Skip to main content
Download PDF

Blood from septic patients with necrotising soft tissue infection treated with hyperbaric oxygen reveal different gene expression patterns compared to standard treatment

BMC Medical Genomics volume 18, Article number: 12 (2025) Cite this article

Abstract

Background

Sepsis and shock are common complications of necrotising soft tissue infections (NSTI). Sepsis encompasses different endotypes that are associated with specific immune responses. Hyperbaric oxygen (HBO2) treatment activates the cells oxygen sensing mechanisms that are interlinked with inflammatory pathways. We aimed to identify gene expression patterns associated with effects of HBO2 treatment in patients with sepsis caused by NSTI, and to explore sepsis-NSTI profiles that are more receptive to HBO2 treatment.

Methods

An observational cohort study examining 83 NSTI patients treated with HBO2 in the acute phase of NSTI, fourteen of whom had received two sessions of HBO2 (HBOx2 group), and another ten patients (non-HBO group) who had not been exposed to HBO2. Whole blood RNA sequencing and clinical data were collected at baseline and after the intervention, and at equivalent time points in the non-HBO group. Gene expression profiles were analysed using machine learning techniques to identify sepsis endotypes, treatment response endotypes and clinically relevant transcriptomic signatures of response to treatment.

Results

We identified differences in gene expression profiles at follow-up between HBO2-treated patients and patients not treated with HBO2. Moreover, we identified two patient endotypes before and after treatment that represented an immuno-suppressive and an immune-adaptive endotype respectively, and we characterized the genetic profile of the patients that transition from the immuno-suppressive to the immune-adaptive endotype after treatment. We discovered one gene MTCO2P12 that distinguished individuals who altered their endotype in response to treatment from non-responders.

Conclusion

The global gene expression pattern in blood changed in response to HBO2 treatment in a direction associated with clinical biochemistry improvement, and the study provides potential novel biomarkers and pathways for monitoring HBO2 treatment effects and predicting an HBO2 responsive NSTI-sepsis profile.

Trial registration

Biological material was collected during the INFECT study, registered at ClinicalTrials.gov (NCT01790698) 04/02/2013.

Peer Review reports

Background

Necrotising soft tissue infections (NSTIs) are severe infections of the soft tissue around the bones that are frequently followed by sepsis, and result in septic shock and multi-organ failure in up to half of patients [1, 2]. The accompanying sepsis is a complex pathophysiological disorder caused by an inflammatory response to infection, with a variable and frequently unpredictable course influenced by factors such as age, pre-existing morbidities, and genetic polymorphisms, all of which influence sepsis susceptibility and the severity of the inflammatory response [3, 4]. While the infection is treated and contained via repeated surgical debridement procedures in conjunction with administration of broad-spectrum antibiotics, the treatment efforts against the sepsis syndrome are focused on supporting failing organs until they resume function, making it supportive rather than restorative. Fluid resuscitation is administered to maintain organ perfusion, vasopressor administration to maintain adequate blood pressure, mechanical ventilation supports failing lungs and dialysis offers sustenance to failing kidneys [3]. Hyperbaric oxygen (HBO2) treatment is considered to operate at several levels in patients with NSTI, including increasing oxygen supply to inflamed hypoxic wounds, potentiate antibiotic treatment, improve leucocyte killing efficacy and possess anti-inflammatory properties [5]. The present study focuses on systemic effects by assessing the cellular responses to HBO2 treatment in blood at the gene expression level. Fluctuations in oxygen levels can trigger cascades involved in metabolism and respiration at the cellular level [6]. Reactive oxygen species (ROS) are important by-products of mitochondrial respiration, and HBO2 treatment may influence the level of ROS as well as antioxidant activity via the induction of redox sensitive transcription factors, growth factors, and hormones [7]. When a single or a few sessions of HBO2 treatment is applied to stressed cells, as in inflammatory conditions, studies have suggested that HBO2 treatment promotes a reduction in oxidative stress [8,9,10]. The cells’ oxygen sensing mechanisms are interlinked with inflammatory pathways [11], and HBO2 treatment has also been reported to alleviate disease-induced inflammation by downregulating inflammatory pathways [12,13,14]. HBO2 treatment might hence have cellular restorative effects in necrotising soft tissue infections and other sepsis conditions. Advanced technologies and computational efforts have revealed that the sepsis syndrome encompasses several endotypes with distinct immune responses, and that these are individually linked to disease severity and risk of death [15, 16]. As HBO2 treatment effects are conditioned by the level of cellular stress, differential responses to HBO2 treatment may occur secondary to the individual response to the infection.

HBO2 treatment is time-consuming, and it is only available in a few medical facilities. Patients with sepsis and NSTI are frequently highly unstable, making patient transfers between hospitals and hospital departments risky. Identifying an immunological profile or a set of candidate biomarker genes that could be used to predict the optimal application and/or timing of intervention with HBO2 treatment could prevent unnecessary and untimely transport and stays outside the intensive care unit (ICU).

The aim of the present study was to (1) identify gene expression signatures related to effects of intervention with HBO2 treatment in patients with NSTI, and (2) investigate if there are sepsis-NSTI profiles that are more amenable to HBO2 treatment, by relating gene expression to clinical disease markers.

Methods

Study design and participants

The paper presents the findings of the prospective observational HBOmic study of NSTI patients hospitalized to Copenhagen University Hospital (Rigshospitalet) in Denmark between February 2013 and March 2017. The patients were included in the international multicenter INFECT trial (ClinicalTrials.gov Number: NCT01790698), and the study design and patient cohort have been thoroughly described previously [2, 17]. A pilot study was performed prior to the current study to validate the quality of the biological specimens, the strategy for sample processing and the sample size needed to detect a difference in blood global gene expression before and after intervention with HBO2 treatment. This work was published together with the statistical analysis plan in the HBOmic study protocol, which also describes the methodology applied in detail [18]. Therefore, this paper only provides a brief overview. The current study included 95 patients who were admitted to the intensive care unit with NSTI as the primary diagnosis. Participants were included following the eligibility criteria and the participant timeline outlined in the study protocol [18]. In short, we selected patients who had blood samples withdrawn before and after HBO2 treatment and had been subjected to surgical debridement with sampling of infected tissue before and after HBO2 treatment during the initial acute phase of the infection. A similar time interval for non-HBO2 patients was applied. Eighty-five patients were given HBO2 treatment. Sixty-nine patients had one blood sample taken before and after the first treatment session with HBO2 (HBOx1 group) and 14 patients had one blood sample taken before the first HBO2 session and one blood sample taken after the second HBO2 session (HBOx2 group). The patients in the intervention groups received 1–6 sessions of HBO2 treatment, with a median of 3 sessions (IQR = 3–3 sessions). Two patients had two blood samples at separate time points taken before receiving the treatment. The remaining 10 patients did not receive HBO2 treatment between blood samples (non-HBO group). The blood samples that served as baseline in the present study were collected after a median of 32 h (IQR = 16–66 h) following first admission to any hospital (primary or specialized referral hospital with HBO2 treatment chamber capacity). These blood samples were taken with a median of 7 min (IQR = 1–12 min) before the first HBO2 session in the HBOx1 group and 7 min and 30 s. (IQR = 3 min 10 s.−13 min) in the HBOx2 group. After intervention, the follow-up sample was collected after a median of 5 min (IQR = 1–8 min 45 s) from the first session of HBO2 in the HBOx1 group and 5 min and 30 s. (IQR = 1 min 15 s.−10 min 45 s) after the second treatment session in the HBOx2 group. The time interval between the baseline sample and the follow-up sample were equal in the HBOx2 group and the non-HBO group with less than a day in between, however the time interval between samples was shorter in the HBOx1 group with a median of 2 h (IQR = 1 h 42 min-2 h 6 min) [Table 1].

Table 1 Time intervals between samples and between samples and the intervention
Full size table

Patient management

Patients were either admitted to or referred to a specialty hospital with the diagnosis NSTI, and all patients were treated in line with our standardised multidisciplinary treatment protocol [2]. In summary, the patients were treated with frequent surgical debridement, broad-spectrum antibiotics (meropenem, clindamycin, and ciprofloxacin), supportive critical care in the ICU (Intensive Care Unit), intravenous polyspecific immunoglobulin G was considered for specific patient groups (with septic shock and Group A Streptococcus) and HBO2 treatment was aimed at being delivered as soon as possible after the initial surgical debridement and for a total of three sessions within 72 h; ideally, the two first treatments were administered within 24 h after admission. Each HBO2 treatment session consisted of breathing 100% oxygen at a pressure of 284 kPa (2.8 absolute atmosphere (ATA)) without air-breaks for 90 min, and with compression and decompression phases of 5–15 min.

Data collection and pre-processing

Demographic variables, including time of sample collection and time of HBO2 interventions were obtained from an electronic data base. Additional clinical variables measured before and after the intervention were obtained from electronic patient records, including plasma levels of leucocytes, thrombocytes, creatinine, C-reactive protein (CRP), lactate dehydrogenase and estimated glomerular filtration rate (eGFR). Sepsis was defined according to The Third International Consensus Definition for sepsis and septic shock (Sepsis-3) [19].

