Direct genetic evidence for involvement of tau in progressive supranuclear palsy

Progressive supranuclear palsy (PSP) was first described by Steele et al.¹ in 1964. Clinically, it is a progressive neurodegenerative disorder presenting in persons older than 40 years as an akinetic, rigid syndrome with vertical gaze palsy associated variably with dysarthria, dysphagia, axial rigidity, evidence of pyramidal tract dysfunction, and a mild dementia (usually of the frontal lobe type). Pathologically, PSP typically shows generalized cerebral atrophy with pallor of the substantia nigra and shrinkage of the globus pallidus. There is neuronal loss and gliosis, as well as the presence of numerous neurofibrillary tangles (NFTs) and neuropil threads. This occurs predominantly in the superior colliculus, periaqueductal gray matter, pretectal areas, the zona compacta of the substantia nigra, and the pallidosubthalamic complex.² In PSP, the NFTs consist of clusters of straight filaments, compared with those comprised of paired helical filaments seen in AD patients. However, there is a considerable overlap with the pathology of other tau-related disorders such as postencephalitic parkinsonism and the ALS-PD dementia complex (APDC) of Guam.³

Because of the frequency of tau pathology in neurodegenerative disease, a number of hypotheses have been proposed in an attempt to explain how tau could lose its normal affinity for microtubules and associate instead to form insoluble aggregates.^4-6 The first direct evidence for a mechanism was provided recently by Conrad et al.⁷ and indicates that a defect within the tau gene itself may be involved, at least in PSP. This study first identified a polymorphic GT repeat in intron nine of the tau gene, and subsequently demonstrated a highly significant overrepresentation of one allele (A0) of this marker in a series of 22 pathologically confirmed patients with PSP. These findings were subsequently confirmed by Higgins et al.⁸ in a set of 24 clinically ascertained PSP patients.

Multiple system atrophy (MSA), along with PSP, is one of the most common causes of the misdiagnosis of idiopathic PD.⁹ The neuropathology of MSA is distinct from PSP, and the presence of ubiquitin-positive glial cytoplasmic inclusions is believed to be an early and specific pathologic marker for MSA.¹⁰ Because glial cytoplasmic inclusions have been found in PSP, and PSP can mimic MSA, we wished to establish whether there was any evidence for association of the A0 allele in MSA.^11,12

In this article we describe the use of a refined methodology (automated fluorescent as opposed to a manual isotopic-based technique) to investigate this marker in a series of 30 clinically ascertained PSP patients and 35 MSA patients collected from a number of European centers.

Methods. Subjects. The total sample population consisted of 30 PSP patients (mean age, 69.4 years; range, 60 to 83 years; 18 men and 12 women), 35 MSA patients (mean age, 63 years; range, 49 to 79 years; 14 men and 21 women), and 70 control subjects (mean age, 66.1 years; range, 52 to 83 years; 34 men and 36 women) recruited from four neurological centers in Italy and one in the United Kingdom. All patients were personally examined by at least two neurologists, and reviewed (clinical notes, video footage, neuroimaging data) by groups from at least two participating recruitment centers. All staff involved in this process were members of the European Study Group on Atypical Parkinsonism (ESGAP) Consortium and had specific interest and experience in movement disorders. The diagnostic criteria used for PSP and MSA were those of Golbe et al.¹³ and Quinn.¹⁴ Control subjects were matched for age (±3 years), sex, ethnicity, and geographic location of origin.

DNA isolation and marker analysis. Total genomic DNA was isolated from approximately 400 µL of whole blood using a Dyna Beads DNA Direct system II kit (Dynal, Liverpool, UK) as per the manufacturer's instructions Individuals' allelotypes were then ascertained by automated analysis of fluorescently labeled PCR products spanning the marker GT repeat sequence. PCR reactions used approximately 50 ng DNA template and were performed under the following conditions: one cycle at 94 °C for 2 minutes, followed by 26 cycles at 93 °C for 30 seconds, 58 °C for 30 seconds, and 72 °C for 40 seconds. Primers TauF2* (5′-GCCTCGCAAATTGCTGGGAT [5′-Hex labeled]) and TauR1 (5′-AGGTGACTGGGTAGAGACAGAGC) were used at a final concentration of 1 µM. Mg2+ concentrations of 0.8 and 2 mM were used with Biotaq and Biotaq Diamond DNA polymerases (Bioline, London, UK) respectively. (Note that Biotaq Diamond was used for certain samples where it was found to provide slightly higher yield and amplification specificity.) The same reaction buffer was used for both enzymes, and was as supplied by the manufacturer. PCR products were analyzed on an ABI Prism 377 DNA Analyser (Applied Biosystems, Warrington, UK). All genotyping was performed blinded to clinical status.

Data analysis. Because our objective was to confirm or refute the findings of Conrad et al.⁷-namely, that the A0 allele was overrepresented in PSP patients compared with control subjects-the frequencies of this allele and of A0/A0 homozygotes were compared with the sum of the other allelic and genotypic frequencies. Given the prior evidence for a direction of effect, this allowed the data in the table to be condensed into 2 × 2 contingency tables and analyzed using one-tailed Fisher's exact tests. This provides the most powerful statistical test for the prior hypothesis of overrepresentation of the A0 allele in PSP patients compared with control subjects.

