Depression is common during and after breast cancer treatment. However, the role of specific therapeutic modalities and related biologic mechanisms remains unclear. Radiation is an essential component of breast-conserving therapy and may contribute to depression in patients with breast cancer through the activation of inflammatory pathways.
Depressive symptoms and inflammatory mediators, including nuclear factor kappa B (NF-κB), were assessed at baseline (before radiation), during radiation, and 6 weeks after radiation in 64 women who had stage 0 through IIIA breast cancer.
No significant increases in depressive symptoms occurred during or after radiation, although a number of patients exhibited moderate-to-severe depression throughout the study. Multivariate analyses of baseline factors predictive of depression revealed that educational status, perceived stress, prior chemotherapy, and peripheral blood NF-κB DNA binding all were independent predictors of persistent depressive symptoms after radiation (all P < .05). Of these factors, only prior chemotherapy was associated with inflammatory mediators, including NF-κB DNA binding, soluble tumor necrosis factor-alpha receptor 2, and interleukin-6, which, in univariate analyses predicted depressive symptoms after radiation (all P < .05). Chemotherapy-treated patients also exhibited an over-representation of gene transcripts regulated by NF-κB.
Depression is a common acute and potentially long-term, debilitating behavioral toxicity of breast cancer and its treatment, occurring in up to 30% of women with breast cancer. Depression also has been associated with increased breast cancer mortality,[7, 8] and clinical trials of interventions that reduce depressive symptoms have reported increased survival in women with metastatic disease.[7, 9] Many factors may contribute to depression, including age at diagnosis, tumor stage, surgery, and chemotherapy.[1, 3, 5, 6] However, data are limited on depression in patients with breast cancer who are receiving radiation.
Recent data indicate that radiation provides a significant survival advantage for patients after breast-conserving surgery. However, predictors of depression during and after radiation remain largely unexplored, as studies are limited by small patient numbers, cross-sectional designs, and retrospective, secondary analyses.[2, 4, 6, 11] Furthermore, among the few longitudinal studies of depression during radiation, patients have been treated in a highly varied manner (lumpectomy vs mastectomy, with or without chemotherapy, and various radiation doses), and within this literature, there are several inconsistencies regarding symptom trajectory and severity as well as clinical and psychosocial predictors of depression. Clinical factors associated with increased depression during radiation include advanced cancer stage, mastectomy, prior chemotherapy, higher radiation duration/dose, younger age, and increased body mass index (BMI).[2, 12] However, data also suggest that depression is related to baseline psychosocial characteristics, including education level, relationship status, and anxiety and distress.[13, 14] Collectively, these reports suggest the need for prospective, longitudinal studies of patients uniformly treated with standardized surgery and radiation controlling for relevant clinical and psychosocial characteristics to clarify primary risk factors for depression, especially depression that persists after treatment.
One proposed mechanism linking depression to cancer treatments, including radiation, is inflammation. Increased inflammatory markers are identified in patients with depression, and the administration of inflammatory cytokines leads to depressive symptoms. Recent data also suggest that blockade of inflammatory cytokines reduces depressive symptoms, and specifically fatigue, in cancer patients.
