Update on tuberculosis biomarkers: From correlates of risk, to correlates of active disease and of cure from disease

Update on Tuberculosis Biomarkers: from correlates of risk, to correlates of active disease and of cure from disease

Delia Goletti^a*, Meng-Rui Lee^b, Jann-Yuan Wang^c, Nicholas Walter^d and Tom HM Ottenhoff^e.

aTranslational Research Unit, National Institute for Infectious Diseases “L. Spallanzani”, Department of Epidemiology and Preclinical Research, Rome, Italy;

bDept Internal Medicine, National Taiwan University Hospital, Hsinchu Branch, Hsinchu, Taiwan

cDept Internal Medicine, National Taiwan University Hospital, Taipei, Taiwan

dDenver Veterans Affairs Medical Center, Denver, USA

eDept Infectious Diseases, Leiden University Medical Center, Leiden, Netherlands

*Corresponding author: Delia Goletti, M.D., Ph.D.

Translational Research Unit,Istituto Nazionale per le Malattie Infettive “L. Spallanzani”, Via Portuense 292, 00149 Rome Italy; E-mail address: delia.goletti@inmi.it. Fax: (+39) 06-5582825; Tel: (+39) 06-55170-906

r d d

Twite han le (optonal-): @ eliagolet2 Words: abstract, 234. Text, 4885;

AUTHORS’ CONTRIBUTIONS: (add authors’ names or initials for each item)

1. Conception of the review and/or literature search and/or critical analysis of the literature: DG, MRL, JYW, NDW, THMO

2. Drafting and/or critical revision of the manuscript: DG, MRL, JYW, NDW, THMO 3. Other, please specify:

DG: wrote several parts of the manuscript, edited and integrated the overall text and the different sections.

MRL wrote mainly the part related to the biomarkers for tuberculosis treatment response: host-based. JYW wrote mainly the part related to the biomarkers for tuberculosis treatment response: host-based. NDW wrote mainly the part related to the biomarkers for tuberculosis treatment response: pathogen-based. THMO wrote several parts of the manuscript, edited and integrated the overall text and the different sections.

4. Final approval of the manuscript before submission (all authors should give final approval of the version of the manuscript submitted and its revisions): DG, MRL, JYW, NDW, THMO

Delia Goletti, MD, PhD is an Infectious Diseases specialist and Head of the Translational Research Unit. She is at the Dept. of Clinical Epidemiology, National Institute for Infectious Diseases L. Spallanzani, Italy. Her research interests are on biomarkers for tuberculosis and cystic Echinococcosis, autophagy in tuberculosis.

Dr. Meng Rui Lee, MD is a chest specialist at the National Taiwan University Hospital, Hsin-Chu Branch and his primary research interests are in mycobacterial lung diseases.

Jann-Yuan Wang, MD: he is currently a professor in the Internal Medicine, College of Medicine, National Taiwan University and a pulmonologist in the National Taiwan University Hospital. His research interests are mycobacterial infection and latent tuberculosis infection.

Dr. Nicholas D. Walter, MD is a pulmonary physician at the Denver VA Medical Center who conducts translational research in mycobacterial genomics.

Tom H. M. Ottenhoff, MD, PhD is Professor in Immunology and Head of the Lab of Infectious Diseases He is at the Dept. of Infectious Diseases, Leiden University Medical Center.

ABSTRACT

Tuberculosis (TB) remains a devastating disease, yet despite its enormous toll on global health, tools to control TB are insufficient and often outdated. TB Biomarkers (TB-BM) would constitute extremely useful tools to measure infection status and predict outcome of infection, vaccination or therapy. There are several types of TB-BM: Correlates of Infection; Correlates of TB disease; Correlates of increased risk of developing active TB disease; Correlates of the curative response to therapy; and Correlates of Protection. Most TB-BM currently studied are host derived BM, and consist of transcriptomic, proteomic, metabolomic, or cellular markers or marker-combinations (“signatures”). Vaccine inducible Correlates of Protection in particular are expected to be transformative in developing new TB vaccines since they will de-risk vaccine R&D as well as human testing at an early stage. In addition, CoP could also help minimizing the need for preclinical studies in experimental animals.

Of key importance is that TB-BM are tested and validated in different well-characterized human TB cohorts, preferably with complementary profiles and geographically diverse populations: genetic and environmental factors such as (viral) co-infections, exposure to non-tuberculous mycobacteria, nutritional status, metabolic status, age (infants vs. children vs. adolescents vs. adults) and other factors impact host immune set points and host responses across different populations.

In this paper we will review the most recent advances in research into TB-BM for the diagnosis of active TB, risk of TB development, and treatment induced TB cure.

r

Sho t ttle :Tube culosis bioma ke sr r r rd

Keywo s: Tube culosis, bioma ke s, latent infecton, actve isease, cu e, incipient tube culosis.r r r d r r

1. Introduction

Tuberculosis (TB) and HIV are leading causes of mortality and morbidity worldwide ¹. In 2016, 1.7 million TB deaths were reported. This number of TB deaths is unacceptably high because with prompt diagnosis and appropriate treatment, almost all TB patients can be cured. Therefore efforts to fight TB must be increased.

After exposure to Mycobacterium tuberculosis (Mtb), an estimated 20-25% of those exposed is thought to become infected. One-fourth of the world’s population is latently Mtb-infected and approximately 3 to 5% of these infected individuals will progress towards developing active TB disease during their lifetime ² (Figure 1).

The risk of reactivation and subsequent disease is significantly increased in individuals with: HIV coinfection ³; therapy with TNF inhibitors ⁴; type-2 diabetes ⁵ and other states of relative immunosuppression (e.g.

transplant patients, the elderly)⁶.

Accurate and fast TB diagnosis can be difficult, due to several limitations in the currently available diagnostic tests. Smear microscopy is mostly available and highly specific but suffers from poor sensitivity ³. Mycobacterial culture, the gold standard, has a long turnaround time (up to 42 days) ^{7, 8}, whereas the recently developed GeneXpert MTB/RIF test (Cepheid Inc., Sunnyvale, USA), although rapid, is expensive, cannot distinguish live from nonviable Mtb and has technical limitations hampering its worldwide use in resource-poor countries.

