Circulating levels of soluble Fas (sCD95) are associated with risk for development of a nonresolving acute kidney injury subphenotype

Acute kidney injury (AKI) is common in the intensive care unit (ICU) and is associated with substantial morbidity and mortality [1‐5]. Development of AKI occurs in response to a variety of toxic, inflammatory, and ischemic events, with the most common predisposing risk being sepsis. Biomarkers of functional (serum creatinine [SCr], blood urea nitrogen) or structural (neutrophil gelatinase-associated lipocalin, kidney injury molecule 1) kidney injury have been shown to have some utility in the early identification of AKI [6‐8]. However, further exploration of the biologic pathways linked to the development of more severe forms of AKI is needed.

AKI is a heterogeneous entity, and increasing evidence shows that the Kidney Disease: Improving Global Outcomes (KDIGO) criteria for AKI severity may not adequately capture this heterogeneity [9‐11]. In clinical trials, this heterogeneity may obscure treatment effects that are present only in subgroups of patients, potentially contributing to the increasing number of negative interventional trials in AKI [12]. We recently identified two AKI subphenotypes (resolving and nonresolving) on the basis of the trajectory of SCr in the first 3 days after hospital presentation [13]. A resolving trajectory, defined as a decrease in SCr of 0.3 mg/dl or 25% from the maximum, was associated with the same risk of hospital mortality as that of subjects with no AKI. A nonresolving trajectory, defined as AKI that did not meet the resolving criteria, was associated with a 60% higher risk of hospital mortality relative to patients with no AKI, even after adjusting for KDIGO stage of AKI and other potential confounders. Traditional prognostic risk factors, such as Acute Physiology and Chronic Health Evaluation III (APACHE III) score, vasopressor use, or sepsis status, did not differ across AKI subphenotypes. Prior work, primarily in animal models, has implicated different biologic processes, such as apoptosis, inflammation, and endothelial dysfunction, in the pathogenesis of AKI [14‐17]. It is not yet known whether these different pathophysiologic processes might contribute to the development of a nonresolving as opposed to a resolving AKI subphenotype.

In this report, we present results of analyses based on testing of whether circulating levels of biomarkers representative of two main biologic pathways, endothelial dysfunction (angiopoietin 1 [Ang-1], angiopoietin-2 [Ang-2], angiopoietin ratio [Ang-2/Ang-1], soluble vascular cell adhesion molecule [sVCAM]) and inflammation/apoptosis (soluble Fas [sFas], soluble tumor necrosis factor receptor 1 [sTNFR-1], interleukin-6 [IL-6], and IL-8), are differentially associated with AKI subphenotypes in a cohort of ICU patients with the systemic inflammatory response syndrome (SIRS). These analyses seek to shed light on the biology of these two distinct AKI subphenotypes, resolving versus nonresolving.

Detailed characteristics of the HMC-SIRS cohort have been published previously [22, 31]. The average age in the HMC-SIRS cohort was 54 years (SD ±16); 65% were men; 77% were Caucasian; 28% had a history of diabetes mellitus; and 9% had chronic kidney disease. Sepsis-3 was the admission diagnosis of 58% of subjects. During the first 3 days of ICU admission, 64% of subjects required mechanical ventilation for any period, and 23% required vasopressors (Table 1). Of 1241 eligible subjects, 868 subjects (70%) developed AKI within 3 days of study enrollment. The incidence of AKI is consistent with prior reports identifying the rate of AKI in ICU populations [32, 33]. Of the subjects with AKI, 502 subjects had a resolving subphenotype, and 366 had a nonresolving subphenotype. The proportion of subjects with a nonresolving subphenotype was greater than in our prior work identifying AKI subphenotypes in post-trauma and mixed medical-surgical ICU populations [13]. Subjects with the resolving and nonresolving subphenotypes had similar characteristics in multiple categories, including age, sex, body mass index, APACHE III score, cirrhosis, chronic kidney disease, need for vasopressors, and maximum SCr during the first 72 h of ICU care. Subjects with the nonresolving subphenotype had a higher rate of Sepsis-3 (66% versus 57%).

