A longitudinal study highlights shared aspects of the transcriptomic response to cardiogenic and septic shock

This study was part of the multicenter prospective observational trial ShockOmics (ClinicalTrials.​gov identifier: NCT02141607, EU grant #602706). Patients were recruited from the ICUs of the Hôpitaux Universitaires de Genève, Université de Genève (Geneva, Switzerland), and Hôpital Erasme, Université Libre de Bruxelles (Brussels, Belgium). The clinical protocol was approved by the ethical committees of the two participating institutions, and informed consent was obtained from the patients or their representatives. In the present study, we included consecutive adult (> 18 years old) patients admitted to the ICU for SS or CS with a SOFA score at admission > 6 and arterial lactate > 2 mmol/L, as previously described [15]. Moreover, patients had to have blood samples, for the analysis of gene expression, collected at three time points: T1, within 16 h of ICU admission; T2, 48 h after study enrollment; and T3, on day 7 from ICU admission or before discharge from the ICU. Exclusion criteria were expected death within 24 h of ICU admission, transfusion of ≥ 4 units of packed red blood cells or infusion of ≥ 1 unit of fresh frozen plasma, active hematological malignancy, metastatic cancer, chronic immunosuppression, pre-existing end-stage renal disease requiring renal replacement therapy, recent cardiac surgery, Child-Pugh C cirrhosis, and terminal illness.

Septic patients without shock were included in the study as negative controls. Admission criteria for these patients were a proven or clinically suspected infection, associated with at least one organ dysfunction but the cardiovascular system, as indicated by SOFA score, and lactate levels < 2 mmol/L. In these patients, blood samples were collected at two time points: T1, within 16 h of ICU admission, and T2, 48 h after study enrollment.

Laboratory and clinical variables measured at the available time points were analyzed using a linear mixed model accounting for fixed effects of time, gender, age, and random effects of patients. We identified the differences between CS, SS, and SC with ANOVA. Demographic variables and variables with one measurement at a single time point were compared using the Wilcoxon rank-sum test or Fisher exact test for categorical variables.

Peripheral blood was collected at the time points foreseen by the study design in EDTA tubes with 400 μL of 2× Denaturing solution (Ambion, Austin, TX, USA) and stored at − 20 °C. Total RNA was extracted from 800 μL of blood with the MirVana Paris Kit and treated with Turbo DNA-free Kit (Ambion). RNA concentration was estimated with a Nanoquant Infinite M200 instrument (Tecan, Austria). RNA quality was assessed on an Agilent Bioanalyzer using the RNA 6000 Nano Kit (Agilent, Santa Clara, CA, USA), and samples with RNA integrity number > 7.5 were considered acceptable for processing.

We prepared sequencing libraries with the TruSeq Stranded Total RNA with Ribo-Zero Globin Kit (Illumina, San Diego, CA, USA) using 800 ng of total RNA input. Final libraries were validated with the Agilent DNA1000 kit and sequenced on a HiSeq2500 platform, producing 50 × 2 base paired-end reads.

We aligned high-quality paired-end reads to the human reference genome (GRCh38) using STAR (version 2.5.2b) [16], and we selected only uniquely mapping reads. We assigned sequencing reads to genes with featureCounts (version 1.5.1) [17] using the gencode (version 25) primary assembly gene transfer file (GTF) as a reference annotation file for genomic feature boundaries.

DESeq2 [18] package built-in functions were used for data preprocessing, exploratory data analysis and analysis of differential gene expression. We studied gene expression changes over time in CS and SS patients and SC separately with a paired analysis, comparing T1 to T2 in SS and SC and T1 to T3 in CS and SS. Genes with padj < 0.01—Benjamini-Hochberg multiple test correction (FDR)—were considered differentially expressed (DEGs) and used for downstream analysis.

We first performed gene set enrichment analysis (GSEA) [19] to identify the biological processes enriched in CS, SS, and SC. For this purpose, the lists of genes were ranked for log₂FC (T2 vs T1 for SS and SC, and T3 vs T1 for CS) and used as input for GSEA, together with the gene set database c5.bp.v6.2.symbols.gmt. As the first step, we selected significant GSEA terms (FDR < 0.1) in at least one type of shock and filtered for the number of DEGs ≥ 5 in CS and SS. GOs common to the two types of shock, as well as specific of shock type (CS vs SS), were manually selected from the dataset. Starting from the identified GO terms, we selected a list of genes showing significant modulation over time in CS and SS, but not modulated in SC.

Gene set enrichment analysis of transcription factor targets (TFT) was performed on the lists of CS and SS genes ranked for log₂FC expression between T1 and T3. Enriched gene sets were filtered for FDR < 0.1 and number of DEGs ≥ 10.

Seventy-nine shock patients were recruited between November 2014 and March 2016 in the frame of ShockOmics trial. Twenty-one septic shock and 11 cardiogenic shock patients (Additional file 1) were eligible for the present study, after excluding 16 patients who did not meet the inclusion criteria and 31 patients who did not have blood samples collected at the three time points (Additional file 2). Five septic patients, not developing shock, were recruited as controls (SC) and followed for two time points.

