Circulating Cell-Free Genomic DNA Is Associated with an Increased Risk of Dementia and with Change in Cognitive and Physical Function

Abstract

Background:

Altered cell homeostasis, seen in cognitive decline and frailty, leads to cell death and turnover, releasing circulating cell-free DNA (ccf-DNA).

Objective:

The goal of this study is to determine if serum genomic cell-free DNA (ccf-gDNA) is associated with physical and cognitive decline in older adults.

Methods:

We used serum from 631 community-dwelling individuals from the Religious Orders Study or Rush Memory and Aging Project who were without cognitive impairment at baseline. ccf-gDNA fragments in serum were quantified using digital PCR. An array of cognitive and physical traits, risk of dementia, global cognition, and frailty at or nearest the time of blood draw were regressed on ccf-DNA, with adjustment for age, sex, race, and education.

Results:

Cross-sectionally, higher ccf-gDNA levels were associated with lower global cognition score and slower gait speed at the evaluation nearest to blood draw. Higher ccf-gDNA levels were associated with increased odds of incident dementia (OR 1.27, 95% CI 1.05, 1.54). Longitudinally, higher levels of ccf-gDNA were associated with steeper general cognitive decline and worsening frailty over eight years of follow up.

Conclusion:

This study demonstrates that ccf-gDNA fragments have utility for identifying persons at higher risk of developing dementia and worsening cognition and frailty.

INTRODUCTION

Numerous epidemiologic studies have shown associations between chronic inflammation and Alzheimer’s disease and related dementias (ADRD), and prior work has demonstrated that chronic inflammation beginning in middle age is strongly associated with dementia in later life and can potentially initiate or mediate neurodegenerative processes [1-3].

In addition to chronic inflammation, cell death is elevated in neurodegenerative diseases such as ADRD [4], although it is less clear what specific cell death processes drive the observed neuronal loss [5, 6]. Age-related mitochondrial dysfunction can impair cellular energetics, triggering cell death, and has been observed in both dementia and the geriatric syndrome frailty [7, 8]. Cell death is an important part of regulating genomic instability and preventing the proliferation of cells with oncogenic mutations. Prior work has shown that mechanisms of cell death are closely related to a cell’s energy status, with apoptosis frequently occurring in cells with sufficient adenosine triphosphate (ATP) stores and necrosis occurring more frequently in cells depleted of ATP [9].

One biologic factor that may contribute to chronic inflammation in ADRD is circulating cell free DNA (ccf-DNA). Cell-free DNA is released into the circulation because of cell-intrinsic homeostatic processes such as apoptosis and necrosis that lead to cell turnover and release intracellular DNA into circulation. Genomic ccf-DNA (ccf-gDNA) levels are a surrogate marker of global cell death and are elevated in the setting of conditions such as myocardial infarction and trauma. Additionally, ccf-DNA levels are also associated with all-cause mortality and with frailty [10, 11].

While recent work has focused on associations between methylation patterns on cell free DNA and cognitive testing scores [12, 13], relationships between serum ccf-gDNA and dementia have not been extensively studied, particularly in longitudinally characterized populations.

Cell-free DNA has been detected in numerous body fluids, including blood, cerebrospinal fluid (CSF), and urine, and is hypothesized to be released in circulation via active and passive processes. Cell death processes such as apoptosis and necroptosis are triggered by many endogenous and exogenous stimuli, and intracellular nuclear and mitochondrial DNA are frequently not fully digested, leading to an accumulation of ccf-DNA. Frail older adults have elevated ccf-DNA in serum, which is thought to be released because of increased cellular senescence and catabolic processes, and its presence in circulation is considered a marker of cellular stress in several chronic conditions [10, 11, 14].

Once in circulation, ccf-DNA can function as a damage-associated molecular pathogen (DAMP), which can itself stimulate DNA sensors and the innate immune system, activating proinflammatory and antiviral responses [15].

