diagnostic setting. This should include multiple operators.
We also acknowledge that this method may increase risk of needle stick injury. Standard biosafety protocols should be strictly adhered to. We were recently funded to develop a device that can eject material from cartridges in a safe manner. Another limitation is MTBDRplus was not feasible on Ultra CEs and we suspect this is due to interference from both rpoB and IS6110/1081 amplicons. Thus, combined with the large volumes (and hence diluted targeted DNA) recovered from non-diamond chambers in Ultra and Xpert, MTBDRplus (and also likely FT) on extract from any Ultra cartridge chamber is in all likelihood not useful for isoniazid or confirmatory rifampicin DST. Finally, although the diamond chamber is a closed system and appears protected against desiccation, we acknowledge that some desiccation may occur over prolonged periods that this requires future systematic evaluation. However, we recommend that extract method is done on an as fresh a cartridge as possible (either at a peripheral or central laboratory), in order to reduce the delays of DR-TB diagnosis. Formal evaluation of CE stability pre-extraction may be useful.
We conclude that dCEs from Ultra at the C Tmin threshold (<25), can be used for genotypic second-line DST (MTBDRsl). Ultra and MTBDRsl on dCE therefore allows for the rapid rule-in detection of XDR-TB on a single specimen.

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on request.