Whole transcriptome analysis to explore the impaired immunological features in critically ill elderly patients with sepsis

The study was approved by the Institutional Review Board of the TCVGH (IRB no. CE20069B), and informed consent was obtained from all participants. Thirty-seven elderly ICU patients were recruited from Taichung Veterans General Hospital (TCVGH) from Dec 2018 to Jan 2022. At the time of enrollment, demographic information was gathered, and baseline blood samples were taken on day-1 and day-8. In order to preserve the transcriptome, two 4 ml aliquots of blood were collected in sterile EDTA vacutainers and immediately transferred to RNAse-free vials containing 10.5 ml RNAlater^® (ThermoFisher, Waltham, MA, USA).

In this study, we extracted RNA with the PAXgene Blood RNA Kit, and the average RNA integrity number (RIN) was 8.31 ± 0.58. The manufacturer's instructions were followed when building the library, and 1,000 ng of fragmented RNA was used in subsequent experiments. 150-bp paired-end reads were used for RNA-Seq on the NovaSeq platform (Illumina, Inc., San Diego, CA, United States), and each sample had at least 50–60 million reads. The RNA-seq dataset was deposited in the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) under accession number GSE216902. The dataset was used for cell proportions and pathway analysis.

The sequential organ failure assessment (SOFA) score is a numerical value of sepsis-related organ failure, i.e., it quantifies the number and severity of failed organs. Each score ranges from 0 to 4, with a higher score indicating worsening organ dysfunction [17]. We used the dynamic change in SOFA scores to define responders (R) and non-responders (NR), and a participant was classified as NR if the decrease in SOFA score from day-1 to day-7 was less than 2 [13, 18, 19].

The sequencing was of high quality, Phred scores of 30 were used, and HISAT2 mapped the sequence reads to the reference genome (GRCh38/hg38) [20]. The R package DEseq2 [21] was used to identify the differentially expressed genes (DEGs), and the read counts were calculated using featureCounts [22]. The average mapping rate and read counts were 87.6 ± 4.8% and 65.5 ± 17.6 million reads, respectively. The gprofiler [23] R package was utilized to carry out functional enrichment and pathway analysis on the basis of the Genomes (KEGG, https://​www.​genome.​jp/​kegg/​) and Gene Ontology (GO, http://​geneontology.​org/​) databases. A corrected p value of less than 0.05 was deemed to be significantly enriched. All differentially expressed genes were functionally annotated using Gene Set Enrichment Analysis (GSEA) [24], and a Cytoscape 3.9.0 [25] visualized enrichment map was created. We visualized GSEA results by employing an enrichment map, which organizes gene sets into a similarity network. The node, link, and node color represent the gene-set, the overlap of member genes, and the enrichment score [24], respectively.

RNA-Seq data were used in MiXCR v3.0.13 [15, 16] to quantify the clonotypes of sepsis patients based upon a previously reported protocol [14]. VDJTools v 1.2.1 [26] software was utilized to calculate the counts of complementarity-determining region-3 (CDR3) and sample diversity following the acquisition of the quantitated clonotypes.

We used the Kolmogorov–Smirnov test to check the normality and found that the distribution of a number of variables, including ventilator day, ICU length of stay and hospital length of stay, were left-skewed. To present the data in the unified format, we presented categorical variables frequencies (percentages) for categorical variables and continuous variables as median (interquartile range, IQR). Differences between two groups were analyzed using the Mann–Whitney U test for continuous variables and Fisher’s exact test for categorical variables. The level of significance was set at 0.05, and statistical analyses were two-sided. R version 4.1.0 was used to conduct all of the data analyses.

