IP-10 release assays in the diagnosis of tuberculosis infection: current status and future directions

Abstract

The current state-of-the-art tests for infection with Mycobacterium tuberculosis – the IFN-γ release assays – rely on accurate measurement of the cytokine IFN-γ. Many other potential biomarkers are expressed in concert with IFN-γ, and IP-10 in particular has shown promising results. IP-10 is produced in large amounts, allowing for the development of new and simplified test platforms, such as lateral flow. In this review, we summarize the results of 22 clinical studies exploring the use of IP-10 as an alternative marker to IFN-γ. The studies report that diagnostic accuracy of IP-10 is on par with IFN-γ, but also that IP-10 may be more robust in young children and in HIV-infected individuals with low CD4 cell counts. We conclude the review by presenting limitations of the published works and outline recent developments and future directions.

Keywords::

• CXCL10
• IFN-γ
• IGRA
• IP-10
• latent TB infection
• tuberculosis

Figure 1. IP-10 induction through adaptive immune responses.

(A) T cells recognize Mycobacterium tuberculosis-specific peptides on the MHC molecules on the surface of the APCs, and this interaction activates the T cells and also the APCs. (B) T cells and the APCs secrete a series of proinflammatory cytokines, which further activates the APCs and leads to (C) IP-10 secretion from activated APCs.

APC: Antigen-presenting cell; IFN: Interferon; TCR: T-cell receptor.

Introduction

Mycobacterium tuberculosis is a virulent intracellular pathogen that survives in macrophages, even in the presence of an intact adaptive immune response. The bacillus has coevolved with primates for thousands of years, which is probably the reason for its success [1]. With approximately a third of the world’s population showing immune recognition of M. tuberculosis antigens, an estimated rate of 8 million new cases of active tuberculosis (TB) and at least 2 million TB-related deaths annually, infection with this bacillus is pandemic [2]. Although the M. tuberculosis genome has been sequenced, the proteome scrutinized and many potential immunogenic proteins identified [1], it has proven difficult to develop an efficient vaccine to substitute the bacillus Calmette–Guérin (BCG) vaccine [3]. Consequently, the global strategy to control the disease centers on identifying and treating active cases to prevent further spread of the bacillus. In resource-rich areas screening and targeted treatment of patients at risk of developing active TB, such as recently exposed children, HIV-infected individuals and patients starting on immunosuppressive treatment, has been mandatory [2,4], and in high endemic regions, chemoprophylaxis is now recommended for all exposed children <5 years of age and HIV-positive individuals [301].

IFN-γ release assays

A decade ago a new generation of tests specific for M. tuberculosis infection were introduced. Two tests are currently commercially available: the whole-blood/ELISA-based QuantiFERON^®-Gold In Tube test (QFT-IT; Cellestis/Qiagen, CA, USA) and the PBMC/ELISPOT based T-SPOT^®.TB test (Oxford Immunotec, Abingdon, UK). These IFN-γ release assays (IGRAs) are in essence cell-mediated immune response (CMI) assays relying on measurement of T-cell responses in vitro to peptide antigens specific for M. tuberculosis[5,6]. The IGRAs are based on five essential components: M. tuberculosis-specific antigens; antigen-presenting cells (APC); T cells; an incubation step; and a readout biomarker of immune recognition (Box 1).

In BCG-unvaccinated otherwise healthy individuals the IGRAs perform similarly to the tuberculin skin test (TST), but the increased specificity has proven superior in BCG-vaccinated individuals and patients with exposure to, or infection with, non-tuberculous mycobacteria (NTM) [5,7–9]. Although IGRAs appear to have higher sensitivity than the TST in immunosuppressed patients, such as persons with HIV infection [9,10] and in the very young and very old, IGRA sensitivity is still compromised in these patients [11–15,302]. The clinical application of IGRAs is a matter of debate [16,302]. New approaches are needed to improve the methods hereunder evaluation of alternative biomarkers.

The aim of this review is to provide an overview of current knowledge on the potential of IP-10 as a diagnostic biomarker for infection with M. tuberculosis and to describe the potential of IP-10 in future improvements of the IGRA.

Discovery of IP-10 as a biomarker for TB infection

In 2006, we screened a series of cytokines and chemokines for their potential as markers of CMI responses to specific M. tuberculosis antigens in QFT-IT supernatants, and found that IP-10 is expressed in very high amounts in patients with active TB, but not in unexposed controls [17]. Further evidence has since substantiated this discovery and >40 additional biomarkers have been screened for diagnostic potential. Of all investigated markers, it appears that IP-10 is the most consistently expressed in response to antigen challenge, rendering it the most promising alternative marker to IFN-γ [18–27] and a promising means of improving IGRAs [28–34].

