Association of genetic variants with fatigue in patients with malignant glioma

Abstract

Background

Fatigue is a consistently reported, severe symptom among patients with gliomas throughout the disease trajectory. Genomic pathways associated with fatigue in glioma patients have yet to be identified.

Methods

Clinical factors (performance status, tumor details, age, gender) were collected by chart review on glioma patients with fatigue (“I have lack of energy” on Functional Assessment of Cancer Therapy-Brain), as well as available genotyping data. Candidate genes in clock and inflammatory pathways were identified from a literature review, of which 50 single nucleotide polymorphisms (SNPs) in 7 genes were available. Clinical factors and SNPs identified by univariate analyses were included in a multivariate model for moderate-severe fatigue.

Results

The study included 176 patients (median age = 47 years, 67% males). Moderate-severe fatigue was reported by 43%. Results from multivariate analysis revealed poor performance status and 2 SNPs were associated with fatigue severity. Moderate-severe fatigue was more common in patients with poor performance status (OR = 3.52, P < .01). For each additional copy of the minor allele in rs934945 (PER2) the odds of fatigue decreased (OR = 0.51, P < .05). For each additional copy of the minor allele in rs922270 (ARTNL2) the odds of fatigue increased (OR = 2.38, P < .01). Both of these genes are important in the circadian clock pathway, which has been implicated in diurnal preference, and duration and quality of sleep. No genes in the inflammatory pathway were associated with fatigue in the current study.

Conclusions

Identifying patients at highest risk for fatigue during treatment allows for improved clinical monitoring and enrichment of patient selection for clinical trials.

Cancer-related fatigue is a distressing, persistent, subjective sense of tiredness or exhaustion related to cancer or cancer treatment that is not proportional to recent activity and interferes with usual functioning.¹ Fatigue and alterations in sleep are consistently reported symptoms among glioma patients. Altered sleep has been reported to occur in up to 54% of glioma patients and fatigue in up to 70% of patients throughout the illness trajectory.^2–4 The prevalence of both symptoms is even higher during and immediately following initial radiation therapy, with greater than 80% of patients reporting fatigue and greater than 90% reporting hypersomnia.⁵ The impact of these symptoms on treatment tolerance and outcome is significant, with reports indicating worse performance status and shorter survival in patients who experience these symptoms.⁶

Many factors are recognized to impact the occurrence of fatigue in brain tumor patients including concomitant medications, underlying neurologic impairment, a variety of comorbid conditions, and directed cancer therapies such as radiation.^7–10 There is a growing body of work exploring underlying genetic mechanisms of fatigue and altered sleep in both cancer and neurologic conditions. The pathophysiologic mechanisms are not completely understood; proposed mechanisms include cytokine dysregulation,^11,12 hypothalamic-pituitary-adrenal (HPA) axis dysfunction,^12–14 and circadian rhythm dysfunction.^15–17

Although fatigue is a common symptom, clinical and genomic pathways associated with risk for more severe fatigue in glioma patients have yet to be identified, other than the identified risk associated with radiation therapy that is frequently administered at initial diagnosis. Based on a review of the literature and a pilot study, a model has been developed proposing that neuroinflammation resulting from cranial radiation may cause alterations in the central molecular clock, leading to altered sleep-wake cycle and production of cytokines and neurotransmitters, ultimately resulting in fatigue.⁹ The purpose of this study was to explore genetic variants associated with moderate to severe fatigue in patients with glioma.

Methods

The population for this study was a subset of patients from a prospective epidemiologic study of newly diagnosed malignant glioma patients consecutively diagnosed and treated at The University of Texas MD Anderson Cancer Center (Houston, TX) between 1992 and 2009 (R01 CA070917, Bondy-PI). The study was approved by MD Anderson Institutional Review Board. Of 1247 patients, 233 had been clinically referred for and completed comprehensive neuropsychologic evaluation at baseline and during the treatment epoch as clinically indicated by the Section of Neuropsychology (also at MD Anderson) and had genotype data available for analysis. The Functional Assessment of Cancer Therapy-Brain (FACT-BR) was included as part of the evaluations. This dataset was also used to evaluate genetic variants associated with altered neurocognitive function at diagnosis, with results reported elsewhere by Liu and colleagues.¹⁸ This same subset was then used to explore self-reported fatigue. Data completion rates based on time from study entry were reviewed (range from 0–14 years after diagnosis, with the majority of patients [n = 176] completing reports within the first 3 years). Self-reports included complete fatigue, SNP, and clinical data. The current analysis included only patients of European descent with a newly diagnosed malignant glioma who had both a completed self-report of fatigue within 3 years of diagnosis and available baseline genotype data (selected 7 genes) as described below.

