Introduction

Huntington’s disease (HD [MIM 143100]) is an autosomal dominantly transmitted, progressive neurodegenerative disease associated with defined expansions in a stretch of CAG repeats in the 5′ part of the HD gene encoding the protein huntingtin (htt) [1]. Expanded CAG block lengths comprising 36–39 repeats may be associated with the clinical presentation of HD, repeat lengths of ≥40 units invariably give rise to clinical manifestation of HD [2]. The expanded CAG repeat is translated into an elongated poly glutamine (polyQ) tract, which apparently leads to neuronal dysfunction and neurodegeneration. Neuropathological features include dramatic loss of neurons and development of astrogliosis in the striatum [3], leading to choreiform movements potentially accompanied by psychiatric and cognitive disturbances. The length of the polyQ tract is the most critical factor for determining age at onset (AO) [4–6], albeit substantial variability remains after controlling for repeat length, particularly when expanded CAG blocks range in lengths ≤46 units [7, 8]. Individuals with identical repeat lengths may exhibit enormous variation in manifestation, progression and outcome of the disease possibly partly due to modifier genes and environmental factors impacting on disease presentation [8]. Defining these modifying genes may have major implications for understanding the pathogenesis. Yet, genetic factors underlying complex traits like AO are notoriously difficult to be identified because many genetic variations may contribute by having small individual effect and, in addition, because of incomplete penetrance. Hitherto, the association of motor AO has been reported with polymorphisms in nine genes [9]. Among these are variations in the NR2A and NR2B sub-units (encoded by GRIN2A, GRIN2B) of the NMDA-type glutamate receptor [10]. An intronic SNP in GRIN2A and the GRIN2B C2664T exchange confer significant effects on AO especially in female HD patients [11]. In addition to these candidate gene studies, evidence for the linkage of AO with the 6q23–24 region was confirmed in HD families [12, 13].

In this study, we pursued a strategy that involves pooling DNAs of patients from the opposite ends of the trait manifestation for determining the association between candidate markers and motor AO in HD. A host of results inter-link apoptotic pathways with the mutant htt protein, and models relating apoptosis to neuronal death in HD have been put forward to explain certain aspects of the pathogenesis [14]. Up-regulation of pro-apoptotic proteins may result in increased cleavage of htt [14], and subsequently arising smaller htt fragments have been proposed to cause cellular stress [15]. Furthermore, a component of the stress response that has been studied in HD represents the activation of mitogen-activated protein kinase (MAPK) signalling pathways [16], which control physiological processes and may initiate apoptotic programmes. Hence, we screened a series of 304 micro-satellites mostly representing genes involved in these apoptosis-related pathways (see Table S1 of the electronic supplementary material).

Results

We screened 151 unrelated patients with clinical diagnosis of HD in an extended association screen with micro-satellite markers as described in detail before by using five DNA pools representing different AO intervals [17, 18; see also supplementary methods]. A total of 304 markers were investigated representing 304 genes involved in apoptosis (see Table S1 of the electronic supplementary material).

Markers for the c-Jun N-terminal kinase (JNK) pathway related genes MAP3K5 (ASK1), MAP2K6, MAPKAPK5 and JUN, and a marker for the cytokine gene IL6 showed obvious inter-pool differences in allele image profiles (AIPs, see Fig. S1 of the electronic supplementary material). Individual genotyping and adjusting the variability for AO attributable to the CAG block length by linear regression confirmed the significant association of AO with markers for ASK1 and MAP2K6.

MAP3K5 (ASK1): fine-mapping and re-screening

Genotyping additional SNPs spread across approximately 230 kb of the corresponding ASK1 region (SNPs 1, 2, 3, 5, 6 and 7 in Fig. 1a, Fig. S3) in the entire HD patient cohort (n = 250) verified the results of the initial screening and localised the association to a region in extensive LD (Fig. 1a) encompassing the inter-genic region between ASK1 and PEX7 and the 5′ regions of both genes. After screening 48 DNAs, re-sequencing of all deviant SSCP patterns revealed exclusively known SNPs without obvious functional consequences (see Table S2 of the electronic supplementary material). Yet the (CGGCG) In/Del polymorphism is situated in the core promoter of ASK1 (rs5880308, SNP4 in Fig. 1a). Furthermore, rs1474988 and rs5880308 are located in the region of the hypothetical LOC7322745 gene with unknown function. But because the latter resides in the intron and rs1474988 is a synonymous polymorphism of apparently no functional significance, we mainly focused on ASK1 and PEX7.

