Brain Advance Access originally published online on December 21, 2007
Brain 2008 131(2):381-388; doi:10.1093/brain/awm312
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Chromosomal profiles of gene expression in Huntington's disease
1MRC Clinical Sciences Centre, Hammersmith Hospital, London, 2Department of Clinical Neuroscience, Division of Neuroscience & Mental Health, Imperial College London, 3MRC Centre for Neurodegeneration Research, Institute of Psychiatry, London, UK and 4Laboratory of Neuropathology, Department of Pathology, University Hospital of Liège, Belgium
Correspondence to: Dr Federico E. Turkheimer, Department of Clinical Neuroscience, Division of Neuroscience and Mental Health, Cyclotron Building, Room 236, Hammersmith Hospital, DuCane Road, London W12 0NN, UK E-mail: federico.turkheimer{at}imperial.ac.uk
 |
Summary |
|---|
 Top Summary IntroductionMethodsResultsDiscussionReferences |
|---|
Recent studies suggested that Huntington's disease is due to aberrant interactions between mutant huntingtin protein, transcription factors and transcriptional co-activators resulting in widespread transcriptional dysregulation. Mutant huntingtin also interacts with histone acetyltransferases, consequently interfering with the acetylation and deacetylation states of histones. Because histone modifications and chromatin structure coordinate the expression of gene clusters, we have applied a novel mathematical approach, Chromowave, to analyse microarray datasets of brain tissue and whole blood to understand how genomic regions are altered by the effects of mutated huntingtin on chromatin structure. Results show that, in samples of caudate and whole blood from Huntington's disease patients, transcription is indeed deregulated in large genomic regions in coordinated fashion, that transcription in these regions is associated with disease progression and that altered chromosomal clusters in the two tissues are remarkably similar. These findings support the notion of a common genome-wide mechanism of disruption of RNA transcription in the brain and periphery of Huntington's disease patients.
Key Words: Huntington's disease; microarrays; histone deacetylase; chromosomal expression; Chromowave
Abbreviations: HD, Huntington's disease; HDAC, histone deacetylase
Received August 15, 2007. Revised October 21, 2007. Accepted November 30, 2007.
|
Introduction |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionReferences |
|---|
Huntington's disease (HD) is an autosomal-dominant neurodegenerative disorder caused by expansion of the CAG repeat region in exon 1 of the HD gene on 4p16 that encodes for the protein huntingtin. Such expansion results in abnormal polyglutamine repeats at the N-terminus of huntingtin (Walker, 2007
) that cannot be cleaved by caspase 3 (Hermel et al, 2004). The mechanisms leading to huntingtin-mediated cell toxicity are still unresolved although it is clear that the mutated protein affects multiple pathways causing cell death.
Mutated huntingtin has been shown to interfere with the transcriptional machinery of the cell. The N-terminal fragments aggregate and sequester several transcription factors leading to decreased availability for binding to DNA promoter regions, and consequent decrease in transcription (Chen-Plotkin et al, 2006). Recent studies in HD models have also shown that mutated huntingtin interferes with the activity of histone acetyltransferase, suggesting that abnormal activity of this enzyme might be a cause of transcriptional dysregulation in HD (Steffan et al, 2001; Sadri-Vakili and Cha, 2006). This observation is supported by the effect of histone deacetylase (HDAC) inhibitors in a number of HD models (Butler and Bates, 2006).
