Cognitive Performance in Breast Cancer Survivors and Markers of Biological Aging

Abstract

BACKGROUND

Biological aging pathways accelerated by cancer treatments may be a mechanism for cognitive impairment in cancer survivors. The goal of the present study was to examine whether indicators of biological aging, namely elevated levels of DNA damage, reduced telomerase enzymatic activity, and shorter peripheral blood mononuclear cell (PBMC) telomere length (TL) would be related to cognitive function in a cohort of breast cancer survivors.

METHODS

We evaluated a cross-sectional sample of 94 women aged 36–69 treated for early-stage breast cancer 3–6 years earlier. Leukocyte DNA damage, PBMC telomerase enzymatic activity, PBMC TL, and the inflammatory marker sTNF-RII were determined from blood samples. Cognitive function was assessed by a neuropsychological test battery and self-report. Linear regression models examined the relationship of biological aging predictors with cognitive outcomes.

RESULTS

Both higher DNA damage and lower telomerase were statistically significantly related to lower executive function scores, adjusting for age, BMI, race, years from treatment, and intelligence score (Standardized Coefficients [B] = −0.23 & 0.30, P’s < 0.05). In addition, lower telomerase activity was associated with worse attention and motor speed (B’s = 0.30 & 0.24, P < 0.05). sTNF-RII and TL were unrelated to any of the neurocognitive domains.

CONCLUSIONS

These results suggest a significant association between measures of biological aging and objective measures of cognitive performance in breast cancer survivors. Future prospective studies are needed to confirm a causal role of biological aging as a driver of declines in cognitive function after cancer treatment.

Condensed abstract

In women treated for early-stage breast cancer on average over 4 years earlier, both high DNA damage and low telomerase activity were related to worse executive function, high DNA damage was associated with worse memory, and low telomerase activity was associated with worse attention and motor speed. These findings provide initial evidence for a link between markers of biological aging and cognitive performance in breast cancer survivors.

Introduction

Breast cancer is the most common cancer in women, with over two-hundred and sixty thousand new cases expected in the United States in 2018.¹ There are estimated to be more than 3 million breast cancer survivors in the US due to substantial advances in detection and treatment.^1,2 However, treatments also increase risk for long-term and late toxicities, including persistent fatigue, pain, and cognitive dysfunction. Further research is needed to better understand the factors that contribute to these adverse secondary health outcomes.^3–6

In this paper, we focus on cognitive dysfunction in breast cancer survivors and its association with processes that are part of biological aging, paralleling the aging related phenotype seen in some cancer survivors.^7,8 Cancer treatments, particularly radiation and some chemotherapeutic agents, work by damaging the DNA of cancer cells, preventing cell replication and causing cell death. However, these same treatments can induce damage to DNA of normal cells,^9,10 causing acute elevations in inflammation,^5,11–13 increasing expression of a marker of cellular senescence, p16^INK4,¹⁴ and having detectable effects on telomerase activity and DNA damage years later;¹⁵ all factors that contribute to accelerated biological aging.^14,16–24 However, few studies have examined whether biological aging plays a role in cognitive dysfunction in cancer patients and survivors.

Conroy and colleagues reported that among breast cancer survivors 3–10 years since treatment, elevated DNA damage was associated with reduced gray matter density particularly in regions showing compromise, suggesting this marker could relate to cognitive function in cancer survivors many years after completion of treatment.⁹ More research has focused on associating various markers of inflammation with cognitive dysfunction in cancer survivors.^13,25–28 The extant literature supports a role of biological aging and inflammation in cancer-related cognitive difficulties, but substantial gaps remain. Despite some initial investigations of the relationship in aging populations,^29–32 no one has yet examined whether cellular markers of biological aging correlate with objective and subjective cognitive function in cancer survivors. The goal of the present study was to test the hypothesis that cellular markers of biological aging and inflammation are related to worse subjective cognitive impairment and objective neuropsychological function in breast cancer survivors.

Methods

Participants

Participants for the current study were recruited from the UCLA Mind Body Study (MBS), a longitudinal, prospective cohort study of women with early stage breast cancer enrolled after the end of primary treatment and prior to initiating adjuvant endocrine therapy if indicated.^33–35 Full details of the design, eligibility/exclusion, recruitment and procedures used in the MBS are described elsewhere, ^33–37 and are summarized here. The MBS was designed to assess cognitive changes due to breast cancer endocrine therapy. Study eligibility for MBS required that women were between the ages of 18 and 65, had received a diagnosis of stage 0-IIIA breast cancer, and were fluent in English. Women were excluded from participation if they had any immune-related conditions such as autoimmune disease, evidence of uncontrolled depression, or a neurological condition.

