GSTM1 and GSTT1 Double Null Genotypes Determining Cell Fate and Proliferation as Potential Risk Factors of Relapse in Children With Hematological Malignancies After Stem Cell Transplantation. On Behalf of the Pediatric Disease Working Party of the European Society for Blood and Marrow Transplantation

Relapse is the major cause of treatment failure in children with hematological malignancies (HMs) undergoing busulfan (BU)- based allogeneic hematopoietic stem cell transplantation (HSCT). Glutathione S-transferases (GSTs) isoforms that participate in BU detoxication and protect cells against stress and cell death may be linked to post-HSCT outcomes. This study aimed to retrospectively evaluate the genetic association of null variants of Glutathione S-transferases GSTM1 and GSTT1 with relapse incidence in children with HMs undergoing BU- containing allogeneic HSCT and to assess the impact of these variants on BU-induced cytotoxicity on the immortalized lymphoblastoid cell lines (LCLs) and tumor THP1 GST-gene edited cell models. GSTM1- and GSTT1- null alleles were genotyped using germline DNA from whole blood prior to a conditioning BU-based regimen. Association of GSTM1- and GSTT1- null variants with relapse incidence was analyzed using multivariable competing risk analysis. BU-induced cell-death studies were conducted in GSTs-null and non-null LCLs and CRISPR-Cas9 gene-edited THP1 leukemia cell lines. vitro functional the variants by continuous follow-up of redox potential within 72h. Apoptosis/necrosis kinetic results demonstrate that BU-induced apoptotic processes are more pronounced in GSTM1-null LCLs. In contrast, primary necrotic cell death was more pronounced in GSTM1-non-null cells when comparing with the GSTM1 -null cells. In addition, primary necrosis was signicantly induced at an earlier stage in GSTM1-non-null cells. These results show that GSTM1-null variants can modulate BU-induced cell death, which were supplemented further by increased activation of known apoptotic markers Caspase − 3 or -7 in GSTM1-null LCLs and THP1 in comparison to GSTM1-non-null cells. Importantly, observed reduced rates of GSTM1-dependent cell death cannot be attributed to the increased baseline cell proliferation. The ndings of higher primary necrosis, lower early apoptosis and lower cell viability in GSTM1-non-null hematopoietic cells compared to GSTM1-null cells treated with BU were unexpected. Contrary to our observations, many studies showed associations between increased expression or activity of GSTs and resistance mechanisms against a range of cytotoxic drugs (44, 45). These results could potentially be explained by not only direct detoxication with GSH but also through negative regulation of pro-apoptotic protein kinases, such as apoptosis signal-regulating kinase 1 (ASK1) (13, 14). For instance, stress conditions cause the release of ASK1 from GSTM1, thereby leading to induction of apoptosis, which was shown in our experiments after induction with BU. In addition, GSTM1-null cells carrying more free ASK1 for phosphorylation activation are expected to have more apoptosis upon BU-induced stress in comparison to GSTM1-null cells which is in accordance with our in vitro results.


Abstract Background
Relapse is the major cause of treatment failure in children with hematological malignancies (HMs) undergoing busulfan (BU)-based allogeneic hematopoietic stem cell transplantation (HSCT). Glutathione S-transferases (GSTs) isoforms that participate in BU detoxi cation and protect cells against stress and cell death may be linked to post-HSCT outcomes. This study aimed to retrospectively evaluate the genetic association of null variants of Glutathione Stransferases GSTM1 and GSTT1 with relapse incidence in children with HMs undergoing BU-containing allogeneic HSCT and to assess the impact of these variants on BU-induced cytotoxicity on the immortalized lymphoblastoid cell lines (LCLs) and tumor THP1 GST-gene edited cell models.
Methods GSTM1-and GSTT1-null alleles were genotyped using germline DNA from whole blood prior to a conditioning BU-based regimen. Association of GSTM1and GSTT1-null variants with relapse incidence was analyzed using multivariable competing risk analysis. BU-induced cell-death studies were conducted in GSTsnull and non-null LCLs and CRISPR-Cas9 gene-edited THP1 leukemia cell lines.

