All blood samples were obtained from National BioService, LLC, a research biobank operating in compliance with the requirements of the International Society for Biological and Environmental Repositories. In total, 20 blood samples were used in this study; 10 healthy control (HC) samples (CAG repeats, n<35), and 10 HD samples (CAG repeats, n>39). For the HD samples, 7 were from symptomatic patients (HD-Sym) and 3 were from presymptomatic patients who had a diagnosis of HD but did not yet show any clinical symptoms (HD-Pre). One HD patient was taking tetrabenazine and one patient was taking sertraline. All samples were negative for human immunodeficiency virus, hepatitis B virus, hepatitis C virus and syphilis ( Supplemental Table 1).

We wanted to identify chromosome conformations that differed between healthy controls (low CAG), presymptomatic HD patients (high CAG, no disease manifestation) and symptomatic HD patients (high CAG, disease manifestation). We focused on a ~225 kb region surrounding the HTT locus from (chr4: 3,033,588 to 3,258,170 as annotated in hg38) for our analysis. Using the CAG repeat expansion tract in exon 1 of HTT (chr4: 3,054,162 to 3,095,930) as the anchor point (“Anchor”), we defined five genomic zones surrounding the anchor to look at chromosome conformations that varied between sample groups ( Figure 1 and Supplemental Table 2). These Zones were chosen based on: the presence of potential EpiSwitch anchoring sites, the presence of known disease-related SNPs (HD and other diseases), and the enrichment of known histone modification sites (H3K4me3, H3K36me3, and H3K27ac) in HD as found in the GWASdvV2 database ( http://jjwanglab.org/gwasdb) ( Figure 1).

A search of the NCBI Gene Expression Omnibus database for previously reported HD epigenetic data was performed in February 2018 ³³ . Peak-called ChIP-seq data for H3K4me3 from 12 (6 HD and 6 control samples) post-mortem prefontal cortex brain samples (bed format) was obtained ( GSE68952) ³⁴ . In addition, Bigwig tracks of ChIP-seq data for H3K27ac and H3K36me3 from HD iPSC-derived neural cell lines and control cell lines were also downloaded ( GSE95342) ³⁵ . The data tracks were loaded into the Integrative Genome Viewer (IGV) ³⁶ version 2.4 alongside the EpiSwitch and reference sequence annotations. Both visual and programmatic (BEDtools) comparisons were performed on the HTT locus to identify the five zones of interest.

Oxford BioDynamics proprietary EpiSwitch pattern recognition algorithm was used to identify high probability chromatin folding interactions with one “end” occurring in the anchor zone proximal to the CAG repeats and the other in any of the five zones of interest. A total of 61 interactions matched these criteria, and for practical reasons, 20 interactions were selected to cover interactions between the anchor site and all the zones of interest. Oxford BioDynamics automated primer design application was used to design oligonucleotide pairs that amplified the expected DNA sequence caused by the interaction when subjected to the chromosome conformation capture (3C) assay.

3C and detection by PCR were performed as described previously ^{25, 28, 30, 37} . Chromatin with intact chromosome conformations from 50 µl of blood sample from each patient sample was extracted using the EpiSwitch assay following the manufacturer's instructions (Oxford BioDynamics Plc). Quality control on all samples was done using the detection of a chromatin loop at the MMP1 locus, a historical internal control for 3C analysis ³⁰ . Pooled 3C libraries for each of the sample types were generated to provide a generalized population sample for each of the C sample subgroups. Real-time PCR experiments were performed in accordance with the Minimum Information for Publication of Quantitative Real-Time PCR Experiments (MIQE) guidelines ³⁸ . Real-time PCR was performed with SYBR green with the CFX-96 (Bio-Rad) machine to identify the interactions with differing PCR product detection patterns between the sample types ³⁹ . Oligonucleotides were tested on control templates to confirm that each primer set was working correctly. A full list of the primers and PCR conditions used in this study can be found in Supplementary File 1. The final nested PCR was performed on each sample in triplicates for the follow up data on individual patients with HD. This procedure permitted the detection of limited copy-number templates with higher accuracy. All PCR-amplified products were monitored on the LabChip ^® GX from Perkin Elmer, using the LabChip DNA 1K Version 2 kit (Perkin Elmer) and internal DNA markers were loaded on the DNA chip according to the manufacturer’s protocol using fluorescent dyes. Fluorescence was detected by laser and electropherogram read-outs translated into a simulated band on gel picture using the instrument software. The threshold of detection for the instrument was set by the manufacturer from 30 fluorescence units and above. All raw gel images for the PCR assays done in this study as well as a description of each set of comparisons can be found in Dataset 1 ⁴⁰ .