In this study, we enrolled 37 elderly patients with sepsis [median age 83 years (IQR: 79–87 years)] (Table 1). With principal component analysis, we observed that a divergence between day-1 and day-8 was apparently better in the responsive patients (responder group, R) than in the nonresponsive patients (non-responder group, NR), indicating regulated recovery from sepsis in the immunocompetent group (Additional file 1: Fig. S1). Overall, among the 37 patients, there were 23 responsive patients (responder group, R) with median age 82 years (IQR: 79.5–86.5 years) and 14 nonresponsive patients (non-responder group, NR) with median age 84 years (IQR: 79–88 years). Table 1 shows the basic characteristics of the study group. Higher albumin (median 3.3 mg/dl, IQR: 3–3.7 mg/dl) and lower C-reactive protein (median 9.16 mg/dl, IQR: 3.9–17.5 mg/dl) were observed in the treatment than in their counterparts. Compared with the responder group, the non-responder group had significantly higher SOFA scores at day-7, longer ventilator days, longer ICU lengths of stay, and higher hospital mortality.

As illustrated in Fig. 1A, using principal component analysis on the top 500 variable genes from all of the samples, we discovered that elderly sepsis patients in the intensive care unit had significantly different transcriptomes on day-1 and day-8. The differential expression genes (DEGs) were then identified by comparing gene expression profiles between day-8 and day-1 in elderly ICU patients using the criteria of p < 0.01 and log fold change > 0.25 or < − 0.25. Three DEGs (two upregulated and one downregulated genes) were found in the R group, while the NR group only had 11 DEGs (five upregulated and six downregulated genes). We found that in elderly sepsis patients in the ICU, the top upregulated genes were innate immunity- and inflammation-relevant genes, namely, ZDHCC19, ALOX15, FCER1A, HDC, PRSS33, and PCSK9, which bind to low-density lipoprotein receptors (LDLRs), leading to LDLR degradation and increasing LDL cholesterol levels (Fig. 1B).

To demonstrate the alteration of biological pathways in elderly ICU patients between day-8 and day-1, we demonstrated the enriched pathway by utilizing the online DAVID v6.8 server. As shown in Fig. 2, Gene Ontology (GO) bubble plots were constructed for the biological processes of the top 20 DEGs from the dataset. We found that the DEGs from the elderly ICU patients with sepsis were enriched in the regulation of transcription, adaptive immune response, immunoglobulin production, negative regulation of transcription and immune response. DAVID functional GO analysis is presented in Fig. 3. When compared to controls, functional annotations of proteins encoded by genes in DEGs showed significantly (P < 0.05) increased or decreased enrichment. These annotations were categorized according to biological processes, cellular components, and molecular functions. There were significant differences between DEG-encoded proteins involved in the regulation of transcription, the immune response, protein phosphorylation, and Ras protein signal transduction at the level of biological processes (Fig. 3), whereas cellular component analysis revealed significant differences between the T-cell receptor complex, cytosol nucleoplasm, and nucleus (Fig. 3). Significant differences were found in the enrichment of encoded proteins associated with protein binding, meta ion binding, translation factor activity, or transcriptional activator activity according to molecular function analysis (Fig. 3). Then, we used the Cytoscape ClueGO plugin to investigate the functional enrichment of the DEGs in the dataset to further refine the biological process from the analysis of DAVID-GO terms [27]. Additionally, the REACTOME pathway analysis from ClueGO revealed that cytokine signaling in the immune system, adaptive immune system, sensory perception, innate immune system, and immunoregulatory interactions between lymphoid and non-lymphoid cells were significantly enriched in many DEGs.

Considering that the R and NR groups recovered from sepsis through a specific T-cell receptor signaling pathway, utilizing the number of unique CDR3, D50, and the inverse Simpson index, we used MiXCR to determine the diversity of the T-cell receptor (TCR) (Fig. 5A–C). On day-8, the R group's CDR3, D50, and inverse Simpson index were significantly higher than on day-1. Moreover, the D50 index was significantly higher in the R group than in the NR group on day-8 (Fig. 5B). Finally, we assessed the diversity of unique CDR3 using the inverse Simpson index, which reflects abundant clonotypes. We observed a similar pattern, with significantly increased diversity of TCR in the R group, particularly on day-8 (Fig. 5C).