Results
Carrying GSTM1/GSTT1 double null genotype was found to be an independent risk factor for post-HSCT relapse in 86 children (adjusted HR: 6.52 [95% Cl, 2.76 -15.42; p= 1.9 x 10 -5 ]). BU induced cell death preferentially in THP1 GSTM1(non-null) and LCLs GSTM1(non-null) as shown by decreased viability, increased necrosis and levels of the oxidized form of glutathione compared to null cells, while GSTT1 non-null cells showed increased baseline proliferation.

Conclusion
The clinical association suggests that GSTM1/GSTT1 double null genotype could serve as genetic strati cation biomarker for the high risk of post-HSCT relapse. Functional studies have indicated that GSTM1 status modulates BU-induced cell death. On the other hand, GSTT1 is proposed to be involved in baseline cell proliferation.

Background
Survival rates of children with hematological malignancies (HMs) undergoing allogeneic hematopoietic stem-cell transplantation (HSCT) have improved over the years achieving 91% estimated two-year overall survival. The improvement is mainly attributed to reduced HSCT-related toxicity and mortality. The incidence of post-HSCT relapse remains a signi cant complication and varies from 12-33% after two years (1). Risk factors that in uence transplant success are on the one hand host-and disease-related, such as disease genetics and remission status before HSCT, and on the other hand, transplant-related, such as conditioning regimen and treatment-related toxicities including for example severe graft-versus-host disease (GvHD), sinusoidal obstruction syndrome (SOS), and infections (2)(3)(4).
A bifunctional alkylating agent Busulfan (BU) is still often used in conditioning regimens prior to HSCT in children and adolescents (5) and is commonly administered along with other chemotherapeutics, e.g. cyclophosphamide (CY), and udarabine (FLU) (6, 7). At least in acute myeloid leukemia (AML), BU has shown lower long-term adverse effects consequently replacing total body irradiation (TBI) in the conditioning regimen (8). In acute lymphoblast leukemia (ALL), although the recently published results showing lower rates of relapse after TBI-containing conditioning, the results obtained with BU in association with FLU and thiotepa were encouraging and indicate an opportunity to nd genetic subgroups of patients who might bene ciate from the TBI-free conditioning (1).
BU is metabolized via conjugation with glutathione (GSH) in the liver, which is predominantly catalyzed by glutathione S-transferase alpha1 (GSTA1) (9). In hematopoietic cells (HCs), where GSTA1 is not expressed (10), other GST isoenzymes, particularly Mu1 (GSTM1, 46% of the BU conjugating activity of GSTA1 (9)) might play the most important role. The role of GSTT1 in BU conjugation is not yet known but has been mostly reported to have combined effects with GSTM1 on clinical outcomes (11,12). In addition to their protective role of the conjugation of BU in HCs, they might contribute to multiple cellular processes such as regulation of cell proliferation and apoptosis through the interaction with protein kinases such as apoptosis signal-regulating kinase 1 (ASK1). Under stress conditions, the interaction of the GSTM1:ASK1 complex is dissociated and results in activation of ASK1 that activates the c-Jun N-terminal kinase (JNK) and mitogen-activated protein kinase p38 (MAPK p38) pathways, leading to upstream cytokine-and stress-induced apoptosis (13,14). However, the impact of apoptosis through kinases on BU-dependent cytotoxicity is poorly understood and even less whether those GST genes naturally knocked down might interfere in the post-HSCT relapse potential. GSTM1 and GSTT1 genes can be homozygously deleted (presented as GSTM1-null and GSTT1-null) and thus completely deprived of the enzyme activities in a high percentage of individuals (the average % in Europe are 51 and 19, respectively) (15). In AML adult patients, Weiss et al. (16) showed a perfect concordance of those variants in malignant and germ-line DNA, which suggests that the germline genotype drives protein expression in malignant cells.
Although these variants have been associated with a higher risk of leukemia development (17), there are con icting reports on the association of the GSTM1null and GSTT1-null variants with relapse in patients with HMs (18-21). To date, there is no evidence available for the association of germline GSTM1-null and GSTT1-null variants with post-HSCT relapse in children with HMs.
