Abstract

Introduction

   Sepsis is the leading cause of death in critically ill patients; it
   results in multiorgan dysfunction, including acute respiratory distress
   syndrome (ARDS). Our study was conducted to determine the role of
   cellular senescence genes and immune infiltration in sepsis and
   sepsis-induced ARDS via bioinformatic analyses.

Experimental Procedures

   Datasets [28]GSE66890 and [29]GSE145227 were obtained from the Gene
   Expression Omnibus (GEO) database and utilized for bioinformatics
   analyses. Gene Ontology (GO) terms and Kyoto Encyclopedia of Genes and
   Genomes (KEGG) enrichment analyses of the differentially expressed
   genes (DEGs) were performed to identify the key functional modules. Two
   machine learning algorithms, least absolute shrinkage and selection
   operator (LASSO) and support vector machine recursive feature
   elimination (SVM-RFE), were used to screen for characteristic genes in
   sepsis and sepsis-induced ARDS. ROC curves were generated to evaluate
   the predictive ability of gene hubs. Differences in immune infiltration
   levels between the disease and control groups were compared via ssGSEA.
   The diagnostic value of the hub genes was verified via quantitative PCR
   (qPCR) in hospitalized patients.

Results

   Four characteristic genes (ATM, CCNB1, CCNA1, and E2F2) were identified
   as biomarkers for the progression of sepsis-induced ARDS. E2F2 showed
   the highest predictive ability for the occurrence of ARDS in patients
   with sepsis. CD56bright and plasmacytoid dendritic cells exhibited high
   infiltration in the sepsis-induced ARDS group, whereas eosinophils,
   MDSCs, macrophages, and neutrophils exhibited low infiltration. In
   addition, ATM expression was lower in patients with sepsis than in
   those without sepsis (n = 6).

Conclusion

   Sepsis-induced ARDS is correlated with circulating immune responses,
   and the expression of ATM, CCNB1, CCNA1, and E2F2 may serve as
   potential diagnostic biomarkers and therapeutic targets in
   sepsis-induced ARDS.

   Keywords: sepsis, cellular senescence genes, immune infiltration, acute
   respiratory distress syndrome, bioinformatics analysis

Introduction

   Sepsis is a life-threatening condition that results in organ
   dysfunction caused by a dysregulated host response to infection. Sepsis
   and septic shock are leading causes of death worldwide; they are fatal
   in approximately 25% of diagnosed patients.[30]^1 Patients with acute
   respiratory distress syndrome (ARDS) are at high risk of death in
   intensive care units (ICUs).[31]^2^,[32]^3 Among the factors that
   influence the development of ARDS, sepsis is a major contributor. ARDS
   caused by sepsis is more severe than ARDS due to nonsepsis causes.
   Specifically, recovery from lung injury is worse, extubation success
   rates are lower, and mortality is greater.[33]^4 Sepsis-related ARDS
   has been demonstrated to be more severe than nonsepsis-related ARDS, as
   sepsis-induced ARDS causes more severe lung injuries, and its mortality
   is greater.[34]^5 Thus, key molecules involved in sepsis-related ARDS
   must be identified to treat sepsis more effectively.

   Numerous studies have been conducted in recent years to determine
   whether genetic factors contribute to ARDS development.[35]^6^,[36]^7
   Clinical and biological markers are useful for predicting ARDS and its
   outcomes in patients with sepsis. Demand for biomarker-targeted
   therapies for ARDS is currently increasing,[37]^8^,[38]^9 and previous
   studies have shown that specific markers may play important roles in
   detecting and preventing sepsis-induced ARDS. Specifically, cyclin B1
   (CCNB1), cyclin B2 (CCNB2), DNA topoisomerase II alpha (TOP2A) and the
   transcription factor fork head box protein M1 (FOXM1) are potential
   targets for gene therapy.[39]^6 In this context, the analysis of gene
   expression profiles in relation to ARDS may yield breakthroughs to
   elucidate the pathological mechanisms underlying ARDS. Thus, cellular
   senescence may be involved in the development and onset of ARDS.
   Notably, inflammation, proliferation, and fibrosis may be causes or
   consequences of senescence. However, studies establishing a link
   between ARDS and cellular senescence are limited. Naturally occurring
   senescence in the early stages of acute injury in other tissues has
   been shown to be beneficial by limiting apoptosis and favouring tissue
   repair.[40]^10

   In this study, we comprehensively evaluated the expression profiles of
   senescence genes in sepsis and sepsis-induced ARDS. Using the GEO
   database, the mRNA expression profiles of patients with sepsis and
   sepsis-related ARDS were analysed via differential expression, immune
   infiltration, gene expression, immune infiltration, and functional
   enrichment analyses. To effectively diagnose sepsis or sepsis-induced
   ARDS, we used two machine learning algorithms to predict outcomes. This
   study aimed to provide useful biological information for identifying
   the key cellular senescence genes involved in sepsis-induced ARDS.

