Host biomarker-based quantitative rapid tests for detection and treatment monitoring of tuberculosis and COVID-19

Summary Diagnostic services for tuberculosis (TB) are not sufficiently accessible in low-resource settings, where most cases occur, which was aggravated by the COVID-19 pandemic. Early diagnosis of pulmonary TB can reduce transmission. Current TB-diagnostics rely on detection of Mycobacterium tuberculosis (Mtb) in sputum requiring costly, time-consuming methods, and trained staff. In this study, quantitative lateral flow (LF) assays were used to measure levels of seven host proteins in sera from pre-COVID-19 TB patients diagnosed in Europe and latently Mtb-infected individuals (LTBI), and from COVID-19 patients and healthy controls. Analysis of host proteins showed significantly lower levels in LTBI versus TB (AUC:0 · 94) and discriminated healthy individuals from COVID-19 patients (0 · 99) and severe COVID-19 from TB. Importantly, these host proteins allowed treatment monitoring of both respiratory diseases. This study demonstrates the potential of non-sputum LF assays as adjunct diagnostics and treatment monitoring for COVID-19 and TB based on quantitative detection of multiple host biomarkers.


INTRODUCTION
Since the end of 2019, COVID-19, the devastating respiratory disease caused by severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) has plagued humans, swiftly resulting in a global pandemic. This has led to over 500 million cases and 6 $ 3 million deaths (July 2022). 1 Currently, reverse transcription polymerase chain reaction (RT-PCR) specific for SARS-CoV-2 based on nasopharyngeal swabs are used for diagnosis. 2 SARS-CoV-2 is transmitted via the respiratory route with an average incubation time of days. 3 Whereas most individuals infected with this virus are asymptomatic or experience mild to moderate disease, 3,4 still a substantial number of patients is hospitalized because of severe respiratory problems. [5][6][7] Moreover, in individuals with severe COVID-19, a proinflammatory cytokine storm can be observed inducing respiratory distress. 4,8 Despite the COVID-19 pandemic, tuberculosis (TB) remains one of the most lethal infectious diseases, mainly in low-income countries, but also presenting an unignorable threat in Europe. 1,9 In 2020, around 10 million individuals developed TB and 1 $ 5 million deaths were attributed to this disease. 9 It is estimated that one-quarter of the global population is latently infected with Mycobacterium tuberculosis (Mtb) and approximately 3-10% of those individuals are at risk of developing active TB during their lifetime. 10 As a poverty-associated disease, TB places a huge burden on health care services of low-and middle-income countries. Of the estimated 10 million patients, 3 $ 6 million active TB cases are not diagnosed or reported. 11 Early diagnosis, followed by prompt and successful treatment will reduce Mtb transmission 11 and prevent disease-associated mortality. 10 Active TB is diagnosed by detection of the pathogen in sputum using microbiological, microscopic or genetic methods, which are often expensive, timeconsuming, resource intense, require specially trained staff, and are less sensitive in HIV co-infected individuals. 12 Besides, sputum has relatively large sampling error resulting in false negative outcomes as it is difficult to obtain, especially from children. [13][14][15][16] Also, sputum cultures carry the risk of infection resulting in unusable data. Another hurdle of the use of sputum-based diagnostics is its lack of point-of-care Using the luminescent upconverting reporter particle (UCP) technology combined with low-cost, fieldfriendly immune-chromatography, we have previously developed and field-evaluated quantitative lateral flow assays (LFAs). [24][25][26][27][28][29][30][31] These UCP-LFAs are suitable for accurate quantification of cytokines, acute phase proteins and antibodies in serum, stimulated whole blood, pleural fluid, saliva, and fingerstick blood. 29,[32][33][34][35] The user-friendly, low complexity UCP-LFAs do not require sophisticated analytical laboratory equipment. Portable battery-operated readers provide full instrument-assisted analysis which also avoid potential operator bias. The UCP-LFAs represent POC alternatives for the more elaborate and time consuming, laboratory-based enzyme linked immunosorbent assay (ELISA). Another advantage of the UCP-LFA is that it permits multiplexing to measure several markers simultaneously allowing a biomarker signature to be assessed in field settings. 27,31 Exploratory proteomics previously identified a promising host protein signature that distinguished active TB patients from other respiratory diseases (ORD) with signs and symptoms suggestive of TB in an African setting. 36 Based on this signature, we have applied C-reactive protein (CRP), apolipoprotein-A1 (ApoA1), inducible protein (IP)-10/C-X-C motif chemokine 10 (CXCL-10), and serum amyloid A (SAA) to the UCP-LFA format. 29 In addition, we have developed UCP-LFAs for interleukin-6 (IL-6), S100 calcium-binding protein A12 (S100A12), and ferritin in view of their role in tuberculous meningitis, 37,38 inflammatory disorders, 39 leprosy, 29,31 and iron homeostasis in Mtb, 36,40 respectively. In this study, we have used UCP-LFAs to rapidly assess serum levels of these host proteins in European TB and COVID-19 patients to investigate to what extent these can identify and discriminate between theold and new respiratory disease.

