Cells, cytokines, chemokines, and cancer stress: A biobehavioral study of patients with chronic lymphocytic leukemia

Abstract

Purpose

Chronic lymphocytic leukemia (CLL) is the most prevalent adult leukemia, with profound disease-related cellular, humoral, and innate immune suppression. The relationship between stress and disease-specific negative prognostic cellular, cytokine, and chemokine markers in CLL patients is studied.

Patients and methods

A single-group, observational design was used. Relapsed/refractory CLL patients (N=96) entering a Phase II trial of an experimental therapy (ibrutinib) were studied. Before the first dose, a validated self-report measure of stress (Impact of Event Scale) was completed and blood was drawn for absolute lymphocyte counts (ALCs) and cytokine and chemokine ELISA assays. Multiple linear regression models tested stress as a concurrent predictor of ALCs, cytokines [tumor necrosis factor alpha (TNFα), a proliferation-inducing ligand (APRIL), B-cell activating factor (BAFF), Interleukins (IL)-6, 10, 16, VEGF] and the chemokine (C-C motif) ligand 3 (CCL3).

Results

Controlling for relevant demographic variables, comorbidities, CLL genetic risk (del17p), and correlates of inflammation, stress predicted higher ALCs (p<.05), and higher levels of TNFα (p<.05), IL-16 (p<.01), and CCL3 (p<.05). Stress was not associated with APRIL, BAFF, IL-6, IL-10, or VEGF.

Conclusions

Novel biobehavioral data for relapsed/refractory patients show stress is related to heightened levels of cellular, cytokine, and chemokine markers that have been previously shown to be associated with progressive disease in CLL. Results indicate that stress is related to immune and inflammatory processes that contribute to cancer cell proliferation and survival. These data provide a first look into these processes.

Introduction

Patients with chronic lymphocytic leukemia (CLL), which accounts for about one-third of adult leukemia in the United States,¹ may be vulnerable to psychological stress as chemo-immunotherapy is not curative. At diagnosis, patients may be asymptomatic excepting for high lymphocyte counts (> 5,000/mm³). Closer inspection finds small, mature appearing lymphocytes with progressive B lymphocyte accumulation in the blood stream, making CLL a disease in which a key biologic system—immunity—is compromised². Treatments may bring remission but relapses are likely to occur with successive treatments often having greater toxicity and reduced effectiveness³. Thus, the relapsing/refractory nature of CLL place patients at risk of stress and perhaps, stress-induced immune impairments.

Previous work has provided empirical support for how stress impacts immunity to initiate or facilitate an ongoing pathological process⁴. The immune system is viewed as acting like a sensory organ, informing the brain of immune challenges, and further, immune activation may be interpreted by the central nervous system as a stressor⁵. Cytokines may be a part of the regulatory loop⁶. Indeed, “sickness behaviors,” such as heightened negative emotions are seen as the products of pro-inflammatory cytokines⁵. Conversely, data from psychological intervention trials show that when patients’ stress lowers, inflammation is reduced and health and cancer outcomes are improved^7-9.

Examining how stress interacts with the immune system in CLL is critical as CLL cells develop in the same environment as normal immune cells¹⁰. Specifically, malignant cells interact with stromal cells as well as T cells in the lymphoid tissue, which collectively create a microenvironment conducive to the survival and proliferation of malignant cells¹¹. CLL cells can be stimulated directly by elements of the microenvironment or through cytokines and chemokines¹². Additionally, lymphatic tissues are the principle site of B-cell receptor (BCR) activation for both normal and malignant B-cells¹³. In response to BCR activation, CLL cells secrete chemokines, a type of cytokine that directs white blood cells to infected tissues to help create a microenvironment conducive to disease progression.¹¹ Additional research has indicated that BCR signaling results in malignant B cell proliferation and may impede the immune system’s ability to destroy malignant cells.¹³ Considering these literatures, we studied how stress might covary with absolute lymphocyte count (ALC), the hallmark CLL indicator¹⁴, and cytokine and chemokine factors important to the interaction of CLL cells and the microenvironment.