Experimental Procedures

Data Source

   RNA-seq profiles and clinical features of the [41]GSE66890 and
   [42]GSE145227 datasets were accessed from the GEO website
   ([43]https://www.ncbi.nlm.nih.gov/gds/?term=). [44]GSE66890 includes 29
   sepsis-induced ARDS samples and 28 sepsis samples collected on the
   platform [45]GPL6244. [46]GSE145227 includes 10 sepsis samples and 12
   normal samples collected on the platform [47]GPL23178. Senescence genes
   were obtained from the “cellular senescence” terms in the KEGG database
   ([48]https://www.kegg.jp/entry/map04218).

Identification of DEGs

   The R (version 4.1.1) package “preprocessCore” was used to homogenize
   the data prior to differential expression analysis. After
   homogenization, DEGs were screened with a threshold of p < 0.05 using
   the “limma” package. Genes that were significantly upregulated or
   downregulated in both GEO datasets were identified as DEGs.

Functional Enrichment Analysis

   The Gene Ontology (GO) resource ([49]http://geneontology.org) provides
   structured, computable knowledge regarding the functions of genes and
   gene products.[50]^11 Gene Ontology function enrichment includes three
   categories: molecular function (MF), biological process (BP) and
   cellular component (CC). The Kyoto Encyclopedia of Genes and Genomes
   (KEGG, [51]http://www.genome.ad.jp/kegg/)[52]^12 is a database of
   biological systems that collects genomic, chemical and systemic
   functional information. Gene Ontology (GO) and Kyoto Encyclopedia of
   Genes and Genomes (KEGG) pathway enrichment analyses were performed and
   visualized via the R package “clusterProfiler.”.[53]^13 Statistical
   significance was set at p < 0.05. Correlation analysis of the hub genes
   with all transcripts was conducted for gene set enrichment analysis
   (GSEA) of the hub genes. The R package “clusterProfiler” was used to
   perform GSEA via correlation analysis to explore the functions of the
   hub genes.

Machine Learning Analysis

   To effectively diagnose sepsis or sepsis-induced ARDS, we used two
   machine learning algorithms to predict outcomes. The LASSO logistic
   regression[54]^14 and SVM-RFE[55]^15 algorithms are highly important
   for identifying key biomarkers. In recent years, these two algorithms
   have been widely used in research to identify diagnostic or prognostic
   factors. The LASSO regression algorithm was applied via the R package
   “glmnet” to determine genes significantly related to the screening
   ability of the disease group and the control group. A support vector
   machine (SVM) is a monitoring technology used for classification and
   regression analysis. To avoid overfitting, the RFE algorithm was used
   to screen hub genes from the metadata. Thus, we used SVM-RFE technology
   to select hub genes. The ROC curve was used to evaluate the predictive
   ability of the hub genes via the R package “pROC.”

Immune Infiltration Assessment

   Single Sample Gene Set Enrichment Analysis (ssGSEA), an extension of
   the Gene Set Enrichment Analysis (GSEA) methodology, is extensively
   utilized in bioinformatics research, particularly in the context of
   immune infiltration studies.[56]^16 The ssGSEA algorithm within the R
   package “GSVA” was used to evaluate the infiltration levels of the 23
   immune cells. The differences in immune infiltration levels between the
   disease and control groups were compared. The Spearman rank correlation
   coefficient was used to evaluate the correlation between the
   infiltration scores of immune cells and the expression levels of the
   hub genes.

qRT‒PCR

   This study was approved by the Medical Ethics Committee of the Central
   Hospital of Wuhan, Tongji Medical College, Huazhong University of
   Science and Technology (WHZXKYL2022-007). Whole blood samples from six
   patients with sepsis and six patients without sepsis were collected
   from the intensive care unit of the Central Hospital of Wuhan, Tongji
   Medical College, Huazhong University of Science and Technology, from
   October 20, 2023 to January 30, 2024. Sepsis was diagnosed on the basis
   of Sepsis version 3.0. Patients with neoplastic diseases or autoimmune
   disorders, as well as those taking oral immunosuppressants, were
   excluded from the study. Quantitative real-time polymerase chain
   reaction (qPCR) was performed to measure the expression levels of four
   characteristic genes (ATM, CCNB1, CCNA1, and E2F2).

