Measuring the human immune response to surgery: multiomics for the prediction of postoperative outcomes

Abstract

Purpose of review:

Postoperative complications including infections, cognitive impairment, and protracted recovery occur in one-third of the 300 million surgeries performed annually worldwide. Complications cause personal suffering along with a significant economic burden on our healthcare system. However, the accurate prediction of postoperative complications and patient-targeted interventions for their prevention remain as major clinical challenges.

Recent findings:

While multifactorial in origin, the dysregulation of immunological mechanisms that occur in response to surgical trauma is a key determinant of postoperative complications. Prior research, primarily focusing on inflammatory plasma markers, has provided important clues regarding their pathogenesis. However, the recent advent of high-content, single-cell transcriptomic and proteomic technologies has considerably improved our ability to characterize the immune response to surgery, thereby providing new means to understand the immunological basis of postoperative complications and to identify prognostic biological signatures.

Summary

The comprehensive and single-cell characterization of the human immune response to surgery has significantly advanced our ability to predict the risk of postoperative complications. Multiomic modeling of patients’ immune states holds promise for the discovery of preoperative predictive biomarkers, ultimately providing patients and surgeons with actionable information to improve surgical outcomes. While recent studies have generated a wealth of knowledge, laying the foundation for a single-cell atlas of the human immune response to surgery, larger-scale multiomic studies are required to derive robust, scalable, and sufficiently powerful models to accurately predict the risk of postoperative complications in individual patients.

Introduction

More than 300 million surgeries are performed annually worldwide [1]. Postoperative complications, such as infections, cognitive impairment, or prolonged recovery occur in approximately 30% of surgeries [2–4], resulting in personal suffering, longer hospital stays, mortality, and significant socioeconomic burden [5,6]. Sufficiently powerful paradigms predicting the risk of postoperative complications would allow for patient-tailored preoperative clinical interventions [7,8], optimization of surgery timing, and selection of the most appropriate surgical approach to mitigate such risk [9,10]. However, current metrics and tools for risk estimation, such as the American Society of Anesthesiologists (ASA) classification, the American College of Surgeons (ACS) NSQIP, or the POSSUM scoring system, show poor to moderate performance as indicated by an area under the receiver operating curve of less than 0.75 [11]. Importantly, these metrics on clinical, demographic, or routine laboratory data do not objectively capture biological factors that drive surgical recovery. In particular, key characteristics of a patient’s immune system, which is at the intersection of most, if not all, biological mechanisms that drive surgical recovery [12], are not included in current prediction metrics and algorithms.

Surgery is associated with significant tissue trauma, triggering a complex inflammatory response that engages the innate and adaptive branches of the immune system [13,14]. Complications, including end-organ damage, infections, and protracted recovery, arise as pro-inflammatory and immunosuppressive responses tilt out of balance [15]. While conventional biological research has primarily focused on the analyses of soluble plasma factors, the development of new technologies that provide high-content data, such as proteomics (proteins) [16], transcriptomics (RNA-sequencing) [17], metabolomics (metabolites) [18], and cytomics (single-cell transcriptomic or proteomic analyses) [19], has created new and powerful avenues for the system-level assessment of the human immune response to surgery and the identification of biological events that drive, and are predictive of, surgical complications.

Here, we will summarize the current understanding of perioperative immunology. We will first discuss key innate and adaptive immune mechanisms implicated in the immune response to surgical trauma and summarize prior evidence of immune dysregulation driving complications. Finally, we will highlight recent technological advances that offer a promising approach for biomarker discovery and the development of accurate algorithms for predicting and, ultimately, preventing postoperative complications.

The immune response to surgical injury, a finely tuned immune balance

In the early postoperative period, tissue injury and other surgical trauma evoke a physiological local and systemic inflammatory response that is essential for pathogen defense as well as wound healing. The immune response to surgery is a finely-tuned balance between proinflammatory immune cell responses, including humoral factors (e.g. cytokines and the complement system) and immunosuppressive mechanisms engaged at the local and systemic level. Dysregulation in either direction increases the risk of postoperative complications: overt proinflammatory immune responses that lead to systemic inflammatory response syndrome (SIRS) are associated with severe morbidity [20] while profound postoperative immunosuppression may predispose a patient to postsurgical wound infection and bacterial sepsis [13,15,21,22].