Blood samples were collected from an arterial line into ethylenediaminetetraacetic acid (EDTA) tubes and immediately transported on dry ice to a freezer in the same building where they were stored at −80 degree Celsius. Blood samples were thawed for RNA extraction and mRNA purification using rRNA depletion. The RNA sequencing was performed on an Illumina Novaseq6000 platform with a targeted sequencing read depth of 20 million reads. Sequential steps of quality control were performed, including RQN (RNA Quality Number) estimation and library quality control on the 5300 Fragment Analyzer from Agilent Scientific Instruments, RNA quantification, and computation of the TIN (transcript integrity number) using the RSeQC tool tin.py v2.6.4. Fastqc, fastp and multiQC for quality of raw and trimmed reads was performed during transcriptome assembly. The first 14 bases and reads shorter than 35 nt were removed with “Trimmomatic” v0.39 [20]. Reads were mapped against the human genome (hg38) with default setting using “STAR” v2.7.3a [21], with the exception of allowance of up to two mismatches during the mapping, and a minimum of five overlap bases to trigger mates merging and realignment.

Bioinformatics and statistical analysis

Analyses of the quantified RNA sequencing data was performed in the statistical software R, v 4.3.0, using the packages; “edgeR” v3.42.4, “FactoMineR” v2.8, “pheatmap” v1.0.12, “org.Hs.eg.db” v3.17.0, “mixOmics v6.24.0 and “WGCNA” v1.71.1 [22,23,24,25,26,27,28], see supplementary methods section for a detailed description (Additional file 1). Data reduction methods were used to visualize data in a lower dimension, to investigate for confounders (including years of sample storage, RQN, TIN, batch of sample shipment, library preparation batch, sequencing batch, age, biological sex, immunoglobulin G treatment -yes/no, and HBOmic Study treatment group), and to visualize differences in gene expression between groups, and for variable selection. We performed differential expression analysis between the intervention groups using a paired design correcting for the time of sample collection (baseline or follow-up). The obtained set of differentially expressed genes were used for hierarchical clustering of the baseline samples and follow-up samples separately, and cluster overlaps was performed to investigate for movement between clusters. Subsequently, differential expression analysis was performed at baseline between the individuals who later moved and the individuals who did not move between clusters in response to HBO2 treatment in a non-paired design. In this model the treatment group (HBOx1 group or HBOx2 group) was included as a covariate. To predict the biological function of genes in each obtained cluster, protein–protein interaction (PPI) networks were constructed, and functional gene annotation performed using the String database (v 12.0) [29]. Categorical data were presented as absolute numbers with percentage (%). Boxplots and scatterplots were produced in R base, v 4.3.0, the scatterplots using the “scatter.smooth” function and simple linear regression. Continuous data were presented as medians with IQR in parentheses. The significance level of associations with clinical parameters between groups were tested with the non-parametric Kruskal–Wallis test by ranks, as data were not normally distributed. For differential expression analysis we used false discovery values below 0.05 as statistically significant.

Results

Patient characteristics

Patient characteristics such as primary site of infection, comorbidity, and biochemistry in the three treatment groups are demonstrated in Table 2. The baseline characteristics were unbalanced with non-HBO participants presenting with more comorbidities and biochemical signs of more severe disease.

Table 2 Patient characteristics at admission
Full size table

Quality control and investigation for confounders

In the sample quality control six samples were excluded due to insufficient amount of total RNA or an electropherogram showing high levels of degradation. We obtained a median of 2493 ng (IQR = 1302–3929 ng) total RNA, a median RQN of 9.3 (IQR = 8.5–9.7) and a median TIN of 70.3 (IQR = 67.4–72.2) per sample. One patient was removed from the study because this individual was an outlier in both the principal component analysis and the hierarchical clustering in both the samples collected before and after the intervention (data not shown), (Fig. 1). According to the patient record, this individual was an Asian national who was visiting as a tourist. After filtering lowly expressed counts 22,461 genes were included in the analysis, of which 6182 could not be annotated. The common and tagwise dispersions and the log2-cpm normalized counts are shown in supplementary results (see Additional file 2, Figure A1 and A2).

Fig. 1
figure 1

Study flow chart. *Two patients had three samples collected: one sample at baseline (baseline non-HBO group), one sample later during the same day (baseline HBOx1 group) and one sample after one session of HBO2 treatment (follow-up HBOx1 group)

Full size image

To look for batch effects or other potential covariates that might cause data bias principal component analysis (PCA) was performed, resulting in exclusion of four samples with TIN < 50 from the differential expression analyses due to possible confounding effects (See Additional file 2, Figure A3).

Gene expression signatures related to effects of intervention with HBO2 treatment

PCA was used to visualize the data in a lower dimension and detect patterns in the data (See Additional file 2, Figure A4). To improve visualization, the PCA was performed separately for the baseline samples (Additional file 2, Figure A5) and the follow-up samples (Fig. 2). The PCA at baseline revealed that the HBO2-treated patients (HBOx1 and HBOx2 groups) were comparable to each other, but the non-HBO2-treated patients formed their own distinct group (see Additional file 2, Figure A5). In the follow-up PCA plot, the two groups of HBO2-treated patients are more separated than at baseline, and the non-HBO group is clearly distinguishable, particularly from the HBOx2 group. The non-HBO group and the HBOx2 group separated mainly on dimension three, which explained 7.8% of the variance, and to a lesser extent on dimension one, which explained 16.7% of the variance (Fig. 2 and Additional file 2, Figure A6).

Fig. 2
figure 2

PCA of individuals at follow-up. a dimension one and two; (b) dimension one and three; (c) dimension two and three. The larger sized dots are the groups mean point

Full size image

A solitary patient in the HBOx2 group was positioned among those patients who had not received the HBO2 treatment in the PCA at follow-up. The patient record revealed that this individual had died shortly after the second treatment session. Two individuals from the non-HBO group did not group with other persons who had not undergone the HBO2 intervention; instead, these two individuals were in close proximity to the patients who had received two HBO2 treatments. According to the patients' records, these two patients had not been referred to HBO2 treatment initially because the surgeon was unsure if they were diagnosed with NSTI or a lesser form of cutaneous infection that is not routinely treated with HBO2. Whereas the remaining patients in the non-HBO group were not treated with HBO2 initially, because they were deemed too unstable by the clinician to enter the pressure chamber.

To examine if the relative position of the individuals in the PCA was associated with disease severity, we plotted the dimension one and three coordinates in the PCA from all follow-up samples against clinical blood test results (Additional file, Fig. 7A). We observed a negative association of dimension three with eGFR and a positive correlation with CRP, as well as a positive association of dimension one with eGFR and a negative correlation with CRP (Fig. 3 and Additional file, Fig. 6A). Because of this association between both dimension one and dimension three with clinical blood tests results, we also plotted these clinical variables against a diagonal between dimension one and three, which did not improve the linearity (plots not shown).

Fig. 3
figure 3

Plots showing linear regression of clinical blood test results and coordinates on dimension one and three. a CRP (C Reactive Protein) level and (b) eGFR (estimated glomerular filtration rate). The red line is coordinates on dimension one, and the blue line is coordinates on dimension three

Full size image

To investigate the effect of treatment on the movement along the PCA dimensions, we calculated the distance travelled (from baseline to follow-up) by each individual in the PCA along dimensions one, two, and three (See Additional file 2, Figure A4). Results revealed that individuals in the HBOx2 group tend to move towards the positive values in dimension one and towards the negative values in dimension three (Fig. 4), that is, they move in the direction of a better clinical output as described by CRP and eGFR. Kruskal Wallis rank sum test found that the difference between groups was significant for dimension one (p = 0.0289) and dimension three (p = 0.0275).

Fig. 4
figure 4

Whisker plot of difference in coordinates between before and after intervention with HBO2 treatment. The y-axis is the difference in coordinates between before and after in the principal component analysis on the three first dimensions

Full size image

Because our groups of interest separated on dimensions one and three, we concentrated our variable selection on explanatory variables related to these dimensions. For variable selection, we utilized the MixOmics package, which was particularly intended for this purpose. The PCA generated by the FactoMineR package and the PCA generated by the MixOmics package are comparable (see Additional file 2, Figure A8). We obtained 30 variables for component one, two, and three respectively. All selected variables on dimension one were protein coding genes with negative loadings (see Additional file 2, Table A1), and the six most contributing variables were NAT10, RRP1B, MYBBP1A, DNAJA3, LAS1L and PPRC1. The selected variables on dimension three all had positive loadings, 27 of them were protein coding gene products, and the six most contributing variables were CEP63, ZDHHC18, KY, ADGRE3, TREM1 and NDE1 (see Additional file 2, Table A2).