Tumor necrosis factor alpha (TNFα) is a cytokine involved in systemic inflammation and is produced by activated macrophages, although other cell types such as CD4+ lymphocytes and natural killer (NK) cells also produce TNFα. TNFα is increased in CLL patients compared to healthy individuals¹⁵, and is correlated with higher white cell counts at diagnosis and shorter progression-free survival^{16, 17}. In vitro data show CLL neoplastic lymphocytes release TNFα spontaneously¹⁵ and further, exposure to TNFα increases the proliferation and viability of leukemic lymphocytes¹⁸.

A proliferation-inducing ligand (APRIL) and the B-cell-activating factor of the TNF family (BAFF) are cytokines that aid in the survival and proliferation of CLL cells^{11, 19}. In vitro data demonstrate APRIL and BAFF aid CLL cell survival through upregulation of CD40 ligands²⁰. Clinical data has shown patients with lower APRIL/BAFF expression in CLL cells had longer survival times compared to those with greater expression of APRIL/BAFF in CLL cells²¹.

Furthermore, inflammatory cytokines, such as Interleukins-6 (IL-6) and -10 (IL-10), are involved in promoting angiogenesis in CLL²². Researchers have demonstrated that B cells and CLL cells produce IL-10 and suggest that IL-10 creates an immunosuppressive niche that allows CLL cells to proliferate^{23, 24}. Elevated levels of IL-6 and IL-10 have been observed in CLL patients compared to controls and are associated with shorter survival times¹⁶. A meta-analysis has shown IL-10 expression to predict worse disease-free survival among those with hematological malignancies²⁵. Also relevant is IL-16, a pro-inflammatory cytokine that contributes to the pathogenesis of CLL through the regulation of CD4+ cell recruitment and activation at sites of inflammation²⁶.

CLL cells produce high levels of angiogenic factors that support cell accumulation and proliferation²⁷. Among the most studied angiogenic factors, vascular endothelial growth factor (VEGF) is secreted by malignant B-cells²⁸. Research has demonstrated that CLL patients with higher levels of VEGF have 3 times the increased risk of progressive disease²⁹. Additionally, malignant B cells secrete chemokine (C-C motif) ligand 3 (CCL3) which fosters the interaction between CLL cells and the leukemia microenvironment³⁰ and has been associated with advanced stage disease³¹.

Patients with relapsed/refractory CLL in a Phase II (open label) trial of a new targeted agent, ibrutinib, were studied. Past research has shown relapsed/refractory patients to have lower quality of life compared to treatment-naïve patients^{32, 33} and may be vulnerable to stress. Previous research has shown a measure of cancer-specific stress³⁴ can predict lower NK cell cytotoxicity, diminished response of NK cells to recombinant interferon gamma, and decreased T-cell blastogenesis in breast cancer patients awaiting chemotherapy³⁵. Therefore, in a CLL sample, heightened stress may be associated with heightened CLL-specific cellular (ALC), cytokine (TNFα, APRIL, BAFF, IL-6, IL-10, IL-16, and VEGF), and chemokine (CCL3) markers while controlling for disease-relevant and behavioral correlates of inflammation.

Methods

Design

A single group, observational design was used. Patients with relapsed/refractory CLL were screened and accrued to a phase II study of ibrutinib at the Ohio State University (OSU) Medical Center.

Participants and procedures

The Institutional Review Board granted ethical approval for the study with an accrual goal of 154 patients. Informed written consent was obtained from 152 patients with relapsed/refractory CLL with accrual occurring from May 2012 to July 2014. Descriptive data are displayed in Table 1. In the clinic on day 1 of initial treatment, patients independently completed a self-report measure of stress (Impact of Event; IES) and provided descriptive information and blood was drawn.

Table 1.

