Lindsay A. Farrer, Tatyana Sherbatich, Sergey A. Keryanov, Galina I. Korovaitseva, Ekaterina A. Rogaeva, Svetlana Petruk, Smita Premkumar, Yuri Moliaka, You Qiang Song, York Pei, Christine Sato, Natalya D. Selezneva, Svetlana Voskresenskaya, Vera Golimbet, Sandro Sorbi, Rangene Duara, Svetlana Gavrilova, Peter H. St. George-Hyslop, Evgeny I. Rogaev. Association Between Angiotensin-Converting Enzyme and Alzheimer Disease. Arch Neurol. 2000;57(2):210–214. doi:10.1001/archneur.57.2.210
Angiotensin-converting enzyme has been reported to show altered activity in patients with neurologic diseases. An insertion-deletion polymorphism in ACE has recently been linked to heart disease, cerebrovascular disease, and AD.
To determine whether the angiotensin-converting enzyme (ACE) is associated with risk of Alzheimer disease (AD).
We investigated the ACE polymorphism as a potential risk factor for AD in 151 patients with AD and 206 ethnically matched controls from Russia and in 236 patients with AD and 169 controls from North America by means of allele association methods and logistic regression.
None of the ACE genotypes was associated with increased susceptibility to AD in the total sample or in subsets stratified by apolipoprotein E gene (APOE) ϵ4 status. However, the D allele was more frequent among AD cases between ages 66 and 70 years compared with controls in both the Russian (P = .02) and North American (P = .001) datasets. In this age group, the effect of D (odds ratio, 11.2; 95% confidence interval, 2.9-44.0) appeared to be independent of and equal or greater in magnitude to the effect of APOE ϵ4 (odds ratio, 7.8; 95% confidence interval, 3.5-7.4).
Our results suggest that APOE and ACE genotypes may be independent risk factors for late-onset AD, but the ACE association needs to be confirmed in independent samples in which the time and extent of vascular cofactors can be assessed.
THE GENETIC factors that predispose individuals to Alzheimer disease (AD) have not yet been fully defined. The proportion of AD that has a major gene component has been estimated to be at least 36%.1 Alzheimer disease associated with mutations in the amyloid precursor protein,2 presenilin 1,3 and presenilin 24,5 genes is transmitted as an autosomal dominant and highly penetrant trait, but, collectively, these defects account for less than 2% of all AD cases.6 Although the apolipoprotein E gene (APOE) ϵ4 allele accounts for an estimated 45% to 60% of the genetic risk for AD,7,8 the observation that ϵ4 is neither necessary nor sufficient to cause the disease8 has prompted the investigation of other candidate susceptibility genes.
Angiotensin-converting enzyme (ACE) is a dipeptidyl carboxypeptidase that catalyzes the formation of angiotensin II by cleaving a dipeptide from angiotensin I.9 Activity of ACE has been reported to be altered in regions of the brain from patients with neurologic diseases. Arregui et al observed a decrease in ACE activity in patients with schizophrenia10 and Huntington disease11 but an increase in activity in patients with AD.12 In contrast, no significant changes were reported with Parkinson disease.13 Angiotensin-converting enzyme has been identified as a normal constituent of the human cerebrospinal fluid. One study indicated that ACE activity in cerebrospinal fluid was decreased by 47% in patients with AD in comparison with age- and sex-matched controls.14 An Alu insertion-deletion polymorphism in the ACE gene results in the genotypes D/D, I/D, and I/I. Homozygosity of the D allele has been linked to risk of myocardial infarction and cardiomyopathy.15- 17D/D has also been reported as more common in subjects with a history of stroke than those without (relative risk, 1.79).18 Since vascular risk factors are likely to have an effect on the presence and severity of AD,19 we evaluated ACE as a potential AD risk factor.
One-hundred fifty-one AD cases were ascertained through the clinical departments and Alzheimer Disease and Related Disorders Center within the Mental Health Research Center in Moscow, Russia. All living patients underwent a standard neurologic examination, a personal interview, psychometric testing, and brain imaging (computed tomography or magnetic resonance imaging). These data were supplemented with information contained in the medical records. The diagnosis of AD was established according to International Statistical Classification of Diseases, 10th Revision,20 and National Institute of Neurological and Communicative Disorders and Stroke–Alzheimer's Disease and Related Disorders Association21 criteria. Patients with evidence of vascular or "mixed" dementia were excluded. The mean age at onset was 64.6 ± 8.8 years (range, 40-88 years). The identity of more than 95% of patients was verified as Moscovite of Russian origin by passport records and interview of relatives. A group of 206 ethnically matched, cognitively normal individuals aged 40 to 91 years (mean, 64.9 ± 11.4 years) was randomly selected from a population study of the frequency of genetic disorders in 3 Russian provinces22 and from hospitals and nursing homes in the Moscow area.
