Disease-Associated Short Tandem Repeats Co-localize with Chromatin Domain Boundaries

daSTRs localize with boundaries exhibiting ultra-high CpG island density.

“Importantly, daSTRs localize with boundaries exhibiting ultra-high CpG island density (blue spheres, Figure 4A).”

The small subTAD encompassing the FMR1 gene dissolves in the ~935 FXS patient compared to the healthy sibling.

“Third, the small subTAD encompassing the FMR1 gene dissolves in the ~935 FXS patient compared to the healthy sibling (Figure 5B-C, lower green arrow).”

Severe topological disruption around the FMR1 daSTR was observed in both FXS patient B cell lines compared to the healthy sibling with normal-length repeats.

“We observed severe topological disruption around the FMR1 daSTR in both FXS patient B cell lines compared to the healthy sibling with normal-length repeats (Figure 5B-D).”

8 out of the 23 neurological daSTR loci were detected at TAD boundaries in human fetal cortical plate tissue.

“Consistent with our observations in human ES cells, we found 8 out of the 23 neurological daSTR loci were detected at TAD boundaries in human fetal cortical plate tissue (Figure 3A).”

All four classes of daSTRs showed enrichment at boundaries compared to other normal-length repeats matched by sequence and gene body location.

“Although statistical power was restricted by the small size of the groups, all four classes of daSTRs showed enrichment at boundaries compared to other normal-length repeats matched by sequence and gene body location (Figure S5A).”

Boundaries containing matched normal-length repeat tracts show a striking increase in CpG island density compared to boundaries without repeats (Odds Ratio = 3.92, Fisher's Exact Test, P = 1.69E-110).

“Boundaries containing matched normal-length repeat tracts show a striking increase in CpG island density compared to boundaries without repeats (Odds Ratio = 3.92, Fisher’s Exact Test, P = 1.69E-110, Figure 4A-B).”

The extent of boundary disruption correlates with the degree of FMR1 silencing and local loss of CTCF occupancy at specific sites.

“The extent of boundary disruption correlates with the degree of FMR1 silencing and local loss of CTCF occupancy at specific sites.”

A domain boundary separating FMR1 from the immediate upstream TAD in the healthy sibling was ablated in the ~935 FXS patient.

“First, a domain boundary separating FMR1 from the immediate upstream TAD in the healthy sibling was ablated in the ~935 FXS patient (Figure 5B-C, green bracket).”

The mean percentage of drawn repeats tracts that are boundary-associated increased from 49.5% in matched repeats to 81.1% in daSTRs.

“The mean percentage of drawn repeats tracts that are boundary-associated increased from 49.5% (bootstrapped 95% CI: 30.3% < μpercent_boundary < 68.6%) in matched repeats compared to 81.1% (bootstrapped 95% CI: 66.2% < μpercent_boundary < 96.0%) in daSTRs, respectively (Figure 2C).”

There is no significant difference between matched repeats and daSTRs when the test is performed on CpG island-depleted boundaries (Fisher's Exact Test: Odds Ratio = 0.271 and P = 0.063; Cochran-Mantel-Haenszel Test: Odds Ratio = 0.364).

“Importantly, there is no significant difference between matched repeats and daSTRs when the test is performed on CpG island-depleted boundaries (Fisher's Exact Test: Odds Ratio = 0.271 and P = 0.063; Cochran-Mantel-Haenszel Test: Odds Ratio = 0.364; Figure 4E)”

There is no significant difference between matched repeats and daSTRs at CpG islands not at boundaries (Fisher's Exact Test: Odds Ratio = 0.531 and P = 0.57; Cochran-Mantel-Haenszel Test: Odds Ratio = 0.407).

“or CpG islands not at boundaries (Fisher's Exact Test: Odds Ratio = 0.531 and P = 0.57; Cochran-Mantel-Haenszel Test: Odds Ratio = 0.407; Figure 4F).”

Fragile X Syndrome patients exhibit severe boundary disruption.

“We found severe boundary disruption in patients with Fragile X Syndrome.”

4 of the 5 daSTR loci that did not exhibit adjacency to quantitatively called domain boundaries were still co-localized at visually evident domain borders.

“Notably, 4 of the 5 daSTR loci that did not exhibit adjacency to quantitatively called domain boundaries were still co-localized at visually evident domain borders (Figure S1C-D).”

A randomization test demonstrated that daSTRs are significantly closer to domain boundaries compared to the genome-wide null distribution of matched repeats (empirical P=0.0001).