All these diagnostic assays, in the case of pulmonary TB, require a Mtb-positive sputum even though many active TB patients, including HIV-coinfected individuals, diabetes patients, and children, often do not present with Mtb positive sputum and thus cannot provide microbiologically positive specimens ⁵. Moreover, in extra- pulmonary disease, sputum-based diagnostics are not useful, and diagnosis relies on samples (tissue or biological fluids as pleural-, cerebral-, synovial-fluids) often collected by invasive procedures. For all of these reasons, there is need to develop highly sensitive and specific diagnostic tests for TB to rapidly identify − or rule out − the presence of active disease.

2. Biomarker Target Product Profiles

The World Health Organization (WHO) recently defined high-priority target product profiles (TPPs) for TB diagnostics ^{9, 10}. These include a rapid non-sputum-based test for detecting TB with the purpose of starting specific TB-therapy on the same day. These tests need to perform in endemic settings with limited laboratory facilities, at low cost, using easily accessible non-sputum based samples such as (finger prick)-blood, urine, or breath. ^{9, 11-16}. Therefore it is urgent to search for biomarkers (BM) that could be used in such tests. Disease- related TB-BM could find application in improved and fast clinical decision making, for example in developing improved tests that more accurately and differentially diagnose TB disease. Conversely, even though not yet identified, future TB-BM of protection could be important to inform and guide TB vaccine research and development by identifying lead vaccine candidates with the most promising performance at an early stage, and allow their selection for further analyses in animal model testing and human clinical evaluation.

There are several types of TB-BM, summarized in Table 1 and Figure 1. TB-BM indicate whether a given host is -or following vaccination has become-: immune (Correlate of Protection or CoP); has TB disease (Correlate of TB disease or CoD); has an increased risk of developing active TB disease (Correlate of Risk or CoR). Most TB-BM currently studied are host derived BM, and can consist of transcriptomic, proteomic, metabolomic, or cellular markers or marker-combinations (“signatures”). Vaccine inducible CoP in particular are expected to be transformative in developing new TB vaccines since they will not only de-risk selection of candidate TB vaccines for human efficacy testing at an early stage, but also help reducing the time and costs involved in large scale human efficacy clinical studies by measuring vaccine immunogenicity and potential efficacy. In addition, CoP could help minimizing the size of, and need for preclinical studies in experimental animals. Once discovered, TB-BM need to be tested and validated in different well-characterized human TB cohorts, preferably with complementary profiles and diverse populations: genetic and environmental factors such as (viral) co-infections, exposure to non-tuberculous mycobacteria, nutritional status, metabolic status, age (infants vs. children vs. adolescents vs. adults) and other factors could impact host immune set points and host responses across different populations.

In this paper we will review most recent advances in research into TB-BM for the diagnosis of active TB, for risk of TB development, and for treatment induced TB cure. For pragmatic reasons, we will distinguish BM related to pathogen vs. host (Figure 2).

3. BM for diagnosing active TB

3A. BM for diagnosing active TB: the pathogen side

Mtb products can be detected directly in blood, sputum or urine. Urine or saliva samples are interesting because sample collection is non-invasive, and sample volumes are relatively large. They are therefore increasingly being used for diagnosis of infectious diseases ^17-21. In particular, urine seems particularly interesting because, as an ultra-filtrate of blood, it is potentially representative of molecules originating from all organs, including the lungs.

Mtb DNA can be detected in blood and urine of pulmonary TB patients with better sensitivity than Mtb culture from the same fluid ²². The Mtb cell-wall component lipoarabinomannan (LAM) has been used to set up a urine test in HIV-infected patients and seem useful in those with low CD4 T-cell counts ^23-25. Mtb Ag85 protein complex is a mycobacterium-specific 30-32 kD family of three proteins (Ag85A/Ag85B/Ag85C) with enzymatic mycolyltransferase activity involved in coupling of mycolic acids to the cell-wall arabinogalactan and in the biogenesis of Mtb cordfactor. The detection of Ag85 in blood and urine is promising, although improvements are needed ^{26, 27}.

Recently a method has been developed to rapidly quantify Mtb-specific peptide fragments of ESAT-6 and CFP- 10 in sera using antibody-labeled and energy-focusing porous discoidal silicon nanoparticles (NanoDisks) and high-throughput mass-spectrometry (MS) ²⁸. The absolute quantification of serum Mtb-antigen concentration

was informative for TB diagnosis and monitoring the effect of anti-mycobacterial treatment in both HIV- uninfected and HIV-infected patients.

In urine, proteomics offer important tools for the discovery of diagnostic and/or prognostic BM. Validation studies are underway to confirm these candidate BM of TB disease ²⁹.

3B. BM for diagnosing active TB: complex host expression signatures

The first application of large scale whole genome transcriptomic analyses in TB was on identifying signatures that discriminated diseased patients from LTBI subjects ³⁰. In this study ³⁰ a strong increase in type I and type II IFN-signaling blood transcripts was found in active TB. This signature originated mostly from neutrophils and macrophages, and dominated the 393 gene signature that was found to discriminate TB vs. LTBI. This seminal finding was confirmed in numerous subsequent genome-wide studies (e.g. ^31-36)

A second important study showed clear differential diagnostic performance of a micro-array based TB diagnostic signature in two different African cohorts, which moreover validated well in published data base cohorts³³. Of particular importance, this CoD signature also worked well in HIV co-infected individuals, and thus may be helpful as a triage test in point-of-care settings, to rule out TB in those with other lung diseases, and select only the relevant individuals for further TB-diagnostic work-up. Another study found a CoD signature in TB-affected children ³¹. Since children are mostly TB culture negative, using blood based CoD signatures may aid in correct diagnosis and subsequent treatment, thus potentially contributing to TB control in these vulnerable population.

Importantly recent studies identified rather small (n=3 or 4) gene signatures differentially diagnostic of active TB ^{37, 38}. If replicated and validated, these signatures should allow now translation to qPCR based simple tests with the potential to find their way into the clinic.

A first metabolomic CoD signature of active TB has been described, with 400 metabolites measured including amino acids, lipids, nucleotides and their metabolic pathways. The results revealed a link between metabolic profiles and cytokine signaling in active TB and allowed the development of a promising diagnostic algorithm based on results from 20 metabolites ³⁹. The study needs a validation. Moreover, it will be of interest, by analogy of the diagnostic value of some CoR, to determine whether this metabolomic CoD might also have predictive value as CoR. If so, this might add another possible technology platform to the armentarium of predictive CoR of developing TB.