The objective of this study is to report cross-sectional and longitudinal associations among genome-derived ccf-gDNA, dementia risk, cognitive decline, and physical decline in a cohort of community dwelling older adults. Our hypotheses are that elevated levels of ccf-gDNA are associated with higher risk of dementia and faster decline in cognitive and physical function.

MATERIALS AND METHODS

Study sample

All data were obtained from the Rush Alzheimer’s Disease Center Religious Orders Study or Rush Memory and Aging Project (ROS-MAP), which has been detailed in prior publications [16, 17]. Briefly, ROS began in 1994 and consists of nuns, priests, and brothers from across the USA, while MAP began in 1997 and consists of community-dwelling older adults from the Chicago metropolitan area. Both studies were approved by an Institutional Review Board of Rush University Medical Center. All participants signed an informed consent, Anatomic Gift Act, and a repository consent to allow their resources to be shared. Yearly assessments are conducted for all participants. Serum samples were obtained from ROS-MAPparticipants during cohort evaluations; we included here 631 samples which were available for analysis at the time of this study (2019–2020). At the study visit from which serum samples were obtained, participants had no history or dementia or MCI.

Biological measures

Genomic ccf-DNA was measured directly in serum using digital PCR. Primers were designed targeting a conserved region of the Ribonuclease P RNA Component H1 (RPPH1) gene, which has been used in previous ccf-DNA experiments as it is readily detectable in plasma and serum and relatively stable [18]. Serum was pre-treated with PBS and heat denatured for 5 min at 90°C. The samples were vortexed until the solid pellet broke apart, approximately 3 s, and then spun down in a microcentrifuge for 15 s to pellet any viscous material.

Master mixes was prepared using the Luna Universal Probe Master Mix, 22% TWEEN and water. Custom assay was ordered from Integrated DNA Technologies (Iowa City IA). 16.4 μL of Master mix was mixed with 4 μL of the prepared serum and the entire volume was loaded into the digital PCR plate and run on the Constellation digital PCR system (Formulatrix, Bedford MA) using the Genomic DNA thermocycling conditions [98°C for 120 s, 40 cycles of {98°C for 10 s, 60°C for 30 s}, 45°C for 30 s]. Amplification thresholds were applied using the Constellation software (Formulatrix, Bedford, MA).

Cognitive measures

A battery of 19 cognitive tests were conducted at each study visit to assess the following cognitive domains: episodic memory, semantic memory, working memory, perceptual orientation, and perceptual speed. A z-score for each test was constructed using all participants, and each cognitive domain z-score was calculated by averaging all z-scores within that domain. A global cognitive function t-score was calculated by averaging the z-score of each of the 19 cognitive tests [19], and standardized to a t-score scale (mean 50, standard deviation 10).

Dementia status was assessed at each study visit using a three-stage process consisting of computer scoring of cognitive tests, clinical judgment by a neuropsychologist, and diagnostic classification by a clinician [20, 21]. Participants were diagnosed based on criteria of the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer’s Disease and Related Disorders Association (NINCDS/ADRDA) as mild cognitive impairment (MCI), and dementia and its causes, especially Alzheimer’s disease dementia. Those without cognitive impairment (CI), i.e., no dementia or MCI, were designated as no cognitive impairment (NCI) [22]. Clinical diagnoses were performed each year shielding examiners from prior years. Clinical diagnosis at time of death was determined without postmortem data [23]. Clinical diagnosis at final follow-up (last available study visit) was used to categorize participants as non-progressors (remained cognitively normal), MCI, and ADRD.

