Materials and methods

A total of 138 Welsh HD patients who had had their CAG repeat numbers determined before4 were analysed for their APOE genotype, blind to repeat number and AO. For 121 of these patients, the sex of the transmitting parent was also known. Polymerase chain reaction and CfoI digestion were carried out as described by Wenham et al,16 followed by polyacrylamide gel electrophoresis of fragments on 6% non-denaturing gels with ethidium bromide staining.

Statistical analysis was performed using the SAS/STAT^®software package (version 6). Comparison of two Pearson correlation coefficients, r₁ and r₂, from samples of sizes n₁ and n₂, was carried out by means of the z statistic,

z=[tanh^-1(r₁)-tanh^-1(r₂)]/[1/(n₁-3)+1/(n₂ -3)]

   ^½      (1)

which follows a standard Gaussian distribution under the null hypothesis r₁=r₂.17 Logistic survival analysis of AO in HD patients was conducted as described by Krawczak et al. 18 Briefly, the annual hazard rate (that is, the probability of disease onset in any given year) is modelled as a logistic function of time (t), pathological repeat length (l), sex (s; 0=female, 1=male), and carriership of a given APOE genotype (c; 0=non-carrier, 1=carrier) by

h(t,l,s,c)=e^F(t,l,s,c)/[1+e^F(t,l,s,c)]      (2)

where

F(t,l,s,c)=β+b_t·t+b_l·l+b_s·s+b_c·c      (3)

is a linear combination of the parameters under study. The probability of AO=α in a given patient with pathological repeat length l₀, sex s₀, and carrier status c₀then equals

h(α,l₀,s₀,c₀)·Π_i=1...α-1[1-h(i,l₀,s₀,c₀)]      (4)

which allows assessment of the significance of a given parameter for AO by means of a stepwise likelihood ratio test. Twice the log likelihood difference between any two nested models approximately follows a χ² distribution with degrees of freedom equal to the difference in parameter number.

Results

As previously reported,4-6 the pathologically increased CAG repeat number was found to be negatively correlated with AO of HD, irrespective of the sex of the transmitting parent (Pearson correlation coefficients: r=−0.771 maternal transmission, p<0.001; r=−0.751 paternal transmission, p<0.001). The length of the CAG repeat on non-disease chromosomes, however, was of significant influence upon AO only when paternally inherited (r=−0.304, p=0.016). For maternally derived normal CAG repeats, a similar, albeit much less pronounced, association (r=−0.152, NS) failed to attain statistical significance. The sex difference between the two correlation coefficients was not statistically significant (z=0.86, two sided p>0.3). No difference was apparent between the lengths of maternally and paternally inherited normal CAG repeats.

A strong gender specific influence on AO was observed for the ratio of pathological to normal CAG repeat number. Here, the negative correlation with AO was stronger when the normal allele was inherited from the mother (r=−0.483, p<0.001) rather than from the father (r=−0.143, NS), a sex difference in correlation which was found to be statistically significant (z=2.05, two sided p=0.04). When allowing for the gender of the patients themselves, a negative correlation between the length of the normal CAG repeat and AO was noted in males (r=−0.327, p=0.014) but not in females (r=−0.039, NS). The sex difference was, however, only of borderline statistical significance (z=1.68, two sided p=0.09). Again, an inverse correlation was observed for the ratio of CAG repeat numbers (male r=−0.143, NS; female r=−0.488, p<0.001) for which the sex difference was statistically significant (z=2.17, two sided p=0.03).

APOE genotypes and mean AOs, classified by patient gender, are summarised in table 1. The APOE genotype distribution in HD patients was not significantly different from that of a control population from the same geographical area (data not shown). Only three genotypes (ε2ε3, ε3ε3, ε3ε4) were present in sufficiently large numbers to allow statistical analysis (n=133). As can be inferred from table 1, genotypes ε3ε3 and ε3ε4 had no obvious effect on AO of HD. Males with genotype ε2ε3, however, had significantly earlier onset than females of the same genotype (t=2.78, 15 df, 0.005<p<0.01) and than males with other genotypes (t=2.33, 51 df, 0.01<p<0.025). The difference in AO between ε2ε3 females and females of other APOE genotypes was of borderline statistical significance (t=1.52, 75 df, 0.05<p<0.1).