Although radiation is known to cause tissue injury and induce a subsequent inflammatory response, only one study has evaluated the relationship between depression and inflammation in women undergoing breast radiation. In that study, soluble interleukin-6 receptor (sIL-6R) levels were significantly elevated in patients with high versus low depression. Many more studies have examined inflammatory mediators of fatigue, which is included in the diagnostic criteria for depression.[3, 11, 15, 17] Nevertheless, results from these studies have been inconsistent, possibly due to varying strategies for measuring cytokines and a lack of longitudinal data.[11, 18, 19] Fatigue during radiation has been associated with increased levels of IL-6, IL-1 receptor antagonist (IL-1ra), and C-reactive protein (CRP).[17, 18] However, other investigators have not reported this relationship after controlling for factors, including BMI.[3, 6, 19]
Although the data suggest a potential relationship among depression, inflammation, and radiation, the inflammatory signaling pathways have not been explored. One candidate pathway involves nuclear factor kappa B (NF-κB). NF-κB is a lynchpin signaling molecule in the inflammatory cascade and is implicated in cancer development and treatment resistance.[20, 21] Fatigued breast cancer survivors have increased activation of NF-κB–regulated genes. Because radiation increases NF-κB pathway activity in breast cancer cells, NF-κB activation may extend beyond the breast to peripheral tissues as a general response to tissue injury, and ultimately, may contribute to behavioral morbidities, including depression. Of note, chemotherapy has been associated with NF-κB activation both in breast cancer tissue and in peripheral blood.[16, 23]
To further explore clinical and inflammatory factors associated with depression in women receiving radiation, we conducted a longitudinal study before, during, and after a standardized course of radiation. The primary objective was to determine which factors were predictive of persistent depressive symptoms after radiation. In addition, we examined the relationship between the clinical factors predictive of depression and inflammation including circulating inflammatory biomarkers and inflammatory gene transcripts. Special emphasis was placed on the potential role of NF-κB and its downstream mediators, tumor necrosis factor (TNF), IL-1, and IL-6.
MATERIALS AND METHODS
From March 2010 to November 2011, patients were recruited from the Emory University Department of Radiation Oncology. Eligible women ranged in age from 18 to 75 years, had stage 0 through IIIA breast cancer, and had undergone breast-conserving surgery with or without neoadjuvant or adjuvant chemotherapy. Women were excluded for medical conditions that could directly influence the immune response, including pregnancy, autoimmune/inflammatory disorders, and uncontrolled cardiovascular, metabolic, pulmonary, or renal disease. Women who had a history of major psychiatric disorder, including schizophrenia, bipolar disorder, or substance abuse/dependence within the past year also were excluded. Patients with depression were not excluded, and the use of antidepressant medications was allowed. Drugs known to affect the immune system (eg glucocorticoids, methotrexate) excluding over-the-counter anti-inflammatory medications, were not permitted. Caucasians and African Americans were enrolled because of the paucity of other racial groups in the community.
All patients were prescribed 50.0 Gy to the whole breast followed by a 10.0-Gy to 16.0-Gy boost to the lumpectomy cavity given in 2.0-Gy fractions with 6.0-megavoltage (MV) photons and/or 18-MV photons using standard tangential field technique. Treatment plans were designed according to International Commission on Radiation Units and Measurements guidelines. Patients underwent clinical and behavioral assessments and peripheral blood sampling at 3 time points: baseline (1 week before radiation) (T1), week 6 of radiation (T2), and 6 weeks after competing radiation (T3). Study procedures were approved a priori by the Emory University Institutional Review Board, and all participants provided written informed consent.
Depression was assessed using the Inventory of Depressive Symptomatology-Self Report (IDS-SR), a 30-item scale that measures all symptom domains used to make a diagnosis of depression included in the Diagnostic and Statistical Manual of Mental Disorders, fourth edition. Higher scores indicate increased severity, and scores ≥33 points indicate moderate-to-severe depression. The IDS-SR has been validated in diverse patient populations including cancer patients. Participants also completed the 20-item Multidimensional Fatigue Inventory to assess fatigue. Because previous studies have demonstrated a relationship between baseline distress and depression, distress was assessed by using the perceived stress scale (PSS), which has been used in multiple populations, including breast cancer patients undergoing radiation.[2, 28]
NF-κB DNA Binding and Downstream Inflammatory Markers
Peripheral blood samples were drawn between 8 and 11 AM (to reduce circadian effects) at all 3 time points. Plasma was separated and stored at −80°C until subsequent batch assay. Peripheral blood mononuclear cells (PBMCs) were isolated using density gradient centrifugation and were stored in freezing serum (90% fetal bovine serum, 10% dimethyl sulfate) at −80°C until nuclear extraction or mRNA isolation.
