A Transcriptomic Approach to Understand Patient Susceptibility to Pneumonia After Abdominal Surgery

Objective: To describe immune pathways and gene networks altered following major abdominal surgery and to identify transcriptomic patterns associated with postoperative pneumonia. Background: Nosocomial infections are a major healthcare challenge, developing in over 20% of patients aged 45 or over undergoing major abdominal surgery, with postoperative pneumonia associated with an almost 5-fold increase in 30-day mortality. Methods: From a prospective consecutive cohort (n=150) undergoing major abdominal surgery, whole-blood RNA was collected preoperatively and at 3 time-points postoperatively (2–6, 24, and 48 h). Twelve patients diagnosed with postoperative pneumonia and 27 matched patients remaining infection-free were identified for analysis with RNA-sequencing. Results: Compared to preoperative sampling, 3639 genes were upregulated and 5043 downregulated at 2 to 6 hours. Pathway analysis demonstrated innate-immune activation with neutrophil degranulation and Toll-like-receptor signaling upregulation alongside adaptive-immune suppression. Cell-type deconvolution of preoperative RNA-sequencing revealed elevated S100A8/9-high neutrophils alongside reduced naïve CD4 T-cells in those later developing pneumonia. Preoperatively, a gene-signature characteristic of neutrophil degranulation was associated with postoperative pneumonia acquisition (P=0.00092). A previously reported Sepsis Response Signature (SRSq) score, reflecting neutrophil dysfunction and a more dysregulated host response, at 48 hours postoperatively, differed between patients subsequently developing pneumonia and those remaining infection-free (P=0.045). Analysis of the novel neutrophil gene-signature and SRSq scores in independent major abdominal surgery and polytrauma cohorts indicated good predictive performance in identifying patients suffering later infection. Conclusions: Major abdominal surgery acutely upregulates innate-immune pathways while simultaneously suppressing adaptive-immune pathways. This is more prominent in patients developing postoperative pneumonia. Preoperative transcriptomic signatures characteristic of neutrophil degranulation and postoperative SRSq scores may be useful predictors of subsequent pneumonia risk.

E very year in high-income countries, 1 in 10 adults undergo a noncardiac surgical procedure, 1 with a mortality rate in Europe as high as 4%. 2 In those aged 45 or over, surgical resection remains the mainstay of treatment for solid organ malignancies, accounting for the bulk of major abdominal surgical procedures performed in this age range.Postoperative nosocomial infections remain a significant cause of morbidity and mortality. 3Over 20% of surgical patients aged 45 or older suffer clinically significant infectious complications, 3,4 with a patient who develops postoperative pneumonia having an almost 5-fold increased risk of 30-day mortality compared to those without infection. 3neumonia risk is amplified following major abdominal surgery as patients often exhibit residual effects from general anesthesia, opioids, and occasionally neuromuscular blockade. 5everal strategies, including postoperative continuous positive airway pressure, have failed to reduce pneumonia incidence. 5he temporal changes in immune function following major abdominal surgery are complex 6,7 with our understanding largely extrapolated from the response to polytrauma, 8,9 major joint replacement 10, or from previous reductionist approaches, assaying alterations to single, or relatively small combinations of cytokines, correlated to clinical outcome. 11,12From these data, it is currently unclear why only some individuals develop postoperative infectious complications despite a similar intraoperative insult and baseline condition.
Cancer drives wide-ranging immunological alterations that facilitate immune tolerance, thus accelerating tumor growth and dissemination, 13 with parallels drawn with what is observed in protracted infection. 14,15The influence of damage-associated molecular patterns (DAMPs) released by tissue damage from major surgery in patients already experiencing chronic DAMPmediated stimulation due to active malignancy, is largely unexplored.Indeed, patients with active cancer are often excluded from analysis, despite this being a highly clinically relevant, vulnerable, and extensive patient cohort.
Distinct sepsis response signatures (SRS) derived from white blood cell transcriptomic profiling have been previously described, [16][17][18] associated with differences in immune function and outcome. 17,19,20It is now increasingly apparent that there are parallels between the immune pathways activated following exposure to circulating pathogen-associated molecular patterns (PAMPs) and DAMPs 21 and that it may be helpful to delineate shared and distinct biological processes between infectious and sterile inflammation. 22n this study, we used a systems-biology approach to define changes in immune pathways and gene networks activated in whole blood samples collected before and after major abdominal surgery in a well-phenotyped cohort of predominantly cancer patients.We hypothesized that similar to PAMPinduced host response endotypes identified in sepsis, [16][17][18] DAMP-mediated transcriptomic patterns and associated endotypes will identify patients with an increased susceptibility to later postoperative pneumonia.

