Predicting nonsense-mediated mRNA decay from splicing events in sepsis using RNA-sequencing data

Abstract

Alternative splicing (AS) and nonsense-mediated mRNA decay (NMD) are highly conserved cellular mechanisms that modulate gene expression. Here, we introduce the NMD pipeline that computes how splicing events introduce premature termination codons to mRNA transcripts via frameshift, then predicts the rate of premature termination codon-dependent NMD. We use whole-blood, deep RNA-sequencing data from critically ill patients to study gene expression in sepsis. Statistical significance was determined as adjusted P < 0.05 and |log₂ fold change| > 2 for differential gene expression and probability ≥0.9 and |DeltaPsi| > 0.1 for AS. The NMD pipeline was developed based on the AS data from Whippet. We demonstrate that the rate of NMD is higher in the sepsis and deceased groups compared with the control and survived groups, which may signify aberrant splicing because of altered physiology in critical illness. Predominance of non-exon skipping events was associated with disease and mortality states. The NMD pipeline also revealed proteins with potential association with sepsis. Together, these results emphasize the utility of the NMD pipeline in studying AS-NMD along with differential gene expression analysis and uncovering proteins associated with sepsis.

Figure 1.. Differential gene expression (DGE) and alternative splicing (AS) data in control versus sepsis and survived versus deceased groups.

(A) Diagram describing RNA-Seq workflow from ICU patients to their DGE and AS information. Created in https://BioRender.com. (B) Volcano plot showing DGE analysis of significantly up- or down-regulated genes in control versus sepsis based on adjusted P-value and log₂ fold change (log₂FC) (red), adjusted P-value alone (blue), log₂FC alone (gray), and not statistically significant results (black). The x-axis (log₂FC) represents log₂ of sepsis/control. (C) Volcano plot showing differential splicing analysis of significantly more or less frequent splicing events in control versus sepsis based on probability and delta percent spliced in (DeltaPsi) (red), probability alone (blue), DeltaPsi alone (gray), and not statistically significant results (black). The x-axis (DeltaPsi) represents the percentage of splicing in sepsis subtracted by the percentage of splicing in control. (D) Proportion of each subtype out of all splicing events in control versus sepsis groups that are then categorized into “splicing” and “transcription-related” groups. AF refers to alternative first exon, and AL refers to alternative last exon. (E) Frequency of each of the four splicing events (from the “splicing” group in Fig 3D) in percentage in control versus sepsis. (B, F) Volcano plot showing DGE analysis of significantly up- or down-regulated genes in survived versus deceased with same color and statistical depictions as (B). The x-axis (log₂FC) represents log₂ of deceased/survived. (C, G) Volcano plot showing differential splicing analysis of significantly more or less frequent splicing events in survived versus deceased with same color and statistical depiction as (C). The x-axis (DeltaPsi) represents the percentage of splicing in deceased subtracted by the percentage of splicing in survived. (H) Proportion of each subtype out of all splicing events in survived versus deceased groups that are then categorized into “splicing” and “transcription-related” groups. (I) Frequency of each of the four splicing events (from the “splicing” group in Fig 3H) in percentage in survived versus deceased.

Figure S1.. Differential gene expression datapoints for the volcano plot in control versus sepsis (Fig 1B).

Figure S2.. Differential splicing analysis datapoints for the volcano plot in control versus sepsis (Fig 1C).

Figure S3.. Differential gene expression datapoints for the volcano plot in survived versus deceased (Fig 1F).

Figure S4.. Differential splicing analysis datapoints for the volcano plot in survived versus deceased (Fig 1G).

Figure 2.. Alternative splicing (AS) and nonsense-mediated mRNA decay (NMD) data in control versus sepsis (Fig 1B–D and H) and survived versus deceased groups (Fig 1E–G and I).

(A) Diagram describing the development of the NMD pipeline from Whippet AS data to NMD outputs. Created in https://BioRender.com. (B) Bar graph showing the percentage of splicing events predicted to induce NMD in control versus sepsis (left) and the percentage of predicted NMD stratified by splicing subtypes (right). (C) Proportion of splicing events of transcripts predicted to cause NMD per splicing subtype in control versus sepsis. (D) Box plot showing the median number of premature termination codons (PTCs) generated per splicing subtype in control versus sepsis. (E) Bar graph showing the percentage of splicing events predicted to induce NMD in survived versus deceased (left) and the percentage of predicted NMD stratified by splicing subtypes (right). (F) Proportion of splicing events of transcripts predicted to cause NMD per splicing subtype in survived versus deceased. (G) Box plot showing the median number of premature termination codons (PTCs) generated per splicing subtype in survived versus deceased. (H) Sankey diagram showing all the genes with P < 0.01 in GO enrichment analysis and their respective biological processes in control versus sepsis. (I) Sankey diagram showing all the genes with P < 0.01 in GO enrichment analysis and their respective biological processes in survived versus deceased.

Figure 3.. NMD pipeline prediction and proteomics data on plasma grancalcin (GCA) in control versus sepsis and survived versus deceased groups.

(A) Diagram describing GCA as the only gene with significant differential splicing in both control versus sepsis (total 2,656 significant differential splicing events) and survived versus deceased (total 866 significant differential splicing events) and with one of the highest absolute RNA-Seq read counts (ARC). Created in https://BioRender.com. (B) Diagram showing prediction on plasma GCA protein level based on differential gene expression, splicing, and NMD data, along with the details on its significant differential splicing event in control versus sepsis. Created in https://BioRender.com. (C) Density map showing the average number of RNA-Seq reads per 150-bp coordinate range of the GCA genome. (B) Dashed line represents the location of the alternative donor (AD) event (from (B)) and respective number of reads in control versus sepsis at the coordinates of the AD event. (D) Violin plot showing the distribution and median ELISA protein concentrations of GCA in each sample in control versus sepsis. (E) Graph showing the correlation data between ELISA concentrations in ng/ml and RNA-Seq read counts of GCA in control versus sepsis. (F) Diagram showing prediction on plasma GCA protein level based on differential gene expression, splicing, and NMD data, along with the details on its significant differential splicing event in survived versus deceased. Created in https://BioRender.com. (G) Density map showing the average number of RNA-Seq reads per 150-bp coordinate range of the GCA genome. (F) Dashed line represents the location of the exon skipping (ES) event (from (F)) and respective number of reads in survived versus deceased at the coordinates of the ES event. (H) Violin plot showing the distribution and median ELISA protein concentrations of GCA in each sample in survived versus deceased. (I) Graph showing the correlation data between ELISA concentrations in ng/ml and RNA-Seq read counts of GCA in survived versus deceased.

Figure S5.. Violin plot showing the distribution and median ELISA protein concentrations of plasma granulin in each sample in control versus sepsis.

Figure S6.. Graph showing the correlation data between ELISA concentrations in ng/ml and RNA-Seq read counts of plasma granulin in control versus sepsis.