ApoE-4 and Age at Onset of Alzheimer's Disease

Subjects.

The sample consisted of 697 individuals affected with AD, from 310 families. Participants were evaluated following a standardized protocol ^[19] at three sites: the University of Alabama at Birmingham, Johns Hopkins University, and Massachusetts General Hospital, Harvard Medical School. Families were recruited from local memory disorder clinics, nursing homes, and the surrounding community, with the only requirement that each family contain at least two living blood relatives with memory problems. As can be seen in Table 1, the majority of the affected individuals were sibling pairs (216 families, 69.7%), but there were 43 larger sibships (13.9%), 37 other pairs (12.6%) (parent-child [19 pairs], half-siblings [3 pairs], first cousin [13 pairs], half first cousins [1 pair], aunt/uncle-niece/nephew [3 pairs]), and 12 more complex families (3.9%). Unaffected relatives, generally siblings, were included when they were available and willing to participate. Therewere a total of 231 unaffected subjects from 120 families (28.7%). Of these families, 66 had 1 unaffected member, 34 had 2, and 20 had 3 or more. An additional 7 study subjecjts with blood available who had unclear phenotypes (e.g., initially unaffected siblings with the apparent onset of memory problems who could not be further evaluated, or additional affected subjects who did not meet criteria for AD) were included in the analyses as "phenotype unknown." All subjects (or, for significantly cognitively impaired individuals, their legal guardian or caregiver with power of attorney) gave informed consent.

One member of the pair, designated the primary proband, was required to meet NINCDS/ADRDA criteria for probable Alzheimer's disease, and the other (and andy additional affecteds in the family) for possible or probable Alzheimer's disease. ^[20] Possible AD was allowed for family members beyond the primary proband primarily because (1) using the criteria in the present study, a diagnosis of probable or possible AD was nearly as specific as probable AD alone (84.2% versus 89.5%, respectively), and captured a far greater proportion of the cases with a pathologic diagnosis of AD (the sensitivity of a probable or possible diagnosis was 82.5% versus 52.5% for probable alone) ^[19]; and (2) it is common in genetic studies to use less stringent criteria for secondary cases because the prior probability of a correct diagnosis is higher given a family history of the disorder under study. Age at onset was determined based on an interview with a knowledgeable informant and review of mecical records.

Diagnostic assessments were made by and MD- or PhD- level clinician at the originating site, based on clinical evaluation and review of medical records. To date, 266 of the cases (39.2%) have been subjected to a consensus procedure involving senior clinicians from all three sites ^[19]; the remainder are in process.

Of the 697 affecteds, 610 (89.8%) met NINCDS/ADRDA criteria ^[20] for a diagnosis of probable AD, 67 (9.8%) met criteria for possible AD, and an additional 2 demented subjects, with insufficient documentation to meet these criteria during life, were included because they met research neuropathologic criteria for definite AD on autopsy. ^[21] Every effort is being made to obtain an autopsy on the participants. Of the 113 autopsies obtained to date on the affecteds (which originally included 681 subjects), 112 (99.1%) met neuropathologic criteria ^[21] for definite AD. Two siblings included in the original smaple were excluded from these analyses because one of the siblings failed to meet criteria for definite AD on autopsy.

Amplification and genotyping of apoE.

Genomic DNA was amplified by polymerase chain reaction (PCR) with the following primers: 5 prime-TCCAAGGAGCTGCAGGCGGCGCA-3 prime and 5 prime-ACAGAATTCGCCCCGGCCTGGTACACTGCCA-3 prime. For each amplification, 20 ng of human genomic DNA, 1 micro M of each primer, 200 micro M of dCTP, dTTP, and dGTP, 25 micro M of dATP, 20 micro Ci (alpha-sup 33 P) dATP (NEN Research, NEN-Dupont, Boston, MA), 10% DMSO, 0.1 micro L of 100x BSA (purified, Boehringer-Mannheim, Indianapolis, IN), 1.6 units Taq DNA polymerase (5 units/micro L; Fisher Biotech, Agawan, MA), and 1x reaction buffer (supplied by vendor with 15 mM MgCl₂) in a final volume of 10 micro L. Reactions were carried out in V-well plates (CoStar, Agawan, MA) in a PTC-100 Programmable Thermal Controller (MJ Research, Watertown, MA) under the following conditions: 5 minutes at 94 degrees C, 34 cycles of 30 seconds at 94 degrees C, 30 seconds at 69 degrees C, 1.5 minutes at 70 degrees C, followed by a final extension step at 70 degrees C for 10 minutes. ApoE isotypes were then determined by cleaving with the restriction enzyme Hha I (5 units) added directly to each well and incubated at 37 degrees C for 6 hours. Ten micro L of 2x stop dye were added to each well and 5 micro l of this mix was then loaded into each well of a 6% nondenaturing polyacrylamide gel. Following electrophoresis at 45 mA for 1.5 hours, the gel was transferred to Whatman 3MM chromatography paper, dried by vacuum, and exposed to Kodak XAR-5 film. Autoradiography was carried out for 1 to 16 hours at -70 degrees C. The resulting genotypes on each autoradiograph were read independently by two different observers and scoring of the alleles was determined, as presented in Hixson and Vernier. ^[22]

