Statin‐based endothelial prophylaxis and outcome after allogeneic stem cell transplantation

Allogeneic haematopoietic stem cell transplantation (alloSCT) often remains the only curative therapy for hematologic malignancies. Although the management of transplant‐associated adverse events considerably improved over the last decades, nonrelapse mortality (NRM) remains a challenge, and endothelial dysfunction was identified as a major contributor to NRM.


| INTRODUCTION
For many malignant and nonmalignant lymphohaematopoietic diseases, allogeneic haematopoietic stem cell transplantation (alloSCT) remains the only curative therapeutic approach. However, the considerable risk of nonrelapse mortality (NRM) remains a major challenge especially for elderly and extensively pre-treated patients.
We and others have shown that endothelial dysfunction plays an important contributing role in several severe complications of alloSCT, in particular in sinusoidal obstruction syndrome/veno-occlusive disease (SOS/ VOD), transplant-associated thrombotic microangiopathy (TA-TMA), and refractory acute graft-versus-host disease (aGVHD). [1][2][3][4] Several efforts have been made to identify markers that predict which patients will likely suffer such complications with the aim to develop riskadapted management strategies. 2,[5][6][7] Serum endothelial biomarkers such as suppression of tumorigenicity 2 (ST2) have been shown to be associated with a high risk of TA-TMA. 8,9 Apart from single biomarkers measured in serum at specific time points post-alloSCT, we established the endothelial activation and stress index (EASIX) based on three basic laboratory parameters that characterize TA-TMA (platelet counts, LDH, and creatinine serum levels). EASIX indicates the risk for endothelial complications post-alloSCT, [10][11][12] but it also appears to estimate endothelial risks independent of transplantation, that is for patients with lower risk MDS, 13 Multiple Myeloma, 14 CAR-T therapy 15 and COVID-19. 16 Statins inhibit the enzyme βhydroxy-βmethylglutaryl-CoA (HMG-CoA) reductase thus lowering the levels of serum cholesterol 17 and are therefore widely used as primary and secondary prophylactic treatment to prevent cardiovascular events. 17 However, there is increasing evidence for pleiotropic effects of statins apart from cholesterol reduction such as immune system modulation. 18 In asthma patients, it was found that statins reduce the serum levels of indoleamine 2,3-dioxygenase (IDO), 19 the rate-limiting enzyme in the catabolism of tryptophan (Trp) to kynurenine (Kyn). 20,21 Statins have been assessed in alloSCT in different settings. 22 Two studies observed reduced higher-grade aGVHD, 23,24 and one study less extensive cGVHD, 24 and no adverse events were observed. 22 However, in another study atorvastatin did not reduce the incidence of aGVHD nor cGVHD, 25 and none of the studies stratified patients according to baseline endothelial risk. Having identified endothelial dysfunction as a major player in NRM post-alloSCT, we changed our in-house standard of care in 2010 by putting all patients undergoing alloSCT on pravastatin and UDCA. Here we retrospectively analysed whether this pharmacological endothelial protection improves outcome of alloSCT patients and whether the effects of statins might be reflected by post-transplantation kinetics of serum markers of metabolic, inflammatory, or endothelial activity.

| Patient cohort selection
We collected data from 1200 transplant patients between 2002 and 2022 ( Figure 1A). Upon changing our in-house policies in January 2010, all patients received pravastatin 20 mg daily in order to reduce calcineurin inhibitor (CNI)associated cardiovascular endothelial damage. This was further amended in 05/2010 by adding 3 × 250 mg ursodeoxycholic acid (UDCA) daily, an FDA-approved drug, which was shown to reduce severe aGVHD, VOD, and elevated bilirubin levels in the context of alloSCT. 26,27 Since it must be assumed that during the total time span between 2002 and 2022 apart from SEP also other transplant modalities most likely changed, a narrower period comprising 2 years before and after the introduction of routine SEP was selected for the purposes of this study to minimize bias introduced by continuous improvement of alloSCT technology, expecting other factors to be stable in this 4-year observation period. Accordingly, the study cohort included 347 patients, 138 patients without statins, and 209 patients on statin treatment (137 with concomitant UDCA treatment) during and post-alloSCT ( Figure 1A).
Written informed consent to prospective blood sample and retrospective data collection was obtained from all patients, and the study was approved by the Institutional Review Board according to the Declaration of Helsinki.

