Association of Short-term Change in Leukocyte Telomere Length With Cortical Thickness and Outcomes of Mental Training Among Healthy Adults

Key Points Question Is naturally occurring short-term change in leukocyte telomere length related to structural plasticity of the brain, and can telomere length be influenced through mental training? Findings In this randomized clinical trial of 298 healthy adults, the mental training intervention did not affect leukocyte telomere length. Naturally occurring change in leukocyte telomere length over 3 consecutive 3-month intervals was significantly associated with cortical thickness change in the left precuneus extending to the posterior cingulate cortex. Meaning This study provides the first evidence to date for an association between short-term change in leukocyte telomere length and brain structure, suggesting that these processes may be mechanistically linked; the mental training used did not influence leukocyte telomere length of healthy, middle-aged adults.


Introduction
The length of telomeres (ie, protective chromosomal caps) functions as a biomarker for an individual's health and aging trajectory. Shortened leukocyte telomere length (LTL) has, for example, been associated with greater susceptibility to age-related diseases, including mild cognitive impairment and Alzheimer disease. [1][2][3][4][5][6] A large-scale, cross-sectional investigation 7 also identified correlations between shorter LTL and smaller volumes of several brain regions associated with the development of Alzheimer disease, including the hippocampus, amygdala, temporal lobe, precuneus, and posterior cingulate. 8,9 Cross-sectional evidence therefore suggests that shorter telomeres are associated with neurodegenerative processes. Less is known, however, about how change in telomere length relates to brain structure.
Shorter telomeres have been associated with genetics 10 and lifestyle factors, such as obesity, 11 but also with psychological and emotional strains, including early life adversity, chronic stress, rumination, and loneliness. [12][13][14][15][16][17] Leukocyte telomere length has therefore been described as a psychobiomarker that reflects the combined result of physiological and psychological burdens on an individual's health and aging profile. 18 Recent evidence suggests that LTL may change more quickly and dynamically than previously assumed. Leukocyte telomere lengthening, for example, has been observed after 6 months of physical training programs, 19,20 as well as after an intensive 1-month mental training intervention. 21 Because telomere lengthening implies a reversal of biological aging processes, these observations have received much attention. However, the biological plausibility of short-term telomere change, herein defined as less than 2 years, remains controversial, particularly for lengthening. [22][23][24] Relating change in LTL with brain structure may provide insight into the biological implications of short-term LTL change. If short-term LTL change reflects biological processes that are generally meaningful for an individual's aging trajectory, such change is unlikely to happen in isolation. Rather, LTL change should be associated with changes in other aging-and health-related markers. Structural brain indices are biomarkers of individual differences in aging and health. 25 The first aim of the present study was therefore to investigate whether naturally occurring aging-or lifestyle-related change in LTL over 9 months was related to structural changes in the brain. Cortical thickness (CT) was selected as our measure of brain structure as a more anatomically specific modality than, for example, volumetric measures, 26,27 and was accordingly expected to be more sensitive to structural changes, including aging-related gray-matter decline. 27,28 The second aim of this study was to assess whether training in different mental practices over the same 9-month period could systematically influence LTL, potentially buffering against agingrelated shortening or facilitating lengthening. Mental training protocols, such as the mindfulnessbased stress reduction program, 29 have been found to reduce several psychological strains that are associated with shorter telomeres, including rumination, loneliness, and stress. 30,31 However, of the 9 studies 21,32-39 that have investigated LTL in association with mindfulness or meditation, as discussed in a review, 40 only 2 found evidence for a change in LTL. One study observed LTL after a 1-month retreat (N = 28), 21 and the other after a 5-year lifestyle intervention (N = 10). 32 The remaining 7 studies used less-intensive or shorter interventions, which may in part explain the absence of LTL change. [33][34][35][36][37][38][39] Evidence for an effect of mental training on LTL therefore appears preliminary and requires replication on a larger scale, which was possible in the present study.
The present investigation was conducted as part of the ReSource Project, 41 a longitudinal mental training study that included training cohorts (TCs) and a retest control cohort (RCC). The RCC was used to address our first aim: to assess potential dynamic associations between LTL and CT. We expected associations with CT change in brain regions previously linked cross-sectionally to LTL. 7 Nonetheless, we conducted analyses on a whole brain level to be able to detect potential aging-or lifestyle-related CT changes more broadly. The main effects of the ReSource Project training on CT have been reported elsewhere. 42 Our second aim focused on a nonoverlapping sample, namely, participants trained in 3 distinct types of mental practices (TCs). These practices were designed to cultivate attention, interoception, and focus on the present moment (Presence); socioaffective capacities, such as compassion (Affect); and sociocognitive skills, such as meta-cognition (Perspective) (Figure 1). 43,44 All 3 training modules had the potential to influence LTL by buffering the experience of acute or chronic stress. 31,45 The Presence module could additionally have influenced LTL by reducing rumination, 46 whereas the Affect and Perspective modules may have had an effect through reduced loneliness or social  A, Key concepts and core exercises taught during the modules Presence (yellow), Affect (red), and Perspective (green). B, Timeline of the ReSource Project and training sequence per cohort. Retest control participants were recruited in 2 cohorts for logistic reasons but were analyzed jointly. We therefore refer to a single retest control cohort in the text. The displayed study timeline was adapted to most accurately reflect the time points of blood sampling. Test phases for other variables may differ slightly. Samples of retest control cohort I were collected after approximately 2 months of no training before T1 and T2; however, given the smaller sample size of this cohort compared with retest control cohort II, combined RCC sampling occurred approximately every 3 months on average; the same is true for magnetic resonance imaging scans. The full ReSource Project design also included follow-up assessments (T4), but these were not included in the present investigation. Adapted from Singer et al. 43,44 isolation. 42,47 We thus predicted leukocyte telomere lengthening or relative maintenance in the TCs compared with the RCC, which, on average, was expected to show aging-related attrition or no change in LTL. In the case of significant training-induced differences in LTL change, we had planned to subsequently analyze how this systematically induced change relates to structural plasticity of regions associated with LTL in the first aim and exploratively at the whole-brain level.