Statistical techniques.

Four basic statistical methods were used to explore the relationship among apoE-4, AD, and its age at onset: association, affected pedigree member, sibling pair, and conventional lod score analyses.

Overall statistics.

The mean age at onset among affecteds in the families was 71.47 (SD +/- 8.66). In males, the mean age at onset was 69.38 (+/- 9.70) and in females it was 72.34 (+/- 8.04). The family mean age at onset varied from 44 to 91, while in the individual subjects the age at onset varied from 42 to 93. The distribution of families by family structure, number of affecteds, and mean age at onset is given in Table 1. It is noteworthy that 257 of the families (83%) had onset over age 65.

The mean age at examination of the unaffected subjects was 69.1 (+/- 12.2), and the range was 31 to 93. Excluding 22 subjects aged 50 and under, most of whom were the children, nieces, or nephews of affected subjects, the mean was 71.8 (+/- 9.5).

The vast majority of families (95%) were Caucasian in origin. When looked at specifically, there were no significant differences in the results between the families of Caucasian and African-American origin.

Association analysis.

The association analysis of affected subjects, based on the primary proband in each family and stratified on individual age at onset, indicated that, as expected, the apoE-4 allele frequency was substantially and significantly elevated compared with controls in the whole group of AD families (p < 0.000001) and in each age-at-onset stratum (Table 2). This significance level is calculated based on the control values of Wilson et al. ^[23] However, as can be seen from the other control values listed, ^[27-29] the difference would be highly significant no matter which control values were selected for comparison.

Of note, the elevation in apoE-4 allele frequency was most marked in the 61 to 65 year age-at-onset group, and fell considerably at later ages; the allele frequency in the 61 to 65 group was 0.62 versus 0.33 in the 81+ group (p < 0.0002). The elevation in apoE-4 gene frequency was also less marked below age 60, but the sample size did not permit dividing this group further to assess the impact at earlier ages.

In order to confirm the association analysis based on allele frequencies, we also performed a modification of the transmission/disequilibrium test (TDT) that allows for less than full information on the parents. ^[34] These results were also highly significant (p < 0.0001), and also showed a peak effect in the early 60s.

The allele frequencies for unaffected subjects, based on one individual selected at random from each of the 120 families with one or more unaffected subjects, and excluding those with age at examination <or=to50, are shown in Table 3. Consistent with their ascertainment through AD families, there is a high background rate of apoE-4 among the unaffected relatives. The apoE-4 allele frequency is particularly high among unaffected subjects aged 60 and under, and then drops off in the 61 to 65 group. This is consistent with the peak in apoE-4 frequency observed at 61 to 65 among affected subjects, as many apoE-4 carriers may be dropping out of the unaffected sample during this age interval due to the onset of AD. This may also explain why there is little difference in apoE-4 allele frequencies between affected and unaffected subjects over the age of 70.

The apoE-4/4 genotype is also more common in unaffected subjects under 60 (frequency = 0.30) than in those over 60 (frequency = 0.11 overall, 0.08 to 0.15 in the various age strate). Nonetheless, the E-4/4 genotype does not invariably result in AD, even in quite elderly individuals with a strong positive family history of AD. For example, there are 18 families in which one or more unaffected subjects with the apoE-4/4 genotype has one or more affected siblings available for comparisons. In 11 of these families (with a total of 15 E-4/4 unaffected subjects), the E-4/4 unaffected subject(s) is (are) older than the age at onset of the affected sibling(s), sometimes by more than 10 years. Five of these E-4/4 unaffected subjects are over age 80. This is consistent with the role of apoE-4 as a risk factor rather than a "deterministic" gene for AD.