Sample characteristics: Demographic, general health, CLL risk, behavioral correlates of inflammation, and stress scores for trial patients (N=152) and subset with immune data (N=96)

Characteristic	Trial patients (n=152)		Patients with cytokine/chemokine data (n=96)
Demographic	M/SD	N (%)	M/SD	N (%)
Age (in years), M (SD)	64.1 (10.80)		64.2 (10.40)
Gender (Male)		107 (71%)		64 (67%)
Married/Partnered (Yes)		131 (86%)		83 (87%)
Distance from cancer center (in miles)	341.44 (438.05)		331.55 (422.58)
Race
Caucasian		147 (97%)		95 (99%)
African-American		5 (3%)		1 (1%)
Education
High School/Technical School or Below		44 (29%)		18 (19%)
Some College/College Graduate		58 (38%)		44 (46%)
Some Graduate School/Graduate Degree		46 (30%)		32 (33%)
Missing		4 (3%)		2 (2%)
Household income (K)
≤ 50		41 (27%)		29 (30%)
51-100		38 (25%)		23 (24%)
>100		46 (30%)		43 (45%)
Prefers Not to Answer		24 (16%)		10 (10%)
Missing		3 (2%)		2 (2%)
General Health
Charlson Comorbidity Index (unadjusted)		2.5 (1.0)	2.49 (0.82)
CLL Risk
Number of prior therapies	3.5 (2.6)		3.27 (2.43)
Del17p		78 (51%)		54 (56%)
Behavioral Correlates of Inflammation
Body Mass Index	26.5 (4.8)		26.13 (5.10)
Smoking Status (Current Smoker)		8 (5%)		5 (5%)
Stress (IES)	11.61 (10.55)		13.55 (10.62)

Abbreviations: Del17p = presence of deletion of 17p; IES = Impact of Event Scale

Funding was sought to conduct immune assays. Of the 152 patients, 144 patients had adequate samples. Patients within sex were rank ordered from highest to lowest on his/her stress score (IES) with every other subject selected for the first batch of cytokine/chemokine assays (n=72). Later funding provided for a second batch where 24 patients with the highest stress scores were selected. A total of 96 patient samples were available for the cytokine/chemokine assays.

Measures

Stress

The Impact of Event Scale-Revised (IES)^{34, 36} assesses stress reactions in the form of intrusive thoughts (8 items), avoidant thoughts or behaviors (8 items), and hyperarousal (6 items) with items tailored to read “cancer stress” or “cancer”. Previous research has found the IES to covary with immunity in breast cancer patients³⁵. Patients rate the frequency of feelings or events in the past week on a scale ranging from 0 (not at all) to 4 (extremely) and all items are summed. Possible scores range from 0 to 88 with higher scores indicative of greater stress. Coefficient alpha reliability was 0.89.

Immune

Absolute Lymphocyte Count (ALC)

Nursing staff collected 6-mL potassium EDTA [lavender-top vacutainer] peripheral blood. Complete blood cell counts with ALC were quantified by the hospital laboratories.

Cytokine and chemokine

Plasma samples were centrifuged and aliquoted into 4 samples and frozen at −80C. Later, samples were thawed, batched by participant, and spun to remove debris. Assessment of plasma levels of cytokines (TNFα, APRIL, BAFF, IL-6, IL-10, and IL-16, VEGF) and chemokine CCL3, were quantified using individual enzyme-linked immunosorbent assays (ELISA) in triplicate per manufacturer’s specifications for each of the cytokines and chemokines (R&D Systems, Minneapolis, MN). Detection limits were 1.6 pg/mL for TNFα, 7.1 pg/mL for APRIL, 2.7 ng/mL for BAFF, 0.7 pg/mL for IL-6, 3.9 pg/mL for IL-10, 6.2 pg/mL for IL-16, 9.0 pg/mL for VEGF, and 10.0 pg/mL for CCL3. The inter- and intra-assay coefficients of variation for all cytokines tested were 6.26% and 14.29%, respectively.