“Finally, we conducted a randomization test and demonstrated that daSTRs are significantly closer to domain boundaries compared to the genome-wide null distribution of matched repeats (empirical P=0.0001, Figure 2D).”

Nearly all disease-associated STRs are spatially placed at boundaries between chromatin domains.

“Here, we discover that nearly all disease-associated STRs (daSTRs) are located at boundaries demarcating 3D chromatin domains.”

The specific loop connecting the TAD immediately upstream of FMR1 was absent in the ~935 FXS patient.

“Second, the specific loop connecting the TAD immediately upstream of FMR1 was absent in the ~935 FXS patient (Figure 5B-C, upper green arrow).”

daSTRs are significantly closer to domain boundaries compared to matched repeat tracts (0 bp versus 5,203 bp median; P = 0.00074 Mann Whitney U Test; P = 0.0037, Kolmogorov-Smirnov test).

“We found that daSTRs are significantly closer to domain boundaries compared to matched repeat tracts (0 bp versus 5,203 bp median; P = 0.00074 Mann Whitney U Test; P = 0.0037, Kolmogorov-Smirnov test) (Figure 2A).”

daSTRs are significantly more enriched at CpG-rich domain boundaries compared to normal-length matched STRs (Fisher's Exact Test: Odds Ratio = 10.4 and P = 1.4E-8; Cochran-Mantel-Haenszel Test: Odds Ratio = 6.64).

“We found that daSTRs are significantly more enriched at CpG-rich domain boundaries compared to normal-length matched STRs (Fisher's Exact Test: Odds Ratio = 10.4 and P = 1.4E-8; Cochran-Mantel-Haenszel Test: Odds Ratio = 6.64; Figure 4D).”

Two classes of boundaries emerge: boundaries with high CpG island density and boundaries depleted of CpG islands.

“Two classes of boundaries emerge: boundaries with high CpG island density and boundaries depleted of CpG islands (Figure S6C-D).”

The effect was largely unchanged (Odds Ratio = 4.23) when performing the Cochran-Mantel-Haenszel test.

“We confirmed that the effect was largely unchanged (Odds Ratio was =4.23) when performing the Cochran–Mantel–Haenszel test.”

Boundaries with daSTRs exhibit a dramatic increase in CpG island density even over the rigorous null of boundaries containing matched, normal-length repeats (Odds Ratio = 24.2, Fisher's Exact Test, P = 1.19E-8).

“Moreover, boundaries with daSTRs exhibit a dramatic increase in CpG island density even over the rigorous null of boundaries containing matched, normal-length repeats (Odds Ratio = 24.2, Fisher’s Exact Test, P = 1.19E-8, Figure 4C).”

The majority of the neurological daSTRs (21 out of 23) are located at domain boundaries in human cortical tissue.

“These results indicate that the majority of the neurological daSTRs (21 out of 23) are located at domain boundaries in human cortical tissue (Figure 3C).”

Boundaries can be stratified into at least two groups: those with high CpG island density and high occupancy of CTCF, and those depleted of CpG island density.

“Boundaries can be stratified into at least two groups: (1) those with high CpG island density and high occupancy of the architectural protein CTCF and (2) those depleted of CpG island density.”

A large number of daSTRs known to undergo paternal instability co-localized to boundaries in mouse sperm and not in mouse oocytes.

“Notably, a large number of daSTRs known to undergo paternal instability (McMurray, 2010; Pearson et al., 2005) co-localized to boundaries in mouse sperm and not in mouse oocytes (Du et al., 2017), suggesting that daSTRs could be at boundaries when germ line expansion takes place (Figure S3).”

• cites McMurray, 2010 sourceNot checked yet

  Not checked in the free run.

• cites Pearson et al., 2005 sourceNot checked yet

  Not checked in the free run.

• cites Du et al., 2017Not checked yet

  Not checked in the free run.

The majority of daSTRs at the human ES cell TAD boundaries reported in Schmitt et al. were also observed at boundaries invariant across the other four ES cell-derived differentiated cell types.

“We observed that the majority of daSTRs at the human ES cell TAD boundaries reported in Schmitt et al. were also observed at boundaries invariant across the other four ES cell-derived differentiated cell types (Figure S2).”

The mean percentage of repeat tracts located at CpG island-depleted boundaries or CpG islands not at boundaries was similar between matched repeats and daSTRs.

“The mean percentage of repeat tracts located at CpG island-depleted boundaries or CpG islands not at boundaries was similar between matched repeats and daSTRs, respectively (Figure 4H, I).”