The immunological basis of IP-10 as a readout marker in CMI assays

IP-10 is a small chemokine expressed by APCs [35,36] and a main driver of proinflammatory immune responses. IP-10 is expressed by cells infected with viruses and bacteria, but can also be induced at high levels as part of the adaptive immune response. In this case, IP-10 secretion is initiated when T cells recognize their specific peptide presented on the APC (Figure 1). IP-10 secretion appears to be driven by multiple signals, mainly T-cell-derived IFN-γ, but also IL-2, IFN-α, IFN-β, IL-27, IL-17, IL-23, and autocrine APC-derived TNF-α and IL-1β. TNF-α is a weak IP-10 inducer per se, but a potent synergistic inducer when acting with the interferons [37–46]. Because of the many different pathways that can lead to IP-10 secretion, IP-10 can be considered a downstream marker to typical readout markers in CMI assays, such as IL-2 and IFN-γ [47,48].

An important distinction of IP-10 compared with the typical readout markers is that it is expressed in much higher levels, and as IP-10 is a 2.5-times smaller molecule than IFN-γ (7.4 vs 18.8 kD), this difference is even larger at the molecular level when determined with solid-phase technologies (such as ELISA). Although the difference in magnitude is very large, the fold change between signal (antigen response) and noise (nil level) appears to be comparable between the two markers. Therefore, IP-10 is a promising alternative marker in CMI assays.

Review of published studies

Studies in active TB patients, healthy controls & exposed individuals

We identified 12 studies in adults and seven studies in children. The studies in adults comprised a total of 645 adult TB patients and 583 controls from both high- (India, Tanzania, Guinea Bissau and South Africa) and low- (Denmark, Sweden, Finland, Switzerland, Greece, Italy, Spain and USA) prevalence settings (Tables 1 & 2). All adult studies confirmed that IP-10 is specifically induced in response to TB-specific antigens in patients with active TB compared with controls [17,19,20,22,49–55]. Seven studies set or used cutoffs for evaluation of derived IP-10 test results (Table 5)[20,51–56], of which five reported result distribution and performed head-to-head comparisons with the QFT-IT [51–55]. These five studies (two from India, one from Denmark, one from Tanzania and one multicenter study from countries in the European Union) comprise a total of 435 patients with active TB and report comparable IP-10 test and QFT-IT positivity rates (74–91 vs 79–100%, respectively) as well as good agreement between the two tests (data available for two studies only; agreement 73–89%; κ: 0.43–0.57) [51–55]. Positivity rates increased significantly when IP-10 and QFT-IT test results were combined (available for four studies only; positivity rates increased to 86–93%) [52–55].

Three studies compared IP-10 test and QFT-IT performance in healthy controls (Table 2)[52–54]. In two studies from a low-prevalence settings (Denmark and Italy), out of 187 nonexposed healthy controls, only 2% were found to be IP-10 test positive (specificity 98%) and 0% were QFT-IT positive (specificity 100%) and there was no indication of cross-reactivity with BCG vaccination [52,53]. The third study, performed in India, included 100 controls and reported positive results in 52% by the IP-10 test and 48% by the QFT-IT, probably reflecting the high prevalence of latent TB infection [54].

Studies of sensitivity to latent TB infection are difficult to interpret owing to the lack of a test-independent gold standard, but can be attempted by using different measures for risk of exposure. The European multicenter study reports such results for the IP-10 test and the QFT-IT in 175 patients with suspected TB but who were finally diagnosed with another disease (Table 2). In this group, the IP-10 test and the QFT-IT were found to be positive in 35 and 27% of patients, respectively (p < 0.05). The increased IP-10 positivity rate could represent either superior sensitivity or impaired specificity of the IP-10 test compared with the QFT-IT. It was shown that the presence of risk factors for TB infection (exposure history, prior TB and stay in a highly endemic area) increased the adjusted odds ratios for having a positive test result from both tests, with no significant difference between tests [53], and larger studies in groups with varying degrees of exposure are therefore needed to clarify this issue.

In conclusion, current evidence suggests comparable accuracy of IP-10 based tests with the QFT-IT in patients with active TB and healthy controls, as well as in TB-exposed individuals.

Children

TB is a major contributor to childhood morbidity and mortality [57,58] with children <5 years of age at highest risk and 40–50% of infected infants developing disease within 1–2 years. Childhood TB is notoriously difficult to diagnose owing to its pauci-bacillary nature and the limited performance of existing tools [59,60].

In contrast to adults, where the use of IGRAs in diagnosing active TB is not recommended, IGRAs may be useful in the diagnostic work-up of active TB in children with negative smear microscopy and culture results. In this case, a positive IGRA in a child indicates recent exposure, and therefore an increased risk of developing active TB.

The reported performance of IGRAs in children is not consistent with varying rates of sensitivity, specificity and proportion of indeterminate results [7,8]; differences that are probably due to heterogeneity in the study population, severity of TB, malnutrition and comorbidities [7,302].

Until now, seven studies comprising 1408 children aged 1–17 years from high- (Nepal, Brazil, Nigeria and Ethiopia) and low-endemic countries (USA, UK and Germany) have assessed IP-10 as a tool for the diagnosis of latent TB infection or active TB (Table 3).