Eukaryotic DNA is wrapped around histones to form chromatin. Nucleosomes, the basic units of chromatin, are composed of eight histones and are arranged to form a structure that facilitates the packaging of chromatin. The transition between tightly protected chromatin to freely accessible DNA is controlled, in part, by modification of the tails of histone proteins. These tails may be modified by adding acetyl groups, phosphates, methyl groups, adenosine diphosphate molecules or ubiquitin proteins (Kuo and Allis, 1998; Cheung et al, 2000; Kouzarides, 2002; Gill, 2004). Acetylation is an important part of the histone-modification code (Kuo and Allis, 1998) because DNA is released by removing the positive charges of the histones through the acetylation of the lysine residues. This results in the loosening of the tightly packed chromatin with subsequent greater access to the DNA by transcription factors and RNA polymerase. The alleged interference of mHtt with histone acetylation is therefore expected to affect transcription globally (Fig. 1).
|
Since it is thought to affect the acetylation state of histones and, thus, chromatin structure that coordinates the expression of gene clusters, we hypothesized that (i) mutated huntingtin affects the transcription of large chromosomal domains, (ii) that the effect of mutated huntingtin is similar in different body tissues and (iii) that chromosomal expression profiles are associated with the disease state. These hypotheses were tested using the genome-scale mRNA measurements of HD tissue available at the time of the project. These consisted in one large microarray study of whole blood in a series of patients and controls (Borovecki et al, 2005
) and in a second microarray study of post-mortem brain tissue from patients and controls (Hodges et al, 2006). We then processed the datasets with Chromowave, a microarray processing tool that allows the identification, with high specificity, of clusters of adjacent genes with homogeneous changes of expression and maps their position on chromosomes (Turkheimer et al, 2006). |
Methods |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionReferences |
|---|
Microarray data
The microarray dataset that analysed global gene expression in blood samples of HD patients, matched controls and pre-symptomatic subjects was taken from a published study that used Affymetrix and Amersham Biosciences oligonucleotide microarrays (Borovecki et al 2005
) and is available from the GEO database. Unfortunately, because of the unavailability of genome alignment data of the Amersham CodeLink arrays, only the data from the Affymetrix platform could be used (series accession GSE1751
[NCBI GEO]
, platform Affymetrix GeneChip HG-U133A). The set contained n = 14 controls, n = 12 HD samples plus n = 5 samples from subjects with genetically determined disease but no overt clinical symptoms. The second dataset consisted in samples of caudate nucleus (n = 14 controls, n = 15 HD), frontal cortex (n = 12 controls, n = 17 HD) and cerebellum (n = 11 controls, n = 17 HD) of genetically confirmed HD patients and age-and sex-matched normal subjects. The dataset is deposited at the Array Express Repository (Experiment E-AFMX-6, platform Affymetrix GeneChip HG-U133A and HG-U133B).
Chromowave analysis
Chromowave (Turkheimer et al 2006) (written in MATLAB 6.5, the Mathworks Inc., Natick MA, USA) normalized the microarrays to the background by dividing intensities by the median value of those genes presented with positive detection. Expression values were then log2 transformed, mapped to their corresponding chromosomal location using the genome alignment information contained in the manifest of each platform (available at http://www.affymetrix.com/support/technical/manual/taf_manual.affx).
Chromowave then applied the wavelet transform to the spatial distribution of the probes and converted the original expression values to wavelet coefficients that are functions of the expression of adjacent genes. Unlike the Fourier transform, that is the classical operator for stationary (periodic) signals, wavelets are a recently introduced mathematical tool for the treatment of signals with ‘non-periodical behaviour’ (e.g. a hammer blow, a plane flyover noise, etc.). Their use is pervasive in areas such as data encoding, transmission and compression including the analysis of gene sequences and functional genomics data (Lio, 2003).
The wavelet transform is an orthogonal mathematical operator which means that the noise level is identical on the original raw data and at all wavelet transform levels. This is advantageous because a cluster of genes with similar expression is transformed into a wavelet transform coefficient that is greater the larger the number of genes in the cluster. Therefore the net effect of the application of the wavelet transform is that genes with individual expression below the noise level, that would otherwise go undetected, are identified when clustered together because their combined energy condenses into a greater wavelet coefficient that arises over the noise.
Unsupervised extraction of chromosomal profiles
The genome-wide ensemble of wavelet coefficients can be used for traditional ways of statistical analysis. In this instance we were interested in extracting the main pattern of chromosomal variation across each of the two datasets, irrespective of the grouping, and to verify afterwards whether major chromosomal variation was indeed associated with the disease state (e.g. unsupervised analysis).
