Prognostic plasma biomarkers of early complications and graft‐versus‐host disease in patients undergoing allogeneic hematopoietic stem cell transplantation

Abstract Early complications post hematopoietic stem cell transplantation (HSCT) such as sinusoidal obstruction syndrome (SOS) and graft versus host disease (GVHD) can be life threatening. Although several biomarkers have been identified to correlate with these complications and their response to treatment, these are yet to be used in clinical practice. Here, we evaluated circulating endothelial cells (CECs) (n = 26) and plasma biomarkers (ST2, REG3α, VCAM1, ICAM1, TIM3) (N = 210) at early time points, to determine their association with early complications post‐HSCT. Elevated CEC counts at the end of conditioning was associated with GVHD, indicating endothelial damage during HSCT. Plasma levels of REG3α, VCAM1, ICAM1, and TIM3 on day 14 (D14) and D14 ICAM1 and D28 ST2 were significantly higher in patients with SOS and aGVHD, respectively. Upon sub‐group analysis, D28 ST2, D14/D28 REG3α, and D14 ICAM1 levels were significantly higher in patients with gastrointestinal GVHD, while D28 ST2 was higher in those with skin/liver GVHD. High ST2 levels on D28 was significantly associated with non‐relapse mortality (NRM) and overall survival. Our results suggest that elevated ST2 levels on D28 could predict the likelihood of developing aGVHD and could influence NRM and OS.

Minimally invasive tests such as plasma or cellular biomarkers could be useful in the diagnosis, prognosis, and response to therapy of these complications.
Biomarkers could be either: (a) diagnostic biomarker (that identifies patients at the onset of clinical disease); (b) prognostic (that identifies the likelihood of a clinical event occurrence in HCT recipients), or (c) predictive (that categorizes patients by their likelihood of response to a particular treatment when measured prior to the treatment) [2].
Several groups have used proteomics strategy to discover candidate plasma biomarkers that were found to be associated with:

Clinical Endpoints
Neutrophil recovery was defined as an absolute neutrophil count ≥0.5 × 10^9/L for three consecutive days while platelet engraftment was defined as platelet count >20,000/mm 3 without platelet transfusion for 7 days. Grading and staging of acute and chronic GVHD was done according to standard CIBMTR criteria [20]. Sinusoidal obstruction syndrome (SOS) also called veno-occlusive disease (VOD) was defined as per McDonald's criteria [21].

Differences in biomarker levels at different time points between
GVHD-and GVHD+ patients and SOS+ and SOS-patients were assessed using Mann-Whitney U test. Receiver operating characteristic (ROC) curves and area under the curves (AUCs) were estimated non-parametrically. A P-value of <0.05 was considered statistically significant. Differences in cumulative incidences of non-relapse mortality (NRM) and SOS/VOD were calculated by log rank test. Overall survival (OS) was estimated by Kaplan Meier method and the difference between groups were calculated by log rank test. Cumulative incidence analysis was used to describe the association between NRM and biomarkers with relapse as a competing risk. All statistical analysis was carried out using GraphPad Prism version 6.0 (San Diego, CA) and IBM SPSS statistics 24.0 (Armonk, NY). Cumulative analysis was done with R software (version 3.6.1)

Patient demographics
The median age of patients in which CEC was enumerated was 15 years (range 2-53; 17 males and 9 females; n = 26). Plasma biomarkers were  Table 1. The outline of the study with number of patients in each analysis is summarized in Figure 1.

Kinetics of CEC levels
Although our original plan was to enumerate CECs at various time points during and after conditioning and post-HSCT, due to technical challenges such as low blood counts and the inability to acquire 1 × 10 6 events, CECs defined as Hoechst  Table 2). An overall comparison of CEC numbers at each time point between myeloablative and reduced intensity conditioning (RIC)/non-myeloablative regimens indicated no significant difference between CECs counts at preconditioning, end of conditioning, and D14. However, at D28 the CEC counts of RIC/non-myeloablative regimens were significantly elevated (P = .015) compared to myeloablative conditioning regimen ( Figure 2C and Table 2).
When CECs counts at each time points were compared with transplant outcomes, there was elevated CEC counts at the end of the conditioning in patients who developed GVHD compared to those without GVHD, although not statistically significant (P = .14). When D28 REG3 compared to those without GVHD ( Figure 3C). Based on these results, we analyzed D28 ST2, D28 REG3 , and D14 ICAM1 in all patients for their potential prognostic value towards aGVHD.

Plasma biomarkers (REG3 , VCAM1, ICAM1, and TIM3) on day 14 associated with SOS/VOD
Among 210 patients, plasma samples collected on D14 were available only for 157 patients. The levels of REG3 , VCAM1, ICAM1, and TIM3 were significantly elevated at D14 in patients who developed SOS compared to those who did not develop SOS ( Figure 4A-E). ROC curve analysis for each biomarker at D14 for SOS resulted in an AUC of more than 0.68 ( Figure S1). Cut-off values were derived from this analysis, which were further used to stratify samples from patients as those with high or low biomarker values. in patients with low ST2 ( Figure 4F). The cumulative incidence of SOS was 12% in patients with high TIM3, compared to 4% in patients with low TIM3, although not reaching statistical significance ( Figure 4G).
For patients with high ICAM1, the cumulative incidence of SOS was F I G U R E 3 Kinetics of plasma biomarkers (ST2, REG3 , TIM3, ICAM1, and VCAM1) in an initial cohort (n = 56) to evaluate the trend in their levels with respect to clinical outcomes. A, kinetic until day 14 for all five biomarkers for patients with VOD (n = 5) and without VOD (n = 51). B, kinetic until day 28 (for ST2 and REG3 ) or until day14 (for TIM3, ICAM1, and VCAM1) for patients with acute GVHD (n = 19) and without acute GVHD (n = 37). C, kinetic until day 28 (for ST2 and REG3 ) or until day14 (for TIM3, ICAM1, and VCAM1) for patients with gastrointestinal GVHD (n = 9) and without acute GVHD (n = 37). Mann-Whitney U test was used to compare the levels of biomarkers at a specific time point in patients with VOD or without VOD, with aGVHD or without aGVHD and with gastrointestinal GVHD or without aGVHD 12.3%, whereas for patients with low ICAM1, it was 2.4% ( Figure 4H).
Similarly, for patients with high VCAM1, the cumulative incidence of SOS was 11%, whereas for patients with low VCAM it was 2.7% ( Figure 4I)

Plasma biomarker (D28 ST2 and D14 ICAM1) levels associated with acute GVHD
D14 ICAM1 and D28 ST2 levels were significantly elevated in patients who developed acute GVHD (all grades) when compared to those without GVHD ( Figure 5). The ROC analysis also demonstrated significant association of D14 ICAM1 and D28 ST2 with acute GVHD (AUC of 0.6 and 0.65, respectively; Figure S2). Table 4 shows the median, 25th, and 75th percentiles of biomarker concentrations of D14 ICAM1 and D28 ST2 in acute GVHD+ and acute GVHD-groups with corresponding AUC, cut-off values, and P-values. These analyses were done including samples from patients with the day of GVHD onset as >14 for ICAM1 and >28 for ST2.
Upon sub-analysis, D28 ST2, D14 and D28 REG3 , and D14 ICAM1 levels were significantly elevated in patients who developed gastrointestinal GVHD versus those who did not (Figures 6A-D). The ROC analysis also demonstrated a significant association of these biomarkers at D14 with gastrointestinal GVHD with an AUC of 0.7 ( Figure S3).
D28 ST2 levels were also significantly elevated in patients who developed liver or skin GVHD versus those who did not develop GVHD (Figure 6E,F). ROC curve analysis of D28 ST2 for liver and skin GVHD resulted in an AUC of more than 0.68 ( Figure S4).

Plasma biomarker levels associated with non-relapse mortality and overall survival
We then analyzed whether or not these biomarkers predict nonrelapse mortality (NRM) or influence overall survival (OS). The incidence of NRM was 34.2% in patients with high ST2 compared with 11% for those with low ST2 values on D28. This was significant when cumulative incidence analysis was done to describe association between D28 ST2 and NRM with relapse as a competing risk (Gray's test P = .001, Figure 7A). The causes of death are included in Table S3. Since D28 ST2 was significantly associated with acute GVHD, we evaluated the role of these biomarkers on 4-year OS based on the cut-off values.
The 4-year OS was 63.2% in patients with high ST2 compared to 85.6% for those with low ST2 (P = .001, Figure 7B).