METHODS
The BIONIC study was a prospective observational study recruiting patients aged 45 or over undergoing elective major abdominal surgery.Ethics approval was granted by the East Midlands -Nottingham 2 Research Ethics Committee (14/EM/ 1223).Inclusion/exclusion criteria and recruitment process are detailed in the Supplementary Methods (Supplemental Digital Content 1, http://links.lww.com/SLA/E786).Data were collected on each patient until hospital discharge and included comorbidities, American Society of Anesthesiology (ASA) physical status classification, indication for surgery, cancer staging and diagnosis, duration of the procedure, intensive care unit admission, and in-hospital mortality.Risk of postoperative pulmonary complications was calculated using the Assess Respiratory Risk in Surgical Patients in Catalonia (ARISCAT) score. 23All patients received standardized perioperative prophylactic antibiotic therapy and were examined daily for the presence of infection.Definitions of infection were agreed prospectively by the investigators and were based on the Centre for Disease Control and Prevention (CDC) criteria (Supplementary Table 1, Supplemental Digital Content 2, http://links.lww.com/SLA/E787. 24

Blood Sampling and Patient Selection
PAXGene (Qiagen, USA) RNA tubes were collected immediately before induction of anesthesia (preoperatively) and then 2 to 6, 24, and 48 hours postoperatively.Paired EDTA samples were drawn, and a differential white cell count was performed.
Patients with core sample sets (preoperative, 24 and 48 hours postoperatively, n = 86) were identified from the total BIONIC cohort (n = 150).Of these, 12 patients subsequently developed a postoperative pneumonia.In a pragmatic manner, taking into account RNA-sequencing cost and statistical power, 25 27 patients with similar clinical characteristics to the pneumonia cohort who remained infection-free were selected as comparators.Samples obtained at 2 to 6 hours postoperatively were available for 11 of the 39 patients (total 126 samples for FIGURE 1.A STROBE diagram outlining the patient selection process for transcriptomic analysis.Patients (n = 12), with complete sample sets, who suffered a postoperative pneumonia were identified for analysis.A cohort (n = 27), who had remained infection-free, were identified as comparators.
RNA-sequencing).The clinical characteristics considered for selection were based on key factors that could affect patient postoperative immune competence including surgery type, operation duration, smoking status, presence of diabetes, existing comorbidities, and laparoscopic surgery.The patient characteristics are outlined in Table 1 and Supplementary Table 2 (Supplemental Digital Content 3, http://links.lww.com/SLA/E788), and in the STROBE diagram (Fig. 1).

RNA-Seq Data Analysis
RNA-sequencing, alignment, determination of gene counts, and gene differential expression are described in detail in the Supplementary Methods (Supplemental Digital Content 1, http:// links.lww.com/SLA/E786).Two samples were removed during QC.All downstream analysis was performed using R v4.0.3.R packages fgsea, and MSigDB v7.4.1 including Canonical pathways (KEGG, BIOCARTA, REACTOME, PID, and Wiki-Pathways) and Gene Ontology (GO) terms were utilized for geneset enrichment analysis (GSEA) by hypergeometric test, while celltype deconvolution was performed using CIBERSORT. 26Patient clusters were defined by hierarchical agglomerative clustering based on a similarity measure (Euclidean distance) and Ward's method or k-means (Hartigan-Wong algorithm).Performance of the gene signatures was evaluated in caret (v6.0.92).Area under curve (AUC) and receiver operating characteristic (ROC) curves of the models were calculated and plotted using MLeval (v0.3).

