Hi-C 2.0: An optimized Hi-C procedure for high-resolution genome-wide mapping of chromosome conformation

Abstract

Chromosome conformation capture-based methods such as Hi-C have become mainstream techniques for the study of the 3D organization of genomes. These methods convert chromatin interactions reflecting topological chromatin structures into digital information (counts of pair-wise interactions). Here, we describe an updated protocol for Hi-C (Hi-C 2.0) that integrates recent improvements into a single protocol for efficient and high-resolution capture of chromatin interactions. This protocol combines chromatin digestion and frequently cutting enzymes to obtain kilobase (kb) resolution. It also includes steps to reduce random ligation and the generation of uninformative molecules, such as unligated ends, to improve the amount of valid intra-chromosomal read pairs. This protocol allows for obtaining information on conformational structures such as compartment and topologically associating domains, as well as high-resolution conformational features such as DNA loops.

Keywords: Chromosome conformation capture; Hi-C; Paired-end sequencing.

Figure 1. Overview of the Hi-C method

(A) Cells fixed with formaldehyde contain protein-mediated DNA-DNA interactions. (B) DNA digestion with DpnII, recognizing GATC, generates a 5′-GATC overhang. (C) Filling in of the 5′overhang with dNTPs and biotin-14-dATP blunts the overhang. Ligation of the blunted ends creates a new restriction site (ClaI), which can be used to assess fill-in efficiency. After ligation, crosslinks are reversed to remove proteins from DNA. (D) Removal of Biotin (green lollipops) from un-ligated ends. DNA is fragmented to 200–300bp DNA fragments to enable paired-end sequencing. Numbers 1–4 indicate the different ligation products observed: 1: valid interaction; 2: partial digest; 3: dangling end; 4: self-circle. Size fractionation results in fragment size reduction, indicated by dotted lines (E) Enrichment of ligation junctions by using the high affinity of streptavidin-coated beads for the incorporated biotin allows for ligation product enrichment prior to adapter ligation. Numbering of fragment types is as in (D).

Figure 2. Quality Control of Hi-C ligation products

(A.1) Quality control of intact genomic DNA after cell lysis and before digestion. (A.2) Hi-C DNA after digestion and ligation (+,+) compared to unligated, digested control (−,+). Size is indicated by the 1Kb Molecular Weight Ladder from NEB (1 and 2). (B) PCR amplification of a specific ligation product to assess ligation efficiency. The PCR product (lane 1), PCR product digested with MboI (lane2), ClaI (lane3), or both ClaI and MobI (lane4). Only properly filled-in ligation products will be digested with ClaI (see cartoon). This allows for a qualitative comparison to MboI digestion, which cuts GATC sites that are present at the ligation junction of both properly filled-in and non-filled-in ligation products. Digestion of the PCR product using ClaI indicates efficient fill-in and the ClaI undigested fraction from the PCR can be used to estimate the fill-in efficiency (red arrow). The molecular weight ladder used is the Low Molecular Weight Ladder from NEB. (C) PCR titration of the final Hi-C library and quantification of the fill-in and ligation efficiency by ClaI digestion. PCR amplification is performed with primers that recognize the PE adaptors that were ligated to the Hi-C library before sequencing. With 6-cycles of PCR amplification enough DNA was produced for sequencing (lane #1). The last lane shows a downward shift of the amplified library after digestion with ClaI, indicative of efficient fill-in.

Figure 3. Possible products generated with Hi-C

(A) Two fragments: A (red) and B (blue), are spatially separated in the linear genome (gray dotted line) or neighboring (red and blue to gray fading). (B) If fragment A and B are in close spatial proximity they can become cross-linked and ligated during the Hi-C procedure (1). Partial digests result from undigested neighboring fragments that were biotinylated (2). Other possible, non-valid products can be derived from non-ligated DNA (dangling-end; 3) or single fragments that have become circularized after ligation (self-circles; 4). The gray arrow indicate the orientation of the paired-end reads in the Hi-C library (C) Dangling ends can be removed from the Hi-C library prior to sequencing, as described in this protocol. Any remaining dangling-ends and self-circles can be filtered out from the sequenced library computationally after mapping and assessing the orientation of the DNA reads. After mapping, valid reads locate to different fragments in the reference genome and are either inward or outward oriented, or directed in the same direction (both pointing left or both pointing right) (1). Unligated partial digestion products cannot be distinguished from valid reads because the two reads will map to two (neighboring) restriction fragments. This category is characterized by an inward read orientation (2). Invalid reads have mapped to the same fragment in the reference genome and can be either inward (dangling ends; 3), outward (self-circles; 4) or same direction (error; 5). Gray arrows indicate the read orientation in the reference genome.

Figure 4. Dangling end removal to increase true valid pair reads

(A) Comparison of frequency of dangling ends, total valid pairs, valid pairs with inward read orientation for short range interactions between nearby fragments (separated by less than 500 bp), and frequency of total inward reads for datasets obtained with in situ Hi-C (SRR1658706, SRR1658593, SRR1658712, SRR1658671, SRR1658648) (Rao et al. [13]) and for datasets obtained with Hi-C 2.0. All datasets were analyzed by 100 bp paired end reads. Datasets from Rao et al. were selected solely based on their read depth that was comparable to datasets obtained with Hi-C 2.0 (100–200 million reads). All data were analyzed through our Hi-C mapping pipeline (available in Github: https://github.com/dekkerlab/cMapping). Hi-C 2.0 and removal of dangling ends results in a more consistent percentage of valid reads. Within the set of valid pair reads, we see a reduction in experiment-to-experiment variation of total amount of inward read pairs. The overrepresentation of inward reads appears due to a large extent to the fact that almost all read pairs between neighboring restriction fragments (separated by less than 500 bp) are inward and this category can be 10–20% of all valid pairs. Hi-C 2.0 reduces the experiment-to-experiment variation of interactions between fragments separated by less than 500 bp. This suggests that at least a subset of interactions between adjacent fragments (interactions separated by less than 500 bp) represent dangling ends. (B) A 10 kb resolution heatmap for chromosome 11 (hg19: 2,540,996–9,878,496 bp) derived from 2 libraries with 300M reads. Libraries were generated with HindII (top triangle) or DpnII (bottom triangle). Color scales are normalized and bins without reads are visualized as gray lines. More unfilled bins (gray) in HindIII are caused by larger fragment sizes (C) An increase in valid pair reads, scored as bins containing at least 1 read (i.e. non-zero), allows for analyses at a higher resolution. The plot compares 2 libraries generated with the same protocol, but with different numbers of valid pair reads (blue: 215 million; red: 140 million). Double arrows indicate that at a higher resolution (smaller bins), adding more valid pair reads (by deeper sequencing) becomes important. These libraries were binned with 20kb as the highest resolution from where dotted lines (red, blue) start extrapolating the data to higher resolutions.

Figure 5. Topological structures obtained at increasing resolution

(A) Heatmaps generated from 100 kb binned Hi-C data for chromosome 14 show the alternating pattern of A and B compartments (yellow/purple) (B) On a sub-chromosomal level, heatmaps at 40 kb resolution show the location of TADs, as indicated by an insulation score on top (gray). (C) Within TADs, DNA loops can form that show up as “dots” of interactions in heatmaps of sufficient resolution (typically 10 Kb bins or less). (D) Interpretation of the topological hierarchy obtained from Hi-C. TADs (gray circles) within the same compartment (A or B) interact more frequently than those located in different compartments. TADs are bordered by insulating proteins (e.g. CTCF, cyan squares). DNA loops form between CTCF sites, enhancers and promoters (red/black circles).