Introduction

Most recently published gene signatures for diagnosing active TB have been developed on microarray and RNA sequencing platforms, which are expensive and measure expression of the entire transcriptome. Although these platforms are ideal for the discovery of transcriptomic signatures, the size of most signatures that have been discovered, including the 11-gene, 48-transcript ACS signature, makes them impractical for clinical use at a point-of-care level. There is an urgent need to develop these signatures into cheap and easy-to-use diagnostic tools especially for use in resource-poor settings. Cheaper and highly targeted gene expression quantification methods such as qRT-PCR assays can make the signatures more accessible for the development of point-of-care diagnostics. This underlies the motivation for translating the ACS correlate of risk signatures from RNA-seq into qRT-PCR, before they were validated on independent clinical cohorts (Zak et al.

2016).

For translational purposes it is believed that fewer transcripts in a signature will have a higher probability of success when implementing such an assay into a hand-held diagnostic device, due to lower costs, and simplicity. Several recent studies have discovered small signatures in a bid to develop cheaper and simpler diagnostic tests for TB disease (Maertzdorf et al., 2016; Roe et al., 2016; Sweeney et al., 2016).

Sweeney and colleagues discovered and validated a three-gene signature that distinguished active TB disease from Mtb infection in published datasets of both HIV-

infected and uninfected individuals with excellent accuracy (Sweeney et al., 2016).

This 3-gene signature also distinguished active TB disease from other diseases, including sarcoidosis and pneumonia. Maertzdorf and colleagues developed and validated a 4-gene signature that can classify TB disease from healthy persons (Maertzdorf et al., 2016). This signature was compared to a 15-gene signature developed in the same cohort; both signatures were applied to publicly available datasets and could distinguish active TB disease from Mtb infection, also with excellent accuracy. In another study, transcript levels of a single gene, BATF2, allows differentiation between active TB disease and healthy HIV-uninfected persons in individuals with diverse ethnic backgrounds (Roe et al. 2016). The performance of these signatures in diagnosing TB disease suggest that signatures based on a few transcripts may provide equivalent diagnostic performance to signatures with large number of transcripts.

We developed a qRT-PCR signature consisting of six transcripts representing six genes that could predict active TB in adolescents before onset of symptoms and diagnosis in the ACS cohort. The signature was independently validated on the GC6- 74 cohort. Referred to as the “ACS 6-gene signature”, it can be measured using six Taqman primer-probe assays (Penn-Nicholson, unpublished data). The 6-gene signature consists of three transcripts, GBP2, FCGR1B, and SERPING1, that are upregulated in progressors or TB patients (Taqman primer-probe assays:

GBP2.Hs00894846_g1, FCGR1B.Hs0234185_m1, and

SERPING1.Hs00934329_m1, respectively) and three transcripts, TUBGCP6, TRMT2A, and SDR39U1, that are downregulated in progressors or TB patients

(Taqman primer-probe assays: TUBGCP6.Hs00363509_g1,

TRMT2A.Hs01000041_g1 and SDR39U1.Hs01016970_g1, respectively), resulting in nine-transcript pairs (Figure 27). The score for the ACS 6-gene signature is derived from computing the ratio between the nine transcript pairs, where each pair contains one transcript that is upregulated in active TB with one that is downregulated in active TB, relative to healthy controls. This pair-ratio format presents two advantages over other signatures. Firstly, the up-down pairing provides a “self-standardisation”

function that eliminates the need for housekeeper transcript-based standardisation to account for variability in input RNA. Secondly, the pair-ratio format allows a signature score to be calculated even if a Ct value for one primer probe is not derived, due to a failed PCR reaction for example. As observed for the pair-wise ACS 16-gene signature, this provides important robustness to the signature (Zak et al. 2016). The sum of the ratio of the transcript pairs and a coefficient “d” (weighting between the pairs, which was originally derived from model parameterisation on the ACS study) results in a binary score (Table 14). Samples are classified as “progressors or TB cases” or “non-progressors or controls” by each pair based on whether this calculation is greater (progressor/case) or less (non-progressor/control) than zero. A final score is given as the summation of the progressor voting pairs divided by the total number of voting pairs. For example, if six out of the nine pairs vote “progressor or TB case” and all nine pairs voted (a Ct value was obtained for all six transcripts), then the score will be calculated as: (6/9)*100 = 66.7%.

Figure 27 Pair structure of the ACS 6-gene signature. The signature comprises nine transcript pairs that each links a transcript that is upregulated in active TB with one that is downregulated in active TB relative to healthy controls. Lines indicate the pairing of the transcripts. Transcripts that are upregulated in progressors are in red nodes and those that are downregulated in progressors are in green.

TRMT2A

TUBGCP6 SDR39U1

FCGR1B SERPING1 GBP2

Table 14: Taqman primer-probe pairs in the ACS 6-gene signature and coefficients for calculating signature scores.

Primer #1 Primer #2 Coefficient d

GBP2.Hs00894846_g1 TUBGCP6.Hs00363509_g1 -2,3 GBP2.Hs00894846_g1 TRMT2A.Hs01000041_g1 -5,7 GBP2.Hs00894846_g1 SDR39U1.Hs01016970_g1 -4,7 FCGR1B.Hs02341825_m1 TUBGCP6.Hs00363509_g1 2,4 FCGR1B.Hs02341825_m1 TRMT2A.Hs01000041_g1 -1,2 FCGR1B.Hs02341825_m1 SDR39U1.Hs01016970_g1 -0,2 SERPING1.Hs00934329_m1 TUBGCP6.Hs00363509_g1 0,7 SERPING1.Hs00934329_m1 TRMT2A.Hs01000041_g1 -2,5 SERPING1.Hs00934329_m1 SDR39U1.Hs01016970_g1 -1,5

In this chapter, we sought to assess the diagnostic, prognostic and treatment response monitoring performance of the ACS 6-gene signature in HIV-infected persons.