A biomarker for prognosis of critically ill surgical patients after sepsis

by Dr R. Stephen Smith, Nicole R. Mercier, Dr Scott C. Brakenridge
The best-known function of glucagon-like peptide 1 (GLP-1) is to promote insulin secretion in a glucose-dependent manner. However, GLP-1 also has effects in other tissues, notable the brain and stomach. Interestingly, GLP-1 is also linked to the immune system, and levels of GLP-1 have been shown to rise rapidly in response to cytokines, such as interleukin 6. Moreover, it is known that critical illness, such as sepsis, causes a disruption of glucose homeostasis. This article discusses how elevated levels of GLP-1 after the onset of sepsis seem to be a strong and independent predictor for the development of chronic critical illness and death in the subsequent 6 months.

Background

Glucagon-like peptide-1 (GLP-1; Fig. 1) is a gut derived incretin hormone that stimulates insulin secretion, suppresses glucagon secretion, inhibits gastric emptying and decreases appetite. Loss of glucose homeostasis is a frequent occurrence in critically ill patients [1–3]. Numerous studies have demonstrated the association between loss of glycemic control and poor outcomes in the critical care setting [4–6]. Hyperglycemia is associated with increased mortality, increased rates of infection and increased length of intensive care unit (ICU) use [5]. However, glycemic control is variable from patient to patient despite implementation of standard glycemic control protocols [6] and the pathophysiologic mechanisms for this variability are not fully understood.
Loss of glycemic homeostasis is a well-known phenomenon following critical illness and is associated with poor outcomes after trauma or sepsis [7–9], and we have previously found that GLP-1 levels were abnormal in seriously injured patients [10]. Levels of GLP-1 during periods of physiologic stress secondary to sepsis has not been extensively studied or documented. Therefore, in our recent study, summarized here, we sought to determine the influence of sepsis on circulating levels of GLP-1 [11]. We hypothesized that abnormal GLP-1 levels would be predictive of poor outcomes.

Methods

Study design
This is an analysis of an ongoing prospective longitudinal, observational cohort study of critically ill trauma and surgical intensive care unit (SICU) patients with sepsis at an academic medical centre. Overall programme study design and protocols have been previously published [12]. GLP-1 biomarker analysis was performed on a group of 157 consecutively enrolled patients between January|2015 and September|2016. Patients included in this study were those SICU patients that screened positive for sepsis and were placed on our unit-specific sepsis protocol. Sepsis screening was performed using the Modified Early Warning Signs-Sepsis Recognition System (MEWS-SRS) [13]. All enrolled patients were managed using a standard protocol based on the Surviving Sepsis guidelines.
Patients
Inclusion criteria were: (1) age ≥18 years; (2) SICU admission; and (3) diagnosis of sepsis, severe sepsis or septic shock [14]. Exclusion criteria included: (1) refractory shock (death <24|hours) or inability to achieve source control; (2) pre-sepsis life expectancy <3|months; (3) goals of care not consistent with aggressive management; (4) severe congestive heart failure (New York Heart Association class 4); (5) Child–Pugh class C liver disease or pre-liver transplant; (6) HIV with CD4+ count <200|cells/mm3; (7) chronic corticosteroids or immunosuppressive agent use; (8) pregnancy; (9) chemotherapy or radiotherapy within past 30|days; (10) severe traumatic brain injury; or (11) spinal cord injury resulting in permanent deficits. Baseline and inpatient clinical data collected prospectively during initial hospitalization included patient characteristics, sepsis severity, clinical and laboratory data, complications and patient disposition.
Post-sepsis patient courses were categorized as ‘early death’, chronic critical illness (CCI) or ‘rapid recovery’ (RAP). Patients with refractory sepsis and death within 24|hours were excluded. Early death was defined as survival for 24|hours, but death before day 14. CCI is defined as an ICU stay of greater than or equal to 14|days with evidence of persistent organ dysfunction (Fig. 2), determined using components of the Sequential Organ Failure Assessment (SOFA) score (cardiovascular SOFA ≥1, or score in any other organ system ≥2) [15]. Inpatient outcomes included in-hospital mortality, hospital and ICU length of stay, ventilator days, organ dysfunction and development of CCI. Six-month outcomes included performance status and mortality, which were determined by follow-up measured by the Zubrod scale.
Biomarker analyses
All subjects enrolled underwent peripheral blood sampling at 12|hours, 1, 4, 7, and 14|days after sepsis onset, and weekly thereafter while hospitalized. These samples were processed and stored for subsequent programme a priori and ad hoc biomarker analysis. Analysis included GLP-1 (Luminex; MilliporeSigma) and interleukin-6 (IL-6) (ELISA; MilliporeSigma). We used IL-6 as a well-established measure of the host innate inflammatory response. We have previously demonstrated that IL-6 is a predictive biomarker for CCI after sepsis [16].
Statistical analysis
Data are presented as either frequency and percentage, or mean and standard deviation, or median and 25th/75th percentiles. Student’s t-test, ANOVA and Kruskal–Wallis tests were used for comparison of continuous variables as appropriate. Chi-squared test and Fisher’s exact test were used for comparison of categorical variables. Measured biomarkers were compared using non-parametric rank tests to determine significant differences between groups at each time point. Two multivariate logistic regression models were constructed to determine if GLP-1 was an independent predictor at 24|hours of the development of CCI, and at day 14 of death or severe disability (Zubrod score 4–5) at 6|months after sepsis onset. The models were designated a priori to model GLP-1 for these outcomes while controlling for both IL-6 (systemic inflammation) and peak glucose level (glucose dysregulation). Adjusted odds ratios with 95% confidence intervals (95% CI) and the area under the receiver operating characteristics curve values (AUC) and Hosmer–Lemeshow goodness-of-fit test were used. Statistical analyses were performed with SAS (v.9.4, SAS Institute).