| Data collection and quantification of biomarker serum levels
aGvHD was graded using standard criteria based on clinical and histological findings. 28 TMA was diagnosed according to international consensus criteria. 29,30 EASIX was calculated as described before. 10 Serum levels of the following biomarkers were measured by ELISA: ST2, CXCL9, IFNg, IL18 (R&D Systems Europe Ltd), IDO, and Trp (Immunodiagnostics). Serum levels of Kyn were measured by liquid chromatographytandem mass spectrometry (LC-MSMS). A method for simultaneous quantification of all metabolites in a single run was used as described. 31 Serum from patients was collected before and on day 100 post-alloSCT and frozen at −80°C at the biobank of Medical Department V at Heidelberg University Hospital until usage. F I G U R E 1 Cohort selection and group characteristics. Timeline of data collection including 1200 alloSCT patients between 2002 and 2022. For analysing the impact of statins, the 347 patients transplanted within 2 years before and 2 years after in-house policy introduced statin-based endothelial prophylaxis (SEP) were used (A); distributions of disease entity, conditioning regimen, donor type, and disease score are shown in the group of patients without statins (left) or on statins (right). p-values were calculated by Fisher's exact or chi-square test (multiple groups). Numbers in pie charts indicate numbers of patients (B); age (left) and EASIX (log2-transformed, right, median indicated as anti-log values) in patients with (n = 209) versus without (n = 138) statins. Mann-Whitney U-test, box plots showing medians, quartiles and outliers (Tukey) (C); serum levels of ST2, IL18, and angiopoietin-2 (Ang2) calculated prior to alloSCT ("pre") in patients with versus without statins.

| Statistical analysis
Categorical variables were compared using Fisher's exact or chi-squared tests, whereas Mann-Whitney U-tests were performed to compare continuous variables in the groups defined by SEP (yes/no).
Multivariable proportional hazard Cox regression was applied to estimate the effect of SEP on overall survival (OS). Multivariable cause-specific proportional hazard Cox regression analyses were applied to model the effect of SEP on the hazard of NRM after alloSCT, non-TAM mortality, non-aGVHD mortality, and non-cGVHD mortality. Additional covariables in multivariable models were patient age (in change of 10 years), recipient gender, donor gender, recipient-donor match, diagnosis, conditioning intensity, disease score and EASIX-pre (log2). Cumulative incidence curves were used to illustrate associations.
The effect of SEP on (log2-transformed) cytokine levels at day 100 was assessed using multivariable linear regression using the same covariables as used for survival analyses.
Cytokine levels in the SEP groups were visualized using boxplots.
All calculations were performed using R version 4.0.4 (R Foundation for Statistical Computing, 2021). Cohort characteristics were visualized using Graph Pad Prism version 9.0.1. Reporting of the study conforms to broad EQUATOR guidelines. 32

| Patient characteristics
The distribution of disease entities, donor types, and a disease score reflecting the overall risk of a patient was not significantly different in the two groups, although there was a tendency toward more mismatches and higher disease score in the "no-statins" group ( Figure 1B, two-sided Fisher's exact test or chi-square test [multiple groups], full data in Table S1). There were more patients receiving an aplasia conditioning regimen and slightly less with myeloablative conditioning (MAC) in the statin group ( Figure 1B, p = .004, chi-square test). Patient age ranged from 19 to 71 years and was similar between the groups ( Figure 1C, Table S1). EASIX pre-transplant (EASIX-pre) was significantly higher in the SEP group ( Figure 1C, p < .001, two-sided MWU test) and was thus included as a covariable in the Cox models. Notably, significant differences in the pre-transplant serum levels of ST2, IL18, and ANG2 ( Figure 1D), known to be associated with severe complications or endothelial activation post-alloSCT 8,33,34 were not observed.

| SEP prophylaxis and outcomes
Tables 1 and 2 illustrate the results of the Cox models with the endpoint overall mortality and post-aGVHD mortality, respectively. SEP was associated with a reduced hazard of death for both endpoints, which was statistically significant only for aGVHD-related mortality (HR 0.59, 95% CI 0.37-0.93, p = .023). Other variables associated with a significantly increased hazard for both mortality endpoints were higher disease score and male recipients.
In a cause-specific proportional hazard Cox model with competing endpoints NRM and relapse, relapse risk was not significantly associated with SEP (Table S2), whereas the hazard of NRM was significantly reduced in the SEP group (Table 3 HR 0.61, 95% CI 0.38-0.96, p = .034). When applying competing risk Cox regression to the subpopulation of patients after aGVHD, the association of SEP with NRM risk was even stronger (Table S3, HR 0.46, 95% CI 0.24-0.89, p = .021). No significant association between SEP and acute and chronic GVHD incidence was observed (Tables 4A,B).

