Identification of genetic variants associated with Huntington's disease progression: a genome-wide association study

TL;DR: A novel measure of disease progression and a genome-wide significant signal on chromosome 5 spanning three genes: MSH3, DHFR, and MTRNR2L2 is generated, suggesting this mechanism as an area for future therapeutic investigation.

Abstract: Summary Background Huntington's disease is caused by a CAG repeat expansion in the huntingtin gene, HTT . Age at onset has been used as a quantitative phenotype in genetic analysis looking for Huntington's disease modifiers, but is hard to define and not always available. Therefore, we aimed to generate a novel measure of disease progression and to identify genetic markers associated with this progression measure. Methods We generated a progression score on the basis of principal component analysis of prospectively acquired longitudinal changes in motor, cognitive, and imaging measures in the 218 indivduals in the TRACK-HD cohort of Huntington's disease gene mutation carriers (data collected 2008–11). We generated a parallel progression score using data from 1773 previously genotyped participants from the European Huntington's Disease Network REGISTRY study of Huntington's disease mutation carriers (data collected 2003–13). We did a genome-wide association analyses in terms of progression for 216 TRACK-HD participants and 1773 REGISTRY participants, then a meta-analysis of these results was undertaken. Findings Longitudinal motor, cognitive, and imaging scores were correlated with each other in TRACK-HD participants, justifying use of a single, cross-domain measure of disease progression in both studies. The TRACK-HD and REGISTRY progression measures were correlated with each other (r=0·674), and with age at onset (TRACK-HD, r=0·315; REGISTRY, r=0·234). The meta-analysis of progression in TRACK-HD and REGISTRY gave a genome-wide significant signal (p=1·12 × 10 −10 ) on chromosome 5 spanning three genes: MSH3, DHFR , and MTRNR2L2 . The genes in this locus were associated with progression in TRACK-HD ( MSH3 p=2·94 × 10 −8 DHFR p=8·37 × 10 −7 MTRNR2L2 p=2·15 × 10 −9 ) and to a lesser extent in REGISTRY ( MSH3 p=9·36 × 10 −4 DHFR p=8·45 × 10 −4 MTRNR2L2 p=1·20 × 10 −3 ). The lead single nucleotide polymorphism (SNP) in TRACK-HD (rs557874766) was genome-wide significant in the meta-analysis (p=1·58 × 10 −8 ), and encodes an aminoacid change (Pro67Ala) in MSH3. In TRACK-HD, each copy of the minor allele at this SNP was associated with a 0·4 units per year (95% CI 0·16–0·66) reduction in the rate of change of the Unified Huntington's Disease Rating Scale (UHDRS) Total Motor Score, and a reduction of 0·12 units per year (95% CI 0·06–0·18) in the rate of change of UHDRS Total Functional Capacity score. These associations remained significant after adjusting for age of onset. Interpretation The multidomain progression measure in TRACK-HD was associated with a functional variant that was genome-wide significant in our meta-analysis. The association in only 216 participants implies that the progression measure is a sensitive reflection of disease burden, that the effect size at this locus is large, or both. Knockout of Msh3 reduces somatic expansion in Huntington's disease mouse models, suggesting this mechanism as an area for future therapeutic investigation. Funding The European Commission FP7 NeurOmics project; CHDI Foundation; the Medical Research Council UK; the Brain Research Trust; and the Guarantors of Brain.

Summary

INTRODUCTION

• Huntington’s disease (HD) is a autosomal dominant fatal neurodegenerative condition caused by a CAG repeat expansion in HTT (1).
• The transition from premanifest to manifest HD is gradual (4, 5), making clinical definition challenging, furthermore psychiatric and cognitive changes may not be concurrent with motor onset (6).
• The need for clinical trials close to disease onset has motivated a raft of observational studies (5, 9, 10).
• TRACK-HD represents the most deeply phenotyped cohort of premanifest and symptomatic disease with annual visits involving clinical, cognitive and motor testing alongside detailed brain imaging (5, 6).
• The authors developed a similar measure in subjects from the REGISTRY study to replicate their findings (9).

Study design and participants

• All experiments were performed in accordance with the Declaration of Helsinki and approved by the University College London (UCL)/UCL Hospitals Joint Research Ethics Committee; ethical approval for the REGISTRY analysis is outlined in (8).
• It provides annually collected high quality longitudinal prospective multivariate data over three years (2008-2011) with 243 subjects at baseline (6) .
• Demographic details of these individuals are shown in Supplementary Information.
• REGISTRY(9) was a multisite prospective observational study which collected phenotypic data between 2003 – 2013 on over 13,000 subjects, mostly manifest HD gene carriers.
• Age, CAG repeat length, UHDRS Total Motor Score (TMS) and Total Functional Capacity (TFC); some patients have further assessments such as a cognitive battery (9), also known as The core data include.