Results

Demographics, clinical course and outcomes
Over an 18-month period, 157 consecutively enrolled critically ill septic patients underwent circulating GLP-1 biomarker analyses. Overall, this cohort represents an older group of patients, with a significant comorbidity burden, a high incidence of early infectionassociated organ dysfunction (severe sepsis/septic shock, 64%) and severe physiologic derangement at 24|hours after sepsis onset. Approximately one-third of patients had pre-existing diabetes mellitus requiring either hyperglycemic control medications. Most of the patients were admitted with an acute intra-abdominal infection. The severity of organ dysfunction in this septic group of patients was high and the incidence of multiple organ failure was 50%. We excluded patients with limited goals of care, refractory shock and death within 24|hours; therefore, in-hospital mortality was lower than many previously reported sepsis cohorts (8%).
Approximately 60% of critically ill septic patients demonstrated rapid recovery from organ dysfunction and were discharged from the ICU within 14|days. Four patients (2.5%) suffered an early death, defined as less than 14|days of sepsis onset. The remaining 55 patients (35%) initially survived the episode of sepsis, but developed CCI, as evidenced by persistent organ dysfunction and a prolonged ICU length of stay (≥14|days). There were no significant differences in hospital admission diagnosis or septic source between the groups of patients. Compared to RAP patients, patients that developed CCI were older, had significantly higher pressor requirements for septic shock, and a greater severity of organ failure. Patients that developed CCI had a higher incidence of comorbidities than RAP patients, but there was no difference in the comorbidity rate of diabetes mellitus between CCI and RAP groups. Compared to RAP patients, those that developed CCI had significantly worse clinical outcomes (ventilator days, ICU days and mortality). Approximately 85% of patients that developed CCI survived to acute hospital discharge, but 90% of these survivors were discharged to facilities associated with poor long-term outcomes (long-term acute care, skilled nursing facility, another hospital, hospice). Post-discharge follow-up showed a 6-month mortality for the CCI group of 40%. In comparison, 6-month mortality for the RAP group was only 5%.
GLP-1 biomarker analyses
Owing to similarity of clinical course, the small number (n|=|4) of early death patients were combined with the CCI group for biomarker analyses. Circulating IL-6 levels were significantly elevated in all septic patients out to 28|days after sepsis onset as compared to healthy matched controls. Additionally, IL-6 levels were significantly elevated in CCI patients as compared to the RAP group at all time points between 24|hours and 28|days after onset of sepsis.
GLP-1 levels were significantly elevated among CCI patients at all measured time points from sepsis onset to 21|days. Maximum daily blood glucose levels were significantly elevated in CCI patients as compared to the RAP group across all measured time points. Accordingly, GLP-1 and blood glucose levels were modestly, but significantly correlated (Spearman correlation|=|0.27, P<0.0001). Logistic regression analysis revealed that GLP-1 level at 24|hours after sepsis onset was an independent predictor of CCI. Additionally, GLP-1 level at day 14 was an independent predictor for death or severe disability at 6|months.