MBS participants underwent comprehensive neurocognitive assessments as well as blood specimen collection at study enrollment, and 6 and 12 months later. Immediately after the 12-month MBS study visit, participants were invited to join a long-term follow-up study that included annual questionnaires that assessed cognitive complaints and other behavioral symptoms on an annual basis. After several years of follow-up, these participants were invited for an in-person assessment that occurred between 3–6 years after initial study enrollment, the timepoint of the current analyses. Of the 190 women in the original MBS cohort, 170 agreed to participate in the follow-up study and 94 of these women ultimately attended an in-person visit that replicated the initial neuropsychological assessments and blood specimen collection for inflammatory markers, as well as the assessment of telomerase, DNA damage, and telomere length. English was not the first language of one participant, and we excluded her neuropsychological data from analyses, resulting in a sample of n=93 for models of neuropsychological domains only, otherwise n=94. All procedures were approved by the University of California, Los Angeles institutional review board, and all participants provided informed consent.

Subjective Cognitive Function

Subjective cognitive function was measured with The Functional Assessment of Cancer Therapy: Cognitive Function (FACT-Cog) version 3,³⁸ a valid and reliable self-report instrument, of cancer-related cognitive difficulties. The FACT-Cog subscale measuring Perceived Cognitive Impairment (range 0–72) was selected as the main outcome as is recommended by the scale’s developers, with psychometric properties and scoring guidelines available from the Functional Assessment of Cancer Therapy website: http://www.facit.org/.

Neuropsychological Testing

All participants were administered a comprehensive neuropsychological battery. Raw test scores were population normalized and transformed into z-scores and averaged to produce the following 6 key domain scores: Learning, Memory, Attention, Visuospatial, Executive Function, and Motor Speed. Additional details on specific tests used in the assessments of each domain can be found in Table 1. Analyses examining cognitive function adjust for age and premorbid estimates of IQ determined by Wechsler Test of Adult Reading (WTAR) scores.³⁹

Table 1.

Neuropsychological tests administered to patients in the Mind-Body Study.

Domain	Test/Measure
Learning	CVLT-II List A Total Trials 1–5 WMS-IV LM I BVMT-R Total Trials 1–3
Memory	CVLT-II List A Long Delay Free Recall WMS-IV LM II BVMT-R Delayed Recall ROCFT 3-minute Delayed Recall
Attention	WAIS-IV Digit Span, Coding, Letter-Number Sequencing, and Symbol Search TMT A PASAT
Visuospatial	ROCFT copy WAIS-IV Block Design
Executive Functioning	TMT B Verbal Fluency40,41
Motor Speed	Grooved Pegboard42

CVLT-II = California Verbal Learning Test-II⁴³; WMS-IV LM = Wechsler Memory Scale, 4^th edition, Logical Memory⁴⁴; BVMT-R = Brief Visuospatial Memory Test, Revised⁴⁵; ROCFT= Rey-Osterrieth Complex Figure test⁴⁶; WAIS-IV = Wechsler Adult Intelligence Scale – 4^th edition⁴⁷; PASAT = Paced Auditory Serial Attention Test⁴⁸; TMT = Trail Making Test⁴⁹

Telomerase

To determine telomerase activity the telomere repeat amplification protocol (TRAP) was performed as previously described.¹⁵ Values are expressed as total telomerase product generated (TPG) per 10,000 cells.

DNA Damage

DNA damage was determined using the comet assay as reported in Singh et al.⁵⁰ with minor modifications⁵¹, and has been previously described.¹⁵ The comet assay is a single cell gel electrophoresis assay that assesses the extent of DNA damage in nucleated white blood cells (WBC) by applying a computerized scoring algorithm. The extent of DNA damage is estimated from approximately 100 comets per sample analyzed by CASP software (CASP, Wroclaw, Poland)⁵² and values are expressed as a mean score from 0–4 (maximum damage/large tail).⁵³

Telomere Length

Telomere length was determined using real-time quantitative polymerase chain reaction (qPCR) methodology, as described in previously published telomere length protocols,^54–56 and in the previous study.¹⁵ Genomic DNA was extracted from peripheral blood mononuclear cells (PBMC). The telomere inter- and intra-assay coefficient of variation (CV) were all < 5%. Using the standard curve method value were determined for telomere DNA repeat sequences (T) and the HGB single-copy gene (S); Telomere length values are expressed as the T/S ratio.

