SEARCH

SEARCH BY CITATION

Keywords:

  • neuropsychological;
  • genetic;
  • longitudinal

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

We hypothesize that quantitative phenotypes related to Alzheimer's disease (AD), rather than the dichotomous disease phenotype, will increase the statistical power for identifying genetic risk factors. Neuropsychological test scores, which allow for the measurement of loss of cognitive function over time, are a particularly promising option for this approach. Using data from a cohort study of prodromal AD in 365 community-recruited subjects with and without memory problems with a baseline and often one or more follow-up administrations of a detailed neuropsychological test battery, we performed both cross-sectional and longitudinal analyses using the known AD gene APOE and four other putative AD genes as predictors. APOE and a promoter polymorphism in insulin degrading enzyme (IDE_4U) showed evidence for association with cross-sectional and longitudinal changes in memory (P = 0.016–0.025) and other cognitive functions. APOE and a polymorphism in the alpha-2-macroglobulin gene (A2M18i) also showed evidence for association with cross-sectional and longitudinal changes in executive functioning (P = 0.010–0.042). In some cases, longitudinal analysis offered stronger evidence for association than could be seen cross-sectionally. These preliminary results suggest that this approach has promised the development of a quantitative phenotype related to AD, but more elaborate methods will be required to address multiple comparisons issues in the setting of correlated data. © 2007 Wiley-Liss, Inc.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

After early successes in Alzheimer's disease (AD) genetics, investigators now face the same frustrations as those seeking genes for other complex disorders. The online database AlzGene (http://www.alzgene.org [Bertram et al., 2007]) documents well over 1,500 polymorphisms covering more than 500 putative AD genes. Often, an initial positive report of association is followed by multiple non-replications. Although most of these are probably false positives, a measurable fraction of the non-replications could be false negatives related to ascertainment issues or power. We hypothesize that quantitative phenotypes related to AD, particularly when examined longitudinally, will not only increase the statistical power for identifying genetic risk factors, but may also shed light on the molecular mechanisms driving disease onset and progression.

Substantial research initiatives have focused on the predictive utility of various cognitive measures to identify individuals at-risk of developing dementia. In particular, the rate of decline in specific cognitive domains appears to be a reliable preclinical indicator of the subsequent development of dementia [Elias et al., 2000; Chen et al., 2001]. Thus, it is possible that a genetic approach that focuses on these early symptoms will be capable of detecting not only AD genes but those that contribute to risk at this critical stage—when prevention and treatment strategies might be the most effective—as well.

Using repeated neuropsychological test battery data from a community-recruited, longitudinal cohort of 367 subjects who initially ranged from cognitively normal to mild cognitive impairment just short of the functional impairment necessary for a diagnosis of dementia [Daly et al., 2000], and closely followed for up to 13 years, we conducted an exploratory genetic association analysis of five putative AD genes. These findings serve to support the role of genetics in the decline of cognitive function and to illustrate the potential of longitudinal phenotypes for AD and related clinical phenomena.

METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

Study Subjects

The sample consisted of 367 individuals recruited in three different waves from the general population through the print media between 1993 and 2005. Detailed description of the recruitment procedure can be found elsewhere [Albert et al., 2001]. Subjects underwent a multistage screening procedure and were required to meet the following criteria: (1) at least 65 years of age (with the exception of seven individuals 57–64); (2) free of underlying medical, neurologic, or psychiatric illness; (3) have a Clinical Dementia Rating (CDR) [Hughes et al., 1982; Morris, 1993] of 0 (normal) or 0.5 (mild symptoms with no or minimal functional impairment); (4) willingness to participate in study procedures.