| SEP effects in endothelial risk groups
As SEP is supposed to prevent endothelial damage, we hypothesized that patients with highest endothelial risk by EASIX-pre would particularly benefit from SEP. To explore this hypothesis, separate competing risk Cox regressions were applied to patients with low (quartile 1, <0.76), intermediate (quartiles 2-3, 0.77-2.69), and high EASIX-pre (quartile 4, >2.69) for NRM for all patients post-alloSCT and post-aGVHD. EASIX-pre values were not significantly different between patients with or without SEP in the three groups ( Figure S1), ruling out that the groups differed in endothelial risk already prior to transplantation. The influence of SEP seemed strongest in the intermediate EASIX-pre group, both when considering all patients and patients after aGVHD (Figure 2A,B). After multivariable adjustment for confounders, SEP was not significantly associated with NRM in patients with low (HR 1.22, 95% CI 0.37-4.00, p = .743) and high EASIX-pre scores (HR 0.90, 95% CI 0.33-2.54, p = .084) (Tables S5 and S6). In contrast, patients with intermediate EASIX-pre (quartiles 2 and 3) seemed to benefit from SEP with a HR of 0.47 (95% CI 0.25-0.92) for NRM, p = .027 (Table 4).
To assess whether UDCA had any synergistic effect with statins we analysed the cumulative incidence of NRM post-alloSCT for statins given as monotherapy and statins in combination with UDCA separately. There was no visible difference between statin monotherapy and the combination with UDCA ( Figure 2C). Similarly, there was no UDCA-mediated augmentation of the protective effect of statins on the TA-TMA cumulative incidence ( Figure S2).

| SEP, IDO activity and tryptophan metabolism
To gain insights into the potential mechanism of action underlying the beneficial effects of SEP on NRM, we determined serum levels of a series of cytokines, enzymes, and metabolites linked to either immune activation or TA-TMA, in all patients, as well as separately for patients with low, intermediate, and high EASIX-pre. Regarding SEP-modulated immune activation post-alloSCT, differences in serum levels of IL18, which is increased in severe chronic GVHD (cGVHD), 33,35 and in interferon-gamma were not observed ( Figure 3A). In contrast, serum ST2 levels tended to be lower in the SEP group when measured on day 100 (p = .055) ( Figure 3A). For CXCL9, we observed significantly higher serum levels in patients on SEP, in particular in those with intermediate (p = .002) and high (p = .02) EASIX-pre ( Figure 3A). We have shown before that only CXCL9, but not IL18 is strongly associated with the activity of the rate-limiting enzyme in tryptophan metabolism IDO. 24 We therefore  To test this hypothesis, we measured the enzyme IDO, as well as tryptophan and its metabolite kynurenine in patients from whom serum was available on day 100 post-alloSCT. As the ratio between these two (Kyn/Trp ratio) is considered to reflect IDO activity better than individual serum levels, we also determined the Kyn/Trp ratio in patients for whom both parameters were available ( Figure 3B). In line with our hypothesis, we observed   higher IDO levels in patients on statins ( Figure 3B,C, median IDO concentration 0.08 vs. 0.22 nM). Tryptophan levels were significantly lower in patients with statins (median 59 vs. 46 nM), while kynurenine levels were not different, resulting in a significantly higher Kyn/Trp ratio in patients on statins (ratio 8 vs. 14). When evaluating the differences in our multivariable model, the association between statins, high IDO activity, and lower tryptophan levels was again significant only in the intermediate EASIX group (Figure 3B).