Acute Transcriptomic Response in Patients Following Major Abdominal Surgery and Signatures of Subsequent Postoperative Pneumonia
Longitudinal RNA-seq analysis of whole blood (Fig. 2A and Supplementary Figure 1A, Supplemental Digital Content 1, http://links.lww.com/SLA/E786)from 39 patients enabled us to capture the evolution of the transcriptomic response to surgery as well as distinct signatures between patients who did, or did not, subsequently develop a postoperative pneumonia [occurring median 6 (IQR 5-9.5) days postoperatively].Principal component analysis (PCA) illustrated that preoperative samples clustered together (gray dots; Fig. 2B) and are clearly distinct from the postoperative samples (Fig. 2B).There were acute transcriptional changes comparing 2 to 6 hours following surgery to preoperative Ellipses indicate a 95% confidence interval for the groups (dotted line: infection-free; continuous line: pneumonia).C, Scatter plots showing the fold enrichment of upregulated differentially expressed genes or downregulated genes (2-6 h vs Preoperative) on the x axis (log 2 scale), and the adjusted P value on the y axis (−log 10 scale) in pathways annotated by Reactome.Hypergeometric enrichment analysis was performed to identify pathways in which DE genes were overrepresented (Supplementary Methods, Supplemental Digital Content 1, http://links.lww.com/SLA/E786).The horizontal dashed line represents the Benjamini-Hochberg-adjusted P value 0.05.samples (red vs. gray circles), with a gradual return to baseline in 24 (cyan) and 48 hours (purple circle) samples.Furthermore, clear differences were observed between 48-hour samples in those who would later be diagnosed with pneumonia versus infection-free patients (Fig. 2B and Supplementary Figure 1B, Supplemental Digital Content 1, http://links.lww.com/SLA/E786,solid vs dashed circle).Transcriptomic differences were not observed when comparing other clinical features including sex, age, duration of surgery, and smoking status (Supplementary Figure 1C, Supplemental Digital Content 1, http://links.lww.com/SLA/E786).At 2 to 6 hours postoperatively, 3639 genes were upregulated while 5043 were downregulated, compared to preoperative samples (fold change > 1.5, FDR < 0.05; Supplementary Table 3, Supplemental Digital Content 4, http://links.lww.com/SLA/E789).Pathway analysis demonstrated innate-immune activation including changes in neutrophil degranulation and Toll-like receptor (TLR) signaling (Fig. 2C) alongside adaptive immune suppression, characterized by downregulation of CD28-mediated T-cell activation and increased expression of T-cell exhaustion markers (Fig. 2C and Supplementary Table 4, Supplemental Digital Content 5, http://links.lww.com/SLA/E790).

Relative Lymphopenia and Differences in S100A8/ 9-high Neutrophils, and Hematopoietic Stem and Progenitor Cells in Patients Who Develop Postoperative Pneumonia
Next, we sought to determine temporal cell-type alterations, specifically exploring the differences between patients ultimately diagnosed with postoperative pneumonia compared to those remaining infection-free.When considering the hospital laboratory leukocyte differential counts generated concomitantly with transcriptomic sampling (Supplementary Methods, Supplemental Digital Content 1, http://links.lww.com/SLA/E786),innate-immune cells including neutrophils and monocytes increased postoperatively, alongside a reduction in lymphocytes (Supplementary Figure 2A, Supplemental Digital Content 1, http://links.lww.com/SLA/E786).No significant differences in hospital laboratory cell count were seen between patients developing pneumonia and those remaining infection-free (Supplementary Figure 2B, Supplemental Digital Content 1, http://links.lww.com/SLA/E786),although the neutrophil:lymphocyte ratio (NLR) was increased at 48 hours in those who developed pneumonia (P = 0.02) (Table 1).To detect if differences in the relative frequencies of cell populations occurred at a more granular level, cell-type deconvolution using CIBERSORTx 26 was performed on the bulk RNAseq samples (Fig. 3A; Supplementary Methods, Supplemental Digital Content 1, http://links.lww.com/SLA/E786).Preoperative samples from patients who ultimately developed a postoperative pneumonia demonstrated reduced numbers of naïve CD4 T cells and B cells, and significantly elevated numbers of mature neutrophils, hematopoietic stem and progenitor cells (HSPCs), and S100A8/9-high neutrophils (Fig. 3B; Supplementary Table 5, Supplemental Digital Content 6, http://links.lww.com/SLA/E791).These differences in naïve CD4 T cells, HSPCs, and S100A8/9-high neutrophils were also present postoperatively at 24 and 48 hours (Fig. 3C and Supplementary Table 5, Supplemental Digital Content 6, http://links.lww.com/SLA/E791).
As the incidence of cancer in this cohort was > 80%, we assessed the influence of this diagnosis on our findings.No significant association was detected between the presence of cancer and a diagnosis of postoperative pneumonia (Supplementary Figure 2C, Supplemental Digital Content 1, http://links.lww.com/SLA/E786).No differences were identified in the abundance of naïve CD4 T cells, HSPCs, or S100A8/9-high neutrophils at any sample time-point in those patients with or without a cancer diagnosis (Supplementary Figure 2D, Supplemental Digital Content 1, http://links.lww.com/SLA/E786).