Physiologically, protein and non-protein components of intracellular compartments and extracellular matrix act as biochemical alarmins for the immune system and are recognized as damage-associated molecular patterns (DAMPs) by pattern recognition receptors (PRRs), such as toll-like receptors (TLRs), expressed primarily on innate immune cell types (e.g. macrophages, monocytes, dendritic cells, and neutrophils) [23,24]. Endogenous ligands, including high-mobility group box 1 protein (HMGB1), adenosine triphosphate (ATP), heat-shock proteins (HSP), or hyaluronic acid, are released from the surgical injury site and bind to TLR2, TLR4, the receptor for advanced glycation end products (RAGE), and modulating receptors that amplify PRR signaling, such as triggering receptor expressed on myeloid cells 1 (TREM-1) [25–27]. Upon activation by DAMPs, TLRs signal mainly through a MyD88-dependent pathway via IRAK1/4 and TRAF6, and, to a lesser extent, via the alternative adaptor protein TRIF, which targets the transcription factor IRF [23]. Downstream of MyD88-dependent signaling, mitogen-activated protein kinases (MAPKs) and IκB kinase (IKK) ultimately promote NF-κB and AP-1 mediated transcription [28] (Figure 1). Subsequently, proinflammatory cytokines IL-1β, IL-6, IL-8, IL-12, and TNFα, chemokines, growth factors, enzymes (e.g. metalloproteases, cyclooxygenases, or nitric oxide synthase), and other mediators of inflammation are released to kickstart an immediate response of the innate immune system, endothelial cells, and tissue adjacent to the surgical trauma in order to remove damaged tissue, to initiate wound healing, and to prevent pathogen intrusion [29].

Figure 1. Balance of proinflammatory and immunosuppressive features in the physiological immune response to surgery.

Alarmins/danger associated molecular patterns (DAMPs) such as HMGB1, ATP, heat shock proteins or hyaluronic acid are released from damaged tissue at the surgery site and recognized by pattern recognition receptors on immune cells. Proinflammatory activity of neutrophils, monocytes, macrophages and dendritic cells with the production of IL-1β, IL-6, IL-8, IL-12, and TNFα is counterbalanced by immunosuppressive processes, such as the induction of Tregs and promotion of Th2 cell immunity, and humoral factors (e.g. IL-10, TGF-β and IL-4).

Simultaneously, postoperative inflammatory signals are accompanied by anti-inflammatory and immunosuppressive adaptations. Numerically and functionally, adaptive immune cells, particularly T lymphocytes, are suppressed in the early postoperative period [30]. While multiple subsets of the T cell population, including Th1 cells and cytotoxic T cells, have been found to be decreased, activity and survival of immunosuppressive regulatory T cells (T_regs) increase upon DAMP signaling [31–33]. Interestingly, postoperative downregulation of Th1 activity, but not Th2 activity, results in a shift towards a Th2-emphasized immune polarization with increased production of immunosuppressive cytokines, such as IL-4 and IL-10 [34,35]. The suppression of T cell function may be linked to the rise of myeloid-derived suppressor cells (MDSCs) in the postoperative period: the induction of arginase 1 (ARG1) in this innate immune cell subset with powerful immunosuppressive capacity leads to an arginine-deficient state inhibitory to normal T cell function [22,36]. MDSCs are induced in a STAT3-dependent manner by inflammatory mediators in the early postoperative period, as their marked expansion profoundly dampens CD4⁺ and CD8⁺ T cell responses, promotes the activity of T_regs, and leads to a rise in anti-inflammatory cytokine levels. Paradoxically, MDSCs have been shown in animal models of acute inflammation, such as burns and sepsis, to promote the activation of certain innate immune cell subsets, suggesting that inhibition of MDSCs could be detrimental to surgical recovery [37,38].

Humoral immune factors as drivers of postoperative complications, do they tell the whole story?