The choice of unsupervised analysis instead of a group comparison was consequential to the hypotheses to be tested. The first hypothesis stated that mutated huntingtin interference would cause a major disruption of the chromatin regulation of RNA. Unsupervised analysis (factor analysis in the instance considered here) allows the identification and quantification (in terms of percentage of total variability) of the major pattern of variation in a data-set. Factor analysis also allows the extraction of subject loadings that can be used in a regression model to verify the association of the pattern extracted from the data with disease states. Finally, one basic assumption of group analysis is that groups need to be homogeneous, clearly not the case in the instance of the data considered in this work.
For this purpose, Chromowave applied the singular value decomposition to the set of wavelet coefficients to extract the first eigenvector, e.g. the main pattern of variation. This pattern was subsequently filtered using a highly conservative threshold that accounted for statistical noise, the number of wavelets and the probe–probe genomic distance; note that in Chromowave the contribution of the individual probes (wavelet transform coefficients at the first level) is zeroed so that only coherent spatial variation of expression is detected. After filtering, the surviving set of coefficients was then passed into the inverse wavelet transform to generate the genome-wide pattern of variation. The contribution of each array to the pattern was calculated as a single number, the ‘case loading’, where a positive value indicates an increased pattern of expression compared to a negative value indicating that a lower pattern is expressed.
Analysis of individual probes expression
In order to verify some of the findings with Chromowave, we also performed traditional probe-by-probe statistical analysis. The normalized expression values were log2 transformed and then a Student t-test was applied to identify differentially expressed genes between groups. The P-values were then corrected for the number of multiple comparisons using the false discovery rate criterion (Reiner et al, 2003) where this was fixed at 5%.
|
Results |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionReferences |
|---|
Analysis of blood samples
The main pattern of chromosomal variation (40% of total variance) obtained from the dataset of blood samples clearly distinguished controls and HD cases (P < 0.000001, Mann–Whitney test) as well as controls and preclinical HD (P = 0.0003, Mann–Whitney test). The distinction between preclinical and clinically manifested HD was less sharp (P = 0.06, Mann–Whitney test), likely because of the low number of preclinical cases (Fig. 2). Large-scale clusters of co-expressed genes appeared to be up-regulated or down-regulated against controls (Fig. 3). Some localized clusters are of interest. One that is considerably down-regulated is on Chr1.p36-p35 and contains MTHFR, CASP9, DFFB, FRAP1 and SHDB genes that have been linked to HD (Kiechle et al, 2002
; Brune et al, 2004; Ravikumar et al, 2004; Chattopadhyay et al, 2005; Majumder et al, 2006). Among these genes, only MTHFR individual expression was significantly down-regulated and passed false discovery rate multiple correction (mean expression ratio HD versus controls = 0.55, P = 0.0037). The repressed cluster on Chr6p21-23 has been previously implicated in the age at onset of HD (Li et al, 2003) (Fig. 3).
|
|
Analysis of post-mortem tissues
The case loadings of the main chromosomal pattern (51% of total variance) obtained with the post-mortem tissues caudate samples demonstrate a significant difference between Hd cases and controls (P = 0.00014, Mann–Whitney test) (Fig. 4). In addition, a Spearman's test found significant positive correlation between HD pathological grade and mRNA expression profile (P = 0.017) where the severity of the disease was graded according to the Vonsattel classification scale (Vonsattel et al, 1985
). Inspection of Fig. 4, illustrates that the pattern is markedly expressed in cases graded 2 and higher, less so in grades 0 and 1. Chromowave analysis of the arrays obtained from frontal and cerebellar samples did not result in any significant chromosomal profile. In caudate, the pattern also extends along large stretches of chromatin but at a lower intensity than the one observed in blood cells (Fig. 5).