Differential Gene Expression and Pathway Enrichment in Patients Developing Postoperative Pneumonia Observed Before and After Surgery
We subsequently sought to identify differences in specific genes and pathways, both pre-and postoperatively that were associated with postoperative pneumonia.The number of genes differentially expressed between pre-and postoperative samples over time reduced in patients remaining infection-free (Fig. 4A; bar plot upper panel).In contrast, the number of differentially expressed genes was sustained in patients who later developed pneumonia (Fig. 4A; bar plot middle panel).This difference is reflected in the principal component analysis (Fig. 2B).The number of genes differentially expressed between those patients that would develop pneumonia compared to those that would remain infection-free was greater preoperatively (131 genes) and at 48 hours (581 genes) than during the more acute phase (2 to 6 or 24 h) (Fig. 4A, B; Supplementary Table 6, Supplemental Digital Content 7, http://links.lww.com/SLA/E792).Gene set enrichment analysis (see Supplementary Methods, Supplemental Digital Content 1, http://links.lww.com/SLA/E786) was performed to explore whether functional gene sets coincident with the acute transcriptomic changes following major abdominal surgery were more evident in those who developed pneumonia compared to those remaining infection-free.These differentially expressed genes were enriched for immune and inflammatory pathways and GO terms (Fig. 4A, right panel and Supplementary Figure 4A, Supplemental Digital Content 1, http://links.lww.com/SLA/E786; Supplementary Table 7, Supplemental Digital Content 8, http://links.lww.com/SLA/E793).Significant neutrophil (Fig. 3A) and myeloid-mediated immunity (Supplementary Figure 3A, Supplemental Digital Content 1, http://links.lww.com/SLA/E786) enrichment was observed in pre-and postoperative samples obtained from those who subsequently developed pneumonia compared to those remaining infection-free (normalized enrichment score [NES] = 2.7, adjusted P = 5.1E-42 for neutrophil degranulation in Fig. 4A; NES = 2.6, adjusted P = 5.6E-39 for myeloid leucocyte-mediated immunity in Supplementary Figure 3A, Supplemental Digital Content 1, http://links.lww.com/SLA/E786).Importantly, this dynamic functional enrichment of gene sets persisted in all previously described contrasts when adjusted for cell proportions of neutrophils, monocytes, and lymphocytes (Supplementary Figure 3B, C, Supplemental Digital Content 1, http://links.lww.com/SLA/E786).
As immature neutrophils have recently been implicated as drivers of the poor outcome SRS1 state, 20 we specifically investigated genes involved in neutrophil degranulation and innate immunity and found that seven of these genes were differentially expressed in preoperative samples (Fig. 4B, left panel; highlighted in red).These genes were all upregulated in both preoperative and 48-hour samples from patients later developing pneumonia when compared to samples from those remaining infection-free (Fig. 4B).As expected, we found that these seven genes form a co-expression/localization network involved in neutrophil maturation and activation (Fig. 4C), with the increased expression persisting in postoperative samples from patients later developing pneumonia (Fig. 4D).