Because we tested eight different associations between biomarker concentrations and a nonresolving AKI subphenotype, we completed a correlation matrix using a Spearman’s correlation test to evaluate collinearity between biomarkers. Other than correlations of 0.66 between sFas and sTNFR-1 and 0.64 between sTNFR-1 and Ang-2, the rest of the correlations between biomarkers were < 0.6 (Additional file 1: Table S2). Thus, we chose the most conservative estimate of a Bonferroni-corrected p value cutoff of 0.05/8 = 0.00625 to test for significance.

In our analysis of a large cohort of critically ill subjects, we confirmed the presence of two AKI subphenotypes based on the trajectory of SCr in the first 3 days of ICU admission. As we previously demonstrated, subjects with a resolving AKI subphenotype have a similar risk of mortality and RRT as that of subjects with no AKI, but subjects with a nonresolving SCr trajectory have a twofold higher risk of death [13]. In contrast to a recently published work in which researchers excluded subjects with KDIGO stage 1 AKI to identify trajectories of AKI, we included all subjects with AKI in our analyses [11]. Minor changes in SCr are important [35], and KDIGO stage 1 AKI includes a large, heterogeneous population of all subjects with AKI (approximately 43% of subjects with AKI in our study were in KDIGO stage 1). To evaluate the pathophysiology of these distinct AKI subphenotypes, we measured plasma biomarkers associated with the development of AKI in key biologic pathways: inflammation, apoptosis, and endothelial dysfunction. We found that higher levels of sFas were associated with an elevated risk of developing a nonresolving AKI subphenotype.

Fas is a type 1 membrane protein that belongs to the tumor necrosis factor receptor 4 superfamily, which activates intracellular signaling after binding of Fas ligand (FasL) [36]. Fas ligation leads to a series of intracellular signaling events, culminating in activation of the death-inducing signaling complexes, which promote the activation of caspase-8-mediated apoptosis. Additionally, Fas ligation is believed to have an inflammatory role through cytokine production and then recruitment of proinflammatory cells [37]. sFas is a truncated form of Fas believed to result from proteolytic cleavage of membrane-bound receptors or alternative splicing of messenger RNA transcripts [38]. We have previously shown that genetic polymorphisms in FAS-related genes are associated with the development of AKI in subjects with acute respiratory distress syndrome (ARDS) [39]. Other studies have implicated the Fas pathway in the development of AKI in non-ARDS populations, such as patients with infection and chronic kidney disease [40‐42]. Moreover, Ko et al. [14] showed in a murine model that a genetic deficiency of functional FasL protects mice from bilateral renal ischemia-reperfusion injury as measured by decreased apoptosis based on caspase 3 immunohistochemical staining, as well as decreases in SCr. Further, these authors also showed that pharmacologic blockade of FasL with an anti-FasL monoclonal immunoglobulin G antibody protected the kidneys of wild-type mice from ischemia-reperfusion injury.

It is well known that septic shock is a strong risk factor for AKI in the critically ill. We found a higher prevalence of sepsis and vasopressor use in the nonresolving AKI subphenotype than in the resolving AKI subphenotype. To account for the confounding of septic shock on the development of a nonresolving AKI subphenotype, we completed a sensitivity analysis limited to subjects with septic shock. We found that the association of sFas with increased risk of a nonresolving AKI subphenotype was unchanged. Of note, biomarker levels of endothelial dysfunction (Ang-1 and sVCAM) were also associated with risk for nonresolving AKI in the septic shock subset. These findings suggest that sFas is linked to the nonresolving AKI subphenotype, not just to the severity of sepsis, and that endothelial dysfunction may play a larger role in the development of AKI in patients with septic shock than in the overall AKI population.

