Telomere shortening reflecting physical aging is associated with cognitive decline and dementia conversion in mild cognitive impairment due to Alzheimer’s disease

Demographic and clinical characteristics of study subjects

The study included 104 CU A-, 26 CU A+, 28 MCI A+, and 35 ADD A+ cases. Table 1 shows the demographic and clinical characteristics of participants by AD cognitive stage. The CU A- and CU A+ participants had a significantly lower mean age than the MCI A+ and ADD A+ participants. The ADD A+ patients had significantly lower educational levels and Mini-Mental State Examination (MMSE) scores than the other groups. The prevalence of men was lower in the ADD A+ group than in the other groups. The Consortium to Establish a Registry for AD (CERAD) and Clinical Dementia Rating-Sum of Boxes (CDR-SB) scores differed by cognitive stage, with better performance in the order of CU A+ > MCI A+ > ADD A+. The Logical Memory (LM) delayed recall score was significantly higher in the CU A- and CU A+ groups than in the MCI A+ and ADD A+ groups. The CU A- group had significantly lower standard uptake value ratios (SUVRs) on ¹⁸F-flutemetamol PET and ¹¹C-Pittsburgh compound-B (PiB) PET, lower prevalence of Apolipoprotein E (APOE) ε4 carrier, and higher CSF Aβ42 levels than the other groups. The ADD A+ patients had significantly higher CSF total tau (t-tau), tau phosphorylated at Thr181 (p-tau), t-tau/Aβ42, and p-tau/Aβ42 levels than the other groups. There were no differences between the groups regarding mean cortical thickness and TL (Table 1).

Associations between telomere length and cognitive function or AD biomarkers

The results of simple linear regression analyses of the association between TL as independent variable and each baseline cognitive function test as dependent variable are shown in Figure 1. Across all A+ participants, higher TL was associated with better global cognition measured with MMSE after adjustment for age, sex, and education (Table 2). Higher TLs were associated with better global cognition, measured with MMSE, as well as better memory function, measured with LM delayed recall, in the MCI A+ group. The relationship between CERAD and TL also showed a positive trend in the MCI A+ group, although it did not reach statistical significance.

TL was not correlated with CSF Aβ42, t-tau, p-tau, t-tau/Aβ42, or p-tau/Aβ42 levels, composite SUVRs on ¹¹C-PiB PET or on ¹⁸F-flutemetamol PET, cortical thickness, or hippocampal volume in all A+ participants as well as in each cognitive stage A+ group (Table 3).

Influence of telomere length on clinical outcomes over time

There were no differences between the TL quartile groups in each cognitive stage group with regard to age, sex, education, body mass index (BMI), homocysteine level, total sleep time, Mini Nutritional Assessment (MNA) and Geriatric Depression Scale (GDS) scores, or the prevalence of hypertension, DM, dyslipidemia, current smoking, current drinking, sarcopenia, APOE ε4 carriers, or for participants that met the World Health Organization PA guidelines (≥ 600 metabolic equivalent minutes of PA per week) [23] (Supplementary Table 1). Thus, the relationship of the baseline TL quartile level (as the explanatory variable) with the four clinical outcome measures was analyzed using linear mixed models with a function of TL quartile group, age, time, and group x time interaction, in each AD cognitive stage group (Supplementary Table 2).

In the lowest TL quartile group of MCI A+ participants, the estimated mean differences on 2-year follow-up evaluation compared with the highest TL quartile group were -9.438 (95% confidence interval [CI] = -14.567 ~ -4.309, P = 0.001), -26.708 (95% CI = -41.576 ~ -11.839, P = 0.001), 3.198 (95% CI = 1.323 ~ 5.056, P = 0.001), and 2.549 (95% CI = 0.527 ~ 4.571, P = 0.014) on MMSE, CERAD, CDR-SB, and Blessed Dementia Scale-Activities of Daily Living (BDS-ADL), respectively (Figure 2A–2D and Supplementary Table 2). Meanwhile, in the lowest TL quartile group of CU A- participants, the estimated mean difference over the same 2-year period compared with the highest TL quartile group was 0.524 (95% CI = 0.201 ~ 0.846, P = 0.002) on the BDS-ADL (Supplementary Table 2).