To evaluate the importance of each of these genes in separating our three treatment groups we estimated the difference in expression level from baseline to follow-up in each treatment group (Fig. 5 and Additional file 2, Table A1-2). We observed a significant difference in change in expression level between groups for NAT10 (N-Acetyltransferase 10), RRP1B (Ribosomal RNA Processing 1B), DNAJA3 (DnaJ Heat Shock Protein Family (Hsp40) Member A3), LAS1L (LAS1 Like Ribosome Biogenesis Factor), PPRC1 (PPARG Related Coactivator 1), CEP63 (Centrosomal Protein 63), ZDHHC18 (Zinc Finger DHHC-Type Palmitoyltransferase 18), ADGRE3 (Adhesion G Protein-Coupled Receptor E3) and NDE1 (NudE Neurodevelopment Protein 1), with larger changes in expression level between baseline and follow-up in the HBOx2 group (Fig. 5), and for TREM1 (Triggering Receptor Expressed On Myeloid Cells 1) with larger changes in the non-HBO group (See Additional file 2, Table A3).

Fig. 5
figure 5

Group-wise comparison of changes in expression level of the top contributing genes in the sparse PCA. Whiskers plots of the difference in expression level of cpm (counts-per-million)-normalized counts between baseline and follow-up of the six most contributing genes to (a) Dimension one (one outlier in the HBOx1 group was removed to improve visualization) and (b) Dimension three. Genes are giving with their gene symbol

Full size image

Sepsis-NSTI profiles and their individual response to HBO2 treatment

To determine whether there are NSTI-sepsis endotypes that respond differently to HBO2 treatment, we first performed a differential expression analysis before and after treatment, including only the 82 patients who were allocated to the HBO2 intervention groups, and then used the top 100 differentiated genes to construct heatmaps with hierarchical clustering dendrograms of patients at baseline (Fig. 6a) and after HBO2 treatment (Fig. 6b).

Fig. 6
figure 6

Hierarchical clustering and heatmaps of the expression levels of genes that were differentially expressed between before and after HBO2 treatment. a Eighty-two samples taken before intervention with HBO2. b Eighty samples taken after intervention with HBO2. Rows are genes and columns are individuals. The bar at the top of each heatmap indicate treatment group, with blue being patients who received one session of HBO2 and red being patients who received two sessions of HBO2. The intensity of the colour at each intersection point the heatmaps indicates the level of expression for that gene in that sample. Red is high level of expression and blue is low level of expression. The heatmap is scaled by row. We named the two distinct clusters of individuals obtained at baseline Baseline Response 1 (BR2) and Baseline Response 2 (BR1), and two clusters obtained at follow-up Treatment Response 1 (TR1) and Treatment Response 2 (TR2). In the hierarchical clustering at follow-up, we named the two obtained gene clusters gene cluster I and gene cluster II respectively

Full size image

We observed two clusters of patients before and after treatment that may reflect two separate endotypes [Figure A10 and A11]. We named the two distinct clusters of patients obtained at baseline: BR1 (Baseline Response 1) and BR2 (Baseline Response-2). Likewise, we denoted the patient clusters after treatment with HBO2 TR1 (Treatment Response-1) and TR2 (Treatment Response-2) (Fig. 6).

We noted that cluster membership was linked to differences in clinical parameters, especially at follow-up. Specifically, CRP was significantly lower in BR2 207 mg/L (IQR = 107–305 mg/L) compared to BR1 246 mg/L (IQR = 193–340 mg/L), (p = 0.0073), and in TR2; 171 mg/L (IQR = 118–233 mg/L) compared to TR1; 261 mg/L (IQR = 220–360 mg/L), (p = 1.144e-05). Estimated GFR was likewise significantly higher in the TR2 cluster; 77 ml/min/1.73m2 (IQR = 37–90 ml/min/1.73m2) compared to TR1; 48 ml/min/1.73m2 (IQR = 30–68 ml/min/1.73m2), (p = 0.03769), whereas there was no difference in estimated GFR between clusters at baseline (see Additional file 2, Figure A11 and A12). This indicates that the BR2 and TR2 endotypes were associated with a better clinical condition than the BR1 and TR1 endotypes.

Focusing on the gene clusters at follow-up we also observed that, the genes in gene cluster I (See Additional file 2, Table A4) were lower expressed in TR1 than in TR2, and the genes in gene cluster II (see Additional file 2, Table A5) were higher expressed in TR1 than in TR2 (Fig. 6b). To predict the biological function of genes in each of these gene clusters, we then constructed PPI networks from gene cluster I and gene cluster II. Ninety-three percent of the genes in gene cluster I were mapped by the String database. Plotting a PPI network with high confidence interaction score of 0.700 we obtained four distinct PPI interaction networks containing 42 nodes and 14 edges. With four expected number of edges, the PPI enrichment p-value was 0.000253 (see Additional file 2, Figure A13). Three of the obtained PPI networks only contained two proteins coding genes, and one larger network contained eight genes with functions related to immune activities (EOMES, RUNX3, RORC, CD247, CD8A, TNFRSF4, LCK, ZAP70). The analysis also revealed four GO biological processes terms and four KEGG pathways that were functionally enriched in the dataset, these were; “T cell activation” (GO:0042110); “Lymphocyte activation” (GO:0046649), “Cell activation” (GO:0001775), “T cell differentiation” (GO:0030217), “Primary immune deficiency” (hsa05340), “Th1 and Th2 cell differentiation” (hsa04658), “Th17 cell differentiation” (hsa04659) and “T cell receptor signalling pathway” (hsa04660) (See Additional file 2, Table A7).

Fifty-nine percent of the genes in gene cluster II were mapped by the String database, thereby obtaining 31 nodes with no interactions with a high confidence interaction score of 0.700. Likewise, no significant enrichments were found.

When we examined the overlap between the clusters before and after treatment, we found that people allocated to BR1 before treatment were also assigned to TR1 after treatment, and the same was true for BR2 and TR2. There were a few exceptions; seven individuals, all from the HBOx1 group, transferred from BR2 to TR1, while nine individuals moved from BR1 to TR2 (Table A6). Of these nine individuals, four belonged to the HBOx2 group (see Fig. 6). Given that there are only five individuals from the HBOx2 group in BR1, it means that 80% of the individuals in BR1 who received two sessions of HBO2 treatment moved to the TR2 group. In comparison, just five out of 33 (15%) individuals from the HBOx1 group moved from BR1 to TR2, indicating that individuals in the HBOx2 group were 5.3 times more likely to change cluster from BR1 to TR2, than individuals who had received one session of HBO2 (Fig. 6 and Table A6).

One genes was found to be significantly differentially expressed between the group of individuals who moved from BR1 and those who did not move from BR1 during treatment with HBO2. Likewise, one gene was found to be significantly differentially expressed between the group of individuals who moved from BR2 and those who did not move from BR2 during treatment with HBO2 (Table 3 and Additional file 2, Figure A14 and Figure A15).

Table 3 Differential expression table
Full size table

Discussion

The current work examines blood global gene expression patterns in three treatment groups of NSTI patients who received standard treatment and one HBO2 treatment session, two HBO2 treatment sessions, or no HBO2 treatment. We found that after two sessions of HBO2 treatment the global gene expression pattern in blood differs from patients that have not received HBO2 treatment in a direction that is associated with improvement in biochemical blood test results. Furthermore, we found that patients who had received two sessions of HBO2 treatment were more likely than patients who had received one session of HBO2 treatment to change from a compromised sepsis endotype to an endotype associated with improved clinical parameters.

Gene expression signatures related to effects of intervention with HBO2 treatment

Using PCA we discovered that patients who received two sessions of HBO2 treatment altered their gene expression pattern in relation to dimension three at follow-up more than the other two groups, resulting in a gene expression pattern that was distinct from patients who did not receive HBO2 treatment. At baseline we found that the non-HBO patients tend to come together in the PCA, whereas the HBOx1 group and HBOx2 group did not. This could be explained by differences in disease severity at baseline, where the non-HBO patients were more critically ill than the two other groups, indicating that dimension one and three retained information on disease severity. This is confirmed by the observed association of dimensions one and three with clinical variables such as CRP and eGFR (Fig. 3, Figure A7). Thus, the movement of patients in the PCA on dimension one and three can be interpreted as an improvement (positive values in dimension one and negative values in dimension three) or deterioration (opposite directions) of the clinical condition of the patients. Interestingly, we found that HBOx2 patients and some HBOx1 patients with a similar gene expression pattern moved in directions of clinical improvement after treatment and showed higher levels of plasma thrombocytes and higher glomerular filtration rate, as well as lower levels of C reactive protein and creatinine level in plasma (Fig. 3 and Additional file 2, Figure A7). This indicates a positive association between treatment mediated changes in gene expression pattern and improvement in clinical markers of inflammation and organ perfusion. The ability of HBO2 treatment to reduce systemic inflammation has long been investigated and demonstrated in various animal models [30,31,32], and human investigations [33,34,35]. This HBO2-mediated reduction in inflammation has also been linked to increased survival in sepsis animal models [36, 37], and may explain some of the observed favourable effects of HBO2 treatment on survival of patients with NSTI [38, 39]. Here we showed how HBO2-mediated reduction in inflammation translates to changes in blood gene expression pattern from before to after HBO2 treatment, and how those changes in gene expression are connected to the improvement of clinical markers.

