Longitudinal change in memory performance associated with HTR2A polymorphism

Neurobiology of Aging 27 (2006) 150–154

Brief communication

Longitudinal change in memory performance associated with HTR2A polymorphism Chandra A. Reynolds a,∗ , M˚arten Jansson b , Margaret Gatz b,c , Nancy L. Pedersen b,c a

Department of Psychology, University of California at Riverside, 1344 Olmsted Hall, Riverside, CA 92521-0426, USA b Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden c Department of Psychology, University of Southern California, Los Angeles, CA, USA Received 28 June 2004; received in revised form 23 November 2004; accepted 20 December 2004

Abstract We present a fresh approach to evaluating association with candidate genes and cognitive change by testing association for parameters describing individual growth curves from twins. Moderate genetic influences on memory in aging adults has been shown in quantitative genetic studies. A recently reported, association of a HTR2A polymorphism with episodic memory in young unrelated adults led us to investigate the association between a nearby polymorphism and longitudinal memory performance in the Swedish Adoption/Twin Study of Aging (SATSA). Analysis of growth curve parameters suggests that both how well individuals perform on figural memory at age 65 years and nonlinear change in figural memory performance across age are associated with HTR2A. Individuals with two copies of the common G allele demonstrated higher figural memory performance longitudinally than those with the less frequent A allele, with performance trajectories differing by 2–6% per year. These findings imply a role for the 5-HT2A serotonin receptor on the formation of episodic memories in older adults. © 2005 Elsevier Inc. All rights reserved. Keywords: Episodic memory; Serotonin receptors; HTR2A; Association; Twins

1. Introduction Whilst anomalous serotonergic functioning has been linked to variety of mental disorders including depression [3] and neuropsychiatric symptomatology in patients with Alzheimer’s disease [8], recent experimental manipulations of serotonin receptor 2A (5-HT2A) activity suggest a role in spatial working memory in primates [23]. In the CNS, 5HT2A receptors populate prefrontal cortex and to a lesser extent limbic regions such as the hippocampus [1], regions linked to memory processes [22]. In humans, a linear loss of 5-HT2A receptors with age has been reported in positron-emission studies of 5-HT2A receptor binding with a reported 16–18% linear loss per decade in the hip-

∗

Corresponding author. Tel.: +1 951 827 2430; fax: +1 951 827 3985. E-mail address: [email protected] (C.A. Reynolds).

0197-4580/$ – see front matter © 2005 Elsevier Inc. All rights reserved. doi:10.1016/j.neurobiolaging.2004.12.009

pocampus and posterior medial prefrontal cortex [17]. However, a nonlinear model may be more appropriate for regions such as the posterior medial prefrontal cortex with the maximal loss at midlife [17]. Recently, a functional genetic variant (H452Y) of the gene coding for 5-HT2A serotonin receptors (HTR2A) was significantly associated with verbal and figural delayed recall in young adults [4]. Specifically, those with the rare tyrosine variant recalled fewer words (males and females) or figures (males only) after a 5 min or 24 h delay than those homozygous for the more common histidine variant. We expand previous work to include a test of association with the −1438 G/A variant on baseline memory performance and on longitudinal memory change over 13 years. In particular, we hypothesize an association between 5-HT2A and longitudinal memory decline given prior evidence of age-associated losses of 5-HT2A receptors in frontal and hippocampal brain regions.

C.A. Reynolds et al. / Neurobiology of Aging 27 (2006) 150–154

151

2. Method

2.3. Data analysis

2.1. Study population

Sibling-based tests of association were conducted on baseline memory performance to avoid biases from population stratification often inherent in case–control analyses with unrelated individuals [6,15]. Analyses of twin pairs, both monozygotic (MZ) and dizygotic (DZ), allowed tests of whether an allelic effect would be observed within pairs who differ on −1438 G/A HTR2A genotypes (DZ pairs) as well between pairs (both MZ and DZ included). Both additive allelic effects and dominance deviations were estimated using a maximum-likelihood model-fitting approach [15] adjusting for age, sex, and education (measured on a four point scale, elementary school to university degree or higher). In addition, we modeled the background genetic and nonshared environmental variation by decomposing the twin variances and covariances. A series of nested models were fit and compared using a difference Chi-square test to determine the significance of: between-pair effects (comprising population stratification plus true allelic association), within-pair effects (true allelic association), equating between- and within-pair effects (if significant, indicates population stratification), dominance deviations, and overall association (any allelic effect, within or between).

