Index | Round
Tables | Selected
Abstracts | News
Search for Late-Onset AD Genes
by Walter Lukiw
Note: This article was revised on 10 August, 1998, to accommodate some factual corrections. See also R. Tanzi's comment, appended below.
22 July 1998. A study published last year by Margaret Pericak-Vance et al. (JAMA 278; 1997) reported a putative locus for late-onset Alzheimer's disease (LOAD) at chromosome 12p11-12, and set off a race to confirm and nail down the gene. The high interest in the search for LOAD genes even attracted a TV news camera crew to this session.
C. Van Broekhoven (abstract 957) underscored the importance of searching for genetic factors for late-onset AD, pointing out that the genes on chromosomes 1, 14 and 21 linked to early-onset Alzheimer's (EOAD) collectively account for ~15% of all EOAD, or less than 5% of all AD cases. She reviewed the APP mutations which affect APP processing and the importance of A-beta42 in AD pathology, as well as the role of PS mutations in EOAD. Out of the 40 or so scattered mutations in PS1, she observed, most mutations are linked to increases in A-beta42 secretion. However, she added, PS1 delta4 mutations do not affect A-beta42 secretion. Van Broekhoven also reviewed presenilin polymorphisms -- many more in PS2 than in PS1 -- and noted PS2 polymorphisms are not associated with AD. In her studies, the candidate locus at chr 12p11 was not totally reconfirmed in LOAD, reflecting the (still) common fnding of false positives or genes which have ony a minor effect in LOAD.
Rudy Tanzi (abstract 961B, not available on-line) presented his group's evidence linking the alpha-2 macroglobuling gene to late-onset AD. 350 sibling pairs from the NIMH genetics screening initiative were screened and showed hits on chromosomes 3,4,6,9 and 12. Chromosome 12 studies focused on the LRP receptor (for secreted APP, alpha-2 microglobulin and APOE). The alpha-2 macroglobulin (A2M) gene product is particularly interesting because it is a pan-protease inhibitor which tightly binds beta-amyloid, preventing fibril formation and neurotoxicity, possibly facilitating beta-amyloid clearance and degradation. D12S391 is near A2M and near the tip of chromosome 12p. Using sib-TDT or SDT, the linkage of A2M to AD had an odds ratio of 3.56 with a p-value of 0.00009, compared to an odds ratio of 3.45 and a p-value of 0.00006 for the APOE-e4 allele. Thus, A2M in this study has as strong an association with AD as does APOE-e4. Tanzi proposed that A2M mutations may lead to lifelong increased accumulations of beta-amyloid. He also noted that LRP and A2M are at the same locus at chromosome 12p and both may be involved in LOAD.
Tanzi's finding received support from Alison Goate's group, which reported modest evidence of linkage to a region of chromosome 12 containing the A2M gene (abstract 959). Goate's study involved 230 families consisting of 510 affected and 97 unaffected individuals using multipoint sib pair analysis (MAPMAKER). Although data points were obtained indicating involvement to the same general area at 12p, it must be significantly smaller than the locus for APOE on chr 19. Goate thought it was encouraging that all groups were detecting linkage to chromosome 12, and said the differences in location were probably due to inaccuracies in pin-pointing a gene using this methodolgy in a disease like late onset AD. "The gene we are detecting could be A2M, LRP or another gene in the region," she said.
In her review of current research, M.A. Pericak-Vance pointed out problems and new approaches in identifying genes associated with late-onset Alzheimer's disease (abstract 956). Individual case /control studies are easy to perform when large data sets are available, and were instrumental in the identification of APOE-e4, for example. However case/control studies are much less successful when anomalies in population selection are factored in. Positive associations can be missed due to population heterogeneity, population-specific effects and chance selection. The transmission disequilibrium test (TDT) can often, but not always, avoid such sampling bias by using internal parental controls. The major drawback is that it can't be used for late-onset diseases, because parents are usually deceased and not available for genotyping. Using unaffected siblings as controls (sib-TDT; as delineated by Spielman and Ewens (1998) and others) compares marker allele frequencies in pairs of affected and unaffected siblings. Such studies to test linkage and association more thoroughly are needed to dissect the inherent complexity.in AD and other late-onset diseases.
Several presenters discussed other candidate genes for LOAD. M.-C. Chartier-Harlin et al. (abstract 958) described their studies examining polymorphisms in APOE gene promoter. The Th1/E4cs polymorphism and the Th1/E47cs polymorphism are highly significant in modulating the APOE4 allele impact on LOAD (each ANOVA < 0.0001). There is a moderate association between the LDL receptor (LRP) locus on chromosome 12 and LOAD, but it is not nearly as significant as APOE with LOAD. Other genes at this chromosome 12 locus may be also in the vicinity and generate linkage disequilibrium and explain part of the association of chromosme 12 with LOAD.