Assessment of host biomarkers for active TB in a European cohort
Previously, host biomarkers were identified by Luminex as discriminatory between TB and ORD in African settings. 37 The aim of the present study was to assess whether these host biomarkers could also allow the identification of TB in a European hospital setting, UCP-LF strips were developed for quantitative measurement of seven cytokines [24][25][26]35,41,42 and used for analysis of banked sera from LTBI and pulmonary TB patients collected in Europe (TB cohort 1, Figure 1). Serum levels for CRP, ferritin, IL-6, IP-10, SAA1/A2, and S100A12 were significantly higher in the TB group (p<0 $ 0001, p = 0 $ 0325, p = 0 $ 0006, p = 0 $ 01, p<0 $ 0001, and p<0 $ 0001, respectively), but no significant difference was found for ApoA1 (p = 0 $ 3244). CRP, SAA1/A2, and S100A12 were the most discriminatory when comparing TB to LTBI (AUCs: 0 $ 87-0 $ 96). Chest X-ray severity did not affect the levels for any of the cytokines ( Figure S1).
The total number of six biomarkers (ApoA1, CRP, ferritin, IL-6, IP-10, and SAA1/A2) scoring above the cutoff value (NUM score), based on the Youden's index for each marker, was calculated. 31 Using a cut-off of R3 positive markers, accuracy for active TB (with LTBI as control group) with a sensitivity of 83% (CI: 66 $ 4 to 92 $ 7) and specificity of 97% (CI: 82 $ 8 to 99 $ 8) (AUC: 0 $ 94; Figure S2) was found. An overview of medians for each marker per test group/comparison and the cut-off values for each biomarker are displayed in Tables 1 and 2. Other NUM score combinations and corresponding AUC and Sn/Sp were evaluated and shown in Table S3. Biomarkers were deleted from the NUM score based on their contribution (AUC); the biomarker with the lowest AUC was removed first and this was repeated until the most discriminatory iScience Article marker for that comparison remained. Noteworthy is that a 2-marker NUM score of CRP and SAA1/A2 (using a cut-off of R1) resulted in an AUC of 0 $ 91 and similar Sn/Sp (83%/97%) as the 6-marker NUM score, in the comparison of LTBI versus TB.
Moreover, comparison of available QuantiFERON data and NUM score results in TB cohort 1 showed that a NUM score based on the levels of six (ApoA1, CRP, ferritin, IL-6, IP-10, and SAA1/A2) as well as two (CRP and SAA1/A2) host proteins, allowed identification of individuals with active TB whose QuantiFERON test was negative ( Figure S3). In addition, whereas QuantiFERON was not able to accurately distinguish between active TB and LTBI, both 2-and 6-marker NUM scores successfully discriminated active from latent TB in QuantiFERON-positive individuals.