All four classes of daSTRs co-localize with CpG island-rich boundaries, including STRs that did not specifically contain CpG dinucleotides in the repeat unit sequence.

“All four classes of daSTRs co-localize with CpG island-rich boundaries, including STRs that did not specifically contain CpG dinucleotides in the repeat unit sequence (Figure S5B).”

An additional 11 out of 27 daSTRs were detected at subTAD boundaries in human ES cells.

“We detected an additional 11 out of 27 daSTRs at subTAD boundaries, including RUNX2 (cleidocranial dysplasia), ZIC2 (holoprosencephaly) and CACNA1A (spinocerebellar ataxia 6) (Figure 1B, Figure S1B).”

daSTRs showed significantly higher enrichment at domain boundaries compared to other normal-length, matched repeat sequences (Odds Ratio = 4.77, Fisher's Exact Test, P = 0.00056).

“Moreover, daSTRs showed significantly higher enrichment at domain boundaries compared to other normal-length, matched repeat sequences (Odds Ratio = 4.77, Fisher’s Exact Test, P = 0.00056, Figure 2B).”

There is enrichment for CTCF at human ES cell boundaries compared to loci internal to domains.

“Consistent with previous reports, we found enrichment for the architectural protein CCCTC-binding factor (CTCF) at human ES cell boundaries compared to loci internal to domains (Figure S6A-B) (Dixon et al., 2012; Nora et al., 2012; Phillips-Cremins et al., 2013).”

• cites Dixon et al., 2012 sourceNot checked yet

  Not checked in the free run.

• cites Nora et al., 2012 sourceNot checked yet

  Not checked in the free run.

• cites Phillips-Cremins et al., 2013 sourceNot checked yet

  Not checked in the free run.

11 out of 27 daSTR loci exhibit striking co-localization with boundaries of TADs in human ES cells.

“Surprisingly, we observed that 11 out of 27 daSTR loci exhibit striking co-localization with boundaries of TADs, including FMR1 (Fragile X syndrome), HTT (Huntington’s disease), DMPK (Myotonic dystrophy 1), FXN (Friedreich’s ataxia), C9orf72 (amyotrophic lateral sclerosis) and ATXN1 (spinocerebellar ataxia 1) (Figure 1A, Figure S1A).”

No difference in boundary structure was observed between FXS patient and control B cells or brain tissue when examining domains distal to FMR1.

“Notably, we did not observe any difference in boundary structure between FXS patient and control B cells or brain tissue when examining domains distal to FMR1 (Figure 5H, Figure S7E-F), suggesting that loss of domain integrity in the FXS patient was specific to the genomic location of the daSTR.”

CTCF occupancy was largely unchanged across the genome in FXS patient and healthy sibling B cell samples.

“CTCF occupancy was largely unchanged across the genome in these samples (Figure 6A).”

Complete boundary ablation was observed in fibroblasts from the same patient who contributed the B cells with ~645 repeats compared to fibroblasts from a race-, gender-, and age-matched individual.

“We also observed complete boundary ablation in fibroblasts obtained from the same patient who contributed the B cells with ~645 repeats (GM04025) compared to fibroblasts from a race-, gender-, and age-matched individual (GM04024) (Figure S7C-D).”

Sanger sequencing confirmed that the loss of occupancy was not due to single nucleotide polymorphisms at the motifs.

“We used Sanger sequencing to confirm that the loss of occupancy was not due to single nucleotide polymorphisms at the motifs (Figure 6F).”

The same boundary disruption and large-scale reorganization of the topological environment surrounding the FMR1 daSTR was observed in human brain tissue from two FXS cases compared to two unaffected males.

“We observed the same boundary disruption and large-scale reorganization of the topological environment surrounding the FMR1 daSTR in human brain tissue, including enrichment of contacts between FMR1 and upstream TAD and depletion of contacts across the downstream TAD (Figure 5E-F, Figure S7A-B).”

There is a strong depletion and enrichment of contacts between the FMR1 gene and downstream and upstream TADs, respectively, in FXS.

“Finally, when plotting the difference between FXS and WT contact maps, it becomes evident that there is a strong depletion and enrichment of contacts between the FMR1 gene and downstream and upstream TADs, respectively (Figures 5B-C, blue and red stripes in fold change maps).”

A doublet of occupied CTCF binding sites approximately 100 kb upstream of the FMR1 daSTR in healthy B-lymphocytes was lost in both FXS patient samples.

“Noteworthy, a doublet of occupied CTCF binding sites approximately 100 kb upstream of the FMR1 daSTR in healthy B-lymphocytes was lost in both FXS patient samples (Figure 6B).”