Influence of telomere length on AD cognitive stage progression

Eight (7.7%) participants from the CU A- group progressed to MCI over two years. Seven (26.9%) participants from the CU A+ group showed cognitive impairment over two years, with five progressing to MCI and two progressing to ADD. Ten (35.7%) participants from the MCI A+ group progressed to ADD within the two-year follow-up. There were no differences between the TL quartile groups in each cognitive stage group with respect to demographic factors, BMI, MNA, homocysteine level, total sleep time, hypertension, DM, dyslipidemia, current smoking, current drinking, sarcopenia, APOE ε4, PA, and GDS (Supplementary Table 1). Therefore, the Cox proportional hazards model was adjusted only for age. This analysis showed that among MCI A+ participants the lowest TL quartile group had a significantly greater probability of progressing to ADD compared with the highest TL quartile group (hazard ratio [HR] = 13.16, 95% CI = 1.11 ~ 156.61, P = 0.041) (Table 4 and Figure 3). Among CU A+ participants, rate of progression to MCI or ADD was higher in the lowest TL quartile group than in the other groups. However, for CU A+ participants the HR of conversion to MCI or ADD did not differ significantly between the lowest and highest TL quartile groups.

Participants

In the present study, 89 participants with amyloid pathology (A+; determined by abnormal amyloid PET findings or CSF Aβ42 levels below the cutoff point) and 104 CU individuals without amyloid pathology (A-) as a control group were selected from participants of the independent Validation cohort of the Korean Brain Aging Study for the Early Diagnosis and Prediction of AD (KBASE-V) [7]. The KBASE-V is an independent nationwide cohort from nine hospitals that reconfirms new potential biomarkers identified in the KBASE study and is intended to identify risk and prognostic factors for AD in additional studies. There were 167 CU, 72 MCI, and 56 ADD participants in the KBASE-V. The amyloid PET or CSF study was performed in 131 CU, 58 MCI, and 40 ADD participants. The number of participants with amyloid pathology proven by amyloid PET or CSF studies was 26, 28, and 36 in the CU A+, MCI A+, and ADD A+ groups, respectively. A patient in the ADD A+ group and a participant in the CU A- group were excluded because of inability to measure leukocyte TL. Finally, we included 26 CU A+, 28 MCI A+, 35 ADD A+, and 104 CU A- participants in the present study.

All of the CU participants had normal age-, sex-, and education-adjusted performance on four memory tests of the Korean version of CERAD (word list immediate recall, word list delayed recall, word list recognition, and constructional praxis recall) [29] and had a global CDR scale score of 0 [30]. The participants with MCI met the core clinical criteria for MCI due to AD established by the National Institute on Aging-Alzheimer’s Association (NIA-AA) workgroups [31], and the following criteria modified from the criteria proposed by Petersen et al. [32]: (1) CDR 0.5; (2) memory complaints compared to the participant’s previous memory function by patients, caregivers, or clinicians; (3) a performance score that was lower than 1.5 standard deviations (SDs) below the age-, education-, and sex-adjusted normative means for one or more of the four memory tests included in the CERAD [29]; (4) the ability to perform independent ADL [33]; and (5) not dementia. The ADD participants in the mild dementia stage met the following inclusion criteria: (1) dementia according to the Diagnostic and Statistical Manual of Mental Disorders, 4^th Edition (DSM-IV-TR) [34]; (2) probable ADD according to the NIA-AA core clinical criteria [35]; and (3) CDR 0.5 or 1. All participants were aged between 55 and 90 years, and had a reliable informant who could provide the requested information to investigators. For all participants, the exclusion criteria included [7]: (1) the presence of major psychiatric illness; (2) significant neurological or medical condition or comorbidities that could affect cognitive functions; (3) contraindications for magnetic resonance imaging (MRI) scans (e.g., pacemaker, claustrophobia); (4) illiteracy; (5) severe visual or hearing difficulty or serious communication or behavioral problems that could make a clinical examination or brain scan difficult; (6) taking an investigational drug; and (7) pregnancy or breastfeeding.