Analysis of host proteins for COVID-19 in a European cohort
In 2020, COVID-19 and SARS-CoV-2 infection posed TB diagnostics with an additional respiratory disease in the differential diagnosis. Therefore, UCP-LFA for the same seven biomarkers were also used to analyze hospitalized Dutch COVID-19 patients (n = 102) and healthy controls (n = 39); the latter sampled either in Figure 1. Evaluation of host biomarkers for TB and LTBI in a European cohort Levels of IL-6, IP-10, ferritin, SAA1/A2, CRP, ApoA1, and S100A12 were measured by UCP-LFA in serum samples of TB patients (n = 30; green dots) and LTBI (n = 29; gray dots) from Europe. Median values for each group are indicated by horizontal bars. Mann-Whitney U tests were performed to determine the statistical significance between groups (pvalues: *p%0 $ 05, **p%0 $ 01, ***p%0 $ 001, ****p%0 $ 0001). iScience Article (n = 12) or before 2020 (n = 27; Figure 2). UCP-LFAs showed that for all seven proteins, serum levels were significantly different between the COVID-19 patients and healthy controls, with pvalues ranging from p = 0 $ 0008 to p<0 $ 0001 ( Figure 2). Of note was that for six markers (CRP, ferritin, IL-6, IP-10, SAA1/A2, and S100A12) higher values were detected in COVID-19 sera whereas for ApoA1, significantly lower levels were detected in the COVID-19 group (p<0 $ 0001). The markers with the highest discriminatory potential between these two groups included CRP, ferritin, and SAA1/A2 (AUCs ranging from 0 $ 94 to 0 $ 98). For ApoA1, ferritin, IP-10, and SAA1/A2, no significantly different levels were observed at hospital admission -between COVID-19 patients with moderate disease, severe disease or fatal outcome (Figures 2 and S1). Cytokines CRP, IL-6, and S100A12 were significantly increased in patients with fatal outcome compared to those with moderate disease. In addition, higher CRP levels were observed in severe iScience Article compared to moderate disease. Moreover, CRP, ferritin, IL-6, IP-10, and SAA1/A2 did not show any significant differences in serum samples at hospital admission of COVID-19 patients who had already received anti-inflammatory treatment before the first sample collection compared to those who had not received any treatment yet ( Figure S4). S100A12, on the other hand, significantly decreased on anti-inflammatory treatment, whereas ApoA1 increased.

DISCUSSION
Early detection and treatment of communicable diseases, particularly those spread via aerosols to the respiratory tract, is vital to stop transmission. As shown in WHO records, migrants traveling to Europe may also iScience Article carry a higher risk of Mtb infection. [43][44][45][46] Thus, in view of the COVID-19 pandemic and the continuous migration to (Western) Europe from areas with multi-and even extensively drug-resistant TB, 47,48 specific tools for screening and (rapid) diagnosis of TB, become even more crucial.
This study describes the performance of the rapid and quantitative measurement of seven markers in the UCP-LFA format. This POC platform is highly adaptable toward implementation of the number and variety of biomarkers and was developed for (simultaneous) assessment of cytokines, acute phase proteins, 49,50 growth factors, 51 antibodies, and complement markers. 52 In this study, LF strips for detection of one biomarker were applied for seven host serum proteins, representing a tentative signature relevant for TB triage, comprising ApoA1, CRP, ferritin, IL-6, IP-10, SAA1/A2, and S100A12.