According to the National Comprehensive Cancer Network, fatigue can be measured as unidimensional construct as the sensation of fatigue or tiredness or using multidimensional scales.¹⁹ For this project, fatigue was measured using the FACT-BR item “I have a lack of energy.” The item is rated from 0 (none) to 4 (very much). If more than one FACT-BR was completed within the 3 years, the highest rating among completed assessments was used as the outcome of interest. Patients were classified based on the rating descriptions as having none-minor fatigue (rating 0–2) or moderate-severe fatigue (rating of 3–4) for the analysis to identify those whose fatigue is more likely to be associated with functional limitations.²⁰ Demographic and clinical factors (Karnofsky Performance Status [KPS], tumor characteristics, treatment history, age, and gender) were collected by chart review. Performance status was further defined as good (KPS ≥ 90) or poor (KPS ≤ 80).²¹

Genomic DNA was extracted from venous blood samples and genotyped as part of the parent study described above and detailed elsewhere.²² Human610-Quad Bead Chips were used according to the manufacturer’s protocols (Illumina). A literature review was performed to identify genes that have been associated with fatigue in the cancer and neurologic patient populations. From that review, 10 genes were identified in association with fatigue: ARNTL2, PER1, CLOCK, PER2, DBP, PER3, FBXL3, TNFα, GNβ3, and IL4.^{11,12,15,16,23–27} These genes all fall into either the circadian or inflammatory pathways. Genotype data from the parent study were then reviewed and compared to this gene list identified from the literature. Genotype data were available for 7 of the 10 identified genes.

Monomorphic SNPs and SNPs with minor allele frequency lower than 0.05 were removed from the analysis. Tagging SNPs that were in strong linkage disequilibrium (r² > 0.80) with other SNPs in the same gene were selected to reduce the number of comparisons. Finally, 50 SNPs in the following 7 genes were included in the analysis: ARNTL2, PER1, CLOCK, PER2, PER3, FBXL3, and IL4.

Statistical Analysis

Descriptive statistics were used for demographic and clinical factors and the fatigue severity rating. Associations of patient demographics and clinical factors with fatigue severity were evaluated by chi-square tests and point biserial correlations. The identification of SNPs that were associated with fatigue severity was performed in 2 steps, according to model building procedures described by Hosmer and Lemeshow.²⁸ First, the available SNPs and clinical factors were evaluated by univariate logistic regression with fatigue severity rating as the outcome variable. Factors with P values ≤ .10 were flagged for inclusion in the second step. Second, all significant factors identified by univariate analyses were included in a multivariate logistic regression model, with backwards selection at P < .05, for fatigue severity. The resulting set of factors at the multivariate level were deemed to be associated with fatigue severity. Analyses were conducted through IBM SPSS Version 23.²⁹

Results

Patient Sample

The sample was comprised of 176 patients, primarily male (67%) with a median age of 47 years (range, 19–75 years), and a documented good KPS (median = 90). The primary diagnosis was grade IV glioma (56%) with patients having undergone a subtotal (40%) or gross total (32%) resection at diagnosis. The majority of patients were on treatment (35%) or had completed treatment (28%) at the time of their highest fatigue rating. Thirty patients (17%) had tumor progression prior to their highest fatigue rating. Fatigue was reported as none-minor in 57% of patients and moderate-severe in 43%. Demographic and clinical characteristics of the sample are summarized in Table 1. Mean time to worst fatigue and fatigue ratings are presented in Table 2.

Table 1.

Demographic and Clinical Characteristics

			Lack of Energy		Univariate Test
			None-Mild	Moderate- Severe	P value
Age		Mean (SD)	45 (13)	48 (13)	.078
		Median	45	49
		Range	19–71	22–75
		N (%)	%	%
Sex	Male	117 (67%)	60.7%	39.3%	.146
	Female	59 (34%)	49.2%	50.8%
Tumor grade / diagnosis	II	15 (9%)	60.0%	40.0%	.920
	III	62 (35%)	56.5%	43.5%
	IV	98 (56%)	57.1%	42.9%
Surgery extent	Biopsy	48 (28%)	58.3%	41.7%	.964
	Subtotal	70 (40%)	55.7%	44.3%
	Gross total	56 (32%)	57.1%	42.9%
Karnofsky Performance Status	Poor (≤80)	23 (15%)	39.1%	60.9%	.030
	Good (≥90)	127 (85%)	63.8%	36.2%
Treatment status*	Treatment naive	49 (28%)	65.3%	34.7%	NA
	Treatment	112 (64%)	51.8%	48.2%
	On treatment**	62 (35%)	51.6%	48.4%
	Post treatment**	50 (28%)	52.0%	48.0%
	Unknown timing	15 (9%)	66.7%	33.3%
Progression status	Unknown	43 (24%)	53.5%	46.5%	NA
	No	80 (45%)	63.8%	36.3%
	Yes	53 (30%)	49.1%	50.9%
	Rating before progression	30 (17%)	50.0%	50.0%
	Rating after progression	23 (13%)	47.8%	52.2%

*Treatment status at the time of fatigue assessment.