Immune proteomic approaches have also been followed to identify potential TB CoD. Luminex-like platform based screening of several dozens of circulating analytes in the plasma or sera of TB cohorts has been used by several groups. Compared to measurements directly in sera, Mtb-antigen-induced profiles (in cell or whole blood supernatants) led to less discriminatory results. Interestingly, a TB-biosignature based on the relative levels of seven serum markers (C-reactive protein, transthyretin, IFN-γ, complement factor H, apolipoprotein- A1, inducible protein-10 and serum amyloid A) seems promising ⁴⁰. Efforts are now ongoing to incorporate these into simple to use and field friendly lateral-flow based point-of-care tests ⁴¹. In agreement with the overall emerging theme in the field of TB CoD and CoR, these markers are mostly representative of innate immunity and inflammation, confirming the above transcriptomic and immune signatures.

A recent high end TB-BM development has been the application of combined positron emission and computed tomography (PET/CT) scanning to measuring pulmonary inflammation and disease progression in TB and LTBI ⁴². One study in HIV-1-infected LTBI showed that almost 30% had pulmonary abnormalities compatible with subclinical active infection; indeed, these individuals had an increased risk of progressing to clinical disease. Thus PET/CT scanning captures a clinically relevant CoR. Although not qualifying as a routine diagnostic test platform, the emerging new insights from this work further support the concept of inflammation as a significant risk factor for TB progression⁴³.

3C. BM for diagnosing active TB: immune protein host signatures in noninvasive specimens Besides the more complex profiles identified above, also several single markers have been investigated as single TB-BM. IFN--inducible protein-10 (IP10) was reported to be increased in the unstimulated plasma of children and adults with active TB 22, 41, 44-49 . IP-10 levels have been evaluated by different methodologies including classical as well as innovative translational lateral-flow technologies based on interference-free, fluorescent upconverting phosphorlabels, in multicenter studies in Africa ^{41, 47}. Interestingly, IP10 was also detected in the urine of adult patients ²² and Ugandan children with active TB ⁵⁰, and IP10 levels decreased following efficacious therapy ²². As has been recently suggested, however, IP-10 should probably be considered as a marker of general inflammation ^{51, 52}.

As mentioned the use of urine or saliva offers the advantage of non-invasive and ease to handle sampling. It has been reported that saliva levels of IL-1β, IL-23, ECM-1, HCC1 and fibrinogen are significantly associated with active TB ^{53, 54}. One study ⁵⁵ showed that some host inflammatory BM were expressed at much higher concentrations in saliva than in the blood.

3D. BM for diagnosing active TB: immune T cell signatures

Flow-cytometry has been proposed as a potential tool to help improving TB diagnosis. Polyfunctional T-cells coproducing IFNγ, TNF and IL2 have been associated with protective T-cell responses in HIV non- progressor subjects ⁵⁷. Studies evaluating the role of polyfunctional T-cells in TB have shown some discrepant results, however. Active TB has been associated with either triple functional IFN⁺TNF⁺IL2⁺ CD4⁺T-cells ⁵⁸, double functional IFN⁺TNF⁺ CD4⁺T-cells ^{59, 60} or monofunctional TNF⁺CD4⁺T-cells⁶¹. Instead, studies measuring activation and memory status of Mtb-specific T-cells gave more consistent results: effector T-cells expand during active Mtb replication, whereas (central) memory T-cells are associated with infection control ^59,

60, 62, 63. In particular, active TB was associated with decreased CD27-surface expression on circulating Mtb- antigen stimulated CD4+ T-cells ^{59, 64-66}. Recently, a novel T-cell activation marker-TB (TAM-TB) assay was

described for the diagnosis of active TB in children ⁶⁵. This TAM-TB assay was validated in an adult population in Tanzania and is based on the ratio of the median fluorescence intensity of all CD4⁺CD27⁺ T-cells over that of Mtb-specific CD4⁺CD27⁺ T-cells (CD27 MFI ratio). This approach has also been tested in an adult population from a low TB endemic country, confirming discrimination between different TB stages ⁶⁴.

Another interesting blood-based study showed that the expression of immune activation markers CD38 and HLA-DR and proliferation marker Ki-67 on Mtb-specific CD4⁺ T-cells was associated with Mtb load. The modulation of these markers accurately distinguished active from LTBI with 100% specificity and over 96%

sensitivity ⁶⁷.

Besides T-cell activation status also other characteristics of circulating human T-cells have been assessed and related to TB disease. For example, high TNFα high producing CD4 and CD8 T-cells are often found in active TB, and qualify as an antigen specific CoD ^{61, 68}. Combined analysis of Mtb-specific CD4 and CD8 T-cell responses was a powerful diagnostic tool for diagnosing TB ⁶⁹. Additional immune profiling studies focused on a TB CoD already identified long ago, namely the ratio of myeloid over lymphoid cells in the blood of patients

70-72. The cumulative findings of these older and more recent studies is that elevated levels of circulating monocytes are an independent CoR for developing active TB in patients’ contacts. It has been postulated that this may be related to enhanced myelopoiesis in response to IFN produced by other cells, which is released in the circulation during TB infection and inflammation ^70-72. This has been confirmed in active TB ⁷³.

3E. BM for diagnosing active TB: serology and volatile compounds

Serological tests such as lateral flow assays are appealing for TB diagnostics due to their simplicity, lack of specimen processing requirements, and potential towards implementation. However, the accuracy of serodiagnostics for TB has been disappointing, leading the WHO to issue a strong recommendation against the use of serological tests commercialized for active TB diagnosis ^{74, 75}. In the WHO policy statement, further research was encouraged, specifically with representative populations with presumptive TB in studies with

prospective follow-up in blinded study designs ⁷⁴. Recently an assay based on antibodies to the trehalose esters of mycolic acids, which are Mtb-specific cell wall glycolipids was reported . This assay showed promising accuracy with a sensitivity and specificity of 87% and 83%, respectively ⁷⁶.

Interestingly, a multiplex TB serology panel using microbead suspension arrays containing a combination of 11 Mtb antigens was shown to have an overall sensitivity of 91% in serum/plasma samples from culture-proven TB patients. Test specificity was 96% compared to chronic obstructive pulmonary disease patients and 91%

compared to the healthy population ⁷⁷. Further studies are needed to validate these results. Finally, analysis of sera for activation markers showed that transthyretin, C-reactive protein and neopterin might discriminate TB patients from subjects with other infectious and inflammatory conditions with high accuracy ⁷⁸, agreeing well with the markers mentioned already in section 3B.