The TAD boundary around the FMR1 daSTR is disrupted in FXS in four unrelated patients, across two cell types and in brain tissue.

“We show that in four unrelated patients, across two cell types and in brain tissue, that the TAD boundary around the FMR1 daSTR is disrupted in FXS.”

All four topological phenotypes occurred in the second FXS patient with ~645 repeats.

“All four topological phenotypes occurred in the second FXS patient with ~645 repeats (Figure 5D).”

Mb-scale TADs co-localized with daSTRs are constitutive across the somatic cell types analyzed in this study.

“Here we find that Mb-scale TADs co-localized with daSTRs are constitutive across the somatic cell types analyzed in this study.”

Many of the daSTR-associated domain boundaries appear to be constitutive across healthy individuals.

“Indeed, many of the daSTR-associated domain boundaries appear to be constitutive across healthy individuals, thus it remains an open question why only a small proportion of human individuals might undergo daSTR expansion at CpG island-rich boundary locations.”

From CTCF ChIP-seq data, three key peaks were identified where CTCF occupancy was differential around FMR1 between samples from healthy controls and those from patients diagnosed with Fragile X Syndrome.

“From the CTCF ChIP-seq data, we identified three key peaks where CTCF occupancy was differential around FMR1 between samples from healthy controls and those from patients diagnosed with Fragile X Syndrome.”

Boundary disruption has been linked to cancer and diseases of limb development.

“Boundary disruption has recently been linked to cancer and diseases of limb development (Flavahan et al., 2016; Hnisz et al., 2016; Lupianez et al., 2015)”

• cites Flavahan et al., 2016 sourceNot checked yet

  Not checked in the free run.

• cites Hnisz et al., 2016 sourceNot checked yet

  Not checked in the free run.

• cites Lupianez et al., 2015 sourceNot checked yet

  Not checked in the free run.

Multiple diverse classes of daSTRs are significantly enriched at domain boundaries compared to their matched, normal-length repeat tracts genome-wide.

“Together, these results indicate that multiple diverse classes of daSTRs are significantly enriched at domain boundaries compared to their matched, normal-length repeat tracts genome-wide.”

The observation of 22 out of 27 daSTR loci at human ES cell boundaries is a conservative estimate.

“Thus, our observation of 22 out of 27 daSTR loci at human ES cell boundaries is a conservative estimate.”

The strong enrichment at boundaries regardless of tissue origin suggests that daSTR co-localization with boundaries might be robust across cell types relevant for the pathology.

“Thus, although the target cell type(s) for many unstable repeat expansion disorders are unknown, the strong enrichment at boundaries regardless of tissue origin suggests that daSTR co-localization with boundaries might be robust across cell types relevant for the pathology.”

• AssumptionRests on an assumption: Domain boundaries called in human ES cells and fetal cortical plate are relevant to the cell types of pathology for all daSTRs.

The majority of daSTRs currently reported as susceptible to unstable expansion in human disease are spatially co-localized to boundaries of 3D genome domains in human ES cells.

“Together, our analyses indicate that the majority of daSTRs currently reported as susceptible to unstable expansion in human disease are spatially co-localized to boundaries of 3D genome domains in human ES cells (Figure 1C).”

• AssumptionRests on an assumption: Mapping daSTRs to normal-length repeat tracts in the hg19 reference genome is representative of disease-associated STR loci.

The combination of ultra-high CpG island density and TAD boundary localization earmarks genomic locations susceptible to STR instability.

“Our results are consistent with a working model in which the combination of ultra-high CpG island density and TAD boundary localization earmarks genomic locations susceptible to STR instability.”

The unique genetic, epigenetic, and topological state at a critical subset of 3D chromatin domain boundaries might earmark locations in the genome with increased susceptibility to STR instability in human disease.

“Together, these data support a working model in which the unique genetic, epigenetic, and topological state at a critical subset of 3D chromatin domain boundaries might earmark locations in the genome with increased susceptibility to STR instability in human disease.”

A subclass of TAD/subTAD boundaries represent hotspots in the genome of ultra-high density of CTCF occupancy and CpG islands.

“These results demonstrate that a subclass of TAD/subTAD boundaries represent hotspots in the genome of ultra-high density of CTCF occupancy and CpG islands.”

Unique genetic and epigenetic features (high CpG island density, high CTCF occupancy) exist at chromatin boundaries where daSTRs become unstable in repeat expansion disorders.