A second set of 236 subjects with AD was ascertained through memory disorder units at the University of Toronto (Toronto, Ontario), Mt Sinai Medical Center (Miami, Fla), and the University of Florence (Florence, Italy). A diagnosis of AD was established according to standard consensus clinical criteria.21 The mean age at onset was 75.2 ± 7.6 years (range, 52-94 years). Blood samples were also obtained from 169 unrelated, cognitively normal individuals aged 38 to 96 years who were recruited from the same communities either through participation in normative studies of aging or as spouse controls for the patients with AD. The mean age at examination of the controls was 70.2 ± 10.2 years. Nonwhites were purposely excluded from the study to guard against false-positive or false-negative associations caused by population (ie, genetic) stratification within or between AD case and control groups as discussed in previous studies that used this dataset.23
Genomic DNA was isolated from peripheral-blood leukocytes by phenol-chloroform extraction. Polymerase chain reaction of APOE exon 4 was carried out by a standard procedure24 with minor modifications. The ACE polymorphism was screened by a modified polymerase chain reaction–based assay.25 Forward primer 5‘-CTGGAGACCACTCCCATCCTTTCT-3‘ and reverse primer 5‘-GATGTGGCCATCATTCGTCAGAT-3‘ were used for amplification in a 20-µL reaction, consisting of 200-ng genomic DNA, 5 pmol of each primer, 0.25 µmol/L of each deoxynucleotide three-phosphate (dNTP), 2 µL of 10 polymerase chain reaction (PCR) buffer (Gibco BRL, New York, NY), and 0.2 U of Taq polymerase. After denaturation for 4 minutes at 94°C, the reaction mixture was subjected to 30 cycles of 36 seconds at 94°C, 26 seconds at 57.5°C, and 38 seconds at 72°C, followed by 1 cycle at 72°C for 7 minutes. The polymerase chain reaction products, 190 base pairs (bp) (D allele) and 480 bp (I allele), were resolved on 2% agarose gels.
A χ2 test was used to compare allele frequencies between AD cases and controls. The Fisher exact test was used to compare genotype frequencies because the χ2 test may not be valid when cell counts are less than 5. Analyses were repeated in subsamples stratified by age (using 65 years as the cutoff) and APOE ϵ4 carrier status. The influence of ACE and APOE genotypes, age, and sex on the odds of developing AD was assessed by logistic regression procedures.26 To accommodate the polychotomous classification of ACE genotype in the regression analysis, 2 indicator variables were constructed representing the D/D, I/D, and I/I genotypes. These variables took on the value of 1 if the subject had the corresponding genotype, and 0 otherwise. The I/I genotype was considered as the referent because previously the D allele has been associated with disease states.15- 18 For APOE, subjects were classified as ϵ4 allele carriers or noncarriers. Age at onset of AD among cases and age at last examination among controls were assigned to the age variable. Interaction between ACE, APOE, age, and sex was evaluated by deriving product terms for each ACE genotype with age and sex.
There were no significant differences in either ACE allele or genotype frequency between cases and controls in the total samples, or in subgroups stratified by APOE ϵ4 status, in Russian subjects or subjects studied at the University of Toronto (Table 1). Logistic regression analysis did not show a significant effect of the D allele or D/D genotype on the risk of AD after controlling for age at examination, sex, and ϵ4 status (data not shown). However, stratification of subjects by age group disclosed a higher frequency of I/I controls compared with AD cases among persons aged 66 years and older. Further analysis of this group indicated that this effect was significant and limited to a subgroup aged 66 to 71 years in both the Moscow (P = .02) and Toronto (P = .001) datasets (Table 2). Moreover, in this age group, significant deviation from Hardy-Weinberg equilibrium was observed in patients but not controls from Moscow (cases, P = .02; controls, P = .77) and Toronto (cases, P = .04; controls, P = .17), as would be expected with a genotypic association with disease. The ACE allele and genotype distributions were not different between cases and controls in any other age groups in either dataset, and a protective effect of the D allele was still absent after adjustment for sex and APOE ϵ4 status (Table 3). By contrast, the presence of at least 1 APOE ϵ4 allele significantly increased the odds of AD in all age groups. In the age group 66 to 70 years, the risk of AD among those having the D/D or I/D genotype compared with I/I subjects was nearly identical in the Moscow and Toronto samples, and of equal or greater magnitude in comparison with the risk associated with APOE ϵ4 (Table 3). Analysis of the combined datasets showed that the ACE D allele increased the odds of AD 11.2 times. Removal of the APOE variable from the model did not change the odds ratio associated with D, suggesting that ACE and APOE are independent risk factors for AD in this age group.
Our results demonstrate an association between AD and the ACE D allele in 2 ethnically and geographically distinct populations, but only among persons between the ages of 66 and 70 years. In this age group, the effect of D appeared to be independent of and equal or greater in magnitude to the effect of APOE ϵ4. This association appears to be robust, especially because it was evident in 2 populations.