Volatile organic compounds (VOCs) in breath may represent novel BM of pulmonary TB, emanating directly from the pathogen (e.g. metabolites of Mtb) or from the infected host (e.g. products of oxidative stress). A breath test based on the detection and quantification of VOCs identified pulmonary TB with high accuracy in symptomatic high-risk subjects ⁷⁹ and in active-TB patients ⁸⁰, and the signals significantly decreased after TB-therapy⁸¹. However, detection of VOCs is still technically difficult because most breath VOCs is excreted in picomolar concentrations and most current analytical instruments cannot detect VOCs at such low levels.

4. Predictive Correlates of TB risk

LTBI is commonly identified by an immune response to Mtb antigens either by tuberculin skin testing (TST) or Interferon (IFN)- release assays (IGRA). Five-10% of LTBI subjects will progress towards active TB in their lifetime (Figure 1) ^82-85. It has recently become clear, however, that LTBI is not a single entity but rather represents a broader spectrum of asymptomatic TB infection states that can be distinguished by different degrees of inflammation, bacterial replication and host immunity ^82-85. Thus, outcome of Mtb infection is not a binary state but rather a continuous spectrum of states with varying host inflammation, bacterial activity and

host immunity⁸⁶. These different states carry significantly different risks in reactivating TB. A major challenge therefore is to identify those that are at increased risk. This would allow targeted treatment of such individuals to prevent disease progression, and/or post-infection vaccination as a form of immunotherapy to strengthen the host’s immune response to control infection. Unfortunately, the predictive value of current IGRA or TST is far too low to qualify them as predictive tests ^{6, 87}. This argues for the importance of better TB-BM that can differentiate ”controller” LTBI from “progressors” towards active disease ^{35, 88}.

There have been several recent encouraging discoveries around TB CoR. ³⁹. Using RNASeq, a whole blood transcriptomic mRNA expression signature was identified in a large prospective cohort of LTBI adolescents around Cape Town comprising the differential expression of 16 human genes in the blood ⁸⁹. This first CoR- signature predicted progression from LTBI to TB disease with 66% sensitivity and 81% specificity one year prior to TB disease. A PCR-adapted 16-gene form of this CoR-signature validated in two additional large prospective African cohorts, consisting of adolescent household contacts of a confirmed TB index case (the GC6-74 “BM for TB consortium” study) ⁸⁹. In these validation cohorts CoR-test sensitivity was 54%, with 83%

specificity. With respect to possible underlying infection biology, it is of interest that all 16 CoR genes are regulated by type I and II interferons. These same genes have also been found to be part of various recently discovered diagnostic TB CoD-signatures, as discussed in section 3B. This suggests that the pathogenic processes that increase TB risk in LTBI and those during TB disease are overlapping. This is also in line with the above discussed continuous spectrum of various LTBI states and active TB disease ^{82-84, 86}. Moreover, this fits well with the observation that the identified CoR displayed good diagnostic performance in independent African cohorts of active TB patients vs. LTBI or other non-TB related lung diseases 30, 32, 33, 89, 90.

Almost all TB-BM discovery studies have focused on HIV-uninfected populations, assuming this would lead to a clearer and easier to interpret signal. In a smaller study we previously reported a simple CoR that predicted onset of TB disease months before clinical manifestation of disease in HIV-infected intravenous drug-users from the longitudinal Amsterdam cohort study. Targeted host-blood gene expression profiling by dcRT-MLPA ⁹¹

found that elevated expression of IL-13 in the absence of AIRE discriminated those who would develop future TB from those that remained TB-free for at least 2 years. Interestingly, individuals carrying the IL-13/AIRE CoR had increased expression of type I IFN related genes, agreeing with the African CoR signature discussed above.

There have been many studies correlating immune parameters to prediction of outcome of TB. Most have been hypothesis-driven, focusing on molecules in the circulation or on T-cell phenotypes following Mtb antigen stimulation. It is beyond the scope of this review to discuss these exhaustively. Instead a few recent important studies will be highlighted as examples of immune based CoR. One of the largest recent such studies measured immunity in BCG-vaccinated infants who either had or had not been booster-vaccinated with virally vectored Mtb Ag85A in MVA85A. In this large Phase IIb efficacy trial ⁹², that infants who would go on to develop active TB had elevated levels of activated CD4 T-cells in the circulation, as determined by HLA-DR expression, compared to infants who did not develop TB in the study period ⁴³. Importantly, this difference was seen at baseline. The results could also be validated in Cape Town cohort of prospectively followed LTBI adolescents: also here elevated CD4 T-cell activation correlated with risk of TB⁴³. These results qualify activated CD4 T-cells in the circulating as an independent CoR. Since most of the cells are supposedly not Mtb-specific, the likely interpretation of this work fits with a general picture of enhanced inflammation, both in the myeloid compartment as discussed above as well as in the CD4 T-cell compartment. Future studies should untangle the relationship and possible causality of these phenomena. There is evidence that concurrent Cytomegalovirus (CMV) infection may be related to the enhanced immune activation. This is particularly interesting as CMV infection also disables host immunity when chronic. Indeed, in the infants studied CMV- directed T-cell responses correlated with risk of progression to TB disease⁴³. However, other factors such NTM infections, other viral infections, environmental factors such as microbiota, and metabolic balances and host genetics may (also) play a role here.

The identification of these CoR, and the future refinement of the signatures obtained, are truly important. They will not only help stratifying risk individuals in clinical trials for TB vaccines and TB drugs, thus accelerating such studies by reducing size, time and costs involved; they will also offer potential advantage to the individuals themselves: preventive treatment can be considered in those testing positively, or testing increasingly positively at various time points e.g. spaced a few months apart to follow possible progression. A first study along these lines has already been initiated (the CORTIS study (ClinicalTrials.gov)). In a next phase the CoR-signatures and their test platforms need to be optimized, refined, decreased in complexity and converted to more practical tests for use in point-of-care settings.

5. BM for TB treatment response

As mentioned above, sputum culture conversion using solid medium is the best-characterized TB-BM indicating successful treatment, having been examined in many studies either as a simple measure (month 2 culture status) or in more complex forms requiring subsequent negative cultures (stable culture conversion).