The dysregulation of perioperative immune mechanisms that drive recovery after surgery is an essential pathophysiological component of postoperative complications, such as infection, protracted recovery, or cognitive decline. For the past decades, many studies have focused on the assessment of circulating humoral factors, such as cytokines and other serum or plasma inflammatory mediators to identify dysregulated immune responses associated with surgical complications. As humoral factors often exert pleiotropic cellular effects, these studies have provided important insight into pro- and anti-inflammatory mechanisms associated with postoperative complications. For example, high levels of IL-6 induce a Th17/T_reg imbalance, as IL-6 promotes specific differentiation of naïve CD4⁺ T cells into Th17 cells [39], inhibits TGF-β-induced T_reg differentiation [40], and can contribute to insufficient immune response to infection [41]. An early postoperative increase in IL-10 is also well known to induce immunosuppression by altering the monocyte activation through STAT3-dependent expression of SOCS3 [42] and sequestering of HLA-DR within cells [43]. A recent study identified a regulatory loop in which IL-10 directly restricts CD8⁺ T cell activation thereby contributing to the development of infections [44]. Many clinical and experimental studies implicated IL-10 as an inhibitor of the deleterious effects of the innate pro-inflammatory immune response in a Th1-dominated environment unless the balance is shifted towards a Th2 response [45].

Circulating humoral factors have also long been targets in the search for diagnostic or predictive biomarkers of infection or sepsis following surgery. Postoperative cytokine levels correlate with the duration and the extent of surgery [46]. Logically, several circulating cytokines (such as IL-6, IL-1α/β, and IL-10) have been suggested as useful markers for predicting the development of postoperative complications [47]. Elevated levels of these cytokines on postoperative day one have been associated with infection following colorectal surgery [48–50] and bone tumor surgery [51], overall postoperative complications in patients with gastric cancer [52], and longer ICU and hospital stays. In a study of genetic predisposition to postoperative sepsis, patients with an AA homozygous genotype of a TNF-α−308 G/A polymorphism in the TNF-α-gene were more likely to develop sepsis compared to non-AA homozygous individuals; the authors attributed this association to an increased capacity to produce proinflammatory TNF-α, IL-6, and IL-8 [53] (Figure 2).

Figure 2. Pathogenesis of adverse surgical outcomes.

Preoperative factors, such as individual susceptibility due to genetic or lifestyle factors, surgical triggers, such as laparoscopic or open approach, or duration of surgery, influence the degree of tissue damage. Following surgical injury, innate immune cells are activated to secrete cytokines, triggering a cascade of immunological events, such as T_reg/Th17 imbalance, CD8⁺ T cell activation, and HLA-DR sequestration. The nuanced individual differences in each of these pre- and perioperative factors and processes in turn determine whether patients will suffer from postoperative complications.

Another promising biomarker of postoperative complications is HMGB1, which is both passively released into the circulation from the nucleus of traumatized necrotic cells and rapidly secreted by stimulated monocytes, macrophages, dendritic cells, and natural killer cells [54] in response to surgery. In observational clinical studies, high serum HMGB1 concentrations within the first hours after surgery was associated with a significantly increased risk of developing complications, including long-term cognitive impairment [55] and acute lung injury [56]. HMGB1 triggers the nuclear translocation of NF-κB in monocytes, which activates the release of pro-inflammatory cytokines, including IL-6 and IL-1β [57]. Multiple preclinical studies have shown that systemic anti-HMGB1 antibody treatment exerts neuroprotective effects by lowering systemic and hippocampal inflammatory responses to surgery and, hence, prevents the development of cognitive impairment [58,59].

While the postoperative patterns of multiple cytokines (e.g. IL-1β, IL-6, IL-10) and alarmins (HMGB1) have been associated with adverse postoperative outcomes, only IL-6 levels have been associated with surgical complications when measured before surgery: elevated IL-6 plasma levels before surgery were observed to forecast adverse postoperative outcomes, including delirium onset [60], survival [61], and acute kidney injury [62]. Interestingly, the COVID-19 pandemic provided an illustration of the consequences of preoperative immune alterations on postoperative outcomes. When invading its target cells, SARS-CoV2 induces an interferon (IFN)-like response and release of key pro-inflammatory cytokines, the “cytokine storm”, such as IFNɣ, IL-1β, IL-33, and IL-6 [63], which are associated with sustained functional changes in circulating immune cells [64–66]. In a large international prospective cohort study, a greater risk of postoperative morbidity and mortality was observed in patients diagnosed with SARS-CoV-2 infection within six weeks prior to surgery (odds ratio > 2.8) [67]. This finding emphasized the link between the preoperative immune environment and postoperative prognosis.

