Immune Profiling to Predict Treatment Response from Extracorporeal Photopheresis in Graft-Versus-Host Disease

Graft-versus-host disease (GVHD) is a major cause of morbidity and mortality in patients receiving stem cell transplant. Extracorporeal photopheresis has demonstrated efficacy in a proportion of GVHD patients that are steroid refractory. However, given the expense, the challenging logistics, and unpredictable GVHD response to ECP, a predictive immune biomarker to assist with the selection of patients to treat with ECP can make this treatment modality more efficacious and cost-effective. One of the challenges of using leukocyte as biomarkers is that any single leukocyte population does not adequately describe the immune system. We have recently described an approach to characterize the immune system in a comprehensive manner. We have used a combination of whole blood quantitative flow cytometry and bioinformatics to generate immune profiles among patients with GVHD who were prescribed ECP. GVHD grading and clinical responses were confirmed by serial physical exams and review of medical records over a 6 month period. Response to ECP was defined as a decrease in clinical grade of GVHD (IBMTR and NIH consensus criteria) at the end of treatment. Peripheral blood drawn prior to initiation of ECP was used to generate a patient's immune phenotype by analyzing 33 lymphoid and myeloid surface marker combinations by flow cytometry. Immune phenotypes from 27 age-matched healthy volunteers were used as controls.

From July 2011 to April 2013, 20 patients were enrolled (3 aGVHD, 17 cGVHD). At the time of this abstract, 15 patients were evaluable for treatment response, with 6 responders and 9 non-responders (median time on treatment 5.28 months, range 0.25 – 6 months). Using multivariate analysis by principal component analysis and unsupervised hierarchical clustering, we identified clustering patterns (immune profiles) in the composite immune phenotypes of the GVHD patients and controls. The healthy controls clustered together into a single immune profile. Four of the six responders clustered closely together and away from the 9 non-responders and healthy controls (Fig 1). Anova analysis identified 21 individual markers that were different among the clusters. Using non-parametric Wilcoxon comparison, we found that patients who responded to ECP had lower numbers of lymphocytes (p=2.5x10^-12), CD4 T cells (p=0.002), CD4⁺CD69⁺ T cells (p=0.02), and CD8⁺CTLA4⁺ (p=1.8x10^-5) than healthy controls and non-responders.

To our knowledge, this is the first report of a bioinformatics approach to comprehensively examine systemic immune profiles of GVHD patients. Our results provide the proof-of-concept for using this method to identify an immune profile as a predictive biomarker in the treatment of GVHD. Our data supports the need for a larger study to definitively define the immune profiles and validate this approach.

Figure 1. Immune profiles of healthy controls and GVHD patients.