Statistical Analysis

   Student’s t test was used to compare the differences between two
   groups. Statistical analysis was performed via R software 4.1.1.
   Differences were considered statistically significant at *p < 0.05, **p
   < 0.01, ***p < 0.001, and ****p < 0.0001.

Results

Identification of DEGs

   We first homogenized the mRNA expression data from the [57]GSE66890 and
   [58]GSE145227 datasets before performing differential expression
   analysis ([59]Sup-Figure 1A-[60]D). We identified 8655 upregulated DEGs
   and 11530 downregulated DEGs in [61]GSE145227 ([62]Figure 1A–B) and
   1092 upregulated DEGs and 958 downregulated DEGs in [63]GSE66890
   ([64]Figure 1C–D). DEGs that were simultaneously upregulated or
   downregulated in both [65]GSE66890 and [66]GSE145227 were selected as
   differentially expressed genes for subsequent analysis, which yielded
   288 upregulated ([67]Figure 1E) and 167 downregulated genes ([68]Figure
   1F).

Figure 1.

   [69]Figure 1
   [70]Open in a new tab

   Identification of DEGs. A total of 8655 upregulated DEGs and 11530
   downregulated DEGs were identified in [71]GSE145227 (A-B), whereas 1092
   upregulated DEGs and 958 downregulated DEGs were identified in
   [72]GSE66890 (C-D), and 288 upregulated genes (E) and 167 downregulated
   genes (F) were identified in [73]GSE66890 and [74]GSE145227.

Functional Enrichment Analysis

   The functional enrichment of the DEGs was performed to identify
   possible biological functions. GO analysis revealed that the DEGs were
   related to neutrophil-mediated immunity in the biological process (BP),
   secretory granule lumen in the cellular component (CC), and kinase
   regulatory activity in the molecular function (MF) ([75]Figure 2A).
   KEGG analysis revealed that genes related to Salmonella infection, the
   cell cycle, cellular senescence, and carbon metabolism were enriched
   ([76]Figure 2B). These results suggest that cellular senescence-related
   functions and pathways play vital roles in sepsis-induced ARDS.

Figure 2.

   [77]Figure 2
   [78]Open in a new tab

   Functional enrichment of DEGs. GO analysis (A); KEGG analysis (B).

Identification of Cellular Senescence Genes Among the DEGs

   Next, we intersected the DEGs with genes related to cellular
   senescence. Eleven cellular senescence genes were identified
   ([79]Figure 3A). We further explored the functions and pathways of
   these genes. GO analysis revealed that these 11 genes were involved
   primarily in the regulation of protein serine/threonine kinase activity
   (BP and MF) and the cyclin-dependent protein kinase holoenzyme complex
   (CC) ([80]Figure 3B). KEGG analysis revealed that the cellular
   senescence, cell cycle, and FoxO signalling pathways were enriched
   ([81]Figure 3C–D).

Figure 3.

   [82]Figure 3
   [83]Open in a new tab

   Identification of DEGs involved in cellular senescence. Eleven genes
   were identified (A). The enriched Gene Ontology (GO) terms in the
   molecular function (MF), biological process (BP) and cellular component
   (CC) categories (B). The results of the KEGG pathway enrichment
   analysis (C); the top 5 KEGG results and their relationships with the
   identified genes (D).

   Volcano plots and heatmaps were generated to depict the differences in
   the expression of cell senescence genes between the [84]GSE66890
   ([85]Figure 4A–B) and [86]GSE145227 datasets ([87]Figure 4C–D). The
   differential expression of the 11 intersecting genes is displayed via
   data from the [88]GSE66890 ([89]Figure 5A) and [90]GSE145227 datasets
   ([91]Figure 5B). Eight genes (CCNA1, CCNB1, CDK1, CAPN1, E2F2, MAP2K2,
   MAP2K3, and CCND3) were upregulated in the sepsis-induced ARDS group,
   whereas three genes (SIRT1, ATM, and RBL2) were downregulated.

Figure 4.

   [92]Figure 4
   [93]Open in a new tab

   Volcano map and heatmap of cell senescence genes in the [94]GSE66890
   (A-B) and [95]GSE145227 (C-D) datasets.