Together, human studies of plasma factors have highlighted important immune dysfunctions that differentiate patients with and without postoperative complications. However, measurement of these factors alone has yielded only weak to moderate associations with clinical outcomes (e.g. AUC < 0.73) [68]. As such, few discoveries have translated into clinically relevant predictive tools. A major impediment has been the lack of high-content, functional assays that can integrate the assessment of plasma factors with single-cell assessment of the complex, multi-cellular immune response to surgery.

Towards an integrated, multiomic, predictive model of postoperative complications

High-content -omic technologies, together with recent computational advances that integrate transcriptomic, proteomic, and metabolomic data have enabled the immune system-wide characterization of human diseases with unprecedented depth and single-cell resolution. Application of these technologies to study the human immune response to surgery is increasingly recognized as a promising approach for the identification of patient-specific immune states predictive of postoperative complications (Figure 3).

Figure 3. Multiomic prediction of postoperative outcomes.

Surgical patients are recruited from major medical centers. 1) Whole blood samples are collected and processed for immunome, transcriptome, proteome, and metabolome analysis. 2) Machine learning algorithms applied to individual data layers predict surgical outcomes. The combined multiomic model shows increased predictive power (black) when compared to individual -omics. 3) Patients are stratified into poor or favorable surgical outcomes based on model output.

Single-cell technologies, such as mass cytometry, or single-cell RNA sequencing are particularly valuable, as the data generated holds substantially more predictive information than the clinical immunological laboratory results often collected before surgery. For example, mass cytometry can quantify over 50 proteomic parameters on a cell-by-cell basis, with a throughput allowing for the single-cell analysis of millions of cells per sample [69]. As such, mass cytometry allows in-depth characterization of the entire immune system both phenotypically and functionally, including the simultaneous identification of multiple immune cell subsets with quantification of activity (endogenous) and reactivity (response to proinflammatory stimuli) of key signaling pathways to produce a set of biological features predictive of the postoperative immune response [13].

Recent studies in patients undergoing total hip arthroplasties have applied this technology to identify perioperative immune signatures of surgical recovery. Specifically, increased immunosuppressive cellular activity preoperatively, including elevated pSTAT3 and pCREB activity within myeloid (M)-MDSCs, was predictive of longer-lasting functional impairment, fatigue, and pain after surgery [13,70]. Mass cytometry is also a valuable tool to profile the effects of perioperative therapeutic interventions on multiple immune cell subsets. As an example, preoperative administration of steroids is common to decrease postoperative nausea and vomiting; its effectiveness to improve surgical recovery, however, remains controversial. In a randomized controlled trial of methylprednisolone administration, mass cytometry analysis revealed major immune cell signaling alterations postoperatively that were predominantly confined to the adaptive immune compartment, while innate immune markers previously associated with surgical recovery were largely preserved in patients who received methylprednisolone [71,72].

While mass cytometry and other single-cell proteomic approaches require the a priori selection of a restricted set of analytes, transcriptomic platforms can offer untargeted analyses. RNA-sequencing (RNAseq) allows for a transcriptome-wide analysis of immune cell activity by detection of messenger, micro, transfer, and ribosomal RNA from the bulk analysis of pooled (bulk RNAseq) or single (single-cell RNAseq) peripheral immune cells. In fully unsupervised analyses of immune cell transcriptional activity in the context of surgery, RNAseq effectively predicted patients who developed pouchitis after ileal pouch–anal anastomosis [73], as well as individuals who developed postoperative recurrence of Crohn’s disease [74].