Data analysis was performed in R (language and environment for statistical computing) version 3.5.1 ( https://www.r-project.org) ⁴¹ . This included stats (version 3.6.0) and dplyr (version 0.7.6) packages for t-tests and R ² analysis & a ggplot2 (version 3.0) ⁴² package for boxplots and regression plots.

HC and HD samples were age (average 36.9 years for HC and 35.3 years for HD) and sex matched (10 male and 10 female), with the majority (70%) of HD cases being symptomatic ( Table 1). All samples were from non-Hispanic or Latino whites. Average CAG repeats lengths were 25.7 for HC and 44.2 for HD ( Table 1, Figure 2). There was no statistical difference in CAG repeat length between HD-Pre and HD-Sym ( Figure 2). The average age at diagnosis for HD samples was 35.3 years and the average disease duration was 3.8 years with 7 out of 10 patients reporting symptoms of irritability, chorea, or both ( Table 1).

Of the 20 interactions that were evaluated ( Supplemental Table 3), we identified nine informative interactions. We identified two constitutive interactions and seven conditional interactions which were present in HD, but not healthy controls. Of the seven conditional interactions, three were present only in HD-Sym, and absent in HD-Pre.

All samples passed internal QC analysis for the MMP1 interaction ( Supplementary Figure 1 and Supplementary Figure 2). Two constitutive (identified in all samples) chromatin loops were identified. Both loops were between the Anchor and Zone 2 with the first loop spanning 28 kb and the second loop spanning 34 kb ( Figure 3).

We identified seven conditional chromosome conformations that could discriminate between the different patient subgroups evaluated in this study. Specifically, we identified two chromosome interactions that were present in HC, but absent in all HD samples ( Figure 4). The first interaction (I1) spanned the anchor and zone 4 and covered 77 kb while the second interaction (I2) spanned the anchor and zone 3 and covered 140 kb. We also identified two chromosome interactions that were present in HC and HD-Pre, but absent in HD-Sym samples. Both interactions (I3 and I4) spanned the anchor and zone 4, and covered 92 kb and 104 kb, respectively ( Figure 5). Last, we identified three chromosome interactions that were present in HD-Sym samples, but absent in HD-Pre and HC samples. The first of these three conditional interactions (I5) spanned the anchor and zone 3, covering 122kb. Notably, this interaction included a SNP ( rs362331) known to be a factor in the predisposition to develop HD. The second and third conditional interactions (I6 and I7) spanned the Anchor and Zone 1 and covered 185 kb and 174 kb, respectively ( Figure 6). Last, we tested the absence or presence of all conditional interactions in individual HD samples. In the HD-Sym samples, we found the presence of at least one of the conditional markers (I5, I6 and I7) in six out of seven samples ( Figure 7 and Supplementary Figure 3– Supplementary Figure 5). The odds ratios of each interaction being associated with symptomatic HD presentation were 30, 9 and 16 for I5, I6 and I7, respectively. A summary of all the interactions that were evaluated in this study are shown in Figure 8 and Supplementary Figure 6.

While it is well-known that individuals with greater than 39 CAG-repeats will get HD, the clinical onset of disease varies widely amongst individual patients and the factors that influence when the disease manifests clinically are less well characterized. Here we used EpiSwitch, an industrial platform for assessing chromatin architecture, to evaluate the epigenomic landscape of the HTT locus in HD patients and healthy, unaffected controls. We identified a set of seven interactions that when taken together as a CCS, could differentiate HD from unaffected controls and more importantly, could differentiate between presymptomatic and symptomatic HD patients. One of these interactions, specific for symptomatic HD, contains a SNP (rs362331) shown to be associated with a predisposing disease haplogroup. When taken together, these results provide an initial indication that a simple, non-invasive blood-based test evaluating a CCS can serve as a surrogate biomarker for assessing disease progression in HD.