The identification of distinct molecular profiles for different AKI subphenotypes has several potential implications. First, our findings suggest that inclusion of subjects with both resolving and nonresolving AKI subphenotypes in clinical trials of treatments for AKI adds a degree of disease heterogeneity that could reduce the efficacy of specific therapeutic interventions based on molecular mechanisms [43]. Second, our findings suggest that the Fas/FasL system may be differentially involved in the development of a nonresolving AKI subphenotype. We do not yet know whether the increased circulating levels of sFas are simply associated with or causal in the development of AKI. However, we do know from animal models that deficiency or blockade of the Fas/FasL system can protect from renal injury [14]. Thus, we favor a model in which increased sFas represents a byproduct of increased Fas pathway activity. Future studies will address the mechanistic role of sFas in the development of AKI. Third, our findings of associations between markers of endothelial dysfunction and the nonresolving AKI subphenotype in patients with septic shock support prior work demonstrating the protective effect of Ang1/Tie2 agonists in animal models of organ dysfunction in sepsis [44], and they suggest that targeting this pathway may have more of an impact in subjects with septic shock at risk for the nonresolving AKI subphenotype. Fourth, the identification of molecular subphenotypes in alternative heterogeneous diseases, such as asthma and lung cancer, has informed novel treatment strategies [45, 46]. Similarly, continued molecular characterization of AKI subphenotypes may identify therapeutic pathways for further investigation.

This study has several important strengths that support its novelty and robustness. First, to the best of our knowledge, this is the first study to link sFas concentrations and the development of AKI. Second, our use of a large, prospective ICU cohort allowed for more precise assessments of the associations between biomarkers and risk for AKI. Indeed, this is one of the largest published studies of biomarkers and risk for AKI. Third, we minimized the potential for type I error by using the Bonferroni correction to account for multiple hypothesis testing. Fourth, the simultaneous evaluation of two distinct pathways implicated in the pathogenesis of AKI allowed for a comparison of the relative strengths of association. Fifth, the association of sFas concentrations with a nonresolving subphenotype persisted in the subgroup of patients with severe AKI (KDIGO stage 2 or 3).

The study has several important limitations. First, our definition of “resolving” AKI required an SCr decrease of only 0.3 mg/dl or 25% from the maximum value. This relatively minimal decrease raises the possibility that a significant number of patients who actually had a nonresolving AKI subphenotype were misclassified as having a resolving AKI subphenotype. However, we have previously shown that this definition is superior to alternative definitions of the subphenotypes with regard to the separation by outcome (mortality) [13]. Even when stratified by KDIGO AKI stage, the patients with the nonresolving AKI subphenotype had a higher mortality and greater need for RRT than those with the resolving subphenotype. Also, misclassification of this type would be expected to add experimental noise and to have biased our results toward the null. Second, our study does not provide insight into the functional significance of elevated sFas in AKI. Future studies are needed to evaluate the relationship of sFas to sFasL and to address mechanistic questions of the role of sFas in AKI. Third, it is unknown if sFas is filtered from the glomerular capillary. In one prior study, researchers reported that sFas concentrations increased with worsening kidney function [47], but it is unknown if this increase was due to activation of the Fas/FasL system, entirely a function of decreased filtration of circulating sFas, or a combination of both. Fourth, clinical factors besides AKI, such as sepsis [48], major trauma [49], or active malignancy [50‐52], have been associated with increased circulating levels of sFas. To account for these additional clinical factors, we excluded from our study patients with major trauma or active malignancy. Additionally, we completed a sensitivity analysis of patients with septic shock to determine if AKI influences sFas levels independently of sepsis.

We have shown that a biomarker of Fas pathway activity, sFas, is associated with the risk of developing a nonresolving AKI subphenotype in critically ill patients without major trauma, severe immune suppression, or active cancer. In contrast, biomarkers of endothelial dysfunction demonstrated an association with this subphenotype only in the subgroup of subjects with septic shock. These findings extend experimental data from animal models, suggesting that activation of the Fas pathway and, to a lesser extent, suppression of the Ang-1 axis play an important role in the pathogenesis of AKI. The continued molecular identification of AKI subphenotypes may allow recognition of subjects at high risk for poor outcomes, might facilitate the identification of novel therapeutic targets, and might allow for targeted enrollment in clinical trials.