Circulating Inflammatory Marker: Soluble TNF receptor II (sTNF-RII)

Previously, we reported sTNF-RII to be elevated after treatment, to be associated with cognitive complaints, and to be higher in those with low telomerase activity and high DNA damage.^13,15,34 Thus we sought to determine the association of sTNF-RII with cognitive function in this follow up study several years later. To do this, blood samples were collected at the study visit by venipuncture into EDTA tubes, chilled, and centrifuged for the collection of plasma. Aliquots of plasma were then stored at −80 C until batch testing could be performed on all samples. Soluble tumor necrosis factor (TNF) receptor type II (sTNF-RII) was assessed using enzyme-linked immunosorbent assays (ELISAs; R&D Systems, Minneapolis, MN) per manufacturer’s protocol; lower limits of detection were 234 pg/mL. All samples were run in duplicate with an average intra-assay and inter-assay precision of < 5%.

Statistical Analyses

Analyses were conducted using IBM SPSS software (v.23). Descriptive statistics were calculated using the entire cohort (N=94). Distribution analyses revealed that DNA damage had modest skew and high kurtosis, such that the majority of individuals had low average damage scores. We created an upper quartile cutoff (≥.85) to indicate high damage, and compared this with the lowest 3 quartiles (<.85), indicating low damage. Decile ranking of the telomerase data was performed to address non-normal distribution of the data and these transformed values are used in statistical analyses, while figures display raw telomerase values.

Linear regression models with adjustments for age, BMI, race, and years from last treatment were performed to test the hypothesis that elevated levels of markers of biological aging namely higher DNA damage, reduced telomerase enzymatic activity, shorter peripheral blood mononuclear cell (PBMC) telomere length (TL), and higher soluble Tumor Necrosis Factor Receptor II (sTNF-RII) would be related to worse subjective cognitive impairment and objective neuropsychological function. Models testing predictors of neuropsychological test scores further adjust for WTAR. Linear regression models examined the relationship of predictors DNA damage (highest quartile/lower three quartiles), telomerase enzymatic activity (deciles), telomere length (T/S ratio), and sTNF-RII - with self-reported cognitive functioning and neuropsychological test domains (Betas (β) reflect unstandardized coefficients and standard error (SE), and B reflects standardized coefficients). To minimize the likelihood of false positive observations we corrected for multiple comparisons with the false discovery rate (FDR) procedure,⁵⁷ setting the FDR rate at .05, and calculating the threshold for tests within each cognitive domain. We report uncorrected p-values, and the FDR-corrected q value threshold.

Results

Demographics and clinical characteristics of the study participants are reported in Table 2. Ages ranged from 36 to 69, with a mean of 56.5, with average time since treatment of 4.4 years. In initial models, as was reported previously¹⁵, we examined the association of demographic and clinical characteristics with cellular aging measures. Age was positively associated with DNA damage and inversely associated with telomerase activity, but was not significantly associated with telomere length. Higher BMI was related to shorter telomere length, consistent with previous reports⁵⁸, but unrelated to DNA damage and telomerase activity.

Table 2.

Demographic characteristics of study participants

	Total N=94
Age, mean (SD) [range]	56.5 (8.1) [36.4–69.5]
Race
White, non-Hispanic	75 (80%)
Hispanic	8 (9%)
Black	4 (4%)
Asian	3 (3%)
Other	4 (4%)
Body Mass Index, mean (SD) [range]	25.7(5.1) [18–42.5]
Education
Post college	47 (50%)
College	29 (31%)
No college degree	18 (19%)
Years from treatment, mean (SD) [range]	4.4 (0.6) [3–6.1]
Cancer Treatment Received
Chemotherapy Alone	11 (11.7%)
Radiation Alone	28 (29.8%)
Both Chemotherapy and Radiation	40 (42.6%)
Surgery Alone	15 (16%)
Received Endocrine Therapy	72%
Post-menopausal	80.9%

Self-Reported Cognitive Function

Neither DNA damage, telomerase activity, nor telomere length were related to FACT-Cog scores of Perceived Cognitive Impairment, adjusting for age, BMI, years from last treatment, and race, all P’s>.21. Likewise, sTNF-RII was not a significant predictor of self-reported cognitive impairments (data not shown).