Memory measures were obtained during four in-person testing sessions (IPT1–IPT4) spanning 13 years from the population-based Swedish Adoption Twin Study of Aging (SATSA) [14]. Twins with data from at least one testing occasion, who reached at least age 50 and had no diagnosis of dementia during the study period, were included. All participants provided informed consent and the study was approved by institutional review boards including the Karolinksa Institute in Stockholm, Sweden, the University of Southern California and the University of California at Riverside in the United States. The sample sizes were 595, 560, 539, and 517 at each testing occasion. Average age was 64.96 years at IPT1 (S.D. = 8.22), 65.44 (S.D. = 8.83) at IPT2, 68.30 (S.D. = 8.99) at IPT3, and 70.79 (S.D. = 10.04) at IPT4. Of 798 twins with memory scores from any IPT, 62% participated at three or more sessions. Blood for genotyping was collected at IPT3. All IPT3 participants were sequenced for the HTR2A promoter polymorphism, −1438 G/A [dbSNP: rs6311] by pyrosequencing as described in previous work [9]. A total of 498 individuals had both genotyping and memory scores. Of 214 genotyped pairs that had complete data on age, sex, and education covariates, 201 (94 MZ, 105 DZ) had at least one memory score for each twin. The sample allele frequencies for the G and A alleles were 0.63 and 0.37, respectively, based on one member of each MZ pair and both members of each DZ pair. Observed genotype frequencies for GG, AG and AA were 0.40, 0.46 and 0.14 respectively. Hardy–Weinberg equilibrium was supported [χ2 (1) = 0.0488, p = 0.83]. 2.2. Measures Memory tasks included Digit Span (sum of forward and backward) [10], Names and Faces [5], and Thurstone’s Picture Memory [18]. Digit Span tasks tap attention and working memory, including central executive processes [7]. Names and Faces tests both immediate and delayed recall of photographs of faces paired with names. Thurstone’s Picture Memory tests delayed recognition memory of line drawings of common items, e.g., truck, table, etc. All scores were transformed into percent correct for ease in presentation. A novel feature of the present analysis is the use of individually estimated growth model parameters as outcome traits based upon phenotypic latent growth models of the four memory tests (full technical details available upon request). Briefly, multilevel regression procedures were used in which Bayesian estimates of the intercept, linear slope, and quadratic trends were obtained based upon a centering of the growth curve at age 65 years [16] and reflect, respectively: memory ability at 65 years, linear rate of memory change at 65 years, and acceleration of memory change over age.

3. Results 3.1. Baseline memory performance No significant allelic effects for Digit Span or Names and Faces immediate or delayed recall were found. For the Thurstone delayed recognition task we observed evidence for association: between-pair effects could be set to zero [χ2 (2) = 0.707, p = 0.735], and marginal significance was observed when within-pair effects were set to zero [χ2 (2) = 5.580, p = 0.055]. However, within and between effects could not be equated [χ2 (2) = 6.10, p = 0.047]. Given that the between effects were neither significant nor equatable to the within-pair effects, in the final model we retained both within-pair additive and dominance effects but set the between-pair additive effects and dominance deviations to zero. Expected means based on the final model parameters obtained with the full sample of twins (see Table 1) suggest that those with the GG genotype perform better (mean = 71.62%) than those with the AG (mean = 62.88%) or AA (mean = 66.77%) genotypes. The findings correspond well to within-pair comparisons of observed scores within DZ pairs who differ on genotype: (1) DZ twins with the GG genotype perform better on the Thurstone Picture Memory task than their cotwins with the AA (median difference = 7.14%) or AG genotype (median difference = 10.71%); (2) those with the AG genotype also perform better than those with the AA genotype (median difference = 3.57%). The proportion of phenotypic and genotypic variance in figural memory scores at baseline explained by