DNA-binding of NF-κB in PBMCs was determined by using an enzyme-linked immunosorbent assay as previously described (Active Motif, Carlsbad, Calif). NF-κB DNA binding was performed in 58, 60, and 57 of the 64 participants at T1, T2, and T3, respectively, based on sample availability/quality. Plasma soluble TNFR2 (sTNFR2), IL-1ra, and IL-6 were assayed in duplicate using sandwich enzyme-linked immunosorbent assay (R & D Systems, Minneapolis, Minn). The mean inter-assay and intra-assay coefficients of variation were ≤10%. CRP levels were measured using the immunoturbidometric method with the Beckman AU 480 chemistry analyzer (Beckman Coulter, Brea, Calif) and the Ultra WR CRP reagent kit (Sekisui Diagnostics, Framingham, Mass). Inter-assay and intra-assay coefficients of variation were <3%.
Total RNA was extracted from PBMCs using RNeasy kits (Qiagen, Valencia, Calif). After extraction, RNA samples were dissolved in RNase-free water, and their concentrations and A260 of 280 ratios were determined using the MBA 2000 System (Perkin-Elmer, Shelton, Conn). Each sample was linearly amplified by using the WT-Ovation RNA amplification system (NuGEN Inc., San Carlos, Calif) and was used for microarray analysis. After hybridization to Illumina HumanHT-12 Expression BeadChips (Illumina, San Diego, Calif), BeadChips were scanned on the Illumina BeadArray Reader to determine probe fluorescence intensity. Raw probe intensities were normalized by using a quantile normalization algorithm.
Wilcoxon rank-sum tests were used to test differences in continuous or ordinal variables between groups defined by binary variables (eg patients who did and did not receive chemotherapy). Fisher exact tests were used to test associations between categorical variables. Spearman correlation coefficients were computed to determine univariate relationships between variables. Multiple linear regression was used to examine associations among relevant variables. Cytokine concentrations were skewed and therefore, log-transformed before analyses.
To identify functional biologic processes over-represented in genes differentially regulated among patients who received chemotherapy versus those who did not (see below) as well as the transcriptional regulatory pathways driving observed differences in gene expression, gene ontology (GO) and transcription factor bioinformatic analyses were conducted. These methods are most accurate with relatively large numbers of genes that have large biologic differences in expression. Therefore, differentially regulated gene transcripts were identified using an effect size with a ≥20% difference (1.2-fold change), corresponding to a <10% false-discovery rate[15, 31] and a cutoff P value of ≤ .05, and then subjected to the stringency of bioinformatic analyses to ensure statistical reliability. The Database for Annotation, Visualization, and Integrated Discovery (DAVID) functional annotation clustering tool was used for GO analysis, which employed a modified Fisher exact test to determine whether a gene list was enriched for genes involved in relevant biologic processes. A network-based transcription factor analysis was conducted with MetaCore software (GeneGo, Inc., St. Joseph, Mich) using an algorithm designed to query a manually curated database that has been determined to be both accurate and comprehensive in identifying transcriptional regulatory pathways and target genes, including those regulated by NF-κB.
Sixty-four women with stage 0 through IIIA breast cancer who met eligibility criteria agreed to participate. Eighteen eligible women refused participation, primarily because of time commitment, blood sampling requirements, or lack of interest. No significant differences were observed between patients who did or did not agree to participate. One patient dropped out before T2, and another dropped out before T3. Clinical and psychosocial characteristics of the participants are listed in Table 1. The higher percentage of women with stage 0/I disease (61%) accounted in part for the lower percentage of patients who received chemotherapy (38%).
Table 1. Patient Characteristics
No. of Patients (%)
Overall, N = 64
No Chemotherapy, N = 40
Chemotherapy, N = 24
Abbreviations: BMI, body mass index; DCIS, ductal carcinoma in situ.
aP < .05 compared with patients who did not receive chemotherapy.
Compared with baseline, depressive symptoms did not significantly increase during or after radiation (mean IDSSR score [± standard deviation]: T1, 15.9 ± 11.5; T2, 15.8 ± 2.2; T3, 14.8 ± 12.1; P value nonsignificant). Nevertheless, several patients exhibited moderate-to-severe depression (IDSSR score ≥33) throughout the study, including 9% at T1, 16% at T2, and 15% at T3. Fatigue severity also did not change during or after radiation. Fatigue and depression scores were highly correlated at all time points (P < .001).