Neuropsychological Domains

On the whole, participants’ neuropsyschological domain scores were normally distributed, with mean scores consistently above zero, indicating generally intact status in most of the sample: Learning $\stackrel{-}{x}$=.46, ±.85; Memory $\stackrel{-}{x}$=.43, ±.74; Attention $\stackrel{-}{x}$=.54, ±.56; Visuospatial $\stackrel{-}{x}$=.11, ±.76; Executive Function $\stackrel{-}{x}$=.43, ±1.12; Motor Speed $\stackrel{-}{x}$=.64, ±1.03. Likewise, very few participants’ domain scores were < −2 z-score (i.e., in the impaired range) – four or fewer in any domain. High DNA damage was associated with having a .23 lower standardized executive function score (P=0.027; q=0.025) compared to low DNA damage, with a similar trend for a relationship of high DNA damage with lower memory scores (P=.06, q=0.013), in adjusted models (See Table 3; Figure 1). Models testing the association of telomerase activity with cognitive domains showed that with each decile decline in telomerase activity there were also a .3 lower standardized attention (P=0.006; q=0.013) and executive function (P=.002;q=0.013) score, and a .24 lower standardized motor speed (P=0.037; q=0.013 See Table 3; Figure 2). Telomere length and sTNF-RII were unrelated to cognitive domains in adjusted regression models (See Table 3).

Table 3.

Multivariate analyses examining aging biology parameters as predictors of neuropsychological test scores, adjusting for age, BMI, race, years from treatment, and intelligence score (WTAR).

	DNA Damage (High vs. Low) N=93			Telomerase (Deciles) N=84			Telomere Length (T/S) N=85			sTNF-RII (pg/mL) N=91
Neuropsychological Test Domain	β(SE)	B	P Value	β(SE)	B	P Value	β(SE)	B	P Value	β(SE)	Beta	P Value
Learning	−.348(.22)	−.18	0.12	−.001(.03)	−.003	0.98	.092(.34)	.029	0.79	.00(.00)	−.08	0.50
Memory	−.371(.20)	−.22	0.06	.019(.03)	.076	0.50	.075(.31)	.027	0.81	.00(.00)	−.05	0.70
Attention	−.231(.15)	−.18	0.12	.058(.02)	.301	0.006	−.155(.23)	−.075	0.50	.00(.00)	−.09	0.45
Visuospatial	−.289(.21)	−.16	0.16	.044(.03)	.162	0.15	.198(.32)	.069	0.54	.00(.00)	−.03	0.78
Executive Function	−.599(.27)	−.23	0.027Ŧ	.118(.04)	.300	0.002	.042(.42)	−.010	0.92	.00(.00)	−.12	0.28
Motor Speed	−.154(.28)	−.07	0.58	.083(.04)	.239	0.037 Ŧ	−.515(.43)	−.134	0.24	.00(.00)	−.11	0.38

Betas (β), unstandardized coefficients; standard error (SE); standardized coefficients (B)

^ŦDenotes criterion for significance not met after correction for multiple testing using False Discovery Rate, q = .025.

Figures 1.

Mean population normalized neuropsychological test score by domain in cancer survivors categorized by low DNA damage and high DNA damage.

Figure 2.

Scatterplot of population normalized attention, executive function, and motor speed scores as a function of telomerase enzymatic activity in breast cancer survivors. Lower scores indicate worse cognitive performance.

Discussion

Cognitive difficulties following cancer treatment are a serious clinical concern and threat to quality of life in cancer survivorship. In the present cross-sectional analyses of a well-characterized cohort of breast cancer survivors several years after completion of treatment, we observed that markers of aging biology including high DNA damage and reduced telomerase activity were associated with worse neuropsychological performance, after covariate adjustments. In particular, higher DNA damage was related to lower executive function scores, and low telomerase activity was related to lower executive function, attention, and motor speed. After adjusting for multiple comparisons, the association of DNA damage with executive function and telomerase with motor speed were no longer considered significant at q=0.025.

Our findings provide evidence of an association of aging biology with cognitive domains commonly affected in both cancer-related cognitive impairment^59–61 and normal aging⁶². The observed relationships are not dramatic; in both the executive function and motor domains the difference between those in the lowest and highest deciles is within one standard deviation. These modest differences in neuropsychological function are consistent with the subtle declines of both age- and cancer-related cognitive changes. Further, even those in the lowest decile are performing in the average range, an unremarkable performance clinically. However, subtle changes in cognitive function are consistent with what is known about cancer-related cognitive difficulties⁵⁹ and can nonetheless result in noticeable changes in daily functioning and quality of life.

In regards to the role of inflammation, our previous work found that sTNF-RII was elevated early after chemotherapy, and was associated with cognitive difficulties.^13,33,34 In the current follow up of 3 to 6 years after treatment, sTNF-RII was unrelated to subjective or objective cognition cross-sectionally; however, we previously reported that both DNA damage and telomerase activity were associated with inflammation.¹⁵ Exposure to elevated inflammation immediately following treatment by chemotherapy and/or radiation could be an early indicator of risk for biological aging, with lasting DNA damage and changes in telomerase activity indicating sustained aging effects. In this regard, several other potential markers of aging might be considered for future research. For example, extensive cell replication cycles or cell stress pathways can induce cellular senescence.^63,64 Future work might consider whether cellular senescence in the periphery and in the brain is associated with cognitive function after treatment. Further research is needed to disentangle the temporal relationship between treatment - inflammation, DNA damage, cell senescence, and cognitive function. As our study is among the first to begin investigating these mechanisms, substantial additional work is needed to address these hypotheses.