Although our results were still significant after correction for multiple testing, this association should be considered tentative until confirmed in larger samples enriched for subjects between 60 and 75 years of age. The age cutoffs were chosen arbitrarily and may not encompass the true risk period associated with ACE D. We investigated the association in the immediately adjacent 5-year age subgroups (ie, ages 61-65 years and 71-75 years), and the results were negative (data not shown). However, there were too few Toronto subjects aged 61 to 65 years and too few Moscow subjects aged 71 to 75 years to draw meaningful conclusions in these particular subsets, and given the large differences in ACE allele frequencies between the 2 datasets, we deemed it unwise to pool subjects in these particular age groups.
The ACE insertion-deletion polymorphism has also been implicated as a genetic risk factor for AD in a recent case-control study involving 3 cohorts from the United Kingdom.27 However, in each cohort, the pattern of association was exactly opposite the one we observed (ie, the I/I genotype conferred an increased risk of AD). Of note, the magnitude of the effect of I/I in these populations (odds ratios ranging from 1.8 to 2.7) was considerably less than the effect of D/D in our samples from North America and Moscow (odds ratios of 13.7 and 11.6, respectively). Furthermore, we detected an association only in a relatively narrow age range, whereas Kehoe et al27 did not evaluate their data by age group.
Our findings could be spurious because of a subtle bias in the selection of controls. This explanation is unlikely for 2 reasons. First, the cases and controls in the Moscow dataset were ethnically homogeneous and from the same geographic region. Similarly, the controls in the Toronto dataset were spouses or from the same community as the cases. Second, our APOE results are consistent with evidence from other studies28- 33 regarding the significance of the ϵ4-AD association at both ends of the age spectrum. Alternatively, the ACE-D/AD association may have been observed simply by chance. The ACE allele and genotype frequencies were not significantly different between cognitively impaired individuals and control subjects recruited from 2 community-based aged populations in England,34 but an AD-ACE association may have been obscured by diagnostic heterogeneity in the case samples. The ACE D allele frequencies were nearly identical in a clinic-based investigation of 49 patients with AD, 41 patients with vascular dementia, and 40 age-matched controls in Tel Aviv, Israel.35 Notably, neither of these studies considered the possibility that the effect of ACE is limited to a specific age interval. Another explanation of our results is that individuals in the 66- to 70-year age group have mixed dementia or pure multi-infarct dementia, or represent a subset of AD cases in which vascular factors play a more prominent role.19 As far as can be determined from clinical and neuroimaging studies in vivo, patients with vascular disease were excluded from our cohort. A final possibility that also reconciles the different patterns of association in the present study and the one by Kehoe et al27 is that an AD susceptibility allele is located elsewhere in ACE or a nearby gene.
Studies showing that the D/D genotype is a risk factor for stroke18,36 are consistent with our finding of a deleterious effect of the D allele on risk of AD. However, it is unclear why homozygosity of the D allele is required for increased risk of stroke, whereas 1 or 2 copies of the D allele promote AD risk. If, as suggested by the study by Doi et al,36D/D increases risk of stroke only among persons younger than 60 years, the lack of an AD-ACE association in our early-onset sample might result from attrition (ie, comorbidity of cerebrovascular disease) of persons with the D allele who would also be vulnerable to AD. Alternatively, the D allele may influence AD risk by promoting small cerebrovascular infarcts, the cumulative effect of which is not manifest until about age 65 years. Although the biological mechanism underlying the association between AD and ACE is unclear, the observation that plasma ACE levels are lowest for I/I individuals and highest for D/D individuals37 suggests that a high level of plasma ACE might promote the development of AD. It is noteworthy that ACE is a key factor in the production of angiotensin II and degradation of bradykinin, which are involved in vascular physiology.
Interestingly, the deleterious effect of the D allele is diminished in our sample of persons older than 70 years. In fact, a comparison of controls aged 71 years and older with controls aged 66 to 70 years showed that the frequency of D homozygotes increased by more than 50% in the Moscow group and nearly doubled in the Toronto dataset. The D/D frequency in the oldest group in each of our datasets is comparable with the frequency observed in a population of French centenarians.38 Galinsky et al39 also observed a depletion of the I/I genotype in persons aged 84 years and older, but in men only. The apparent benefit of D homozygosity for longevity might explain the sudden diminution of the deleterious effect of this genotype on AD risk after age 70 years. Secular trends in ACE genotype frequencies also suggest the possibility that the D/D genotype reflects better survival with the disease than a true risk factor for the disease. Thus, the association between ACE and AD may be an example of a gene with pleiotropic age- and sex-dependent effects on disease and survival.
Accepted for publication June 25, 1999.
This work was supported by grants from the HHMI (75195-546801), INTAS-RFBR (95-0871), RFBR (95-5-4a), Russian Home Genome Program, European Commission Copernicus, Medical Research Council of Canada, Alzheimer Association of Ontario, and the National Institutes of Health, Bethesda, Md (AG09029 and TW00886).
Reprints: Evgeny I. Rogaev, PhD, Laboratory of Molecular Brain Genetics, Mental Health Research Center, Zagorodnoe sh 2/2, 113152, Moscow, Russia (e-mail: email@example.com).