Physical measures

Physical measures (grip strength, gait speed, body mass index, and fatigue) were obtained at each study visit to create a frailty composite z-score. The continuous frailty z score used in the ROSMAP study has previously been compared to a discretized frailty measure, corresponding to the Frailty Phenotype developed by Fried [24] and both methods of measuring frailty demonstrated significant hazard ratios for incident AD [25]. With regards to the frailty z-score using BMI instead of weight loss, as is used in the traditional Frailty Phenotype method, BMI has been shown to have a strong correlation with frailty [26]. The 2007 Buchman paper also compared the frailty z-score utilizing BMI to the cut point approach using BMI<20 and found similar findings [25]. Grip strength was measured using the Jamar hydraulic hand dynamometer (range 0–200 kg) and calculated using the average of four trials [27]. Gait speed was defined as the time it took a participant to walk 8 feet (2.4 m) at their usual pace, measured using a stopwatch (seconds). Body mass index (BMI) was calculated using weight in kilograms divided by height in meters squared. Fatigue was assessed using two items from the CESD questionnaire [28]: “Everything I did was an effort” and “I could not get going” by summing up the total number of positive responses, giving a total score range of 0–2. Z-scores were calculated for grip strength, gait speed, BMI, and fatigue using all participants at baseline. These z-scores were then combined to create a frailty composite z-score, where larger values indicate greater levels of frailty [25, 29]. Due to the presence of missing data for the frailty z-score at the visit where cell-free DNA was measured, an available frailty score from the preceding or following 2 study visits was used.

Additional physical measures were collected to assess motor function [30, 31]. A global motor function composite z-score was constructed using 10 sub-tests: Purdue Pegboard Test (number of pegs), finger-tapping test (taps/10 s), gait speed, number of steps to cover 8 feet, 360 degree turn time (s), number of steps to complete a 360 degree turn, leg stand (s), toe stand (s), grip strength, and pinch strength (kg).

Adjustment variables

Analyses were adjusted for age, sex, race and years of education. Race was self-reported as White, Black or African American, American Indian or Alaska Native, Native Hawaiian or Other Pacific Islander, Asian, Other, or Unknown. A binary race variable was constructed and defined as 0 = White and 1 = any other race; this binary race variable was used in all analyses since there were few participants who reported race other than White or African American.

Analyses

We first described the sample using descriptive statistics. Cognitive and physical measures were regressed on ccf-DNA, with adjustment for age, sex, race, and years of education. We regressed global cognition and frailty, as well as domain-specific cognitive indices (episodic memory, semantic memory, working memory, perceptual speed, and perceptual orientation) and other physical outcomes (grip strength, gait speed, motor composite score), on ccf-gDNA using linear regression. We regressed MCI and dementia on ccf-gDNA using logistic regression. We evaluated associations of ccf-gDNA with changes in cognitive (global, episodic memory, semantic memory, working memory, perceptual speed, and perceptual orientation) and physical outcomes (frailty, gait speed, grip strength, hand strength, and motor composite score) using mixed effects models with random effects for participants and time [32]. Analyses were performed in Stata version 15.1. This study was approved by the Johns Hopkins University Institutional Review Board.

RESULTS

Demographic and clinical characteristics of the 631 study participants are in Table 1. The average age of participants at time of blood collection was 80 years (range 54 to 93), 80% were female, 75% were White, and on average had completed post-secondary education (16 years) (Table 1). Circulating cell-free genomic DNA (ccf-gDNA) ranged from 0 to 150 copies per microliter with a median value of 20 copies per microliter. Missing data was most common for physical assessment measures (Table 1).

Table 1.

Descriptive data of participants^a

Characteristic	Baseline	Clinical Status at Final Visit (N = 631
Demographics	NCI (N = 631)	Non-progressor	MCIb	ADRDc
Participants, N (%)	631	362 (57.4%)	135 (21.4%)	134 (21.2%)
Age, mean (SD)	79.9 (7.0)	79.0 (7.4)	79.9 (6.8)	82.3 (5.6)
Sex (Female), N (%)	509 (80.7%)	297 (82.0%)	103 (76.3%)	109 (81.3%)
Race (white), N (%)	475 (75.3%)	266 (73.5%)	95 (70.4%)	114 (85.1%)
Years of Education, mean (SD)	15.8 (3.5)	15.6 (3.5)	15.8 (3.6)	16.3 (3.3)
Years of follow-up, median (range)		7 (0–19)	8 (2–14)	9 (1–16)
Cognitive testing
Global cognitive function (t-score)	52.1 (9.6)	54.6 (8.8)	49.7 (9.1)	47.9 (9.8)
Physical measures
Frailty z-score, mean (SD) d	−0.10 (0.52)	−0.19 (0.50)	−0.01 (0.55)	−0.00 (0.51)
Gait speed (m/s)	0.65 (0.20)	0.67 (0.19)	0.61 (0.18)	0.63 (0.21)
ccf-gDNA (copies/ul)	21.4 (14.1)	20.6 (13.5)	20.5 (12.1)	24.4 (17.1)