Because GSTM1 and GSTT1 are the main remaining GSTs in HCs, we hypothesized that the absence of either or both proteins should affect BU cytotoxicity through conjugation -dependent or -independent ways, interfering in the HSCT outcomes. Hence, a genetic association study based on germline GSTT1and GSTM1-null variants was undertaken. Further, we conducted in vitro functional analyses to understand the role of these variants in survival and BU-induced apoptosis and necrosis of the immortalized and tumor lymphoblastoid cell lines (LCLs).

Clinical association study
Patients and treatment Pediatric patients with ALL, AML or myelodysplastic syndrome (MDS) who had undergone allogeneic HSCT between 2000-2013 were enrolled in the study.
The Institutional Review Board or ethics committees approved the study and all patients and/or parents provided informed consent. The present study is a subset of the multicentric study under the umbrella of the European Society for Blood and Marrow Transplantation (EBMT) (Clinicaltrials.gov identi er: NCT01257854) (22). I.v. BU (Busulfex, Otsuka Pharmaceuticals, Saint-Laurent, Montreal, QC, Canada or Busilvex, Pierre Fabre Laboratory, Paris, France) administration was given as a 2h infusion to the patients, every 6 h for a total of 16 doses. The rst BU dose was age-and weight-based and pharmacokinetic (PK)-guided dose adjustment was performed in order to obtain a cumulative area under the curve (CumAUC) between 59.2-98.56 mg*h/L as reported previously (22).
The primary diagnosis of HMs was made at the referring institution. Patients were considered to be in remission after chemotherapy if they presented < 5% blasts in the normal cellular bone marrow. Relapse in MDS was de ned as > 5% and ≤ 20% of blasts at the bone marrow examination after engraftment and/or reappearance of major dysplastic features associated with cytopenias and/or mixed chimerism > 5% and/or detection of the same cytogenetic abnormality present at diagnosis. Relapse in AML and ALL was de ned as the presence of blasts in the bone marrow > 5%, con rmed by ow cytometry; detection of the gene fusion present at diagnosis; or according to minimal residual disease (MRD) results after the transplantation if available. Disease remission status was de ned by the number of bone marrow remission or relapse events before HSCT.
Cumulative relapse incidence was de ned according to the standard guidelines of EBMT and as detailed in our recent report (22).

Genotyping and statistical analysis
Genotyping of GSTM1-null and GSTT1-null variants was performed on germline DNA, extracted from whole blood or peripheral mononuclear cells of all patients before the rst HSCT as described by Lin et al. (23).
Pearson Chi-Square test was used to analyze the differences in demographics between groups with and without GST -null variants. Estimated cumulative relapse incidence by competing risk analysis with non-relapse mortality as a competing event and the difference among groups were estimated by Gray's test (24). The Fine-Gray model was used for competing risk regression in multivariable analysis to obtain adjusted p-values for all the variables in relation to the genotype groups (25). The potential risk factors with a p-value ≤ 0.25 in the univariable competing risk analysis were retained in the multivariable analysis by including the GST genotype factor with the lowest p-value. The nal multivariable analysis included: diagnosis (ALL, AML and MDS), disease status [1st complete remission (CR), a higher degree of CRs and absence of CR], conditioning regimen (standard regimen with two alkylating agents and intensi ed regimen with three alkylating agents), AUC after the rst dose of Busulfan (1st BU dose AUC categorized into below 3.7, between 3.7 and 6.16, and above 6.16 mg*h/L) and BU CumAUC (below 59.2, between 59.2 and 98.6, and above 98.6 mg*h/L) as categorical variables.  Table 1. THP1 GSTM1and GSTT1knock out cell lines (THP1 GSTM1(−/−) and THP1 GSTT1(−/−) ) were prepared from parental THP1 representing non-null genotype for GSTM1 (THP1 GSTM1(+/+) ) and GSTT1 (THP1 GSTT1(+/+) ) using CRISPR/Cas9 gene-editing method. Plasmid PX458 containing 5`-TGATACTGGGGTACTGGGAC-3g RNA (GSTM1) or 5`-TGAAGGACGGGGACTTCACC-3` gRNA (GSTT1) (prepared by GeneScript, The Netherlands) was transfected into THP1 cells. 10,000 cells were uorescence-activated cell sorted (FACS) in 24-well plates based on the presence of green uorescence protein (GFP) 48 h post-transfection. After 48 to mutations using Sanger sequencing and con rmed by Western blot for the success of gene knockout. Five clones of the same genotype were pooled in a population.