Fig. 1
figure 1

Schematic representation of the a ASK1–PEX7 and b MAP2K6 gene region and pair-wise normalised linkage disequilibrium (LD) as calculated using the Haploview programme. Vertical and horizontal lines represent exons and introns, respectively. The vertical black arrow indicates the position of the originally investigated micro-satellite marker (M). Haplotype blocks are boxed, comprised SNPs in bold face. Each square represents the magnitude of pair-wise LD. Dark grey shading indicates stronger pair-wise LD; light grey squares represent intermediate and white squares represent weak LD (R 2). Asterisks indicate SNPs showing significant association with AO

Full size image

MAP2K6 fine-mapping

Genotyping seven additional SNPs (Fig. 1b, Fig. S3) in the expanded HD cohort revealed 5′ of the originally associated micro-satellite as a haplotype block comprising SNPs in intron 7 of the MAP2K6 gene, rs2521354 and rs2074028, which are strongly associated with AO in HD patients (Fig. 1b).

AO-modifying effect

The SNPs in extensive LD in the inter-genic region between ASK1 and PEX7 show strong correlation with AO in HD as demonstrated in the box plot of the (CGGCG) In/Del genotypes (rs5880308, Fig. 2). Adding the genotypes to the effect of CAG block lengths resulted in significant increase of the R 2 value (0.577 to 0.587, p = 0.010, Table 1). Stratification of patients according to sex revealed no sex-specific differences (Fig. 3a).

Fig. 2
figure 2

Distributions of the ASK1 (CGGCG) In/Del genotypes (rs5880308) and AO for 250 Huntington’s disease patients

Full size image
Fig. 3
figure 3

Sex-specific distributions of the a ASK1 (CGGCG) In/Del genotypes (rs5880308) and b MAP2K6 rs2521354 genotypes and AO for 250 Huntington’s disease patients. Box-plots of female (light grey) and male (dark grey) AO are presented for each genotype, boxes span the inter-quartile range from first to third quartile with a horizontal line showing the median; the range is also indicated by the horizontal lines

Full size image
Table 1 Linear regression analysis of polymorphisms in candidate genes affecting AO in HD
Full size table

Figure 3b depicts the box plots for the AO in the cohort stratified by the significantly associated MAP2K6 rs2521354 SNP genotypes and sex. A genotype-dependent effect on AO is obvious in males where the AA genotype is associated with later AO. Controlling this effect by linear regression confirmed the correlation using the logarithmically transformed AO as the dependent variable and rs2521354 SNP genotypes as independent variables. Sex-specific addition of genotype variations to the effect of CAG repeat lengths resulted in an increased R 2 value only in males (0.617 to 0.626, p = 0.050). A total of 4.9% additional variance was attributed to the rs2521354 genotype variations in males. Together, ASK1 and MAP2K6 genotype variations explain 6.3% additional variance in AO for HD in males. When including the previously published GRIN2A and GRIN2B genotype variations [11], 5% additional variance in AO is explained for the total cohort, and 6% and 7.7%, for males and females, respectively.

Real-time PCR and reporter gene analyses

In transient transfection assays in the human cell lines with constructs expressing the ASK1-(CGGCG)5 and ASK1-(CGGCG)4 alleles of SNP4 and vectors expressing E2F1, no significant differences were observed in reporter gene activity (Fig. 4). Similarly, no differences were evident in ASK1 or PEX7 gene expression comparing individuals with common haplotypes (data not shown).

