TAU as a Susceptibility Gene for Amyotropic Lateral Sclerosis–Parkinsonism Dementia Complex of Guam

Background  A Guam variant of amyotrophic lateral sclerosis (ALS-G) and parkinsonism dementia complex (PDC-G) are found in the Chamorro people of Guam. Both disorders have overlapping neuropathologic findings, with neurofibrillary tangles in spinal cord and brain. The cause of ALS-G–PDC-G is unknown, although inheritance and environment appear important. Because neurofibrillary tangles containing tau protein are present in ALS-G–PDC-G, and because mutations in the tau gene (TAU) cause autosomal dominant frontotemporal dementia, TAU was examined as a candidate gene for ALS-G–PDC-G.

Methods  TAU was evaluated by DNA sequence analysis in subjects with ALS-G–PDC-G, by linkage analysis of TAU polymorphisms in an extended pedigree from the village of Umatac, and by evaluation of linkage disequilibrium with polymorphic markers flanking and within TAU.

Results  Linkage disequilibrium between ALS-G–PDC-G and the TAU polymorphism CA3662 was observed. For this 2-allele system, PDC and ALS cases were significantly less likely than Guamanian controls to have the 1 allele (4.9% and 2% vs 11.5%, respectively; Fisher exact P = .007). DNA sequence analysis of TAU coding regions did not demonstrate a mutation responsible for ALS-G–PDC-G. Analysis of TAU genotypes in an extended pedigree of subjects from Umatac showed obligate recombinants between TAU and ALS-G–PDC-G. Linkage analysis of the Umatac pedigree indicates that TAU is not the major gene for ALS-G–PDC-G.

Conclusions  The genetic association between ALS-G–PDC-G implicates TAU in the genetic susceptibility to ALS-G–PDC-G. TAU may be a modifying gene increasing risk for ALS-G–PDC-G in the presence of another, as yet, unidentified gene.

A UNIQUE FORM of amyotrophic lateral sclerosis (ALS) exists in the Chamorro people of the western Pacific island of Guam.¹ Guam ALS (ALS-G) is clinically similar to typical ALS found in other populations,² as both disorders have neuropathologic changes in the spinal cord with characteristic degeneration of the upper and lower motor neurons. However, unlike typical ALS, in ALS-G, neurofibrillary tangles (NFTs) occur in the hippocampus, entorhinal cortex, and neocortex.^2-4 A second disorder in the same population is parkinsonism dementia complex (PDC-G), an extrapyramidal syndrome with cognitive decline.^5,6 In PDC-G, as in ALS-G, extensive neocortical and hippocampal NFT pathology occurs with some spinal cord degeneration. About 5% of affected subjects have both ALS-G and PDC-G. Because of the clinical and neuropathologic similarities between ALS-G and PDC-G, and because both are found exclusively among Western Pacific groups including the Chamorro population, these 2 disorders may be different manifestations of a single disease.

The prevalence of ALS-G on Guam, when first described in the early 1950s,^7,8 was substantially higher than that of ALS in white populations (50-80 per 100 000 vs 1-6 per 100 000, respectively).^7,8 Similar prevalence rates were subsequently observed for PDC-G.⁹ The highest prevalence of ALS-G on Guam was in the small southern coastal village of Umatac (250 per 100 000).^7,8,10,11 In subsequent decades, the prevalence of both of these disorders declined, although neither has completely disappeared. Estimates of ALS-G prevalence for the 1970s and 1980s range from 30 per 100 000¹¹ to a low of less than 5 per 100 000.¹² Likewise, the prevalence of PDC-G has declined, but to a lesser extent.^9,13 During this same period, the age at onset for ALS-G increased from 47.6 to 51.9 years, and for PDC-G, from 42.1 to 52.2 years.¹⁴ Ongoing studies indicate that PDC-G is still common on Guam, but ALS-G is quite rare. Although the cause of ALS-G–PDC-G is unknown, the disease clusters in families and, therefore, inheritance may contribute to ALS-G–PDC-G susceptibility.^5,7,10,15-17 The change in incidence and onset age suggests that gene-environment interactions are important.