DNA and proteins of the selected cell lines (LCLs, THP1 GSTM1(−/−) and THP1 GSTT1(−/−) cell models) were extracted using DNeasy Blood and Tissue Kit (Qiagen, Hilden, Germany) and standard protein extraction protocol for western blot using RIPA lysis buffer (Sigma-Aldrich, Germany), respectively. The intracellular concentration of extracted proteins was measured using Bradford assay from Bio-Rad (Hercules, CA) according to the manufacturer's recommendations.
Aliquots containing 20 µg of proteins, sample reducing agent and LDS sample buffer (Thermo Fisher Scienti c, USA) were subjected to electrophoresis by using Invitrogen Novex Tris-Glycine Gels (Thermo Fisher Scienti c, USA). Dry transfer to a nitrocellulose membrane was performed with the iBlot dry blotting system (Thermo Fisher Scienti c, USA). A membrane was blocked using 5% milk in PBS and 0.05% Tween 20. The following primary antibodies were used for protein labeling: ß-Actin Mouse monoclonal antibody ( DNA samples from LCLs were genotyped for GSTM1-null and GSTT1-null variants using multiplex real-time PCR ampli cation in the presence of SYBR Green I and genotype discrimination by melting curve analysis in a StepOnePlus™ Real-Time PCR System (Applied Biosystems™, Foster City, CA, USA) with BCL2 (BCL2 apoptosis regulator) gene as an internal control as described earlier (26). The genotyping method used cannot differentiate the heterozygous individuals from homozygous non-null carriers (furtherly marked as GSTM1(+) and GSTT1(+)) except when using Sanger sequencing.  Table 1) and at baseline in THP1 GSTT1(−/−) and THP1 GSTT1(+/+) using RealTime-Glo MT Cell Viability Assay (Promega, USA).
Annexin V/PI assay (BD Biosciences) was used to measure live, early and late apoptotic and necrotic cells. Prior to FACS, ten GSTM1(+) and ten GSTM1(-/-) LCLs (Supplementary Table 1) were treated for 48 h with BU (1% DMSO) at 250, 500 and 1000 µM and two samples of each cell line were used as controls (1% DMSO and untreated). One million cells were labeled according to the manufacturer's protocol. FACS analysis was performed using the CyAN ADP system (Beckman Coulter, UK). Results were analyzed by Kaluza analysis software, version 1.3 (Beckman Coulter, UK). Apoptosis and necrosis were followed for 72h in six GSTM1(+) and six GSTM1(-/-) LCLs (Supplementary Table 1) treated with 500 µM BU using RealTime-Glo MT Annexin V Apoptosis and Necrosis Assay (Promega).
Chemiluminescent signals were measured using Victor3 (Perkin Elmer, Inc, USA). All BU-based data were normalized relative to the negative controls with 1% DMSO.

Statistical analyses in in-vitro functional studies
The cell-based experiments (IC 50 distribution, end-point apoptosis and necrosis, real-time monitoring of viability, apoptosis and necrosis, Caspase3/7 activities; and [GSSG/GSH T ] ratios) were performed at least in duplicates and results are reported as observed means ± SD strati ed by GST-null and GST-nonnull variants. Statistical differences between genotypes were assessed using Mann-Whitney, t-tests, or two-way ANOVA according to the normality of the distribution and compared to untreated controls using GraphPad Prism 7 software (RRID: SCR_002798). We considered p < 0.05 to be statistically signi cant in all analyses. Results 1. GSTM1-and GSTT1-double nullgenotypes are associated with higher relapse incidence.