The study was performed in accordance with the International Harmonization Conference guidelines on Good Clinical Practice, and was approved by the institutional review board of each center. All participants, as well as legal representatives of ADD patients, provided written informed consent prior to study initiation.

Clinical assessment

All participants underwent physical and neurological examinations and thorough diagnostic procedures including assessment of participants’ cognition, abnormal behaviors, ADL, demographic characteristics, family history, current medications, vascular risk factors, and other comorbidities through the MMSE [29], GDS [36], BDS-ADL [37], and CDR yearly. The participants also underwent the CERAD every year and more detailed neuropsychological tests, including the Wechsler Memory Scale-Fourth edition Korean version LM I, II and recognition test, every two years [7]. Brain MRI and laboratory tests that included blood chemistry; lipid panel; complete blood count; serum levels of folate, vitamin B12, 25-hydroxy vitamin D, and brain-derived neurotrophic factor; C-peptide; glycated hemoglobin (HbA1c); homocysteine; adiponectin; venereal disease research laboratory test; thyroid function test; and APOE genotyping were performed at baseline.

Hypertension was defined as systolic blood pressure ≥ 140 mmHg, diastolic blood pressure ≥ 90 mmHg, or use of antihypertensive medication [38]. DM was defined based on current treatment with insulin or oral hypoglycemic medication, 8-h fasting plasma glucose ≥ 126 mg/dl, or HbA1c ≥ 6.5% [39]. Dyslipidemia was defined as total cholesterol ≥ 200 mg/dl, low-density lipoprotein cholesterol ≥ 130 mg/dl, high-density lipoprotein cholesterol < 40 mg/dl, triglyceride level ≥ 150 mg/dl, or the use of lipid-lowering drugs [40]. Participants’ weight and height were measured while they were wearing light clothing. BMI was calculated as their weight (kg) divided by the square of their height (m²). Participants underwent bioelectrical impedance analysis to measure the appendicular skeletal muscle mass index and sarcopenia was diagnosed according to the Asian Working Group criteria [41]. Nutritional status was evaluated using MNA [42]. PA was assessed using the International PA Questionnaire [43], and total sleep time was assessed using the Pittsburgh Sleep Quality Index [44].

Brain MRI

Brain MRI data were obtained from all participants using a 3.0 T MR scanner, which captured 3D T1-weighted and T2-weighted SPACE sagittal images of 0.8-mm thickness without gap as well as diffusion tensor imaging, axial fluid-attenuated inversion recovery imaging, and resting-state functional MRI. The MRI protocols were based on the AD Neuroimaging Initiative phase 2 MRI protocols [7, 45]. The 3D T1-weighted MRI parameters were as follows: repetition time (TR) = 2300 ms, echo time (TE) = 2.14 ms, inversion time (TI) = 900 ms, Flip Angle (FA) = 9°, and voxel resolution of 0.8 × 0.8 × 0.8 mm³ in the Skyra and Trio Tim scanners (Siemens, Washington DC, USA); TR = 7.32 ms, TE = 3.02 ms, TI = 400 ms, FA = 11°, and voxel resolution of 0.8 × 0.8 × 0.8 mm³ in the General Electric (GE) Discovery MR750 scanner (GE Healthcare, Milwaukee, WI, USA); and TR = shortest (6.8 ms), TE = shortest (3.1 ms), FA = 9°, and voxel resolution of 0.8 × 0.8 × 0.8 mm³ in the Achieva scanner (Philips Healthcare, Andover, MA, USA).