|
|
Comparison between blood and post-mortem samples
Striking similarities between the two patterns, particularly involving the whole of Chr4, Chr5, Chr8, Chr10, Chr12, Chr19 and Chr20 are observed between blood and post-mortem tissues (Fig. 6). For instance, note that both patterns contain localized groups of down-regulated genes in a telomeric region of
8 MB of Chr4p around the HD gene confirming the involvement in the disease of genes on this region other than the HD (Farrer et al, 1993; Li et al 2003; Djousse et al, 2004).
|
Individual probes analysis
In both datasets, a very large number of transcripts displayed significant change and survived the false discovery rate correction. Both original references (Borovecki et al 2005
; Hodges et al 2006) contain exhaustive lists of these individual transcripts. Here we focus on the sub-set of genes that showed the largest up/down-regulation in those chromosomal locations identified by Chromowave and whose deregulation was consistent in both blood and striatum. This set, listed in Table 1, contains the DRD2 receptor gene, the hallmark of movement disorders, and genes vital to calcium homeostasis (NRGN, TRPM2, CACNA1A, CACNG4), G-protein signalling (RGS4, HRAS), exocytosis (EXOC7, VAMP2) and endocytosis (DNM1). Of interest is also the up-regulation of genes that induce apoptosis (FAS), transducers of Interleukin6 (IL6ST), whose levels are increased in the plasma of HD patients (Dalrymple et al, 2007), and the down-regulation of dynamin synaptic proteins (DNM1) previously involved in the disease (DiProspero et al, 2004).
|
|
Discussion |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionReferences |
|---|
The results extracted by Chromowave using two previously published expression microarray datasets supported the view that coordinated repression and expression of large chromosomal regions may explain the pathogenetic mechanism that underlie HD (Borovecki et al 2005
; Hodges et al 2006). The transcriptional repression of these regions is consistent with the effect of mHtt on the acetylation state of histones and the consequent compaction of chromatin regions (Butler and Bates, 2006; Sadri-Vakili and Cha, 2006). Notably, the effect size of the transcriptional changes, particularly in the whole blood samples, was comparable to the mRNA down-regulation previously observed in a set of low grade and anaplastic gliomas (Turkheimer et al 2006) where marked reduction in expression was secondary to loss of genetic material, as proved with FISH analysis but also to epigenetic control. Interestingly, transcriptional changes were quite evident in the blood samples of pre-clinical HD subjects (4/5 were outside the normal range) but this finding was not replicated in the caudate samples of subjects at the low end of the Vonsattel scale. The latter result however may be due to the lower quality of RNA in post-mortem samples.
In the analysed samples, the observed large domains of repressed transcription co-exist with sizeable domains of increased transcription. The increased activity of a large number of transcripts, that has been previously reported (Borovecki et al 2005; Hodges et al 2006), can be also explained with the particular chromosomal model adopted here. For example, in the dataset from blood samples of HD patients the repressed regions Chr1p34, Chr17q21 and ChrXp11.2 contain HDAC genes (HDAC1, HDAC5 and HDAC6, respectively) whose inactivation could result in increased acetylation of the relative histones and local increases in transcription. Also, the increased transcription observed in some chromosomes can be explained by the changes in cell population within the affected tissue (i.e. astrocytosis and microglial activation associated with nerve cell loss). For instance, the chromosomal region chr6p, which shows repressed transcription in the blood cells of HD samples is largely active in the HD striatal samples, likely because of the presence of microglial activity in caudate of HD patients (Sapp et al, 2001) and whose transcriptional signature contains MHC molecules, which are encoded by genes on Chr6p (Horton et al, 2004). Individual probe analysis of the microarray data for the caudate samples confirmed up-regulation of MHC Class I molecules but not of MHC Class II nor tumour necrosis factors, interleukins (IL 1–33) or interferon (type 1–3) (data not shown).