Eighty-six children with malignancies aged 5 months -18 years (female/male, 44/42), who received myeloablative conditioning containing with four-timesdaily i.v. BU followed by HSCT, were enrolled in this study. The patients` baseline characteristics at the time of their HSCT are summarized in Table 1 Regarding the genotype frequency, 49 patients were GSTM1-null (57.0 %), 24 GSTT1-null (27.9 %) and 9 had null genotypes in both GSTM1 and GSTT1 genes (10.5%). Characteristics of these patients according to GSTM1-null and GSTT1-null variants are shown in Supplementary Table 2.
2. LCL sensitivity to BU is associated withGSTM1but not withGSTT1genotypes Signi cantly higher cell viability after treatment with BU was observed in LCLs with GSTM1-null genotype (1.8-fold, p = 0.013) and THP1 GSTM1(−/−) cells (1.5fold, p = 0.0006) compared to GSTM1-non null by 48h endpoint ( Figs. 2A and 2B, respectively) and the results were con rmed by 72h-kinetic measurements in LCLs ( Supplementary Fig. 2). GSTT1-null, alone or in combination with GSTM1-null, did not show a signi cant association with BU-IC 50 in LCLs and THP1 GSTT1(−/−) cell lines ( Fig. 2C and 2D, respectively). No difference in baseline cell proliferation was seen between GSTM1-null and GSTM1-non-null cells (Fig. 3A), while the proliferation of GSTT1-null cells was signi cantly decreased in comparison to GSTT1-non-null carriers in LCLs carrying GSTT1-null Cell death mechanisms were further followed by kinetic plots. We observed higher apoptotic rates in GSTM1-null cells (THP1 GSTM1(−/−) , LCLs) through the whole 72h of follow-up after BU treatment when compared to GSTM1-non-null cells from unrelated individuals (Supplementary Figs. 3A and 3C, p < 0.0001 and p = 2.6E-05, respectively). In contrast, we observed lower necrotic rates in GSTM1-null cells (THP1 GSTM1(−/−) , LCLs) that were increasing after 26 h of BU treatment when compared to GSTM1-non null cells (Supplementary Figs. 3B and 3D, p < 0.001 and 1.4E-05, respectively). Apoptosis at the same time-points was lower in these cells accounting for the faster cell death mainly as the result of primary necrosis.

Discussion
Our clinical association study performed in 86 children with HMs undergoing HSCT following BU-based conditioning regimens demonstrated that patients harboring homozygous deletions in both GSTM1 and GSTT1 genes presented a high risk of relapse (HR 7.2 [95% Cl, 2.2-23.9; p = 0.002]). After adjustment for known risk factors (diagnosis, disease status, the intensity of conditioning regimen and BU exposure), the association remained signi cant demonstrating that the deletion of both GST genes is an independent risk factor for relapse (adjusted HR 6.52 [95% CI, 2.8-15.4; p = 1.9 x 10 − 5 ]). Although it is a small cohort, this is the rst report on the risk of post-HSCT relapse in relation to the germ-line GSTM1and GSTT1-null variants in children with HMs. Until now, only one study conducted in BU/CY-based HSCT settings although in adults showed increased relapse rates in patients carrying GSTM1-null genotype, while no association was identi ed with GSTT1-null genotype (27). Concerning non-transplant-based studies in pediatric or adult patients, a similar association between GSTM1/GSTT1 double null carriers and increased risk of relapse (19, 28-30), lower complete remission rate (31) and lower event-free survival was demonstrated (18, 19, 32-38). There are nevertheless a few studies showing no such an association (20,39), in which the small number of patients or the different treatment regimens may have mainly precluded de ning a relationship between GST variants.
Based on the known detoxifying role of GSTs, our results from the clinical association are contradictory. Although GSTA1 is the main enzyme involved in BU detoxi cation, GSTM1 is also highly expressed in the liver and recognized as involved in BU conjugation (9,11,22,40), precluding the BU to cross-link with the DNA strands. Functional variants of the genes coding for GSTs may then interfere in HSCT by affecting BU metabolism. It is known that low BU exposure (CumAUC < 59 mg×h/L) is associated with graft failure and relapse (5,22,41) whereas high BU exposure (CumAUC > 98.6 mg×h/L) could reduce post-HSCT relapse in leukemia at the cost of an increase in organ toxicities, and therefore transplantation-related mortality (5,22,(41)(42)(43). However, at the level of HCs, less is known about the direct effect of BU.