As in adults, the IP-10 levels after antigen stimulation were higher compared with IFN-γ levels, and IP-10 levels in unstimulated samples were generally higher in children with active or latent TB compared with children with no disease and no signs of infection with M. tuberculosis, as determined by exposure risk or a positive TST or QFT-IT test.

Antigen-stimulated IP-10 responses were higher in children with active TB and in high-risk groups. The IP-10 test positivity rate was assessed in three studies comprising 40 children with active TB (22 children from Ethiopia [61] and 11 from Germany [62] and seven from the USA [63]). Two studies found a high rate of positive responders comparable to adult studies of 82% (18/22) [61] and 91% (10/11) [62], whereas Lighter et al. found a lower positivity rate of 43% (3/7) [63].

In 14 children from Germany with latent TB, defined as a positive TST or QFT-IT result, the IP-10 positivity rate was 86% [62]. In a study from Nigeria, the IP-10 test and the QFT-IT detected comparable proportions of positive children in high-, moderate- and low-risk groups [64]. In unexposed or low-risk controls, the specificity of IP-10 varied depending on the setting; in 44 children from low-endemic countries [62,63] and in 86 children with negative TST and QFT-IT from Ethiopia [61], IP-10 positivity rates were low (0–5%) [61–63], whereas in 23 community controls from Nigeria, the positivity rate was higher (26%) [64], indicating similar specificity to the QFT-IT. Agreement of the IP-10 test and the QFT-IT was assessed in three studies and ranged from 0.50 to 0.96 [25,61,62].

The IGRAs appear to perform suboptimally in children <5 years of age compared with adults and older children [7,65]. As prevalence of latent TB infection increases with age, the impact of age alone on test sensitivity is difficult to interpret and studies addressing the possible influence of age on test performance are needed.

Lighter et al. found that mitogen-induced IFN-γ, but not IP-10 responses, increased with age [63]. Lighter-Fischer et al. found that children <5 years of age with latent TB infection had reduced IFN-γ, but not IP-10, responses to TB antigens [25]. By contrast, Ruhwald et al. found an association between age and rate of positivity rate for the IP-10 test and TST but not for QFT-IT in a group of Nigerian children with different exposure risk [64]. Alsleben et al. found no effect of age on IP-10 mitogen responsiveness in 48 German children also with varying exposure [62].

In children, lymphadenopathy caused by NTM is a differential diagnostic problem and in eight children with NTM infection both QFT-IT, and IP-10 tests were negative, suggesting that IP-10, as IFN-γ, discriminates between M. tuberculosis and NTM infections [62].

In summary, the studies suggests that IP-10 has comparable diagnostic accuracy to QFT-IT for diagnosing TB in children, with some indication that IP-10 may be less affected by age than IFN-γ in the QFT-IT.

HIV-infected individuals

HIV-infected individuals with severe immune suppression have a high risk of progression to active TB disease and would therefore benefit from accurate diagnosis and prophylactic treatment of latent M. tuberculosis infection [66]. However, many studies have shown that the IGRAs, in particular the QFT-IT, render many indeterminate and possibly also false-negative results in HIV-infected individuals with a low CD4 cell count [13,14,65–72]. Therefore, alternative approaches in this group of patients are necessary, but only few studies have evaluated IP-10 for this purpose.

We identified three studies comprising a total of 143 HIV-infected TB patients, and these report rather divergent findings (Table 4)[55,56,73]. Two small studies from India, comprising a total of 78 HIV-infected TB patients, both found a marginally higher IP-10 test positivity rate compared with the QFT-IT (86 vs 61% and 86 vs 70%, respectively) [56,73]. By contrast, a study in 65 HIV-infected TB patients from Tanzania found equal positivity rates for the tests (63% for both), which was significantly lower compared with HIV uninfected individuals (82 and 80%, respectively) [55].

Three studies from India included HIV-infected patients with no signs of active TB. Two studies (n = 208) both reported more positive results for the IP-10 test than the QFT-IT test (87 vs 32% [56] and 45 vs 28% [10], respectively), while one study (n = 50) found no positive results by either test [73].

Interestingly, all of the aforementioned studies in HIV-infected individuals with active TB found that IP-10 positivity rates were less dependent on CD4 cell count than QFT-IT positivity rates were. One large study in 170 otherwise healthy HIV-infected individuals from a TB-endemic setting, found that very low IP-10 responses to mitogens (<200 pg/ml) only occurred in individuals with CD4 cell count <100 cells/µl [55]. However, comparisons of indeterminate results are difficult to make due to the lack of a validated cutoff for an indeterminate IP-10 response (see below). The immunological mechanisms underlying these suggested differences between IFN-γ and IP-10 in HIV-infected patients are as yet unknown, but are probably due to a combination of both the higher magnitude of response and the broader induction profile of IP-10 compared with IFN-γ [37–46]. Sutherland and colleagues recently demonstrated that coincident with the loss of total CD4 count, the TB antigen-specific immune response changed from a polyfunctional CD4 response to a monofunctional IFN-γ or TNF-α CD8 T-cell response [74]. As IP-10 is inducible by a wide panel of CD4 and CD8 T-cell cytokines (particularly IFN-γ and TNF-α [75]), it could be that IP-10 picks up and combines these weaker cytokine responses from the monofunctional cells in a stronger common downstream signal.

