Comprehensive serum cytokine analysis identifies IL-1RA and soluble IL-2Rα as predictors of event-free survival in T-cell lymphoma

This study assessed blood cytokine levels as prognostic factors in untreated patients with T-cell neoplasms. Soluble IL-2Rα and IL-1RA were predictive of inferior event-free survival in untreated T-cell lymphoma. These serum biomarkers have potential use for prognosis and targeted therapy.

Abstract

Background

T-cell malignancies are heterogeneous in their clinical presentation and pathology, and have a poor prognosis. New biomarkers are needed to predict prognosis and to provide insights into signal pathways used by these cells. The goal of this study was to evaluate pretreatment serum cytokines in patients with newly diagnosed T-cell neoplasms and correlate with clinical outcome.

Patients and methods

We evaluated 30 cytokines in pretreatment serum from 68 untreated patients and 14 normal controls. Significantly elevated cytokines were correlated with patterns of abnormalities, event-free survival (EFS) and overall survival (OS).

Results

Our data demonstrated significantly elevated levels (versus controls) of seven cytokines—epidermal growth factor (EGF), IL-6, IL-12, interferon gamma-induced protein (IP)-10, soluble interleukin (sIL)-2Rα, monokine induced by gamma interferon (MIG), and IL-1RA—in all T-cell neoplasms (P < 0.05). In the angioimmunoblastic subset, all seven cytokines except IP-10 and in the peripheral T-cell lymphoma (TCL)-not otherwise specified subset, only IP-10, sIL-2Rα, MIG, and IL-8 were statistically elevated compared with control. Of these, elevated cytokines all but EGF were predictive of an inferior EFS; IL-1RA, sIL-2Rα, and MIG predicted an inferior OS. In a multivariate analysis, sIL-2Rα [hazard ratio (HR) = 3.95; 95% confidence interval (CI) 1.61–8.38] and IL-1RA (HR = 3.28; 95% CI 1.47–7.29) levels remained independent predictors of inferior EFS. TCL cell lines secreted high levels of sIL-2Rα and expressed the IL-2Rα surface receptor.

Conclusions

This report describes the cytokines relevant to prognosis in patients with untreated TCL and provides the rationale to include serum IL-1RA and sIL-2Rα as biomarkers in future trials. Inhibition of these cytokines may also be of therapeutic benefit.

introduction

T-cell lymphoma (TCL) comprises ∼10% of all non-Hodgkin lymphoma (NHL) in the United States and up to 20% of cases in Asia [1]. TCL are currently classified by WHO criteria [2, 3], with the most common subtypes being peripheral TCL-not otherwise specified (PTCL-NOS), angioimmunoblastic TCL (AITL), anaplastic large cell (ALCL), and the predominant subsets of cutaneous TCL (CTCL), Sézary syndrome (SS) and mycosis fungoides (MF). The long-term outcome of the non-CTCL groups remains poor with ∼30% of patients being cured [4]. Induction therapy is typically with cyclophosphamide, doxorubicin, vincristine, and prednisone (CHOP) or CHOP plus etoposide (CHOPE), and eligible patients typically receive autologous stem cell transplant in the first remission [5]. There have been three new agents recently approved by the US Federal Drug Administration—pralatrexate, belinostat, and romidepsin—all for relapsed TCL [6], and these are now being moved up to first line in clinical trials.

The variable remission and cure rates and these new therapeutic approaches provide a need to develop new biologic and genetic biomarkers to predict prognosis and direct therapy. Recent studies of tumor cell mutations in TCL have identified important mutations in p53-related genes (TP63) that predict inferior survival in PTCL [7] and improved survival in anaplastic lymphoma kinase (ALK)-negative TCL cases with rearrangement of DUSP22 [8]. The results of gene expression profiling have also deepened our understanding of the tumor biology of TCL [9]. Recent reviews have discussed the importance of genetic abnormalities [8, 10, 11] and the nonmalignant cells in the microenvironment of TCL [12]. Distinctly lacking in previous studies are those focusing on the secretion of cytokines into the blood of these patients. Patients with aggressive TCL often present with B-symptoms suggesting serum cytokine elevations. Cytokine elevations have been found in studies of B-cell NHL [13–16] and Hodgkin lymphoma (HL) [17]; however, comprehensive data on multiple cytokines are lacking in TCL. This study of pretreatment serum of 68 patients with newly diagnosed T-cell neoplasms demonstrates that multiple cytokines are indeed dysregulated in this type of NHL, are prognostic of inferior outcome, and are biologically relevant.