Intervention Design
Data were acquired from participants in the RCC 4 times in intervals of approximately 3 months. The TC participants were tested at similar intervals in which data acquisition followed training in 1 of 3 distinct modules termed Presence, Affect, and Perspective. Training cohorts 1 and 2 practiced the 3 modules in counterbalanced order to account for potential sequence effects. Training cohort 3 completed only the 3-month Affect training to isolate the specific effects of the Presence module from the Affect module ( Figure 1B). Further details have been reported elsewhere. 43  The detailed trial protocol (Supplement 1) has also been published elsewhere. 44

Sampling of LTL
Blood was collected at 4 points (T0-T3) and frozen at −80°C until assay. Genomic DNA was extracted from whole blood (QIAamp DNA blood mini kit; Qiagen). Leukocyte telomere length was measured using quantitative polymerase chain reaction assay as previously described, 52 and reported as T/S ratios (ie, the relative ratio of telomere repeat copy number to single-copy gene; interassay coefficient of variability, 2.3%). All DNA samples were run twice, each with triplicate wells. If the difference between the 2 runs was greater than 7%, samples were run a third time and the average of the 2 closest values was used for data analysis. DNA samples from the same participant were assayed as 1 batch, using the same reagent lots, and run on the same assay plate.

Magnetic Resonance Imaging Acquisition
Magnetic resonance imaging (MRI) acquisition was carried out on a 3-T scanner (Siemens Verio; Siemens) with a 32-channel head coil. T1-weighted images were acquired using a 3-dimensional