Procedures

• For both studies, atypical severity scores were derived with a combination of principal component analysis (PCA) and regression of the predictable effects of the primary gene HTT CAG repeat length.
• Details differed however, due to differences in nature of the two data sets.
• This model regressed the observed values on clinical probability of onset statistic (CPO) derived from CAG repeat length and age, and its interaction with follow-up length.
• Principal Component Analyses (PCA) of the random slopes was then used to study the dimensionality of these age and CAG-length corrected longitudinal changes.
• Further methodological detail, including control for potential demographic confounders, is given in Supplementary Methods and a flow chart is given in Figure 1.

Statistical and genetic analysis

• Data analyses were performed using SAS/STAT 14·0 and 14·1 primarily via the MIXED, FACTOR and GML procedures (11).
• Genotypes for the REGISTRY subjects were obtained from the GeM-HD Consortium (8), where details of their genotyping, quality control, curation and imputation are provided.
• Association analyses were performed with the mixed linear model (MLM) functions included in GCTA v1·26(12).
• Because of the relatively small sample sizes, analyses were restricted to SNPs with minor allele frequency >1%.
• Gene-wide p-values were calculated using MAGMA v1·05, a powerful alternative to SNP-based analyses which aggregates the association signal inside genes while taking linkage disequilibrium (LD) between SNPs into account (15), using a window of 35kb upstream and 10kb downstream of genes (16).

RESULTS

• The authors performed individual PCA of each domain and found that first PC scores were highly correlated between the domains (P < 0·0001 in all cases, Supplementary Information.).
• The first PC of this combined analysis accounted for 23.4% of the joint variance, and was at least moderately correlated (r>0·4) with most of the variables that contributed heavily to each domainspecific first PC (Supplementary Tables 3 and 4).
• Notably, the genic associations at the MSH3 locus in the TRACK-HD sample also remain significant after correcting for AAO (http://hdresearch.ucl.ac.uk/data-resources/), as does the association with rs557874766 (p=6·30x10 -6 ).
• Msh3 is required for both somatic expansion of HTT CAG repeats and for enhancing an early disease phenotype in mouse striatum (32), Msh3 expression level is associated with repeat instability in mouse brain, (whereas DHFR is not) (30) and expansion of CAG and CTG repeats is prevented by msh3Δ in Saccharomyces cerevisiae (33).
• This indicates that either their progression measure developed in TRACK-HD is an excellent reflection of disease pathophysiological progression or that this is a locus with a very large effect size, or, most likely, both.

Author contributions and declarations

• DJHM collected data, undertook analysis, and wrote the first draft of the ms.
• DL undertook the statistical analysis of phenotype, co-wrote the ms.
• LJ helped secure funding, supervised data analyses, co-wrote the ms.
• DL reports grant funding from CHDI via University College London (UCL), and personal fees from Roche Pharmaceutical, Voyager Pharmaceutical, and Teva Pharmaceuticals.
• DJHM, KL, AD, AFP, SM, LJ, RR, PH, and SJT declare no competing interests.

Figure & Table legends

• After establishing that brain imaging, quantitative motor and cognitive variables are correlated and follow a similar trajectory, the authors scored the TRACK-HD subjects using principal component 1 as a Unified progression measure, and used this measure to look for genomewide associations with HD progression.
• Assessing progression in Huntington’s disease (A) Graphical illustration of the trajectory of HD symptoms and signs over time, annotated to show what time period the different measures of onset and progression discussed in this paper cover, also known as Figure 2.
• (B) Manhattan plot of REGISTRY GWA analysis showing suggestive trails on chromosome 15 in the same area as the GeM GWAS significant locus (8), and chromosome 5 in the same area as the TRACK progression GWAS.
• (iii) Repair of the strand break leads to expansion of the CAG repeat.
• The p-values in columns 2 – 4 refer to the association between the pathway indicated and rate of progression described in this paper (TRACK- TRACK-HD study; REGISTRYREGISTRY study; META- meta-analysis).