materials and methods

patient selection

The studies from which these samples were derived were reviewed and approved by the Human Subjects Institutional Review Board at Mayo Clinic and the University of Iowa. All participants provided written informed consent for the use of research blood and outcome data. Blood samples were collected and processed in the same manner and cryopreserved without thawing before this analysis. The samples for this study were from patients with confirmed untreated T-cell neoplasms who enrolled in the University of Iowa/Mayo Clinic Specialized Program of Research Excellence Molecular Epidemiology Resource [18]. Clinical management was per standard of care with follow-up for relapse, retreatment, and death. Control serum (n = 14) samples were obtained from an ongoing case–control study enrolling over the same timeframe with similar sample processing [19].

human TCL cell lines

Six human TCL cell lines were utilized for these studies. SUDHL1 and SR786 cell lines were purchased from Leibniz Institute DSMZ—German Collection of Microorganisms and Cell Cultures; Karpas 299 was purchased from American Type Culture Collection. SeAx, HuT 78, and MyLa were generous gifts from Dr Robert Gniadecki (University of Copenhagen). All cell lines were grown in RPMI 1640 supplemented with 10% fetal bovine serum. JAK1/JAK2 inhibitors, INCB018424 (ruxolitinib) and CYT387 (momelotinib), were purchased from ChemieTek (Indianapolis, IN) and JAK3 inhibitor, WHI-P154, from Santa Cruz Biotechnology.

cytokine analysis of patient serum and TCL cell line supernatants

Serum cytokine levels in patients and healthy controls and supernatants from TCL cell lines were measured using a standard 30-cytrokine kit (Invitrogen, Camarillo, CA) as previously described [15]. Data were acquired using the STarStation software (Applied Cytometry, Dinnington, Sheffield, UK).

flow cytometric analysis

For the detection of surface receptors for IL-2Rα, IL-1RA, and monokine induced by gamma interferon (MIG) (CXCL9), 0.5 × 10⁶ cells were washed and incubated with fluorochrome-conjugated CD25, IL1-R1, and CD183 (CXCR3) antibodies or isotype controls (BD Biosciences) for 30 min at 4°C, and then run on a FACSCalibur with data analysis carried out using the FlowJo software.

statistical analysis

Event-free survival (EFS) was defined as time from diagnosis until relapse, retreatment, or death due to any cause; overall survival (OS) was defined as time from diagnosis until death due to any cause. Associations between cytokines and clinical characteristics were assessed using Wilcoxon rank-sum tests, Spearman correlation coefficients, and Fisher's exact tests, as appropriate. Associations between cytokines and time-to-event end points were carried out using Cox proportional hazards models and Kaplan–Meier curves. The primary test for this study was the univariate association of cytokines between cases and controls (cytokine deregulation) and the univariate association of cytokines and EFS/OS. Cytokines used for analysis were log2-transformed and median-adjusted to correct for any plate effects with the level of significance set at P < 0.05. Cytokines significantly different between cases and controls were assessed for prognostic significance. The cytokine values for TCL cases were dichotomized into values above or below the median level for the purposes of running models for EFS and OS. Cytokines identified as significantly associated with EFS were further examined by adjusting for International Prognostic Index (IPI), examining association with OS, as appropriate. Multivariable stepwise selection models were run with the significant cytokines alone, and additional models were run forcing IPI into the model selection.

results

patient characteristics

Sixty-eight patients with new, untreated T-cell neoplasms were identified for this study between 2002 and 2010. The neoplasms were classified [3] as follows: 23 patients (34%) with PTCL-NOS; 11 (16%) AITL; 6 (9%) ALCL (3 ALK+ and 3 ALK−); 5 (7%) T-cell large granular lymphocytic leukemia; 4 (6%) extranodal NK/T-cell lymphoma nasal type; 8 (12%) CTCL (4 MF, 3 SS, and 1 CTCL NOS); 2 (3%) subcutaneous panniculitis-like TCL; 8 (13%) other TCL—1 case each of CD8+ non-large granular lymphocytic leukemia and T-cell prolymphocytic leukemia; three cases of CD4+ lymphoproliferative disorder—NOS; two cases of enteropathy-associated TCL; and two cases of T lymphoblastic leukemia/lymphoma. The median age was 58 years (range, 21–85); 59% (40/68) were male; 72% (46/68) had stages III/IV; 15% (10/68) had performance status ≥2; 46% (28/68) had an elevated LDH; and 21% (14/68) had B-symptoms. The IPI scores were 43% (29/68) 0–1, 22% (15/68) 2, and 35% (24/68) 3–5. Initial treatment was with chemotherapy in 76% (52/68) and in 51% (35/68) receiving an anthracycline-based regimen. Three (4%) patients received only radiation therapy; 16% (11/68) were observed; and in 2 the regimen was not available.