We compared BU-related cell death mechanisms in LCLs and THP1 with and without GSTM1 and/or GSTT1 genes after exposure to BU. We demonstrated that only GSTM1-null (but not GSTT1-null) is associated with higher resistance to BU as determined by higher BU-IC 50 values of GSTM1-null LCLs and THP1 (GSTM1−/−) in comparison to GSTM1-non-null cells. This could be due to a change in the redox equilibrium as demonstrated by lower levels of oxidized GSH, lower primary necrosis and higher early apoptosis. An increase of GSTM1-null LCL`s viability was con rmed either by continuous follow-up of redox potential within 72h. Apoptosis/necrosis kinetic results demonstrate that BU-induced apoptotic processes are more pronounced in GSTM1-null LCLs. In contrast, primary necrotic cell death was more pronounced in GSTM1-non-null cells when comparing with the GSTM1 -null cells. In addition, primary necrosis was signi cantly induced at an earlier stage in GSTM1-non-null cells. These results show that GSTM1-null variants can modulate BU-induced cell death, which were supplemented further by increased activation of known apoptotic markers Caspase − 3 or -7 in GSTM1-null LCLs and THP1 in comparison to GSTM1non-null cells. Importantly, observed reduced rates of GSTM1-dependent cell death cannot be attributed to the increased baseline cell proliferation.
The ndings of higher primary necrosis, lower early apoptosis and lower cell viability in GSTM1-non-null hematopoietic cells compared to GSTM1-null cells treated with BU were unexpected. Contrary to our observations, many studies showed associations between increased expression or activity of GSTs and resistance mechanisms against a range of cytotoxic drugs (44,45). These results could potentially be explained by not only direct detoxi cation with GSH but also through negative regulation of pro-apoptotic protein kinases, such as apoptosis signal-regulating kinase 1 (ASK1) (13,14). For instance, stress conditions cause the release of ASK1 from GSTM1, thereby leading to induction of apoptosis, which was shown in our experiments after induction with BU. In addition, GSTM1-null cells carrying more free ASK1 for phosphorylation activation are expected to have more apoptosis upon BU-induced stress in comparison to GSTM1-null cells which is in accordance with our in vitro results.
However, the observed paradox in increased cell death of GSTM1 well expressed cells upon BU treatment could additionally be explained by ndings of the study of DeLeve et al. (46) demonstrating that in murine hepatocytes BU is cytotoxic also through oxidative stress caused by BU metabolites (BU glutathione S-conjugate thiophenium ion, GS + THT) and by the depletion of GSH in addition to DNA alkylation. The toxic metabolites of BU/GSH metabolism are mainly oxidized by Flavin-containing monooxygenases (FMOs, e.g. FMO3) and cytochromes (CYPs, e.g. CYP3A4) (47) to water-soluble non-toxic metabolites (e.g. sulfolane, (48)). However, CYP3A4 and FMO3 are mainly expressed in the liver (accounting for 54% of overall tetrahydrothiophene [THT] disappearance, the metabolite of BU), and less in LCLs, as observed in our laboratory (data not shown) and by others (https://www.proteinatlas.org). After RNA sequencing in LCLs, very low or no gene expressions of CYP 2D6, 2C19, 2C9, 2B6, 2C8, 4A11, 3A4, FMO1 and FMO3 were identi ed. In this context, the oxidative burst caused by electrophilic molecules from BU-GSH conjugation (49,50) in addition to the absence of CYP3A4 and FMO3 could be a reason for the lower sensitivity of GSTM1-null HCs to BU, as observed in LCLs and THP1. In contrast, higher total expressions of CYPs and FMOs in hepatocytes (47) could explain why GSTA1slow BU metabolizing individuals in addition to the absence of GSTM1 activity show potentially more treatment-related toxicities (e.g. SOS (51) and aGvHD (52)) than carriers with normal GST`s enzyme activities. A hypothetical comparative model of the difference in BU fate between hepatocytes and lymphocytes is suggested in Supplementary Fig. 4 and warrants further investigation.