Covariates

Older age, male sex, presence of multiple comorbidities, number of prior treatments, and presence of del17p were included as covariates as these factors are predictive of poorer outcomes in CLL^37-41. Four areas were considered. 1) Demographic: age and gender. 2) Comorbidities: The Charlson Comorbidity Index (CCI)⁴² is a measure of risk of death from current diseases or conditions. It has 19 items (e.g. liver disease, diabetes mellitus), each weighted from 1 to 6 based on severity and mortality risk, which are summed for a total score. All patients received 2 points for having cancer; a score > 2 indicates the presence of comorbid conditions. 3) CLL severity and risk: Number of prior therapies and genetic risk (presence of del17p). 4) Behavioral correlates of inflammation⁴³ were also considered: body mass index (BMI) and smoking status.

Analytic Plan

Preliminary analyses contrast patients in the cytokine/chemokine assay subgroup (n=96) and those not (n=56) on descriptive characteristics, covariates, and stress (IES), using independent-sample t-tests for continuous variables and chi-square tests for categorical variables. For cytokine and chemokine values below detectable limits, the MICE package⁴⁴ for statistical software R was used to generate random values between 0 and the minimum detectable level to increase variance. Multiple imputation has been previously used to impute cytokine and chemokine values⁴⁵. Spearman correlations between each covariate and outcome variable were examined and only covariates associated (p < 0.10) with an outcome were included in analyses.

Analyses used multiple linear regressions (MLRs) to test the concurrent relationship between stress and ALC, cytokines, and chemokine responses following the entry of relevant covariates. For each outcome, step 1 included only the associated covariates as predictors while step 2 included associated covariates with the addition of IES as predictors of cellular, cytokine, and chemokine responses. The increment in the squared multiple correlation (R²) from step 1 to step 2 provided variance attributable to stress beyond the covariates. In addition, standardized regression betas (β) in step 2 indicated the magnitude and direction of the influence of stress on outcomes. General assumptions of linear regression including multicollinearity, variance of errors, and normality of error distributions were examined. If positively skewed, variables were log transformed.

Results

Preliminary

Minimum detectable levels were imputed using the MICE package for TNFα (5 patients), APRIL (6 patients), and IL-10 (2 patients). Patients with (n=96) and without (n=56) cytokine/chemokine data did not differ (ps ≥.18) on any demographic, general health, CLL risk, or behavioral correlates of inflammation variables. Patients with cytokine/chemokine data reported greater IES scores, [t(150) = 3.04, p<0.01].

Intercorrelations between covariates and ALC, cytokine, and chemokine markers were inspected. No outcome covaried with current smoking status. For the remainder, the following p<.10 associations were used in MLR analyses: ALC (age: ρ=.232), TNFα (prior therapies: ρ=.224, gender: ρ=−.190), APRIL (prior therapies: ρ=.218, gender: ρ=−.194), BAFF (CCI: ρ=−.203,), IL-6 (prior therapies: ρ=.338, gender: ρ=.192, del17p: ρ=−.177), IL-16 (gender: ρ=−.269, prior therapies: ρ=.229, BMI: ρ=−.200, del17p: ρ=.178),VEGF (prior therapies: ρ=−.271), CCL3 (gender: ρ=−.216, BMI: ρ=−.199). IL-10 was not associated with any covariates.

Primary

Results of MLRs are displayed in Table 2. Results indicated stress was a significant predictor of higher ALC (β=.208, p=.037) (Figure 1). Additionally, the R² change between steps 1 and 2 indicated that stress accounted for 4.3% of the variance in ALC, controlling for age. Models examining cytokines and chemokines showed stress to be a significant predictor of TNFα (β=.251, p=.016) and IL-16 (β=.413, p<.001), and CCL3 (β=.230, p=.027) such that higher levels of stress were associated with increased levels of these three outcomes (Figure 1). The R² change indicated stress accounted for 6.3%, 17.0%, and 5.2% of the variance in TNFα, IL-16, and CCL3, respectively. Analyses with stress and IL-6, IL-10, and VEGF were not significant (ps>.17).

Table 2.

Results of multiple linear regression analyses with standardized betas. Step 1 enters covariates associated (p < 0.10) with outcomes, step 2 enters cancer-specific stress to predict log-transformed cellular, cytokine, and chemokine markers.