We used sparse PCA to characterize the genes that contributed to the different expression patterns in the groups. The model identified at least 10 genes (NAT10, RRP1B, DNAJA3, LAS1L, PPRC1, CEP63, ZDHHC18, ADGRE3, TREM1, NDE1) that had significantly larger changes in expression level between baseline and follow-up in the group who received two sessions of HBO2 compared to the two other groups. In the following we discuss possible implications of the observed regulation of four of those genes that has previously been described in relation to immune function and sepsis. We found that the heat shock protein DNAJA3 was upregulated in the HBOx2 group and downregulated in the two other groups. DNAJA3 has previously been identified as a ROS-associated biomarker predicting survival in sepsis, where it was found to be downregulated in non-survivors [40]. The expression levels of the two genes NAT10 and ADGRE3 were also found to be higher following two sessions of HBO2 treatment (Additional file 2, Table A1). A recent study demonstrated that NAT10 expression decreased in septic mice, and the decreased level of NAT10 contributed to the progression of sepsis by exacerbating neutrophil pyroptosis [41]. Also, an RNA sequencing study of LPS-treated cells found that NAT10 decreased during LPS-induced inflammation, and functional enrichment analysis suggested that NAT10 activated the NF-κB signalling pathway by regulating ROS generation [42]. ADGRE3 has been discovered as one of seven predictive markers for categorising patients into two different sepsis phenotypes in a prospective cohort study of global gene expression in peripheral blood leucocytes from 265 sepsis patients admitted to ICU, and ADGRE3 expression was lower in the immunocompromised phenotype, which was related to increased 14-day mortality [15]. This indicates that the change in gene expression of these genes in the HBOx2 group is related to known immune adaptive cellular activity in sepsis. On the other hand, TREM1, a signalling receptor that has been linked to septic shock and may be important in systemic infections by increasing inflammation [43, 44], was also found to increase in the HBOx2 group. However, this gene also increased in expression level in the nonHBO group. The other six genes found in the sparse PCA have not, as far as we are aware, been linked to outcomes in sepsis or severe infections. Nonetheless, our findings suggest that the expression level of these 10 genes are increased when two sessions of HBO2 are administered to patients with sepsis due to NSTI, and that this change in the expression level is associated with lower CRP plasma levels and higher eGFR.

Sepsis-NSTI profiles and their individual response to HBO2 treatment

The hierarchical clustering analysis found two distinct clusters of patients both before and after intervention with HBO2 treatment (Additional file 2, Figure A9 and A10). In both conditions, one patient cluster (BR1/TR1) had blood test results associated with a worse clinical status, demonstrating that these two groups were more seriously afflicted by the illness than their counterparts BR2/TR2 (see Additional file 2, Figures A11 and A12). This is in line with other studies that identified sepsis endotypes using gene expression data and found two to four clusters with one immune-adaptive endotype and one immune-compromised endotype [15, 16, 45,46,47,48]. The association to blood test results were most pronounced in the clusters TR1 and TR2 obtained after intervention with HBO2 treatment. To examine the molecular properties of TR1 and TR2, we constructed PPI networks from the genes in the two discovered gene clusters. We focus this discussion on networks comprising genes previously linked to inflammation and the sepsis syndrome, considering that genes in gene cluster I were lower expressed in TR1 than in TR2, whereas genes in gene cluster II were higher expressed in TR1 than in TR2 (Fig. 6).

Gene cluster I contained a network with genes known to be involved in immune functions, which included eight protein-coding genes (EOMES, RUNX3, RORC, CD247, CD8A, TNFRSF4, LCK, ZAP70). EOMES (Eomesodermin) as well as GZMA (Granzyme A) and FGFBP2 (Fibroblast Growth Factor Binding Protein 2), two other genes in our gene cluster I, have previously been linked to sepsis development in a co-expression module analysis investigating genome-wide alterations in mRNA and long non-coding RNAs in sepsis, where these three genes were among 11 downregulated genes shared by six sepsis microarray datasets [49]. A different study investigating transcriptomic profiles in children with immune paralysis as a consequence of sepsis also found that EOMES downregulation was associated with immune paralysis and worse prognosis [50]. In line with this, the two T cell surface glycoproteins, CD8A (CD8 subunit Alpha) and CD247 (CD247 Molecule) have been shown to be downregulated in a sepsis response signature characterised by relative immune suppression, endotoxin tolerance, T-cell exhaustion, and metabolic derangement [15]. CD247 alone is considered a molecular target of septic shock as low expression of this gene has been associated with haemorrhage, necrosis, septic shock and a negative correlation to prognosis [51]. The pivotal role of low expression of CD247 and other genes in our gene network I has been confirmed in more studies. CD247 together with LCK (LCK Proto-Oncogene, Src Family Tyrosine Kinase) and ZAP70 (Zeta Chain Of T Cell Receptor Associated Protein Kinase 70), were among seven genes identified as potential new biomarkers for sepsis in a study comparing whole blood RNA samples from 89 sepsis patients with 67 healthy controls, where the relative expression of these three genes were lower in sepsis patients [52]. Another study with a similar design likewise found CD247 to be central in sepsis, and single cell sequencing revealed that this molecule was primarily expressed by T cells, decreased in sepsis and positive correlated with survival rate. The study also found ZAP70 in the same co-expression network of sepsis associated genes with decreased expression [53]. ZAP70 is an adaptor protein in kinase pathways that upon inhibition prevents T cell activity, proper metabolic activity, T cell survival, and proliferation in sepsis [54]. Additional genes in our gene cluster I have likewise been related to immunosuppression when present at low levels in sepsis. TNFRSF4 encodes a receptor of the tumour necrosis factor superfamily named OX40, which has been suggested to reverse sepsis induced immunosuppression upon stimulation in a model of polymicrobial sepsis [55, 56]. Finally, the transcription factors encoded by the genes RORC (RAR Related Orphan Receptor C) [57] and RUNX3 (Runt-related transcription factor 3) both have regulatory roles in the formation of immunosuppressive regulatory T cells (Tregs) [58, 59]. Changes in RUNX3 and RORC expression in peripheral blood have been linked to inflammatory disorders [59, 60], but the implication in sepsis remains to be elucidated. When considered together, previous studies suggests that decreased expression of this gene network in gene cluster I in TR1 may be linked to T helper cell exhaustion, immunosuppression, and a poor prognosis in sepsis. Likewise, the enriched terms and pathways in this lowly expressed gene set were lymphocyte and T cell activation as well as T cell differentiation.

At a confidence interaction level of 0.700, gene cluster II did not establish a PPI network. However, IL-10 (Interleukine-10), a crucial cytokine in the anti-inflammatory immunosuppressive response and a major predictor of severity and catastrophic outcome in sepsis, was included in this gene set and was therefore shown to be more expressed in TR1 than TR2. At protein level IL-10 has previously been demonstrated to be positively correlated with serum CRP in patients with severe infections [61], which is consistent with our results, where plasma CRP levels were significantly higher in TR1.

Hence, both the clinical blood test results and the gene expression pattern in TR1 indicate that this phenotype is an immune-compromised profile, as compared to TR2. The subsequent overlaps analysis revealed that intervention with two sessions of HBO2 treatment may induce adaptive gene expression changes that cause individuals to transition from the immunosuppressive phenotype TR1 to a more immune-adaptive phenotype TR2, as individuals who received two sessions of HBO2 treatment were 5.3 times more likely than individuals who received one session of HBO2 treatment to transition to the TR2 phenotype. Interestingly, we have previously shown that most of the genes (EOMES, RUNX3, RORC, CD247, CD8A, LCK, ZAP70) that were contained in the larger PPI network of gene cluster I, and hence is lowly expressed in the immune-suppressive endotype TR1, were upregulated following intervention with HBO2 treatment [14]. Likewise, the terms and pathways that are enriched in gene cluster I (T cell activation, Th1 and Th2 cell differentiation, Th17 cell differentiation) were also enriched in upregulated genes following intervention with HBO2 treatment [14].

To identify the genetic signature of patients who responded to treatment by transitioning endotype, we performed a differentially expression analysis of the difference in gene expression between this group and the patients assigned to the same endotype that did not respond by changing endotype. In the group who transitioned from the immunosuppressive endotype to a more immune-adaptive endotype we found that one gene, MTCO2PI2, was significantly higher expressed (Table 3 and Additional file 2, Figure A14). MTCO2PI2 (Mitochondrially Encoded Cytochrome C Oxidase II Pseudogene 12). This gene had a higher expression level in the group that transitioned from the immune-compromised phenotype (Table 3). MTCO2PI2 is a pseudogene that has received scant attention in the literature; however, it is predicted to be involved in mitochondrial electron transport, where it contributes to cytochrome-c oxidase activity as part of the respiratory chain complex IV [provided by Alliance of Genome Resources, Apr 2022], which is also considered a primary cellular target of HBO2 treatment [7]. In the group who transitioned to a worse endotype, one pseudogene, CD177P1, which encodes an uncharacterized protein was significantly higher expressed, when compared to the individuals who did not change endotype.

