Detection of differentially culturable tubercle bacteria in sputum using mycobacterial culture filtrates

Rapid detection of tuberculosis (TB) infection is paramount to curb further transmission. The gold standard for this remains mycobacterial culture, however emerging evidence confirms the presence of differentially culturable tubercle bacteria (DCTB) in clinical specimens. These bacteria do not grow under standard culture conditions and require the presence of culture filtrate (CF), from axenic cultures of Mycobacterium tuberculosis (Mtb), to emerge. It has been hypothesized that molecules such as resuscitation promoting factors (Rpfs), fatty acids and cyclic-AMP (cAMP) present in CF are responsible for the growth stimulatory activity. Herein, we tested the ability of CF from the non-pathogenic bacterium Mycobacterium smegmatis (Msm) to stimulate the growth of DCTB, as this organism provides a more tractable source of CF. We also interrogated the role of Mtb Rpfs in stimulation of DCTB by creating recombinant strains of Msm that express Mtb rpf genes in various combinations. CF derived from this panel of strains was tested on sputum from individuals with drug susceptible TB prior to treatment. CF from wild type Msm did not enable detection of DCTB in a manner akin to Mtb CF preparations and whilst the addition of RpfABMtb and RpfABCDEMtb to an Msm mutant devoid of its native rpfs did improve detection of DCTB compared to the no CF control, it was not statistically different to the empty vector control. To further investigate the role of Rpfs, we compared the growth stimulatory activity of CF from Mtb, with and without Rpfs and found these to be equivalent. Next, we tested chemically diverse fatty acids and cAMP for growth stimulation and whilst some selective stimulatory effect was observed, this was not significantly higher than the media control and not comparable to CF. Together, these data indicate that the growth stimulatory effect observed with Mtb CF is most likely the result of a combination of factors. Future work aimed at identifying the nature of these growth stimulatory molecules may facilitate improvement of culture-based diagnostics for TB.

The global tuberculosis (TB) epidemic remains a leading cause of death with an estimated 10 million people afflicted with the disease and 1.5 million deaths annually 1 . Moreover, a large proportion of people harbour the infectious agent, Mycobacterium tuberculosis (Mtb), in lung lesions without any symptoms, providing a massive reservoir of the pathogen that can reactivate to cause active disease and drive further transmission. TB is curable however, the 6 month protracted treatment regimen is difficult to implement and challenges related to patient adherence result in the development of drug resistant forms of the pathogen that cannot be eradicated using the current standard first line drugs 2 . Although several clinical trials have investigated novel treatment shortening therapies, published trials have yet to show broad benefit, possibly due to limited understanding of mycobacterial growth states during active and asymptomatic or subclinical disease.
Several studies recently have provided evidence that sputum from TB patients with active disease is characterised by a large fraction of non-replicating bacteria [3][4][5][6] . A commonly used diagnostic method for detection of Mtb in patient sputum is enumerating mycobacterial colony forming units (CFU) of culture dilution suspensions Scientific Reports | (2021) 11:6493 | https://doi.org/10.1038/s41598-021-86054-z www.nature.com/scientificreports/ on solid agar media. However, this methodology is limited to detecting only replicating mycobacterial populations and is unable to account for the non-replicating population, which has important implications for TB diagnostics. This is illustrated by prior work that demonstrated the presence of differentially culturable tubercle bacteria (DCTB) in sputum that do not grow on solid media and are detected only in liquid media supplemented with culture filtrate (CF) 3,6 . This demonstrates that TB infection presents with a much higher level of bacterial phenotypic complexity than previously thought and a deeper understanding of the clinical relevance of DCTB populations is vital for the advancement of improved TB diagnosis and development of novel drug regimens 4 . The growth stimulatory activity of CF has been ascribed to the presence of resuscitation promoting factors (Rpfs), a group of secreted enzymes in Mtb 6,7 however, use of this approach is limited by the availability of freshly prepared Mtb CF in diagnostic laboratories. To investigate alternative approaches for detection of DCTB, we explored the use of CF from Mycobacterium smegmatis (Msm), a non-pathogenic close relative of Mtb, with the hypothesis that Msm CF would provide a more safer and practical approach in routine laboratories. Bioinformatic analysis identified three rpf genes with a duplicate rpfE homolog (rpfA, rpfB, rpfE1 and rpfE2) in Msm compared to the five homologous (rpfA-rpfE) in Mtb 8 ( Figure S1). In prior work, we generated a mutant of Msm devoid of all Rpfs 9 and herein, we exploited this strain to generate complemented mutant strains that carry Mtb Rpfs in various combinations. These recombinant strains were used to generate CF for detection of DCTB in sputum. We found that Mtb CF was superior at detecting DCTB when compared to CF from Msm and whilst the addition of Mtb Rpfs to Msm did improve recovery of DCTB, the bacterial yield was not comparable to that observed with Mtb CF. It has also been demonstrated that fatty acids and cyclic-AMP (cAMP) are able to stimulate the growth of DCTB in an in vitro model of mycobacterial dormancy 10 . Hence, we also tested various fatty acids and cAMP for the ability to resuscitate DCTB and found that no single agent was able to stimulate bacterial growth in a manner comparable to Mtb CF.

Results
Construction of Msm mutant strains expressing Mtb rpf genes. To assess the role of Mtb Rpfs in stimulation of DCTB, we generated strains of Msm that expressed Mtb rpf genes. The inclusion of antibiotics during the culture of these strains to ensure that rpf-expressing plasmids are retained was problematic as the antibiotic would be carried into the resulting CF, leading to inhibition of bacterial growth in sputum cultures. To address this, we used integrating plasmids as these carry an integrase to facilitate plasmid integration into the bacterial chromosome at phage attachment sites 11,12 . However, these integrases can also spontaneously excise vectors in the absence of antibiotic selection. To overcome this, we generated versions of conditionally integrative plasmids that lacked the integrase gene and then co-electroporated these with the suicide plasmid that provided the integrase in trans (Fig. 1). With this approach, the integrase is present during the initial integration event and is then lost during subsequent cell division cycles. We used a mutant of Msm, deleted for all four of the Msm rpf genes to build recombinant strains expressing the different Mtb rpf genes. The resulting strains were then used to generate CF from strains expressing rpf genes from Mtb for use in the MPN assay for the detection of DCTB. Wild type Msm, carrying the Msm Rpfs was also used as a control. The strains generated and predicted complement of Rpf proteins from the resulting CF are shown in Fig. 1.
To confirm that the rpf-like genes from Mtb are expressed in Msm from the modified integrative plasmids, we conducted transcriptional analysis using quantitative real-time PCR. Recombinant Msm strains expressed all Mtb rpf genes provided on integrative plasmids. With the exception of rpfC Mtb , all Mtb rpf genes were expressed at equivalent levels, Fig. 2. As we had used native Mtb promoters, we expected some difference in rpf gene expression in recombinant Msm strains. As the Mtb genes were expressed in Msm, we also assessed expression of Msm rpf-like genes for comparative purposes. The rpfA Msm gene was expressed at the highest level relative to the other Msm rpf-like genes and the rpfE2 Msm gene was expressed at levels similar to the rpfABDE Mtb genes.
We anticipated that both recombinant Mtb and Msm Rpfs would be secreted as analysis of the domain structure of these proteins confirmed the presence of a SecA-dependent signal sequences in all of them, except Mtb RpfC ( Figure S1). We further confirmed the presence of these signal peptides using Signal P analysis, which identified clear Sec-dependent signal peptides in all proteins, except RpfC ( Figure S2). In addition, both SecA1 and SecA2 are conserved in Msm and Mtb ( Figure S3), which when combined with the demonstrated secretion of Rpfs in prior work 13 , suggests that recombinant Rpfs are most likely secreted from the cell and these proteins, or the products of their catalytic activity, are present in the CF used in this study.