| DISCUSSION
Continuous efforts have been made to reduce NRM post-alloSCT to render this potentially curative therapy amenable to an increasing number of patients with hematologic malignancies. 36 This is reflected by the increasing proportion of elderly patients at transplantation over the past decades. 36 Nonetheless, NRM post-alloSCT remains higher compared with other treatments and therefore warrants further research. Here, we report that statin-based endothelial protection during and post-alloSCT significantly reduces the hazard of NRM and OS. 10 The reduction in NRM was not due to a lower incidence of aGVHD, which is in line with previous studies that showed no effect of atorvastatin on the incidence of aGVHD nor cGVHD. 25 Of note, atorvastatin was highly efficient in delaying GVHD onset in a mouse model mimicking allogeneic transplantation, 37 which indicates that murine GVHD models might not reflect the complexity of the human alloSCT setting. While statins had no effect on aGVHD onset, we found that NRM was more strongly reduced in statintreated patients who suffered aGVHD compared to those without aGVHD. There is broad evidence that aGVHD is accompanied by extensive endothelial damage and dyfunction [38][39][40] and hence often co-occurs with TA-TMA. 41 This might explain why patients with aGVHD benefitted more from statin treatment than those who never suffered aGVHD. Furthermore, this observation suggested that patients with low endothelial risk might not necessarily benefit from endothelial prophylaxis. We therefore grouped patients according to endothelial risk prior to transplantation using EASIX and found the strongest effects in patients with an intermediate score. The lack of benefit of SEP in the high-risk group suggests that for patients in whom endothelial damage is already manifest as indicated by a high EASIX-pre score this damage cannot be reversed by pharmacological endothelial protection. Patients with a low EASIX-pre most likely do not have any apparent endothelial problem with or without statins, and therefore, the beneficial effect of SEP is less prominent (e.g., trend for lower NRM post-aGVHD in low EASIX-pre patients). These observations emphasize that the effects of endothelial prophylaxis might be highly dependent on risk factors. Studies aiming at the reduction of endothelial toxicity or of endothelial complications should therefore stratify endothelial risk cohorts separately. This might also explain discrepant statin effects observed in different studies in the context of kidney transplantation 42 and al-loSCT. 22 Our data suggest that estimation of endothelial risks with EASIX pre-alloSCT might help distinguish patients who will benefit from SEP and patients who might need additional endothelial protection.
With the aim to get insights into potential mechanisms by which statins lower the hazard of NRM post-alloSCT, we measured cytokines and metabolites in the serum of patients 100 days post-transplantation. As pleiotropic effects described for statins included also modulation of the immune system, we hypothesized that statins might either affect serum concentrations of activating cytokines such as those associated with cGVHD (CXCL9, IL18), 33,35 or metabolites associated with IDO activity, which was found to be lower in asthma or COPD patients. 19,43 While F I G U R E 3 Immune modulatory cytokines and tryptophan metabolism. Left: Forest plot showing the log2-fold changes of effect size and 95% confidence intervals (95% CI) derived from multivariate analyses to assess the impact of statins on cytokine levels for ST2, CXCL9, IL18, and interferon gamma including statins and the following covariables: age (10 years), recipient gender, donor gender, donor type, disease entity, conditioning regimen, disease score, and EASIX-pre. The analysis was performed for all EASIX groups ("all"), as well as for low, intermediate, and high EASIX groups separately. Right: ST2 serum levels measured on day 100 in patients with versus without statin treatment. Median serum levels were 363 pg/ml versus 226 pg/ml, p-value 0.016, Mann-Whitney U-test. (A). Left: cartoon illustrating the role of IDO in the metabolism from tryptophan to kynurenine modified from. 21,54 IDO: indoleamine 2,3 dioxygenase, TDO: tryptophan-2,3dioxygenase, KMO: kynurenine 3 monooxygenase, KAT: kynurenine aminotransferase, KYNU: kynureninase, 3HAO: 3-hydroxyanthranilic acid 3,4-dioxygenase; Right: Forest plot showing the log2-fold changes of effect size and 95% confidence intervals (95% CI) derived from multivariate analyses to assess the impact of statins on cytokine levels for IDO, tryptophan, kynurenine, and the Kyn/Trp-ratio (RatioKT) including besides statin use the following covariables: age (10 years), recipient gender, donor gender, donor type, disease entity, conditioning regimen, disease score, and EASIX-pre. The analysis was performed for all EASIX groups ("all"), as well as for low, intermediate, and high EASIX groups separately (B). Serum levels of IDO, Trp, Kyn, and Kyn/Trp ratio on day 100 post-alloSCT in patient on versus off statin treatment. Numbers below plots indicate sample size per group, p-values from univariate analyses, Mann-Whitney U-test, see B for multivariate analysis (C).