Strength and limitations

One limitation of this study is the lack of some commonly used clinical markers of disease severity. Endotypes discovered in sepsis are frequently related to long term outcomes such as ICU admission length, days on a ventilator and death. In the current study we focused on individual differences in response to treatment with HBO2, and therefore we included clinical parameters within a narrow time range from the intervention. The retrospective design limited our access to clinical data on the included patients, and data such as arterial blood gas test measures, including the important parameter lactate were not available before and after intervention. Important prognostic measures such as the disease severity scores SAPS II, and SOFA are calculated using the worst measured value of a collection of clinical and biochemical indicators within 24 h. Because most patients in our study had their first HBO2 session within the first 24 h of admission and our samples were collected close to the intervention, these severity scores may have been calculated using measurements taken both before and after HBO2 treatment, rendering these variables inapplicable in our study design. The clinical markers of disease severity in our study were obtained from electronic patient records and was limited to a panel of measurements that was frequently acquired in this patient group, and consistently measured as routine parameters over the study period of 2013 −2017. Both a low measurement of eGFR and high CRP in plasma are known risk factors in sepsis. Sepsis-associated acute kidney injury is independently associated with increased mortality, as demonstrated in a recent retrospective cohort study of 84,528 ICU sepsis admissions [62]. Decreased eGFR has been found to be associated with increased mortality in a cohort of 802 septic patients admitted to ICU in both an unadjusted analysis and in an analysis adjusted for age, sex, comorbidities and creatinine level [63], and is hence a valid measure for showing that the treatment response phenotype TR1 has a worst prognosis in our study. CRP is an accepted sepsis biomarker [64], and high CRP values are associated with bacterial infections, severe sepsis and septic shock [64, 65]. However, CRP is not an independent predictor of mortality in sepsis, and CRP may be within the range observed in apparently healthy individuals, even in septic patients with high in-hospital mortality [66]. Another limitation of the study is the imbalance in disease severity across groups at baseline, with more severely ill patients in the nonHBO group. Biologically this restricts our study of effects of HBO2 treatment to the patient group which is less severely ill. Moreover, this imbalance between groups, as well as the small number of samples in the nonHBO group, made it difficult to isolate effects of HBO2 treatment, i.e., using interaction effects in the differential expression analysis was not applicable due to the biological imbalance between groups, and the small sample size of the nonHBO group implies less power of the differential expression analysis in this group compared to the intervention groups, making group-wise comparisons in gene expression difficult. We therefore used PCA to explore the effects of HBO2 treatment on the blood global gene expression.

PCA is useful for presenting patterns in multidimensional data because it condenses data into fewer components that account for the most variance in the data, thereby capturing the signal in the data. The disadvantages of PCA are the low interpretability of the resultant dimensions, that are orthogonal projections of the features from the original data, which cannot tell what the most important features in the dataset are. To approach this, we used sPCA for variable selection. The subsequent univariate analysis showed that the sPCA model had selected variables that had larger changes in the HBOx2 group, which indicates that the model had successfully selected important variables. However, we only presented the six most contributing variables to the two dimensions of interest. An overrepresentation analysis of these 30 genes against all genes expressed in our dataset, did not reveal any significant results (data not shown), and therefore we applied a univariate approach with the arbitrary cut off of “top-six”. We cannot rule out that some of the other twenty-four genes have an equally important influence, and therefore we decided to display them all in the supplementary results (see Additional file 2, Table A1 and A2).

Similarly, when using hierarchical clustering it is necessary to determine both the distance matrix and the linkage criteria, and there is no strong theoretical basis for these decisions, which again tenants output variation. Due to space complexity, the approach is not suitable for huge datasets, thus we only included the top 100 differentially expressed genes in the study, which again is another discretionary option, and this decision is projected further into the number of genes included in the PPI network analysis. However, the hierarchical clustering and the belonging dendrogram is both easy to interpret and sensitive to outliers which makes it applicable in a study like ours. In our study design, one strength is that we employed both PCA and hierarchical clustering to investigate relationships between group-wise changes in expression pattern and clinical blood test results. Both techniques detected the same outlier and provided indication for the beneficial effect of two HBO2 treatment sessions. Thus, the applied omics approach renders a data driven holistic representation of the complex molecular mechanisms that underpin HBO2 treatment intervention in this heterogeneous group of severe ill NSTI patients.

Conclusion

In summary, in this study of differences in blood global gene expression between NSTI sepsis patients who were and were not treated with HBO2, both the PCA and hierarchical clustering results demonstrated adaptive changes in gene expression pattern in response to HBO2 intervention, both in relation to clinical markers of disease severity and the known molecular function of influential genes, and these effects were most pronounced in the group that was treated with two sessions of HBO2. The study did not uncover a complex sepsis-NSTI endotype that was more likely to respond to intervention with HBO2, but it did identify a few genes that, when highly expressed, may regulate the efficacy of adjuvant HBO2 in patients with sepsis caused by NSTI.

Data availability

The gene expression data generated and analysed in the current study have not been deposited in any repositories. The Danish National Committee on Health Research Ethics does not approve deposition of genomic data from living individuals in public repositories. Data sharing may only take place after entering into a personal data processing agreement between Data Controller and Data Processor, which has been approved by the Danish Capital Region at Knowledge Centre for Data reviews.

Abbreviations

ADGRE3:

Adhesion G Protein-Coupled Receptor E3

BR1:

Baseline response 1

BR2:

Baseline response 2

CD8A:

CD8 subunit Alpha

CEP63:

Centrosomal Protein 63

cpm:

Counts per million

CRP:

C-reactive protein (mg/L)

DNAJA3:

DnaJ Heat Shock Protein Family (Hsp40) Member A3

eGFR:

Estimated glomerular filtration rate (ml/min/1.73m2)

EOMES:

Eomesodermin

FGFBP2:

Fibroblast Growth Factor Binding Protein 2

GO:

Gene Ontology

GZMA:

Granzyme A

HBO2 :

Hyperbaric oxygen

HBOx1:

Group that had received two sessions of hyperbaric oxygen treatment

HBOx2:

Group that had received two sessions of hyperbaric oxygen treatment

ICU:

Intensive care unit

IQR:

Interquartile range

KEGG:

Kyoto Encyclopedia of Genes and Genomes

kPa:

Kilopascal

K Y:

Kyphoscoliosis Peptidase

LAS1L:

LAS1 Like Ribosome Biogenesis Factor

LCK:

LCK Proto-Oncogene Src Family Tyrosine Kinase

LogFC:

Log2 Fold Change

min:

Minuts

MTCO2PI2:

Mitochondrially Encoded Cytochrome C Oxidase II Pseudogene 12

MYBBP1A:

MYB Binding Protein 1a

NAT10:

N-Acetyltransferase 10

NDE1:

NudE Neurodevelopment Protein 1

non-HBO:

Group that had not received hyperbaric oxygen treatment

NSTI:

Necrotizing soft tissue infections

PCA:

Principal component analysis

PPI:

Protein protein ineraction

PPRC1:

PPARG Related Coactivator 1

RORC:

RAR Related Orphan Receptor C

RQN:

RNA Quality Number

RRP1B:

Ribosomal RNA Processing 1B

RUNX 3:

Runt-related transcription factor 3

SAPS II:

Simplified Acute Physiology Score

Sec:

Seconds

SOFA:

Sequential Organ Failure Asessment

TIN:

Transcript Integrity Number

TNFRSF4:

TNF Receptor Superfamily Member 4

TR1:

Treatment response 1

TR2:

Treatment response 2

TREM1:

Triggering Receptor Expressed On Myeloid Cells 1

v:

Version

ZAP70:

Zeta Chain of T Cell Receptor Associated Protein Kinase 70

ZDHHC18:

Zinc Finger DHHC-Type Palmitoyltransferase 18

References

  1. Stevens DL, Bisno AL, Chambers HF, Dellinger EP, Goldstein EJ, Gorbach SL, et al. Practice guidelines for the diagnosis and management of skin and soft tissue infections: 2014 update by the infectious diseases society of America. Clin Infect Dis. 2014;59:147–59. https://doiorg.publicaciones.saludcastillayleon.es/10.1093/cid/ciu296.

    Article  PubMed  Google Scholar 

  2. Madsen MB, Skrede S, Perner A, Arnell P, Nekludov M, Bruun T, et al. Patient’s characteristics and outcomes in necrotising soft-tissue infections: results from a Scandinavian, multicentre, prospective cohort study. Intensive Care Med. 2019;45:1241–51. https://doiorg.publicaciones.saludcastillayleon.es/10.1007/s00134-019-05730-x.