Quantification of DCTB in sputum specimens using different mycobacterial CF preparations. To assess the capacity of CF from the various mycobacterial strains generated to detect DCTB, we established a cross-sectional observation cohort of HIV-negative individuals with drug susceptible TB. The participant disposition flow chart for our cohort is given in Fig. 3A. We screened 400 individuals and enrolled 57 individuals. A large number of screen failures were due to co-incident HIV infection and these were excluded from our study due to the low bacterial burden that is expected to prevail in sputum. A summary of participant demographics and laboratory data is included in Table 1. All included participants were sputum smear positive and the majority had medium or high bacterial loads by GeneXpert, with a median time to culture positivity in the MGIT system of 5.5 days.
Sixteen specimens were used for methods optimization as we had contamination of the CF with bacteria, requiring iterative revision of the filtering protocol. Two specimens yielded contamination on CFU plates, 1 specimen gave no Mtb on MPN assays and 2 specimens were negative for MGIT, Smear, GeneXpert, MPN and CFU after enrolment (Fig. 3A). After these exclusions, 36 specimens were tested and analysed for DCTB using the approach outlined in Fig. 4A. As reported previously 3 , we were able to demonstrate that inclusion of Mtb CF resulted in detection of higher number of specimens with DCTB (24/36), together with a higher DCTB quantum The presence of the integrase on mycobacterial vectors that integrate phages at attachments sites, facilitates the first integration event but then can also lead to excision of the vector in the absence of antibiotic selection. For DCTB assays, CF that is free of antibiotic was required as antibiotics will inhibit the growth of sputum-derived bacteria. To facilitate this, the integrase was first removed from vectors expressing Mtb Rpfs and was provided in trans on another suicide vector during transformation. With this approach, the integrase is present for the first integration event and is then lost as it does not carry a mycobacterial origin of replication.    . When compared to the no CF control, inclusion of CF from wild type Msm in MPN assays did not yield a comparable increase in DCTB quantum and appeared to be indistinguishable from assays with no CF. The addition of RpfCD Mtb to Msm yielded a significant increase in the mean quantum of DCTB, Fig. 4B, but this was not associated with detection of DCTB in more specimens when compared to the no CF control (both yielding 13/36 DCTB positive specimens). Following this, we next asked if CF derived from Msm containing only Mtb Rpfs was able to stimulate growth of DCTB, as Msm would be a more tractable source of Rpfs in the diagnostic setting. For this, we used our previously reported mutant of Msm deleted for all four Msm rpf-like genes 9 , Msm Δrpf, and inserted Mtb rpf genes into this genetic background in different combinations. We found that only the addition of RpfAB Mtb and RpfABCDE Mtb to Msm Δrpf yielded a marginal but statistically significant increase in the quantum of DCTB recovered, together with a statistically non-significant increase in the number of positive specimens detected (15/36 for both compared to 13/36 for the no CF control). However, whilst the mean DCTB values for CF containing RpfAB Mtb and RpfABCDE Mtb were higher than the empty vector control (mean DCTB yield of 0.8 log and 0.7 log for RpfAB Mtb and RpfABCDE Mtb compared to 0.6 log for the empty vector control), the difference was not statistically significant. Addition of RpfCD Mtb and RpfDCE Mtb to the Msm Δrpf mutant did not yield CF preparations that were able to stimulate more DCTB when compared to the no CF control (Fig. 4B).

Scientific Reports
Data from our heterologous complementation experiments suggested that addition of Mtb Rpfs to Msm does not yield growth stimulatory effects comparable to that seen with the Mtb CF, suggesting that other factors could also contribute to the growth stimulatory effect. To further investigate this, we compared DCTB yields using Mtb CF derived from wild type Mtb or a mutant defective for the five rpf genes. There was no statistical difference in the quantum of DCTB detected with CF derived from these two strains.