**The majority of patients on treatment were receiving chemotherapy postradiation (56%) and the majority in the post-treatment group completed radiation and chemotherapy (84%).

Table 2.

Fatigue Characteristics

Days to highest fatigue rating from diagnosis Mean (SD) 206.2 (242.2) Median (Range) 130 (-70, 1002)
FACT-BR item: “I have lack of energy”
	Rating	n	%
	0	7	4.0	None-Mild 100 (57%)
	1	28	15.9
	2	65	36.9
	3	50	28.4	Moderate-Severe 76 (43%)
	4	26	14.8

Predictors of Fatigue

Results from univariate analyses indicated that clinical factors associated with report of significant fatigue included age (P = .078) and poor performance status (P = .030). There was no difference in the proportion of patients reporting mild versus moderate-severe fatigue based on treatment status or tumor recurrence. Analysis of SNPs revealed rs922270 (ARNTL2) (P = .033), rs3792603 (CLOCK) (P = .059), rs2253820 (PER1) (P = .089), and rs934945 (PER2) (P < .100) to be associated with fatigue severity (Table 3). One hundred fifty patients had both clinical factors and genotyping available for multivariate analysis. Results from multivariate analysis revealed poor performance status and 2 SNPs were associated with fatigue severity. Moderate-severe fatigue was more common in patients with poor KPS (OR = 3.52, P < .01). The presence of each additional minor allele in SNP rs934945 decreased odds of fatigue (OR = 0.51, P < .05). The presence of each additional minor allele in SNP rs922270 increased odds of fatigue (OR = 2.38, P < .01). These results suggest an effect of variants in the circadian pathway on fatigue in brain tumor patients. Full results are provided in Table 4.

Table 3.