A Preoperative Signature Composed of Genes Involved in Neutrophil Degranulation Is Associated With the Development of Postoperative Pneumonia
To assess the performance of the 7 neutrophil-related gene signature we had identified, k-means clustering analysis was performed across the preoperative samples (Fig. 5A).Two k-means clusters were extracted and analyzed to detect if cluster assignment could distinguish between those who did or did not develop pneumonia.All patients in cluster 2 (5 out of 5) developed pneumonia compared to 6 out of 33 patients in cluster 1 (P = 0.00092; Fig. 5A, B).Higher expression of these seven genes was also associated with postoperative pneumonia at later time-points (24 and 48 h) but not in the acute phase (2-6 h) (Fig. 5B).Similarly, when patient clusters were defined with hierarchical clustering, 2 distinct clusters were identified (Supplementary Figure 4, Supplemental Digital Content 1, http://links.lww.com/SLA/E786).Both preoperatively and at 48 hours, higher proportions of patients in cluster 2 developed pneumonia (Fig. 5C).
Next, the performance of the 7 neutrophil-related gene signature in predicting patient outcomes was evaluated.The presence or absence of pneumonia was used as the outcome and models were constructed using the random forest method, based on the gene expression levels at each time-point (Supplementary Methods, Supplemental Digital Content 1, http://links.lww.com/SLA/E786).This gene signature achieved an AUC of 0.70, 0.71, and 0.81 preoperatively, at 24 and 48 hours, respectively (Fig. 5D), but showed poor predictive performance 2 to 6 hours postoperatively.
Validation was performed using 2 publicly available microarray datasets 8,27 (Supplementary Methods Supplemental Digital Content 1, http://links.lww.com/SLA/E786).In a cohort of critically injured polytrauma patients, 8 whole blood transcriptome was analyzed at 3 different time-points; the acute phase (0-2 h; within 2 h of traumatic injury), 24 and 72 hours after injury.Good performance was observed using our seven neutrophil-related gene signature in predicting infection with AUC of 0.76 at 72 hours, but not at the acute phase time-point (0-2 h; AUC = 0.58) (Fig. 5E).In an independent prospective cohort of elective surgery patients, which included patients who later developed a postoperative infection, 27 our 7 neutrophil-related gene signature also showed good predictive performance, up to 3 days before clinical diagnosis 27 (AUC 0.78 at day 3, 0.79 day 2 and 0.69 day 1; Fig. 5F).

Predictive Performance of a Quantitative Transcriptomic Sepsis Response Score for the Early Identification of Postoperative Pneumonia
The results of our analysis indicate potential predictive utility of preoperative whole blood transcriptomics involving neutrophil function.Given our recent work developing a quantitative transcriptomic sepsis response score 18 (SRSq; ranging from 0 to 1) that in part reflects neutrophil dysfunction and altered granulopoiesis 20 with a greater risk of adverse outcome, 18 we hypothesized that SRSq might be associated with postoperative pneumonia risk.We found that SRSq scores were increased in postoperative samples relative to preoperative samples (Fig. 6A) and that SRSq was higher at 48 hours in those patients who ultimately developed pneumonia compared to those remaining infection-free (P = 0.045, Fig. 6B).Furthermore, SRSq predictive performance at 48 hours postoperatively showed an AUC of 0.69 (Fig. 6C).It is important to note that SRSq differences were not evident preoperatively or at earlier postoperative time-points (Fig. 6B, C).Using an extended (19 gene) methodology to calculate SRS, 18 we observed more significant differences between pneumonia and infection-free patients at 48 hours (P = 0.0095, Supplementary Figure 5, Supplemental Digital Content 1, http://links.lww.com/SLA/E786).