Assessment of the ability of fatty acids and cyclic-AMP (cAMP) to resuscitate DCTB. We next set out to evaluate whether other molecules present in the CF are able to contribute to the growth stimulatory effect seen with sputum specimens and tested the ability of free fatty acids and cAMP to stimulate DCTB. For this component of the study, we obtained sputum (n = 21) from a second collection of patients with drug susceptible TB of whom, 62% were HIV-coinfected ( Table 1). The participant disposition flow chart and demographics, together with laboratory data, are shown in Fig. 3B and Table 1 respectively. Twenty specimens were smear positive, the majority had high or medium bacterial load as assessed by GeneXpert, with a median time to positivity on MGIT culture of 6 days. Of these, 4 specimens were used to interrogate the effect of cAMP and select fatty acids, whilst the remainder were used for further analysis of a broader range of fatty acids.
In prior work, cAMP appeared to facilitate recovery of DCTB from an in vitro model of differential culturabilty, together with fatty acids such as arachidonic acid, whereas addition of stearic acid did not yield the same effect 10 . Hence, we first tested these agents (as outlined in Fig. 5A) in a small selection of sputum specimens and found that neither cAMP nor arachidonic acid facilitated recovery of DCTB in a manner comparable to Mtb CF (Fig. 5B). We also noted that stearic acid did not enhance recovery of DCTB (Fig. 5B). Following this, we opted to test a broader range of fatty acids, with the hypothesis that sputum resident organisms may respond to different stimuli when compared to differentially culturable bacteria generated in vitro. We chose a variety of www.nature.com/scientificreports/ chemically distinct molecules and first optimised the concentrations using the previously reported in vitro model of M. smegmatis differential culturability 14 . The concentration that yielded the best bacterial recovery (detailed in Table S1) were then applied to sputum DCTB assays (Fig. 5A). Certain fatty acids such as palmitoleic, petroselenic and linolenic acid yielded a marginal, statistically non-significant increase in bacterial recovery when compared to media alone (Fig. 5C). However, no fatty acid yielded growth stimulation in a manner comparable to CF from Mtb. We noted that in this collection of sputum specimens, the yield of DCTB with CF from Mtb was higher when compared to yields shown in Figure 4B. As these specimens were collected from two distinct cohorts of individuals, at different times, these differences may relate to sputum collection procedures or individual characteristics of the participants.  Statistical comparisons were conducted between CFs and the no CF or cAMP/fatty acid assays. There was an increase in bacterial recovery from MPN assays containing CF, which was not significant, most likely due to the small samples size (n = 4), no significant increase in DCTB recovery was noted in the media control and cAMP or fatty acid assays. Error bars depict the standard error of the mean. (C) Histogram depicting mean DCTB counts from assays using CF from Mtb and standard media supplemented with fatty acids. The mean DCTB counts were compared to that obtained from the media controls assay using a Wilcoxon test to determine the growth stimulatory affect of fatty acids. P-values depict the comparison of the CFs to the no CF or fatty acid assays. Graphs were generated using Graphpad Prism software, Version 6.

Discussion
A growing number of studies point to the presence of a differentially culturable population of tubercle bacteria in sputum which require CF to grow in liquid or on solid media 3,5,15,16 , this growth stimulatory effect of CF has been ascribed to the presence of Rpfs 6,15 . A single Rpf was identified in Micrococcus luteus as an essential secreted protein able to stimulate the growth of dormant bacterial cultures and represented a group of bacterial growth modulators 17 . Subsequent studies have highlighted a role for these proteins in exit from dormancy in vitro and during virulence and reactivation in the murine model of TB infection [18][19][20][21] . It has been shown that these proteins exert their effects through breakdown of the peptidoglycan component of the cell wall, leading to the release of growth stimulating muropeptides suggesting that the growth stimulatory effect can be applied to bacterial cultures in a paracrine manner, through supplementation of growth media with the Rpf proteins themselves 22,23 . Consistent with this, it has been demonstrated that mycobacterial Rpfs are secreted into the culture media hence, making the CF amenable for the identification of DCTB 13 .