Results. Use of the fluorescence-based methodology permitted successful, rapid, and reliable allelotype determination for all individuals in the sample population. Typical examples of electrophoreograms are shown in the figure. Despite the presence of five known alleles, A0 through A4 containing 11 to 15 GT repeats respectively, only the considerably more frequent A0, A3, and A1 alleles were detected in our sample. All raw data relating to allelic and genotypic frequencies within each of the sample subgroups are displayed in the table. However, to summarize, a highly significant (93.3% versus 76.4%; p = 0.0067; odds ratio [OR] = 4.33; 95% confidence interval [CI], 1.36 to 13.60) overrepresentation of the A0 allele was found in the PSP group compared with control subjects. This was due to increased numbers of homozygous A0 individuals within the PSP group (86.7% compared with 61.1% in control subjects). The difference in genotypic frequency was also significant (p = 0.02; OR = 4.14; 95% CI, 1.19 to 14.48). The A3 allele was found to be completely absent from the PSP group, hence all non-A0 alleles in this group were of the A1 type. No significant differences in the frequency of any allele were observed when the MSA group was compared with control subjects. Allelic distribution within control subjects was in Hardy-Weinberg equilibrium Comparison of genetic data with gender and subject recruitment center showed no evidence of stratification within our sample population.

• Download figure
• Open in new tab
• Download powerpoint

Figure. Subsections of typical electrophoreograms produced by the ABI 377. The four most common genotypes are shown as indicated. bp = base pair.

Discussion. We compared a group of patients with PSP to appropriate control subjects and found a significant change in the allelic distribution of a polymorphic marker within the tau gene. There are a number of possible explanations for this. First, the results may represent a chance finding. This seems unlikely given the level of significance and the fact that this comprises the fourth independent report of this finding.^7,8,15 The second explanation is that it is a spurious result due to stratification effects within our sample population. This possibility was examined, and no gender or intraregional effects were found. Third, the GT repeat sequence itself could exert a direct biological effect involved in the pathogenesis of PSP. Similarly, although this cannot be excluded completely, the nature of the polymorphism and its position approximately seven kilobases downstream of exon nine and five kilobases upstream of exon 10 makes this somewhat unlikely.⁷ This leaves the final and most likely explanation that the effect is due to linkage disequilibrium between the marker we have investigated and another unknown site that exerts a direct functional effect. Current data are insufficient to give an exact indication of the distance between the marker and such a site, although the size of the tau gene in conjunction with its known role in the pathology of PSP would suggest that a location within the tau gene itself seems most likely.

As we know nothing regarding the position of the putative biologically relevant locus, it would be purely hypothetical to discuss the likelihood of mechanism such as differential phosphorylation or transcription and sensitivity to redox state. However, we can at least speculate on the mode of action and importance of the effect of this locus. First, the fact that the overrepresentation of the A0 allele appears to be due to increased numbers of A0 homozygotes immediately suggests that a recessive mode of action seems most likely, as was concluded by Higgins et al.⁸ However, given that we know nothing about the degree of linkage disequilibrium between the marker and the biologically relevant locus, combined with the fact that the A0 allele is already the most common, and that to date of a total of 92 PSP patients (current study combined with three previous reports^7,8,15) no individuals have been found in whom the A0 allele is absent, the possibility that it acts in a dominant or more likely a complex manner cannot be excluded. Such modes of action could be explained either by reduced or age-related penetrance, or by an environmental trigger.

Regardless of the mode of action, we can generate some estimate of the importance of our findings relative to the general disease population. This is given by the attributable fraction (AF), and is defined as the proportion of disease that could be prevented if the risk factor could be removed. In PSP, the AF of the A0/A0 genotype is 0.65 (calculated using the combined data of our study and those from the study of Conrad et al.⁷).

Our study used the diagnostic criteria for PSP of Golbe et al.¹³ These have recently been largely superseded by those of Litvan et al.,¹⁶ with the main difference being the requirement for prominent instability with falls within the first year. Had this criterion been adopted in our study, some patients would have been excluded. Clearly, although it is not possible to draw conclusions, we believe that the remarkable concordance between our genetic data in a clinically ascertained set and those of Conrad et al.⁷ in a pathologically ascertained set may represent a potential example of how genetic data may begin to refine clinical diagnostic criteria.

It seems highly probable that genetic variation at or near the tau gene locus on chromosome 17q21 plays a major and perhaps predominant role in predisposing individuals to develop PSP.¹⁷ Furthermore, given the high probability that the investigated marker is relatively close to the biologically relevant locus, the identification of this locus and an understanding of its mechanism of action will further our understanding of the pathogenesis of PSP. A sibling case-control study in PSP would be the optimum method to perform a tau haplotyping study as the next step toward identifying the PSP locus.¹⁸ Although there has been a report of an association between the A0 allele and familial PD, this has only been in abstract form and awaits confirmation.¹⁵ The absence of an association of the A0 allele with MSA, AD, and the APDC of Guam suggests that the marked effect found by ourselves, by Conrad et al.,⁷ and by Higgins et al.⁸ may be specific to PSP. Additional work is needed to establish whether the A0 allele is relevant to other neurodegenerative disorders with tau pathology, such as corticobasal degeneration and chromosome 17-linked frontotemporal dementia.¹⁹

Acknowledgment

The authors thank the Diabetes Research Group, Department of Medicine, University of Birmingham, UK, for the use of its ABI Prism 377.

Appendix