APM, sib-pair, and lod score analyses.

APM, sib-pair, and lod score analyses were then performed to confirm and further clarify the association findings. The overall APM statistic was 11.90 (using the 1/sqrt [p] weighting function), with a corresponding significance level of p < 0.0001. The sib-pair sharing was 0.53, also highly significant (p = 0.005). As expected, these results strongly support an effect of apoE on AD. The lod score analysis results are given in Table 4. All the dominant and recessive models provide similar lod scores (but at varying recombination fractions), ranging between 2.25 and 2.77. The co-dominant model provides lower overall lod scores.

Because all of these methods would be expected to give positive results solely on the basis of the established association between apoE-4 and AD, an additional set of analyses were performed to see whether the positive results were confined to families in which an apoE-4 allele was present. The families were divided into two groups: those families with no apoE-4 alleles (apoE-4 absent) and those in which at least one affected individual carried one or two apoE-4 alleles (apoE-4 present). If the observed positive results in the APM, sib-pair, and lod score analyses are due solely to the association of apoE-4 with AD, we would expect the positive findings to be negligible in the families in which apoE-4 is absent. The APM and sib-pair results, along with the peak lod scores (for any model, maximized over the recombination fraction) are presented in Table 5. The strong positive findings in the apoE-4 present group in all three analyses are consistent with the association between apoE-4 and AD. In the apoE-4 absent families, there is a significant APM result, a nominally significant sib-pair result, and a positive but nonsignificant lod score. Because only the APM result, which is notably subject to false-positives, is strongly significant, these findings could be the result of chance, or could be due to the protective effect of apoE-2 that has been observed in some datasets. ^[34] Because of the rarity of the E-2 allele, the present sample is too small to distinguish these possibilities.

Effect on age at onset.

To determine if the linkage and association results were limited to a specific subgroup of families, the families were stratified, by the mean age at onset of the affecteds, into six age strata, and APM, sib-pair, and lod score analyses were conducted (Table 6). The most striking result is the strength of the lod score results in the 61 to 65 year age group. This is consistent with the apoE-4 allele frequency peak within the same age group noted above.

Restricting all of these analyses to those families in which the spread in age at onset among the affecteds was <or=to10 years (74% of the overall sample) made only minimal differences in the results.

One affected person was then chosen at random from each family and categorized by apoE genotype in order to calculate the mean age at onset for each genotype. Individuals with the apoE-4/4 genotype had a significantly earlier mean onset (66.4 years versus 71.3 to 73.6 years) than any of the other genotypes (2/3, 3/3, 2/4, 3/4) (p < 0.001, using Student's t test; for this comparison all genotypes except apoE-4/4 were pooled). Individuals with one copy of the apoE-4 allele (genotypes 2/4 and 3/4) did not differ in mean onset age from those with none (genotypes 2/3 and 3/3). Interestingly, as noted in Table 6, the maximum lod score for linkage of apoE to AD with age at onset 61 to 65 was obtained assuming a recessive inheritance. This is consistent with the finding that the possession of two but not one copy of apoE-4 was associated with earlier onset of AD.

To determine if the effect of the apoE-4 allele on age at onset could be observed within families, as well as in the overall sample, we looked at the difference in the number of apoE-4 alleles between the individual in each family with the earliest onset and the individual with the latest onset. If apoE-4 has no effect on age at onset, the expected value of this difference is 0. The observed mean difference was 0.56 alleles (+/- 0.83), which is highly significant (p < 0.00001, Student's t test). We also looked specifically at the 40 pairs in which one individual had genotype 3/3 and the other had 4/4. In 37 of these pairs (93%), the 4/4 individual had an earlier onset. This was significant not only in the sample overall, but also within each age stratum (p < 0.0005), with the exception of the subjects over 80 years of age, in which only a trend was observed (p < 0.07).

Effect on age at onset.

The strongest effect of the apoE-4 allele occurs in the 61 to 65 age group, where the apoE-4 allele frequency is 0.62 (see Table 2). The results of the APM, sib-pair, and lod score analyses are also most strongly significant in the stratum of 33 families with mean age at onset 61 to 65 (see Table 6). This confirmation of the association findings is particularly important in the absence of independent age-stratified controls in the present study, as indicated above.