Neuropsychological Test Battery

All subjects were administered a battery of neuropsychological tests, from which 12 tests with 16 test scores were selected based upon previous work in this and other studies suggesting they might be sensitive to change in the prodromal phase of AD [Albert et al., 2001; Aggarwal et al., 2005]: (a) four memory tests: California Verbal Learning Test (CVLT) [Delis et al., 1987], the Delayed Word Recall Test [Knopman and Ryberg, 1989], Rey Ostereith Complex Figure [Rey, 1941], and the Free and Cued Selective Reminding Test [Grober and Buschke, 1987]; (b) four tests of executive function: Trail Making Test, Part B [Reitan, 1958], Self-Ordering Test [Petrides and Milner, 1982], Alpha Span Test [Craik, 1986], and Digit Span Backward [Wechsler, 1988]; (c) one language test: Controlled Word Association Test for letters and for categories [Benton and Hamsher, 1976]; (d) three tests of sustained attention: Digit Span Forward [Wechsler, 1988], Trail Making Test, Part A [Reitan, 1958], Cued Reaction Time [Baker et al., 1985]. The individuals who administered the neuropsychological test battery were different than those who conducted the interview used to generate the CDR ratings, and the test scores were not used in the assignment of the CDR ratings.

Longitudinal Measures

Subjects were assessed at baseline and up to two additional visits at varying intervals with the neuropsychological test battery described above. Of the 367 with complete neuropsychological testing data at baseline, 235 subjects had at least two visits and 64 had three visits. The variable number of follow-up visits is a reflection of the ascertainment via three cohorts which began at different times and thus, some subjects have simply not reached a third follow up visit. The time between first and last testing was approximately 4.6 years (s.d. = 2.9, range 1–3).

Genotyping

A total of six polymorphisms from five genes were genotyped (Table II). In addition to the known AD gene APOE, we selected four candidate genes based on prior findings of our research group (alpha-2-macroglobulin [A2M; rs3832852 (“18i”]): [Saunders et al., 2003]; ubiquilin 1 [UBQLN1; rs12344615 (“UBQ8i”)]: [Bertram et al., 2005], the online database AlzGene (insulin degrading enzyme [IDE; rs2251101 (“IDE_7”) and “IDE_U4”] and brain-derived neurotrophic factor [BDNF; rs6265 (“V66M”): http://www.alzgene.org). APOE was genotyped as described previously [Blacker et al., 1997]. Single nucleotide polymorphisms (SNPs) were genotyped using high-efficiency fluorescent polarization detected single base extension (HEFP-SBE, on a “Criterion Analyst AD,” Molecular Devices, Inc.). Briefly, PCR primers were designed to yield products between 200 and 400 bp in length and added to ∼10 ng of genomic DNA using individually optimized PCR conditions (primer sequences and cycling conditions are available on request). PCR primers and unincorporated dNTPs were degraded by the direct addition of exonuclease I (0.1–0.15 U/rxn) and shrimp alkaline phosphatase (1 U/rxn). The single base extension step was carried out using Thermosequenase (0.4 U/rxn) and the appropriate mix of R110-ddNTP, TAMRA-ddNTP (3 mM), and all four unlabeled ddNTPs (22 or 25 µM) to the Exo1/SAP treated PCR product. Genotyping of the 5 bp ins/del polymorphism in A2M was done in 96-well format by PCR followed by amplicon separation via capillary electrophoresis on a MegaBACE-1000 sequencing/genotyping instrument (Amersham Biosciences, Piscataway, NJ). Genotypes were determined using the manufacturer's allele-calling software (“Genetic Profiler,” v1.5).

Statistical Analysis

A dominant (i.e., allele carrier-status) genetic model was used for all analyses. This was done primarily to ensure adequate numbers of subjects at each time point, but also to ensure a uniform approach to the analysis. For baseline NP tests, linear regression was used to estimate the effect of genotype on testing outcome. For longitudinal NP testing data, linear mixed effects models (random intercept and random slope) were constructed using SAS 9.1. Random intercepts allow for the detection of genetic differences in function at the time of ascertainment, essential given that some subjects entered the study with greater or lesser degrees of impairment. Random slopes allow us to detect genetic effects on the rate of decline. All analyses were adjusted for baseline age, gender, and highest attained education level. Age and education, both in years, were modeled linearly. Each variable was assessed for normality, and aside from the Trail Making Tests (parts A and B), all tests were approximately normally distributed. The Trail Making Tests were log-transformed, resulting in approximately normal measures. Given the exploratory nature of this study, P-values were not adjusted for multiple-comparisons. Hardy–Weinberg Equilibrium was assessed using standard Chi-square test.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