In comparison to sera from individuals with latent TB, we found significantly increased levels of CRP, ferritin, IL-6, IP-10, SAA1/A2, and S100A12 in those from active TB patients. CRP, SAA, and ferritin, all acute phase proteins (APPs) synthesized in the liver, 53 are associated with TB 54-60 and inflammation. 36,61,62 Once secreted by monocytes, endothelial cells or fibroblasts, IP-10 can act as a chemotactic mediator for both innate and adaptive immune cells, 63 and is recognized as a marker in HIV-positive TB patients. 64 IL-6 is a proinflammatory cytokine produced by macrophages. 53,65 In line with previous literature, these five markers were also elevated in COVID-19 patients compared to healthy controls. 5,65-70 Furthermore, two additional markers ApoA1 and S100A12, also showed diagnostic potential for COVID-19 confirming the earlier reported downregulation of ApoA1 and the upregulation of S100A12 in COVID-19 patients in France and China. [71][72][73] ApoA1 is believed to play an important role in modulating inflammation as it can inhibit monocyte activation by binding to T cells 74 resulting in decreased ApoA1 serum levels during inflammation. 75 On the other hand, the phagocytic protein S100A12 can exhibit proinflammatory effects and is found in high concentrations at sites of inflammation. 39 When levels of host proteins were compared between COVID-19 and TB patients, five (ApoA1, CRP, ferritin, SAA1/A2, and S100A12) showed promising discriminatory potential as all but ApoA1 were significantly higher in COVID-19 patients' sera. IL-6 and IP-10 seemed promising biomarkers successfully discriminating TB from LTBI and COVID-19 from healthy controls. However, these two markers did not show potential in distinguishing TB from COVID-19 disease. Although CRP is generally described as a biomarker for bacterial infection, 76,77 higher levels of this acute phase protein were observed in COVID-19 patients compared to TB. The excessive levels of these biomarkers could possibly be explained by the acute and exorbitant nature of local and systemic inflammation and immune activation observed in COVID-19 patients. [78][79][80] Despite the fact that certain host proteins are detectable in healthy as well as diseased individuals, the quantitative nature of the UCP-LFA allows the discrimination of TB and COVID-19 at POC because levels varied significantly between the test groups. In this respect, it should be noted that for each comparison (TB versus LTBI; COVID-19 versus HC; TB versus COVID-19) a distinct cut-off was required per biomarker. Besides allowing quantification of host biomarkers at POC, the addition of pathogen-specific biomarkers to a signature based on proteins that are not disease-specific (such as the host proteins evaluated here), increases the diagnostic potential. This was recently demonstrated for leprosy diagnostics in which the simultaneous detection of the cytokines described in this study and anti-Mycobacterium leprae PGL-I IgM 29 into a multi-biomarker test (MBT) allowed the discrimination of patients with both paucibacillary and multibacillary leprosy from controls in high-but also in non-endemic areas. 31 In the case of Mtb infection, however, a specific and sensitive antibody has not yet been identified. 81,82 This study aimed to evaluate whether serum levels of the selected host proteins can be used to detect TB and COVID-19 using in-sample validation. These markers need to be validated in an independent cohort, in which both patient groups are recruited prospectively at the same site and time. However, in-sample validation using NUM scores was assessed. This approach, based on serum levels of ApoA1, CRP, ferritin, IL-6, IP-10, and SAA1/A2, yielded an 83% sensitivity and 97% specificity for detection of TB versus LTBI. Noteworthy is that not all six markers might be necessary, as a 4-marker NUM score combining CRP, IL-6, IP-10, and SAA1/A2 in this study yields a sensitivity of 87%, which nears the WHO-recommended sensitivity (90%) for triage TB tests. 83 Similarly, using a NUM score based on 7 host proteins (additionally including S100A12), COVID-19 patients were separated from healthy controls with sensitivity of 93% and specificity of 100%, respectively. A 3-marker NUM score (CRP, ferritin, and SAA1/A2) resembles the ll OPEN ACCESS 10 iScience 26, 105873, January 20, 2023 iScience Article above-mentioned test performance with sensitivity of 95% and specificity of 97%. The seven markers could also distinguish TB from COVID-19 with 91% sensitivity and 87% specificity. A combination of CRP and SAA1/A2 might already be sufficiently discriminatory, with Sn/Sp of 94%/80%.