ADRD, Alzheimer disease related dementias; ccf-gDNA, circulating cell-free genomic DNA; MCI, mild cognitive impairment; NCI, no cognitive impairment.

^aMissing data as follows: Overall, 4/631 (0.63%) were missing cognitive status at last visit and could not be categorized into non-progressor, progression to MCI, or progression to dementia. One person in the ADRD group was missing education. Global cognition: NCI missing 37 (10.22%); MCI missing 15 (11.03%); ADRD missing 9 (6.67%). Frailty: NCI missing 78 (21.55%); MCI missing 25 (18.38%); ADRD missing 14 (10.37%). Gait speed: NCI missing 75 (20.72%); MCI missing 26 (19.12%); ADRD missing 19 (14.07%). Two individuals in the NCI group were missing ccf-gDNA measurements and were not included in further analyses.

^bMCI category includes individuals with MCI alone and MCI plus another condition contributing to cognitive impairment.

^cADRD category includes individuals with Alzheimer’s disease dementia alone as well as Alzheimer’s disease dementia plus another condition contributing to cognitive impairment.

^dFrailty is composite variable consisting of grip strength, timed walk, body mass index, and fatigue. Larger values correspond to higher frailty.

Cross-sectional associations with ccf-gDNA and cognitive and physical outcomes

Unadjusted analyses did not show associations between ccf-gDNA and cognitive and physical outcomes (Supplementary Table 1). After adjustment for age, sex, race, and education, ccf-gDNA was associated with global cognition (β=−14; 95% CI: −21, −7.4) and frailty (β=0.3; 95% CI: 0.03, 0.6) (Table 2). There were no associations with individual cognitive domains or other physical outcomes.

Table 2.

Adjusted cross-sectional associations of ccf-gDNA with cognition score and frailty^a

Outcomes	CCF-gDNA Fragment β (95% CI)
Global cognitionb	−14 (−21, −7.4)
Frailtyc	0.3 (0.03, 0.6)

CI, confidence interval; ccf-gDNA, circulating cell-free genomic DNA.

^aRegression of ccf-gDNA on cognitive and physical outcomes. Adjusted for race, sex, age, education. Coefficients represent the difference in the outcome per 100 copies per uL difference in the ccf-gDNA predictor.

^bGlobal cognition score is a composite of 19 cognitive tests. Larger values correspond to better scores on tests.

^cFrailty is composite variable consisting of grip strength, timed walk, body mass index and fatigue. Larger values correspond to higher frailty.

Longitudinal associations with ccf-gDNA and physical and cognitive measures

The odds ratio for development of dementia was 1.27 (95% CI: 1.05, 1.54), indicating that individuals with higher levels of ccf-gDNA had 27% higher hazard of developing dementia (Table 3). We next examined associations between ccf-gDNA levels and slopes of cognitive and physical decline (Table 4). After demographic adjustment, elevated levels of ccf-gDNA were associated with steeper declines in global cognition (β=−0.11; 95% CI: −0.19, −0.03) as well as worsening frailty (β=0.13; 95% CI: 0.01, 0.25). Among cognitive outcomes, elevated levels of ccf-gDNA were associated with steeper declines in working memory (β=−0.16; 95% CI: −0.28, −0.04), episodic memory (β=−0.13; 95% CI: −0.23, −0.03), and perceptual speed (β=−0.10; 95% CI: −0.20, −0.002). Among physical outcomes, elevated levels of ccf-gDNA were also associated with steeper declines in grip strength (β=−0.18; 95% CI: −0.30, −0.06).