Univariate Analyses Results Among SNPs

Pathway	Gene	SNP ID	Location	Chr:	Position	MAF	MA	Sig.	OR
Clock	ARNTL2	rs922270	Intron	12:	27396218	0.155	C	.033	1.869
Clock	CLOCK	rs3792603	Intron	4:	55996815	0.215	G	.059	1.599
Clock	PER1	rs2253820	Coding	17:	7988894	0.152	A	.089	.604
Clock	PER2	rs934945	Coding	2:	238819792	0.187	A	.100	.614
Clock	FBXL3	rs700361	Intron	13:	76497955	0.101	C	.128	.549
Clock	CLOCK	rs11932595	Intron	4:	56018354	0.411	G	.149	1.363
Clock	CLOCK	rs11931061	Intron	4:	56033550	0.327	G	.149	.703
Clock	ARNTL2	rs17497606	Intron	12:	27385716	0.186	T	.175	1.471
Clock	PER3	rs707467	Intron	1:	7784271	0.179	G	.190	.690
Clock	FBXL3	rs9573983	Intron	13:	76492516	0.066	G	.191	.561
Clock	PER3	rs10462020	Coding	1:	7803270	0.195	G	.215	1.409
Clock	ARNTL2	rs7300833	Intron	12:	27441977	0.098	G	.229	.630
Clock	PER2	rs10462023	Intron	2:	238849320	0.407	A	.300	1.259
Clock	PER3	rs228642	Intron	1:	7803233	0.412	C	.353	.806
Clock	ARNTL2	rs11048980	Intron	12:	27394814	0.091	C	.370	.721
Clock	PER1	rs2585408	3’ Flanking	17:	7984478	0.423	T	.439	1.186
Clock	ARNTL2	rs2306073	Intron	12:	27447104	0.322	A	.441	1.196
Clock	ARNTL2	rs2289709	3’ Flanking	12:	27464900	0.134	A	.451	1.283
Clock	ARNTL2	rs1471327	Intron	12:	27420353	0.125	T	.454	.786
Clock	ARNTL2	rs1443857	Intron	12:	27436627	0.148	A	.459	1.242
Inflammation	IL4	rs2243300	5’ Flanking	5:	132031985	0.076	T	.503	.757
Clock	ARNTL2	rs10842913	Intron	12:	27460330	0.094	A	.520	1.291
Clock	ARNTL2	rs7306410	Intron	12:	27416863	0.182	C	.522	1.191
Clock	ARNTL2	rs2587052	Intron	12:	27460479	0.229	T	.530	1.178
Clock	ARNTL2	rs10506018	Intron	12:	27387124	0.17	G	.551	.841
Clock	ARNTL2	rs11048994	Intron	12:	27422045	0.191	A	.588	.868
Clock	PER3	rs697686	Intron	1:	7790158	0.344	A	.592	.880
Clock	ARNTL2	rs1037921	Coding	12:	27444833	0.054	C	.592	1.348
Clock	ARNTL2	rs3751221	Intron	12:	27429922	0.17	C	.613	1.166
Clock	PER3	rs10127838	Intron	1:	7804367	0.224	T	.614	1.135
Clock	CLOCK	rs1522113	Intron	4:	56026528	0.053	A	.617	.787
Clock	ARNTL2	rs1256955	Intron	12:	27460292	0.192	C	.629	1.143
Clock	ARNTL2	rs11048977	Intron	12:	27388716	0.198	A	.656	.886
Clock	CLOCK	rs9312661	Intron	4:	56037083	0.378	A	.716	.920
Clock	PER3	rs228682	Intron	1:	7796286	0.411	C	.785	1.063
Clock	ARNTL2	rs4964058	Intron	12:	27414033	0.469	G	.817	1.052
Clock	PER3	rs228729	5’ near gene	1:	7785635	0.332	A	.823	.951
Clock	PER2	rs6431590	Intron	2:	238829867	0.308	G	.828	.951
Clock	PER1	rs2518023	5’ Flanking	17:	7997331	0.088	A	.842	.930
Clock	PER2	rs4663868	Intron	2:	238825830	0.08	T	.856	1.071
Inflammation	IL4	rs2243288	Intron	5:	132045843	0.166	G	.865	.955
Clock	PER3	rs228654	Intron	1:	7837168	0.117	T	.872	1.058
Clock	ARNTL2	rs16931937	Intron	12:	27439458	0.076	C	.882	.944
Clock	PER2	rs2304673	Intron	2:	238850661	0.148	C	.901	1.039
Clock	FBXL3	rs599115	Intron	13:	76487726	0.322	A	.941	1.016
Clock	ARNTL2	rs4964059	Intron	12:	27420486	0.354	C	.953	1.013
Clock	ARNTL2	rs4964060	Intron	12:	27424634	0.414	A	.964	.990
Inflammation	IL4	rs2243248	5’ Flanking	5:	132036543	0.059	G	.974	.985
Clock	PER3	rs10462018	Intron	1:	7802214	0.147	T	1.000	1.000
Clock	ARNTL2	rs11048998	Intron	12:	27437644	0.133	C	1.000	1.000

Chr, chromosome; SNP, single nucleotide polymorphism; MA, minor allele; MAF, minor allele frequency; Sig., significance

Table 4.

Clinical Factors and SNPs on Risk of Moderate/Severe Fatigue in Glioma Patients (N = 136)

Factor	B	SE	P value	OR	95% CI
Karnofsky Performance Status
Poor (≤80) (Ref = Good (≥90))	1.26	049	.01	3.52	(1.35, 9.18)
SNPs
rs922270 (ARTNL2)	0.87	0.33	.01	2.38	(1.24, 4.54)
rs934945 (PER2)	−0.68	0.35	.05	0.51	(0.26, 1.01)

Ref, reference group; SNP, single nucleotide polymorphism

Discussion

This report is the first that we are aware of that evaluates genotypes associated with fatigue in patients with malignant glioma during the first 3 years of diagnosis. Fatigue is common; 43% of all patients with malignant gliomas reported moderate-severe fatigue during the first 3 years of diagnosis in this cohort. Importantly, the identified risk factors for moderate-severe fatigue experienced by patients with malignant gliomas included definable molecular, genetic, clinical, and demographic factors.

