Prognostic Value Of Plasma Biomarkers Compared To Urinary Proteome Patterns

Graft-versus-host disease (GvHD) is a major complication and cause of morbidity and mortality after hematopoietic stem cell transplantation (HSCT). Prediction of acute GvHD is essential to optimize treatment. An effective method for early investigator independent diagnosis of aGvHD is under investigation in a phase III clinical trial. A urinary proteomic pattern – aGvHD-MS17 – is evaluated in a preemptive treatment trial, administering steroids at the time of pattern positivity. Here we present our data for early diagnosis of GvHD by monitoring easily accessible samples like urine for the presence of a specific proteom pattern and compare the data to published plasma biomarkers.

Plasma and urine were collected from patients after HSCT at defined time points (Weissinger et al., Blood 2007). Capillary electrophoresis and mass spectometry (CE-MS) analyses were done according to Weissinger et al., Blood 2007 and Leukemia, 2013. Two peptides found to be highly specific in urine for development of aGvHD were β2-microglobulin (β2m) and CD99. For both, antibodies are available and thus they can be analyzed in plasma samples from the same patients with ready-to-use ELISA-kits (Pasnecey et al., 2010). Plasma biomarkers were first analyzed in a test set based on 34 patients, followed by validation in 31 patients. As a further contribution to harmonization, published plasma biomarkers elafin, soluble IL-2-receptor alpha (IL-2sRα), REG3α, ST2 (IL-1 receptor 4) and sTNFRI were added to the ELISA-panel. Using the angiogenesis panel for bioluminex analyses, IL-8 and hepatocyte growth factor (HGF) were analyzed in addition and the outcomes were compared to the urinary proteomic analyses.

While the urine derived CE-MS pattern was able to predict an upcoming GvHD prior to the development of clinical signs, most of the plasma biomarkers could not predict pending GvHD. The test set for analysis of plasma consisted of 20 patients with aGvHD, from whom samples from the time of clinical diagnosis of aGvHD were used, and 14 controls without aGvHD at any time. Receiver operated characteristic curves were generated, areas under the curve (AUC) were more than 80% for β2m, CD99, sIL2Rα and sTNFRI (p<0.01). The validation set consisted of samples from 31 patients at different time points (7 samples each patient). In the validation set some markers did not reach their established cut-off point for aGvHD-diagnosis prior to or at the day of clinical diagnosis of aGvHD. Especially REG3α and elafin did not turn positive in any samples of patients with acute GvHD. Both markers presumably are organ-specific, but also limiting the analysis to patients with isolated GvHD of the intestine (REG3α) or skin (elafin) did not increase the sensitivity of these markers. CD99 and β2m discriminated patients with aGvHD from controls and CD99 was even specific for chronic GvHD and could distinguish between patients with acute and chronic GvHD with high sensitivity (89%) and specificity (95%); AUC was 86% (p<0.001). Additional problems with the ELISA-technique were the need to dilute samples prior to testing due to highly diverse concentrations of the biomarkers. The concentrations sometimes varied by the factor of 10 or more which required preparation of each sample in different dilutions. This made diagnosis based on ELISA data more time consuming and costly as envisioned. Urine on the other hand has shown to be stable and to provide a reliable proteome pattern. Furthermore, urinary proteom analysis can be normalized within each sample by using internal standards.

Prediction of pending aGvHD in patients post-allogeneic HSCT is possible with urine monitoring using CE-MS analyses with a sensitivity of 82% and specificity of 77% for aGvHD (Weissinger et al., Leukemia 2013). In contrast, published plasma biomarkers in our hands lack reliable sensitivity and specificity in the validation set. In addition, the markers detected in plasma do not appear to predict development of aGvHD. Biomarkers identified in urine (CD99 and β2m) can be found in plasma as well and are predictive for aGvHD-development. CD99 is also predictive for cGvHD development which can be explained by its function as an activation marker of (donor) T-cells. In conclusion, our results demonstrate that plasma monitoring post-allogeneic HSCT seems less reliable than urinary proteomics for prediction of aGvHD.

No relevant conflicts of interest to declare.