Interestingly, no associations were found between cellular aging marker and subjective cognitive functioning similar to the findings of Conroy and colleagues (2013).⁹ Subjective cognitive functioning in cancer survivorship is complex and does not consistently tightly correspond to neuropsychological performance,⁶⁵ but may also be associated with other factors such as mood and stress.⁶⁶ So despite our observed relationships between aging markers and neuropsychological performance, the multi-factorial nature of subjective cognitive functioning may be less strongly linked to specific biological processes. Given the cross sectional nature of this study, it will be important to continue examining these relationships in longitudinal study designs.

Contrary to our hypothesis, PBMC telomere length was unrelated to cognitive function domain scores. A lack of association suggests that this measure of biological age is not necessarily capturing the specific pathways of aging that occur after cancer treatment. Consistent with this, we did not see an effect of cancer treatment on telomere length in our previous report,¹⁵ similar to others.¹⁴ Thus, cancer treatments may not accelerate aging by contributing to blood cell telomere length shortening per se, but rather via induction of DNA damage and cell senescence.¹⁴ Indeed, telomere length shortening is driven by cell replication, a process often halted during cancer treatments.⁶⁷ Cellular senescence, on the other hand, is reached through either cell replication cycles (that shorten telomeres) or cell stress pathways (e.g., extensive damage to DNA). It is also possible that a one-time sampling approach fails to capture temporal dynamics. Further research should consider within-individual changes in telomere length from pre to post treatment as an indicator of cellular aging and assess the extent to which this is predictive of cognitive function.

Regardless of pathway to senescence, senescent cells no longer divide, and express very low levels of telomerase activity. Thus, our measure of telomerase activity may reflect the extent of senescence within PBMCs, regardless of whether it is replicative or stress induced senescence.²¹ Telomerase can also repair DNA damage, and helps resist stress induced growth arrest,^20–23 suggesting the enzyme is important for defense against cell aging independent of its role in protecting telomeric ends.

Our analyses were performed in a cross-sectional sample of women after completing treatment for breast cancer, and therefore causal pathways cannot be confirmed. We recognize that an alternative interpretation could also be made, which is that cognitive function may impact telomerase activity and contribute to DNA damage by modifying behavior. Future research should consider additional assessment of these markers earlier after the course of treatment including before, during, and after cancer treatment, which may yield important information about early indicators of vulnerability to developing cognitive difficulties that can signal the need for intervention or prevention. Additional study limitations include a relatively well-educated sample with generally intact cognitive function, limited diversity in terms of racial/ethnic groups, and a relatively young cohort of women, resulting in a possible sampling bias. The lack of representation of higher risk groups, including older women and a wider range of cognitive impairment, may contribute to smaller detected effects. Future studies in more diverse samples of age, race and ethnicity, and socioeconomic backgrounds are needed. Another limitation is the small sample size that limits our ability to measure smaller effects, raising the possibility of a type II error. Larger sample sized studies are needed to determine whether effects that were subthreshold are detectable with more power. Along this line, the reported medium effects for the associations of DNA damage with executive function and of telomerase with motor speed scores did not survive the FDR correction for multiple tests.

In light of our findings, future work is warranted to further investigate the mechanistic role of aging as a key factor contributing to elevated inflammation and poorer cognitive function, a symptom commonly experienced by cancer patients years after completing treatment. Future research might consider examining aging pathways in current interventions that target cognitive function, diet, and physical activity. Such behaviors are known to modify inflammatory signaling^68–73, telomerase activity⁷⁴, and cellular aging pathways,⁷⁵ but have not yet been tested in the context of cancer and aging. Other possible avenues of research may be manipulating mechanisms of cellular senescence with pharmaceutical agents and employing mouse models^76–78 to characterize plausible targets for intervention.

These results add support for an accelerated aging model of cancer-related cognitive impairment, and harmonize with other studies in non-cancer populations pursuing the link between these markers and cognition.^29,30 Together our findings point to an aging-like effect of cancer treatments on cellular biology and further connect this to cognitive function. An important next research objective will be examining comprehensive and well-powered models that include repeated assessments of neuropsychological function, cellular aging markers, inflammation, and neuro-inflammatory factors.