cytokine levels in all patients and the PTCL-NOS and AITL subtypes

We analyzed the serum from all 68 patients and 14 controls, and found seven serum cytokines—epidermal growth factor (EGF), IL-6, IL-12, interferon gamma-induced protein (IP)-10, soluble interleukin (sIL)-2Rα, MIG, IL-1RA—to be significantly elevated in patients versus controls (all P < 0.05; Table 1A and Figure 1). In the PTCL-NOS subset (n = 23), only four cytokines (IP-10, sIL-2Rα, MIG, and IL-8) were statistically elevated. In contrast, in the AITL patients (n = 11), all seven cytokines, except IP-10, were significantly elevated (Table 1A and Figure 1).

Table 1.

Serum cytokine levels in (A) patients with PTCL compared with control and (B) cytokine secretion by malignant human T-cell lymphoma cell lines

(A)
Cytokine	Controls (n = 14)	All TC neoplasms (n = 68)		PTCL-NOS (n = 23)		AITL (n = 11)
	Median	Median	P-value	Median	P-value	Median	P-value
EGF	58.2	140.2	0.008	94.8	0.085	161.2	0.025
HGF	524.9	672.0	0.156	667.2	0.125	771.3	0.067
IFN-α	119.8	168.5	0.097	161.2	0.145	204.2	0.218
IL-6	5.1	12.5	0.028	17.5	0.077	21.0	0.015
IL-12	288.6	466.3	0.015	373.1	0.072	841.2	0.001
Eotaxin	124.7	124.9	0.966	135.6	0.963	114.2	0.529
MIP-1α	36.8	44.3	0.07	46.4	0.082	57.1	0.139
MIP-1β	126.0	166.3	0.321	156.4	0.605	243.4	0.262
MCP-1	714.2	823.3	0.427	920.3	0.309	874.5	0.366
IL-1RA	217.5	363.7	0.043	331.2	0.178	517.9	0.007
IP-10	21	30.8	0.021	34.4	0.007	27.2	0.528
IL-2Ra	352.6	813.6	0.019	1012.7	0.047	2594.4	0.017
MIG	40.6	99.8	0.002	112.7	0.004	389.6	<0.001
IL-8	50.4	64.2	0.055	77.7	0.02	71.6	0.147
(B)
TCL cell lines	EGF (pg/ml)	IL-6 (pg/ml)	IL-12 (pg/ml)	IL-1RA (pg/ml)	IP-10 (pg/ml)	sIL-2Rα (pg/ml)	MIG (pg/ml)
Karpas299	8	15	35	62	3	4823	0
SUDHL1	6	5	7	79	0	8032	0
SR786	6	23	34	361	1	5471	0
Hut 78	6	14	43	134	22	1963	0
SeAX	7	34	141	210	6509	838	0
MyLa	7	374	118	80	487	5309	0

A: Values in bold are statistically significant at P < 0.05.

B: >25 pg/ml was considered positive secretion.

TCL, T cell lymphoma; PTCL-NOS, peripheral T-cell lymphoma-not otherwise specified; AITL TCL, angioimmunoblastic T-cell lymphoma.

Figure 1.

Dot-plot of serum levels of the cytokines found to be significantly (P < 0.05) elevated in all patients with T-cell neoplasms as well as the PTCL-NOS and AITL TCL compared with controls. PTCL-NOS, peripheral T-cell lymphoma-not otherwise specified; AITL TCL, angioimmunoblastic T-cell lymphoma.

To determine whether the cytokine elevations in TCL patient serum could be derived from tumor cell secretion (autocrine), supernatants from six human TCL lines were studied (Table 1B). Karpas 299, SUDHL1, SR786, HuT 78, SeAx, and MyLa secreted high levels (cutoff of >25 pg/ml) of sIL-2Rα (range, 838–8032 pg/ml) and modest levels of IL-1RA (range, 62–361 pg/ml). None of the cell lines secreted MIG or EGF. IL-6 and IP-10 were secreted by SeAx and MyLa; IL-12 was secreted by Hut 78, SeAx, and MyLa (Table 1B). Thus, autocrine secretion is a potential source of the elevated serum cytokine levels of IL-1RA, sIL-2Rα, IL-6, and IL-12 in TCL patients.

