Shared regulation and functional relevance of local gene co-expression revealed by single cell analysis

Abstract

Most human genes are co-expressed with a nearby gene. Previous studies have revealed this local gene co-expression to be widespread across chromosomes and across dozens of tissues. Yet, so far these studies used bulk RNA-seq, averaging gene expression measurements across millions of cells, thus being unclear if this co-expression stems from transcription events in single cells. Here, we leverage single cell datasets in >85 individuals to identify gene co-expression across cells, unbiased by cell-type heterogeneity and benefiting from the co-occurrence of transcription events in single cells. We discover >3800 co-expressed gene pairs in two human cell types, induced pluripotent stem cells (iPSCs) and lymphoblastoid cell lines (LCLs) and (i) compare single cell to bulk RNA-seq in identifying local gene co-expression, (ii) show that many co-expressed genes – but not the majority – are composed of functionally related genes and (iii) using proteomics data, provide evidence that their co-expression is maintained up to the protein level. Finally, using single cell RNA-sequencing (scRNA-seq) and single cell ATAC-sequencing (scATAC-seq) data for the same single cells, we identify gene-enhancer associations and reveal that >95% of co-expressed gene pairs share regulatory elements. These results elucidate the potential reasons for co-expression in single cell gene regulatory networks and warrant a deeper study of shared regulatory elements, in view of explaining disease comorbidity due to affecting several genes. Our in-depth view of local gene co-expression and regulatory element co-activity advances our understanding of the shared regulatory architecture between genes.

Loss of long-range co-expression is a common trait in cancer

“…Local high co-expression is a characteristic that has been reported for normal tissues, both in bulk RNA-Seq (21) and single cell analysis (22), reinforcing the conclusion that gene order in eukaryotes is not random (23, 24). However, this phenomenon has been observed in gene pairs within a local vicinity of less than 1Mb (25, 26).…”

“…Which mechanisms are underlying this behavior in cancer? High local gene co-expression has been previously reported for normal human tissues between genes that are located near each other (within 1 Mb), regardless of strand or transcription orientation (22,23,25,26), showing that gene transcription is influenced by its physical location. Moreover, there is evidence that the transcription process contributes to the formation of gene clusters and influences chromatin organization at a fine scale (61,62).…”

Widespread enhancer co-activity identified by multimodal single cell analysis

“…When taking an enhancer-centric perspective, we find that enhancer pairs are more likely to associate with a higher proportion of genes when more genes are present in their vicinity, with as much as 94.5% genes being significantly associated with an enhancer pair when 10 to 15 genes are present in its vicinity (Figure 2d, Spearman R = 0.3, p-value < 2.2e -16 ). This illustrates the presence of genomic regions with high enhancer and gene activity and the high sharing of enhancers across genes, as previously observed 23 . For instance, two enhancers in chr6 (chr6:26104800-26105400 and chr6:26189200-26191000) display significant associations with 21 out of 22 of their neighbouring genes, many of which are found within co-expression gene clusters encoding for Histone proteins 12,25 (Supplementary Table 1).…”

“…Indeed, enhancer co-activity could occur as a consequence of the chromatin being open, which in itself could occur stochastically or due to regulation of other nearby genes. Given the high levels of co-expression found between nearby genes 12 and the sharing of enhancers between co-expressed genes we previously observed 23 , enhancer co-activity is likely influenced by local gene co-expression. Finally, while gene expression and enhancer activity in the same cell are indicative of their relationship, gene transcription is a highly dynamic process and could be partially decoupled in time with nearby enhancer activation 39 , i.e.…”

RNA degradation heavily impacts mRNA co-expression

“…Early bulk co-expression experiments have shown that promoter sharing can be a major source of gene co-expression (Gu et al, 2011). Yet, more recent single cell studies have shown that shared target-regulator relationships are unlikely to result in co-expression and that most--but definitely not all--shared transcription factors fail to enforce co-expressed behavior among target genes (Ribeiro et al, 2021; Yin et al, 2021). Solving the interplay between regulatory architectures and effective co-expression is key for novel strategies that apply combined approaches to study transcriptional regulation (Jeong et al, 2021).…”