Figure 5.

   [96]Figure 5
   [97]Open in a new tab

   Differential expression of 11 cellular senescence genes in [98]GSE66890
   (A) and [99]GSE145227 (B). *p < 0.05, **p < 0.01, ***p < 0.001, and
   ****p < 0.0001.

Identification of Hub Genes

   Two machine learning algorithms were used to select biomarker genes.
   Using the LASSO regression algorithm, we screened for four hub genes as
   diagnostic markers ([100]Figure 6A). Using the SVM-RFE algorithm, we
   identified 11 hub genes ([101]Figure 6B). Finally, we selected four
   overlapping features (ATM, CCNB1, CCNA1, and E2F2) between the two
   algorithms ([102]Figure 6C). These four genes may be involved in the
   progression of sepsis-induced ARDS.

Figure 6.

   [103]Figure 6
   [104]Open in a new tab

   Machine learning algorithms. LASSO (A): SVM-RFE (B): overlapping
   features between these two algorithms (C).

Predictive Ability of the Hub Genes

   [105]Figure 7A shows the log2-fold change values of the four hub genes
   in both datasets. The correlations between the hub genes based on the
   [106]GSE66890 dataset are shown in [107]Figure 7B. Notably, CCNA1 is
   positively correlated with CCNB1 and E2F2 and negatively correlated
   with ATM. The AUC value of the ROC curve indicates that E2F2 has the
   highest predictive ability (AUC = 0.749) for the occurrence of ARDS in
   patients with sepsis ([108]Figure 7C).

Figure 7.

   [109]Figure 7
   [110]Open in a new tab

   Predictive ability of the four hub genes. The log2(FC) values of 4
   genes in [111]GSE66890 and [112]GSE145227 (A). Correlations between the
   hub genes in [113]GSE66890 (B). The AUC value of the ROC curve for
   predicting the occurrence of ARDS in sepsis patients (C).

Immune Infiltration Analysis

   Because the immune system plays a crucial role in sepsis-induced ARDS,
   we assessed the infiltration levels of 23 immune cells in each patient.
   [114]Figure 8A shows the relationship between each type of immune cell.
   We found that CD56bright structural killer cells and plasmacytoid
   dendritic cells were highly infiltrated in the sepsis-induced ARDS
   group, whereas eosinophils, MDSCs, macrophages, and neutrophils were
   poorly infiltrated ([115]Figure 8B). Immune cell infiltration was
   correlated with the four hub genes ([116]Figure 8C).

Figure 8.

   [117]Figure 8
   [118]Open in a new tab

   Immune infiltration analysis. Relationships between each immune cell
   type (A). Infiltrating immune cells in the sepsis group and
   sepsis-induced ARDS group (B). Correlations of the 4 hub genes with
   immune cell types (C). *p < 0.05, **p < 0.01 and ***p < 0.001.

Exploration of the Hub Genes

   Correlation analysis between the hub genes and all transcripts was
   performed based on data from [119]GSE66890. The expression levels of
   the 50 genes most positively correlated with each hub gene are shown in
   [120]Figure 9. GSEA was conducted on the basis of the correlation
   results. The 20 terms most enriched for the 50 genes correlated with
   each of the four hub genes are shown in [121]Figure 10. For example,
   genes positively correlated with CCNA1 were enriched for functions
   involved in membrane trafficking, neutrophil degranulation, and
   vesicle-mediated transport terms and depleted for functions related to
   signal transduction and postmitotic nuclear pore complex (NPC)
   reformation. These results may facilitate in-depth investigations of
   the involvement of hub genes in sepsis-induced ARDS.

Figure 9.

   [122]Figure 9
   [123]Open in a new tab

   The expression of the top 50 genes positively correlated with each hub
   gene.

Figure 10.

   [124]Figure 10
   [125]Open in a new tab

   The top 20 terms of the 4 hub genes identified via GSEA.

Preliminary Validation Analysis of Hub Genes

   We performed RT‒PCR analysis of peripheral blood mononuclear cells
   (PBMCs) from 6 sepsis patients and 6 nonsepsis patients to verify the
   expression of the four hub genes. The results revealed that only ATM
   expression significantly differed between the two groups ([126]Figure
   11), which was consistent with the results of the bioinformatics
   analysis. Further exploration of ATM genes in sepsis is therefore
   important.

Figure 11.