    Article  CAS  PubMed  Google Scholar 

  3. Carre JE, Singer M. Cellular energetic metabolism in sepsis: the need for a systems approach. Biochim Biophys Acta. 2008;1777:763–71. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.bbabio.2008.04.024.

    Article  CAS  PubMed  Google Scholar 

  4. Evans L, Rhodes A, Alhazzani W, Antonelli M, Coopersmith CM, French C, et al. Surviving sepsis campaign: international guidelines for management of sepsis and septic shock 2021. Intensive Care Med. 2021;47:1181–247. https://doiorg.publicaciones.saludcastillayleon.es/10.1007/s00134-021-06506-y.

    Article  PubMed  PubMed Central  Google Scholar 

  5. Levett D, Bennett MH, Millar I. Adjunctive hyperbaric oxygen for necrotizing fasciitis. Cochrane Database Syst Rev. 2015;1:CD007937. https://doiorg.publicaciones.saludcastillayleon.es/10.1002/14651858.CD007937.pub2.

    Article  PubMed  Google Scholar 

  6. Hadanny A, Efrati S. The hyperoxic-hypoxic paradox. Biomolecules. 2020;10:958. https://doiorg.publicaciones.saludcastillayleon.es/10.3390/biom10060958.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  7. Schottlender N, Gottfried I, Ashery U. Hyperbaric oxygen treatment: effects on mitochondrial function and oxidative stress. Biomolecules. 2021;11:1827. https://doiorg.publicaciones.saludcastillayleon.es/10.3390/biom11121827.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  8. Lin HC, Wan FJ, Wu CC, Tung CS, Wu TH. Hyperbaric oxygen protects against lipopolysaccharide-stimulated oxidative stress and mortality in rats. Eur J Pharmacol. 2005;508:249–54. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.ejphar.2004.12.021.

    Article  CAS  PubMed  Google Scholar 

  9. Grimberg-Peters D, Buren C, Windolf J, Wahlers T, Paunel-Gorgulu A. Hyperbaric oxygen reduces production of reactive oxygen species in neutrophils from polytraumatized patients yielding in the inhibition of p38 MAP kinase and downstream pathways. PLoS ONE. 2016;11:e0161343. https://doiorg.publicaciones.saludcastillayleon.es/10.1371/journal.pone.0161343.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  10. Hedetoft M, Jensen PO, Moser C, Vinkel J, Hyldegaard O. Hyperbaric oxygen treatment impacts oxidative stress markers in patients with necrotizing soft-tissue infection. J Investig Med. 2021;69:1330–8. https://doiorg.publicaciones.saludcastillayleon.es/10.1136/jim-2021-001837.

    Article  PubMed  PubMed Central  Google Scholar 

  11. Eltzschig HK, Carmeliet P. Hypoxia and inflammation. N Engl J Med. 2011;364:656–65. https://doiorg.publicaciones.saludcastillayleon.es/10.1056/NEJMra0910283.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  12. Liang F, Kang N, Liu X, Yang J, Li Z, Tan JW. Effect of HMGB1/NF-kappaB in hyperbaric oxygen treatment on decreasing injury caused by skin flap grafts in rats. Eur Rev Med Pharmacol Sci. 2013;17:2010–8.

    CAS  PubMed  Google Scholar 

  13. Tan J, Zhang F, Liang F, Wang Y, Li Z, Yang J, et al. Protective effects of hyperbaric oxygen treatment against spinal cord injury in rats via toll-like receptor 2/nuclear factor-kappaB signaling. Int J Clin Exp Pathol. 2014;7:1911–9.

    PubMed  PubMed Central  Google Scholar 

  14. Vinkel J, Rib L, Buil A, Hedetoft M, Hyldegaard O. Key pathways and genes that are altered during treatment with hyperbaric oxygen in patients with sepsis due to necrotizing soft tissue infection (HBOmic study). Eur J Med Res. 2023;28:507. https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s40001-023-01466-z.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  15. Davenport EE, Burnham KL, Radhakrishnan J, Humburg P, Hutton P, Mills TC, et al. Genomic landscape of the individual host response and outcomes in sepsis: a prospective cohort study. Lancet Respir Med. 2016;4:259–71. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/S2213-2600(16)00046-1.

    Article  PubMed  PubMed Central  Google Scholar 

  16. Scicluna BP, van Vught LA, Zwinderman AH, Wiewel MA, Davenport EE, Burnham KL, et al. Classification of patients with sepsis according to blood genomic endotype: a prospective cohort study. Lancet Respir Med. 2017;5:816–26. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/S2213-2600(17)30294-1.

    Article  PubMed  Google Scholar 

  17. Madsen MB, Skrede S, Bruun T, Arnell P, Rosen A, Nekludov M, et al. Necrotizing soft tissue infections - a multicentre, prospective observational study (INFECT): protocol and statistical analysis plan. Acta Anaesthesiol Scand. 2017. https://doiorg.publicaciones.saludcastillayleon.es/10.1111/aas.13024.

    Article  PubMed  Google Scholar 

  18. Vinkel J, Rib L, Buil A, Hedetoft M, Hyldegaard O. Investigating the Effects of hyperbaric oxygen treatment in necrotizing soft tissue infection with transcriptomics and machine learning (the HBOmic Study): protocol for a prospective cohort study with data validation. JMIR Res Protoc. 2022;11:e39252. https://doiorg.publicaciones.saludcastillayleon.es/10.2196/39252.

    Article  PubMed  PubMed Central  Google Scholar 

  19. Singer M, Deutschman CS, Seymour CW, Shankar-Hari M, Annane D, Bauer M, et al. The third international consensus definitions for Sepsis and Septic Shock (Sepsis-3). JAMA. 2016;315:801–10. https://doiorg.publicaciones.saludcastillayleon.es/10.1001/jama.2016.0287.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  20. Bolger AM, Lohse M, Usadel B. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics. 2014;30:2114–20. https://doiorg.publicaciones.saludcastillayleon.es/10.1093/bioinformatics/btu170.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  21. Dobin A, Davis CA, Schlesinger F, Drenkow J, Zaleski C, Jha S, et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics. 2013;29:15–21. https://doiorg.publicaciones.saludcastillayleon.es/10.1093/bioinformatics/bts635.

    Article  CAS  PubMed  Google Scholar 

  22. Robinson MD, McCarthy DJ, Smyth GK. edgeR: a Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics. 2010;26:139–40. https://doiorg.publicaciones.saludcastillayleon.es/10.1093/bioinformatics/btp616.

    Article  CAS  PubMed  Google Scholar 

  23. Lê S, Josse J, Husson F. FactoMineR: An R package for multivariate analysis. J Stat Softw. 2008;25:1–18. https://doiorg.publicaciones.saludcastillayleon.es/10.18637/jss.v025.i01.

    Article  Google Scholar 

  24. Lê Cao K-A, González I, Déjean S. integrOmics: an R package to unravel relationships between two omics datasets. Bioinformatics. 2009;25:2855–6. https://doiorg.publicaciones.saludcastillayleon.es/10.1093/bioinformatics/btp515.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  25. Lê Cao K-A, Boitard S, Besse P. Sparse PLS discriminant analysis: biologically relevant feature selection and graphical displays for multiclass problems. BMC Bioinformatics. 2011;12:253. https://doiorg.publicaciones.saludcastillayleon.es/10.1186/1471-2105-12-253.

    Article  PubMed  PubMed Central  Google Scholar 

  26. Yao F, Coquery J, Lê Cao K-A. Independent Principal Component Analysis for biologically meaningful dimension reduction of large biological data sets. BMC Bioinformatics. 2012;13:24. https://doiorg.publicaciones.saludcastillayleon.es/10.1186/1471-2105-13-24.

    Article  PubMed  PubMed Central  Google Scholar 

  27. Langfelder P, Horvath S. WGCNA: an R package for weighted correlation network analysis. BMC Bioinformatics. 2008;9:559. https://doiorg.publicaciones.saludcastillayleon.es/10.1186/1471-2105-9-559.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  28. Carlson M, Falcon S, Pages H, Li N. org. Hs. eg. db: Genome wide annotation for Human. R package version. 2019;3:3.

  29. Szklarczyk D, Franceschini A, Wyder S, Forslund K, Heller D, Huerta-Cepas J, et al. STRING v10: protein-protein interaction networks, integrated over the tree of life. Nucleic Acids Res. 2015;43:447–52. https://doiorg.publicaciones.saludcastillayleon.es/10.1093/nar/gku1003.

    Article  CAS  Google Scholar 

  30. Slotman GJ. Hyperbaric oxygen in systemic inflammation … HBO is not just a movie channel anymore. Crit Care Med. 1998;26:1932–3. https://doiorg.publicaciones.saludcastillayleon.es/10.1097/00003246-199812000-00002.