However, trials including the REMoxTB trial ⁹³, have shown that patients developed recurrent TB despite negative sputum cultures at month 2. Conversely, PET-CT scanning showed that certain TB patients that had completed TB chemotherapy and were cured nevertheless still had active TB lesions post treatment ⁹⁴. Depending on the clinical setting, treated patients relapse ¹. Thus better BM predictive and indicative of TB treatment outcome are needed ³⁶.This is a priority for the TB research field and has the potential to impact clinical practice.

An ideal treatment response biomarker would be a surrogate for relapse, i.e., a marker that is validated and has regulatory approval as an early substitute outcome in clinical trials⁹⁵. Unfortunately, the TB community is currently far from such a validated surrogate marker for relapse. As reviewed in more detail below, existing culture-based BM have limited accuracy for predicting meaningful clinical outcomes. This presents a

challenge for new biomarker development. As reported above, since culture is an inadequate reference standard, new BM can only be meaningfully validated based on their ability to predict relapse versus durable non-relapsing cure. This section will review several promising investigational BM for treatment response based on either pathogen or host characteristics. Importantly, however, none have yet been evaluated rigorously yet for prediction of relapse.

5A. BM for TB treatment response: pathogen-based.

Measures of bacterial burden

The existing standard biomarker of treatment response is enumeration of the burden of Mtb culturable from sputum. Bacterial burden is measured as colony forming units⁹⁶, by culture conversion⁹⁷ or by time to detection in liquid culture. Unfortunately, culture-based BM of treatment response have proven poorly predictive of relapse among individuals, and inconsistently predictive among groups. A meta-analysis of patients treated with a rifampin-based regimen for drug-susceptible TB found that 2-month sputum culture had a sensitivity of 40% for detection of subsequent relapse⁹⁸. The positive predictive value of a 2-month sputum culture was 9%. In a recent large Phase-III clinical trial, intermediate culture-based BM were actually negatively associated with relapse. Experimental arms had significantly faster time to culture negativity yet had higher relapse rates than the control arm⁹³.

Several investigational approaches employ other methods of monitoring the sputum burden of Mtb. The first is quantification of “differentially culturable” bacteria (i.e., viable but non-culturable on standard agar plates, growing only with growth factor supplementation.) Reportedly, up to 90% of Mtb in the sputum of treatment- naïve patients may fail to grow on standard agar. It has been proposed that quantification of differentially culturable bacilli may elucidate drug-tolerant populations that persist late into treatment⁹⁹.

Sputum bacillary burden can also be quantified in a culture-independent manner based on abundance of Mtb nucleic acids. Bacterial messenger RNA (mRNA) degrades within minutes implying that detection of mRNA signals the presence of viable Mtb. Several studies have quantified the decline in individual mRNA in

sputum^{100, 101}. Ribosomal RNA (rRNA) has a longer half-life and is more abundant than mRNA, resulting in a longer “detection phase”. The Molecular Bacterial Load (MBL) Assay is a standardized method for quantification of Mtb 16S rRNA. MBL has demonstrated equivalence to culture on solid agar for monitoring early bactericidal activity¹⁰². DNA is relatively refractory to degradation and therefore does not distinguish between viable and dead Mtb¹⁰³. An additional approach is measurement of sputum LAM by ELISA. While sputum LAM performed poorly as a diagnostic test, it is currently being explored as a measure of bacterial burden during treatment¹⁰⁴.

A challenge common to all measures of bacillary burden is that the latter decreases precipitously with initiation of effective TB treatment. Quantitative culture and nucleic acid quantification show that the sputum burden of Mtb declines >99% during the initial 3-5 day bactericidal phase⁹⁶. Markers of bacterial burden therefore effectively measure bactericidal activity but are generally at the margins of detectability during the crucial subsequent sterilizing phase¹⁰⁵. Relapse is caused by the very small subpopulations of residual Mtb that survive this largely sterilizing phase of treatment. Therefore, prediction of relapse will likely require a biomarker capable of quantifying sterilizing drug activity as well as survival of very small subpopulations of residual Mtb ¹⁰⁵.

Measures of Mtb physiologic state

An alternative to measuring bacillary burden is measuring drug impact on the pathogen’s physiologic state.

Experimental data indicate that accumulation of triacylglycerol lipid--bodies indicates a non-replicating, drug- tolerant bacterial phenotype that may be the cause of treatment failure and relapse¹⁰⁶. Among patients in Malawi, sputum acid-fast bacillus lipid-body staining was significantly higher early in treatment among patients who proceeded to unfavourable outcomes (failure and relapse) than among those that were cured¹⁰⁶.

Another investigational measure of Mtb physiologic state is large-scale, pathogen-targeted transcriptional profiling¹⁰⁷. This elucidates bacterial molecular adaptations to drug stress in the bacterial population surviving early killing¹⁰⁸. Current challenges for this investigational approach include a relatively short detection phase

and technical challenges of quantifying very low abundance sputum-mRNA. As with other investigational BM, the association of Mtb transcriptional changes with relapse remains to be evaluated.

In summary, early bactericidal activity is readily measurable via serial enumeration of bacillary burden in sputum. A biomarker that predicts relapse will likely need to measure sterilizing effect and/or survival of very small subpopulations of residual Mtb. To date, this goal remains elusive.

5B. BM for TB treatment response: host-based BM.

The goal of TB-treatment is to clear Mtb from the host, preventing any relapses. Although treatment primarily targets the pathogen, host biomarkers may be used to measure, and preferably predict treatment response as outlined above. In a first large-scale prospective gene-expression study of TB patients undergoing treatment, the above mentioned type I and type II IFN-signaling signature normalized following curative treatment ¹⁰⁷. Another study analysed the transcriptomic profiles of relapsing vs. non relapsing patients who remained relapse-free for two years. Higher cytolytic responses were found to predict relapse, despite successful standard TB treatment, and this was confirmed in a larger patient cohort ¹⁰⁹. Also host immunological BM, especially cytokines and chemokines, have been used as biomarkers to predict or measure treatment responses. Cytokines involved in TB pathogenesis, such as TNF-α, interleukin-10, IFN- and IP-10, have attracted particular attention. TNF-α levels have been reported to decrease with treatment ^110,

111. High initial IL-10 level was reported to be in association with shorter survival ¹¹². IP-10 showed significant reduction at treatment-completion ¹¹³. Cytokines involved in tissue destruction, such as matrix metalloproteinases (MMP) and tissue inhibitor of matrix metalloproteinase (TIMP), represent another interesting group of molecules. Among patients who remained sputum culture positive at the 2^nd week of treatment, MMP-2, MMP-8, MMP-9 and TIMP-2 concentrations were higher in initial sputum specimens ¹¹⁴. In another study, high plasma MMP-8 levels at month 2 were associated with Mtb culture persistence ¹¹⁵. Additionally, apoptosis-associated BM may also be relevant as one study revealed that higher serum decoy

receptor 3 (DcR3) and monocyte chemotactic protein (MCP)-1 were independently associated with poorer six- month survival in active TB patients ¹¹⁶.