Multivariate analysis, including baseline clinical, psychosocial, and inflammatory variables, revealed that educational status (P = .007), perceived stress (P = .03), chemotherapy (P = .03), and NF-κB DNA binding (P = .04) all were independently predictive of persistent depression 6 weeks after radiation. Age, race, initial cancer stage, income, marital status, hemoglobin, BMI, endocrine therapy, and antidepressant use were not associated with depressive symptoms after radiation and did not improve the multivariate model.
It is noteworthy that many of the factors that were predictive of depression were inter-related. For example, both lower educational status and prior chemotherapy treatment were associated with higher mean ± standard deviation PSS scores (education: high school graduates or less, 16.8 ± 7.7; college graduates, 12.4 ± 6.8 [P = .03]; treatment: chemotherapy, 17.3 ± 6.8; no chemotherapy, 12.1 ± 7.2 [P = .008]). However, among these factors, only chemotherapy was associated with inflammatory mediators, including NF-κB DNA binding, sTNFR2, and IL-6 (see below).
As indicated above, the receipt of prior chemotherapy was the only disease-related or treatment-related factor that was associated with depression after radiation. Figure 1 indicates that although neither women who received chemotherapy nor those who received radiation alone (without chemotherapy) exhibited an increase in depressive symptoms during or after radiation, chemotherapy-treated patients had significantly higher depression scores at all time points compared with those who did not receive chemotherapy (all P < .01). Similar results were observed for fatigue. In addition, before radiation, 25% of women who received chemotherapy versus 0% of those who did not receive chemotherapy exhibited moderate-to-severe depression (P < .005), while after radiation, 29% of the chemotherapy group exhibited moderate-to-severe depression versus 5% of the no chemotherapy group (P = .02). The median time between the last cycle of chemotherapy and the first day of radiation was 11.9 weeks (range, 3.6-21.9 weeks). No difference in depression or fatigue severity was observed at any time point between women who received neoadjuvant chemotherapy (n = 17) versus adjuvant chemotherapy (n = 7) or between those who received anthracycline-based (n = 12) versus nonanthracycline-based (n = 12) chemotherapy (all P values were nonsignificant). It is noteworthy that, after radiation, significantly more women in the chemotherapy group were receiving antidepressants (8 women; 33.3%) than women in the no chemotherapy group (3 women; 7.5%; P = .01). Women in the chemotherapy group were more likely to be younger than 50 years of age (46% vs 13%; P < .01), to have stage II/III disease (75% vs 18%; P < .001), and to have tumors that were either hormone receptor-negative or human epidermal growth factor 2 (Her2)-positive (P < .01) (see Table 1).
NF-κB DNA Binding and Downstream Inflammatory Mediators
As previously indicated, multivariate analysis revealed that NF-κB DNA binding at baseline was associated with depressive symptoms after radiation. Univariate analyses revealed that baseline NF-κB DNA binding also was correlated significantly with depressive symptoms at baseline and during radiation, but only in patients who previously received chemotherapy (baseline: r = 0.64; P < .005; during radiation: r = 0.63; P = .03) (Fig. 2). Throughout the study, NF-κB DNA binding also was correlated with fatigue in women who received chemotherapy but not in those who did not receive chemotherapy (all P < .05)
To better understand how chemotherapy is related to inflammatory mediators downstream of NF-κB, we conducted multivariate analyses controlling for age, BMI, PSS, antidepressant use, hormone therapy, initial cancer stage, educational level, income, race, and marital status. Baseline concentrations of both IL-6 (P = .002) and sTNFR2 (P = .003) were significantly higher in the chemotherapy group versus the no chemotherapy group (P < .05) (Table 2). CRP levels also were increased in the chemotherapy group, but the difference was not statistically significant (P = .06). Univariate analyses of the correlation between baseline inflammatory mediators and behavioral endpoints revealed a significant correlation between baseline IL-6 and both depression and fatigue at T3 (depression: r = 0.32; degrees of freedom [df] = 60; P = .01; fatigue: r = 0.44; df = 61; P < .01). Moreover, baseline sTNFR2 and CRP levels were correlated with fatigue (but not depression) at T3 (TNFR2: r = 0.31; df = 60; P = .02; CRP: r = 0.38; df = 61; P = .002). No correlations between behavioral endpoints and baseline IL-1ra levels were observed. Radiation treatment itself, as noted above, was not associated with increases in depression or fatigue, and no significant increases in peripheral inflammatory markers were observed during or after radiation in multivariate analyses controlling for the variables noted above (all P values were nonsignificant).