Discussion

Role of GLP-1 in glucose metabolism
Gut hormones facilitate metabolism of glucose through stimulation of insulin secretion and various other mechanisms [17]. In normal physiology, plasma levels of GLP-1 increase soon after oral intake, whereas plasma levels of GLP-1 are low during fasting. However, the levels, trajectory and role of GLP-1 in patients with critical illness remains unclear. In murine models, increases in GLP-1 levels correlate with increased insulin secretion and reduced blood glucose levels after a glycemic challenge [18]. The GLP-1 receptor is expressed in pancreatic islet alpha and beta cells and in a variety of peripheral tissues including the central and peripheral nervous systems, heart, kidney, lung and the gastrointestinal tract. Activation of incretin receptors on the pancreatic beta cells causes a rapid increase in insulin secretion [19], and, during normal physiologic states, this response occurs in a glucose-dependent manner [20]. GLP-1 inhibits glucagon secretion and promotes glucose metabolism through neural mechanisms that contribute to glucose regulation. Antagonism of GLP-1 decreases insulin secretion and causes an increase in glucose levels [21]. Additionally, it has been shown that administration of exogenous GLP-1 stimulates insulin secretion and causes glucagon suppression that results in decreased blood glucose levels [22].
GLP-1 receptors are abundant in many extra-pancreatic tissues, including the gastric, hepatic, cardiac and neural tissues. GLP-1 is an enterogastrone – meaning that it decreases gastric emptying in a dose dependent manner [23]. GLP-1 also exhibits cardiovascular effects through increased inotropicity and chronotropicity as well as increasing blood pressure [24]. GLP-1 receptors are also present in hepatocytes and appear to affect both glycogen and glucose metabolism [25]. GLP-1 receptors are also found in the brain and nerves. Administration of GLP-1 into the cerebrum and ventricles of animals produces delayed gastric motility and gastric acid secretion [26].
Critical illness and loss of glucose homeostasis
It has been previously established that illness results in loss of glucose homeostasis. The hyperglycemia that occurs after trauma and serious illness impacts a number of physiologic systems including immune function, wound healing and other areas of metabolism. Acutely and critically ill patients with hyperglycemia have poorer outcomes [27–30]. The mechanisms responsible for glucose dysregulation in the ICU are complex, but is it seems apparent that levels of GLP-1 play an important role in glucose homeostasis. Deane et al. demonstrated that the exogenous administration of GLP-1 analogues in critically ill patients alters glucose metabolism [31–33]. Our evaluation of septic patients confirms these previous observations. Both GLP-1 levels and glucose levels were consistently and persistently elevated in patients that develop CCI. GLP-1 and blood glucose levels showed significant correlation in this group. The most straightforward explanation for this finding is that elevated GLP-1 levels were in response to high blood glucose. This is the well-documented function of the incretin system. GLP-1 receptors present on pancreatic islet cells increase insulin secretion. However, extra-pancreatic functions of GLP-1 appear to be involved in the response to sepsis. Our regression models demonstrated that an elevated GLP-1 level at 24|hours is an independent predictor for the development of CCI. Elevated GLP-1 levels 14|days after the onset of sepsis is a strong and independent predictor of death or severe disability in the subsequent 6 months. Surprisingly, this analysis showed that an elevated GLP-1 level is a better predictor of outcome than IL-6. Patients that developed CCI were older, more likely to require pressor support and had more medical comorbidities. However, the group that went on to CCI did not have a greater incidence of pre-sepsis diabetes. Again, this is suggestive of factors other than simple hyperglycemia causing a compensatory elevation of GLP-1 levels.
GLP-1 secretion has been demonstrated to rapidly increase in response to cytokines. For example, Ellingaard et al. showed that administration of IL-6 stimulates GLP-1 secretion from intestinal L cells and pancreatic alpha cells [34]. This mechanism increased insulin secretion and improved glycemic control. This group concluded that IL-6 mediates crosstalk between insulin sensitive tissues, intestinal L cells and pancreatic islet cells to adapt to changes in insulin demand. LeBrun et al. noted that GLP-1 suppresses inflammation and promotes gut mucosal integrity. Furthermore, this group demonstrated that GLP-1 levels increased rapidly after the administration of lipopolysaccharide (LPS) in mice. This increase in GLP-1 was detected before measurable changes in cytokine levels or LPS. A similar response was noted after gut ischemia [35]. Lebherz et al. have demonstrated the predictive value of GLP-1 levels in critically ill patients [36]. This group measured GLP-1 levels in critically ill patients admitted to an ICU, patients with chronic kidney disease on hemodialysis and a control group without acute inflammation or kidney disease. Critically ill patients had a 6-fold increase in GLP-1 levels compared to the control group. Those requiring hemodialysis exhibited a fourfold greater GLP-1 level compared to controls. Lebherz concluded that both chronic and acute inflammatory states, including sepsis, increase circulating GLP-1 levels. This group further demonstrated in vivo that serum from critically ill patients had a strong potential for increasing GLP-1 secretion. Importantly, this group concluded that elevated GLP-1 levels were an independent predictor of mortality in critically ill and end-stage renal disease patients [36].
It is convenient to hypothesize that elevated GLP-1 levels are simply another compensatory biomarker of dysregulated glucose control after an acute, severe pro-inflammatory stressor. Another possibility is that GLP-1 elevation is representative of, or even contributory to, persistent low-grade inflammation and catabolism after sepsis. We have shown previously that a pathophysiologic syndrome of persistent inflammation, immunosuppression and catabolism is the mechanism driving the development of CCI, dismal long-term outcomes and mortality after sepsis [15, 16]. We have hypothesized that persistent kidney dysfunction perpetuates inflammation and immunosuppression through the local and systemic release of damage-associated molecular patterns, cytokines and the expansion of myeloid derived suppressor cells and also through metabolic reprograming (i.e. aerobic glycolysis) [37, 38]. After sepsis (or severe trauma) persistently elevated levels of GLP-1 may represent an underlying shift in metabolic programming towards aerobic glycolysis.

Conclusions

Elevated circulating GLP-1 levels within 24|hours of sepsis are a strong predictor of early death or the development of CCI. Elevated GLP-1 levels appear to be a better predictor of a poor outcome than IL-6. The predictive value of GLP-1 appears to be independent of other factors, such as hyperglycemia. Among early survivors, persistently elevated GLP-1 levels at day 14 are strongly predictive of death or severe functional disability at 6|months. Persistently elevated GLP-1 levels may be a marker of ongoing metabolic dysfunction and a non-resolving catabolic state. Future work should focus on elaborating the relationship of elevated GLP-1 to these underlying mechanisms.

The authors

R. Stephen Smith*1 MD FACS, Nicole R. Mercier2 MS, Scott C. Brakenridge1
MD FACS
1Department of Surgery, University of Florida, Gainesville, FL USA
2University of Arkansas for Medical Sciences, Little Rock, AR, USA

*Corresponding author
E-mail: Steve.Smith@surgery.ufl.edu

The advent of matrix-assisted laser desorption ionization has allowed the technique of mass spectrometry imaging to be used for relatively large biomolecules, enabling visualization of their location and distribution within tissues, as well as the identification of, and any changes in, disease biomarkers. Dr Shannon Cornett, applications development manager from Bruker Daltonics, discusses recent technological advances in MSI and their impact on clinical diagnostics research.

What is mass spectrometry imaging?

Mass spectrometry imaging (MSI) was originally introduced more than 50|years ago as a tool to study semiconductor surfaces. Using mass spectrometry (MS), MSI enables the visualization of the spatial distribution of molecules – biomarkers, metabolites, peptides or proteins – by their molecular masses. In practice, mass spectra are collected on an array of spatial coordinates until the entire sample is scanned. By choosing a peak in the resulting spectra that corresponds to the compound of interest, the MS data is used to map its distribution across the sample, creating pictures of the spatially resolved distribution of a compound.
In 1988, Franz Hillenkamp and Michael Karas described a method whereby they used a laser to irradiate a crystalline admixture of a biomolecule and small organic acid. Called matrix-assisted laser desorption/ionization (MALDI), they laid the foundation of a technology that opened new possibilities across science – from organic and polymer chemistry to proteomics, microbiology and drug development.
The pioneering work of Richard Caprioli and colleagues in the late 1990s demonstrated how MALDI-MS could be applied to visualize distributions of large biomolecules (such as proteins and lipids) i n cells and tissue to reveal greater insight into how molecular expression is changed by diseases like cancer. MSI can be used with different ionization techniques, including secondary ion mass spectrometry (SIMS), MALDI and desorption electrospray ionization (DESI). Today MALDI is the leading technology in clinical and biological applications of MSI, and it is widely considered the most effective for imaging tissue samples.

What are some of the current applications of MSI in clinical research?

MALDI imaging has continuously gained acceptance in clinical research. Significant technological and methodological improvements have contributed to enhance the performance of MALDI imaging recently, pushing the limits of throughput, spatial resolution and sensitivity. This has stimulated the spread of MALDI imaging across various biomedical research areas such as oncology, neurological disorders, cardiology and rheumatology, just to name a few.
MALDI imaging applications include protein characterization, glycoprotein analysis, quality control applications, polymer analysis and ultra-high throughput screening. Approximately 85% of MALDI imaging used in clinical research relates to cancer studies. Other major clinical research applications include Parkinson’s disease, Alzheimer’s, diabetes and non-alcoholic fatty liver disease, and well as detecting tumour margins.

What are the most recent technological advances in MALDI-MS imaging?

Because the information contained in each pixel is an unlabelled chemical fingerprint (or mass spectrum) of those particular cells, advances have been driven by: (a) faster acquisition of pixels; and (b) increasing the number of molecular features detected at each pixel location. Most recently, a novel combination of MALDI imaging with post-ionization (PI) has demonstrated significant enhancement in sensitivity. Studies show that this new combination, named MALDI-2, increases the sensitivity for many small molecules and lipids by up to three orders of magnitude. Further, some classes of compounds are only detectable with MALDI-2, expanding the range of applications for MALDI imaging even more. It is particularly significant for drug metabolism and pharmacokinetics (DMPK) studies.
A major limiting factor in modern drug development is use of the ‘wellstirred model’, which homogenizes organs and tissues before analysis and quantitation with liquid chromatography-mass spectrometry (LC-MS) studies. This approach is well-suited to providing exact amounts of drugs and metabolites within a target organ, but not readily compatible with pathology methods that seek to describe physiological effects of drug compounds. MALDI imaging has made a significant impact towards the conversion of plasma to tissue models by pinpointing the exact location of drugs and metabolites in tissue. The novel PI source enhances molecular imaging for pharma studies by increasing overall sensitivity, enabling quantitation for a wider range of dosing levels. Additionally, the increased variety of molecular classes expands the applicability of imaging to many more pharma projects that involve both xenobiotics and endogenous molecules.

The identification of biomarkers for disease diagnostics and prognosis is a key area of clinical research. How will these recent advances benefit biomarker identification?

Cancer cells and other diseased tissues have significant genetic and epigenetic modifications that influence the genomic expression cascade. Whether you are looking at the proteome, lipidome or metabolome, the spatial distribution of compounds contains valuable information for understanding your sample. If certain compounds are highly spatially concentrated or if molecules co-distribute in specific compartments, this vital information is lost when examining only homogenized samples. OMICS-based biomarker discovery becomes more complete when contextualized with spatial information to provide important clues into intercellular communications networks that are integral to cancer growth.
The combination of the novel MALDI-2 source mentioned previously provides the best opportunity to combine region-specific information from MALDI imaging with deep 4D OMICS coverage for biomarker discovery and molecular characterization. One advantage is the greater degree of molecular information that can be used to detect spatially significant region of interests (ROIs) in tissue sections that share common molecular signatures. For example, one study of micro-proteomic characterization of tumour subpopulations in breast cancer analysed MALDI Images of lipids to identify and target tumour subpopulations of specific molecular phenotype for laser capture microdissection (LCM). Pro¬tein extraction and tryptic digestion of small microdissected material was followed by proteomic analysis. Analysis of proteomics data of each molecular phenotype provided a more comprehensive mechanistic under¬standing of cell-type specific biological processes in situ to complement the workflow.