    Article  CAS  PubMed  Google Scholar 

  31. Vinkel J, Arenkiel B, Hyldegaard O. The mechanisms of action of hyperbaric oxygen in restoring host homeostasis during sepsis. Biomolecules. 2023;13:1228. https://doiorg.publicaciones.saludcastillayleon.es/10.3390/biom13081228.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  32. Gouveia D, Chichorro M, Cardoso A, Carvalho C, Silva C, Coelho T, et al. Hyperbaric oxygen therapy in systemic inflammatory response syndrome. Vet Sci. 2022;9:33. https://doiorg.publicaciones.saludcastillayleon.es/10.3390/vetsci9020033.

    Article  PubMed  PubMed Central  Google Scholar 

  33. Woo J, Min JH, Lee YH, Roh HT. Effects of hyperbaric oxygen therapy on inflammation, oxidative/antioxidant balance, and muscle damage after acute exercise in normobaric, normoxic and hypobaric, hypoxic environments: a pilot study. Int J Environ Res Public Health. 2020;17:7377. https://doiorg.publicaciones.saludcastillayleon.es/10.3390/ijerph17207377.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  34. Capó X, Monserrat-Mesquida M, Quetglas-Llabrés M, Batle JM, Tur JA, Pons A, et al. Hyperbaric oxygen therapy reduces oxidative stress and inflammation, and increases growth factors favouring the healing process of diabetic wounds. Int J Mol Sci. 2023;24:7040. https://doiorg.publicaciones.saludcastillayleon.es/10.3390/ijms24087040.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  35. De Wolde SD, Hulskes RH, Weenink RP, Hollmann MW, Van Hulst RA. the effects of hyperbaric oxygenation on oxidative stress. Inflamm Angiogenesis Biomol. 2021;11:1210. https://doiorg.publicaciones.saludcastillayleon.es/10.3390/biom11081210.

    Article  CAS  Google Scholar 

  36. Yang HW, Choi S, Song H, Lee MJ, Kwon JE, Lee HAR, et al. Effect of hyperbaric oxygen therapy on acute liver injury and survival in a rat cecal slurry peritonitis model. Life (Basel). 2020;10:283. https://doiorg.publicaciones.saludcastillayleon.es/10.3390/life10110283.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  37. Halbach JL, Prieto JM, Wang AW, Hawisher D, Cauvi DM, Reyes T, et al. Early hyperbaric oxygen therapy improves survival in a model of severe sepsis. Am J Physiol Regul Integr Comp Physiol. 2019;317:160–8. https://doiorg.publicaciones.saludcastillayleon.es/10.1152/ajpregu.00083.2019.

    Article  CAS  Google Scholar 

  38. Wilkinson D, Doolette D. Hyperbaric oxygen treatment and survival from necrotizing soft tissue infection. Arch Surg. 2004;139:1339–45. https://doiorg.publicaciones.saludcastillayleon.es/10.1001/archsurg.139.12.1339.

    Article  PubMed  Google Scholar 

  39. Devaney B, Frawley G, Frawley L, Pilcher DV. Necrotising soft tissue infections: the effect of hyperbaric oxygen on mortality. Anaesth Intensive Care. 2015;43:685–92. https://doiorg.publicaciones.saludcastillayleon.es/10.1177/0310057X1504300604.

    Article  CAS  PubMed  Google Scholar 

  40. Bime C, Zhou T, Wang T, Slepian MJ, Garcia JG, Hecker L. Reactive oxygen species-associated molecular signature predicts survival in patients with sepsis. Pulm Circ. 2016;6:196–201. https://doiorg.publicaciones.saludcastillayleon.es/10.1086/685547.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  41. Zhang H, Chen Z, Zhou JA, Gu J, Wu H, Jiang Y, et al. NAT10 regulates neutrophil pyroptosis in sepsis via acetylating ULK1 RNA and activating STING pathway. Commun Biol. 2022;5:916. https://doiorg.publicaciones.saludcastillayleon.es/10.1038/s42003-022-03868-x.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  42. Zhang Z, Zhang Y, Cai Y, Li D, He J, Feng Z, et al. NAT10 regulates the LPS-induced inflammatory response via the NOX2-ROS-NF-κB pathway in macrophages. Biochim Biophys Acta Mol Cell Res. 2023;1870:119521. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.bbamcr.2023.119521.

    Article  CAS  PubMed  Google Scholar 

  43. Arts RJ, Joosten LA, van der Meer JW, Netea MG. TREM-1: intracellular signaling pathways and interaction with pattern recognition receptors. J Leukoc Biol. 2013;93:209–15. https://doiorg.publicaciones.saludcastillayleon.es/10.1189/jlb.0312145.

    Article  CAS  PubMed  Google Scholar 

  44. Bouchon A, Facchetti F, Weigand MA, Colonna M. TREM-1 amplifies inflammation and is a crucial mediator of septic shock. Nature. 2001;410:1103–7. https://doiorg.publicaciones.saludcastillayleon.es/10.1038/35074114.

    Article  CAS  PubMed  Google Scholar 

  45. Zhou W, Zhang C, Zhuang Z, Zhang J, Zhong C. Identification of two robust subclasses of sepsis with both prognostic and therapeutic values based on machine learning analysis. Front Immunol. 2022;13:1040286. https://doiorg.publicaciones.saludcastillayleon.es/10.3389/fimmu.2022.1040286.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  46. Antcliffe DB, Burnham KL, Al-Beidh F, Santhakumaran S, Brett SJ, Hinds CJ, et al. Transcriptomic signatures in sepsis and a differential response to steroids. From the VANISH randomized trial. Am J Respir Crit Care Med. 2019;199:980–6. https://doiorg.publicaciones.saludcastillayleon.es/10.1164/rccm.201807-1419OC.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  47. Sweeney TE, Azad TD, Donato M, Haynes WA, Perumal TM, Henao R, et al. Unsupervised analysis of transcriptomics in bacterial sepsis across multiple datasets reveals three robust clusters. Crit Care Med. 2018;46:915–25. https://doiorg.publicaciones.saludcastillayleon.es/10.1097/CCM.0000000000003084.

    Article  PubMed  PubMed Central  Google Scholar 

  48. Maslove DM, Tang BM, McLean AS. Identification of sepsis subtypes in critically ill adults using gene expression profiling. Crit Care. 2012;16:183. https://doiorg.publicaciones.saludcastillayleon.es/10.1186/cc11667.

    Article  Google Scholar 

  49. Liu X, Hong C, Jiang Y, Li W, Chen Y, Ma Y, et al. Co-expression module analysis reveals high expression homogeneity for both coding and non-coding genes in sepsis. BMC Genomics. 2023;24:418. https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12864-023-09460-9.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  50. Snyder A, Jedreski K, Fitch J, Wijeratne S, Wetzel A, Hensley J, et al. Transcriptomic Profiles in Children With Septic Shock With or Without Immunoparalysis. Front Immunol. 2021;12:733834. https://doiorg.publicaciones.saludcastillayleon.es/10.3389/fimmu.2021.733834.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  51. Yang Q, Feng Z, Ding D, Kang C. CD3D and CD247 are the molecular targets of septic shock. Medicine (Baltimore). 2023;102:34295. https://doiorg.publicaciones.saludcastillayleon.es/10.1097/md.0000000000034295.

    Article  Google Scholar 

  52. Gong FC, Ji R, Wang YM, Yang ZT, Chen Y, Mao EQ, et al. Identification of potential biomarkers and immune features of sepsis using bioinformatics analysis. Mediators Inflamm. 2020;2020:3432587. https://doiorg.publicaciones.saludcastillayleon.es/10.1155/2020/3432587.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  53. Tian Y, Wang C, Lu Q, Zhang C, Hu L, Ling J, et al. Screening of potential immune-related genes expressed during sepsis using gene sequencing technology. Sci Rep. 2023;13:4258. https://doiorg.publicaciones.saludcastillayleon.es/10.1038/s41598-022-23062-7.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  54. Wakeley ME, Gray CC, Monaghan SF, Heffernan DS, Ayala A. Check point inhibitors and their role in immunosuppression in sepsis. Crit Care Clin. 2020;36:69–88. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.ccc.2019.08.006.

    Article  PubMed  Google Scholar 

  55. Unsinger J, Walton AH, Blood T, Tenney DJ, Quigley M, Drewry AM, et al. Frontline Science: OX40 agonistic antibody reverses immune suppression and improves survival in sepsis. J Leukoc Biol. 2021;109:697–708. https://doiorg.publicaciones.saludcastillayleon.es/10.1002/jlb.5hi0720-043r.

    Article  CAS  PubMed  Google Scholar 

  56. Sherwood ER, Williams DL. Reversal of sepsis-induced T cell dysfunction: OX-40 to the rescue? J Leukoc Biol. 2021;109:689–91. https://doiorg.publicaciones.saludcastillayleon.es/10.1002/jlb.3ce0720-468.

    Article  CAS  PubMed  Google Scholar 

  57. Khader SA, Gopal R. IL-17 in protective immunity to intracellular pathogens. Virulence. 2010;1:423–7. https://doiorg.publicaciones.saludcastillayleon.es/10.4161/viru.1.5.12862.