In conclusion, the IP-10 test in HIV-infected patients with TB performs slightly better or comparable to QFT-IT, and both tests are negatively affected by advanced HIV infection. In HIV-positive but otherwise healthy high-risk individuals more IP-10 results are positive, a finding that may reflect that the IP-10 test is either more sensitive or less specific in this population. Therefore, although the IP-10 test performance does seem to be impaired in HIV-infected individuals, it may be more robust than the QFT-IT in individuals with low CD4 cell counts. However, current evidence is limited and larger studies are needed.

Immunosuppressive therapy

The use of IGRAs and TST in screening for latent TB infection before initiation of immunosuppression with, for example, TNF-α inhibitors, and also that IGRAs are more specific and in some instances more sensitive than TST in this vulnerable group of patients is well recognized [76]. However, false-negative IGRA results occur, and test performance is impaired, in patients treated with prednisolone and possibly also other drugs used to treat inflammatory diseases [77].

So far, only one study has evaluated IP-10 for diagnosing latent TB infection in patients receiving immunosuppressive treatment [78]. In this study, 52 patients from Taiwan with rheumatoid arthritis (RA) were treated with methotrexate and one or more of the immunosuppressive drugs used for RA; of these, 24 had latent TB (defined by a positive TST). In addition, 18 patients with active TB were included. The study reported that, in contrast to IFN-γ, unstimulated and antigen-dependent IP-10 levels were significantly higher in TST-positive compared with TST-negative results in RA patients, and higher in those with a positive compared with a negative QFT-G result.

Interestingly one patient developed active TB despite chemoprophylaxis, and in this case, unstimulated levels of IP-10 increased until the diagnosis of active TB and declined again during effective treatment [78].

The area under the curve (AUC) for detecting active TB using IP-10 was 0.71 and 0.79 for ESAT-6 and CFP-10 respectively. The AUC for detecting latent TB infection (TST positive) was 0.88 and 0.78, respectively. Interestingly, baseline IP-10 had similarly discriminatory power. IP-10 may be useful for diagnosing active and latent TB in patients receiving immunosuppressive treatment; however, the current data are very limited and further studies are needed.

Limitations of the current studies & technical aspects

One limitation of the studies evaluating diagnostic tests for latent TB infection is the lack of a gold standard. In some of the clinical studies, TST and/or QFT-IT have been used as gold standard, but in order to compare diagnostic accuracy between tests, a gold standard independent of the test under evaluation must be used. For example, if a TB patient has a positive IP-10 but a negative QFT-IT result, the IP-10 response will be deemed false positive if QFT-IT is the gold standard.

Another important issue is the need for a proper measure of indeterminate test. One study defined an arbitrary cutoff for an indeterminate IP-10 test [52], which has also been applied in other studies [10,53,55]. The rates of indeterminate results were, in general, comparable with the IGRAs, but there seemed to be a tendency towards fewer indeterminate and more false-negative results for the IP-10 test, suggesting a need for an adjustment of the cutoff.

The IGRAs utilize phytohemagglutinin (PHA) as mitogen positive control. However, PHA may not be the optimal choice for an IP-10 test. PHA is a powerful inducer of IFN-γ, whereas IP-10 is induced at relatively lower levels. The difference is probably due to the fact that PHA acts directly on the T cells but not on the APCs. PHA-induced IP-10 responses are indirectly induced via T-cell cytokines released into the cell-culture supernatant, leading to APC activation. A more suitable positive control mitogen for an IP-10-based test could be the lectin concavalin A [79], which leads to IP-10 induction through direct activation of the monocytes [80].

A future cutoff for indeterminate test should ideally be based on a cost-effective assessment of the risk of progression in false-negative patients with latent TB and weigh in the required resources needed to retest, as well as counseling of patients with persistently indeterminate tests. Such cutoffs are likely to be setting specific.

A major limitation in this field relates to the technical aspects of the assays used for IP-10 detection. In the 22 clinical studies investigating the diagnostic potential of an IP-10 release assay included in this review, IP-10 has been measured with a variety of assays: from different brands of multiplexing assays to commercial or in-house ELISA kits (Table 5). Furthermore, samples have been measured in different dilutions that might not have been optimal for the dynamic range of the standard curve (Table 5). Cutoffs vary from study to study, and as cutoffs are linked both to the readout platform and the sample dilution, a more stringent approach is needed before studies can be compared. Future studies should measure and report IP-10 results in ways that allow comparison between studies. Adherence to Standards for Reporting of Diagnostic Accuracy criteria, and thorough descriptions of sample assay ranges and dilutions, will accelerate the field, as cutoffs will be identified and applied between studies.