The resolution of near-isobaric ions has also been a challenge – what have the new technologies been able to do for imaging these types of molecules?

Quantification MSI (qMSI) remains a challenging but necessary aspect of MALDI imaging when applied to DMPK. Numerous factors such as the prevalence of isobaric endogenous compounds, chemical background from tissue matrix, and ion suppression often leads qMSI to have low reliability. However, to fully realize MALDI imaging as a versatile and highly applicable technique, quantification results must be accurate and reliable.
This is especially important in pharmacology, where the distribution of drugs and their metabolites in tissue is as important as the absolute quantification. The concentration of drugs present in disease sites determines the efficacy of the dosage and the impact of side effects, which in turn also illustrates the efficiency of any drug delivery methods. Pharmacology research is guided by determining the pharmacokinetics and the pharmacodynamics of the drug, and such studies often require screening large cohort of samples. Speed, sensitivity, spatial resolution, and specificity often determine the efficiency of a qMSI method.
The combination of the novel MALDI-2 source with timsTOF fleX can separate near-isobaric ions by their ion mobilities, significantly improving targeted compound specificity and sensitivity for quantitation in a complex molecular environment. MALDI spectra can be particularly complex in the lower mass-to-charge ratio (m/z) range owing to isobaric and near-isobaric interference from matrix ions and higher charge state analyte ions such as dimers from species to be quantified tracking of these ions that might affect the linear dynamic range. Results show that a combination of parallel accumulation and selective elution of ions by parallel accumulation– serial fragmentation (PASEF) and matching quadrupole isolation also improves sensitivity.

Mapping of steroids by traditional MSI has also been difficult. Why is it important to study steroids and how does MALDI imaging help?

Steroids are a biologically important class of compounds, and there is a growing interest in studying steroid distributions using MALDI imaging. As important components of cell membranes, steroids affect membrane fluidity and cell signalling. Hundreds of steroids can be found in plants, animals and fungi. Due to their nonpolar core structure, steroids do not ionize well by traditional MALDI imaging without specialized on-tissue derivatization protocols.
Steroids are one such analyte class that benefits strongly from this technology. We observed a sensitivity boost by up to 2–3 orders of magnitude depending on analyte and concentration [3]. It is a night-and-day difference.

What do you envisage for the future of MSI?

MALDI imaging has proven to be a powerful MS tool for mapping the distribution of molecules from a thin sample, ranging from small metabolites to large proteins, without molecular tags or labels. Application areas for MALDI imaging are diverse and growing, driven by the label-free nature of the technique and the ability to differentiate compounds by molecular weight, and also by collisional cross section.
The technology is suited for both targeted and untargeted studies. Untargeted discovery studies that use MALDI imaging are found throughout clinical research where the goal is to capitalize on the regional specificity to uncover novel biomarkers of disease and treatment. MALDI imaging is also revolutionizing pre-clinical drug discovery pipelines by providing direct distribution monitoring of targeted therapeutic compounds and their metabolites. Further, the label-free nature of the technique makes it possible to mine untargeted pharmacodynamic data from the same targeted data sets. Newer applications surrounding plants, polymers and microbes also are emerging. Eventually, we believe MALDI imaging has the potential to impact patient treatment. The technology is still evolving and we see a growing number of applications that can benefit from MALDI imaging as instrumentation continues to advance.

The interviewee

Dr Shannon Cornett PhD applications development manager
Bruker Daltonics, 40 Manning Rd, Billerica, MA 01821, USA
For further information visit Bruker Daltonics: www.bruker.com

Precise • Cost-efficient • Time-saving