    Article  PubMed  PubMed Central  Google Scholar 

  58. Kondĕlková K, Vokurková D, Krejsek J, Borská L, Fiala Z, Ctirad A. Regulatory T cells (TREG) and their roles in immune system with respect to immunopathological disorders. Acta Medica (Hradec Kralove). 2010;53:73–7. https://doiorg.publicaciones.saludcastillayleon.es/10.14712/18059694.2016.63.

    Article  PubMed  Google Scholar 

  59. Korinfskaya S, Parameswaran S, Weirauch MT, Barski A. Runx transcription factors in T cells-what is beyond thymic development? Front Immunol. 2021;12:701924. https://doiorg.publicaciones.saludcastillayleon.es/10.3389/fimmu.2021.701924.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  60. El-Karaksy SM, Raafat HA, Abadir MN, Hanna MO. Down-regulation of expression of retinoid acid-related orphan receptor C (RORC) in systemic lupus erythematosus. J Recept Signal Transduct Res. 2016;36:207–12. https://doiorg.publicaciones.saludcastillayleon.es/10.3109/10799893.2015.1075042.

    Article  CAS  PubMed  Google Scholar 

  61. Han H, Ma Q, Li C, Liu R, Zhao L, Wang W, et al. Profiling serum cytokines in COVID-19 patients reveals IL-6 and IL-10 are disease severity predictors. Emerging Microbes Infect. 2020;9:1123–30. https://doiorg.publicaciones.saludcastillayleon.es/10.1080/22221751.2020.1770129.

    Article  CAS  Google Scholar 

  62. White KC, Serpa-Neto A, Hurford R, Clement P, Laupland KB, See E, et al. Sepsis-associated acute kidney injury in the intensive care unit: incidence, patient characteristics, timing, trajectory, treatment, and associated outcomes. A multicenter, observational study. Intensive Care Med. 2023;49:1079–89. https://doiorg.publicaciones.saludcastillayleon.es/10.1007/s00134-023-07138-0.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  63. Linné E, Elfström A, Åkesson A, Fisher J, Grubb A, Pettilä V, et al. Cystatin C and derived measures of renal function as risk factors for mortality and acute kidney injury in sepsis – A post-hoc analysis of the FINNAKI cohort. J Crit Care. 2022;72:154148. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.jcrc.2022.154148.

    Article  CAS  PubMed  Google Scholar 

  64. Castelli GP, Pognani C, Meisner M, Stuani A, Bellomi D, Sgarbi L. Procalcitonin and C-reactive protein during systemic inflammatory response syndrome, sepsis and organ dysfunction. Crit Care. 2004;8:234. https://doiorg.publicaciones.saludcastillayleon.es/10.1186/cc2877.

    Article  Google Scholar 

  65. Ryoo SM, Han KS, Ahn S, Shin TG, Hwang SY, Chung SP, et al. The usefulness of C-reactive protein and procalcitonin to predict prognosis in septic shock patients: a multicenter prospective registry-based observational study. Sci Rep. 2019;9:6579. https://doiorg.publicaciones.saludcastillayleon.es/10.1038/s41598-019-42972-7.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  66. Wasserman A, Karov R, Shenhar-Tsarfaty S, Paran Y, Zeltzer D, Shapira I, et al. Septic patients presenting with apparently normal C-reactive protein: a point of caution for the ER physician. Medicine. 2019;98:13989. https://doiorg.publicaciones.saludcastillayleon.es/10.1097/md.0000000000013989.

    Article  Google Scholar 

Download references

Acknowledgements

We would like to thank all collaborators from the Personalized Medicine in Acute Infectious Diseases (PERMIT)/Personalized Medicine in Infections: From Systems Biomedicine to Precision Diagnosis and Stratification Permitting Individualized Therapies (PERAID) consortium and the Systems Medicine to Study NSTI (INFECT) group. We would also like to thank Azenta Life Sciences' European division for their assistance with RNA sequencing, and Leonor Rib from Biotech Research and Innovation Centre (BRIC) for data pre-processing. Finally, we would like to express our gratitude to each patient who took part in the trial.

Funding

Open access funding provided by Copenhagen University The HBOmic study was supported by the PERMIT project (grant 8113-00009B), which is funded by Innovation Fund Denmark and EU Horizon 2020 under the ERA (European Research Area in Personalized Medicine) frameworks PerMed (project 2018–151) and PERAID (grant 8114-00005B) (project 90456) funded by Innovation Fund Denmark and Nordforsk. OH also received a research grant from Denmark's Ellab-Fonden. The funders had no role in the study's data collection, analysis, or interpretation. Similarly, the funders and sponsors had no involvement in the study's design, manuscript preparation, review and approval, or decision to submit the paper for publication.

Author information

Authors and Affiliations

  1. Department of Anaesthesiology, Centre of Head and Orthopedics, Copenhagen University Hospital, Rigshospitalet, Inge Lehmanns Vej 6, Copenhagen, 2100, Denmark

    Julie Vinkel & Ole Hyldegaard

  2. Department of Clinical Medicine, University of Copenhagen, Copenhagen, Denmark

    Julie Vinkel & Ole Hyldegaard

  3. Institute for Biological Psychiatry, Mental Health Centre Sct. Hans, Roskilde, Denmark

    Alfonso Buil

Authors
  1. Julie Vinkel
    View author publications

    You can also search for this author inPubMed Google Scholar

  2. Alfonso Buil
    View author publications

    You can also search for this author inPubMed Google Scholar

  3. Ole Hyldegaard
    View author publications

    You can also search for this author inPubMed Google Scholar

Contributions

OH was local PI on PERMIT and PERAID study and conceived the primary study program. JV, AB, and OH contributed to the study concept and design. JV and OH were in charge of obtaining renewed informed consent from study participants and record keeping as well as management of biological material. JV and AB contributed to data analysis. JV, AB and OH undertook the data interpretation and drafted the manuscript. All authors have read the journal's authorship agreement and policy on disclosure of potential conflicts of interest and approved the final manuscript.

Corresponding author

Correspondence to Julie Vinkel.

Ethics declarations

Ethics approval and consent to participate

The study was conducted out in accordance with the World Medical Association's Code of Ethics (Declaration of Helsinki). Biological material was gathered as part of the INFECT project [18]

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

12920_2024_2075_MOESM1_ESM.docx

Additional file 1. Supplementary methods. Description: A detailed description of the computational methods applied in the paper, including relevant references.

12920_2024_2075_MOESM2_ESM.docx

Additional file 2. Supplementary results. Description: Additional results from data normalization, differential expression analysis, principal component analysis, hierarchical clustering and gene annotations, including graphical representations and gene tables. Supplementary results. Description: Additional results from data normalization, differential expression analysis, principal component analysis, hierarchical clustering and gene annotations, including graphical representations and gene tables. Figure A1; Visualization of tagwise and common dispersion estimates from all samples. Figure A2; Plot of log2-CPM normalized counts from all samples. Figure A3; Principal component analysis of gene expression of all samples with transcript integrity scores labelled. Figure A4; Principal component analysis and scree plot and of the first three component of the principal component analysis made from all the samples. Figure A5; Principal component analysis and scree plot of the first 3 component of the principal component analysis made from all the baseline samples. Figure A6; Scree plot of the principal component analysis made from all the follow-up samples. Figure A7; Scatterplots showing correlations between coordinates in the PCA and clinical blood test results at follow-up. Figure A8; Scree plot and plot of the first three component of the principal component analysis made from all the follow up samples (MixOmics package). Table A1; Table of gene contributing to PC1 in the sparse PCA of follow-up samples. Table A2; Table of gene contributing to PC3 in the sparse PCA of follow-up samples. Table A3; Statistical results for group-wise differences shown in Figure 5b and Figure 5c. Figure A9; The optimal number of clusters for variables (genes). Figure A10; The optimal number of clusters for individuals (patients). Figure A11; Whiskers plot demonstrating the association between clinical blood tests and cluster membership at baseline. Figure A12; Whiskers plot demonstrating the association between clinical blood tests and cluster membership at follow-up. Table A4; Table with genes contained in gene cluster I of the two main clusters obtained from hierarchical clustering of samples taken after HBO2 treatment. Table A5; Table with genes contained in gene cluster II of the two main clusters obtained from hierarchical clustering of samples taken after HBO2 treatment. Table A6; Overlaps counts presented as a matrix. Figure A13; Full protein-protein interaction String network of gene cluster I obtained from hierarchical clustering of the follow-up samples. Table A7; Functional enrichments in the PPI network obtained from gene cluster I. Figure A14; MA-plot and volcano plot of differentially expressed genes (Baseline response 2). Figure A15; MA-plot and volcano plot of differentially expressed genes (Baseline response 1).

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Vinkel, J., Buil, A. & Hyldegaard, O. Blood from septic patients with necrotising soft tissue infection treated with hyperbaric oxygen reveal different gene expression patterns compared to standard treatment. BMC Med Genomics 18, 12 (2025). https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12920-024-02075-3

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12920-024-02075-3

Keywords

BMC Medical Genomics

ISSN: 1755-8794

Contact us