The NFTs in ALS-G–PDC-G are bundles of paired helical filaments composed of aggregated hyperphosphorylated tau. These NFTs are ultrastructurally and biochemically indistinguishable from NFTs in Alzheimer disease^18-20 and numerous other neurodegenerative disorders. Mutations in TAU, the gene that encodes tau protein, cause frontotemporal dementia with parkinsonism chromosome 17 type (FTDP-17).^9,21,22 This is an autosomal dominant disease with clinical and neuropathologic features that overlap with ALS-G–PDC-G.^9,23 Thus, TAU is a candidate gene for ALS-G–PDC-G. Some FTDP-17 TAU mutations are missense changes that alter the interactions of tau with microtubules.^24,25 Other mutations affect TAU alternative splicing regulation and result in a change in the ratio of the tau isoforms produced.^21,26-28 Mutations affecting splicing include both missense and silent mutations in TAU exon 10, and intronic mutations in the sequence closely flanking this exon.^21,26 In 1 autosomal dominant FTDP-17 family (the hereditary dysphasic dementia [HDDD2] kindred), genetic linkage analysis conclusively identifies the TAU region of chromosome 17 as the location of the responsible gene.²⁹ However, no mutation has been identified in this family despite extensive sequence analysis of all coding regions and intronic sequences surrounding each exon. Presumably, the mutation is in regulatory sequences either deep within an intron or in flanking regulatory sequences. Genetic changes in TAU also contribute to susceptibility to progressive supranuclear palsy (PSP). A positive genetic association between nonfamilial PSP and a polymorphism in intron (I) 9 (I9) of TAU was found³⁰ and has been confirmed in multiple studies.^31-33 Again, as with the HDDD2 family, no PSP susceptibility allele has been identified in coding regions or in flanking intronic sequences, and the susceptibility site is presumed to be in a regulatory sequence.

To investigate the role of TAU in ALS-G–PDC-G, we studied Chamorro subjects from the entire island of Guam and subjects from the village of Umatac.¹⁰ Subjects from Umatac were studied because of the high incidence of ALS-G–PDC-G in this village. TAU was evaluated as a candidate gene for ALS-G–PDC-G by DNA sequence analysis of coding regions and closely flanking intronic regions. Because some TAU mutations or susceptibility sites may be in cryptic regulatory sequences, we also performed linkage and association studies. The DNA sequence analysis and linkage analysis show that TAU mutations are not the major cause of ALS-G–PDC-G. However, association studies show that a TAU variant(s) confers susceptibility to ALS-G–PDC-G.