	Cellular	Cytokine							Chemokine
Step 1	ALC (n=96)	TNFα (n=88)	APRIL (n=94)	BAFF (n=95)	IL-6 (n=95)	IL-10 (n=74)	IL16 (n=86)	VEGF (n=96)	CCL3 (n=92)
Age	.253*	–	–	–	–	–	–	–	–
Gender	–	−.221*	−.168	−.152	.178	–	–	–	−.125
Number of prior therapies	–	.151	.267**	.264*	.321**	–	–	−.262*	–
Del17p	–	–	–	–	−.078	–	.150	–	–
CCI	–	–	–	−.016		–	–	–
BMI	–	–	–	−.041		–	–	–	−.117
Adjusted R2	.054	.073	.081	.066	.121	–	.011	.059	.019
Step 2
Age	.262*	–	–	–		–	–	–	–
Gender	–	−.218*	−.171	−.160	.177	–	–	–	−.107
Number of prior therapies	–	.167	.263*	.256*	.320**	–	–	−.258*	–
Del17p	–	–	–	–	−.078	–	.173	–	–
CCI	–		–	−.018	–	–	–	–	.
BMI	–	–	–	−.034	–	–	–	–	−.126
Stress (IES)	.208*	.251*	−.049	−.119	−.008	.215	.413**	.051	.230*
Adjusted R2	.088	.105	.074	.071	.111	.033	.173	.052	.062
R2 Change	.043*	.063*	.002	.014	.000	–	.170**	.003	.052*

Note.

^**= p <.01,

^*= p <.05,

dashes indicate covariates that were not correlated (p>.01) with each outcome variable

Abbreviations: Del17p = presence of deletion of 17p; CCI = Charleson Comorbidity Index; BMI = Body Mass Index; IES = Impact of Event Scale

Figure 1.

Scatterplot of cancer-specific stress scores from CLL patients (N=92) and A) absolute lymphocyte counts B) TNFα C) IL-16, D) CCL3 at treatment initiation with line of best fit, in natural log.

Discussion

Novel data from patients with relapsed/refractory CLL show increased stress is associated with multiple poor prognosis CLL biomarkers. Heightened stress covaried with higher ALC, a salient marker of CLL burden, along with TNFα and IL-16. In the biobehavioral literature there are few studies of stress and chemokines^{46, 47} and none of CCL3, which also covaried with stress. Effects were observed with outcome-specific controls for age, gender, BMI, comorbid illnesses, prior treatments, and/or CLL genetic risk.

Meta analyses of studies with adults experiencing general stressors, show heightened stress to covary with increases in lymphocyte numbers^{48, 49}, whereas meta-analyses of stress from acute stressors show a lowering of lymphocyte numbers⁵⁰. Both human lymphocyte data and leukemic animal paradigms⁵¹ illustrate the adverse effects of chronic stress, an effect that may be especially pernicious in CLL.

The association between stress and TNFα is particularly important in CLL as malignant cells release TNFα spontaneously¹⁵ and TNFα increases their proliferation and viability⁵². The TNFα finding adds to the biobehavioral cancer literature, as most studies have found depression rather than stress to be related to higher levels of TNFα^53-55. Stress and depression can co-occur in patients; future research may examine the independent and synergistic effects of depression and stress on inflammatory markers such as TNFα⁵⁶.

Higher levels of stress were also associated with higher levels of IL-16. It has been suggested that IL-16 may mediate communications between B cells and T cells within lymph node follicles^{57, 58} and further, IL-16 may suppresses effector T-cell function^59-61. To our knowledge, the only other stress/IL-16 data come from a study of a naturalistic stressor (students’ academic examinations) in which IL-16 was significantly elevated pre-examination in contrast to post examination⁶². Interestingly, no covariates were needed for the IL-16 analyses and stress accounted for substantial variance (17%) in IL-16 prediction. Future research might explore the pathway of heightened stress, increased IL-16, and suppression of effector T-cell function which may be relevant to autoimmune diseases⁶³ or other cancers⁶⁴ in which Il-16 plays an important role.