In a previous study we demonstrated the utility of CF derived from Mtb to detect DCTB in individuals from whom the sputum specimen yielded negative cultures using standard techniques 3 . This suggested that CF-supplementation in the diagnostic setting could improve TB detection in difficult to diagnose cases with paucibacillary TB however, use of Mtb CF is limited by the requirement to perform these processes in containment laboratories and need for expert skills to culture the pathogen with subsequent separation of the CF from the bacteria. In this context, Msm provide a tractable avenue to generate CF but it remained unclear if this would yield a similar growth stimulatory effect. As the genome of Msm lacks homologues for rpfC and rpfD, we first added the rpfC Mtb and rpfD Mtb genes into the genome of wild type Msm, to create a CF that was comparable to that derived from Mtb. Thereafter, we tested various combinations of Mtb rpf-genes in the Msm Δrpf mutant for growth stimulatory capabilities. None of the CF preparations from Msm yielded growth stimulatory effects that were comparable to CF from Mtb.
The addition of CF containing RpfAB Mtb and RpfABCDE Mtb yielded a small but significant increase in recovery of DCTB and given that addition of CF with RpfCDE Mtb did not yield comparable effects, the increased growth stimulatory effect by the former CF may be due to RpfAB Mtb . These results should be interpreted with caution as in cases where addition of select Mtb Rpfs appeared to significantly increase the yield of DCTB compared to the no CF control, these differences were not significantly different when compared to CF from the no vector control strain. In our prior work, we demonstrated that the combinatorial effect of RpfA and RpfB from Msm played an important role in biofilm formation 9 , suggesting that these two proteins may have an essential function in bacterial communication.
In addition to the Rpf proteins, cAMP and fatty acids have also been reported to stimulate the growth of dormant Msm 10 . These factors in CF, together with other as yet unidentified molecules may work in concert to allow for paracrine stimulation of DCTB in sputum specimens. We tested the effect of cAMP and a variety of fatty acids and whilst these appear to differentially facilitate recovery of differentially culturable Msm in vitro, they had no benefit in recovery of DCTB in sputum. When combined with the observations on recombinant CFs (with and without Mtb Rpfs), our data suggests that the growth stimulatory effect observed in sputum specimens cultured with Mtb CF is most likely the result of a combination of factors. Further work to identify the interplay between these factors and how they enhance bacterial growth may yield useful new supplements for diagnostic media and would facilitate the development of faster culture-based diagnostics for TB.

Materials and methods
All methods were performed in accordance with the relevant guidelines and regulations for growth of Mtb and handling of human specimens. All procedures were conducted in a BioSafety Level III laboratory, registered with the South African Department of Agriculture Forestry and Fisheries (Registration Number: 39.2/NHLS-20/010). All procedures were approved by the Institutional BioSafety Committee of the University of the Witwatersrand (approval number: 20200502Lab).
Bacterial strains and culture conditions. Bacterial strains and plasmids used in this study are listed in Table 2. E. coli and M. smegmatis strains were grown as previously described 9 .

