AUTHORS
Michael Martens, PhD, Denggang Fu Sr, DF2, Amin Alousi, MD3, Iskra Pusic, MD MSCI4, George Carrum, MD5, Asad Bashey, MD, PhD6, Stephanie J. Lee Jr, MD, MPH7, John Levine, MD, MS8, Sameem Abedin, MD9, Yongxia Wu, MD, PhD10, Mehdi Hamadani, MD11, Xue-Zhong Yu Sr, MD, MSc12
AFFILIATIONS
1 Medical College of Wisconsin, Milwaukee, Wisconsin, United States
2 Medical College of Wisconsin, Milwaukee, Wisconsin, United States
3 MD Anderson Cancer Center, Houston, Texas, United States
4 Washington University in St. Louis, St. Louis, Missouri, United States
5 Baylor College of Medicine, Houston, Texas, United States
6 Northside Hospital, Atlanta, Georgia, United States
7 Fred Hutchinson Cancer Center, Seattle, Washington, United States
8 Mount Sinai Medical Center, New York, New York, United States
9 Medical College of Wisconsin, Milwaukee, Wisconsin, United States
10 Medical College of Wisconsin, Milwaukee, Wisconsin, United States
11 Medical College of Wisconsin, Milwaukee, Wisconsin, United States
12 Medical College of Wisconsin, Milwaukee, Wisconsin, United States
ABSTRACT
Background: Acute graft-versus-host disease (aGVHD) is a major contributor to mortality following allogeneic hematopoietic cell transplant (HCT), prompting extensive investigation into potential biomarkers to guide risk stratification and preemptive therapeutic interventions. We and others have demonstrated that many cytokines and chemokines are involved in GVHD pathogenesis. Despite the availability of several validated prognostic biomarkers (e.g., IL-6, ST2, CXCL10, and MCP-1) and potential therapeutic targets (e.g., TNFα, IL-1β, IL-6, and IL-23) for aGVHD post-HCT, additional biomarkers that improve risk stratification and identify novel therapeutic targets are in demand. In the present study, we comprehensively profiled a panel of 71 cytokines and chemokines in a multicenter prospective HCT cohort to identify biomarkers associated with aGVHD onset.
Methods: Blood samples were collected at baseline and at Days 7, 14, 21, 28, 42, 56, and 90 post-HCT from 276 participants in the Blood and Marrow Transplant Clinical Trials Network 1202 trial. Fluorescent intensities (FIs) and plasma concentrations of 71 cytokines and chemokines were assayed at each collection time using the Luminex 200 platform. Concentrations were calculated from the Fis using the standard curves. It is worth noting that ST2 was not included in the 71-panel. Associations of baseline cytokine levels with grade II-IV aGVHD onset were evaluated using univariable logistic regression models. Landmark Cox regression models were used to evaluate associations of Day 90 levels with chronic GVHD (cGVHD), relapse, and death within 2 years of HCT. Areas under the receiver operating characteristic curve (AUCs) and C statistics were used to summarize to abilities of cytokines to predict GVHD, relapse, and death. We also investigated the predictive power of multivariable models, composed of prognostic cytokines identified in univariable analysis, by using leave-one-out cross validated (LOOCV) AUCs and C statistics. FIs and plasma concentrations were log transformed for modeling to reduce skewness. A p-value cutoff of 0.01 determined statistical significance.
Results: Of the 71 cytokines and chemokines evaluated, baseline levels of Fractalkine (CX3CL1) and TNFβ were associated with grade II-IV aGVHD onset by Day 100 in univariable logistic models, with higher Fractalkine levels found to be deleterious and higher TNFβ being protective from aGVHD onset; Fractalkine odds ratios (ORs)=3.531 and 4.839 for log FI and concentration; TNFβ ORs=0.181 and 0.295 for log FI and concentration. A multivariable logistic model including both Fractalkine and TNFβ FIs yielded a LOOCV AUC of 0.651 (99% confidence interval (CI): 0.565-0.737) with sensitivity 0.611 and specificity 0.612. Day 90 levels of Fractalkine, G-CSF, and IL-1RA were associated with subsequent cGVHD in landmark Cox models. Higher levels of Fractalkine and G-CSF were protective of cGVHD onset, while higher IL-1RA increased the risk of this outcome; Fractalkine hazard ratios (HRs)=0.418 and 0.356 for log FI and concentration; G-CSF HRs=0.428 and 0.496 for log FI and concentration; IL-1RA HRs=1.838 and 1.752 for log FI and concentration. The Day 90 concentration of G-CSF showed the highest discriminative ability with a C statistic of 0.579 (99% CI: 0.517-0.641). Day 90 levels of I-309 (CCL1), M-CSF, and TNFα were associated with death in landmark Cox models, with higher levels of each marker indicating a higher mortality risk (I-309 HRs=1.981 and 2.292 for log FI and concentration; M-CSF HRs=5.126 and 3.370 for log FI and concentration; TNFα HRs=4.081 and 4.399 for log FI and concentration). The Day 90 TNFα concentration produced the best discrimination with a C statistic of 0.622 (99% CI: 0.545-0.700). No associations of Day 90 cytokines and chemokines with subsequent relapse were detected in any models.
Conclusions: Among the cytokines and chemokines examined, new prognostic markers for aGVHD (CX3CL1 and TNFβ), cGVHD (CX3CL1, G-CSF, and IL-1RA), and mortality (CCL1, M-CSF, and TNFα) were identified. These findings may improve post-transplant risk stratification and provide candidate biomarkers for future validation and therapeutic investigation.
DISCLOSURES
Xue-Zhong Yu Sr, MD, MSc: No relevant conflicts of interest to declare.