Dynamic Interplay between Structural Variations and 3D Genome Organization in Pancreatic Cancer

Abstract Structural variations (SVs) are the greatest source of variations in the genome and can lead to oncogenesis. However, the identification and interpretation of SVs in human cancer remain technologically challenging. Here, long‐read sequencing is first employed to depict the signatures of structural variations in carcinogenesis of human pancreatic ductal epithelium. Then widespread reprogramming of the 3D chromatin architecture is revealed by an in situ Hi‐C technique. Integrative analyses indicate that the distribution pattern of SVs among the 3D genome is highly cell‐type specific and the bulk remodeling effects of SVs in the chromatin organization partly depend on intercellular genomic heterogeneity. Meanwhile, contact domains tend to minimize these disrupting effects of SVs within local adjacent genomic regions to maintain overall stability. Notably, complex genomic rearrangements involving two key driver genes CDKN2A and SMAD4 are identified, and their influence on the expression of oncogenes MIR31HG, MYO5B, etc., are further elucidated from both a linear view and 3D perspective. Overall, this work provides a genome‐wide resource and highlights the impact, complexity, and dynamicity of the interplay between structural variations and high‐order chromatin organization, which expands the current understanding of the pathogenesis of SVs in human cancer.


Hi-C library preparation and sequencing
We used 40 ml 2% formaldehyde solution to crosslink the seed tissue for 15 min at room temperature in vacuum. Then, we added 4.324 ml of 2.5 M Gly to quench the cross-linking reaction. The supernatant and tissues were removed from the precipitate.
We then performed centrifugation at 4000 rpm at 4°C for 20 min. After that, the pellet was resuspended in 1 ml extraction II buffer (10 mM Tris-HCl pH 8, 1% Triton X-100, 0.1 mM PMSF, 0.25 M sucrose, 10 mM MgCl 2 , 5 mM mercaptoethanol, and 13 protease inhibitors). Later, the solution was centrifuged for 10 min at 14000 rpm at 4°C. The pellet was resuspended in 300 μl extraction buffer III (10 mM Tris-HCl pH 8, 1.7 M sucrose, 2 mM MgCl 2 , 1 μL protease inhibitor, 0.1 mM PMSF, 5 mM b-mercaptoethanol, 0.15% Triton X-100). Afterward, another 300 μl of clean 3 / 31 extraction buffer III was loaded and centrifuged at 14000 rpm for 10 min. The supernatant was removed, and the pellet was washed with 500 µl ice-cold 1x CutSmart buffer and centrifuged at 2500 g for 5 min each time. The remaining pellets contained the nuclei. Then, we used restriction enzyme buffer to wash the pellet twice and moved it to a safe-lock tube. The next step was to solubilize the chromatin with dilute SDS and incubate it at 65°C for 10 min. Then, the SDS was quenched by Triton X-100 overnight. Next, the nuclei were digested by 4 cutter restriction enzymes (400 units MboI) at 37°C on a rocking platform. The next step was to mark the DNA ends by biotin-14-dCTP and then perform blunt-end ligation between cross-linked fragments. Thus, the ligation enzyme could ligate the proximal chromatin DNA. We incubated the nuclear complexes with proteinase K at 65°C, and then the nuclear complexes were reverse cross-linked. DNA was then purified by phenol-chloroform extraction. We used T4 DNA polymerase to remove biotin-C from unligated fragments. Next, we used sonication to shear the fragments to 200-600 base pairs and repair the fragments with a mixture of T4 DNA polymerase, Klenow DNA polymerase and T4 polynucleotide kinase. Biotin-labeled DNA fragments were then 4 / 31 enriched by streptavidin C1 magnetic beads. The Hi-C library from the beads was sequenced on the Illumina HiSeq X Ten platform with 150 bp paired-end reads. Raw reads were trimmed to 50 bp and then filtered by fqtools plus (https://github.com/annoroad/fqtools_plus) to discard the reads with adapters (> 5 bp adapter nucleotide) and a high N ratio (>5%) and low-quality reads.

Differential interaction and calculation of chromatin interaction
To analyze the difference between interaction matrices in the BxPC3, PANC1, and HPDE6C7 samples, the different matrices were computed by subtracting the z-score matrices of sample-paired intra-interaction matrices. To measure the distance-dependent decay in the cis-interactions of a genome, we used 1 Mb interaction matrices. The intra-chromosome interaction frequency between each bin with the same distances on the reference genome was calculated as real distance in space (https://github.com/dekkerlab/cworld-dekker). Interaction frequencies were log10 transformed to fit a linear model. The slope of each model was outputted as the corresponding finalized interaction decay exponents (IDE) value (https://github.com/dekkerlab/cworld-dekker).

Cis and trans interactions in PDAC
We first analyzed the genome-wide interaction ratio of each chromosome in the three cell lines and found that most of the interactions were intrachromosomal, whereas only a few interactions were inter-chromosomal (Supplementary Figure 3b and 3c).
Next, we analyzed the correlation between chromosome size, spatial position and inter-chromosomal interactions. We found that small, gene-rich chromosomes (chromosomes 16, 17, 19, 20, 21, 22) preferentially interacted with each other, suggesting that chromosome proximity strongly influenced contact probability

Compartment switching
Genome-wide comparative analysis identified common and specific compartment shifts in each PDAC cell line compared with HPDE6C7. The proportions of stable A and B in the genomes of BxPC3 and PANC1 were 33.03% and 29.39%, respectively, and the common A-to-B and B-to-A ratios were 7.8% and 3.53%, respectively (Supplementary Figure 4a). Interestingly, the specific A-to-B and B-to-A switching in BxPC3 cells accounted for 7.45% and 6% of all genome, respectively.
Correspondingly, specific A-to-B and B-to-A transitions occurred in 8.46% and 4.33% of the genome in PANC1, respectively, suggesting the cell type specificity of A/B switching.