Sample Characteristics

The average age of study participants at baseline was 72.5 years (s.d. = 5.5). Approximately 57% of the cohort was female and the average number of years of education was 15.5 (s.d. = 2.9). Table I displays these demographic data in greater detail, along with the mean scores from the neuropsychological battery. In all, 15 neuropsychological tests covering memory, executive function, language, and sustained attention were included in addition to the CDR Sum of Boxes (CDR-SB). Table II shows the genotype distribution of each of the six polymorphism studied here as 0, 1, or 2 copies of the denoted allele (i.e., the minor allele for biallelic polymorphisms). Allele frequency ranged from 0.08 to 0.44 (APOE-2 allele and IDE_U4, respectively). APOE-4 allele frequency is somewhat higher than in population samples, reflecting ascertainment based on memory impairment. All genotypes were found to be in Hardy–Weinberg Equilibrium (data not shown).

Table I. Mean (s.d.) and Range of Scores for the Neuropsychological Test Battery at Baseline
DemographicMean (s.d.)Range (min, max)
Age (years)72.5 (5.5)(57, 87)
Education (years)15.5 (2.9)(5, 24)
Gender (% female)57
CDR0.92 (0.90)(0, 3.5)
 PercentCumulative percent
 0.031.231.2
 0.5–1.550.681.8
 2.0–3.518.2100.0
Domain/testMean (s.d.)Range (min, max)
 Memory
 CVLT Total Score48.36 (11.36)15, 75
 Delayed Word Recall5.61 (2.12)0, 10
 Rey Ostereith Complex Figure35.58 (14.36)0, 72
 Selective Reminding Free Recall41.68 (8.36)10, 59
 Executive function
 Trail Making Test, Part B4.52 (0.45)3.14, 6.06
 Self Ordering Test11.05 (5.08)0, 31
 Alpha Span4.49 (0.90)2, 8
 Digit Span Backward5.31 (1.42)0, 8
 Language
 Letter fluency44.43 (13.07)9, 77
 Category fluency19.18 (5.30)4, 39
 Sustained attention
 Digit Span Forward6.91 (1.30)4, 39
 Trail Making Test, Part A4.08 (0.32)2.89, 5.17
 Cued Reaction Time394.09 (50.40)280, 602
Table II. Genotype Distribution
PolymorphismAlleleFrequencyCopies of minor allele
012
APOE20.08306560
 30.7712151233
 40.15264976
A2M18iDEL0.172399413
BDNFA0.2122110919
IDE_U4A0.4511215975
IDE_7C0.3016214527
UBQ8iC0.172401028

Genetic Associations

Table III presents all associations between genes and neuropsychological tests that show a P-value of less than 0.05 for the longitudinal analysis. In addition to the longitudinal analysis, the baseline effects are also shown. Out of the 10 associations with P-values less than 0.05, all but one were attributable to APOE, A2M18i or IDE_U4. In particular, three associations, including the most significant association observed (CLVT total P = 0.01), are attributable to the APOE-4 allele. Beyond this expected result, three of the significant associations are attributable to A2M18i, and three are attributable to IDE_U4. The other polymorphism investigated in IDE (IDE_7) did not reveal any associations of interest. The memory domain, and particular the total score on the CVLT, represented two associations below a P-value of 0.05. Four components of executive function are represented in Table III as well, with two Self-Ordering, two Digit-Span Backward and one Trail Making Part B.

Table III. Results of the Nominally Significant Associations as Determined by the P-value From the Longitudinal Analysis
OutcomeGeneBaselineLongitudinal
Effect sizeaP-valueEffect sizebP-value
  • Results from the corresponding baseline analysis are shown as well.

  • a

    The regression coefficient from a linear regression model using baseline function only.