At ICU admission, there were no significant demographic differences between the CS and SS group, including age, BMI, severity of illness (APACHEII), and needs for norepinephrine. In both groups, about 20% of patients died later in the period from the second to the fourth week, after study enrollment. SOFA and lactate levels decreased over the measured 1 week of ICU stay in all patients. We assessed SOFA score in patients classified according to mortality (alive or dead at 28 days), showing that a decreasing trend of SOFA can be appreciated also in patients who died (Additional file 3). Laboratory results from the blood collected at the three time points of interest, however, revealed significant differences between cardiogenic and septic shock patients for C-reactive protein (CRP) level, lymphocyte count, hematocrit, and fibrinogen (p < 0.05) (Table 1). Diastolic blood pressure, mean arterial pressure, and temperature were significantly different as well (Table 1). SC compared to SS patients showed significant lower severity scores (APACHE II and SOFA) at ICU admission (Table 2) and did not develop any circulatory failure later during the course of the ICU stay. Clinical and laboratory variables in SC and SS patients are available in Additional file 7.

Total RNA libraries were sequenced in several batches, producing 28.61 M ± 6.49 M, 31.32 M ± 7.81 M, and 28.47 ± 4.37 raw read pairs on average for CS, SS, and SC groups, respectively. Ribosomal depletion was effective for all samples; the rRNA rate on mapped data was negligible in both groups (0.77 ± 1.25%, 0.80 ± 0.92%, and 1.38 ± 1.44% for CS, SS, and SC, respectively). The percentages of reads mapping to exons (85.20 ± 5.66% exonic rate) and DNase efficiency (3.02 ± 1.96% intergenic rate) were satisfactory in all samples. We obtained on average 12.19 ± 2.82, 14.12 ± 4.20, and 13.29 ± 1.62 million of uniquely and unambiguously mapped fragments for the CS, SS, and SC groups, respectively.

To explore whole gene expression modifications in CS, SS, and SC patients across the time points of the study, we first performed a principal component analysis (PCA) separately in each group of patients (Fig. 1a–c). PCA was performed on the 2000 most variable genes across samples. This analysis revealed that SS patients mostly clustered together according to time point on PC1, suggesting that the largest gene expression variation in the dataset was related to the time point (Fig. 1b). Also, in SC patients, we could appreciate a difference between the time points (Fig. 1c). In CS, on the contrary, patients clustered according to T3 only, but not to T1 and T2, suggesting minor gene expression changes between these two time points (Fig. 1a). Following this analysis, we independently identified genes that are differentially expressed (DEGs) in CS, SS, and SC patients over time. We found that SS patients modulated a higher number of genes compared to CS patients both in the comparison T2 vs T1 (CS = 12, SS = 3474) and T3 vs T1 (CS = 1073, SS = 6173). In the SC group, we detected 130 DEGs in the comparison T2 vs T1.

Based on the exploratory analysis, we analyzed with GSEA the genes identified in each group, with the aim to pinpoint overrepresented classes of genes and Gene Ontology (GO) terms that describe the underlying biological processes. We used as input for GSEA the genes of the comparison between T1 and T2, separately in SS and SC. In CS, we analyzed the comparison T1 vs T3, because this is the time frame in which relevant gene expression differences were observed. We first selected GO terms significantly enriched in at least 1 type of shock, and we found a total of 315 downregulated and 78 upregulated biological processes (Additional file 4). GO terms of the inflammatory response and pattern recognition receptors (PRR) were downregulated over time, whereas GO terms related to DNA replication were upregulated in both CS and SS. We selected a list of inflammatory, PRR, and DNA replication genes showing significant modulation over time in CS and SS but not modulated in SC (Fig. 2, Table 3). In detail, we observed a negative expression trend for alarmins (S100A8, S100A9, S100A12), components of the inflammasome (NAIP, NLRC4), genes related to interleukin receptors (IL10RB, IL17RA, IL4R), transcription factors (CEBPB, PPARG, RBPJ, BCL6), Toll-like receptors (TLR1, TLR4, TLR8), and C-type lectin receptors (CLEC5A, CLEC6A). A positive expression trend was observed for genes essential for DNA replication (MCM2, MCM3, MCM5, MCM7). We also estimated the gene expression modifications in SS and CS comparing T1 to T3. In SS patients only, we found increasing expression of genes involved in defense response to bacteria and lymphocyte-mediated immunity, as well as decreasing expression of genes involved in platelet function and carbohydrate catabolic processes (Additional file 5, Additional file 8). Immunoglobulin genes encoding the heavy constant chains (IGHA1, IGHA2, IGHG1, IGHG2, IGHG3, IGHG4, IGHGP, IGHM) and the variable heavy and light chains (IGHV, IGKV, IGLV gene classes) were upregulated over 1 week of observation in both types of shock (Table 4).

We searched for overrepresented transcription factor targets in the lists of genes ranked according to log₂FC in the CS and SS groups separately. GSEA results are as shown in Additional file 6. Then, we focused on the common enriched transcription factors: we observed that genes with a negative expression trend were enriched in target sequences of CCAAT-enhancer-binding protein beta (CEBPB), whereas positively regulated genes were enriched in target sequences of the members of the E2F family of transcription factors (Table 5, Fig. 3).