   [127]Figure 11
   [128]Open in a new tab

   Expression of the ATM gene in sepsis patients and nonsepsis patients.
   *p < 0.05, P = 0.0327.

Discussion

   Owing to the lack of effective risk prediction, accurate genetic
   diagnosis, and molecular interventions, sepsis-induced acute
   respiratory distress syndrome adds to the global impact of sepsis, a
   life-threatening condition. The prognosis of patients with
   sepsis-induced ARDS is poor due to limited effective prevention and
   treatments.[129]^17 Because appropriate treatment for sepsis-induced
   ARDS relies heavily on early prediction, identification, and diagnosis,
   bioinformatic analysis can provide a useful tool for exploring the
   underlying mechanisms of sepsis-induced ARDS via large-scale gene
   expression data.[130]^18 To investigate the transcriptomic consequences
   of sepsis-induced ARDS, we conducted a bioinformatics analysis of human
   sepsis-related ARDS samples and sepsis samples. In total, 288 DEGs were
   upregulated, and 167 DEGs were downregulated. The differentially
   expressed transcripts were primarily involved in the cell cycle and
   cellular senescence, suggesting that cellular senescence-related
   functions and pathways play a vital role in sepsis-induced ARDS.

   Pulmonary diseases may benefit from cellular senescence or may have
   detrimental effects. The senescence process can limit DNA damage during
   acute injury, preserve senescent cells, and prevent apoptosis.
   Following the removal of senescent cells, proliferation occurs to
   regenerate the injured tissue. Senescent cells may prevent recovery
   from lung injury if cell turnover slows because they decrease the
   proliferation rate and prolong proinflammatory states. Abundant
   evidence supports that senescence plays a role in chronic and
   age-related lung diseases, but less evidence is available for
   ALI.[131]^10 In recent years, animal and cell experiments have
   demonstrated that cellular senescence plays an important role in
   sepsis-induced lung injury. In a two-hit mouse model generated via
   caecal ligation and puncture (CLP) followed by sublethal Pseudomonas
   aeruginosa (PA) lung infection, a relatively high mortality rate was
   observed because of increased lung inflammation and cellular
   senescence.[132]^19 LPS can induce cellular senescence in alveolar
   epithelial cells, suggesting that LPS-induced cellular senescence plays
   a critical role in limiting tissue repair when sepsis occurs.[133]^20
   Sepsis-induced ARDS has a multifactorial pathogenesis, among which
   cellular senescence is believed to play a crucial role.

   In the present study, 11 cellular senescence genes were identified. The
   differential expression of these 11 genes was investigated via data
   from the [134]GSE66890 and [135]GSE145227 datasets. Eight genes,
   namely, CCNA1, CCNB1, CDK1, CAPN1, E2F2, MAP2K2, MAP2K3, and CCND3,
   were upregulated in the sepsis-induced ARDS group, whereas three genes,
   SIRT1, ATM, and RBL2, were downregulated. KEGG analysis revealed that
   the functions and pathways of these genes were enriched in the cellular
   senescence, cell cycle, and FoxO signalling pathways. These findings
   increase our understanding of sepsis-induced acute respiratory distress
   syndrome. To effectively diagnose sepsis or sepsis-induced ARDS, we
   used the LASSO regression algorithm and SVM-RFE algorithm to predict
   disease outcomes and identified four biomarkers (ATM, CCNB1, CCNA1, and
   E2F2) across both algorithms. These four genes may be involved in the
   progression of sepsis-induced ARDS. The AUC value of the ROC curve
   indicated that E2F2 had the highest predictive ability (AUC 0.749) for
   the occurrence of ARDS in patients with sepsis.

   Targeted E2F2 knockdown causes disproportionate entry into the cell
   cycle upon T-cell stimulation in mice. This study suggests that E2F2
   plays a critical role in regulating apoptosis and may help in the
   design of new approaches for treating immune-mediated diseases. Both in
   vitro and in vivo, E2F2 provides genomic stability to activated T
   lymphocytes and plays a specific antiapoptotic role.[136]^21 In
   LPS-induced acute lung injury, the processes of resolving lung injury
   and regulatory T (Treg) cells appear to be important mediators.[137]^22
   Although E2F2 has not been studied previously, we speculate that E2F2
   may be involved in the progression of ARDS by regulating T cells or
   antiapoptotic properties.