  • b

    The regression coefficient from a linear mixed effects model using longitudinal data.

CVLT totalAPOE−3.050.0109−3.070.0102
Trail Making Part BA2M18i0.120.01260.120.0103
Category fluencyIDE_U41.320.01621.210.0205
CVLT totalIDE_U42.460.03442.650.0228
Cued Reaction TimeIDE_U4−11.100.0636−12.690.0254
Self-ordering totalA2M18i1.040.05761.020.0390
Self-ordering totalUBQ8i−0.630.2600−1.010.0401
Digit Span BackwardA2M18i−0.170.2900−0.290.0422
Digit Span BackwardAPOE−0.230.1642−0.280.0478
Trail Making Part AAPOE0.070.03850.070.0493

Proof of Concept—APOE

Using APOE as a standard of comparison, we used graphs to further explore neuropsychological tests that showed association with both APOE and any other gene. Figures 1 and 2 show the mean scores of the neuropsychological tests by visit and genotype. In both figures, the APOE-4 carriers are compared to the non-APOE-4 carriers for the given neuropsychological test is at the top, and a similar comparison is made by carrier status for the minor allele of the other gene. Figure 1 shows CVLT total scores for APOE (APOE-4 allele carrier vs. non-carrier) and IDE_U4. Visually, the pattern of decline in CVLT Total scores is similar across APOE and IDE_U4. IDE_U4 homozygotes for the major allele (IDE_U4 = 0) have lower mean scores at baseline compared to all other genotypes—a trend that continues over visits 2 and 3. The same effect is apparent in APOE-4 allele carriers. Figure 2 displays one of the executive function tests, Digit-Span Backward, as a function of APOE and A2M18i over all three visits. Carrying at least one copy of the minor allele of A2M18i, subjects have a lower mean Digit Span Backward score that is consistent across all three visits. This effect is similar to that seen for APOE-4 carriers.

thumbnail image

Figure 1. Mean CVLT total by visit and genotype (APOE-4 carrier and IDE_U4).

Download figure to PowerPoint

thumbnail image

Figure 2. Mean digit span backward by visit and genotype (APOE-4 carrier and A2M18i).

Download figure to PowerPoint

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES

The majority of studies involving older populations “at-risk” for dementia are cross-sectional and therefore limit the investigation of individual rates of decline over time [Wilson et al., 2002]. The use of longitudinal neuropsychological tests allowed us to assess the impact of potential genetic risk factors at baseline, as well as over time. Further, we were able to determine whether genetic effects were longitudinally consistent over an average of 4.6 years. The findings from this study highlight the potential utility of longitudinal measures of AD-related traits, particularly in the context of genetic research. In the majority of gene-outcome associations observed here, stronger signals were observed with longitudinal than baseline-only analysis. It is likely that longitudinal phenotypes will prove more powerful for gene-mapping efforts than cross-sectional approaches, although a combination of simulation studies and ongoing research experience will be required to provide a definitive answer.

The APOE locus has proven to be a consistent and reliable tool for genetic research. It provides a proof of concept for methodological research and offers a real example of genetic susceptibility, including an idea of how genes might be contributing to disease risk. In an attempt to make greater use of APOE as a standard for comparison, we used a graphical approach to examine polymorphisms that showed comparable effects to APOE on the same neuropsychological test. This approach lends further support to a potential genetic effect for IDE_U4 and A2M18i, which resemble APOE in the pattern of baseline differences and patterns of change over time.

We are careful to point out that the present study was exploratory in nature so the findings were not assessed for statistical significance in the context of multiple testing. Therefore, the results are to be interpreted with extreme caution until more research can explore these hypotheses directly. However, these results might serve as the basis of hypotheses for other genetic studies of cognitive decline, or help to prioritize follow-up of genes showing positive association to the AD disease phenotype.