Hot topics & unsolved questions

Improving the IGRA using IP-10

The role of IGRAs in diagnosing and managing TB infection and disease is still controversial, and all of these concerns will also need to be addressed for a potential IP-10 test [9,81,82]. Two of the greatest obstacles for the IGRAs are the inability to discriminate between active and latent TB infection, as well as suboptimal sensitivity [83]. In immunocompromised individuals and children, who have the greatest risk of progression to active disease and would benefit the most from an accurate diagnosis of latent TB infection, the IGRA performance is compromised.

Several approaches have been attempted to overcome the shortcomings of the current IGRAs. IP-10 has been shown to be specifically induced, not only in response to the TB-specific antigens used in the QFT-IT and QFT-G [78] tests, but also in response to purified protein derivative [25] and alternative TB antigens, including selected region of difference 1-specific peptides [21,50,51,56], peptides spanning the mycobacterial protein TB10.4 [49], whole M. tuberculosis sonicate [21], cytomegalovirus, HIV [84] and peptides specific for Mycobacterium leprae[85]. Alternative or additional antigens appear to have the potential for improving IGRA sensitivity [86,87] and enable discrimination between active and latent TB [88–91]; in this context, IP-10 may be a favorable biomarker for measuring responses to novel antigens with limited immunogenicity because of the high levels of biomarker produced [51].

Several studies have demonstrated that the discordance between IP-10 and IFN-γ test results can be utilized by a combination of tests to increase diagnostic sensitivity, apparently without a compromise in specificity [52–55]. Other cytokines and chemokines (including IL-2, TNF-α, MIP-1b and MCP-1) have shown promise as diagnostic markers, but the potential added benefit of these has not yet been systematically assessed. We have studied MCP-2 and found a lower sensitivity of 71% for active TB at 97% specificity in healthy controls, but no significant added effect by combination with either IP-10 or IFN-γ [52]. In a subsequent evaluation in HIV-infected patients with confirmed TB, we observed significant lower sensitivity of MCP-2 compared with IFN-γ and IP-10, and found no added effect by combination [Aabye MG, Ruhwald M, Unpublished Data].

Tweaking of the incubation step, for example by increasing incubation temperature [49], prolonging the incubation period [23,92,93], improving culture conditions with nutrients [201] or addition of proinflammatory cytokines (e.g., IL-12 [85]), blocking of anti-inflammatory cytokines (e.g., IL-10 [49,94]) or adding ‘survival’ cytokines (e.g., IL-7 [95,96]), have also shown potential for augmentation of the antigen-specific immune response.

IGRAs offer some practical advantages compared with the TST. IGRAs can be performed as a day-to-day assay with no need for the patients to return and IGRAs have an objective readout. However, IGRAs do require direct access to a laboratory that can perform the ELISA or a cold chain for transporting samples to the laboratory, which may limit their use, for example, in field studies or in low-resource settings. In this regard, IP-10 offers a significant advantage. The high levels of IP-10 secreted allow for the use of smaller blood volumes, less sample material, as well as simpler readout systems. We have developed a method using dried plasma or whole blood in filter paper [Aabye MG, Ruhwald M, Unpublished Data]. Biomarkers are stable in dried blood and plasma at ambient temperature [97], therefore the filter paper technology enables cheap and simple sample transport over long distances by normal mail, and centralized analysis vis à vis the neonatal screening for metabolic diseases [84].

Recently, two independent groups demonstrated CMI assays based on a finger-prick volume of blood (50 µl) using quantitative PCR detection of IP-10 mRNA [98,99]. Berit Lange and coworkers have explored the potential of a simple field-friendly lateral flow rapid test for IP-10 detection in QFT-IT supernatants at point of incubation [100]. IP-10 is thus a top candidate biomarker for a point-of-care test for detection of TB infection.

Other potential uses of IP-10 in TB management

Blood levels of IP-10 have potential as a marker of inflammation vis à vis C-reactive protein. A series of papers have demonstrated that unstimulated IP-10 levels in blood or plasma are increased with infection and disease, such as TB [101–104], hepatitis C viral infection [105–107] and bacteremia [108–110]. IP-10 is produced at the site of inflammation and the levels in the blood appear to correlate with the extent of the inflammatory process. Compared with the high increase seen after in vitro antigen stimulation, blood levels in sick patients are lower.

Interestingly, IP-10 levels in blood and urine appear to decline with successful treatment [104,110–116], and increase in patients where there is disease relapse or progression to active disease [78,104]. IP-10 levels also remain consistent in a whole-blood sample for several days at 37°C, for example in the nil tube of the QFT-IT test [53] [Aabye MG, Ruhwald M, Unpublished Data]. Therefore, measuring the level of IP-10 in the unstimulated sample appears to contribute additional information about the extent of infection, such as severity of disease and risk of progression, and could potentially also be used as a measure of treatment efficacy. This potential added benefit of IP-10 calls for exploration in prospective studies.

Conclusion

In conclusion, IP-10-based tests appear to perform comparably to the QFT-IT in most patient groups, but may provide some additional information, for example in young children and HIV-infected individuals with low CD4 cell count. However, more studies with a broader consensus on several technical aspects in IP-10 assay methodology and using validated cutoffs, as well as studies in exposed individuals, are needed before conclusions can be drawn.