Full text

1
Identification of genetic variants associated with Huntington’s disease
progression: a genome-wide association study
Davina J Hensman Moss*
1
, MBBS, Antonio F. Pardiñas*
2
, PhD, Prof Douglas Langbehn
3
,
PhD,
Kitty Lo
4
, PhD, Prof Blair R. Leavitt
5
, MD,CM, Prof Raymund Roos
6
, MD, Prof
Alexandra Durr
7
, MD, Prof Simon Mead
8
, PhD, the REGISTRY investigators and the
TRACK-HD investigators, Prof Peter Holmans
2
, PhD, Prof Lesley Jones
§2
, PhD, Prof Sarah J
Tabrizi
§1
, PhD.
* These authors contributed equally to this work
§
These authors contributed equally to this work
1) UCL Huntington’s Disease Centre, UCL Institute of Neurology, Dept. of Neurodegenerative
Disease, London, UK
2) MRC Centre for Neuropsychiatric Genetics and Genomics, Cardiff University, Cardiff, UK
3) University of Iowa Carver College of Medicine, Dept. of Psychiatry and Biostatistics, Iowa,
USA
4) UCL Genetics Institute, Div. of Biosciences, London, UK
5) Centre for Molecular Medicine and Therapeutics, Department of Medical Genetics,
University of British Columbia, Vancouver, British Columbia, Canada
6) Department of Neurology, Leiden University Medical Centre, Leiden, Netherlands
7) ICM and APHP Department of Genetics, Inserm U 1127, CNRS UMR 7225, Sorbonne
Universités, UPMC Univ Paris 06 UMR S 1127, Pitié- Salpêtrière University Hospital, Paris,
France
8) MRC Prion Unit, UCL Institute of Neurology, London, UK
Corresponding authors:
Sarah J Tabrizi at s.tabrizi@ucl.ac.uk
Lesley Jones at JonesL1@cardiff.ac.uk

2
ABSTRACT
Background Huntington’s disease (HD) is a fatal inherited neurodegenerative disease, caused by a
CAG repeat expansion in HTT. Age at onset (AAO) has been used as a quantitative phenotype in
genetic analysis looking for HD modifiers, but is hard to define and not always available. Therefore
here we aimed to generate a novel measure of disease progression, and identify genetic markers
associated with this progression measure.
Methods We generated a progression score based on principal component analysis of prospectively
acquired longitudinal changes in motor, behavioural, cognitive and imaging measures in the
TRACK-HD cohort of HD gene mutation carriers (data collected 2008 – 2011). We generated a
parallel progression score using 1773 previously genotyped subjects from the REGISTRY study of
HD mutation carriers (data collected 2003 – 2013). 216 subjects from TRACK-HD were genotyped.
Association analyses was performed using GCTA, gene-wide analysis using MAGMA and meta-
analysis using METAL.
Findings Longitudinal motor, cognitive and imaging scores were correlated with each other in
TRACK-HD subjects, justifying a single, cross-domain measure as a unified progression measure in
both studies. The TRACK-HD and REGISTRY progression measures were correlated with each
other (r=0·674), and with AAO (r=0·315, r=0.234 respectively). A meta-analysis of progression in
TRACK-HD and REGISTRY gave a genome-wide significant signal (p=1.12x10
-10
) on chromosome
5 spanning 3 genes, MSH3, DHFR and MTRNR2L2. The lead SNP in TRACK-HD (rs557874766) is
genome-wide significant in the meta-analysis (p=1.58x10
-8
), and encodes an amino acid change
(Pro67Ala) in MSH3. In TRACK-HD, each copy of the minor allele at this SNP is associated with a
0.4 (95% CI=0.16,0.66) units per year reduction in the rate of change of the Unified Huntington’s
Disease Rating Scale (UHDRS) Total Motor Score, and 0.12 (95% CI=0.06,0.18) units per year in

3
the rate of change of UHDRS Total Functional Capacity. The associations remained significant after
adjusting for AAO.
Interpretation The multi-domain progression measure in TRACK-HD is associated with a
functional variant that is genome-wide significant in a meta-analysis. The strong association in only
216 subjects implies that the progression measure is a sensitive reflection of disease burden, that the
effect size at this locus is large, or both. As knock out of Msh3 reduces somatic expansion in HD
mouse models, this highlights somatic expansion as a potential pathogenic modulator, informing
therapeutic development in this untreatable disease.
Funding sources The European Commission FP7 NeurOmics project; CHDI Foundation; the
Medical Research Council UK, the Brain Research Trust, the Guarantors of Brain.
Research in context
Evidence before this study
Huntington’s disease (HD) is universally caused by a tract of 36 or more CAG in exon 1 of HTT.
Genetic modifiers of age at motor onset have recently been identified in HD that highlight pathways,
which if modulated in people, might delay disease onset. Onset of disease is preceded by a long
prodromal phase accompanied by substantial brain cell death and age at motor onset is difficult to
assess accurately and is not available in disease free at risk subjects. We searched all of PubMed up
to Oct 31st 2016 for articles published in English containing “Huntington* disease” AND “genetic
modifier” AND “onset” which identified 13 studies, then “Huntington* disease” AND “genetic
modifier” AND “progression” which identified one review article. Amongst the 13 studies of
genetic modification of HD onset most were small candidate gene studies; these were superseded by
the one large genome wide genetic modifiers of HD study which identified three genome-wide
significant loci, and implicated DNA handling in HD disease modification