The ALS-G and PDC-G cases and controls were ascertained from the University of Guam, Mangilao, registry, recruitment at Guam Memorial Hospital, Oka, Tamining, and from the National Institute of Neurological and Communicative Disorders and Stroke Guam Intramural Research Program that operated from 1956 to 1983. Subjects from the current study were examined by a trained research assistant who obtained a medical history and performed a brief examination that included neurologic screening and cognition testing with the Cognitive Abilities Screening Instrument (CASI). The CASI was pilot tested on Guam to ensure cultural fairness, to assess the extent of variation due to age and education, and to develop cutoff points for screening. Information was collected on risk factors, family history, medication use, medical and neurologic history, cognitive and motor symptoms, and functional performance (of activities of daily living). All subjects who failed screening with the CASI or whose history or brief examination suggested neurologic disease underwent detailed evaluation. In the more extensive evaluation, a neurologist examined mental status and performed a structured neurologic examination that included the Unified Parkinson's Disease Rating Scale. A psychometrist administered a standardized test battery for which normative data have been obtained from elderly Chamorro subjects who lacked significant medical or neurologic diagnoses. Laboratory data such as blood test results were reviewed. Neuroimaging studies were recommended (and could be ordered by personal physicians) if subjects had unusual clinical pictures or focal neurologic findings. Consensus diagnoses were made by 3 neurologists, who reviewed all clinical information. Diagnoses were made at 2 levels: descriptive syndromes (such as dementia, parkinsonism, ALS, stroke, or other conditions) and suspected etiologic diagnosis (such as ALS-G or PDC-G). Standard clinical diagnostic criteria were used wherever possible. Parkinsonism arising after neuroleptic exposure or in severe stages of dementia was considered to be secondary. The diagnosis of PDC-G required the insidious onset and gradual progression of primary parkinsonism and dementia. Modifying factors such as stroke or other brain diseases were taken into account. Dementia was diagnosed with criteria of the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition.³⁴ The pure dementia syndrome on Guam resembled Alzheimer disease clinically, and many patients met standard criteria for probable Alzheimer disease. Subjects from the National Institute of Neurological and Communicative Disorders and Stroke studies were all cases of clinical and autopsy-documented ALS-G and/or PDC-G. Ethnic Chamorro controls were diagnosed as being normal if they had a CASI score of 80 or higher, were functionally independent, and lacked motor weakness, tremor, or gait difficulty on brief screening. Subjects whose CASI scores were below 80 were designated controls if detailed neurologic and psychometric evaluation did not show evidence of dementia, mild cognitive impairment, or a neurologic disorder. Genomic DNA was prepared as previously described.^35,36 DNA was either from archived frozen brain samples from previous studies or from fresh blood samples obtained by the ongoing project.

Subjects were genotyped for a previously described short tandem repeat polymorphism (STRP) Tau-CA³⁰ and 3 new STRPs in or near TAU (Figure 1). The STRPs located 5′ and 3′ to TAU were identified by searching for dinucleotide repeat sequences in the bacterial artificial chromosome clones HCIT104N19 (Genbank accession No. AC003662) and HRPC843B9 (Genbank accession No. AC004139), which contain TAU 5′ and 3′ flanking sequences, respectively. The STRP CA3662 was amplified with the primer pair 3662-23F 5′ GGC CGG TAA GAG ATC AGC AAA C and 3662-23R 5′ AAC AGC AAG CAG TAT ATA CC, yielding a product of approximately 230 base pairs (bp) in size. The STRP Tau-GT was amplified with the primer pair GTF 5′ CTT CAC TCT CGA CTG CAG C and GTR 5′ GAC AGC GGA TTT CAG ATT CGG, yielding a product approximately 262 bp in size. The STRP CA4139 was amplified with the primer pair 4139 CA1F 5′ CGA GAT GGT ACC ACT GCA CTC C and 4139 CA1BR 5′ CTG ATA GCA TGT CTT CTA GAA C, yielding a product approximately 140 bp in size. Genotypes were determined by previously described methods.³⁷

Six ALS-G and 5 PDC-G cases were used for TAU mutational analysis. Six of the sequenced cases were documented by autopsy, including 1 case each of ALS-G and PDC-G from Umatac. Both strands of the gene were sequenced for all coding exons (exons 0, 1-4, 4a, and 5-13), including at least 7 bp of flanking intronic DNA, 50 bp of the 5′ untranslated region, and 70 bp of the 3′ UTR. Primer pairs for amplification and sequencing were previously described,³⁸ with the exception of the following new primer sets: exon 10 (10GF 5′ GTC AGT GTG GCC GAA CAC and 10CR 5′ GGC TAC ATT CAC CCA GAG G) and exon 11 (11EF 5′ TGC TTC TCA TTG AGT TAC ACC and 11ER 5′ TTG TCT TGG GCA GCA TGG CC).