The 3D T1-weighted MRI data were processed using CIVET pipeline v2.1 (http://mcin-cnim.ca/ neuroimagingtechnologies/civet/) [46]. The N3 intensity nonuniformity correction algorithm was used to calibrate the intensity difference due to an inhomogeneity in a magnetic field [47]. Corrected T1-weighted images in native space were aligned to the Montreal Neurological Institute 152 standard space [48]. Non-brain tissue was excluded using the BET algorithm [49]. The registered images were classified into white matter, gray matter, and CSF using an advanced neural-net classifier [46]. The inner surfaces of the cortex were extracted from the partial volume corrected white matter mask using deformable spherical mesh, and then the outer surface of the cortex was automatically extracted using the constrained Laplacian-Based Automated Segmentation with Proximities algorithm [50]. Using the Euclidean distance between the linked vertices of the inner and outer surfaces, cortical thickness values in native space were calculated [51]. After intensity inhomogeneity correction, the corrected T1-weighted images were segmented into the left and right sides of hippocampus using FMRIB’s Integrated Registration and Segmentation Tool [52]. Hippocampal volumes were normalized for total intracranial volume.

Amyloid PET

Of the total 193 participants in this study, 181 underwent amyloid PET at baseline. There were 116 participants who underwent ¹⁸F-flutemetamol PET, 64 who underwent ¹¹C-PiB PET, and one participant with the historical ¹⁸F-florbetapir PET result. The PET methods for each of the tracers have been previously described [7]. The SUVR was obtained by using the pons as a reference region on ¹⁸F-flutemetamol PET and the cerebellar gray matter as the reference region on ¹¹C-PiB PET. Composite SUVR values were formed by averaging the SUVR values for frontal, temporal, parietal, occipital, anterior cingulate, and posterior cingulate/precuneus cortices. Based on previous work, elevated Aβ PET was defined as composite SUVR ≥ 0.634 on ¹⁸F-flutemetamol PET and composite SUVR > 1.21 on ¹¹C-PiB PET [7].

Cerebrospinal fluid analysis

At baseline, 123 of the total 193 participants underwent spinal fluid testing in the morning. CSF was collected in 15-mL polypropylene transfer tubes (Falcon, Corning Science, NY, USA) and immediately centrifuged at 2000 g for 10 min at room temperature (RT). The supernatant (~10 mL) was frozen in dry ice and transferred to Inha University’s laboratory, where CSF biomarkers were measured. After thawing at RT, the shipped CSF samples were mixed with a pipette with a polypropylene tip, and 0.4-mL CSF sample aliquots were frozen in polypropylene tubes (Sarstedt AG and Co., Nümbrecht, Germany) and stored at -80°C until analysis. Aβ42, t-tau and p-tau were measured using the multiplex xMAP Luminex platform with INNO-BIA AlzBio3 kits (Fujirebio Europe, Ghent, Belgium). Based on previous work, participants who underwent CSF studies were deemed to have AD pathology when the CSF Aβ42 was 433.68 pg/ml or lower [7].

Telomere length assay

Leukocyte TL was examined once at baseline. To this end, DNA was extracted from whole blood using G-DEX^TM IIb RBC lysis buffer and G-DEX^TM IIb Cell lysis buffer (Intron, MA, USA). DNA hydration was performed with 300 μL of DNA hydration solution (QIAGEN, Hilden, Germany). TL analysis was carried out using a nonradioactive TeloTAGGG TL Assay (Roche Boehringer-Mannheim, Grenzach-Wyhlen, Germany) as described by the manufacturer. Approximately 2-4 μg of DNA from each sample was digested with Hinf I/RsaI enzyme mix and isolated by gel electrophoresis. DNA fragments were transferred to a nylon membrane (Millipore, Bedford, MA, USA) by Southern transfer and hybridized to a digoxigenin (DIG)-labeled probe specific for telomeric repeats. The membrane was incubated with DIG-specific antibodies conjugated to alkaline phosphatase, and the probe was visualized by chemiluminescence using an image analyzer (ImageQuant LAS 4000, GE Healthcare, Little Chalfont, UK). Mean telomeric repeat-binding factor lengths were determined by comparison with molecular weight standards.