4
Added value of this study
We examined the prospective data from TRACK-HD and developed a measure of disease
progression that reflected correlated progression in the brain imaging, motor and cognitive symptom
domains: there is substantial correlation among these variables. We used the disease progression
measure as a quantitative variable in a genome-wide association study and in only 216 people from
TRACK-HD detected a locus on chromosome 5 containing three significant genes, MTRNR2L2,
MSH3 and DHFR. The index variant encodes an amino acid change in MSH3. We replicated this
finding by generating a parallel progression measure in the less intensively phenotyped REGISTRY
study and detected a similar signal on chromosome 5, likely attributable to the same variants. A
meta-analysis of the two studies strengthened the associations. There was some correlation between
the progression measures and AAO of disease but this was not responsible for the association with
disease progression. We also detected a signal on chromosome 15 in the REGISTRY study at the
same locus as that previously associated with AAO.
Implications of all the available evidence
The progression measures used in this study can be generated in asymptomatic and symptomatic
subjects using a subset of the clinically relevant parameters gathered in TRACK-HD. We use these
measures to identify genetic modifiers of disease progression in HD. We saw a signal in only 216
subjects, which replicates in a larger sample, becoming genome-wide significant, thus reducing the
chance of it being a false positive. This argues for the power of better phenotypic measures in
genetic studies and implies that this locus has a large effect size on disease progression. The index
associated genetic variant in TRACK-HD encodes a Pro67Ala change in MSH3, which implicates
MSH3 as the associated gene on chromosome 5. Notably, altering levels of Msh3 in HD mice
reduces somatic instability and crossing Msh3 null mice with HD mouse models prevents somatic
instability of the HTT CAG repeat and reduces pathological phenotypes. Polymorphism in MSH3 has
been linked to somatic instability in myotonic dystrophy type 1 patients. MSH3 is a non-essential

5
neuronally expressed member of the DNA mismatch repair pathway and these data reinforce its
candidacy as a therapeutic target in HD and potentially in other neurodegenerative expanded repeat
disorders.

Frequently asked questions

Q1. What contributions have the authors mentioned in the paper "Identification of genetic variants associated with huntington’s disease progression: a genome-wide association study" ?

In this paper, the authors used the TRACK-HD data to generate a novel unified disease progression measure for use in a genetic association analysis and developed a similar measure in subjects from the REGISTRY study. 

Q2. What is the effect size of Msh3 on the pathological phenotypes?

altering levels of Msh3 in HD mice reduces somatic instability and crossing Msh3 null mice with HD mouse models prevents somatic instability of the HTT CAG repeat and reduces pathological phenotypes. 

Q3. What is the significant SNP in the meta-analysis?

The most significant SNP in the meta-analysis is rs1232027, which is genome-wide significant (p=1.12x10 -10 ), with the p-value of rs557874766 being 1.58x10 -8 . 

Q4. What is the significance of the progression measure in TRACK-HD?

The progression measures used in this study can be generated in asymptomatic and symptomatic subjects using a subset of the clinically relevant parameters gathered in TRACK-HD. 

Q5. What is the significance of the GWAS SNPs in the TRACK-HD gene?

Despite the lack of co-localisation between the TRACK GWAS and MSH3 expression signal, several of the most significant GWAS SNPs were associated with decreased MSH3 expression and slower progression (Supplementary Information). 

Q6. What was the primary component used in the TRACK-HD cohort?

The cross-domain first principal component was used as a unified Huntington’s disease progression measure in the TRACK-HD cohort (Figure 1 and 2B). 

Q7. What are the genetic modifiers of age at motor onset in Huntington’s disease?

Genetic modifiers of age at motor onset have recently been identified in HD that highlight pathways, which if modulated in people, might delay disease onset. 

Q8. What is the evidence for MSH3 in Huntington’s disease?

A complete list of genes in the Pearl et al. (20) pathways is given in http://hdresearch.ucl.ac.uk/data-resources/.DISCUSSIONThe evidence from their study suggests that MSH3 is likely to be a modifier of disease progression in Huntington’s disease. 

Q9. What could be the reason for the signal on chromosome 5?

the signal on chromosome 5 could be due to the coding change in MSH3, or to expression changes in MSH3, DHFR or both, and both effects may operate in disease. 

Q10. What is the likely association between MSH3 and HD?

MSH3 is a neuronally expressed member of a family of DNA mismatch repair proteins (29); it forms a heteromeric complex with MSH2 to form MutSβ, which recognises insertion-deletion loops of up to 13 nucleotides (30) (Figure 4D). 

Q11. What is the likely cause of the changes in CAG repeat size in HD?

Changes in CAG repeat size occur in terminally differentiated neurons in several HD mouse models and in human patient striatum, the brain area most affected in HD, and notably, somatic expansion of the CAG repeat in HD patient brain predicts onset (31).