Transcriptional changes are tightly coupled to chromatin reorganization during cellular aging

by Jana M. Braunger, Louis V. Cammarata, Trinadha Rao Sornapudi, Caroline Uhler, and G.V. Shivashankar

Abstract

Human life expectancy is constantly increasing and aging has become a major risk factor for many diseases, although the underlying gene regulatory mechanisms are still unclear. Using transcriptomic and chromosomal conformation capture (Hi-C) data from human skin fibroblasts from individuals across different age groups, we identified a tight coupling between the changes in co-regulation and co-localization of genes. We obtained transcription factors, cofactors, and chromatin regulators that could drive the cellular aging process by developing a time-course prize-collecting Steiner tree algorithm. In particular, by combining RNA-seq data from different age groups and protein-protein interaction data we determined the key transcription regulators and gene regulatory changes at different life stage transitions. We then mapped these transcription regulators to the 3D reorganization of chromatin in young and old skin fibroblasts. Collectively, we identified key transcription regulators whose target genes are spatially rearranged and correlate with changes in their expression, thereby providing potential targets for reverting cellular aging.

Structure of this repository

This github repository consists of the code we used for the analyses described in our paper:

1. R:
   • RNAseq_processing.Rmd: This file contains the RNAseq data processing, the age group definition, as well as the differential gene expression analysis (Fig. 1).
   • DE_gene_robustness.R: This R script was used for obtaining DE genes of subsets of individuals to identify robust DE genes (Supplementary Fig. 2A).
   • Expression_pattern_DE_genes.Rmd: This markdown contains the heatmaps to show the expression over the five age groups of selected genes (DE genes between two consecutive age groups (Supplementary Fig. 3) and signature DE genes (Fig. 4A1/B1/C1/D1)).
   • TF_expression_trajectories.Rmd: This markdown was used to visualize the expression of selected TFs over the five age groups (Supplementary Fig. 10).
   • TF_clustering.Rmd: This markdown was used to cluster the bridge TFs based on a variety of features (Fig. 5C and Supplementary Fig. 14) and create the boxplots for Fig. 5A and Fig. 5B.
2. Python:
   • Steiner_networks: This folder contains the python script to run the prize-collecting Steiner tree (pcst_utils.py), as well as the jupyter notebook used for the analyses in Fig. 2 (Steiner_networks.ipynb) and Supplementary Fig. 7 (Steiner_networks_robustness.ipynb).
   • Process_HiC_data: Inter- and intrachromosomal Hi-C data was processed separately using process_hic_contacts_intra.py and process_hic_contacts_inter.py to filter out repetive chromosome regions such as the centromeres and log-transform and z-score the data. Additonally, global Hi-C analyses were performed (global_hic_analyses.ipynb).
   • LAS: The LAS algorithm was run for each chromosome pair (LAS_intra.py, LAS_inter.py) and then the python script combine_all_submatrices.py was used to check whether submatrices found in one sample also have a high score in the others and only keeping submatrices whose score was above the threshold in both replicates.
   • Intermingling_difference_maps: The jupyter notebook intermingling_difference_maps.ipynb was used to create the binarized intermingling maps for young and old and create intermingling difference maps for selected genes (Fig. 3 and Fig. 4). The python script create_intermingling_networks.py was used to create interactive networks with dash cytoscape (Fig. 5D, Supplementary Fig. 14 and 15).
3. utils: This folder contains the juicer tools jar file used for
   • the pre command to convert the data from the .pair to the .hic format
   • the dump command for normalization

Dependencies needed to run the code

1. R (version 4.2.1) with the following packages:
   • tidyverse (version 1.3.2)
   • tximeta (version 1.14.1)
   • DESeq2 (version 1.36.0)
   • biomaRt (version 2.52.0)
   • UpSetR (version 1.4.0)
   • pheatmap (version 1.0.12)
2. Python (version 3.7.13) with the following packages:
   • networkx (version 2.4)
   • OmicsIntegrator 2 (version 2.3.10)
   • gseapy (version 0.10.8)
   • upsetplot (version 0.6.1)