Processing of Structural Data
The T1-weighted MRIs were processed using FreeSurfer, version 5.1.0, to generate cortical surface models for measurements of CT. 53 We chose the most general cross-sectional image processing procedure to enable baseline data analysis for cross-sectional study goals of the ReSource Project (eg, as used by Valk et al 54 ) before the completion of data acquisition, which spanned more than 2 years. The exact processing steps of cortical reconstruction and volumetric segmentation are described in previous publications. [55][56][57] Briefly, T1-weighted images were intensity normalized and skull stripped, followed by a tessellation of the gray matter/white matter cortical boundary and automated topologic correction. Surface deformation was performed along intensity gradients, placing borders of the inner (gray matter/white matter) and outer (gray matter/cerebrospinal fluid) cortical interfaces at the location where the greatest shift in intensity defined the transition to the other tissue class. Cortical thickness was calculated as the shortest distance from the gray matter/ white matter boundary to the gray matter/cerebrospinal fluid boundary at each vertex on the tessellated surface and is reported in millimeters. 58

Preprocessing of Dependent Variables
Change in LTL was calculated as difference scores between each participant's measurements from a set of consecutive testing times (ie, T1 minus T0, T2 minus T1, and T3 minus T2). Specifically, we calculated LTL change as the Verhulst D value 59 (herein termed DLTL) to correct for regression to the mean instead of the common practice of controlling for baseline LTL because it has recently been suggested that the latter approach inflates type I error rates. 60 For statistical analysis of CT, we generated participant-specific CT change maps (ΔCT) by subtracting vertexwise thickness maps of subsequent measurement times. Thickness data at each vertex were normalized before change calculation by regressing out the effects of global thickness to emphasize relative region-specific change patterns.
All difference scores diverging more than 3 SDs from the sample mean difference score were defined as outliers. There were no outliers in the ΔCT data analyzed in the present study. Seven DLTL outliers were winsorized to the respective upper or lower boundaries of 3 SD. In addition, MRI data were excluded if they did not pass quality control by 2 independent expert raters (S.L.V. and B.C.B.) owing to excessive movement or artifacts in the T1-weighted MRI images (eTable 5 and eTable 6 in Supplement 2). Five scans were excluded because of low image quality.

Statistical Analysis
Data analysis was conducted between September 23, 2016, and June 21, 2019. All longitudinal analyses were performed using multivariate linear mixed models, which are robust to unbalanced and incomplete data in longitudinal designs, 61 allowing the inclusion of all eligible data points for a given analysis. Body mass index, age, and sex were selected as covariates owing to their established association with LTL 11,62,63 and brain structure. 64 We also included the variable time point to control for potential effects of the time of measurement on ΔCT. Detailed model descriptions can be found in eAppendix 1 in Supplement 2. All analyses were conducted with an α threshold of .05 or less.
The whole-brain, linear mixed-model analysis of CT data was performed using SurfStat for Matlab, 65

Participants
Sample sizes and reasons for missing cases of all variables relevant to the current investigation are described in Figure 2 as well as eTable 5 and eTable 6 in Supplement 2. Of the 362 individuals randomly assigned to TCs or RCC, 30 participants dropped out before study initiation (initial sample, 332  Figure 3A). In repeating the analysis with the added variables age, body mass index, and sex, the same region was identified (mean t 161 = 3.22; P < .001; r = 0.246; 61% overlap with precuneus, 24% overlap with PCC).
In subsequent follow-up cluster-level tests, the association between DLTL and ΔCT in the identified region was also significant when analyzed with an additional random slope term, as well as when controlling for change in leukocyte cell type counts (eAppendix 3 and eTable 1 in Supplement 2). Further follow-up analyses assessed the consistency of the observed phenomenon by testing whether the association between DLTL and ΔCT was reliable across the 3 intervals of change. Linear models showed that, when analyzed separately, change in LTL and CT were significantly associated at 2 of the 3 intervals (T0-T1: t 54 = 2.60; P = .006; r = 0.334; T1-T2:   Figure 3B). In an exploratory analysis of change from baseline to 9 months, however, we did not find a significant association (t 56 = 0.46; P = .31; r = 0.06) (eFigure 1 in Supplement 2).
Leukocyte telomere shortening and lengthening manifested themselves in a similar number of observations ( Figure 3A, scatterplot). Examining these separately in follow-up analyses revealed that LTL change was positively associated with ΔCT regardless of direction, that is, leukocyte telomere shortening with a greater tendency for cortical thinning (t 77 = 2.38; P = .01; r = 0.262), and lengthening with a greater tendency for cortical thickening (t 77 = 2.42; P = .009; r = 0.266) ( Figure 3C). Corresponding to the observed association of DLTL with ΔCT in the left hemisphere, exploratory analyses identified a contralateral subthreshold association with ΔCT in the right precuneus/PCC (eAppendix 4 and eFigure 3 in Supplement 2).