Longitudinal analysis of both TB and COVID-19 cohorts, indicated that biomarker analysis allows immunomonitoring of treatment for both groups. CRP, ferritin, IL-6, and S100A12 all declined during TB treatment, whereas ApoA1 levels increased over time. Furthermore, in line with the use of IL-6 inhibitors for treatment of COVID-19 patients attempting to mediate inflammation, 65,80 IL-6 levels declined significantly in these patients on treatment. Of interest, baseline IL-6 levels were significantly higher in patients with fatal outcome compared to those with moderate disease. Nevertheless, contradictory effects on mortality in several clinical trials using IL-6-blocking agents were reported. 65 Markers described to be valuable in predicting clinical outcome for COVID-19 in other studies included CRP, SAA, ferritin, and S100A12. 62,69,73,84,85 For two of those markers, CRP and S100A12, increased levels were indeed detected in our study at hospital admission in patients with fatal outcome compared to those with moderate disease.
Although some patients had been treated with anti-inflammatory medication before hospital admission, this did not affect biomarker levels for CRP, ferritin, IL-6, IP-10, and SAA1/A2. However, only S100A12 and to a lesser extent also ApoA1 already showed a significant effect, arguing for the potential of these proteins as biomarkers for monitoring of early treatment effect.
Our study shows that the UCP-LFA format cannot only provide (adjunct) rapid diagnostics for (triage of) chronic diseases such as TB and leprosy, 31,34,36 but also for more acute diseases including COVID-19. Of note, we demonstrated that in contrast to QuantiFERON-TB Gold which detects Mtb infection but cannot discriminate between active TB and LTBI, 86 the host proteins assessed here showed significant differences between active TB and LTBI (AUC: 0 $ 88-1 $ 00). Application of these biomarkers in UCP-LFA as adjunct diagnostic tools for triage of TB, can be useful to assess whether further diagnostic testing is warranted thereby reducing the costs for referrals for SARS-CoV-2 PCR and/or GeneXpert.

Limitations of the study
It should be noted that the COVID-19 cohort studied concerned patients hospitalized in 2020 who were all severely ill but admitted at various COVID-19 stages, which was reflected by the detected range in host biomarker levels. Therefore, in areas endemic for both TB and COVID-19, it will be feasible to triage TB (accepting lower sensitivity) but challenging to diagnose TB based on the studied biomarkers, because these reflect an individual's disease and inflammation state which may be comparable for these diseases. Consequently, the outcome of non-disease specific, host biomarker-based diagnostics should always be considered in the context of the individual's clinical presentation and the burden of diseases in an area.
Future studies should thus be focused at simultaneous recruitment of TB as well as all ORD, including other, non-European settings. To this end, the UCP-LFA platform can facilitate replacement of biomarkers to generate more optimal signatures for various use-cases including discrimination of TB and COVID-19.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

ACKNOWLEDGMENTS
We thank all patients and healthy volunteers for taking part in this study. We thank Corine Prins for sampling TB cohort 2. Formats of the diagnostic platform evaluated in this study were also assessed in parallel studies aimed at user-and field-friendly diagnostics for active tuberculosis: EDCTP funded projects AETBC (IP_2009_32040) and Screen-TB (DRIA2014-311).

DECLARATION OF INTERESTS
The authors declare that they have no conflict of interest. Mann-Whitney U tests and Kruskal-Wallis tests were performed to determine the statistical significance between two and three independent groups, respectively. Wilcoxon matched pairs signed rank tests and Friedman tests with Dunn's correction were performed to determine the statistical significance between two and three paired timepoints, respectively. Plot receiver operating characteristic (ROC) curves were created and sensitivity (Sn), specificity (Sp) and the area under the curve (AUC) were calculated to evaluate test performance. A cut-off for positivity for each biomarker was determined by calculating the maximal Youden's index. 97 For each of the individuals tested, an extra parameter (NUM score), 31 was calculated, representing the number of biomarkers that scored above the threshold of positivity based on the Youden's index. Three comparisons were made: TB vs. LTBI, COVID-19 vs. healthy controls, and COVID-19 vs. TB. For each comparison, a NUM score was calculated with the number of biomarkers used ranging from 1 to 7. ll OPEN ACCESS