Outcomes

We used 4 summary scores to index disease progression, namely, the MMSE, the CDR-SB, the CERAD, and the BDS-ADL. The MMSE (range 0-30) evaluates the orientation, registration and recall of three words, attention and calculation, language, repetition, and complex commands [29]. The CDR assesses dementia severity along 5 levels of impairment (rated as 0, 0.5, 1, 2, or 3) in each of 6 domains: memory, orientation, judgment and problem solving, community affairs, home and hobbies, and personal care. The CDR-SB score (range 0-18) is the sum of the ratings in each of the 6 domains [30]. The CERAD total score is the sum of the scores of its 7 sub-tests: verbal fluency (animal naming), Boston naming (15 items), word list learning, constructional praxis, word list recall, word list recognition, and constructional praxis recall [29]. The BDS-ADL score (range 0-17) was determined using an 11-item questionnaire assessing activities of daily life, including performing household tasks, coping with small sums of money, remembering short lists of items, finding one’s way indoors, finding one’s way around familiar streets, interpreting surroundings, recalling recent events, tendency to dwell on the past, eating, using the toilet, and dressing [37]. Increases in scores represent worsening on the CDR-SB and BDS-ADL, and improvement on the MMSE and CERAD.

We also investigated cognitive stage transitions (from CU to MCI or dementia, and from MCI to dementia) over two years. The diagnosis of MCI was based on the core clinical criteria for MCI due to AD established by the NIA-AA workgroups [31] and the criteria modified from those proposed by Petersen et al. [32], as mentioned earlier [7]. The diagnosis of dementia was based on DSM-IV-TR criteria for dementia [34]. The diagnosis of ADD was based on the core clinical criteria for probable ADD established by the NIA-AA workgroups [35].

Statistical analyses

For comparisons among the CU A-, CU A+, MCI A+, and ADD A+ groups, we used chi-square tests for categorical variables and one-way analysis of variance (ANOVA) for age and education level. When statistically significant overall differences were detected in the ANOVA test, pairwise comparisons of means between the diagnosis groups were performed by Tukey’s method. Analysis of covariance (ANCOVA) was performed for statistical analysis of GDS with age, sex, and education level as covariates. ANCOVA was performed for statistical analysis of MMSE, CERAD, LM delayed recall test, CDR-SB, CSF Aβ42, t-tau, p-tau, t-tau/Aβ42, and p-tau/Aβ42 levels, the composite SUVRs on ¹¹C-PiB PET and on ¹⁸F-flutemetamol PET, mean cortical thickness, average score of right and left hippocampal volumes, and TL with age, education level, sex, and GDS as covariates. When an overall statistically significant difference was detected in the ANCOVA test, pairwise comparisons of the means between diagnosis groups were performed by Bonferroni post hoc analysis. Linear regression adjusted for age, sex, and education was used to examine the relationship between TL as independent variable and each baseline cognitive function test as dependent variable in all A+ participants and within each group. Pearson’s correlation coefficients were evaluated to examine the relationship between TL and AD biomarkers in all A+ participants and within each A+ cognitive stage group.

TL was divided into quartiles in each of the CU A-, CU A+, MCI A+, and ADD A+ groups. We used chi-square tests for categorical variables and the Kruskal Wallis test for continuous variables to compare between TL quartile groups the factors associated with TL and AD progression in each AD cognitive stage group. We used linear mixed models with a random subject effect to analyze the relationship between baseline TL quartile level as the explanatory variable and the four clinical outcome measures. The fixed effects included TL quartile group, age, time, and group × time interaction. We used the 1^st order autoregressive (AR1) covariance structure in the mixed model, which was selected by the Akaike information criterion (AIC) of the model. The AIC values of AR1 were the lowest in most tested correlation structures (Supplementary Table 3).

To examine the effects of baseline TL on patient’s progression through the different cognitive stages, a Cox proportional hazards model was used after controlling for age with proportional hazard assumption checked by log-log plotting. Survival curves according to TL quartile groups did not intersect and were found to be parallel in each of CU A-, CU A+, and MCI A+ groups. Data are presented as HRs and 95% CIs. Survival was defined as the time between entering the KBASE-V study and the progression of cognitive status (from CU to MCI or dementia, or from MCI to dementia) or a censoring event such as withdrawal from the study or the last completed follow-up examination. Statistical analyses were performed using SPSS 19.0 (SPSS, Chicago, IL, USA). P < 0.05 was considered significant.