Increased plasma cytokine concentrations in tuberculosis patients
Initially, we determined seven candidate cytokines in plasma samples from acute tuberculosis patients (n = 56) and household contacts (Contacts; n = 40). Six cytokines, i.e., IL-6, IP-10, IL-10, IL-22, IFN-γ, and GM-CSF, were significantly higher in plasma samples from patients with tuberculosis (Fig. 1a). IL-8 was not different between the study groups (Fig. 1a). IL-6, IP-10, IL-10, and IL-22 were detectable in the majority of tuberculosis patients (> 50%), whereas IFNγ and GM-CSF were measurable only in minor subsets of both study groups (IFNγ: 41.1% in patients, 25.0% in contacts; GM-CSF: 21.4% in patients, 2.5% in contacts). Hence, we focused on IL-6, IP-10, IL-10, and IL-22 for further analyses. Comparison of different cytokines in individual donors detected strong correlation between IL-6 and IP-10 (r = 0.62, p < 0.0001) as well as for IL-22 that showed strong correlation with IL-6 (r = 0.50, p < 0.0001), IL-10 (r = 0.54, p < 0.0001), and IP-10 (r = 0.47, p < 0.0001) (Fig. 1b). Moderate, but significant, correlations were detected between the other cytokines (Fig. 1b). These results identified concomitantly increased plasma cytokine candidates in patients with acute tuberculosis.
IL-6 and IP-10 plasma concentrations discriminate tuberculosis patients from household contacts
Next, we analyzed the capacity of different cytokine candidates to discriminate tuberculosis patients from contacts. IL-6 was the most potent factor for discrimination in ROC analyses [area under the curve (AUC): 0.94, p < 0.0001]. IP-10 had also strong capacity to discriminate between the study groups (AUC: 0.88, p < 0.0001), whereas IL-22 (AUC: 0.79, p < 0.0001) and IL-10 (AUC: 0.69, p = 0.0020) were less efficient for discrimination (Fig. 1c). Combinations of different cytokines (i.e., sums) did not increase discriminatory power as compared to IL-6 alone (Supplementary Fig. 2). We concluded that single plasma cytokine levels, namely of IL-6 and IP-10, were most potent for classification of study participants. Next, we analyzed if M. tuberculosis sputum burden as a measure of disease severity was associated with plasma cytokine levels.
Tuberculosis patients with paucibacillary disease manifestation have lower IL-6 and higher IL-10 plasma concentrations
M. tuberculosis sputum burden was determined and a subgroup of patients presented with negative sputum smear and culture prior to therapy (‘paucibacillary’; TB-Pb). These TB-Pb patients (n = 20) were not included in initial analyses, but were now compared to sputum-positive tuberculosis patients (TB-Sp; n = 40). In addition, we separated a subgroup of tuberculosis patients, whose disease outcome was death. (TB-Dc; n = 19). The majority of TB-Dc patients were part of the TB-Sp group (n = 16), but three TB-Dc patients were TB-Pb (Fig. 2; Table 1). No differences were detected for IP-10 and IL-22 between the subgroups (Fig. 2a). Lower IL-6 plasma levels were detected in TB-Pb as compared to TB-Sp patients (p = 0.049), whereas no IL-6 differences were seen for TB-Dc (Fig. 2a). Notably, IL-10 was significantly higher in TB-Pb and TB-Dc patients (p = 0.0016 and p = 0.0006, respectively) (Fig. 2a). Due to opposing results for IL-6 and IL-10, we calculated IL-6/IL10 ratios for individuals from all subgroups. IL-6/IL10 ratios were significantly higher in TB-Sp patients as compared to the TB-Pb (p < 0.0001) and TB-Dc patients (p = 0.017; Fig. 2b). Actually, similar IL-6 and IL-10 plasma levels were detected in TB-Pb, whereas median levels for IL-6 were approximately tenfold higher than for IL-10 in TB-Sp patients (Fig. 2b). Marked differences in concomitant IL-6 and IL-10 levels between subgroups prompted us to determine the discriminatory power of different factors. IL-6 alone showed discriminatory capacity between TB-Sp and TB-Pb (AUC: 0.79, p = 0.0003), and IL-10 was effective for discrimination of TB-Sp and TB-Dc (AUC: 0.82, p < 0.0001) Table 2. Notably, IL-6/IL-10 ratios showed strongest capacity to discriminate TB-Sp and TB-Pb (AUC: 0.88, p < 0.0001), suggesting that opposing IL-6 and IL-10 levels characterize these tuberculosis patient subgroups (Table 2).