Inspection of table 1 indicates that the average pathological CAG repeat length was slightly higher among ε2ε3 males than among females of the same genotype. In order to test whether this discrepancy fully explains the observed difference in AO, data from the 17 ε2ε3 patients were subjected to logistic survival analysis (table 2). Two models were considered which both allowed for repeat length as a covariate of AO, but which differed with respect to whether the gender of a patient was considered (model I) or not (model II). Male ε2ε3 patients turned out to face an annual hazard rate that was twice as high on a logit scale than that of ε2ε3 females. Computation of twice the log likelihood difference between models I and II showed that this difference was highly significant (χ²=7.314, 1 df, p=0.007). Similarly, the annual hazard rate of male ε2ε3 carriers was 1.3 times higher on the logit scale than that of male HD patients with other APOE genotypes (table 3). This difference was also found to be statistically significant when tested by means of twice the log likelihood difference between the respective models, III and IV (χ²=4.816, 1 df, p=0.028). Logistic survival analysis did not, however, show any significant difference between the annual hazard rates of female ε2ε3 carriers and non-carriers (χ²=0.464, 1 df, p>0.4) which implies that their differences in AO may be explicable in terms of different pathological repeat length.

Discussion

In trinucleotide repeat disorders where the expanded repeat is expressed within a protein, the extent of the pathological expansion shows consistent negative correlation with AO2-6 and, at least for Machado-Joseph disease, the length of the normal allele is also critical for AO.19 The length effect of the normal allele in HD has been assessed by a number of different groups.2-6 An association of the normal repeat with AO has been reported previously in the Welsh population4where an effect was found in maternally inherited HD, that is, when the normal repeat came from the father. The present study reproduces this finding and reports further data highlighting the importance of the normal repeat for AO.

Undoubtedly, the phenomenon of anticipation in HD is related to the progressive expansion of the pathological CAG repeat upon transmission from parent to child, with greater expansions occurring in the paternal germline. In the 37-52 repeat range, Duyaoet al 5 found an additional sex of transmitting parent effect on AO which was independent of repeat length. It was observed that, on average, maternally inherited HD has AO two years later than paternally inherited HD with the same pathological repeat length, but no significant effect of the normal allele was apparent.5 Our finding of a negative correlation between AO and the length of the normal repeat in paternally derived cases, however, appears to imply an additional effect of the normal allele reducing AO.

The ratio of pathological to normal repeat length may be as important for AO as the length of the normal repeat itself; the larger the disparity between the two repeat lengths, the earlier AO in paternally (but not in maternally) inherited HD, and in female (but not in male) patients. The evidence that HD pathology proceeds at the protein level is strong20 21 and inclusions immunoreactive to huntingtin, the product of the HD gene, have been found both in mice transgenic for exon 1 of the HD gene with expanded repeats and in HD brain.22 23 Huntingtin interacts with molecules such as HAP1 and GAPDH in a polyglutamine tract length dependent manner in that longer repeats bind these proteins more strongly.24 25 If the disease phenotype in HD results from a greater binding affinity or higher dissociation constants preventing the release of interacting proteins from any complexes formed with the pathologically expanded polyglutamine tracts, then a longer normal polyglutamine tract (that is, a smaller ratio of normal to expanded CAG repeat) may compete better with the expanded protein, allowing fewer aberrant interactions and thus delaying AO. The observed sex specificity may reflect gender differences in the availability of various proteins for interaction. Alternatively, an imprinted modifier gene in linkage disequilibrium with HD might drive the interaction, thereby rendering the effect dependent upon sex of the transmitting parent.

While no sex difference in AO is observed for patients with APOE genotypes ε3ε3 or ε3ε4, the ε2ε3 genotype is associated with significantly earlier onset of HD in men than in women. Also, ε2ε3 males had significantly earlier onset than other male patients while for female patients this relationship appeared to be reversed. It cannot be excluded that this is because of linkage disequilibrium with a nearby gene, but APOE genotype itself is known to affect pathology in disease and traumatic injury. In AD and in head injury, for example, ApoE is associated with the deposition of β-amyloid and such deposition is more prevalent in carriers of the ε4 allele.26 Furthermore, a sex difference has been shown in prevalence and age of onset for AD: women with the ε3ε4 genotype have AO as early as ε4ε4 homozygotes while in men the ε3ε4 genotype gives the same (late) onset as ε3ε3.27Carriership of ε4 has also been implicated in Pick disease, corticobasal degeneration, and progressive supranuclear palsy15 and motor neurone disease.12 These findings suggest that ApoE is likely to affect neuronal function in the brain over and above the formation of neurofibrillary tangles and neuritic plaques, characteristic of AD. Expression of ApoE rises dramatically in response to brain injury and damage26 and it may be expressed in the damaged areas of the brains of HD patients or be associated with the observed inclusions.22 24Although ApoE genotype is associated with late onset Alzheimer disease, the mechanism for this is unknown and thus it is difficult to assess what the mechanism behind this observed effect in HD might be.