The future of genetic research involves the increasingly available large-scale genotyping platforms, where hundreds of thousands of SNPs are available for association analysis. Coupled with rich longitudinal datasets that measure multiple phenotypes, there is great potential for uncovering disease-susceptibility variants. However, the increased numbers of phenotypes and genetic markers carry a substantial price—the number of tests balloons, creating a severe multiple-testing problem. Several approaches to address this problem have been in development and are discussed elsewhere [Hirschhorn and Daly, 2005]. The majority of approaches focus on multi-stage designs using case-control datasets, where one binary outcome is recorded per individual. Data from the present study, and other cohort studies with multiple phenotypes observed over time, do not readily fit into this mold, as the majority of the information would have to be discarded in order to apply these methods, ultimately reducing the statistical power to detect any genetic effects. In order to carry this and similar longitudinal cohort studies into the large-scale genetic association era with any appreciable power, methods are needed to circumvent the multiple-testing problem, and to make maximal use of all measured traits.

With regard to optimizing multiple traits and/or repeated measures, principal components analyses and related methods have been used to reduce the dimensionality of neuropsychological data [Albert et al., 2001; Wilson et al., 2002] and other datasets that include multiple correlated variables. However, while these composite phenotypes may be relevant for clinical or diagnostic purposes, it remains unclear as to how useful they will be for genetic research, as they incorporate no information related to heritability. An alternative strategy geared towards constructing composite phenotypes based upon maximizing heritability as opposed to phenotypic variance has been developed in the context of family-based genetic association tests [Lange et al., 2004], and has been successfully applied to a family-based genome-wide association study investigating body mass index [Herbert et al., 2006]. This method accommodates any combination of longitudinal and/or repeatedly measured traits with the goal of finding the maximally heritable aggregate trait to subsequently test for association. This method too is presently being extended to the analysis of unrelated samples, which will enable this and other cohort studies to make the most out of potentially powerful multidimensional longitudinal measures.