To validate these findings, we interrogated published data from the independent cohort of elective surgery patients 27 and found that SRSq could distinguish those who later developed infection from matched patients remaining infection-free at one (P = 1.4 × 10 −5 ) and at 2 (P = 2.8 × 10 −7 ) days before the clinical diagnosis of infection (Fig. 6D).In critically injured polytrauma patient, 8 we found that SRSq scores differed significantly between patients who subsequently developed infection compared to those remaining infection-free, but only at later time-points (after 24 h) (Fig. 6E).Consistently, SRSq performed well as a predictor of infection in both cohorts (Fig. 6F, G; AUC of 0.76, 0.95, and 0.75 at days 1, 2, and 3, respectively, in the prospective cohort of elective surgery patients; AUC of 0.76 and 0.82 at 24 and 48 h, respectively, in the polytrauma patient cohort), but not at the acute phase timepoint (Fig. 6G; 0-2 h; AUC = 0.35).The differentially expressed genes are colored in orange (upregulated) and cyan (downregulated).The differentially expressed genes involved in neutrophil degranulation and/or myeloid leucocyte-mediated immunity pathway are labeled and highlighted in red.C, Network showing the interactions between seven differentially expressed genes (dashed circles; highlighted in Fig. 3B, left panel) with their related genes using GeneMANIA with a default setting.The purple edge color represents co-expression, and blue edge represents co-localization.The edge width indicates the interaction weight.The size of the nodes for the related genes indicates the interaction score assessed by GeneMANIA.The bar chart showing the enriched GO terms assessed by GeneMANIA using the network genes (D).The dynamic changes of the seven differentially expressed genes comparing pneumonia patients (in orange) to infection-free patients (in cyan) across different time-points.*P < 0.05, **P < 0.01, ***P < 0.001, n.s.= not significant.are rarely used in clinical practice. 28Of these, ARISCA, 23 which is derived from variables including age, oxygen saturation, previous respiratory infection, anemia, abdominal or thoracic surgery, duration of operation, and emergency surgery, is seen to be the best performing 28 and demonstrated an AUC of 0.64 in our cohort.When combined with SRSq, this had an AUC of 0.63 preoperatively and 0.62 at 48 hours postoperatively.ARISCAT combined with our neutrophil-related gene signature demonstrated an AUC of 0.72 preoperatively and 0.79 at 48 hours postoperatively (Supplementary Figure 6, Supplemental Digital Content 1, http://links.lww.com/SLA/E786), a comparable value to the gene model utilized alone.
Given similarities between uncontrolled DAMP-mediated stimulation (polytrauma) and a more controlled DAMP-mediated stimulation (major abdominal surgery), a 63-gene panel 29 reported to distinguish patients with a complicated outcome following polytrauma was assessed in our cohort.This demonstrated an AUC of 0.52 preoperatively and 0.68 at 48 hours postoperatively.The addition of ARISCAT had no effect on discriminatory capacity (Supplementary Figure 6, Supplemental Digital Content 1, http://links.lww.com/SLA/E786).