Mental Training Intervention
Likelihood ratio tests showed no significant main effect of the training module by time interaction (χ 2 = 3.26 3 ; P = .35) or module alone on (χ 2 = 2.20 3 ; P = .53) DLTL (Figure 4A and B). We could therefore not pursue the hypothesis that systematic training-induced differences in LTL change would be associated with changes in brain structure. Estimated mean change in T/S ratios in model As a follow-up analysis, we explored whether the combined 9-month training period affected DLTL; this was not the case (F 1,196 = 1.88; P = .17) ( Figure 4C; eAppendix 1 in Supplement 2).
Additional exploratory analyses showed that the effect of training remained nonsignificant even when considering practice frequency or participant personality or when focusing on withinparticipant differences (eAppendix 4, eTable 2, and eTable 3 in Supplement 2). Checking the residuals of the above models through visual inspection and diagnostic tests did not reveal any evidence for influential cases, systematic error, or multicollinearity (eAppendix 5 in Supplement 2).

Discussion
With the present investigation, we sought to improve our understanding of the neural and psychological processes associated with short-term change in LTL through 2 steps. We first related DLTL to structural brain changes over 3 consecutive 3-month intervals in an RCC. Second, we assessed the effect of 3 distinct, 3-month mental training modules on DLTL.
Repeated measures of DLTL in RCC participants were associated with ΔCT in the left precuneus/ PCC. When analyzed separately, the association was significant over 2 of the 3 consecutive, 3-month  intervals. Precuneus and PCC were previously shown to be linked to LTL in a large, cross-sectional, population-based study. 7 The longitudinal association between short-term changes in LTL and structural indices of the left precuneus/PCC noted in our study suggests a dynamic association between the 2 phenomena.
To our knowledge, there is only 1 other study associating change in LTL and brain structure.
Leukocyte telomere shortening and loss of structural integrity were found to correlate in older adults at a single, 2.9-year follow-up measurement. 72 In contrast, we observed similar rates of leukocyte telomere shortening and lengthening. Shortening was related to a greater tendency for cortical thinning and lengthening was related to a greater tendency for thickening, even when analyzed separately. The identified association was therefore not predominantly driven by aging-related decline. In an exploratory analysis, DLTL from baseline to the 9-month follow-up measurement showed no association with ΔCT, and we cannot say with certainty why this occurred. Our results may, however, indicate that short-term LTL change does not follow a continuous trajectory but rather represents transient change.
Telomere fluctuations that average out over time have previously been observed, 73 but there is currently no model for the potential underlying biological processes. One mechanism that could be responsible for the associated changes in LTL and CT may be the action of the cellular enzyme telomerase. Telomerase counteracts telomere shortening by replenishing telomeric DNA during cell division. 22,74 Increases in telomerase activity have been associated with short-term telomere lengthening, most notably following 6-month physical training protocols, 20,75 and may also have mediated the lengthening observed in this study. Likewise, cortical thickening could be facilitated by an increased telomerase-associated proliferative capacity of critical support cells in the brain capable of undergoing mitosis in the adult, resulting in greater tissue volume, as previously proposed. 7 Heightened telomerase activity may therefore lead to associated increases in LTL and CT, assuming that central and peripheral telomerase activity are related. In line with this reasoning, telomerase activity was previously found to correlate positively with hippocampal volume, although not with LTL, in a small sample of patients with major depressive disorder. 76 While it is unclear which neurobiological mechanisms exactly underlie cortical thickening in the age group evaluated herein, these authors described the potential gliogenesis-enhancing and neuroprotective effects of telomerase in more detail. Concomitant telomere shortening and cortical thinning, if not reflective of aging-related decline, could similarly result from accelerated telomere attrition and decreased cellular proliferation under lowered telomerase activity. In the present study, we could not assess whether such fluctuations in telomerase activity occurred and, if so, whether they result from endogenous processes or external influences, such as the lifestyle factors stress or physical exercise. 20,77 The same lifestyle variables have also been directly associated with differences in telomere length 15,20 and brain structure, 78,79 and therefore provide an explanatory framework for the observed concomitant change that may not involve mediation through telomerase activity.