Longitudinal datasets are an extremely useful and informative approach to the study of cognitive decline, but the utility of such data for genetic research remains to be seen. Using the methods available, we explored the associations of APOE and four potential candidate genes for AD in an effort to lay groundwork for the coming large-scale approaches on horizon. We found that, using APOE as a standard for comparison, we could explore other associations and determine how similar or different they are to the effect of APOE. We also outlined the issues inherent in using data of this type for larger-scale genomic studies, and highlighted areas of promise for addressing those issues in this sample, in other samples focused on cognitive decline and AD, and more for other complex diseases.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. REFERENCES
  • Aggarwal NT, Wilson RS, Beck TL, Bienias JL, Bennett DA. 2005. Mild cognitive impairment in different functional domains and incident Alzheimer's disease. J Neurol Neurosurg Psychiatry 76: 14791484.
  • Albert MS, Moss MB, Tanzi RF, Jones K. 2001. Preclinical prediction of AD using neuropsychological tests. J Int Neuropsychol Soc 7: 631639.
  • Baker EL, Letz R, Fidler AT. 1985. A computer administered neurobehavioral evaluation system for occupational and environmental epidemiology. J Occup Med 27: 206212.
  • Benton AL, Hamsher K. 1976. Multilingual Aphasia Examination. Iowa City: University of Iowa Press.
  • Bertram L, McQueen MB, Mullin K, Blacker D, Tanzi RE. 2007. Systematic meta-analyses of Alzheimer's disease genetics association studies: The AlzGene database. Nat Genet 39: 1723.
  • Bertram L, Hiltunen M, Parkinson M, Ingelsson M, Lange C, Ramasamy K, Mullin K, Menon R, Sampson AJ, Hsiao MY, Elliott KJ, Velicelebi G, Moscarillo T, Hyman BT, Wagner SL, Becker KD, Blacker D, Tanzi RE. 2005. Family-based association between Alzheimer's disease and variants in UBQ LN1. N Engl J Med 352: 884894.
  • Blacker D, Haines JL, Rodes L, Terwedow H, Go RC, Harrell LE, Perry RT, Bassett SS, Chase G, Meyers D, Albert MS, Tanzi R. 1997. ApoE-4 and age at onset of Alzheimer's disease: The NIMH genetics initiative. Neurology 48: 139147.
  • Chen P, Ratcliff G, Belle SH, Cauley JA, DeKosky ST, Ganguli M. 2001. Patterns of cognitive decline in presymptomatic Alzheimer disease: A prospective community study. Arch Gen Psychiatry 58: 853858.
  • Craik FIM. 1986. A functional account of age differences in memory. In: KlixF, HagendorfH, editors. Human memory and cognitive capabilities. Holland: Elsevier Press. pp 409422.
  • Daly E, Zaitchik D, Copeland M, Schmahmann J, Gunther J, Albert M. 2000. Predicting conversion to Alzheimer disease using standardized clinical information. Arch Neurol 57: 675680.
  • Delis D, Kramer J, Kaplan E, Ober B. 1987. The California Verbal Learning Test. New York: Psychological Corp.
  • Elias MF, Beiser A, Wolf PA, Au R, White RF, D'Agostino RB. 2000. The preclinical phase of Alzheimer disease: A 22-year prospective study of the Framingham Cohort. Arch Neurol 57: 808813.
  • Grober E, Buschke H. 1987. Genuine memory deficits in dementia. Developmental Neuropsychology 3: 1336.
  • Herbert A, Gerry NP, McQueen MB, Heid IM, Pfeufer A, Illig T, Wichmann HE, Meitinger T, Hunter D, Hu FB, Colditz G, Hinney A, Hebebrand J, Koberwitz K, Zhu X, Cooper R, Ardlie K, Lyon H, Hirschhorn J, Laird NM, Lenburg ME, Lange C, Christman MF. 2006. A common genetic variant 10 kb upstream of INSIG is associated with adult and childhood obesity. Science 312: 279283.
  • Hirschhorn JN, Daly MJ. 2005. Genome-wide association studies for common diseases and complex traits. Nat Rev Genet 6: 95108.
  • Hughes C, Berg L, Danziger W, Coben L, Martin R. 1982. A new clinical scale for the staging of dementia. Br J Psychiatry 140: 566572.
  • Knopman DS, Ryberg S. 1989. A verbal memory test with high predictive accuracy for dementia of the Alzheimer type. Arch Neurol 46: 141145.
  • Lange C, Van Steen K, Andrew T, Lyon H, DeMeo DL, Raby B, Murphy A, Silverman EK, MacGregor A, Weiss ST, Laird NM. 2004. A family-based association test for repeatedly measured quantitative traits adjusting for unknown environmental and/or polygenic effects. Stat Appl Genet Mol Biol 3: article 17.
  • Morris J. 1993. The Clinical Dementia Rating (CDR): Current version and scoring rules. Neurology 43: 24122414.
  • Petrides M, Milner B. 1982. Deficits in subject-ordered tasks after frontal and temporal lobe lesions in man. Neuropsychologia 20: 249262.
  • Reitan RM. 1958. Validity of the Trail Making Test as an indicator of organic brain damage. Percept Motor Skills 8: 271276.
  • Rey A. 1941. L'examen psychologique dans les cas d'encephalopathie traumatique (psyhcological examination of cases of traumatic encephalopathy). Archives de Psychologie 28: 286340.
  • Saunders AJ, Bertram L, Mullin K, Sampson AJ, Latifzai K, Basu S, Jones J, Kinney D, MacKenzie-Ingano L, Yu S, Albert MS, Moscarillo TJ, Go RC, Bassett SS, Daly MJ, Laird NM, Wang X, Velicelebi G, Wagner SL, Becker DK, Tanzi RE, Blacker D. 2003. Genetic association of Alzheimer's disease with multiple polymorphisms in alpha-2-macroglobulin. Hum Mol Genet 12: 27652776.
  • Wechsler D. 1988. The Wechsler Adult Intelligence Scale—Revised. New York: Psychological Corporation.
  • Wilson RS, Beckett LA, Barnes LL, Schneider JA, Bach J, Evans DA, Bennet DA. 2002. Individuals differences in rates of change in cognitive abilities of older persons. Psychol Aging 17: 179193.