The majority of participants were married, Caucasian, and postmenopausal, consistent with the parent study (Collado-Hidalgo et al., 2006
). Fatigued women were younger (51.2 vs. 62.2 years old; p
= .003), less likely to have received radiotherapy (64% vs. 100%; p
= .043), and longer post-diagnosis than non-fatigued women (3.03 vs. 2.14 years post-diagnosis; p
= .08). Fatigued women were also significantly more likely to have a current diagnosis of depression (50% vs. 0%; p
= .01) and marginally more likely to have a past depression diagnosis (55% vs. 20%; p
= .10). There were no other group differences in demographic, medical, or treatment-related characteristics.
3.1. Differential gene expression
A total of 437 transcripts showed >30% difference in expression in leukocytes from fatigued vs. non-fatigued survivors (119 relatively up-regulated in fatigued survivors, 318 up-regulated in non-fatigued survivors; see and Supplementary Table 1
). Prominent among genes showing up-regulation in fatigued survivors were genes encoding pro-inflammatory cytokines (IL1A, IL1B, IL6, OSM
), genes involved in chemokine signaling (CXCL2, CXCR5, CCL20
) and transcriptional activation (IER3, ZNF331, NR4A2, NR4A3
), and the vascular growth factor VEGFA
. Transcripts showing up-regulation in non-fatigued survivors included genes involved in cytotoxic T lymphocyte responses (TRGC2, TIGIT, CX3CR1, CMKLR1, GZMH
), the apoptosis promoter NLRC4
, the hemoglobin-regulator HP
, the RNA splicing regulator CROP
, and the glucoamylase MGAM
. We found no differences in expression of consensus leukocyte subset marker genes (i.e., CD3Z, CD4, CD8A, CD19, CD14, CD16
, or CD56
). Confirmatory RT-PCR analyses of four transcripts up-regulated in fatigued survivors and two up-regulated in non-fatigued survivors corroborated microarray findings (Supplementary Table 2
Fig. 1 Differential gene expression in fatigued versus non-fatigued breast cancer survivors. Affymetrix Human Gene 1.0 ST high-density oligonucleotide arrays identified 119 genes relatively up-regulated in fatigued survivors, and 318 genes up-regulated in non-fatigued (more ...)
Consistent with the hypothesis that fatigue is associated with increased activity of immune- and inflammation-related genes, Gene Ontology analyses identified inflammation (GO annotation GO:0006954 and GO:0002526), immune response (GO:0006955; GO:0002376), and cytokine (GO:0005125) functional annotations as over-represented in genes showing selective up-regulation in fatigued survivors (all p < .001).
3.2. Role of NF-κB and GR signaling
To test hypotheses about NK-κB and GR transcription factor activity, we conducted TELiS bioinformatic analyses of transcription factor response elements in the promoters of differentially expressed genes. Results showed a significantly greater density of NF-κB response elements in promoters of genes upregulated in fatigued vs. non-fatigued survivors (; average difference 2.28-fold ±.09 across nine combinations of promoter length and scan stringency, p < .0001), and a significant under-representation of glucocorticoid response elements in the promoters of genes up-regulated in fatigued vs. non-fatigued survivors (; average .45-fold ±.07 across nine combinations of promoter length and scan stringency, p = .007). This resulted in a net 5.10-fold skew in pro- vs. anti-inflammatory transcription factor-binding motifs in the promoters of genes up-regulated in fatigued survivors (difference from neutral 1.0-fold, p = .0004).
Fig. 2 Transcriptional activity of NF-κB and GR signaling pathways. Results of TELiS bioinformatics analyses showed significantly greater density of NF-κB response elements (2A), and significantly lower density of glucocorticoid response elements (more ...)
Analyses of covariance were used to remove all variation in gene expression profiles attributable to age, time since diagnosis, and radiation treatment prior to promoter-based bioinformatic analyses. Results continued to show a 2.37-fold greater prevalence of NF-κB response elements and a 0.87-fold lower prevalence of glucocorticoid response elements in promoters of genes up-regulated in fatigued survivors (net pro- vs. anti-inflammatory skew 2.72-fold, p = .041). Bioinformatic indications of increased NF-κB also remained significant in analyses controlling for current depression and depression history (as measured by SCID; both differences >1.5-fold, p < .02). Indications of GR desensitization were attenuated by control for current depression (ns after adjustment) but continued to indicate reduced GR activity after controlling for history of depression (0.77-fold, p = .082).
To determine whether reduced GR transcriptional activity was mediated by decreased cortisol output, serum cortisol levels were assayed in blood samples drawn concurrently with transcriptionally profiled leukocytes. Results showed no significant difference between groups (mean = 29.9 µg/dl ± 4.3 for fatigued vs. 33.9 ± 3.6 in non-fatigued, p = .50). To determine whether reduced GR transcriptional activity might be secondary to reduced expression of the GR gene, we analyzed NR3C1 mRNA levels in microarray gene expression profiles. Average levels of GR transcripts differed by <1% across groups (p = .80).