The present study identified significantly less widespread association between change in LTL and brain structure than was reported by King and colleagues. 7 Likely reasons for this divergence include King and colleagues' 7 region of interest-based, cross-sectional study design compared with our whole-brain longitudinal analysis approach, as well as their older and considerably larger sample (N = 1960), which sensitized their analyses to associations with very small effect sizes of R 2 less than 1%. Owing to our comparatively smaller participant sample, the present study was not equipped to detect similarly sized effects. Nonetheless, a specific association between LTL change and thickness change in the left precuneus/PCC was identified. The precuneus/PCC region is a prominent metabolic hub and central node of the default mode network. 80 The observed change in CT may thus, for example, reflect a particularly strong susceptibility to telomerase-related synaptic plasticity grounded in this region's unique metabolic and connectional properties. Moreover, it is notable that precuneus/PCC structure and function are impaired in age-related neurodegenerative disorders, such as Alzheimer disease. 8,9,81 In cross-sectional studies, shorter telomeres have been associated with similar age-related conditions. 3-6 The specific coupling of short-term LTL change with ΔCT in the Contrary to our predictions, we found no effect of contemplative mental training on change in LTL. This result contrasts with the findings of 2 preceding interventions that identified telomere lengthening in relation to meditation. 21 The first study found telomere lengthening in men with low-risk prostate cancer following a relatively broad, 5-year lifestyle intervention. 32 The differing participant population and intervention design may account for the diverging outcomes. The other study found telomere lengthening in healthy, middle-aged participants after a short, but intense, 1-month residential training retreat, which may have unique benefits. 31 In contrast, our results provide what we believe to be the largest body of evidence that longitudinal contemplative mental training does not systematically lengthen leukocyte telomeres of healthy adults. Because we detected no mean attrition in the RCC over the entire 9-month study period, we cannot make any conclusions regarding potential relative maintenance of telomere length through mental training.
Effects of the ReSource Project training modules on CT alone have been reported. 42 Because we found evidence for an association between CT and LTL on the individual participant level, but no effect of training on LTL, it appears that the mechanism underlying the herein observed association is independent of the capacities trained during the intervention. Other work showed that the ReSource Project intervention reduced physiological responses to acute stress and perceived social connectedness. 45,47 Our present findings therefore provide indirect evidence that alleviating these psychophysiological strains does not lead to short-term telomere lengthening in healthy individuals.

Limitations
The present study has limitations. The effect of measurement error on short-term LTL change, particularly lengthening, is controversial. 23,24 In the present sample, given our interassay coefficient of variability of 2.3% and mean LTL T/S ratio of approximately 1.0, measurement error should, on average, be 0.023. On an individual participant level, observed change ranged from −0.259 to 0.251 and is therefore unlikely predominantly attributable to measurement error, although lengthening or shortening classifications of particularly small change values could be distorted. Regarding the association with CT, measurement error should increase the risk of type II rather than type I error. The herein detected robust associations at multiple intervals of change are therefore unlikely attributable to measurement error. Rather, stronger associations may be found if measurement error was minimized. A further limitation is that, by evaluating structural change using CT, we were unable to examine potential associations with subcortical or allocortical regions, such as the hippocampus, which has been associated with LTL in cross-sectional studies. 82

Conclusions
Our findings contribute to the evidence that LTL changes more dynamically on the individual level than previously thought and indicate that short-term LTL change is associated with structural brain alteration. Further studies will need to identify the long-term implications of such changes in relation to cellular aging and the development of neurodegenerative disorders, as well as how to activate protective processes that influence LTL. In contrast to our hypotheses and some earlier reports, LTL of healthy adults was not influenced through contemplative mental training over 9 months.