E. Tourneier-Lasserve et al. (abstract 960) discussed recent findings relating to CADASIL (cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy), an adult-onset autosomal dominant condition that involves recurrent ischemic strokes, vasculature abnormalities and associated subcortical dementia. EM reveals dense granular osmophilic deposits in the vasculature. CADASIL has been linked to the human homolog of Notch 3, which was just recently cloned: a 33-exon, 30-kb monster gene. Screening of 55 unrelated CADASIL patients showed that at least 70% of all mutations were located in the 5' end of the Notch 3 gene in exons 3 and 4, corresponding to the first five EGF-like repeats. Nearly all of these mutations led to creation or destruction of a cysteine residue in the EGF-type repeats The resultant mutations may produce an aberrant dimerization of Notch proteins into granual, osomophilic and aggregated forms and may underly the pathophysiological mechanism which leads to the CADASIL phenotype.
The most novel finding was presented by Fred van Leeuwen et al. (abstract 961C, not available on-line), who described his theories on "gene information assembly line slippage" generating mutant RNA forms which are subsequently translated into mutant dysfunctional proteins. Loss of 2 bases (GA) garbles the RNA message.(see TINS 21:331-335 ,1998) ultimately generating "+1" proteins. Altered "GAGAG" motifs are found in the RNA messages for beta-APP, tau, ubiquitin B, apoE4, PS-1, PS-2, MAP 2b, NF-L, NF-M, and NF-H (neurofilaments light, medium and heavy chain proteins, respectively), indicating potential involvement of all of theses species in the "molecular misreading" hypothesis. Van Leeuwen's theory offers an explanation of a wide range of "pathological" molecules in the pathophysiology of neurodegeneration. The molecular mechanism for "molecular misreading" is still not known.
To address the question whether the +1 proteins a cause or a consequence of AD neurodegeneration, the following data were reported. In young Down's syndrome cases, APP+1 is present when amyloid deposition is not apparent, suggesting that "molecular misreading" is an early event in neurodegeneration. Data suggest that mRNA surveillance may be a common mechanism in which the accuracy (of mRNA surveillance) may decrease during aging and hence account for the accumulation of aberrant proteins in the aging brain . The biological basis for +1 protein generation, accumulation and degradation are currently unknown, but should fuel abundant future research. This intriguing, refreshingly new and welcome hypothesis could explain a lot of "non genetically linked" or sporadic/idiopathic neurodegenerative disease. (See also Van Leeuwen On-line Journal Club.)
Comment from R. Tanzi -(7 August 1998)
Dr. Lukiw does not mention that Dr. Peggy Vance reported that she replicated our significant genetic association of A2M and AD using the same A2M polymorphism (exon 18 deletion) in the same AD families (NIMH AD family set N = ~350) that we used and with the same family-based assocation analysis technique. She also presented data indicating that the much smaller set of Duke families (N = ~60) they tested subsequently did not yield a significant p-value. It is not clear that the Duke families had sufficient power to yield a significant p-value since the test they used (the S-TDT) can only use one affected and one unaffected individual per family. Dr. Vance presented no power calculations. Thus, this was not shown to be a sufficiently sized confirmatory family set (in contrast to the large set of NIMH AD families which were positive in both our and Dr. Vance's analyses). When I questioned Dr. Vance as to whether the family-based association p-value remained significant for the association of A2M and AD in the combined NIMH and Duke AD families, she answered affirmatively.
I would also like to point out that in Dr. Vance's linkage analyses, heterogenity-based tests yielded the highest score of 4.6 at the marker D12S358 which resides right next to A2M. This is consistent with linkage of AD to A2M in the Duke family set. Also, Lindsay Farrer and Peter Hyslop presented a peak NPL score of 2.5 at D12S358 (by Genehunter analysis) further supporting the candidacy of A2M. They also presented another impressive linkage peak at markers near the gene for LRP on the long arm of chromosome 12. Thus, overall, the data presented by the groups of Hyslop (and Farrer), Vance (and Haines), and Goate (and Hardy) primarily supported and certainly did not refute linkage of A2M to AD as suggested by Dr. Lukiw in his summary. The only negative data presented at the entire meeting was the lack of a significant p-value for A2M and the smaller Duke family set (60 families versus 350 in the NIMH set which was positive in two seperate analyses-our and Duke;s). A proper replication set will require at least a couple of hundred discordant sib-pairs.
It should also be mentioned that while both Hyslop and Farrer stated that a case-control study did not reveal an association between A2M and AD, no actual data were presented. We did not perform a case-control study, only a family-based association (since it is not prone to type I error due to population stratification). However, Dr. Bill Rebeck presented data from Dr. Brad Hyman's showing that another A2M polymorphism (V1000I)was associated with AD in a large case-control study.
Finally with regard to the marker D12S391 mentioned by Dr. Lukiw in his summary, it maps roughly 20 cM below A2M (quite far genetically) and was previously published as the best linked chromosome 12 marker to AD by the Duke group. Our linkage data, along with those of Goate, and those of Hyslop did not show significant genetic linkage of this marker or to AD. In fact, our data and Goate's were quite negative at this locus. In a poster from Dr. Schellenberg's group, negative data were also present for D12S391 and the entire Duke region of chromosome 12. Dr. Schellenberg did not present data for markers near A2M. This news summary states that the lack of confirmation for Dr. Vance's linkage of D12S391 to AD by Goate, Schellenberg and our group may be due to problems with case-control studies. This is erroneous, because this marker was assessed only by traditional genetic linkage analyses (LOD, MLS etc) and not by "association" methods usually applied to candidate genes.