Using Mtb- specific stimulation assays, Mtb-specific inducible cytokines profiles can be measured. A newer generation IGRA, QuantiFERON-TB Gold Plus, includes novel CD8-specific antigens compared with the previous generation IGRA (QuantiFERON-TB Gold), which contained Mtb antigens-specific for CD4+

lymphocytes ¹¹⁷. A recent study suggested that while CD4 and CD8 responses decreased significantly during the first half of treatment, only CD8 responses decreased significantly during the latter half of treatment ¹¹⁸.

Other studies used proteomic approaches to identify TB-BM of therapeutic responses. By simultaneously measuring 1,030 proteins, several candidate proteins expressing significant changes during intensive treatment phase were explored ¹¹⁹. Compared to MTB proteomic responses, such host proteomic responses to TB treatment remain poorly studied ¹²⁰.

6. Future Perspectives and Concluding remarks

Thanks to the emergence of powerful unbiased omics and immune profiling platform technologies, advanced bioinformatics (e.g. ¹²¹ and deep immune cell phenotyping and functional characterization, significant progress has been made in the discovery and first validation of TB CoR, CoD and CoP. Validation studies in well- characterized clinical cohorts are now needed. Such studies will be essential in refining and improving the first-generation correlates which we have in hand and to understand whether these refined signatures are sufficiently powerful to justify their development towards use in clinical practice.

REFERENCES r

1 WHO. Tube culosis fact sheet, htp://www.who.int/me iacent e/factsheets/fs104/en/. 2017.d r

M dd rd r

2 Houben R , Do PJ. The Global Bu en of Latent Tube culosis Infecton: A Re-estmaton Using Mathematcal M do elling.PLoS M de . 2016; 13(10): e1002152.

r r d r d d

3 Getahun H, Gunnebe g C, G anich R, Nunn P. HIV infecton-associate tube culosis: the epi emiology an r

the esponse. Clin.Infect.Dis. 2010; 50 Suppl 3 :S201-7.

d r r r d d d d

4 Cantni F, Golet D. Biologics an tube culosis isk: the ise an fall of an ol isease an its new resu gence.r J.Rheumatol.Suppl. 2014; 91 :1-3.

r rd M M M M r M r r

5 Gi a i E, Sane Schepisi , Golet D, Bates , waba P, Yeboah- anu D, Ntoumi F, Palmie i F, aeu e M , Zumla A, Ippolito G. The global ynamics of iabetes an tube culosis: the impact of mig aton an d d d r r d policy implicatons. Int.J.Infect.Dis. 2017; 56 :45-53.

r M r M d

6 Zellwege JP, Sotgiu G, Block , Do e S, Altet N, Blunschi R, Bogyi , Bothamley G, Bothe C, Co ecasa L,

r r rd r r M r

Costa P, Dominguez J, Dua te R, Floe A, F esa I, Ga cia-Ga cia J , Golet D, Halm P, Hellwig D, Henninge

d r r rr

E, Heykes-U en H, Ho n L, K uczak K, Lato e I, Pache G, Rath H, Ringshausen FC, Ruiz AS, Solovic I, Souza-

M d r r

Galvao L, Wi me U, Wite P, Lange C, TBNET. Risk Assessment of Tube culosis in Contacts by IFN-gamma

r r r r r d

Release Assays. A Tube culosis Netwo k Eu opean T ials G oup Stu y. Am.J.Respi .C it.Ca er r r M de . 2015;

191(10): 1176-84.

r r d r rd M

7 Lag ange PH, Thanga aj SK, Dayal R, Despan e A, Ganguly NK, Gi a i E, Joshi B, Katoch K, Katoch V ,

r M r r M M d M r r dr

Kuma , Lakshmi V, Lepo te , Longuet C, alla i SV, uke jee D, Nai D, Raja A, Raman B, Ro igues C,

r d r r r r d

Sha ma P, Singh A, Singh S, So ha A, Kabee BS, Ve net G, Golet D. A toolbox fo tube culosis iagnosis: an

d r d r r

In ian multcent ic stu y (2006-2008): mic obiological esults.PLoS One .2012; 7(8): e43739.

M d d d r

8 Davies PD, Pai . The iagnosis an mis iagnosis of tube culosis. Int.J.Tube c.Lung Dis.r 2008; 12(11):

1226-34.

r r r r d r r r d r

9 WHO. High-p io ity ta get p o uct p ofles fo new tube culosis iagnostcs. Repo t of a consensus meetng. 2014;:htp://www.who.int/tb/publicatons/tpp_ epo t/en/.r r

r M r M M r M

10 Denkinge C , Kik SV, Ci illo D , Casenghi , Shinnick T, Weye K, Gilpin C, Boehme CC, Schito ,

r M M d r r r r

Kime ling , Pai . Defning the nee s fo next gene aton assays fo tube culosis .J.Infect.Dis. 2015 ; 211 Suppl2: S29-38.

r r r r r r d rr d r rr d d d

11 Bioma ke s Defnitons Wo king G oup. Bioma ke s an su ogate en points: p efe e efnitons an

r r

conceptual f amewo k. Clin.Pha macol.Ther r .2001 ; 69(3): 89-95.

r d d r r r r r r

12 Ethe i ge A, Lee I, Hoo L, Galas D, Wang K. Ext acellula mic oRNA: a new sou ce of bioma ke s. Mutat.Res .2011 ; 717(1-2): 85-90.