C.A. Reynolds et al. / Neurobiology of Aging 27 (2006) 150–154

152

Table 1 HTR2A allelic effects for figural recognition memory (Thurstone Picture Memory): full and final model estimates Effect

Grand mean Gw Aw AGw Gb Ab AGb NMZ NDZ

Baseline score

Intercept at 65 years

Slope at 65 years

Quadratic

Full

Final

Full

Final

Full

Full

69.00 1.21 −1.21 −6.31 −0.30 0.30 0.83 102 111

69.20 1.21 −1.21 −6.31 – – –

72.24 1.24 −1.24 −3.84 −0.25 0.25 1.48 102 111

72.77 1.25 −1.25 −3.83 – – –

−0.428 0.010 −0.010 0.151 −0.017 0.017 −0.056 101 111

Final −0.466 – – – – – –

−0.0264 −0.0028 0.0028 0.0372 −0.0011 0.0011 −0.0151 89 104

Quadratic, VE constrained Final −0.0340 −0.0028 0.0028 0.0371 – – –

Full −0.0263 −0.0029 0.0029 0.0372 −0.0011 0.0011 −0.0150 89 104

Final −0.0339 −0.0028 0.0028 0.0371 – – –

Note: Grand mean: the average trait score adjusted for any allelic, age, sex, and education effects; A: additive allelic effect for the HTR2A A allele; G: additive allelic effect for the HTR2A G allele; AG: heterozygote effect; w: within-pair; b: between-pair; N: pair sample size, MZ: monozygotic; DZ: dizygotic. To find the effect for the AA or GG genotype the grand mean is added to twice the A or G allelic effect, respectively. To find the effect for the AG genotype the grand mean is added to AG heterozygote effect.

the HTR2A polymorphism is 1.44% and 1.79%, respectively (see Table 1). 3.2. Longitudinal change To utilize the longitudinal information, we next examined allelic association with empirical Bayes parameters of intercept at age 65, linear slope at age 65 years and a quadratic term, reflecting curvature of memory change over age. No significant associations were observed for Digit Span or Names and Faces immediate or delayed recall. However, for the Thurstone delayed recognition task we observed evidence of within-pair association for intercept and quadratic terms but not for linear slope. For the intercept, between-pair effects could be set to zero [χ2 (2) = 1.958, p = 0.376], within-pair effects could not be set to zero [χ2 (2) = 6.053, p = 0.048], and within and between effects could not be equated [χ2 (2) = 7.998, p = 0.018]. The withinpair dominance parameter was nonsignificant if dropped singly [χ2 (1) = 2.475, p = 0.116]. However, the within-pair additive allelic effect became nonsignificant when setting the between effects and within pair dominance effects to zero. Therefore, similar to the baseline memory results described above, in the final model within-pair additive and dominance effects were retained but the between-pair additive effects and dominance deviations were set to zero. Expected means of the intercepts based on the final model parameters (see Table 1) suggest that those with the GG genotype perform better at age 65 years (mean = 75.27%) than those with the AG (mean = 68.89%) or AA (mean = 70.27%) genotypes. The proportion of phenotypic and genotypic variance in figural memory scores at age 65 years (intercept) explained by the HTR2A polymorphism is 1.38% and 2.20%, respectively. For the quadratic parameter for the Thurstone task, dominance deviations for the heterozygous genotype AG were significant [χ2 (2) = 12.222, p = 0.002], between-pair effects could be set to zero [χ2 (2) = 3.996, p = 0.136], within-pair effects could not be set to zero [χ2 (2) = 11.457, p = 0.003], and between- and within-pair effects could not be equated

[χ2 (2) = 15.126, p = 0.001]. Instability of the background variance was evident across the nested submodels with small but consistent decreases to the background nonshared environmental variance as within-pair and between pair allelic association were considered in the model. In order to properly estimate effect sizes (see Table 1), we refit the association models holding background nonshared environmental variation constant (to the value estimated under no allelic association). The within-pair effect remained significant (p = 0.004) and gives credence to a genuine allelic effect on figural memory change. Expected means of the quadratic term based on the final model additive and dominance parameters suggest that those with the GG genotype shows a greater negative acceleration (−0.04%) than those with the AG (0.00%) or AA (−0.03%) genotypes. The proportion of phenotypic and genotypic variance in nonlinear figural memory change explained by the HTR2A polymorphism is 2.12% and 5.88%, respectively. Fig. 1 represents the expected growth curve over age for the Thurstone Picture Memory test by genotype given the best-fitting models for intercept (within-pair additive allelic

Fig. 1. Expected longitudinal change by 5-HTR2A (−1438 G/A) genotypes for Thurstone delayed recognition task. Note: Curve based on age, sex and education adjusted association parameters where age was centered at age 65 years, males were coded 0, and the lowest education level was coded 0.