DNA sequence analysis identified three 2-allele polymorphic sites in subjects with ALS-G–PDC-G. One, in I9, is a 2-allele C/A polymorphism 47 bp (I9-47) from the 3′ end of the intron. The I9-47 polymorphism was genotyped by the following restriction digest assay. A 313-bp fragment was amplified by polymerase chain reaction using primers 10CR (5′ GCT ACA TTC ACC CAG AGG) and 10F (5′ AAG TGG AGG CGT* CCT TGC GGC CAA GC), where the underlined GC differs from the normal-sequence AG found in TAU. This sequence change creates a BglI site when a C is present at the base followed by the asterisk, but not when an A is present. Digestion with BglI produces 287- and 26-bp fragments for the C allele and an undigested 313-bp allele for the A allele. Two additional variable sites, one in I9 (I9-176) and another in I11 (I11 + 90) were genotyped by DNA sequence analysis as described herein.

Tests for possible frequency differences between cases and controls or between observed and expected frequencies were performed with χ² tests. Because previous results demonstrated that combining alleles to collapse cells into a 2 × 2 contingency table can give spurious results,^39,40 all contingency table analyses were performed on full 2 × m contingency tables, where m is the number of alleles, haplotypes, or other groups being compared. A Fisher exact test was performed for the STRP, and exact P values were estimated. Pairwise tests were performed when the 3 populations differed (ALS-G, PDC-G, and Guam controls).

Linkage analysis was performed for single markers plus the disease by means of a disease allele frequency of 0.01 and a dominant mode of inheritance with a 0% sporadic rate. Age-dependence penetrance for the carrier genotypes was assumed by means of a cumulative normal function with a mean of 49.4 years and a variance of 96.2 years. These values corresponded to the mean and variance observed in the affected members of the entire 1167-subject Umatac pedigree. Linkage analysis of the full pedigree with its many inbreeding loops was impossible for computational reasons. Linkage analysis of the largest subcomponent of the pedigree that both traced back to a single founder couple and that could be feasibly analyzed was performed by means of the program FASTLINK,⁴¹ version 4.1 (Rockefeller linkage analysis software; available at: http://linkage.rockefeller.edu/soft/list.html), which implements computationally efficient methods for automatic loop breaking.⁴² For a single marker on this subcomponent of the pedigree, the programs Unknown and M link (within the FASTLINK software) took 11 hours and 11.35 days to run, respectively, with the use of a Digital Alpha XP1000 500-MHz workstation (Compaq Computer Corporation, Houston, Tex).

TAU was first evaluated as a candidate gene for ALS-G–PDC-G by comparing STRP allele frequencies of Chamorro cases with those of Chamorro controls. Cases were 49 subjects with ALS-G (41 with autopsy confirmation), 86 subjects with PDC-G (49 with autopsy confirmation), and 3 subjects with ALS-G–PDC-G (2 with autopsy confirmation). Chamorro controls (n = 78) were healthy nondemented subjects with no family history of ALS-G–PDC-G. Subjects for the genetic association studies were from the entire island of Guam, except that subjects from Umatac were not used. Four STRP sites were used including 2 that are within TAU (Tau-GT and Tau-CA) and 2 that closely flank the 5′ and 3′ end of TAU (CA3662 and CA4139, respectively; Figure 1). Allele and genotype frequencies for CA3662 and Tau-CA are in Table 1 and Table 2. Allele frequencies for CA3662 significantly differed between ALS-G, PDC-G, and controls (P = .007 among the 3 groups), where there was an underrepresentation of the rare allele (allele 1) and excess of the common allele in both case groups when compared with controls (Table 1). Pairwise tests performed between PDC-G cases and controls and ALS-G cases and controls were also significant (P = .04). These results demonstrate that both ALS-G and PDC-G have an association with a TAU allelic variant.

There was no statistically significant difference in allele frequencies for the Tau-CA (Table 2), Tau-GT, or CA4139 (P = .09 and P = .27, respectively; data not shown) when Chamorro subjects with ALS-G and PDC-G were compared with Chamorro control subjects. The lack of association between ALS-G–PDC-G and these sites indicates either a lack of power, or that the regions of TAU tested by these markers are not in linkage disequilibrium with the disease.