The observation of large congruencies between chromosomal profiles in blood and brain tissue samples validates further the hypothesis that the mechanism of interference of mutated huntingtin with histone acetyltransferases generates a similar pattern of transcriptional repression in tissues that are affected by the disease. Furthermore, the similarities between the expression profiles in blood and tissue samples identify the genes that are involved in the pathogenesis of HD. The most striking similarities are observed in chr4, chr8, chr10, chr12, chr19 and chr20 where changes in expression extend throughout the chromosome. In other chromosomes such as chr5, chr14, chr15, chr16, chr17 and chr22 and chrX the correspondence between blood and tissue is more limited but still extends to or more than 50% of the chromatin.
Finally, the significant association between genome-wide patterns extracted with Chromowave and clinical progression brings quantitative evidence to the suggestion made by Borovecki et al (2005) that analysis of peripheral blood samples can provide a minimally non-invasive and easily accessible approach to monitor disease progression and therapy efficacy in HD patients. In particular, chromosomal patterns of expression from microarray peripheral data are promising tools for the assessment of biological efficacy of histone deacetylase inhibitors (Butler and Bates, 2006).
In conclusion, this is the first time that the putative large-scale effects of mutated huntingtin on global changes in gene expression have been visualized in peripheral blood and brain tissue. The ability of Chromowave of detecting expression changes of gene clusters and map their position on chromosomes allowed a step forward in understanding the widespread effects of mutated huntingtin on the human genome.
|
References |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionReferences |
|---|
Borovecki F, Lovrecic L, Zhou J, et al. Genome-wide expression profiling of human blood reveals biomarkers for Huntington's disease. Proc Natl Acad Sci USA (2005) 102:11023–8.
Brune N, Andrich J, Gencik M, et al. Methyltetrahydrofolate reductase polymorphism influences onset of Huntington's disease. J Neural Transm Suppl (2004) 105–10.
Butler R, Bates GP. Histone deacetylase inhibitors as therapeutics for polyglutamine disorders. Nat Rev Neurosci (2006) 7:784–96.[CrossRef][Web of Science][Medline]
Chattopadhyay B, Baksi K, Mukhopadhyay S, Bhattacharyya NP. Modulation of age at onset of Huntington disease patients by variations in TP53 and human caspase activated DNase (hCAD) genes. Neurosci Lett (2005) 374:81–6.[CrossRef][Web of Science][Medline]
Chen-Plotkin AS, Sadri-Vakili G, Yohrling GJ, et al. Decreased association of the transcription factor Sp1 with genes downregulated in Huntington's disease. Neurobiol Dis (2006) 22:233–41.[CrossRef][Web of Science][Medline]
Cheung P, Allis CD, Sassone-Corsi P. Signaling to chromatin through histone modifications. Cell (2000) 103:263–71.[CrossRef][Web of Science][Medline]
Dalrymple A, Wild EJ, Joubert R, et al. Proteomic profiling of plasma in Huntington's disease reveals neuroinflammatory activation and biomarker candidates. J Proteome Res (2007) 6:2833–40.[CrossRef][Web of Science][Medline]
DiProspero NA, Chen EY, Charles V, Plomann M, Kordower JH, Tagle DA. Early changes in Huntington's disease patient brains involve alterations in cytoskeletal and synaptic elements. J Neurocytol (2004) 33:517–33.[CrossRef][Web of Science][Medline]
Djousse L, Knowlton B, Hayden MR, et al. Evidence for a modifier of onset age in Huntington disease linked to the HD gene in 4p16. Neurogenetics (2004) 5:109–14.[CrossRef][Web of Science][Medline]
Farrer LA, Cupples LA, Wiater P, Conneally PM, Gusella JF, Myers RH. The normal Huntington disease (HD) allele, or a closely linked gene, influences age at onset of HD. Am J Hum Genet (1993) 53:125–30.[Web of Science][Medline]
Gill G. SUMO and ubiquitin in the nucleus: different functions, similar mechanisms? Genes Dev (2004) 18:2046–59.