Fig. 4
figure 4

Comparison of the responsiveness of the ASK1-(CGGCG)5 and ASK1-(CGGCG)4 promoter constructs to E2F1 in U2OS, HeLa and T98G cell lines. Results are expressed as “x-fold” increase over the pcDNA3-HA control expression (see supplementary methods)

Full size image

Discussion

Our results demonstrate significant evidence of association of variations in a genomic region comprising ASK1 and PEX7 at 6q23.3, a replicated linkage region for AO in HD [12, 13] and for MAP2K6 at 17q24.3, respectively. The region at 6q23.3 comprises a haplotype block of high LD spanning ASK1 (MAP3K5) and PEX7 located 30 kb apart in head to head orientation. Peroxisomal assembly proteins encoded by PEX genes are required for import of matrix proteins into peroxisomes. Refsum disease, a peroxisomal disorder of branched chain lipid metabolism, is caused by PEX7 mutations [19]. MAPKs are involved in a cascade of three sequentially phosphorylated and, hence, activated kinases [20]. Apoptosis signal-regulating kinase 1 (ASK1) is a MAPK kinase kinase (MAP3K) that is activated by various types of stress such as reactive oxygen species (ROS), endoplasmatic reticulum (ER) stress, calcium influx and immunological stress. ASK1 selectively activates two different sub-groups of MAP kinase kinases (MAPKK), SEK1 (MKK4) and MKK3/MAPKK6 (MKK6, encoded by MAP2K6), which in turn activate stress-activated protein kinase (SAPK, also known as JNK; c-Jun amino-terminal kinase) and p38 sub-groups of MAP kinases, respectively [21]. Different MAPKs were shown to play a role in HD: elongated htt activates JNK in HN33 hippocampal neuronal cell lines whilst wild-type htt did not [22]. ER stress-triggered by proteasomal dysfunction activates ASK1 via the formation of an IRE1–TRAF2–ASK1 complex. Furthermore, ASK1 − /− primary neurons are defective in polyQ-, proteasome inhibitor- and ER stress-induced JNK activation and cell death [23]. These findings suggest that ASK1 is a key element in ER stress-induced cell death that plays an important role in the neuropathological alterations in polyQ diseases, which in turn renders it a prime candidate gene in this region. Both ASK1 and MAP2K6 are upstream activators of the p38 signaling pathway. ASK1 directly phosphorylates and activates MAP2K6. We hypothesise that sequence variations of ASK1 and MAP2K6 lead to changes in the levels and/or phosphorylation states of p38 and p38-regulated proteins in ways that contribute to the observed delaying effects in motor AO of HD.

The (CGGCG) In/Del genotype, however, is not predicted to exert potential allelic differences in cis-acting regulatory function in transcription by the Matinspector database (http://www.genomatix.de) [24]. Recent studies show that ASK1 is directly regulated by E2F1 and that the DNA-binding domain and the trans-activation domain of E2F1 are necessary for efficient E2F1 trans-activation of the ASK1 promoter [25]. Yet, quantitative gene expression analysis using reporter gene analysis and real-time PCR do not show any differences in expression.

Although intronic SNP rs2521354 in MAP2K6 may cause altered splicing, we cannot infer from the currently available data that this is a functional locus. Nevertheless, a recent study showed sex-related differences in p38 MAP kinase in mice after trauma and haemorrhage, the effect was dependent on testosterone and estradiol exposure [26]. Furthermore, gender-specific differences in p38 MAP kinase production have been observed in rodent models of myocardial ischaemia and reperfusion [27].

In summary, we report a specific ASK1–PEX7 haplotype block within the 6q23–24 region that may harbour biologically significant variations modifying AO in HD. Uncovering MAP2K6 as a putative modifier gene in HD highlights the potential role of the p38 signalling pathway as a key regulatory component in HD pathogenesis. Although our findings are preliminary and need to be replicated in a larger cohort, the results add to the evidence that, besides the HD mutation itself, alterations of other cell pathways may contribute to the inherited variability of the HD phenotype. Sex should be taken into account when modelling for modifier genes in HD as suggested previously.