M d M r d r d r r r d

13 Nalejska E, aczynska E, Lewan owska A. P ognostc an p e ictve bioma ke s: tools in pe sonalize oncology. Mol.Diagn.Ther .2014 ;18(3): 273-84.

r r r r r r r M d r M rd

14 Buschmann D, Habe be ge A, Ki chne B, Spo n af , Rie maie I, Schelling G, Pfaf W. Towa reliable bioma ke signatu es in the age of liqui biopsies - how to stan a ize the small RNA-Seq wo kfowr r r d d rd r .

d

Nucleic Aci s Res .2016 ;44(13): 5995-6018.

r r r d r r

15 Ballman KV. Bioma ke : P e ictve o P ognostc? J.Clin.Oncol .2015 ;33(33): 3968-71.

r r r r

16 St imbu K, Tavel JA. What a e bioma ke s? Cu .Opin.HIV.AIDSrr .2010 ;5(6): 463-6.

r r r M r rr r

17 Southe n TR, Racsa LD, Alba ino CG, Fey PD, Hin ichs SH, u phy CN, He e a VL, Sambol AR, Hill CE,

r M rd M r r

Ryan EL, K af CS, Campbell S, Sealy TK, Schuh A, Ritchie JC, Lyon G ,3 , ehta AK, Va key JB, Ribne BS,

r r r rd M r rr d r B antly KP, St ohe U, Iwen PC, Bu E . Compa ison of FilmA ay an Quanttatve Real-Time Reve se

r r r r r r r d d

T ansc iptase PCR fo Detecton of Zai e Ebolavi us f om Cont ive an Clinical Specimens.J.Clin. ic obiolM r . 2015 ; 53(9): 2956-60.

d r rd rd M

18 van e Zee A, Roo a L, Bosman G, Ossewaa e J . Molecula Diagnosis of U ina y T act Infectons byr r r r

r

Semi-Quanttatve Detecton of U opathogens in a Routne Clinical Hospital Setng. PLoS One. 2016 ; 11(3):

e0150755.

dr r r r d d M

19 An ies AC, Duong V, Ly S, Cappelle J, Kim KS, Lo n T y P, Ros S, Ong S, Huy R, Ho woo P, Flaman ,

r r d

Sakuntabhai A, Ta antola A, Buchy P. Value of Routne Dengue Diagnostc Tests in U ine an Saliva Specimens. PLoS Negl T op.Dis r .2015 ;9(9): e0004100.

r M r d d r r

20 G een C, Hugget JF, Talbot E, waba P, Reithe K, Zumla AI. Rapi iagnosis of tube culosis th ough the detecton of mycobacte ial DNA in u ine by nucleic aci amplifcaton metho s r r d d . Lancet Infect.Dis .2009 ; 9(8):

505-11.

r

21 Iwasaki H, Chagan-Yasutan H, Leano PS, Koizumi N, Nakajima C, Tau ustat D, Hanan F, Lacuesta TL,

r r d d d d r r d

Ashino Y, Suzuki Y, Glo iani NG, Telan EF, Hato i T. Combine antbo y an DNA etecton fo ea ly iagnosis

r r d r

of leptospi osis afe a isaste .Diagn. ic obiol.Infect.DisM r .2016 ; 84(4): 287-91.

r r rd d

22 Cannas A, Calvo L, Chiacchio T, Cuzzi G, Vanini V, Lau ia FN, Pucci L, Gi a i E, Golet D. IP-10 etecton in

r d d

u ine is associate with lung iseases . M B C Infect.Dis .2010 ;1 0: 333,2334-10-333.

r rr r M M d r

23 Ke khof AD, Ba DA, Schutz C, Bu ton R, Nicol P, Lawn SD, eintjes G. Disseminate tube culosis

d r d r d d d

among hospitalise HIV patents in South Af ica: a common con iton that can be api ly iagnose using

r d

u ine-base assays .Sci.Rep .2017 ;7(1): 10931,017-09895-7.

r r M r M M

24 Lawn SD, Ke khof AD, Bu ton R, Schutz C, Boulle A, Vogt , Gupta-W ight A, Nicol P, eintjes G.

r r d d r r M r r d

Diagnostc accu acy, inc emental yiel an p ognostc value of Dete mine TB-LA fo outne iagnostc

r r d r r d r

testng fo tube culosis in HIV-infecte patents equi ing acute hospital a mission in South Af ica: a

r r

p ospectve coho t. B C M M de .2017 ;15(1): 67,017-0822-8.

M d M M r r

25 inion J, Leung E, Talbot E, Dhe a K, Pai , enzies D. Diagnosing tube culosis with u ine

r r d

lipoa abinomannan: systematc eview an meta-analysis. Eu .Respi .J r r .2011 ;38(6): 1398-405.

r r r r M

26 Kashyap RS, Rajan AN, Ramteke SS, Ag awal VS, Kelka SS, Pu ohit HJ, Tao i G , Daginawala HF. Diagnosis

r d d r r d d

of tube culosis in an In ian populaton by an in i ect ELISA p otocol base on etecton of Antgen 85

r r d

complex: a p ospectve coho t stu y. B C Infect.Dis M .2007 ; 7: 74.

r d r

27 Bentley-Hibbe t SI, Quan X, Newman T, Huygen K, Go f ey HP. Pathophysiology of antgen 85 in patents

r r r d

with actve tube culosis: antgen 85 ci culates as complexes with fb onectn an immunoglobulin G. Infect.Immun .1999 ;67(2): 581-8.

d M r r d M r

28 Liu C, Zhao Z, Fan J, Lyon CJ, Wu HJ, Ne elkov D, Zelazny A , Olivie KN, Caza es LH, Hollan S , G aviss

r M r r d r d d

EA, Hu Y. Quantfcaton of ci culatng ycobacte ium tube culosis antgen pept es allows api iagnosis

d d r r

of actve isease an t eatment monito ing. P oc.Natl.Aca .Sci.U.S.A r d .2017 ;114(15): 3969-74.