New technologies are emerging, and a new generation of rapid and user-friendly tests appears eminent. If proven reliable for large-scale use, these new tests will probably be combined with a novel generation of M. tuberculosis-specific antigens with higher sensitivity, and antigens that are able to discriminate between patients with active TB and latent TB infection. In addition, several other uses of IP-10 measurements could be studied more thoroughly, such as the potential of the acute-phase reactant properties of plasma IP-10 in monitoring treatment responses, and predicting mortality and morbidity. Finally, dynamic patterns (conversions and reversions) of immune responses to TB antigens might provide useful information and need to be better understood.

Expert commentary

The IGRAs were introduced in the early 2000s as a new generation of very specific tests for infection with M. tuberculosis. It has been established that the IGRAs’ performance is superior to that of the TST, but sensitivity is still compromised in high-risk individuals, including HIV-infected individuals, children and patients treated with immunosuppressive drugs. New measures are needed to improve the IGRA method, for example the use of alternative biomarkers. From studies screening >40 alternative biomarkers, it appears that IP-10 is the most consistently expressed in response to TB antigens, rendering it the most promising alternative marker to IFN-γ in TB diagnosis.

Until now, 22 articles including authors from 22 countries have explored the diagnostic potential of IP-10. The studies conclude that IP-10 is a specific marker for TB infection expressed in high magnitude compared with IFN-γ. IP-10 appears to perform comparably to the QFT-IT in diagnosing active TB and several studies suggest that IP-10 is less influenced by advanced HIV infection and this is possibly also the case for young age. The majority of studies use supernatants from the QFT-IT; however, IP-10 has been measured with different ELISA and multiplexing assays, and interpreted using different cutoff points and very diverging study designs. This complicates accumulation of knowledge and direct comparison of the performance of an IP-10 test between studies. Therefore, there is a need for consensus in the field.

Some of the major problems of the IGRAs also seem to apply to the IP-10 test. However, particularly from the technical aspect, IP-10 might offer advantages over IFN-γ and, several innovative developments point towards a new generation of smaller, simpler and cheaper diagnostics tests based on IP-10. Further to this, IP-10 is a top candidate biomarker for a point-of-care test.

When the technical issues are resolved and the best candidate test has been identified, larger prospective studies investigating the predictive value of a positive IP-10 test should be carried out.

Five-year view

The IGRAs have set the stage for a move from the unspecific TST toward specific and objective tests for infection with M. tuberculosis. The current commercially available IGRAs show very high specificity, but have insufficient sensitivity. IP-10 is quantitative and qualitatively different to the currently used marker, IFN-γ, which opens the way for assay optimization and simplification.

In terms of improved accuracy, we believe that the intense search for efficient TB vaccine antigen candidates will also lead to the discovery of novel diagnostic antigens. Such diagnostic antigens are expected to both complement the currently known antigens for higher sensitivity, but also aid in differentiation between latent TB infection and active TB. These novel antigens will probably be less immunogenic than those such as ESAT6, wherein IP-10 is probably a more robust readout marker for such a diagnostic test.

In terms of improved ease of use we expect a panel of recently suggested low- and high-tech platforms to be optimized and evaluated for IP-10 detection. Recent work from our own group demonstrated that antigen-stimulated blood can be dried on filter paper and sent by regular mail to large centralized laboratories vis à vis neonatal screening for metabolic disorders. This method also has potential to markedly alleviate IGRA research in low-resource settings and can probably be put to use very soon.

In the future, IP-10 will possibly be quantified using a lateral-flow device, therefore enabling rapid diagnosis in places with little infrastructure, or as suggested by others in a microassay based on IP-10 mRNA detection with quantitative PCR following antigen stimulation. In addition, integration of the incubation step with the IP-10 readout step could lead to novel fully integrated fast and accurate lab-on-a-chip diagnostic platforms.

We therefore believe it is realistic that a new generation of IP-10 release assays will be designed within the next 5 years, and we expected that these have higher sensitivity, increased robustness and much greater ease of use.

Table 1. Selected studies assessing the performance of IP-10 in patients with confirmed or clinically assessed active tuberculosis.

Study (year)	Country	Study subjects	Results	Ref.
Ruhwald et al. (2008)	Switzerland and Denmark	80 confirmed TB patients	Similar positivity rate between IP-10 test (74%) and QFT-IT (81%) Increased when combining tests (88%)	[52]
Aabye et al. (2010)	Tanzania	92 confirmed TB patients	Similar positivity rates between the IP-10 test (75%) and the QFT-IT (74%). Increased when combining tests (86%)	[55]
Kabeer et al. (2010)	India	175 confirmed TB patients	Similar positivity rate between IP-10 test (91%) and QFT-IT (87%) and increased when combining tests (93%)	[54]
Kabeer et al. (2011)	India	17 confirmed TB patients	Similar positivity rate between IP-10 test (94%) and QFT-IT (100%) Combination of tests not reported	[51]
Ruhwald et al. (2011)	Italy, Spain, Greece, Denmark, Sweden, Finland (nine centers)	168 confirmed TB patients	Equal positivity rate between IP-10 test (79%) and QFT-IT (79%) Increased when combining tests compared with either test alone (87%)	[53]

QFT-IT: QuantiFERON^®-TB Gold In tube; TB: Tuberculosis.