Table 2. Chemotherapy Status and Baseline Inflammatory Biomarkers
Overall, N = 58
No Chemotherapy, N = 38
Chemotherapy, N = 20
Abbreviations: SD, standard deviation.
aP < .01 compared with patients who did not receive chemotherapy.
On the basis of the correlations between chemotherapy, NF-κB DNA binding, and persistent depression observed 6 weeks after radiation, we examined the expression of genes in the chemotherapy group versus the no chemotherapy group at T1. In total, 340 gene transcripts were differentially expressed: 128 were up-regulated and 212 were down-regulated in the chemotherapy group versus the no chemotherapy group. Figure 3 indicates that the top 2 biologic processes represented in the GO analysis were the immune and defense responses. MetaCore analysis of genes that were differentially expressed in the chemotherapy group versus the no chemotherapy group revealed an over-representation of genes regulated by NF-κB family transcription factors, including RelA, the p65 subunit of NF-κB (Z = 99.9; P = 5.7 × 10−89), and the NF-κB complex (Z = 89.4; P = 2.3 × 10−71). Transcription factor analysis of all differentially expressed genes identified 48 NF-κB–regulated genes, corresponding to 60 NF-κB–regulated gene transcripts (Fig. 2). Of 60 NF-κB–regulated transcripts, 38 were increased (30% of 128 up-regulated gene transcripts), and 22 were decreased (10.4% of 212 down-regulated transcripts) in the chemotherapy group.
Prior chemotherapy was associated with significantly higher depression scores before, during, and after breast cancer radiation independent of chemotherapy type or whether it was received as neoadjuvant or adjuvant treatment, underscoring a persisting effect of chemotherapy up to several months after the last cycle of treatment. Only the women who received chemotherapy had increased expression of NF-κB–regulated gene transcripts and increased plasma levels of the downstream inflammatory mediators IL-6 and sTNFR2. In addition, baseline NF-κB DNA binding independently predicted depression after radiation when the analysis was controlled for multiple clinical factors, including BMI, initial cancer stage, age, endocrine therapy, and antidepressant use. Thus, chemotherapy-induced inflammation may be an important mechanism by which patients with breast cancer develop depression, and those who previously received chemotherapy may be the most at risk for both increased inflammation and depression during and after radiation.
Significant increases in the expression of NF-κB–regulated gene transcripts were observed in the women who received chemotherapy, and NF-κB DNA binding was associated with depressive symptoms in the chemotherapy group but not in the no chemotherapy group throughout the study. Chemotherapy may activate NF-κB through the destruction of rapidly proliferating malignant and nonmalignant cells, leading to an inflammatory response. Moreover, chemotherapy can directly activate NF-κB signaling pathways in multiple cell types. Relevant to depression, NF-κB activation induces inflammatory cytokines, which can access the brain in humans and activates a central inflammatory response associated with altered metabolism of serotonin involved in depression. Stress-induced NF-κB activation leads to depressive-like behavior in rodents and inhibits neurogenesis in brain regions involved in depression.[15, 36] Neurogenesis is an important component of antidepressant action and may explain why some patients exhibited significant depressive symptoms after radiation despite antidepressant treatment. In terms of potential mechanisms that may explain the lingering effects of chemotherapy, NF-κB can undergo epigenetic modification leading to persistent activation and may explain the increased NF-κB gene transcripts reported in fatigued versus nonfatigued breast cancer survivors several years after completing treatment.