Table 3.

ccf-gDNA as predictor of final cognitive status

	Odds Ratio (95% CI)
Mild cognitive impairment	1.10 (0.85, 1.41)
Dementia	1.27 (1.05, 1.54)

Coefficients represent odds ratios for the risk of cognitive status (MCI or dementia) per unit difference in the winsorized ccf-gDNA predictor (standardized to mean 0, variance 1)

Table 4.

Unadjusted and adjusted longitudinal changes in cognitive and motor outcomes relative to baseline ccf-gDNA

Outcome	Slope on ccf-gDNA
	Unadjusted β (95% CI)	Adjusted β (95% CI)
Global cognition	−0.07 (−0.15, 0.01)	−0.11* (−0.19, −0.03)
Working memory	−0.12* (−0.24, 0.00)	−0.16* (−0.28, −0.04)
Episodic memory	−0.10* (−0.20, 0.00)	−0.13* (−0.23, −0.03)
Semantic memory	−0.05 (−0.15, 0.05)	−0.07 (−0.17, 0.03)
Perceptual speed	−0.07 (−0.17, 0.03)	−0.10* (−0.20, 0.00)
Perceptual orientation	−0.07 (−0.19, 0.05)	−0.08 (−0.20, 0.04)
Frailtya	0.10 (−0.02, 0.22)	0.13* (0.01, 0.25)
Gait speed (m/s)	−0.01 (−0.26, 0.24)	0.02 (−0.23, 0.27)
Grip strength	−0.21* (−0.35, −0.07)	−0.18* (−0.30, −0.06)
Motor composite scoreb	0.02 (−0.12, 0.16)	0.03 (−0.11, 0.17)

^aFrailty is composite variable consisting of grip strength, timed walk, body mass index and fatigue. Larger values correspond to higher frailty.

^bMotor composite score is calculated using the following items: Purdue Pegboard Test, Finger-tapping test, Time to cover 8 feet, Number of steps required to cover 8 feet, 360 degree turn time, Number of steps to complete a 360 degree turn, Leg stand, Toe stand, Grip strength, Pinch strength. Model-estimated slope parameters from latent growth models of each outcome domain regressed on baseline ccf-DNA levels. Adjusted for race, sex, age, education.

^*indicates p < 0.05

DISCUSSION

We found that elevated baseline ccf-DNA measurements in NCI individuals are associated with lower cognition and higher frailty, and with risk of incident dementia, faster cognitive decline, and progression of frailty. Our longitudinal analysis demonstrated that for every 100 copies/μl increase in ccf-gDNA, there was a steeper decrease in global cognitive score of 0.11 points per year. For every 100 copies/μl increase in ccf-gDNA, there was a steeper increase in frailty z-score per year of 0.13 points per year. This study is a first step towards characterizing the role of ccf-gDNA as a blood-based biomarker of cognitive decline and progression of frailty and further underscores the importance of inflammation in the geriatric syndromes of dementia and frailty.

Numerous previous studies have demonstrated links between frailty and cognitive decline, including studies of the ROS-MAP cohort [25, 33, 34]. Recently, a study using UK Biobank data demonstrated that frail individuals developed dementia at twice the rate of non-frail controls [35]. Both geriatric syndromes are characterized by chronic inflammation and this study proposes that a common mediator of this inflammation may be ccf-gDNA.

Cell-free DNA has shown great promise in oncology to contribute to personalized medicine in its ability to identify individuals who may benefit from specific therapeutic regimens or predict progression free survival [36, 37]. Similarly, development of ccf-gDNA as a non-invasive biomarker associated with future cognitive or physical decline has the potential to identify individuals who will benefit from targeted preventative measures.