Table 2. Selected studies assessing the specificity of IP-10 in unexposed healthy controls.

Study (year)	Country	Study subjects	Results	Ref.
Healthy controls
Ruhwald et al. (2008)	Denmark	86 high-school students with no history of exposure to TB	No positive results by either test	[52]
Kabeer et al. (2010)	India	50 healthy controls with no close family contact with TB and a negative TST and QFT-IT result	No positive results by either test, but inclusion criteria was a QFT-IT negative result	[73]
Ruhwald et al. (2011)	Denmark and Italy	101 healthy students with no history of exposure to TB, or prior TB disease or treatment	Similar negativity rate between IP-10 test (95%) and QFT-IT (99%)	[53]
Frahm et al. (2011)	USA	26 healthy controls with a negative TST and QFT-IT result	96% IP-10 test negative and 100% QFT-IT negative, but inclusion criteria was a QFT-IT negative result	[20]
Kabeer et al. (2010)	India	100 controls with no symptoms indicative of TB and no history of a close family contact with TB, but from high endemic setting	More negative results by the QFT-IT (52%) than by the IP-10 test (48%)	[54]
Diseased controls
Ruhwald et al. (2011)	Italy, Spain, Greece, Denmark, Sweden, Finland (nine centers)	175 patients initially suspected of TB but finally diagnosed with another disease	More positive results by IP-10 test (35%) than by QFT-IT (27%). Increased AOR of a positive result by either test with the presence of risk factors for exposure (born in a TB endemic country, prior TB)	[53]

AOR: Adjusted odds ratio; QFT-IT: QuantiFERON^®-TB Gold In tube; TB: Tuberculosis; TST: Tuberculin skin test.

Table 3. Selected studies assessing the performance of IP-10 in children.

Study (year)	Country	Study subjects	Results	Ref.
Yassin et al. (2011)	Ethiopia	22 active TB, 136 probable TB, 131 unlikely TB, 335 healthy high risk and 156 low risk of TB infection	Positivity rate was 81% for IP-10, 72% for QFT-IT and 79% for TST. High levels of IP-10 were found in children with active TB and close contact with TB. Up to 37% indeterminate QFT-IT. No effect of HIV on IP-10 responses	[61]
Lighter et al. (2009)	USA	Seven active TB, 12 with high risk, 87 moderate risk and 35 low risk of TB infection	Comparable positivity rates between biomarkers in risk groups. Negative effect of young age on mitogen-induced levels of IFN-γ but not IP-10. IP-10 responses after antigen stimulation correlated with exposure risk	[63]
Petrucci et al. (2008)	Nepal and Brazil	259 children were included	Positive correlation between IP-10 release and QFT-IT and TST results	[120]
Ruhwald et al. (2008)	Nigeria	59 high risk, 38 moderate risk and 23 low risk of TB infection	Comparable proportion of IP-10 test and QFT-IT positivity rates in high-risk (both 71%), medium-risk (8 and 21%, respectively) and low-risk (13 and 26%, respectively) groups. Association with young age and negative IP-10 test and TST results, but not QFT-IT results	[64]
Alsleben et al. (2011)	Germany	11 active TB, 22 high risk of TB infection and 14 with NTM infection	Positivity rates in TB patients was similar (91%) for the IP-10 test and the QFT- IT. No false-positive results in children with NTM infection	[62]

NTM: Non-tuberculous mycobacteria; QFT-IT: QuantiFERON^®-TB Gold In tube; TST: Tuberculin skin test.

Table 4. Selected studies assessing the performance of IP-10 in immunosuppressed patients with active tuberculosis or tuberculosis exposure.

Study (year)	Country	Study subjects	Results	Ref.
HIV-positive patients
Goletti et al. (2010)	India	28 confirmed TB with HIV coinfection (39% with CD4 cell count <200 cells/µl)	Borderline more positive by the IP-10 test (86%) than QFT-IT (61%). Negative trend of decreasing positivity rate with decreasing CD4 cells count for the QFT-IT, but not for the IP-10 test	[56]
Aabye et al. (2010)	Tanzania	65 confirmed TB patients with HIV coinfection (median CD4 cell count: 272 cells/µl)	Lower positivity rate in HIV-infected than uninfected individuals. Equal positivity rate of the IP-10 test (63%) and the QFT-IT (63%). Increased when combining tests (71%) compared with either test alone. Negative trend of decreasing positivity rate with decreasing CD4 cell count for the QFT-IT, but not for the IP-10 test	[55]
Kabeer et al. (2010)	India	50 confirmed TB patients with HIV coinfection (median CD4 cell count 86 cells/µl)	More IP-10 test positive (86%) than QFT-IT positive (70%)	[73]
Kabeer et al. (2011)	India	170 HIV-infected individuals with no signs of or history of TB infection (22% with CD4 cell count <200 cells/µl)	More IP-10 test positive (45%) than QFT-IT positive (38%). Increased when combining tests (48%) compared with the QFT, but not with the IP-10 test	[10]
Patients receiving immunosuppressive treatment
Chen et al. (2011)	Taiwan	56 TB patients, 24 TST-positive using QFT	IP-10 responses were higher in TST positive and in patients with active TB compared with patients with negative TST	[78]

QFT-IT: QuantiFERON^®-TB Gold In tube; RA: Rheumatoid arthritis; TB: Tuberculosis; TST: Tuberculin skin test.

