Transplant biomarkers ready for the clinic?

Biomarkers promise to refine the prediction of allogeneic stem cell transplantation (SCT) outcomes. In this issue of Blood, a phase 3 clinical trial reported by Abu Zaid et al brings us closer to routine biological profiling of major complications that occur after allogeneic SCT.¹ 

In the last few years, pioneering studies, notably by Paczesny and colleagues, have discovered a range of molecules that can be assayed in plasma, which has proven to be strongly related to some key transplant complications that define transplant survival.²  To reach this point basic research has been required to identify potential biomarkers in an unbiased manner, together with rigorous correlative statistics to find strong and true associations between biomarker variations and posttransplant events, such as the overall outcome (survival or death) and specific complications, such as graft-versus-host disease (GVHD). Initially, biomarker research focused on acute GVHD. First, an unsupervised and giant array of markers was pared down to a shortlist of a dozen or so candidates strongly consistent with acute GVHD outcome. Subsequently, the most reliable and strongly correlative markers were selected. Validation of candidate biomarkers required correlative analysis of biomarkers with large groups of patients developing training sets and subsequent validation sets. Initial trials were performed within a single institute.³  Over time, new and better biomarkers were discovered, and further studies have validated some key biomarkers: suppression of tumorigenicity-2 (ST2) as a predictor for acute GVHD, and in particular, steroid refractory GVHD and nonrelapse mortality (NRM),⁴  Reg3α as a predictor of gastrointestinal GVHD,⁵  chemokine (C-X-C motif) ligand 9 (CXCL9), associated with chronic GVHD,⁶  and L-ficolin in association with sinusoidal obstruction syndrome/veno-occlusive disease (SOS/VOD).⁷ 

Although many factors are already known to determine transplant outcome (age, comorbidity, donor-recipient compatibility, and investigator-determined issues, such as choice of conditioning regimen and posttransplant GVHD prophylaxis), they are incomplete guides for predicting outcomes. Biomarkers promise to further refine our ability to determine the likely outcome of the individual patient, making possible the prevention of complications by individualizing the treatment approach.²  For this to become a reality, it has to be demonstrated that biomarkers are robust and sure predictors of outcomes across a variety of transplant approaches performed in any center (see figure). In this issue, Paczesny’s group has moved the field closer to this goal. Collaborating with the Blood and Marrow Transplant Clinical Trials Network, they prospectively measured a set of previously validated biomarkers in patients participating in a multicenter trial comparing GVHD prophylaxis with tacrolimus and sirolimus vs tacrolimus and methotrexate. The study involved 304 patients transplanted in 23 US transplant centers.⁸  Blood samples for biomarker analysis were collected at fixed time points between days 28 and 365 after SCT in 211 individuals. Critically, the trial found no significant difference in GVHD between the 2 study arms, although there was slightly more SOS/VOD in the tacrolimus/sirolimus set. After multivariate analysis in a proportional hazards model with time-dependent coordinates, they identified 4 biomarkers associated with outcome high day 28 ST2 and Tim3 correlating with NRM and survival, low L-ficolin correlating with SOS/VOD, and high CXCL9 correlating with chronic GVHD.

The predictive value of a biomarker is dependent upon the quality of the statistical analysis, the reliability of the clinical readout, and above all, the relevance of the biomarker to the biological process underlying the clinical event. This study followed the required norms of statistical evaluation involving a sufficiently large, prospectively studied, patient cohort. Multivariate analysis and allocation of proportional hazards ensured the identification of independent determinants of outcome. The study was conducted under the rigor of a well-organized clinical trial. The biological relevance of the winning markers is also a strength, and the mechanisms are discussed in the paper: ST2 in plasma is the soluble form of the interleukin-33 (IL-33) receptor, acting as a decoy receptor for IL-33 and preventing the binding of IL-33 to T cells. The observation by this group that an antibody to ST2 prevents GVHD in a mouse model strongly links this molecule with the GVHD process.⁹  Tim-3 is present on activated T cells as well as in a soluble decoy form. Plasma Tim3 can limit the interaction of cellular Tim3 and its ligand, blocking its regulatory role in cytotoxic T-cell function. The role of CXCL9 as a gatekeeper for tissue distribution of alloreactive T cells in chronic GVHD is supported by the high levels of CXCL9 seen in oral, ocular, and mucosal chronic GVHD. Low levels of L-ficolin correlate with diminished hepatic clearance of mitochondria. However, the relationship of L-ficolin with the underlying mechanism of SOS/VOD is unclear.

The design and execution of this study incorporated several limitations. First, not all the patients in the clinical trial had biomarkers measured. Furthermore, the choice of the day 28 collection of the first sample for analysis eliminated any possibility of exploring acute GVHD prediction, because a third of the patients had already developed GVHD by this time. Although the results were validated across different GVHD prophylaxis protocols, the study groups were otherwise uniformly treated. It has yet to be determined if biomarker predictions can span SCT given for a variety of disease indications with diverse donor types and different strengths of conditioning regimens. Furthermore, no study has yet combined NRM predictors with relapse predictors (notably sensitive molecular analysis of residual disease) to provide a comprehensive biological profiling of all the determinants of posttransplant disease-free survival where the indication is for malignant disease.

How will biomarkers change the way we do transplants in the future? Apollo gave Cassandra the unenviable gift of divining the future, with the proviso that nobody would believe her. To avoid Cassandra’s fate, biomarker prediction must be proven to work in diverse transplant conditions, before it can achieve general acceptance. Furthermore, without the ability to modify outcomes, the precise determination of fate is neither a gift to the patient nor to their physician. How much we can control the destiny of the transplant will depend upon the outcome of further trials where biomarkers are used to make decisions between treatments designed to avoid, for example, acute or chronic GVHD or SOS/VOD. There is still a long way to go, but this paper is a sound basis for new trials designed to further extend biomarkers to larger and more diverse transplant populations and to explore ways to modify predicted outcomes by individually directed treatment approaches.