The genetically-determined different cell fate after BU exposure might explain the apparently discordant results between the relapse incidence in patients carrying GSTM1-null genotype (in combination with GSTT1-null) and the cellular resistance to BU in GSTM1 null-LCLs and THP1 GSTM1(−/−) . The higher rates of necrosis in GSTM1-non null cells might predict a pro-in ammatory cell death of the malignant cells, resulting in enhanced immunogenicity (53). Unlike the other chemotherapeutic regimens including autologous transplantation, the e cacy of the allogeneic transplantation relies on the graft-versus-leukemia effect, especially in HMs (54,55), but that theory should be further explored.
Another relevant observation is the signi cantly increased post-HSCT relapse in GSTT1-null when combined with GSTM1-null genotype in children with HMs. The link between GSTT1 and post-HSCT relapse is not clear yet. Our in vitro observations cannot be attributed to the BU-related differences in IC 50 values or [GSSG/GSH T ] ratios. Other pharmacogenomics studies also demonstrated that genetic variations in GSTT1 are not associated with BU clearance or liver toxicity (11,51,56,57). Nevertheless, we observed faster baseline proliferation in GSTT1-non-null LCLs/THP1 and a slightly higher baseline increase of Caspase 3/7 activation compared to those with GSTT1-null genotype, indicating GSTT1 potential involvement of BU-independent mechanisms in the relapse development.
The results of the present clinical study are limited by the retrospective study design and relatively small pediatric sample size with no clinical validation cohort. However, the sample size of 86 patients has at least 80% power with 10% of observed combined GSTM1/GSTT1 double null variants frequency and relapse incidence with the estimated observed effect size of ≥ 7.0 and alpha value of 0.05. The primary diagnosis of HMs was made at the referring institution and was not centrally reviewed. Well-known risk factors such as somatic genetic/cytogenetics abnormalities, the donor DNA and the initial response to the treatment (e.g. MRD) were not available. However, as described in Supplementary Table 2 similar characteristics were present between the GST genetic subgroups (p-values > 0.05). The GST-null variants were not associated with the status of the disease before HSCT and we are assuming that the germline genotype impact on protein expression was present in malignant cells as shown by Weiss et al. (16). The majority of cases in our study underwent a BU-CY conditioning regimen, however, it is not known if this association is speci c to a BU-CY conditioning regimen only or unspeci c to other chemotherapeutics used in HSCT setting (e.g.Thio or Mel) (7). For instance, active metabolites of CY (e.g. acrolein) are also eliminated by GSH conjugation catalyzed by GSTs (48). This needs to be evaluated in the future with a focus on whether GSTs play a major role in determining clinical outcomes. This aspect is currently being evaluated by our group using a cohort from multiple centers with the usage of multiple conditioning regimens. Furthermore, the transplantrelated mortality or combined toxicities were not associated with the GSTM1and GSTT1-null variants (data are not shown), suggesting compensation of BU conjugation by other GSTs, especially GSTA1, which is mainly expressed in hepatocytes and other somatic cells.

Conclusions
In summary, we report that GSTM1/GSTT1 double null genotypes could serve as genetic biomarkers for identifying pediatric patients with HMs at higher risk of relapse after an allogeneic HSCT following BU-containing conditioning. On the other hand, the absence of those markers might predict the patients who more likely will respond to the chemotherapy-based conditioning. Functional studies indicated different mechanisms of cell death upon exposure to BU based on the presence or absence of GST-null alleles and the in vivo impact of those ndings must be further explored.       THP1GSTT1(-/-) (D.) cell models after the treatment with 500 μM BU. Statistical analysis was performed by the two-way ANOVA considering 250 or 500 μM BU concentration (genotype factor); t-tests between GST(-/-) variants in each condition separately were used; no statistically signi cant differences were observed between GST(-/-) variants in LCLs and THP-CRISPR-Cas9 models in either 1% DMSO or medium; P-values below 0.05 were considered statistically signi cant. GSTM1-null and GSTT1-null are presented as GSTM1(-/-) and GSTT1(-/-), respectively. GSTM1 non-null and GSTT1 non-null genotypes are presented as GSTM1(+) and GSTT1(+), respectively.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. 003JurkovicMlakarM1T1SuppMaterial.docx