We also examined the association of cytokines with standard clinical characteristics (supplementary Table S1, available at Annals of Oncology online). Using the median value as a cutoff, we found serum IL-6, IL-1RA, and sIL-2Rα levels to be significantly higher in the 14 patients with B-symptoms (P < 0.01, P < 0.02, and P < 0.03, respectively) than the values for patients without B-symptoms (supplementary Table S1, available at Annals of Oncology online).

cytokine elevations as predictors of survival

Associations between cytokine levels and clinical outcome were examined using a dichotomous variable (above versus below the median) for each cytokine. Elevated levels of sIL-2Rα (HR = 3.95), MIG (HR = 2.84), IL-6 (HR = 2.41), IL-12 (HR = 1.91), IL-1RA (HR = 3.16), and IP-10 (HR = 2.22) were strongly associated with inferior EFS (supplementary Table S2A, available at Annals of Oncology online; Figure 2A–C). Of these, MIG (HR = 2.24), IL-1RA (HR = 3.28), and sIL-2Rα (HR = 3.68) were also strongly associated with an inferior OS (supplementary Table S2B, available at Annals of Oncology online; Figure 2D–F). Elevated levels of sIL-2Rα were significantly associated with inferior EFS and OS in the PTCL-NOS subtype (Figure 3A and C). In the AITL subtype, elevated sIL-2Rα was significantly associated with an inferior EFS but not with OS (Figure 3B and D).

Figure 2.

The clinical outcome of untreated patients with T-cell neoplasms (n = 68) by IL-1RA, sIL-2Rα, and MIG levels. Event-free survival (A–C) and overall survival (D and E) was analyzed by IL-1RA, sIL-2Rα, and MIG levels.

Figure 3.

Clinical correlation of PTCL-NOS (n = 23) and AITL (n = 11) TCL patients by sIL-2Rα. Event-free survival and overall survival were analyzed by sIL-2Rα level in PTCL-NOS (A and C) and AITL (B and D) TCL patients. PTCL-NOS, peripheral T-cell lymphoma-not otherwise specified; AITL TCL, angioimmunoblastic T-cell lymphoma.

The correlations between the seven cytokines found elevated in PTCL patients are summarized in Table 2A with R-values >0.6 considered to be notable. In a multivariate analysis of cytokines and clinical factors, only IL-1RA and sIL-2Rα were independent predictors of inferior EFS (Table 2B). When the IPI score was included in the multivariate analysis, only IL-1RA remained independent (Table 2C).

Table 2.

(A) Correlations of cytokine levels with each other; (B) Multivariate model for EFS resulting from including all seven cytokines in a stepwise selection procedure and, (C) multivariable model for EFS resulting from forcing the International Prognostic Index score into the model and using stepwise selection for the seven cytokines

(A)
Cytokine	EGF	IL-6	IL-12	IL-1RA	IP-10	IL-2Rα	MIG
EGF	1.00
IL-6	0.41	1.00
IL-12	0.38	0.49	1.00
IL-1RA	0.58	0.67	0.72	1.00
IP-10	0.02	0.19	0.38	0.30	1.00
IL-2Rα	0.27	0.53	0.49	0.49	0.38	1.00
MIG	0.29	0.47	0.69	0.59	0.48	0.68	1.00
(B)
Parameter	Hazard ratio	95% Hazard ratio confidence limits	P-value
IL-2Rα >median	2.99	(1.43, 6.25)	0.0037
IL-1RA >median	2.16	(1.07, 4.37)	0.032
(C)
Parameter	Hazard ratio	95% Hazard ratio confidence limits	P-value
IPI	1.661	(1.258, 2.191)	0.0003
IL-1RA above median	2.679	(1.363, 5.264)	0.0043

Correlation (R > 0.6) values were considered significantly.

EFS, event-free survival; IPI, International Prognostic Index.

detection of cell surface receptors for IL-2Rα, IL-1RA, and MIG in TCL cell lines

We next analyzed the expression of cell surface receptors used by sIL-2Rα, IL-1RA, and MIG on malignant TCL cell lines (n = 6). All TCL cell lines expressed high levels of the surface receptor for IL-2Rα and low to very low levels of the MIG receptor CXCR3; and none of the cell lines expressed the IL-1RA receptor IL-1R1 (supplementary Table S3A–C, available at Annals of Oncology online).