Both the allele frequency data and the APM results suggest that the peak falls off fairly steeply above 65, although a considerable effect is still apparent up to age 70. It is somewhat more difficult to assess the effect among subjects 60 and younger. The peak in the sib-pair and lod score results is also strongly focused in the 61 to 65 age-at-onset group. However, the allele frequency and APM data suggest that the effect remains strong below age 61. Unfortunately, the number of families with mean onset substantially under age 60 is too small to determine whether this finding results primarily from families with onsets close to age 60. The finding of peak association, APM, sib-pair, and lod score results in the 60s is consistent with the elevated apoE-4 gene frequency (and apoE-4/4 genotype frequency) in the unaffected subjects tapering off after age 60, as family members pass through the peak age of risk associated with apoE-4.

Our data indicate that having two copies of apoE-4 also appears to lower the age at onset of AD, confirming previous findings. ^[7,8,11,12] Whether selecting one individual from each family, or looking at differences within families, having two copies of the apoE-4 allele are associated with significantly lower age at onset. This effect was largely confined to individuals with the apoE-4/4 genotype. In contrast, having one apoE-4 allele was not associated with significantly earlier age at onset. The specific association of two but not one copy of the apoE-4 allele with earlier age at onset may explain why recessive as opposed to dominant models yielded the highest lod scores in AD patients with onset age 65 and under.

It is not possible to determine from the data collected to date whether apoE-4 acts as a modifier of age at onset, in individuals otherwise at risk, or as an independent risk factor whose more pronounced effect in this age group might be due to some intrinsic biological effect. In any case, the observed peak does not appear to be due solely to apoE-4's independent effect on mortality. Although there is some drop off in the frequency of the apoE-4 allele with age, numerous population-based studies suggest that the drop probably occurs beyond age 70. ^{[23-25,27-29,35,36]} In addition, because at least some of the studies of apoE gene frequencies at advanced ages ^[28,35] included primarily independent or cognitively unimpaired individuals, some fraction of the observed drop may be due to the underrepresentation of individuals with AD. ^[12,27]

Implications for diagnostic and predictive testing.

What are the implications of these findings for diagnostic or predictive use? These findings suggest that the addition of apoE testing might be somewhat helpful as an ancillary test for the differential diagnosis of demented patients in the 61 to 70 age range. In this age range, cases with AD represent a smaller proportion of the total number of dementia cases than at older age ranges. ^[37] Thus, competing diagnostic entities increase the difficulty of a correct diagnosis. Particularly within the 61 to 70 age group, a test indicating genotype apoE-4/4 might increase the probability somewhat that an individual with a progressive dementia indeed has AD, but it would by no means guarantee a diagnosis of AD. In particular, the likelihood would not increase sufficiently to obviate the need for a thorough neurologic and psychiatric evaluation, including laboratory and imaging studies. Thus, the use of apoE genotyping for diagnostic purposes, even in this limited age range where it may prove to be more informative, should not be adopted in clinical practice until it has been subjected to rigorous study under carefully monitored conditions.

As for predictive testing, although a finding of an apoE-4/4 genotype does suggest a higher probability of developing AD, and with an earlier onset, it remains premature to offer apoE testing as a predictive test for AD. Too little is known concerning the effect of the interaction of apoE-4 and other genetic and environmental risk factors (e.g., family history, head trauma, gender) on the development of AD, and on death from other causes. Before we can make use of these and other findings for predictive testing, it is critical to explore all facets of the apoE-4 effect, both as it changes with age and as it relates to other risk factors. These conclusions are similar to those of the American College of Medical Genetics/American Society of Human Genetics Workgroup on APOE ^[38] and the National Institute on Aging/Alzheimer's Association Working Group on ApoE Genotyping in Alzheimer's Disease. ^[39]

Implications for the genetics of late-onset AD.

These findings also underscore how much more there is to learn concerning the genetics of the late-onset form of AD. While virtually all of AD with onset before age 60 appears to be accounted for by the three known early-onset genes, ^[1,2,3] the picture is far less clear after age 60. While there is considerable evidence that genetic factors play a substantial role in late-onset AD, ^[40-43] apoE-4-the only identified gene with an effect in late-onset disease-appears to account for less than half of the genetic risk for AD beyond age 60. ^[44,45] The present study suggests that the apoE-4-associated risk for AD is focused on patients with age at onset under 70 years, yet the majority of AD occurs over the age of 70. Thus, it remains critical to search for additional genes involved in the development of AD beyond age 70.