DISCUSSION
Early transplant related complications such as GVHD and RRT influences the success of HSCT, which is the only currently available cure for various hematological disorders. Biomarkers have been evaluated by various groups during and post HSCT as diagnostic/prognostic tool for RRTs and GVHD. [18,8,9,22,2] However, these biomarkers are still  observed an approximately four fold increase in median CEC counts in patients who developed GVHD compared to those without GVHD at the end of the conditioning. This is in contrast with previous reports, [8,11] where increased CEC levels were reported in patients without GVHD compared to those with GVHD at the end of conditioning and at engraftment. This could be explained by the different methodology used (CellSearch System) and the immunophenotypic definition of CECs defined as CD146 + /CD105 + /DAPI + /CD45 − . Furthermore, CECs could be influenced by multiple factors such as conditioning regimen [9,23], GVHD prophylaxis [8,9], immunosuppressive treatments [24,25], and infections [8]. However, our observation of elevated CEC levels at the end of conditioning in patients who developed GVHD was not statistically significant, probably due to the small numbers evaluated. A potential limitation of enumerating CECs is the relatively low numbers of blood cells in circulation post-conditioning that makes acquisition of rare CECs cumbersome. Also, lack of consensus on the immunophenotypic definition of CECs and lack of standardized methodology to enumerate them, makes validation of CECs as biomarkers across different centers challenging.
We observed all five plasma biomarkers that we measured to be significantly elevated on D14 in patients with SOS. While the potential prognostic significance of ST2 [18], VCAM1 [18], and ICAM1 [26] toward SOS is consistent with previous reports, association of two .02

F I G U R E 6
Prognostic value of plasma levels of ST2, Reg3 , and at ICAM1 toward organ specific GVHD. A, Reg3 concentrations at D14 in patients with gut GVHD (n = 29) and without GVHD (n = 106). B, Receiver operator characteristic (ROC) curve for Reg3 at D14 comparing patients with and without gut GVHD. E, ST2 concentrations at D28 in patients with liver GVHD (n = 16) and without GVHD (n = 111). F, ST2 concentrations at D28 in patients with skin GVHD (n = 26) and without GVHD (n = 111). Mann-Whitney U test was used to compare the levels of biomarkers in patients with gut/liver/skin GVHD and without GVHD F I G U R E 7 High ST2 levels were correlated with NRM and OS. The cumulative incidence of NRM by 48 months (4 years) stratified by: A, day 28 ST2 levels (high vs low, median cutoff of 53 ng/mL); B, Kaplan-Meier curve for OS stratified by day 28 ST2 levels (high vs low, cutoff of 53 ng/mL) NRM. Thus, the predictive values of these biomarkers could potentially aid clinicians in assessing and managing HSCT complications effectively.
To the best of our knowledge, this is the first report where we have evaluated biomarkers comprehensively in HSCT patients and identified biomarkers that can predict early complications arising due to HSCT. While we derived cut-off values for these biomarkers to risk stratify patients for HSCT complications, this need to be validated in an independent cohort or in large multicenter trials for future clinical applications. Moreover, based on the cut-off values, it is possible to define a high-risk population that would benefit from a pre-emptive intervention that will need still to be defined (eg, cortecosteroids at 1 mg/kg/day for 7-10 days with rapid taper). The major limitation of this study is the significant overlap in biomarker levels between groups with and without HSCT related complications. Further, many groups have reported several cut-off values for ST2 (33.9 ng/mL [18]; 740pg/mL [29]; 3230 ng/mL [30]) and REG3 (151 ng/mL [19]; 1989 pg/mL [30]) which makes establishing reproducible cut-off values for biomarkers a challenge. We also believe evaluating the biomarkers prospectively at this and earlier time points will better validate the association of these biomarkers with outcomes and could potentially explain their predictive values.