DISCUSSION
In this prospective observational transcriptomic analysis of patients undergoing major abdominal surgery, pathway analysis demonstrated postoperative innate-immune activation, including  subsequently developed a nosocomial pneumonia and those remaining infection-free.The expression of 7 genes, characteristic of neutrophil maturation and activation showed good predictive performance preoperatively and at 48 hours postoperatively in identifying patients who later develop pneumonia.Analysis of this novel gene set and the SRSq score in independent cohorts undergoing major surgery or following polytrauma showed good predictive performance in identifying patients who later developed nosocomial infection.The addition of a clinical score (ARISCAT) did not improve the discriminatory capacity.
A recent analysis of perioperative microarray data demonstrated a series of differing gene signatures, derived by statistical means rather than biological plausibility, that distinguish between patients who develop postoperative infection up to 3 days before clinical diagnosis. 27This study, 27 previous studies describing the utility of SRS/SRSq in infectious diseases, [16][17][18] and the gene signatures described here involving neutrophil and myeloid-mediated immunity emphasize the informativeness of transcriptomic patterns for delineating biological function and pathway activation in acute illness.They also highlight the opportunity to develop point-of-care testing for relatively small combinations of genes, over time for prognostication and targeted intervention. 27espite differences in the pathophysiology between polytrauma (shock and acute tissue damage) and major abdominal surgery (controlled tissue damage), there are key similarities in the transcriptomic patterns demonstrated with Hypoxia-Inducible Factor and TLR upregulation and T cell downregulation. 8,9,29here is also evidence that, as in polytrauma, 9,29,30 those patients who have complicated outcomes demonstrate a failure to quickly restore immune homeostasis.Further exploration of this phenomenon will allow the analysis of shared biological processes between differing forms of sterile inflammation and how this relates to later nosocomial infection susceptibility. 22xamining the content of the gene sets and SRS/SRSq implicates mature, S100A8/9 high, degranulating, IL1R2 + , immature, and PADI4 + neutrophils as the underlying drivers of SRS membership/ SRS. 20Our signature is based on differentially expressed genes characteristic of neutrophil maturation and activation, rather than an unsupervised analysis.S100A8/9-high neutrophils, identified from bulk-RNA deconvolution, are known to act as an endogenous ligand for TLR-4 31 and the receptor for advanced glycation end product (RAGE, 32 and have been negatively correlated with sepsis recovery, 20 as well as being implicated in SARS-CoV-2 induced viral pneumonitis. 33We observed a heightened abundance of these 2 cell types in both the pre-and postoperative samples of pneumonia patients, implying that the functional effects of S100A8/9-high neutrophils may predispose to postoperative pneumonia. Preoperative neutrophil degranulation is evident in those with an increased risk of nosocomial pneumonia.The conventional view of neutrophils as homogeneous, short-lived cells playing a passive role in coordinating inflammation is being revisited. 34During malignancy, deranged myelopoiesis drives the expansion of polymorphonuclear (PMN) myeloid-derived suppressor cells from the bone marrow, impairing immune surveillance. 35Elevated concentrations of these heterogeneous, poorly defined neutrophil-like populations are associated with a worse prognosis. 36,37In this context, neutrophil degranulation has been implicated in tumor progression and the facilitation of metastatic dissemination 38 and may identify patients with a more severe disease process than is apparent when assessed by conventional preoperative clinical scoring.Indeed, elevations in NLR are associated with worse survival in many solid tumors. 39ollowing lessons learned from cohort heterogenicity in the sepsis literature 40 pneumonia 24 was selected rather than all cause nosocomial infection as an enrichment strategy 41 ensuring our clinical endpoint was as robust and homogeneous as possible.We acknowledge that inter-individual variation in the perioperative immune response will play a role in susceptibility to other nosocomial infectious complications such as intra-abdominal abscess, secondary to anastomotic breakdown or surgical site infection, 42 but this signal will be diluted by the contribution of surgical complications to these types of infection.
The overall effects of both volatile and intravenous anesthesia are thought to be immunosuppressive. 6,7However, patients recruited to this study received a consistent anesthetic strategy, most commonly neuraxial analgesia (either an epidural (59%) or a single-shot intrathecal injection (8%)) with an intravenous anesthetic induction and maintenance with a volatile anesthetic (sevoflurane or desflurane).Intraoperative dexamethasone (6.6 milligrams) for prophylaxis of emesis was widely used in this cohort; however, this does not appear to influence the incidence of postoperative nosocomial infection, including pneumonia. 43More recently, there has been an interest in the influence of anesthetic choice on cancer outcome. 44Cancer-associated immunological changes are characterized by T-cell exhaustion, the presence of circulating myeloid-derived suppressor and T reg cells, and are thereby predominantly immunosuppressive. 14,15These effects of cancer on immune function may have influenced the differential transcriptomic patterns described in this study.
This study has further limitations.Sampling only blood may miss compartmentalized responses that can occur in sterile insults. 45Peripheral myeloid expansion occurs with malignancy 46 and following physiological stress 34 meaning that comparisons between quiescent preoperative samples and postoperative samples should be interpreted with caution, despite adjusting for paired differential leukocyte cell counts.
This study demonstrates the potential for SRSq scores postoperatively and transcriptomic signatures characteristic of neutrophil degranulation pre-and postoperative to not only act as useful predictors for the subsequent risk of pneumonia but also to offer value in understanding the underlying pathobiology of postoperative pneumonia development.However, as this study was intended to be hypothesis generating, findings, specifically these preoperative changes described, require replication in a larger prospective cohort.Additional work specifically focusing on neutrophil heterogeneity using microscopy and single-cell analysis techniques is warranted, while the identification of the shared pathways between the immune response to malignancy and susceptibility to infection will be beneficial in this high-risk perioperative cohort.