Cloning of shuttle plasmids for integration into M. smegmatis. To prevent spontaneous excision
of the integrated vector in the absence of antibiotic selection, previously generated integrating vectors carrying the rpf genes 21 were manipulated to remove the integrase gene. Plasmid pMV-RPFAB was digested with MluI and BstBI and the 6313 bp blunt ended fragment was cloned into pTTP1B at the Acc65I site to generate pTT-RPFAB. The resulting plasmid was subsequently partially digested with NcoI, and the 10,937 bp fragment was circularised to yield plasmid pTT-RPFABΔi (Table 2). Plasmids pH-RPFCD and pH-RPFCDE 21 were digested with HindIII/BamHI and HindIII/BsrGI respectively and the resulting 9496 bp and 10,907 bp fragments were blunt ended and circularised to yield plasmids pH-RPFCDΔi and pH-RPFCDEΔi ( Table 2). The empty vector control, pHINT-attP-hyg, containing no rpf genes was generated by digesting pHINT with HindIII/NdeI and the resulting 4781 bp fragment, was blunted and circularised ( Table 2). The integrase protein was provided in trans, by generating suicide plasmids from pHINT and pTTP1B which lack the phage attachment sites attP L5 and attP Tweety , respectively ( Table 2). The pHINT was digested with EcoRI/ NdeI, and pTTP1B was digested with HindIII and partially with AccI to generate the 4204 bp fragment and the 4533 bp fragment respectively. These fragments were blunt ended and circularised to yield plasmid pHINT-int-bla and pTTP1B-int-bla (   Table S1.

Gene expression analysis by real time, qRT-PCR.
To monitor the expression of the rpf genes, all recombinant strains were grown to early exponential phase (OD 600nm 0.6). Gene expression analysis was conducted as previously described 9 using primers indicated in Table S2.
Recruitment of participants to obtain sputum specimens. Ethics approval for participant recruitment was obtained from the Human Research Ethics Committee of the University of the Witwatersrand, South Africa, with clearance number M110833, subsequently revised to M200164 (SNT study). Individuals 18 years and older attending the Perinatal HIV Research Unit (PHRU), Soweto, South Africa were approached for participation into the study. Fifty seven HIV sero-negative patients with drug sensitive TB, as determined from either an auramine stained smear or by GeneXpert obtained from the public sector (National Health Laboratory Service, Johannesburg, South Africa) were eligible for enrolment in the study. Patients willing to participate were approached and written informed consent was obtained using Informed Consent Forms reviewed and approved by the Human Research Ethics Committee of the University of the Witwatersrand. After written informed consent was granted, a spot sputum was collected for analysis by the most probable number (MPN) assay (otherwise referred to as limiting dilution assays) as described previously 3 . The MPN assay is a limiting dilution series based on a Poisson distribution for the quantification of bacterial growth in liquid media, described in further detail in the Supplementary Information. Sputum samples were serially diluted in a 48 well microtitre plate in media with Optimization of fatty acid concentrations to use in sputum DCTB assays. The concentrations of fatty acids for use in sputum were optimised using an in vitro model of dormancy that generates differentially culturable Msm as previously described 14 . Msm was pre-cultured in 20 ml of nutrient rich broth for 16 h at 37 °C in an orbital shaker (250 rpm). Pre-cultures were sub-cultured into modified Hartman's-de Bont medium. Trace elements solution was prepared as previously described 24 . The OD 600nm of the pre-culture was adjusted to 0.9 and 20 ml was added to 130 ml of the Hartmans de Bont media and incubated at 37 °C for 14 days without shaking. The unsaturated fatty acids kit (Sigma Catalogue number UN10-1KT) was used for the resuscitation of differentially culturable Msm. Each fatty acid was initially resuspended in 1 ml of dimethyl sulfoxide to create a working stock. The starved Msm was added to Middlebrook media (supplemented with ADC-no oleic acid was used) and the OD 600nm adjusted to 0.1. Thereafter, 9 ml aliquots of these cells was used to test the different concentrations of fatty acids. Fatty acids were tested at the following concentrations: 0.125 µM, 0.25 µM, 0.5 µM, 1 µM, 2.5 µM, 5 µM, 7.5 µM, 10 µM, 12.5 µM, 15 µM, 17.5 µM, 20 µM, 22.5 µM, 25 µM, 27.5 µM, 30 µM and 32.5 µM.
The OD 600nm of each culture was assessed after 72 h and the bacterial recovery was measure as the fold increased in OD 600nm compared to the inoculum. For cAMP (3 mM), steric acid (4 µM), arachinodic acid (1.6 µM) and Linoleic acid (1.7 µM) the concentrations used were derived from previous similar work 10 .