   Both CCNA1 and CCNB1 are cell cycle regulatory proteins that are
   involved in mitosis. A study reported that errors in mitosis can either
   kill the cell through apoptosis or cause potentially pathogenic
   mutations.[138]^23 GO functional analysis revealed that DEGs associated
   with the mitotic cell cycle and cellular senescence, such as CCNA1 and
   CCNB2, were the most significantly enriched biological processes among
   the upregulated genes. The abnormal cytoplasmic expression of cyclin B1
   and T-cell cells is associated with immune responses specific to cyclin
   B1.[139]^24 A growing body of evidence suggests that the immune system
   plays a key role in recovery from lung diseases and acute lung
   injury.[140]^25 ARDS is driven by the activation of innate immunity via
   the binding of cell injury-associated endogenous molecules to pattern
   recognition receptors.[141]^26 However, neither CCNA1 nor CCNB2 are
   directly implicated in ARDS; therefore, we speculated that CCNA1 and
   CCNB1 might be closely related to ARDS development.

   The ataxia-telangiectasia mutated (ATM) gene causes a rare autosomal
   recessive disorder that results in a defective repair of double-strand
   breaks in DNA.[142]^27 Ataxia-telangiectasia (A-T) is an autosomal
   recessive, progressive, multisystem disease caused by mutations in the
   ATM gene (11q22.3). This gene, which encodes ATM kinase, a kinase
   involved in cellular stress response signalling, is widely expressed.
   More than one hundred proteins related to the DNA damage response, cell
   cycle regulation, and other pathways are correlated with ATM
   expression. ATM plays several important roles in neuroprotection,
   innate immunity, inflammation, metabolism (eg, insulin signalling),
   longevity, and fertility.[143]^28 Several factors can contribute to the
   development of ataxia-telangiectasia (A-T), including recurrent sinus
   infections, immunodeficiency, and aspiration due to impaired swallowing
   caused by neurodegenerative progression.[144]^27^,[145]^29 The immune
   deficiency in A-T is highly variable and affects both B and T cells.
   Therefore, we speculated that ATM downregulation might be related to
   sepsis-induced ARDS. In addition, ATM gene expression was significantly
   decreased in patients with sepsis (p = 0.0327).

   Overall, our study revealed that the four identified cellular
   senescence genes exhibited marked differences and significant
   correlations with each other and with immune cell infiltration during
   the development of sepsis-induced ARDS in human transcriptomic studies.
   Recent studies have shown that LPS-induced ALI activates alveolar
   macrophages rather than causing systemic activation and that aged mice
   exhibit impaired adaptive immunity.[146]^30 Ageing and immune
   regulation are also inextricably linked. Immune regulation plays a
   crucial role in sepsis-induced ARDS.[147]^31^,[148]^32 Multiple immune
   cells, including neutrophils and monocytes, are considered closely
   involved.[149]^33 Therefore, to explore the underlying immune
   mechanisms, we performed GSEA and immune infiltration assays. According
   to the GSEA results, the high-risk groups presented dysregulated
   expression of genes that were significantly enriched in neutrophil
   degranulation and innate immune pathways. We found that CD56bright
   structural killer cells and plasmacytoid dendritic cells were highly
   infiltrated in the sepsis-induced ARDS group, whereas eosinophils,
   MDSCs, macrophages, and neutrophils were poorly infiltrated.

   To further explore the pathogenesis of ARDS, correlation analysis
   between the four hub genes and all transcripts was performed on the
   [150]GSE66890 dataset. These analyses revealed that CCNA1 was
   positively correlated with membrane trafficking, neutrophil
   degranulation, and vesicle-mediated transport terms and negatively
   associated with signal transduction and postmitotic nuclear pore
   complex (NPC) reformation terms. These results may facilitate in-depth
   investigations of the hub genes involved in sepsis-induced ARDS.

   Although this study produced some positive results, it also has several
   limitations. First, our findings were not verified in cells or tissues;
   we did not combine expression profiles to confirm our findings, and the
   sample size was limited. Furthermore, mRNA expression profiles were
   obtained from blood samples in this study and not from other important
   cells in ARDS, such as the epithelial and endothelial cells of the
   lungs. Hence, further studies involving other ARDS cell types with
   experimental verification and diverse samples are necessary.

Conclusions

   In conclusion, the results of the present study revealed that ATM,
   CCNB1, CCNA1, and E2F2 can predict the progression of ARDS in sepsis
   patients. This study provides a comprehensive understanding of the
   genes involved in sepsis-induced ARDS pathogenesis.

Acknowledgments