C.A. Reynolds et al. / Neurobiology of Aging 27 (2006) 150–154

effects), linear slope (no association) and quadratic (withinpair additive and nonadditive allelic effects) parameters. As shown in Fig. 1, the highest performance at all ages is predicted for the GG genotype with a 2–6% advantage over those with the AG genotype and a 4–5% advantage over those with the AA genotype.

4. Discussion The findings of the present study extend previous reports of allelic association of the H452Y HTR2A variant with delayed verbal and figural recall performance in younger adults [4] to the −1438 G/A variant with delayed figural recognition performance and longitudinal change in older adults. A feature of the current study is the use of individual growth trajectory parameters as outcomes, an elegant way to summarize individual longitudinal profiles of scores into a few parameters that describe change over age. Indeed with this method we show that different aspects of the growth curve may show differential association, i.e., association with memory performance at 65 years as well quadratic/nonlinear trends over age but no association for the linear rate of change in performance. Also consistent with the previous report of the H452Y HTR2A variant is the specificity of the association of the −1438 G/A variant with figural memory after a delay (Thurstone) but not with immediate recall, measured with Names and Faces in our study and with both verbal and figural immediate recall tasks in the prior study. We also found no association for attention and working memory (Digit Span) or for another delayed recall task (Names and Faces). The lack of association for the Names and Faces recall tasks is perhaps explainable by high task difficulty leading to low proportion correct on average. The specificity of allelic association with delayed figural recognition performance could represent a functional brain–behavior relationship that longitudinally is driven by serotonin 2A receptor losses in brain regions identified with episodic memory, i.e., hippocampus and prefrontal regions [19]. Conceivably a loss in serotonin 2A receptors would affect early neural processes involved in episodic memory storage, such as serotonin-induced synaptic changes [11]. Previously we reported a significant association between the HTR2A A/A genotype and depressed mood in elderly men although not women [9]. Relationships between depressive symptoms and recall memory performance have been reported in large population-based studies [2,21], and it has been suggested that similar etiological influences may play a role in their co-occurrence. Similarly, research indicates associations between affective disorders and Alzheimer‘s disease (AD), with differential depletion of 5HT2A serotonin receptors reported between AD, depressed, and normal elderly controls [12,20] suggesting potentially different dysfunctions of the serotonergic system in this case.

153

Use of longitudinal data from twins allowed a test of true allelic association accounting for population stratification and permitted examining trajectories of change rather than simply baseline differences. A further substantive contribution of the present study is that nonlinear change in figural recognition performance over age is genotype dependent with indication of both additive and nonadditive genotypic effects. The most dramatic change in memory across age is noted for the AG heterozygote, which exhibits steep decline over age in comparison to the homozygotes AA and GG. The pattern of change for those with the AA genotype parallels the GG genotype, though at every age those with the AA genotype exhibit worse performance. Our findings imply a role for the 5-HT2A serotonin receptor on the formation of episodic figural memories as well as changes in memory functioning in the second half of the life-span.

Acknowledgements Analytical work was supported by Grant No. R01 AG17561. The Swedish Adoption/Twin Study of Aging (SATSA) is supported by Grant Nos. R01-AG04563 and AG10175, The MacArthur Foundation Research Network on Successful Aging, and the Swedish Council for Social Research (97:0147:1B).