While it is known that the poly-Q repeat tract expansion and production of mHTT are the underlying causes of HD, the molecular events leading to the development of clinical symptoms are less well characterized. Several studies have looked at SNPs within the HTT locus as a potential contributor to disease onset. One recent SNP genotyping study of HD patients identified ~41 SNPs heterozygous in at least 30% of the patients, including the rs362331 C/T SNP in exon 50 of the HTT gene ⁴³ . Perhaps more biologically relevant is that when the rs362331 SNP is allele-selectively knocked down using anti-sense oligonucleotides, siRNAs or miRNA, a dramatic reduction in the levels of mHTT protein is achieved both in vitro and in vivo, suggesting that this SNP and its surrounding genomic landscape play an important role in regulating mHTT levels ^{44– 46} . In this study, we observed a chromosome conformation (I5) that was present in HD-Sym patients, was absent in HD-Asy and HCs, and overlapped with the rs362331 SNP. While requiring further study, this observation raises the interesting possibility that the production of neurotoxic mHTT in patients that have increased poly-Q tracts and a genetic predisposition to the early development of HD by the presence of the rs362331 SNP may be regulated at the level of higher-order chromatin structure. Another outstanding question in HD is how the disease is inherited in cases where neither parent has received a diagnosis. The two main prevailing hypotheses posit that 1) the carrier parent could have passed away from another factor before the onset of the disease and 2) “unstable” CAG repeat tracts expand with each generation. A third possibility also exists, in that at mid-range (35-50) repeats, individuals could be carriers without manifestation of the disease, but their progeny might be unable to compensate for the genetic defect through undefined mechanisms and will develop the disease. The HD patients evaluated in this study all had CAG repeats in this mid-range, raising the possibility that potential compensatory mechanisms in disease development may be mediated through differences in genomic architecture.

HD is rare, in that there exists a simple test to definitively diagnose the disease, HTT gene sequencing and measurement of CAG repeat number. For clinical care and clinical trials, there are also several tests to measure disease severity, such as the Unified Huntington’s Disease Rating Scale, the Shoulson–Fahn Scale, and the Mini–Mental State Examination ^{47– 49} . While these assessments measure different elements of an HD patients physical and mental well-being as a surrogate for disease severity, they are all subjective in nature and most are not specific for HD. What is missing are concrete molecular tools to monitor disease progression.

As of the time of this writing, there are 22 therapeutic agents for treating HD in different stages of preclinical and clinical development, half of which are in Phase 2 or Phase 3. Once further validated, the CCS reported here could be used in clinical trials as a surrogate outcome biomarker to assess the therapeutic efficacy of the drug in question. In addition to monitoring a symptomatic patient’s response to a particular therapy in clinical trials, another advantage of the approach described here lies in the information that can be obtained for pre-symptomatic patients. For most patients with HD, the pre-symptomatic period can last decades. Five of the seven (i3–i7) interactions identified here clearly separate presymptomatic HD patients from symptomatic ones, and when further validated could serve as an “early warning” indicator test for the onset of HD symptoms in presymptomatic carriers.

This study gives first evidence of detectable conditional differences in chromatin architecture specific for the manifestation of HD and correlated with known disease haplotypes. The major strength of this study lies in its unique approach, which is based on the latest developments in understanding the regulatory role of genomic architecture. While there have been several historical studies in HD aimed at developing disease progression biomarkers based on clinical, imaging and molecular measures ⁵⁰ , to the best of our knowledge this is the first time that the assessment of higher-order chromatin structures in a clinically accessible biofluid has been applied in HD. With the successful application of EpiSwitch in another neurodegenerative condition, amyotrophic lateral sclerosis, as well as other non-neurological conditions such as melanoma, diffuse large B-cell lymphoma, chronic myelogenous leukemia, breast cancer, and rheumatoid arthritis, the results presented here further validate the use of regulatory conditional CCS as disease-related biomarkers ^{25, 28– 31} . A notable limitation of this study was the small sample size.

While the data presented here offer a novel insight into the clinical progression of HD, this study was intended to be a proof-of-concept and not powered for statistical significance. A follow-up study using a larger patient cohort will be required for validation of these initial results.