Table 5. Overview of assays used to detect IP-10 and potential cutoffs linked to the assays.

IP-10 assay	Assay range^†(pg/ml)	Studies using this assay	Dilution factor^‡	Cutoff (pg/ml)	Validation study (year)
Luminex^® human IP-10 kit (Biosource/Invitrogen, UK)	5–18,000^§	[17,20,22,52,53,55,64]	×4, ×2, ×8, ×10	673 pg/ml (455 pg/ml also suggested) Ruhwald et al. (2008) [52]	Ruhwald et al. (2011) [53]; Aabye et al. (2010) [55]; Frahm et al. (2011) [20]; Ruhwald et al. (2008) [64]
ELISA DuoSet^® (R&D systems, MN, USA)	5–500	[21,50,51,56,78,121]	×10, ×5, ×2	698 pg/ml Goletti et al. (2010) [56]
ELISA IP-10 construction kit (Antigenix America, NY, USA)	Unknown	[61,120]	×2	3128 pg/ml Yassin et al. (2011) [61]
Luminex LincoPlex^®, (Linco Research, MO, USA)	3.1–10,000^§	[19,25,63]	×4, ×2	732 pg/ml^¶ Lighter et al. (2009) [63]
ELISA OptEIA™, (BD Biosystems, NJ, USA)	5–500	[10,54,70,73]	×20	300 pg/ml Kabeer et al. (2010) [73]; Kabeer et al. (2009) [70]	Kabeer et al. (2011) [10]

^†The range of the assay in undiluted sample.

^‡Factor in bold was used to set the cutoff.

^§Varies depending on the batch of kit and number of analytes included in the multiplex.

^¶Cutoff set in children.

Box 1. The five essential components in the IFN-γ release assays test.

• • Antigens: the specificity of the IFN-γ release assays is primarily linked to the antigens. The commercial tests use overlapping 16–26-mer peptides of the Mycobacterium tuberculosis proteins ESAT-6, CFP-10 and TB7.7. Several promising antigen alternatives are emerging.

• • Antigen presenting cells (APCs): APCs express MHC molecules with specificity for the antigens to T cells. If the T-cell receptor recognizes the peptide antigen on the MHC, both the T cell and the APC become activated. If the person tested has a poor fit between HLA genotype and antigens used in the test, the IFN-γ release assays are expected to perform poorly [117].

• • T cells: a high number of central memory T cells specific for the peptides presented on the APCs ensure a strong immune activation in vitro; the presence of regulatory T cells dampens this response [118,119].

• • Incubation: for immune activation to occur, the cells need time and heat, so the commercial tests incubate for 16–24 h at 37°C.

• • A readout biomarker: the commercial tests monitor the degree of immune activation by measuring the release of IFN-γ. If the release is higher than a cutoff, the test is deemed positive; if there is little or no response, the test is deemed negative [5,30]. As in any immune response, several cytokines and chemokines are expressed as a result of the immune activation, including the chemokine IP-10 [19,20].

Key issues

• • Tuberculosis (TB) is one of the most important public health problems, with over 8 million new cases and 2 million deaths annually.

• • Current focus is on identifying and treating active TB, but more accurate and cheaper tests for latent TB would enable early detection and targeted treatment of infected individuals at risk of developing active TB.

• • The IFN-γ release assays have operational and diagnostic qualities for latent TB, which makes them superior to the century-old tuberculin skin test; however, there is room for improvement, especially in terms of increased sensitivity, ease of use and cost.

• • IP-10 is an alternative diagnostic biomarker expressed at very high levels compared with the current marker, IFN-γ.

• • The studies included in this review suggest that IP-10 performs at least on par with IFN-γ, and is unaffected by bacillus Calmette–Guérin vaccination and non-tuberculous mycobacteria infection.

• • IP-10 has qualities that open the way for further developments, not available for IFN-γ, and several innovations have shown potential.

Financial & competing interests disclosure

M Ruhwald holds a research grant from the Danish National Advanced Technology Foundation, MG Aabye holds a Career PhD Scholarship from Copenhagen University and P Ravn is supported by The Danish Council for Independent Research – Medical Sciences. M Ruhwald, P Ravn and MG Aabye are registered as inventors on pending and issued patents disclosing the use of IP-10 as diagnostic biomarker for infection with Mycobacterium tuberculosis. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript.