Circulating cytokines in sarcoidosis: Phenotype-specific alterations for fibrotic and non-fibrotic pulmonary disease

Abstract

Aims

Sarcoidosis is a granulomatous disease of unknown etiology marked by tremendous clinical heterogeneity. Many patients enter remission with good long-term outcomes. Yet, chronic disease is not uncommon, and this important phenotype remains understudied. Identified alterations in local and circulating cytokines—specifically targeted for study, and often in the acute phase of disease—have informed our growing understanding of the immunopathogenesis of sarcoidosis. Our aim was to evaluate a broad panel of circulating cytokines in patients with chronic sarcoidosis. Among those with chronic disease, pulmonary fibrosis occurs in only a subset. To gain more insight into the determinants of the fibrotic response, we also determined if the phenotypes of fibrotic and non-fibrotic pulmonary sarcoidosis have distinct cytokine profiles.

Results

In patients with sarcoidosis compared to controls, IL-5 was decreased, and IL-7 was increased. Both of these comparisons withstood rigorous statistical correction for multiple comparisons. GM-CSF met a nominal level of significance. We also detected an effect of phenotype, where IL-5 was significantly decreased in non-fibrotic compared to fibrotic pulmonary sarcoidosis, and compared to controls. Compared to controls, there was a trend towards a significant increase in IL-7 in fibrotic, but not in non-fibrotic pulmonary sarcoidosis. In contrast, compared to controls, there was a trend towards a significant increase in GM-CSF in non-fibrotic, but not in fibrotic pulmonary sarcoidosis.

Conclusions

In a comprehensive evaluation of circulating cytokines in sarcoidosis, we found IL-5, IL-7, and GM-CSF to be altered. These findings provide a window into the immunopathogenesis of sarcoidosis. IL-7 is a novel sarcoidosis cytokine and, as a master regulator of lymphocytes, is an attractive target for further studies. By observing an effect of phenotype upon cytokine patterns, we also identify specific immune alterations which may contribute to clinical heterogeneity.

1. Introduction

Sarcoidosis is a clinically heterogeneous, granulomatous condition of unknown etiology. While the lungs are most often affected, multi-organ disease is common, and the heart, skin, eyes, liver, and/or nervous system can be affected [1]. Many patients enter remission after a limited duration of disease, experiencing minimal long-term or permanent organ damage [2,3]. However, chronic disease develops in a subset of patients, often marked by high morbidity and an increased risk of death. Outcomes are particularly poor for those who develop fibrotic pulmonary sarcoidosis. While there is no cure for sarcoidosis, immunosuppression is often employed to control inflammation, although the clinical response is variable [4].

The adjusted annual incidence of sarcoidosis is much higher for African Americans compared to European Americans, and African Americans are more likely to develop chronic disease [5–7]. These observations suggest that clinical phenotypes are mediated by distinct, genetically determined immune responses. Indeed, evolving evidence suggests that fibrotic pulmonary sarcoidosis has immune alterations distinct from non-fibrotic pulmonary disease [8–11].

The initial pathogenesis of sarcoidosis centers around antigen presenting cells interacting with naive CD4+ T cells, and leading to a T helper 1 (Th1)-polarized inflammatory response. Engaged CD4+ T cells release cytokines which drive the accumulation of macrophages, which organize to form granulomas [3]. TNF-α and IFN-γ are fundamental to granulomatous inflammation, and both are increased in sarcoid tissues [12]. While the oligoclonal expansion of CD4+ T cells suggests the presentation of antigen, the inciting antigen remains unidentified. In most patients, the activated immune response down regulates after the acute phase of disease, resulting in clinical remission. The mechanisms of persistent inflammation in a subset of patients, and the determinants of an extensive fibrotic response in an even smaller subset, are not known. Most studies of sarcoidosis have targeted acute disease. In contrast, subjects in our cohort predominantly have chronic disease, where events in the immunopathology may differ from those of acute inflammation.

In addition to TNF-α and IFN-γ, multiple other cytokines also are increased in sarcoidosis. While the full array of clinically important cytokine alterations is not known, the biology of cell signaling is no doubt complex and is seemingly concerted. Identified alterations in local cytokines, often specifically targeted for study, have provided insight into the immunopathogenesis [13–16]. Knowledge about corresponding circulating cytokines is more limited, yet there is increasing interest in the circulating immune system in sarcoidosis, where alterations in circulating monocytes, dendritic cells, T regulatory cells, and natural killer T cells have been demonstrated [17–23]. To better elucidate cytokine alterations in sarcoidosis, we evaluated a broad and relatively unbiased array of circulating cytokines in our cohort.

Combining clinically heterogeneous subjects into a single study cohort may impair detection of immune alterations, which may be phenotype-specific. Phenotyping enhances detection of potentially important alterations and allows comparison among phenotypes to reveal mechanisms of clinical diversity, and has been advocated for studies of sarcoidosis [24–26]. Beyond capturing the phenotype of chronic sarcoidosis, we compared cytokines in those with fibrotic and non-fibrotic pulmonary disease, as the clinical features of these phenotypes suggest that there are important differences in their underlying immunopathologies.

2. Materials and methods

2.1. Subjects

Serum samples were collected from 54 adults with biopsy-established sarcoidosis. Details regarding organ involvement, treatment, and smoking status were available for review. Eighteen controls were matched for age, gender and race. Race was self or clinician designated; while imperfect, it was the only available measure of race that was non-invasive and clinically feasible. Exclusion criteria for cases and controls included pregnancy, active infection, cancer, chronic inflammatory conditions other than sarcoidosis, or a glomerular filtration rate <40 mL/min/body surface area. Immunosuppression included use of ≥10 mg/daily of prednisone or the equivalent corticosteroid dose, or any other non-steroidal immunosuppressive therapy. Without selection bias, cases and controls were recruited consecutively in outpatient clinics via an institutional translational research program, and all patients provided informed consent. This study was approved by the University of Chicago Institutional Review Board.

2.2. Phenotypes

Fibrotic pulmonary sarcoidosis was established by the presence of honeycombing, bronchial distortion, and/or linear fibrosis on computerized tomography imaging. Non-fibrotic pulmonary sarcoidosis was established by non-fibrotic parenchymal or bronchial abnormalities on computerized tomography imaging.

2.3. Multiplex cytokine assays

IL-1β, IL-2, IL-4, IL-5, IL-6, IL-7, IL-8, IL-10, IL-12 (p70), IL-13, IL-17, G-CSF, GM-CSF, IFN-γ, MCP-1, MIP-1β, and TNF-α were measured using Bio-Plex human cytokine multiplex kits (Bio-Rad Inc., Hercules, CA) per manufacturer instructions. Normalization was performed to control for variance between plates.

3. Theory/calculation

Four cytokines (IL-1β, IL-2, IL-6, and IL-8) performed poorly, with less than 25% of values registering above background fluorescence, and were excluded from further analysis. GM-CSF performed well, but demonstrated significant background staining, which potentially diminished the signal when background fluorescence was subtracted. The data were non-parametric, and were expressed as medians with interquartile ranges. Differences were initially tested using the non-parametric 2-tailed Mann–Whitney U test. Cytokine data were log transformed for multi-variate logistic regressions, which controlled for potential influences of gender, race, smoking, and immunosuppression, as well as cytokine–cytokine interactions. Correlations were measured with the Spearman’s correlation test.

Pairwise Spearman correlations were calculated for each possible pairing of the 13 analyzed cytokines, resulting in 78 pairwise correlations. We expected to find correlations among cytokines, making a strict Bonferroni correction inappropriate, as the variables were not independent. Thus, we used a modified Bonferroni correction to account for the number of cytokines tested and their degree of correlation with each other as outlined in the following equation, where p_corr is the corrected p-value and p_obs is the observed p-value:

$$p_{\text{corr}}=p_{\text{obs}}\times 13\left[1-\frac{{\Sigma }_{i=1}^{78}\mid {\mathit{rho}}_i\mid }{78}\right]$$

This resulted in a p-value of <0.005 to define significance. For phenotype analyses, we controlled the family-wise type I error rate at 0.05 using a Bonferroni correction for the number of comparisons (three), which resulted in a calculated p-value of ≥0.017 to define significance.

4. Results

4.1. Subject and control demographics

Subject characteristics are shown in Table 1. Our cohort was enriched with African Americans and women. Many but not all subjects were on immunosuppression. Nearly all had chronic disease, with documented sarcoidosis activity for at least 2 years at the time of the study [27]. Approximately half of the patients with pulmonary disease had pulmonary fibrosis. There were no significant differences in demographic data between cases and controls except for a trend towards a higher rate of smoking among controls.

Table 1.

Demographic and clinical data for cases and controls.

Characteristic	Controls (N = 18)	Sarcoidosis subjects (N = 54)	p-Value
Age-yr Mean Range	52 37-74	50 31-75	0.523
Gender - no.(%) Female	13 (72)	40 (74)	1.000
Ancestry - no. (%) African American European American Hispanic	13 (72) 5(27) 0	43 (80) 10 (10) 1 (2)	0.611
Active Smoking - no. (%) Yes No Unknown	3(17) 13 (72) 2(11)	3 (5) 50 (93) 1(2)	0.100
Immunosuppression a - no. (%) Yes No Unknown	-	31 (57) 22 (41) 1 (2)	-
Years since diagnosis - no. (%) <2 years (acute disease) >2 years (chronic disease)	-	2 (4) 52 (96)	-
Sub-phenotype no. (%) Lofgren’s syndrome Fibrotic pulmonary sarcoidosis Non-fibrotic pulmonary sarcoidosis	-	0(0) 19 (35) 21 (39)	-

^aPreclnisone =10 mg/daily, or any other immunosuppressive therapy.

4.1.1. Levels of IL-5, IL-7, and GM-CSF were different for cases and controls

Uni-variate analyses were performed for cytokines with good bioassay performance. We identified IL-5, IL-7, GM-CSF, and G-CSF levels as potentially significantly different (p < 0.05) between cases and controls (Fig. 1). These four cytokines were then entered into a regression model for multi-variate analysis, where we found significant differences between cases and controls for IL-5 (p = 0.0047) and IL-7 (p = 0.0023). IL-5 was decreased, whereas IL-7 was increased in cases compared to controls. GM-CSF and G-CSF did not meet strict criteria for significance, with p-values of 0.068 and 0.082, respectively. To control for the effects of gender, race, and smoking, these variables were then included in the regression model. The p-value for G-CSF increased significantly when these variables were entered into the regression, and G-CSF was therefore not carried forward for further analysis. The p-values for IL-5, IL-7, and GM-CSF were not significantly different in the full regression analysis, suggesting that demographic data and smoking status did not account for the differences in these cytokines between cases and controls. Further, we did not find associations between treatment with immunosuppression and levels of IL-5, IL-7, or GM-CSF. In addition to IL-5 and IL-7, we included GM-CSF for phenotype analysis, with the hypothesis that there could be phenotype associations not evident in combined analysis.

Fig. 1.

Cytokines in combined analyses. IL-5, IL-7, and GM-CSF in controls and cases are shown. P-values were calculated by Mann–Whitney testing. Central tendencies are represented by the median, with interquartile ranges shown.

4.1.2. Cytokine patterns vary according to phenotype (Fig. 2)

Fig. 2.

Cytokines in phenotype analyses. IL-5, IL-7, and GM-CSF in controls and cases of phenotypes are shown. Patients were stratified by phenotypes defined in the methods. P-values were calculated by Mann–Whitney testing. Central tendencies are represented by the median, with interquartile ranges shown. Pulm = Pulmonary.

IL-5 was significantly decreased in non-fibrotic compared to fibrotic pulmonary sarcoidosis (p = 0.013), and compared to controls (p = 0.0004). Compared to controls, there was a trend towards a significant increase in IL-7 in fibrotic (p = 0.052), but not in non-fibrotic pulmonary sarcoidosis. In contrast, compared to controls, there was a trend towards a significant increase in GM-CSF in non-fibrotic (p = 0.054), but not in fibrotic pulmonary sarcoidosis.

4.1.3. Cytokines were correlated with each other, but cytokine–cytokine interactions did not influence case-control associations

There were many significant correlations among cytokines, with TNF-α and IL-12, TNF-α and IL-17, TNF-α and IFN-γ, IL-5 and IL-13, and GM-CSF and IL-4 demonstrating the highest degree of correlation (Fig. 3a). These correlations were also highly significant. Regression analyses were performed to control for potential cytokine–cytokine interactions which could have influenced the cytokine associations we detected between cases and controls: the p-values of these associations were not significantly different when models were run of IL-5, IL-7, and GM-CSF and their correlated cytokines (data not shown). Interestingly, when correlation matrices were done separately for fibrotic and non-fibrotic pulmonary disease, the patterns of correlations were different according to phenotype, and only four out of 23 correlation pairs were shared (Fig. 3 b and c).

Fig. 3.

Matrix of cytokine correlations. Multiple positive correlations among cytokines are demonstrated in our combined cohort (3a). Distinct patterns of cytokine correlations are observed for fibrotic (3b) and non-fibrotic (3c) pulmonary sarcoidosis. There are no significant negative correlations. Spearman’s Rho value 100× is indicated in each pair box for which a significant correlation was detected. Light shade represents a p-value <0.05–0.0001. Dark shade represents a p-value <0.0001.

4.1.4. Peripheral lymphocyte counts did not correlate with levels of IL-7

IL-7 has a hematopoietic effect, and levels increase during lymphopenic conditions to regulate lymphocyte counts towards normal [28,29]. In this way, IL-7 has been designated by some to be a “master regulator” of lymphocytes [29]. Peripheral lymphopenia is common in chronic sarcoidosis [30]. To evaluate whether alterations in IL-7 in sarcoidosis derive from altered lymphocyte counts, we correlated IL-7 with lymphocyte counts in the 47 subjects with a white blood count differential performed at the time of the serum sample collection. We found no significant correlation between levels of IL-7 and lymphocyte counts (Fig. 4).

Fig. 4.

IL-7 levels and lymphocyte counts. IL-7 levels were not correlated with lymphocyte counts (p = 0.566) in the 47 patients with a leukocyte differential performed at the time of the serum sample collection.

5. Discussion

In this study, we found altered levels of circulating IL-5, IL-7, and GM-CSF in patients with chronic sarcoidosis. Circulating IL-5, a T helper 2 (Th2) cytokine, was decreased in sarcoidosis compared to controls. This suggests that circulating cytokines may not merely be due to “spillover” from local sites of active inflammation, but rather may reflect primary alterations, such as suppression of Th2 cells, in the circulating immune system [31]. In contrast to IL-5, IL-7 and GM-CSF were increased in sarcoidosis compared to controls. Historically, IL-7 levels have not been studied in sarcoidosis. However, the up-regulation of IL7 expression in sarcoid lung tissue, alterations in the IL7R gene in patients with sarcoidosis, and increased circulating IL-7 in female patients with severe sarcoidosis have been recently reported, and are of particular interest in light of our current findings [32–35]. IL-7 has a critical role in lymphocyte regulation and homeostasis, and therefore is an attractive target for further studies in sarcoidosis, a lymphocyte-mediated disease. Finally, GM-CSF regulates and activates monocytes, and GM-CSF seems to be up-regulated in sarcoid lung tissue [16]. Previous studies on circulating GM-CSF levels have yielded mixed results, which may relate to different experimental techniques [13,18]. Our result suggests that GM-CSF—an important signal to direct antigen presenting cells, including macrophages—is indeed elevated and potentially important in sarcoidosis. While GM-CSF was not as strong in our analysis, this may have been a function of high background staining.

We are not able to replicate some previously published results of circulating cytokine alterations. Methodological differences may account for our different, yet not contradictory findings: whereas many studies have employed single-cytokine ELISA tests, used plasma or culture supernatant samples, or had cohorts of newly diagnosed disease, we employed a multiplex assay, used serum samples, and had a cohort of chronic disease [18,36]. Our intent was not to detect all possible alterations in circulating cytokines, for which dedicated singleplex assays would likely perform better for a given cytokine. Rather, our aim was to advance our understanding of cell signaling in sarcoidosis through the evaluation of a broad array of cytokines and chemokines, including the study of potentially novel cell signals.

Our findings withstood controlling for variables predicted to be potentially confounding. Clinical and demographic variables did not influence the cytokine associations we observed, and cytokine levels did not correlate with use of immunosuppression. While we found many significant correlations among cytokines, controlling for these correlations did not significantly alter our findings. Finally, IL-7 levels did not correlate with lymphocyte counts, suggesting that increased IL-7 in sarcoidosis is a primary phenomenon.

We found specific cytokine patterns for fibrotic and non-fibrotic pulmonary sarcoidosis. The determinants of pulmonary fibrosis in a subset of patients with chronic disease are not known. However, it is increasingly recognized that pulmonary fibrosis in sarcoidosis likely occurs via specific inflammatory processes which are not shared by all patients, and fibrosis-specific immune and genetic alterations recently have been reported [8–11]. Indeed, we found substantial differences in the patterns of cytokine correlations between these phenotypes. It has been suggested that fibrotic sarcoidosis may result from the transition from a Th1 to a Th2-biased response [37]; our finding of significantly lower levels of IL-5, a Th2 cytokine, in fibrotic compared to non-fibrotic pulmonary sarcoidosis does not support such a transition. Certainly more work in this domain is indicated. Our finding of a trend towards increased IL-7 in fibrotic but not non-fibrotic pulmonary sarcoidosis raises the possibility that this cytokine may have a role in the fibrotic response to sarcoid inflammation.

Most of our subjects were female and African American. The female predominance in our cohort exceeds the higher rate of disease often reported for women [5]. We speculate that women may experience increased disease-associated morbidity compared to men, or for other reasons may more readily seek medical care services for sarcoidosis [38–40]. A predominance of African American patients in our cohort in is line with the ancestral demographics of the catchment area of our medical center. In addition, sarcoidosis can be more severe in African Americans, and disease severity should relate to utilization of tertiary care services, such as those offered by our University Medical Center [5,12,41]. Studies with more gender balanced cohorts and varied ancestral groups will help to generalize our findings. Nonetheless, as we have done here, it is important to study patient groups most likely to be severely affected by sarcoidosis.

Our findings in IL-5, IL-7, and GM-CSF in patients with chronic sarcoidosis provide a window into the immunopathogenesis of disease. They also support the biomarker potential of the circulating immune system in sarcoidosis. Correlating serum and lung cytokines will help to elucidate the similarities and differences between systemic and tissue-specific cytokine alterations, and are important next steps. In addition, repeating our studies in subjects with acute disease will provide insight into the relationship between cytokine alterations and the phase of disease. Finally, by observing an effect of phenotype upon cytokine patterns, we identify specific immune alterations which may contribute to the clinical heterogeneity of sarcoidosis. Further defining the differences between fibrotic and non-fibrotic pulmonary sarcoidosis is important, and will advance our understanding of the fibrotic phenotype, an uncommon yet particularly severe manifestation of disease.