Previous research has suggested that specific downstream inflammatory mediators of NF-κB are associated with distinct behavioral morbidities of cancer treatment in long-term breast cancer survivors. For example, fatigue, but not depression, has been associated with elevated levels of sTNFR2, and it has been demonstrated that TNF inhibitors reduce fatigue in patients with advanced cancer.[16, 39] Increases in CRP and IL-1ra also have been correlated with fatigue after controlling for sleep and depression, and fatigue has been associated with sIL-6R independent of depression. In the current study, fatigue and depression were correlated significantly with each other at all time points. Baseline NF-κB DNA binding activity and IL-6 were correlated significantly with depression and fatigue after radiation. Nevertheless, consistent with previous studies, sTNFR2 and CRP were correlated with fatigue, but not with depression after radiation, supporting the notion that activation of fundamental inflammatory signaling pathways like NF-κB may be common to symptoms of both depression and fatigue, whereas more nuanced correlations between specific cytokines and specific symptoms may exist, especially after treatment.
Findings from the current study indicate that radiation did not contribute to depressive symptoms at the time points measured and did not appear to exacerbate preexisting depression. Regarding the trajectory of depressive symptoms and inflammation, no significant increases in depression or inflammatory markers were observed during or after radiation. In other studies of patients who did not receive chemotherapy, greater increases in depressive symptoms have been observed, although, compared with our study, there was more variability in radiation dose and areas treated.[3, 17] The results of the current study were unexpected but are important, because previous research has not adequately addressed the effect of multimodality therapy. On the basis of our data, the contribution of chemotherapy to persistent depression after radiation far outweighs any contribution from radiation alone.
Regarding clinical relevance, almost 30% of breast cancer patients who received chemotherapy plus radiation (compared with 5% of patients who received radiation alone) exhibited depressive symptoms (IDSSR scores ≥33), the severity of which would qualify them for a clinical trial of antidepressant medication. Given the impact of depression on quality of life as well as mortality, these results highlight the importance of psychiatric screening of breast cancer patients who have completed radiation and have previously received chemotherapy.
Several limitations and strengths of the study warrant consideration. Regarding limitations, the relatively small number of participants limits generalizability, although 42% of patients were African American, indicating that our findings may be largely independent of race. In addition, because participants were not assessed before chemotherapy, the chemotherapy effect on depression and the trajectory of this symptom during chemotherapy and surgery cannot be determined. It is also possible that the chosen time points missed acute inflammatory or behavioral changes because of radiation treatment. Moreover, factors like advanced cancer stage that contribute to the decision to treat with chemotherapy may be the same factors that are associated with depression, albeit initial cancer stage was not an independent predictor of depressive symptoms and, when it was included in multivariate analyses, did not affect the results. Furthermore, there may be an interaction between chemotherapy and radiation leading to persistent behavioral changes that are not observed in patients who forego radiation. A comprehensive, prospective, longitudinal study tracking patients from diagnosis through treatment with chemotherapy and surgery with or without radiation is needed to fully address this question. Regarding strengths, all patients underwent lumpectomy and received standard radiation treatment. Moreover, use of the IDS-SR, a validated index with established and defined cutoff scores for pathology, allowed us to determine the prevalence of moderate-to-severe cases of depression warranting treatment.
In summary, the activation of inflammatory signaling pathways, including NF-κB, appears to be a potential mechanism by which chemotherapy is linked to depression, thereby identifying a subgroup of patients at high risk for depression and a potential biologic pathway for intervention. Future longitudinal studies assessing breast cancer patients from diagnosis through each component of multimodal treatment are needed to clarify the time sequence linking chemotherapy, inflammation, and the development of depression.
This study was supported by the Loan Repayment Program Award L30 CA171103 and the Exploratory/Development Grant 1R21 CA155511-01A1 from the National Institutes of Health, National Cancer Institute. The Cooper Family Foundation Breast Cancer Initiative, The Robbins Scholar Award from Winship Cancer Institute, The Cancer Control and Population Sciences Pilot Project Grant from the Winship Cancer Institute, and the Radiation Therapy Oncology Group (RTOG) also provided support for this study.