M r r r d r d r

29 Young BL, lamla Z, Gqamana PP, Smit S, Robe ts T, Pete J, The on G, Goven e U, Dhe a K, Blackbu n J.

d r r r r

The i entfcaton of tube culosis bioma ke s in human u ine samples. Eu .Respi .J r r .2014 ; 43(6): 1719-29.

rr M r M M r r

30 Be y P, G aham C , cNab FW, Xu Z, Bloch SA, Oni T, Wilkinson KA, Banche eau R, Skinne J,

M M

Wilkinson RJ, Quinn C, Blankenship D, Dhawan R, Cush JJ, ejias A, Ramilo O, Kon O , Pascual V,

r rr r r d r dr d r r

Banche eau J, Chaussabel D, O'Ga a A. An inte fe on-in ucible neut ophil- iven bloo t ansc iptonal

r r

signatu e in human tube culosis. Nature. 2010 ; 466(7309): 973-7.

d r r M r r M r r M 31 An e son ST, Kafo ou , B ent AJ, W ight VJ, Banwell C , Chagaluka G, C ampin AC, Dock ell H ,

r M rd M r rd M r

F ench N, Hamilton S, Hibbe L, Ke n F, Langfo PR, Ling L, lotha R, Otenhof TH, Pienaa S, Pillay V,

r d r M r d

Scot JA, Twahi H, Wilkinson RJ, Coin LJ, Hey e man RS, Levin , Eley B, ILULU Conso tum, KIDS TB Stu y

r d d r d r r

G oup. Diagnosis of chil hoo tube culosis an host RNA exp ession in Af ica. N.Engl.J. eM d .2014 ; 370(18):

1712-23.

r M rr M d rd

32 Bloom CI, G aham C , Be y P, Rozakeas F, Re fo PS, Wang Y, Xu Z, Wilkinson KA, Wilkinson RJ,

dr rr M r M r r r d

Ken ick Y, Devouassoux G, Fe y T, iya a , Bouv y D, Valey e D, Go ochov G, Blankenship D, Saa atan M , Vanhems P, Beynon H, Vancheeswa an R, Wick emasinghe r r M , Chaussabel D, Banche eau J, Pascual V, Ho r

M rr r r d r d r r r

LP, Lipman , O'Ga a A. T ansc iptonal bloo signatu es istnguish pulmona y tube culosis, pulmona y

r d d r

sa coi osis, pneumonias an lung cance s. PLoS One. 2013 ;8(8): e70630.

r M r r d r M r r

33 Kafo ou , W ight VJ, Oni T, F ench N, An e son ST, Bangani N, Banwell C , B ent AJ, C ampin AC,

r M d r rd M r rd M d M

Dock ell H , Eley B, Hey e man RS, Hibbe L, Ke n F, Langfo PR, Ling L, en elson , Otenhof TH,

M r d d d r

Zgambo F, Wilkinson RJ, Coin LJ, Levin . Detecton of tube culosis in HIV-infecte an -uninfecte Af ican

d d r r r d

a ults using whole bloo RNA exp ession signatu es: a case-cont ol stu y. PLoS M de . 2013 ; 10(10):

e1001538.

r d d r

34 Otenhof TH. New pathways of p otectve an pathological host efense to mycobacte ia. T en s r d M ric obiol .2012 ; 20(9): 419-28.

d r

35 Otenhof TH. The knowns an unknowns of the immunopathogenesis of tube culosis. Int.J.Tube c.Lungr Dis .2012 ;16(11): 1424-32.

r r r r

36 Otenhof TH, Ellne JJ, Kaufmann SH. Ten challenges fo TB bioma ke s. Tube culosis (E inbr d ). 2012 ; 92 Suppl1: S17-20.

r M r d r r d r

37 Sweeney TE, B aviak L, Tato C , Khat i P. Genome-wi e exp ession fo iagnosis of pulmona y

r r

tube culosis: a multcoho t analysis. Lancet Respi . er M d .2016 ;4(3): 213-24.

M r d r r r d r

38 ae tz o f J, Repsilbe D, Pa i a SK, Stanley K, Robe ts T, Black G, Walzl G, Kaufmann SH. Human gene

r r d r r

exp ession p ofles of susceptbility an esistance in tube culosis. Genes Immun .2011 ; 12(1): 15-22.

r rd r d M r d r r r M r d

39 Weine J,3 , Pa i a SK, ae tz o f J, Black GF, Repsilbe D, Telaa A, ohney RP, A n t-Sullivan C,

r r r d r

Ganoza CA, Fae KC, Walzl G, Kaufmann SH. Bioma ke s of infammaton, immunosupp ession an st ess

d r r d r r

with actve isease a e eveale by metabolomic p ofling of tube culosis patents. PLoS One. 2012 ; 7(7):

e40221.

r d M r r r M

40 Chegou NN, Suthe lan JS, alhe be S, C ampin AC, Co stjens PL, Geluk A, ayanja-Kizza H, Loxton AG,

d r d r r M r d dd M d r M

van e Spuy G, Stanley K, Kotze LA, van e Vyve , Rosenk an s I, Ki , van Hel en PD, Dock ell H ,

r r r r r r

Otenhof TH, Kaufmann SH, Walzl G, AE-TBC conso tum. Diagnostc pe fo mance of a seven-ma ke se um

r r r d d r r r r d

p otein biosignatu e fo the iagnosis of actve TB isease in Af ican p ima y healthca e clinic aten ees

d

with signs an symptoms suggestve of TB. Tho ar x. 2016 ;71(9): 785-94.

r M d d d r r

41 Co stjens PL, Tjon Kon Fat E , e Doo CJ, van e Ploeg-van Schip JJ, F anken KL, Chegou NN,

r d M r d r r M d M

Suthe lan JS, Howe R, ih et A, Kassa D, van e Vyve , Sheehama J, Simukon a F, ayanja-Kizza H,

r M r r r d r

Otenhof TH, Walzl G, Geluk A, AE-TBC conso tum. ult-cente evaluaton of a use -f ien ly late al fow

d r d d d r

assay to ete mine IP-10 an CCL4 levels in bloo of TB an non-TB cases in Af ica. Clin.Biochem. 2015;.

M r M r M d r

42 Esmail H, Lai RP, Lesosky , Wilkinson KA, G aham C , Coussens AK, Oni T, Wa wick J , Sai -Ha tley Q,

r rr rr r r

Koegelenbe g CF, Walzl G, Flynn JL, Young DB, Ba y Iii CE, O'Ga a A, Wilkinson RJ. Cha acte izaton of

r r d r d r r d

p og essive HIV-associate tube culosis using 2- eoxy-2-[18F]fuo o-D-glucose posit on emission an

d r

compute tomog aphy.Nat. e .M d 2016; 22(10): 1090-3.

r d M dr d rr M M r M

43 Fletche HA, Snow en A, Lan y B, Ri a W, Sat I, Ha is SA, atsumiya , Tanne R, O'Shea K,

d rd d M r rr d M

Dheena hayalan V, Boga us L, Stock ale L, a say L, Chomka A, Ha ington-Kan t R, anjaly-Thomas ZR,