Previous reports exploring gene associations with self-report of fatigue have indicated that specific cytokine genes, including IL4 and TNFα, were associated with more severe fatigue in patients with a variety of solid tumor malignancies, including a small subset of patients with brain cancer.²³ At present, there is limited understanding of the neural processes that may mediate or be mediated by the effects of peripheral inflammation, although it has been postulated that this may be through impact on the dopaminergic system, resulting in an increase in dopaminergic tone. Another theory proposes that alterations in proinflammatory cytokine production through regulation of the HPA axis modulates fatigue.³⁰

In this report no significant association between fatigue and the inflammatory gene IL4 was elucidated. Rather, SNPs in the circadian pathway genes ARNTL2 (rs922270) and PER2 (rs934945) were associated with fatigue. Genes from the PER and ARNTL families are expressed at opposite time of day³¹ and the encoded proteins perform antagonistic roles as part of the core clock negative feedback loop. PERs including PER2 in combination with CRYs form a protein complex repressing the activity of the transcription factors ARNTL1/2/CLOCK.³² TNFα was shown to induce the expression of ARNTL2 in fibroblasts,³¹ potentially linking the effect of inflammation on the circadian clock. Furthermore PER2 is expressed in various regions of the brain and is part of the core circadian transcriptional complex.³³ Loss of PER2 affects sleep timing³⁴ and may affect parameters of sleep homeostasis.³⁵ The identified PER2 SNP encodes a missense mutation in the CRY binding domain of the protein³⁶ that could lead to an abrogated clock mechanism. A recent study demonstrated that sleep duration and disruption in persons with human immunodeficiency virus is associated with clock gene variants, including ARNTL2 and PER1.³⁷ Associations with winter seasonal affective disorder and dementia due to Alzheimer’s disease have also been reported with these circadian pathway genes.^38–40

A circadian clock and fatigue relationship has also been previously reported in a population-based sample of the Health 2000 dataset from Finland, in which clock gene variants were associated with less fatigue in males, and with shorter sleep duration, lower energy levels, and less social activity in females.^41,42 The difference in the associated self report between males and females in the Finland study may reflect sex differences in how fatigue is reported. The use of a single measure “lack of energy” in this study may have allowed for the increased reporting in males, resulting in a nonsignificant difference in reporting by sex, unlike earlier oncology studies in which higher fatigue has been reported in females,⁴³ including our report in brain tumor survivors.⁴ Future studies should be conducted with a measure designed to more fully interrogate the dimensions of fatigue, to ascertain if there are differences in severity by sex and if these are varied across the disease epoch.

We have previously hypothesized that fatigue and daytime hypersomnia in brain tumor patients may be related to impact on the pineal gland as a consequence of radiation therapy.⁹ The hypothesis is further supported by previous reports that indicate a somnolence syndrome—with symptoms including fatigue, excessive drowsiness, incoordination and inability to concentrate—occurs postradiation in many patients with malignant glioma.^7,8 The ability to identify patients at highest risk for fatigue during and after treatment is important for clinical care and evaluation of clinical response. Identification would allow for improved clinical monitoring, enrichment of patient selection for clinical trials targeting fatigue, and identification of patients who may benefit from intervention and supportive clinical care.

There are several limitations to this report. Given the retrospective nature of this study, a validated measure of fatigue and criteria to classify fatigue severity were not available. In addition, a single rating of fatigue was assessed at heterogeneous times within the first 3 years after diagnosis. As noted in the introduction, fatigue as a unidimensional construct has been supported in the literature and was used for this preliminary work. Detailed therapy and outcome information (e.g., type and dose of adjuvant therapy, toxicities, use of stimulants or sleep promoting agents) that may be associated with fatigue was not available. Fatigue in the brain tumor population is a pervasive symptom occurring in the majority of patients. The primary goal of this study was to explore whether polymorphisms associated with a more severe fatigue phenotype could be identified, as has been done in other cancer and neurologic populations. SNPs were chosen based on their association with fatigue in these other populations. Although a genome-wide association study would have allowed for a more robust analysis, a much larger patient sample would be needed. This preliminary study supports that there is variability in the severity of fatigue and that individual risk of fatigue as a consequence of the disease and therapy may be varied based on underlying genetic susceptibility. Given the potential importance in identifying patients at high risk of severe fatigue to either modulate or develop treatment or preventative measures, future studies validating these findings in a prospective study are warranted.

Conclusions

This initial report adds to the body of literature concerning fatigue in glioma patients, with results indicating that specific genetic, clinical, and demographic characteristics of glioma patients are associated with risk for fatigue and that the occurrence of specific polymorphisms may be associated with more severe fatigue. It is important to recognize that fatigue is a common and complex symptom and standards for clinical evaluation and management exist and should guide clinical care. This report should be considered preliminary and further validation of the findings is needed to confirm the reported association. The implications of these findings include improved understanding of the underlying biologic basis of fatigue in this patient population and the potential for improved symptom management in this vulnerable patient population.

Funding

Financial Support: R01 CA070917 (Bondy for genomic analysis).

Conflict of interest statement. The authors declare no potential conflicts of interest.