The genetic association results suggest that TAU could be either a susceptibility allele or a major gene for ALS-G–PDC-G. To identify potential TAU mutations responsible for ALS-G–PDC-G, all coding exons with some flanking intronic regions and portions of the 3′ and 5′ untranslated regions were sequenced for 6 subjects with ALS-G and 5 subjects with PDC-G, including 1 case of each type from Umatac. No sequence variants were identified in coding regions. Three sequence variants were found in intronic regions. Two sites were in I9, a G/A polymorphism at −176 (numbers are relative to the first nucleotide of E10) and a C/A polymorphism at −47. The third polymorphic site was a G/A polymorphism in I11 at +90 nucleotide (after exon 11). To determine whether these variable sites are potential mutations or benign polymorphisms, Chamorro controls were also examined for these 3 sites. For each site, both alleles were also found in Chamorro and white control subjects, indicating that these variants are not causative mutations.

Subjects for family linkage analysis were from Umatac (Table 3, Figure 2). As described in 1969,¹⁰ the familial relationships for the entire village of Umatac dating back to 1830 could be represented as a single pedigree of 1450 people in 262 sibships over 8 generations, including 53 subjects affected with ALS-G–PDC-G. Subsequently, additional subjects with ALS-G–PDC-G were ascertained, including the 16 described in Table 3. This includes 2 PDC-G cases and 1 dementia case identified by this study in the past 3 years. Figure 2 represents only the portion of the pedigree containing affected subjects for whom DNA was available. Also included are relatives needed to connect the sampled affected subjects. For each sampled affected subject, parental relationships, shown as different colored lines, were traced back to 6 couples who could not be connected in the earliest known generations (not shown). The complexity of the pedigree reflects the inbred nature of this village. Because the present Umatac population was descended from a small number of original families (8 in 1830) and the community was genetically isolated, and because most of the affected subjects are concentrated in a subset of the Umatac families,¹⁰ ALS-G–PDC-G in this village may be the result of a genetic founder. The identification of new cases of PDC-G demonstrates that this disease has not disappeared from Umatac.

Fourteen affected Umatac individuals (8 PDC-G cases, 5 ALS-G cases, and 1 dementia case) were genotyped for the 4 TAU STRP sites (Figure 2). Genotypes for the Tau-GT and CA4139 STRP markers are displayed on the pedigree. Multiple discordant genotypes are present between affected subjects for both disease types, indicating obligate recombinants between TAU and both ALS-G and PDC-G. For example, ALS-G subjects 286 and 184 are 4/6 and 2/5, respectively, for CA4139 and thus do not share a common genotype for TAU. Both are descended from the same 2 couples (couple 1/2 pink and couple 040/041 orange). For CA4139, PDC-G subjects 500 and 538 (genotypes 2/6 and 2/4, respectively) are discordant with subject 721 (genotype 5/7), even though subjects 500 and 721 are descendents of couple 4/3 (blue). Assuming a simple major gene model of inheritance, these discordances indicate obligate recombination events between TAU and ALS-G–PDC-G.

Linkage analysis of the Umatac pedigree was also performed with tau-GT, the most informative STRP in TAU. Because of the extensive inbreeding in this pedigree, only a subset of the sampled subjects could be included in the analysis (Figure 2). All resulting lod scores were negative, with a lod score of −0.025 at a recombination fraction of 0. Thus, the pattern of marker genotypes among cases combined with the linkage analysis indicates that TAU is not the major gene responsible for ALS-G–PDC-G in the Umatac kindred.