Concentration of ccf-gDNA in our cohort is higher than what is observed in other studies using similar quantification methods (mean = 21.4 copies/μl=21,400 copies/mL), though a factor that may be playing a role in this increase is that this study has older participants compared to the other studies and ccf-gDNA is known to increase with age [10]. Healthy controls with a mean age of 45 years old in the Madsen et al. paper showed a mean concentration of ccf-gDNA of 1300 copies/ml [38], while the Devonshire et al. study measuring RPPH1 in plasma from healthy adults (mean age 53) showed ccf-gDNA levels ranging from 800 to 1600 copies/ml [18]. In previous studies examining ccf-DNA in older adults, a DNA purification step is commonly included in the methods to remove proteins and other circulating factors that can inhibit DNA quantification, and this purification step may inadvertently bias the results and analysis due to column factors that can lead to loss of cell free DNA (pH, buffers, etc.).

This work expands on prior dementia-related studies examining relationships between cell-free DNA and cognition by utilizing a dataset that not only has a large study sample but is well phenotyped and characterized with regards to cognitive and physical function. Although there have been inconsistent results regarding associations between total cell-free DNA content and cognitive test scores, with the study by Pai et al. [39] showing increased ccf-DNA associated with poor test scores and the pilot study by Konki et al. [12] showing no significant associations, our study is concordant with the findings from Pai et al. Differences in sample size and study design regarding cognitive outcome measures and timing of serum measurements with respect to trajectory of cognitive decline may have contributed to these apparent inconsistencies.

Additionally, this study demonstrates associations between longitudinal trajectory and cognitive and physical decline using easy to obtain peripheral blood as opposed to prior work focused on measuring ccf-DNA in CSF [40]. Furthermore, our study demonstrates feasibility of directly quantifying cell-free DNA in serum without the need for purification steps.

Interestingly, gait speed in the study cohort was significantly lower than 1 m/s, which is typically considered a slow speed and is associated with longitudinal development of frailty and cognitive decline [41]. However, the average gait speeds measured in in this study are consistent with prior reports demonstrating slower gait speeds in older women and greater frequency of gait speeds ranging from 0.6–0.8 m/s in adults over age 65 [42]. In the current study, 80% of the participants were women. Prior studies have used 0.6 m/s as gait speed cut off indicating absence of overt dysmotility in older adults [43].

A consideration when examining our findings is that the population studied is predominantly female and White, which may limit generalizability. However, follow up studies validating these findings in other samples can further inform how ccf-gDNA measurements can be utilized within the field of precision medicine, as a panel of other blood-based biomarkers identifying individuals at highest risk of developing incident dementia and frailty. In addition, the variable sample processing over the course of the ROSMAP study, though comparable to sample processing in other cell free DNA studies, may result in genomic DNA contamination, however future studies utilizing specialized blood collection tubes for cell-free DNA analysis can mitigate this effect. The study also has important strengths. The extraordinarily high follow-up rates ensure high internal validity of associations.

Next steps are to identify tissue origin of ccf-gDNA in the ROS-MAP population using known methylation signatures from genomic DNA, which can inform on the specific tissues that are undergoing increased cell death and homeostatic processes and examine associations with ADRD and frailty. Also, associations between ccf-gDNA and plasma ADRD biomarkers, specifically subtypes of phosphorylated tau and amyloid-β, can be examined to identify whether prediction of cognitive trajectory can be further improved when these two blood-based markers are studied in conjunction. Further work will examine relationships between ccf-DNA from serum and CSF and to assess whether similar associations are observed with cell-free DNA of CSF origin. These findings would further demonstrate the utility of measuring ccf-DNA in blood as a non-invasive substitute for measuring cell-free DNA in CSF and identifying individuals at elevated risk for dementia. Additionally, given the role of ccf-DNA as a pro-inflammatory DAMP, future studies will examine mechanisms by which ccf-gDNA contributes to chronic inflammation, potentiating the inflammatory pathology seen in ADRD.