References [1] Burnet PWJ, Eastwood SL, Lacey K, Harrison PJ. The distribution of 5-HT1A and 5-HT2A receptor mRNA in human brain. Brain Res 1995;676(1):157–68. [2] Comijs HC, van Tilburg T, Geerlings SW, Jonker C, Deeg DJ, van Tilburg W, et al. Do severity and duration of depressive symptoms predict cognitive decline in older persons? Results of the Longitudinal Aging Study, Amsterdam. Aging Clin Exp Res 2004;16(3):226–32. [3] Cryan JF, Leonard BE. 5-HT-sub(1A) and beyond: the role of serotonin and its receptors in depression and the antidepressant response. Hum Psychopharmacol: Clin Exp 2000;15(2):113– 35. [4] de Quervain DJF, Henke K, Aerni A, Coluccia D, Wollmer MA, Hock C, et al. A functional genetic variation of the 5-HT2A receptor affects human memory. Nat Neurosci 2003;6(11):1141– 2. [5] DeFries JC, Plomin R, Vandenberg SG, Kuse AR. Parent-offspring resemblance for cognitive abilities in the Colorado adoption project: biological, adoptive, and control parents and 1-year-old children. Intelligence 1981;5(3):245–77. [6] Fulker DW, Cherny SS, Sham PC, Hewitt JK. Combined linkage and association sib-pair analysis for quantitative traits. Am J Hum Genet 1999;64(1):259–67. [7] Hester RL, Kinsella GJ, Ong B. Effect of age on forward and backward span tasks. J Int Neuropsychol Soc 2004;10(4):475– 81. [8] Holmes C, Arranz MJ, Powell JF, Collier DA, Lovestone S. 5HT2A and 5-HT2C receptor polymorphisms and psychopathology in late onset Alzheimer’s disease. Hum Mol Genet 1998;7(9):1507– 9.

154

C.A. Reynolds et al. / Neurobiology of Aging 27 (2006) 150–154

[9] Jansson M, Gatz M, Berg S, Johansson B, Malmberg B, McClearn GE, et al. Association between depressed mood in the elderly and a 5-HTR2A gene variant. Am J Med Genet 2003;120B(1):79–84. [10] Jonsson C-O, Molander L. Manual till CVB-skalan (Manual of the CVB Scales). Stockholm: Psykologi F¨orlaget; 1964. [11] Kandel ER. The molecular biology of memory storage: a dialogue between genes and synapses. Science 2001;294(5544):1030– 8. [12] Meltzer CC, Price JC, Mathis CA, Greer PJ, Cantwell MN, Houck PR, et al. PET imaging of serotonin type 2A receptors in latelife neuropsychiatric disorders. Am J Psychiatry 1999;156(12):1871– 8. [14] Pedersen NL, McClearn GE, Plomin R, Nesselroade JR, Berg S, DeFaire U. The Swedish Adoption/Twin Study of Aging: an update. Acta Genet Med Gemellol: Twin Res 1991;40(1):7– 20. [15] Posthuma D, de Geus EJ, Boomsma DI, Neale MC. Combined linkage and association tests in mx. Behav Genet 2004;34(2):179– 96. [16] SAS. SAS online documentation for windows, version release 8. Cary, NC: SAS Institute Inc.; 1999.

[17] Sheline YI, Mintun MA, Moerlein SM, Snyder AZ. Greater loss of 5-HT(2A) receptors in midlife than in late life. Am J Psychiatry 2002;159(3):430–5. [18] Thurstone LL. Primary mental abilities. Chicago: University of Chicago Press; 1948. [19] Tulving E. Episodic memory: from mind to brain. Annu Rev Psychol 2002;53:1–25. [20] Versijpt J, Van Laere KJ, Dumont F, Decoo D, Vandecapelle M, Santens P, et al. Imaging of the 5-HT2A system: age-, gender-, and Alzheimer’s disease-related findings. Neurobiol Aging 2003;24(4):553–61. [21] Vinkers DJ, Gussekloo J, Stek ML, Westendorp RGJ, van der Mast RC. Temporal relation between depression and cognitive impairment in old age: prospective population based study. BMJ 2004;329(7471):881–90. [22] Wall PM, Messier C. The hippocampal formation—orbitomedial prefrontal cortex circuit in the attentional control of active memory. Behav Brain Res 2001;127(1-2):99–117. [23] Williams GV, Rao SG, Goldman-Rakic PS. The physiological role of 5-HT2A receptors in working memory. J Neurosci 2002;22(7):2843–54.