In this study, TAU was evaluated as a candidate gene for ALS-G–PDC-G. The underlying hypothesis for this work is that inheritance contributes to susceptibility to this disease complex. The following lines of evidence support a genetic hypothesis. First, close relatives of affected subjects are at greater risk for ALS-G–PDC-G than are close relatives of controls, and approximately 40% of probands have an affected relative.^7,8,17 Second, families have been described with multiple affected subjects in 2 or more generations.¹⁷ The most dramatic of these pedigrees consists of subjects from Umatac, where the disease can be traced back 4 generations (Figure 2),¹⁰ a pattern consistent with a genetic cause. Third, segregation analysis of an ALS-G–PDC-G registry of patients and relatives rejected an environmental factor–only hypothesis.⁴³ However, an inheritance-only model is also not consistent with how ALS-G–PDC-G occurs. An environmental hypothesis is supported by the fact that, for both ALS-G and PDC-G, during the past 30 to 40 years, the age at onset has increased^13,14 and the incidence decreased,^12,44 suggesting a change in exposure to an environmental factor. Candidates for nongenetic factors include aluminum deposition,^45,46 excess aluminum absorption caused by low dietary calcium,⁴⁷ and 2-amino-3-(methylamino)-propanoic acid, a neurotoxin found in traditional foods,⁴⁸ although the evidence that these factors are important in causing ALS-G–PDC-G is equivocal.^49-51 The most parsimonious hypothesis is that ALS-G–PDC-G is the result of interaction between genetic and environmental susceptibility factors, and exposure to this unknown environmental factor has decreased in the past 30 to 40 years.

The mode of inheritance of susceptibility to ALS-G–PDC-G is difficult to predict. Visual inspection of the Umatac pedigree (Figure 1)¹⁰ suggests autosomal dominant inheritance with near-complete penetrance in males and reduced penetrance in females. Recessive inheritance is unlikely, since the mean inbreeding coefficient was similar for Umatac sibships with and without affected subjects.¹⁰ Formal segregation analysis of probands and close relatives for subjects in a registry from the entire island rejected both dominant and recessive models, but was consistent with a 2-allele additive major locus model.⁴³ However, the penetrance of this major locus is low, with a maximum liability of 0.1 for homozygous carriers. Certainly, more complex models are possible, and a major gene hypothesis can also include modifier genes.

The linkage disequilibrium results with CA3662 demonstrate that TAU is a susceptibility gene for the disease. Although the results from the other 3 sites tested were not significant, data from Tau-CA are intriguing (Table 2). For this site, the 142 allele (also called the Ao allele) was previously reported to be elevated in white patients with PSP, with a frequency of 0.98 in cases and 0.75 in white controls, making it the high-risk allele for PSP.³⁰ For Chamorro subjects, both cases and controls had high frequencies of the 142 allele, comparable with that of white patients with PSP. Allele 142 frequencies were higher in ALS-G–PDC-G cases than Chamorro controls, but this difference did not reach statistical significance. Interestingly, Japanese control subjects also have a high 142 allele frequency of 0.98³² to 1.0,⁵² suggesting that Chamorros are more similar to Asians than to whites. Previous work⁵³ attempting to identify a genetic association between ALS-G–PDC-G and TAU-related polymorphisms failed to detect the linkage disequilibrium observed here. This is not surprising, since the markers used in the previous study were not in proximity to TAU. None of the markers previously used are in the genomic sequence available for TAU (Figure 1), indicating that these markers are at least 70 kb upstream of the 5′ end of TAU and 200 kb downstream from the 3′ end of the gene. Regions of linkage disequilibrium do not typically extend over distances this great, and thus, failure to detect disequilibrium with distant markers does not exclude the involvement of TAU in ALS-G–PDC-G. Also, the sample size used in the previous study was small (23 subjects with PDC-G, 19 control subjects, and no subjects with ALS-G) compared with the subject panel used here (Table 1 and Table 2).⁵³