IL18 and IFNg showed no difference on day 100 with or without statin treatment, we found higher d100-Kyn/Trp ratios in the serum of patients on statins. In line, IDO serum levels were also higher in the statin group. Higher IDO activity in statin patients might also result from the association between statins and CXCL9 levels in our multivariate analyses, as we had shown a strong association between IDO and CXCL9, but not IL18 in a previous study. 31 Tryptophan metabolism has been shown to play a pivotal role in the immune system with diverse effects. 20,44 Increased IDO activity enhances tryptophan catabolism and leads to an accumulation of kynurenine and its downstream metabolites, which was shown to induce Treg differentiation by Aryl hydrocarbon receptor (AhR) pathway activation. 45,46 Our results are consistent with reports from other, nonmalignant diseases in that statins affect tryptophan metabolism by increasing IDO activity. 19 While the immunosuppressive and anti-inflammatory effects of IDO are beneficial in the context of asthma or auto-immune diseases, 19,45,47 it was suggested that high IDO activity provides an immuno-evasive tumour microenvironment that promotes cancer progression. 44,48 These reports raised the question whether SEP might negatively affect graft-versus-leukaemia/graft-versus-tumour effects (GVL/ GVT). Reduced GVL/GVT would be reflected by a higher relapse rate. However, we did not detect an increase in relapse rates in patients receiving statins, and relapses post-aGVHD seemed to be even less frequent in the statin group, although this observation must be confirmed in independent patient cohorts. It has been shown before that mild cGVHD, but not aGVHD, is associated with lower incidence of relapse in patients with acute myeloid leukaemia (AML) post-alloSCT. 49 We also did not observe any change in the incidence of cGVHD in patients with versus without statins, further supporting that the immune activation mediating mild cGVHD and GVL/GVT is not hampered by statin treatment.
Another serum marker that differed in the statin group was the soluble IL33 receptor ST2. Serum ST2 is a biomarker of cardiovascular injury and risk of mortality. 50,51 Furthermore, ST2 levels were found to be associated with increased 6-month NRM post-alloSCT. 52 We show here that ST2 levels were lower in statin-treated patients. The effect was less prominent in multivariate analyses (p = .05 in the high EASIX group), suggesting that ST2 might not be the only driver of the observed effects of statin treatment. ST2 levels have also been associated with the occurrence of TA-TMA. 8 In line, TA-TMA incidence was reduced in the statin group, as reported before. 9 We could not detect an additional benefit of UDCA prophylaxis on TA-TMA incidence suggesting that UDCA and statins might act via different mechanisms.
Whether and how higher IDO activity on day 100 post-alloSCT is caused by statins, and whether these alterations are responsible for the reduced NRM under statin treatment will have to be determined in functional in vivo experiments.
The limitations of our study include the retrospective approach, the single-centre data, heterogeneity of the patient cohort, and the missing validation cohort. Our basic assumption requires that no significant clinical changes were introduced other than SEP during the recruiting period between 2008 and 2011.
In summary, we report that statins administered as prophylaxis against potential endothelial damage were associated with reduced NRM after alloSCT and aGVHD resulting in better OS of patients. Patients with intermediate endothelial risk seem to particularly benefit from this prophylaxis when compared to patients with low or very high endothelial damage. Possible mechanisms mediating these effects might be the (trend of) reduced ST2 levels measured on day 100, which are in line with a reduced incidence of TA-TMA, and the altered tryptophan metabolism through higher activity of IDO on day 100. Further functional studies are required to establish a mechanistic link between statins, ST2, tryptophan metabolism, and NRM in the alloSCT setting.
In conclusion, our study suggests that implementation of SEP in the standard of care after alloSCT is associated with reduced NRM in a subgroup of patients with an intermediate endothelial risk profile (EASIX intermediate quartiles) and highlights the importance of endothelial risk stratification when assessing the impact of endothelial prophylaxis in heterogenous patient cohorts.

ACKNOWLEDGMENTS
We thank M. Heß for his technical support with cytokine and metabolite measurements. We thank the Med V biobank for serum collection and preparation. C.P. receives a Max-Eder grant from the German Cancer Aid (70114435) and is supported by the Olympia Morata Program of Heidelberg University. TL: Deutsche Krebshilfe (70113529). Open Access funding enabled and organized by Projekt DEAL.