effect of JAK pathway inhibition on sIL-2Rα secretion in TCL cell lines

We used pharmacological inhibitors of JAK1/JAK2 or JAK3 and determined their effect on autocrine secretion of sIL-2Rα in the SUDHL1 and Karpas 299 TCL cell lines. We measured sIL-2Rα by ELISA in culture supernatants treated with the JAK1/JAK2 inhibitors, ruxolitinib and momelotinib, or the JAK3 inhibitor, WHI-P154 (WHI-P), for 24 h. Ruxolitinib, momelotinib, and WHI-P partially inhibited autocrine sIL-2Rα production in Karpas 299 and SUDHL1 cells at sublethal doses when compared with cells grown in medium alone (supplementary Table S4A and B, available at Annals of Oncology online).

discussion

Blood cytokines have been demonstrated to be abnormally elevated compared with control in a number of different types of NHL and HL [13–15, 17, 20], but these studies have previously not been carried out in patients with T-cell neoplasms. In this study of 68 newly diagnosed patients, we aimed to understand which of the 30 cytokine panels of cytokines were elevated and associated with prognosis. We found seven cytokines—EGF, IL-6, IL-12, IP-10, sIL-2Rα, MIG, and IL-1RA—to be elevated above control (all P < 0.05), with all but EGF predictive of inferior EFS. In addition, IL-1RA, sIL-2Rα, and MIG predicted an inferior OS. In a multivariate analysis that included clinical factors, IL-1RA remained an independent predictor of inferior EFS. In addition, cell line studies demonstrated that autocrine secretion is a potential cause of the elevated cytokine levels for all the cytokines, except EGF and MIG. These cells can also express the cognate receptors for sIL-2Rα and MIG and signal through the JAK/STAT pathway.

Although there has not been a comprehensive analysis like this report on pretreatment serum cytokine levels in newly diagnosed T-cell neoplasms, there have been studies on cytokine expression in malignant T-cells. Galectin-1 has been shown to be derived from CTCL cells and can induce a Th2 cellular cytokine profile [21]. Foss et al. [22] studied lymph nodes from 11 patients with AITL for IL-6 and IL-1β expression by in situ hybridization and found expression of both cytokines in all cases; serum levels were not reported. A case of AITL was found to have elevated IL-6 that decreased with effective chemotherapy [23]. Patients with extranodal NK/TCL, nasal type were found to have elevated levels of sIL-2Rα, IL-6, and IL-10 [24]. These results are similar to our observations with regard to sIL-2Rα and IL-6, but not to IL-10. Wang et al. [25] demonstrated the importance of IL-10 on the microenvironment through promotion of alternative macrophage polarization. It has long been recognized that sIL-2R levels predicted inferior outcomes in CTCL [26, 27] and studies of sIL-2R levels in TCL have also been consistent with our findings. sIL-2R elevations were predictive of an inferior outcome in a study of 36 AITL patients [28] and in a separate study of 30 patients with PTCL-NOS [29]. In a recent review of cytokine signaling in CTCL [30], the importance of IL-2-related cytokines and signaling through the JAK–STAT pathway was noted. Indeed, all of this work has provided the rationale for ongoing trials of JAK/STAT inhibitors as therapy for relapsed TCL.

These serum cytokine abnormalities are certainly not unique to TCL, but have been described in other types of B-cell NHL [13, 15], Waldenstrom's macroglobulinemia [16], and HL [17]. Specifically, sIL-2R levels have been shown to predict the outcome in patients undergoing autologous stem cell transplant for B-cell NHL [31]. Recent studies of pretreatment of sIL-2R in patients with untreated diffuse large B-cell lymphoma treated with modern RCHOP-based regimens demonstrated that high levels of sIL-2R were predictive of an inferior OS [32–34].

The strengths of our comprehensive study of serum cytokine abnormalities in untreated patients are the use of a multiplex assay with excellent quality control and the inclusion of a relatively large (for PTCL) group of untreated, well-characterized patients with mature follow-up. A shortcoming of our study was the small number of patients within each of the T-cell neoplasm subtypes. It is important to build serum studies of the cytokines; we found to be elevated into prospective randomized clinical trials where patients are treated in a uniform manner to confirm our results and further integrate cytokine levels into prognostic scoring systems. In addition, knowledge of the cytokine profile will enable trials to learn if adding a specific anticytokine therapy to standard chemotherapy will augment the response to treatment and decrease the symptom burden of these TCL patients.

funding

This work was supported in part by the National Cancer Institute at the National Institutes of Health (P50 CA97274 to MG, R01 CA127433 to TEW, and R01 CA92153 to JRC] and the Predolin Foundation.

disclosure

The authors have declared no conflicts of interest.