Circulating proteins and metabolites released from immune and non-immune cells are components of the regulatory network connecting immune cells with other biological systems implicated in the physiological response to surgery (e.g. neuro-endocrine and cardiovascular systems). Therefore, examination of the plasma proteome and metabolome are essential elements of the multiomic analysis of local and systemic immune responses predictive of postoperative complications. Recent developments in antibody-based (Olink [75]) or aptamer-based (Somalogic [76]) platforms have revolutionized the field of proteomics, allowing for analysis of up to 7000 plasma proteins. Application of high-content proteomic platforms to analyze patient samples collected before and after surgery has enabled the identification of predictive signatures of postoperative complications, increased length of stay, and risk of discharge to post-acute facility [77]. In addition, targeted and untargeted metabolomic platforms profile the molecular end-products of cell biological processes and provide a readout of cellular activity on a biochemical level. As such, a recent metabolomic study characterizing altered metabolite levels of anaerobic glycolysis as well as amino acid, nitrogenous, lipid, and bile acid metabolism after surgery identified metabolic signatures of postoperative complications following liver transplantation [78].

Integrating the knowledge garnered from individual -omic technologies to generate a holistic understanding of all immunological processes implicated in the pathogenesis of postoperative complications creates fundamental challenges in the bioinformatics field. Combining transcriptomic, proteomic, metabolomic, and mass cytometry features into multiomic predictive models requires the development of new statistical tools specifically adapted to high-dimensional datasets, which often contain more features than the number of samples collected: a computational dilemma referred to as “the curse of dimensionality” [79]. Regularized regression methods such as the Elastic Net (EN) algorithm are useful for the identification of predictive models of clinical outcomes and selection of key predictive features while accounting for the dimensionality of the dataset. Application of an EN algorithm was highlighted in a multiomic study that combined mass cytometry with plasma proteomics analysis of peripheral blood samples to characterize the immunological properties of a nutritional intervention known to reduce the incidence of postoperative complications (an Arginine-rich supplement) in patients undergoing abdominal surgery [80]. In addition, recent application of a stacked generalization algorithm that combines multiple regularized regression models built on individual -omic datasets has developed as a valuable approach for -omic data integration, while improving overall predictive performance [81]. The stacked generalization algorithm and other multiomic data integration methods have been successfully implemented in several recent studies for the prediction of various clinical outcomes, including the development of insulin resistance [82], the onset of spontaneous labor [83], preterm birth [84], survival in pancreatic cancer [85], and COVID-19 severity [86]. Cost and access to technology are clear limitations of multiomic approaches which have thus far often restricted the study sample size. However, multiomic approaches provide the means to cast an exploratory net and “find the needle in a haystack” for the validation of reduced predictive models that contain a limited number of features in large prospective studies.

Conclusion

A detailed understanding of the human immune response to surgery is a necessary first step towards identification of mechanistic biomarkers for the accurate prediction of postoperative complications. Recent technological advances allowing for an immune system-wide assessment of single-cell events in patients undergoing surgery form the foundation for a comprehensive map of the pre- and postoperative immune states predictive of surgical recovery. Multiomic modeling in studies integrating mass cytometry, plasma proteomic, metabolomic, and transcriptomic data hold promise for the discovery of pre-operative signatures predictive of postoperative outcomes. However, large-scale multiomic studies, as well as innovative machine learning approaches [87], are urgently needed to test the generalizability of identified signatures and to improve the performance of predictive models. Importantly, as most surgeries are non-urgent and have flexibility in timing and approach, the accurate prediction of postoperative complications can guide clinical decision making and the implementation of pre-operative interventions, such as prehabilitation programs or targeted immune therapeutics, to improve patient outcomes after surgery.

Key points:

• Surgery induces significant tissue trauma that triggers a complex immune reaction with a finely-tuned balance of pro- and anti-inflammatory response mechanisms; however, imbalanced immune activity after surgery is associated with postoperative complications.

• Postoperative patterns of multiple humoral factors have been associated with adverse postoperative outcomes, but only IL-6 levels have been associated with surgical complications when measured before surgery.

• Innovative single-cell technologies (e.g. mass cytometry and single-cell RNA sequencing) that allow in-depth characterization of the entire immune system recently demonstrated an association between increased preoperative immunosuppressive cellular activity and longer-lasting functional impairment, fatigue, and pain after surgery.

• Recent developments in high-content proteomic platforms that analyze patient samples collected before and after surgery have enabled the identification of predictive signatures of postoperative complications.

• Large multiomic analyses are needed for the characterization of integrated biological signatures that can predict poor surgical outcomes and ultimately guide clinical decisions, including the implementation of preoperative interventions to improve surgical outcomes.