Although the genetic association studies indicate that TAU could be a susceptibility gene for ALS-G–PDC-G, TAU does not appear to be the major gene responsible for the disease. Direct DNA sequencing of all TAU coding exons present in brain tau isoforms did not demonstrate any mutations or polymorphisms, either in controls or in affected subjects, consistent with DNA sequencing studies by others.⁵³ However, absence of variants in these exons does not exclude TAU as a major gene or a modifying susceptibility gene, because genetic variation in noncoding regions not directly adjacent to exons can cause disease. This has clearly been demonstrated for the HDDD2 family, where linkage analysis has clearly demonstrated that TAU is the major gene responsible for disease,²⁹ yet no mutation in exons or in closely flanking intronic regions are known for this family. Another example is PSP, where TAU is clearly a modifying locus as demonstrated by linkage disequilibrium analysis,³⁰ yet, as for HDDD2, no coding or closely flanking intron variants are known. For both PSP and the HDDD2, presumably sequence variants in noncoding regulatory regions that control gene expression or alternative splicing are pathogenic. TAU regulatory mutations are known, and these cause FTDP-17 by affecting alternative splicing of exon 10. These mutations are located in intron 10, close to the 3′ end of TAU exon 10,^21,22,27 and in exon 10.⁵⁴ Since regulation of exon 10 alternative splicing is complex and not completely understood,⁵⁴ additional cis-acting sequences may exist elsewhere in the gene, and these regulatory elements are candidate locations for mutations and susceptibility sites. Regulatory sites that control other aspects of TAU gene regulation scattered throughout the gene are also potential locations for a mutation-susceptibility site. Thus, the fact that there are no mutations or polymorphisms associated with ALS-G–PDC-G in TAU coding regions does not exclude this gene from being involved in Guam neurodegenerative disease.

The pattern of TAU segregation in the Umatac pedigree is also not consistent with TAU being the major gene for ALS-G–PDC-G (Figure 2). Numerous obligate recombinants demonstrate that no specific TAU allele is required to develop ALS-G–PDC-G, though a specific TAU allele increases susceptibility to the disease. This conclusion is based in part on the hypothesis that all cases of ALS-G–PDC-G are the result of the same inherited allele(s) or mutation. This is a reasonable assumption considering that this is a rare disease found only among the Chamorros from the Mariana Islands, Japanese from the Kii Peninsula in Japan,⁵⁵ and the isolated Auyu and Jakai people of West New Guinea,⁵⁶ although no autopsy information is available for the latter population. In addition, the Chamorro population, originally 50 000 to 150 000 at the time of the first contact with Europeans in 1521, was reduced to approximately 2500 in 1830s and further reduced by a smallpox epidemic in 1856.⁵⁷ This population history suggests that ALS-G–PDC-G may come from a single or small number of genetic founders, particularly in the high-incidence village of Umatac.

The linkage disequilibrium results presented here indicate there may be a polymorphic site(s) in or closely linked to TAU that increases risk for ALS-G–PDC-G. The polymorphisms tested are unlikely to be the actual pathogenic site. Presently, it is not possible to predict the actual location of the critical TAU susceptibility site. Because of intrinsic properties of linkage disequilibrium, lack of significant results for a particular marker does not exclude the region containing that marker. Thus, even though significant results were obtained with a marker 5′ to TAU, presently the entire gene must be considered for the susceptibility site. The critical regulatory sequence affected may be within an intron or in flanking regions on either side of the gene. Since the flanking genes, CRFR and KIAA1267, are unrelated to TAU in terms of sequence homology or obvious predicted function, TAU is not part of a cluster of related genes. Thus, the susceptibility site is part of the TAU gene including flanking regulatory sites most likely located between CRFR and KIAA1267. Since the entire gene is 130 kb and flanking regions include 50 kb 5′ to the gene and 3 kb 3′ to the gene, significant additional sequencing comparing affected subjects with controls is required. Identification of the major gene(s) for ALS-G–PDC-G will require linkage analysis of ALS-G–PDC-G in large families such as the Umatac pedigree.

Accepted for publication May 4, 2001.

This study was supported by grant PO10135316 from the National Institute on Aging, National Institutes of Health, Bethesda, Md, and by the Department of Veterans Affairs, Washington, DC.

We thank the subjects who contributed immensely to this work. We also thank the staff at the University of Guam for assistance in this work.

Corresponding author and reprints: Gerard D. Schellenberg, PhD, GRECC 182-B, Veterans Affairs Puget Sound Health Care System, 1660 S Columbian Way, Seattle, WA 98108 (e-mail: zachdad@u.washington.edu).