Move over culture Rapid and accurate quantification of Mycobacterium tuberculosis in sputum samples takes just one day Centre for Clinical Microbiology meeting Wednesday 11 th January 2012 Dr Isobella Honeyborne

Culture of Mycobacterium tuberculosis to measure response to treatment Solid culture 7H10/7H11 is not a gold standard Sputum serial colony counts (SSCC) – used in clinical trials to measure response to treatment Contamination with other organisms Viable, non-culturable bacteria not detected Long assay turnaround time Clumping of bacilli may mean 1 colony does not mean 1 bacilli Not practical in a routine laboratory setting Expensive Selective antibiotics reduce count by 1 log 10 Automated liquid culture TTP can be used to estimate bacterial number – decontamination process kills 1 log 10 of the bacteria present. Suffers from most of the above problems

What would be useful for clinical trials? A method for quantification of M. tuberculosis bacilli in sputum samples that is: Culture-free Rapid Relatively cheap.

What’s available and why hadn’t quantitation been developed previously? Molecular techniques for detection of M. tuberculosis rRNA & DNA are used diagnostically but do not quantify. GeneXpert (Cepheid) recently launched – 1 st automated platform for TB molecular diagnostics. DNA is not a measure of live bacteria RNA for quantification – seems like a useful candidate.

Inhibitors & variable RNA loss  Sputum samples vary in the inhibitors carried forward and the efficacy of the RNA extraction THEREFORE  Specific gene expression cannot be used to quantify directly  mRNA is not very adundant

Assay specifications We used a duplex RT-qPCR assay with dual labelled probes to detect 16S rRNA and the novel internal control. 16S rRNA is a very abundant molecule RT and qPCR in same tube - so add RNA directly into reaction – performs RT and then straight into the qPCR.

Internal Control To account for RNA loss and inhibitors:  Developed a robust internal control (IC). Internal control detection following RNA extraction and effect of inhibitors MUST correlate with the M. tuberculosis gene of interest.

Correlation between internal control and M. tuberculosis and 16S rRNA 51 M. tuberculosis-negative sputa spiked with 10 7 M. tuberculosis bacilli and 50ng I957bp internal control 15 sputa divided into 1mL aliquots and spiked with a serial dilution of bacilli. c Bacilli mL -1 sputum < 10 2 Mean CT  SD 6.6  0.9* † 9.6  0.9 † 12.6      1.3 >27.2

Accuracy of assay Based on Mean CT at each dilution  STD DEV: 100% (n=86) spikes were calculated within 1 log 10 of original bacilli number ( ) 98% (84/86) spikes were within 0.5 log 10 ( ). Detect bacilli by this assay with fold inhibition/RNA loss (compared to the best extraction)

Bacterial decline as measured by MBL Mean 1 log 10 decline in first 3 days of treatment Those that later relapse are associated with higher bacterial load pre- treatment Every 1 log 10 increase in baseline bacterial load was associated with 3.62 increased odds ratio of relapse.

Mathematical modelling of data Solid culture studies – found biphasic decay in bacterial load – reflecting killing of different populations of bacteria

The way the bacteria respond to drugs in the first 3 days could tell us something about the site of the bacteria in the lung and the predominant bacterial phenotype – persister/actively dividing. Different populations of bacteria in a sample?

Assay uses Could be used to measure response to novel M. tuberculosis antibiotics in EBA-type trials. Monitor longitudinal treatment response in clinical trials. Maybe able to develop a relapse hazard analysis based on presenting bacterial load. Modify to detect M. tuberculosis in paediatric stool and gastric aspirate.

Summary Developed a robust assay for measuring true bacterial load, within 0.5 log 10, in sputum samples. A result can be obtained on the same day as expectoration rather than waiting 3 weeks for solid culture-based assays. Mathematical modelling found that rRNA in vivo does have a half-life reflective of bacterial decline in response to treatment. Assay may be able to identify patients who are at risk of relapse by ascertaining their bacterial load on day 0

Acknowledgments Gerhard Walzl Stephen Gillespie Katharina Ronacher-Mansvelt Timothy D McHugh Paul Van Helden Felicity Perrin Nora Carroll Anna Bateson Selina Bannoo Laura Wright Funders