IL-6 and IP-10 plasma concentrations decline during anti-mycobacterial treatment of tuberculosis patients
Next, plasma samples of TB-Sp patients were compared prior to treatment and at follow-up (i.e., 6 and 16 weeks after treatment start). IL-10 and IL-22 concentrations showed no significant decline during treatment and IL-10 even increased between week 6 and 16 (Fig. 3a). In contrast, IL-6 and IP-10 concentrations declined during treatment (p < 0.0001, p = 0.003, respectively) (Fig. 3a). Both, IL-6 and IP-10, showed decreased concentrations at week 6 (p < 0.0001, p = 0.013, respectively) and IL-6 further declined between week 6 and 16 (p = 0.006; Fig. 3a). Since IL-6 and IP-10 plasma concentrations correlated prior to treatment (Fig. 1b), we next analyzed correlation during treatment course. IL-6 and IP-10 correlated in tuberculosis patients prior to treatment (r = 0.4, p = 0.011) and at week 6 (r = 0.34, p = 0.030). However, the correlation between IL-6 and IP-10 vanished until week 16 (r = 0.03, p = 0.84). These results suggested differences in time courses between plasma cytokines during treatment and identified IL-6 and IP-10 as an early marker of treatment response in tuberculosis patients.
Rapid and slow treatment responders differ in decline of IL-6 plasma concentrations between week 6 and 16
Treatment efficacy in tuberculosis patients can be assessed by M. tuberculosis detection in sputum samples. The samples were analyzed at baseline as well as 6, 9, 12, and 16 weeks after treatment initiation. TB-Sp patients with negative sputum samples already after 6 or 9 weeks were classified as ‘rapid’ treatment responders, whilst those with any positive test after week 6 were classified as ‘slow’ treatment responders (Supplementary Figure 1). Rapid and slow responders showed no differences in IL-6, IP-10, IL-22, and IL-10 plasma concentrations prior to treatment (Supplementary Figure 3a). Time course measurements detected no differences for IL-22 and IL-10 during treatment for both rapid and slow responders and IP-10 showed a significant decrease for the slow responder group (p = 0.002; Supplementary Figure 3b). Notably, rapid and slow responders differed in IL-6 changes during treatment (Fig. 4a, b). Rapid responders had significantly declined IL-6 concentrations at week 6 (p = 0.004) and no differences between week 6 and 16 (Fig. 4a). In contrast, slow responders had significantly declined IL-6 plasma concentrations also between week 6 and 16 (p = 0.0003) (Fig. 4b). To compare individual IL-6 changes between BL and week 6 as well as week 6 and week 16 for rapid and slow responders, we calculated fold-changes (i.e., BL vs. week 6; week 6 vs. week 16) for each individual donor. Study group comparisons detected similar fold-changes between week 0 and week 6 (median fold-change RR: 1.97; SR: 1.87; p = 0.799; Fig. 4c), but only slow responders showed increased fold-changes between week 6 and 16 (median fold-change RR: 1.012; SR: 1.931; p = 0.033; Fig. 4c). We concluded that although the interindividual variability in initial IL-6 levels and changes was marked (Fig. 4a, b), study groups of rapid and slow responders differed in the time interval of IL-6 normalization. Whereas IL-6 plasma levels normalized in rapid responders within the first 6 weeks, this process was still ongoing between week 6 and 16 in slow responders.