Hermel E, Gafni J, Propp SS, et al. Specific caspase interactions and amplification are involved in selective neuronal vulnerability in Huntington's disease. Cell Death Differ (2004) 11:424–38.[CrossRef][Web of Science][Medline]
Hodges A, Strand AD, Aragaki AK, et al. Regional and cellular gene expression changes in human Huntington's disease brain. Hum Mol Genet (2006) 15:965–77.
Horton R, Wilming L, Rand V, et al. Gene map of the extended human MHC. Nat Rev Genet (2004) 5:889–99.[CrossRef][Web of Science][Medline]
Kiechle T, Dedeoglu A, Kubilus J, et al. Cytochrome C and caspase-9 expression in Huntington's disease. Neuromolecular Med (2002) 1:183–95.[CrossRef][Web of Science][Medline]
Kouzarides T. Histone methylation in transcriptional control. Curr Opin Genet Dev (2002) 12:198–209.[CrossRef][Web of Science][Medline]
Kuo MH, Allis CD. Roles of histone acetyltransferases and deacetylases in gene regulation. Bioessays (1998) 20:615–26.[CrossRef][Web of Science][Medline]
Li JL, Hayden MR, Almqvist EW, et al. A genome scan for modifiers of age at onset in Huntington disease: the HD MAPS study. Am J Hum Genet (2003) 73:682–7.[CrossRef][Web of Science][Medline]
Lio P. Wavelets in bioinformatics and computational biology: state of art and perspectives. Bioinformatics (2003) 19:2–9.
Majumder P, Chattopadhyay B, Mazumder A, Das P, Bhattacharyya NP. Induction of apoptosis in cells expressing exogenous Hippi, a molecular partner of huntingtin-interacting protein Hip1. Neurobiol Dis (2006) 22:242–56.[CrossRef][Web of Science][Medline]
Ravikumar B, Vacher C, Berger Z, et al. Inhibition of mTOR induces autophagy and reduces toxicity of polyglutamine expansions in fly and mouse models of Huntington disease. Nat Genet (2004) 36:585–95.[CrossRef][Web of Science][Medline]
Reiner A, Yekutieli D, Benjamini Y. Identifying differentially expressed genes using false discovery rate controlling procedures. Bioinformatics (2003) 19:368–75.
Sadri-Vakili G, Cha JH. Mechanisms of disease: Histone modifications in Huntington's disease. Nat Clin Pract Neurol (2006) 2:330–8.[CrossRef][Web of Science][Medline]
Sapp E, Kegel KB, Aronin N, et al. Early and progressive accumulation of reactive microglia in the Huntington disease brain. J Neuropathol Exp Neurol (2001) 60:161–72.[Web of Science][Medline]
Steffan JS, Bodai L, Pallos J, et al. Histone deacetylase inhibitors arrest polyglutamine-dependent neurodegeneration in Drosophila. Nature (2001) 413:739–43.[CrossRef][Medline]
Turkheimer FE, Roncaroli F, Hennuy B, et al. Chromosomal patterns of gene expression from microarray data: methodology, validation and clinical relevance in gliomas. BMC Bioinformatics (2006) 7:526.[CrossRef][Medline]
Vonsattel JP, Myers RH, Stevens TJ, Ferrante RJ, Bird ED, Richardson EP Jr. Neuropathological classification of Huntington's disease. J Neuropathol Exp Neurol (1985) 44:559–77.[Web of Science][Medline]
Walker FO. Huntington's disease. Lancet (2007) 369:218–28.[CrossRef][Web of Science][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  C. L. Benn, T. Sun, G. Sadri-Vakili, K. N. McFarland, D. P. DiRocco, G. J. Yohrling, T. W. Clark, B. Bouzou, and J.-H. J. Cha Huntingtin Modulates Transcription, Occupies Gene Promoters In Vivo, and Binds Directly to DNA in a Polyglutamine-Dependent Manner J. Neurosci., October 15, 2008; 28(42): 10720 - 10733. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||