While the function of CCL3 is not yet fully clarified³¹, it has been hypothesized that increased CCL3 secretion may induce trafficking and homing of T cells to CLL cells in the tissue microenvironments⁶⁵. Moreover, by attracting immune cells for interaction with CLL cells and their microenvironment, CCL3 creates a circumstance in which CLL cells interact with T cells providing survival and proliferation signals⁶⁶.

Some cytokines did not covary with stress: VEGF, BAFF, APRIL, IL-6, and IL-10. It is perhaps noteworthy that number of prior therapies was a significant negative predictor for all outcomes, except for IL-10 which was not associated with any covariates. This highlights the general importance of including treatment variables in tests of stress. It also suggests possible limiting factor(s) – the disease, repeated treatments, or some combination – in detecting stress effects on immunity for relapse/refractory CLL patients as manifest in these specific assays. When stress was a significant predictor, number of prior therapies was not correlated (entered) for ALC, IL-16, and CCL3 and was entered but was not significant for TNFα analysis. Data such as these are important for discerning under what specific conditions effects of stress may be detected. While the use of these covariates may account for null findings, results may be specific to CLL as previous research suggests stress is associated with increased levels of inflammatory cytokines in solid tumor cancers. For example, IL-6 may contribute to tumor suppression rather than immune suppression in CLL⁶⁷.

The strengths and limitations of the study are considered. The data come from complete accrual to a phase II trial of an experimental therapy, an important study context for biobehavioral studies. When both effectiveness and toxicities are unclear, insights into psychological or behavioral factors and their interaction with disease-specific biomarkers add to the overall contribution of a phase II trial. Internal validity was strong, but as a single-site trial at a comprehensive cancer center, patients generally have more resources, are younger, and are less diverse^{68, 69}. Low incidence blood cancers tend to receive treatment at regional centers, making the sample geographically diverse (patients traveled over 300 miles on average to receive treatment). Limited funds prevented analysis of all trial patients (N=152) but a sample of 96 was comparatively large for an exploratory study⁷⁰. The latter sample differed from the remainder only in terms of having a higher mean level of stress (IES) which was anticipated because of oversampling patients with higher stress for the second batch of assays. Of the covariate candidates, age, gender, and number of prior therapies were important, while del17p, comorbidity, and BMI played little, if any role in analyses, providing knowledge relevant to future studies. The cytokine/chemokine outcomes were chosen for their relevance to CLL, but there are others important for study¹¹, such as those for the T cell compartment (CD8+, CD57), NK cells, and response to Bruton’s tyrosine kinase (BTK) inhibition. Lastly, as an exploratory study, findings are not adjusted for multiple outcomes in parallel analyses.

Regarding the clinical contribution, other studies using the IES to assess cancer stress find higher stress covaries with higher levels of CLL signs/symptoms (e.g., fatigue, enlarged nodes, infections, and others)⁷¹. Stress is also an important individual difference variables that is predictive of future quality of life⁷¹ and is a moderator of intervention effectiveness. Patients with moderate to severe stress, anxiety, and/or depressive symptoms are in need of early, evidence based, psychological treatment to address current difficulties and lower the likelihood of poorer quality of life which may otherwise follow⁷³. Selected intervention trials have resulted in increased T cell immunity⁷⁴, greater production of Th1 cytokines (IL-2, IL-12, IFN–γ)^{75, 76}, and decreases in Th₂ cytokines (IL-4, IL-10)⁷⁶.

In conclusion, novel biobehavioral data for relapsed/refractory CLL patients show psychological stress is related to heightened levels of cellular, cytokine, and chemokine markers that are associated with progressive disease. Understanding the association between psychological stress and host factors, particularly immunity, is needed. Although previous research has examined stress associations with animal models⁵¹, this is the first biobehavioral study to show stress to covary with four key biomarkers in patients with CLL. Data are consistent with the hypothesis that stress may be a negative interface to an already weakened immune system. Replication and longitudinal data will be needed to clarify the trajectories of these responses and any relevance to CLL relapse.