“These data uncover unique genetic and epigenetic features (high CpG island density, high CTCF occupancy) at chromatin boundaries where daSTRs become unstable in repeat expansion disorders.”

Loci susceptible to pathologic, unstable repeat expansion are significantly closer to domain boundaries than expected genome-wide by matched, normal-length repeat sequences.

“Together, our data indicate that loci susceptible to pathologic, unstable repeat expansion are significantly closer to domain boundaries than expected genome-wide by matched, normal-length repeat sequences.”

The topological reconfiguration of the genome could be related to pathologic silencing of the FMR1 gene.

“suggesting that the topological reconfiguration of the genome could be related to pathologic silencing of the FMR1 gene.”

Additional architectural proteins are affected by repeat expansion.

“suggesting that additional architectural proteins are affected by repeat expansion.”

These observations add a 3D component to the locus-specific evidence that STR instability is linked to alterations of CTCF occupancy.

“Our observations add a 3D component to the locus-specific evidence that STR instability is linked to alterations of CTCF occupancy (Brock et al., 1999; Cleary et al., 2010; Filippova et al., 2001; Libby et al., 2008; Sopher et al., 2011).”

• cites Brock et al., 1999 sourceNot checked yet

  Not checked in the free run.

• cites Cleary et al., 2010 sourceNot checked yet

  Not checked in the free run.

• cites Filippova et al., 2001 sourceNot checked yet

  Not checked in the free run.

Loss of domain integrity in the FXS patient was specific to the genomic location of the daSTR.

“suggesting that loss of domain integrity in the FXS patient was specific to the genomic location of the daSTR.”

These results reveal a new link between mutation-length daSTRs and the reconfiguration of 3D genome folding in repeat expansion disorders.

“Together, these results reveal a new link between mutation-length daSTRs and the reconfiguration of 3D genome folding in repeat expansion disorders.”

Disruption of a CTCF binding site at the Atxn7 locus leads to increased somatic and germline repeat instability in a transgenic mouse model of spinocerebellar ataxia 7.

“Disruption of a CTCF binding site at the Atxn7 locus leads to increased somatic and germline repeat instability in a transgenic mouse model of spinocerebellar ataxia 7 (Libby et al., 2008).”

• cites Libby et al., 2008 sourceNot checked yet

  Not checked in the free run.

Boundary disruption correlates with the pathological transcription silencing of FMR1 in FXS.

“These results suggest that boundary disruption correlates with the pathological transcription silencing of FMR1 in FXS.”

FMR1 TAD boundary disruption is correlated with disruption of CTCF occupancy.

“We additionally show that FMR1 TAD boundary disruption is correlated with disruption of CTCF occupancy.”

Occupancy of a doublet CTCF motif over 100 kb from the FMR1 daSTR is lost in FXS patient samples at the genomic location of boundary disruption.

“These results demonstrate that occupancy of a doublet CTCF motif over 100 kb from the FMR1 daSTR is lost in FXS patient samples at the genomic location of boundary disruption.”

Studies in model organisms show that mutations in genes encoding key machinery involved in DNA replication, repair, and recombination are linked to STR instability.

“Studies in model organisms show that mutations in genes encoding key machinery involved in DNA replication, repair, and recombination are linked to STR instability (Iyer et al., 2015; McMurray, 2010; Mirkin, 2007; Pearson et al., 2005).”

• cites Iyer et al., 2015 sourceNot checked yet

  Not checked in the free run.

• cites McMurray, 2010 sourceNot checked yet

  Not checked in the free run.

• cites Mirkin, 2007 sourceNot checked yet

  Not checked in the free run.

daSTRs localized between constitutive TADs are likely to be placed at boundaries in the specific cell type(s) of origin relevant to each repeat expansion disorder.

“Thus, daSTRs localized between constitutive TADs are likely to be placed at boundaries in the specific cell type(s) of origin relevant to each repeat expansion disorders.”

• AssumptionRests on an assumption: Domain boundaries called in human ES cells and fetal cortical plate are relevant to the cell types of pathology for all daSTRs.

Inherited human disorders driven by unstable repeat expansion might also be linked to patterns of higher-order genome folding.

“Here, we provide evidence suggesting that inherited human disorders driven by unstable repeat expansion might also be linked to patterns of higher-order genome folding.”

Although daSTRs exhibit diverse features, they share a common spatial placement at chromatin domain boundaries with ultra-high density CpG islands.

“We demonstrate that although daSTRs exhibit diverse features, they share a common spatial placement at chromatin domain boundaries with ultra-high density CpG islands.”