FIGURE 2 .
FIGURE 2. Sustained acute response in major abdominal surgery patients is associated with later postoperative pneumonia diagnosis.A, Schematic diagram of a longitudinal RNA-seq analysis of whole blood isolated from patients undergoing major abdominal surgery.B, Principal component analysis of gene expression in samples pre-and postoperatively in 2 to 6, 24, and 48 hours.Each dot represents a sample, labeled with patient ID.Colors indicate different time-points.Shapes indicate the clinical status of patients after operation (circle: infection-free; circle plus: pneumonia).Ellipses indicate a 95% confidence interval for the groups (dotted line: infection-free; continuous line: pneumonia).C, Scatter plots showing the fold enrichment of upregulated differentially expressed genes or downregulated genes (2-6 h vs Preoperative) on the x axis (log 2 scale), and the adjusted P value on the y axis (−log 10 scale) in pathways annotated by Reactome.Hypergeometric enrichment analysis was performed to identify pathways in which DE genes were overrepresented (Supplementary Methods, Supplemental Digital Content 1, http://links.lww.com/SLA/E786).The horizontal dashed line represents the Benjamini-Hochberg-adjusted P value 0.05.

FIGURE 3 .
FIGURE 3. Specific cell populations and pathways showing differential abundance and enrichment before and after abdominal surgery, and relationship with subsequent development of pneumonia.A, Heatmap of CIBERSORTx absolute scores of 23 different cell types in 126 bulk RNA-seq samples.The absolute proportion of each cell type (CIBERSORTx absolute scores) was obtained from deconvolution using a single-cell RNA-seq reference panel (Supplementary Methods, Supplemental Digital Content 1, http:// links.lww.com/SLA/E786).B, Box plots outline the proportions of 23 different cell types in preoperative samples from patients who did (orange) or did not develop pneumonia (cyan).P value was calculated by 2-tailed Wilcoxon signed-rank test.P < 0.05, **P < 0.01, ***P < 0.001, NS = not significant.C, Box plots of CIBERSORTx absolute scores of indicated cell types in samples from patients who did (orange) or did not develop pneumonia (cyan) across different time-points after surgery.

FIGURE 4 .
FIGURE 4. Neutrophilic inflammation may predispose patients to postoperative pneumonia.A, Bar plots outlining the number of differentially expressed genes in each contrast as indicated on the y axis (left panel), and a heatmap (right panel) displaying the enrichment of MSigDB canonical pathways using fgsea (Supplementary Methods, Supplemental Digital Content 1, http://links.lww.com/SLA/E786).The top 5 pathways in each contrast were selected for visualization.B, Volcano plot showing differentially expressed genes in preoperative samples (left panel) and 48-hour samples (right panel).Log 10 adjusted P values are plotted against the log 2 fold change of the gene expression (pneumonia vs infection-free).The horizontal dashed lines represent the FDR-adjusted P value of 0.05.The differentially expressed genes are colored in orange (upregulated) and cyan (downregulated).The differentially expressed genes involved in neutrophil degranulation and/or myeloid leucocyte-mediated immunity pathway are labeled and highlighted in red.C, Network showing the interactions between seven differentially expressed genes (dashed circles; highlighted in Fig. 3B, left panel) with their related genes using GeneMANIA with a default setting.The purple edge color represents co-expression, and blue edge represents co-localization.The edge width indicates the interaction weight.The size of the nodes for the related genes indicates the interaction score assessed by GeneMANIA.The bar chart showing the enriched GO terms assessed by GeneMANIA using the network genes (D).The dynamic changes of the seven differentially expressed genes comparing pneumonia patients (in orange) to infection-free patients (in cyan) across different time-points.*P < 0.05, **P < 0.01, ***P < 0.001, n.s.= not significant.

FIGURE 5 .
FIGURE 5. Predictive performance of a signature composed of genes involved in neutrophil degranulation for detection of postoperative infections.A, Cluster plot showing K-means analysis of preoperative samples with the proportion of variance explained by each component shown.Two major clusters were highlighted in orange and cyan.The patient IDs together with the clinical status (1: pneumonia; 0: infection-free) were shown next to each dots.Bar plots of the number of K-means (B) or hierarchical (C) cluster1 and cluster2 patients who did (in orange) or did not develop pneumonia (in cyan) across different time-points.P values were calculated by 2-tailed Fisher's exact test.*P < 0.05, **P < 0.01, ***P < 0.001, NS = not significant.ROC curves of prediction models constructed using the seven neutrophil-related gene expressions in the BIONIC cohort (D); a cohort with critically injured patients (E)8 and a cohort with patients who develop postoperative infection up to 3 days before clinical diagnosis (F).27

FIGURE 6 .
FIGURE 6. Predictive performance of the transcriptomic sepsis response score for detection of postoperative infection.Box plots of the SRSq scores for samples in each time-point (A) or for samples from patients with (orange) or without (cyan) pneumonia (B).SRSq was quantified using a predefined and previously validated seven gene set through SepstratifieR package. 18P value was determined using a linear model, ***P < 0.001.C, Receiver operating characteristic (ROC) curves of prediction models constructed using SRSq scores for samples in the BIONIC cohort.AUC: Area under the (ROC) Curve.D, Bar plots showing the SRSq scores in a cohort with patients who develop postoperative infection up to 3 days before clinical diagnosis assayed at days 1 (n = 46), 2 (n = 32), and 3 (n = 20). 27E, Bar plots showing the SRSq scores in a cohort with critically injured patients 8 assayed at 0 to 2 (n = 26), 24 (n = 26), and