Four Biomarkers Linked to Activation of Cluster of Differentiation 8-Positive Lymphocytes Predict Clinical Outcomes in Pediatric Acute Liver Failure

Abstract

Background and aims: Immune dysregulation contributes to the pathogenesis of pediatric acute liver failure (PALF). Our aim was to identify immune activation markers (IAMs) in PALF that are associated with a distinct clinical phenotype and outcome.

Approach and results: Among 47 PALF study participants, 12 IAMs collected ≤6 days after enrollment were measured by flow cytometry and IMMULITE assay on blood natural killer and cluster of differentiation 8-positive (CD8⁺ ) lymphocytes and subjected to unsupervised hierarchical analyses. A derivation cohort using 4 of 12 IAMs which were available in all participants (percent perforin-positive and percent granzyme-positive CD8 cells, absolute number of CD8 cells, soluble interleukin-2 receptor level) were sufficient to define high (n = 10), medium (n = 15), and low IAM (n = 22) cohorts. High IAM was more frequent among those with indeterminate etiology than those with defined diagnoses (80% versus 20%, P < 0.001). High IAM was associated with higher peak serum total bilirubin levels than low IAM (median peak 21.7 versus 4.8 mg/dL, P < 0.001) and peak coma grades. The 21-day outcomes differed between groups, with liver transplantation more frequent in high IAM participants (62.5%) than those with medium (28.2%) or low IAM (4.8%) (P = 0.002); no deaths were reported. In an independent validation cohort (n = 71) enrolled in a prior study, segregation of IAM groups by etiology, initial biochemistries, and short-term outcomes was similar, although not statistically significant. High serum aminotransferases, total bilirubin levels, and leukopenia at study entry predicted a high immune activation profile.

Conclusion: Four circulating T-lymphocyte activation markers identify a subgroup of PALF participants with evidence of immune activation associated with a distinct clinical phenotype and liver transplantation; these biomarkers may identify PALF participants eligible for future clinical trials of early targeted immunosuppression.

FIG. 1.

Consort diagram of participant selection process for derivation, validation, and combined cohorts. Phases 2 and 3 PALF Study Group participants were eligible for inclusion in the validation and derivation cohorts as described. *Validation cohort initially described in Bucuvalas et al.⁽⁶⁾

FIG. 2.

Unsupervised hierarchical analysis of IAM expression results and development of an LD equation to determine the degree of IA in PALF at presentation. In (A), 12 IAMs (absolute number of CD8 and NK cells and their percent of lymphocytes, percent of CD8 and NK cells expressing perforin or granzyme, median fluorescent intensity of perforin/granzyme expression on NK cells, and IL2SR level) were available in 41 participants and subjected to cluster analysis. Gap analyses identified two IA cluster groups associated with low (group 2) and high (group 1) degrees of IA. In (B), four biomarkers (percent perforin-positive CD8 cells, percent granzyme B-positive CD8 cells, percent CD8⁺ lymphocytes, and IL2SR) were included in the analysis, which increased the number of participants with PALF to 47. Gap analyses identified three IA cluster groups with low (group 2), medium (group 1), and high (group 3) degrees of IA (also see Supporting Fig. S2). In (C), the discrimination between the expression levels (medians and interquartile ranges) for the four IAMs is displayed for the 47 participants assigned to low, medium, or high IA based on the cluster analysis of HA-4. In (D), the equation to calculate the LD score and the interquartile range for LD1 scores for assignment to IA group are displayed. Abbreviations: Gran, granzyme; MCF, mean channel fluorescence; Perf, perforin.

FIG. 3.

Clustering of participants in the combined cohort of 118 participants. The expression levels for the four IAMs including percent granzyme and percent perforin expressing CD8 cells, number of circulating CD8 lymphocytes, and serum levels for IL2SR were subjected to unsupervised cluster analysis defining three groups of participants in the 118 participants enrolled in the derivation and validation cohorts. Gap analyses identified three IA cluster groups with low (group 2), medium (group 1), and high (group 3) degrees of (A). Association between group assignments to high versus combined medium or low IA and 21-day survival with the native liver censored for death, liver transplant, or death or liver transplant is shown in (B). Abbreviations: Gran, granzyme; Perf, perforin.

FIG. 4.

Prediction of high IA using routine clinical laboratory tests at study entry and liver tissue CD8 immunohistochemistry. In (A), a ROC curve is shown following a multivariable logistic regression analysis of 77 participants with the complete set of the clinical parameters ALT, total bilirubin, and WBC obtained at study entry. A value > 3.3542 was used to predict high IA. In (B), CD8 immunohistochemistry of liver tissue available from PALF participants (n = 10) in the derivation cohort was subjectively scored by investigators blinded to clinical information and thereafter compared to IA group status. There were no medium IA group specimens in those with liver tissue available for analysis. Abbreviation: Tbili, total bilirubin.