Prediction of acute graft versus host disease and relapse by endogenous metabolomic compounds in patients receiving personalized busulfan-based conditioning

Abstract

Busulfan-based conditioning is the most commonly used high-dose conditioning regimen for allogeneic hematopoietic cell transplant (HCT). The alkylating agent busulfan has a narrow therapeutic index, with busulfan doses personalized to a target plasma exposure (targeted busulfan). Using a global pharmacometabonomics approach, we sought to identify novel biomarkers of relapse or acute graft versus host disease (GVHD) in a cohort of 84 patients receiving targeted busulfan before allogeneic HCT. A total of 763 endogenous metabolomic compounds (EMCs) were quantitated in 230 longitudinal blood samples before, during, and shortly after intravenous busulfan administration. We performed both univariate linear regression and pathway enrichment analyses using global testing. The cysteine/methionine pathway and the glycine, serine, and threonine metabolism pathway were most associated with relapse. The latter be explained by the fact that glutathione-S-transferases conjugate both busulfan and glutathione, which contains glycine as a component. The D-arginine and D-ornithine metabolism pathway and arginine and proline metabolism pathway were most associated with acute GVHD. None of these associations were significant after correcting for false discovery rate (FDR) with a strict cutoff of FDR-adjusted P < 0.1. Although larger studies are needed to substantiate these findings, the results show that EMCs may be used as predictive biomarkers in HCT patients.

Graphical Abstract

Introduction

Allogeneic hematopoietic cell transplantation (HCT) is a potentially curative procedure, with its most frequent indication being hematologic malignancies.¹ In allogeneic transplantation, grafting of hematopoietic stem cells from one individual to another provokes immunologic reactions involved in engraftment of the donor cells, graft-versus-host disease (GVHD), control of a malignancy, the development of tolerance, and immune reconstitution.² These immunologic reactions are influenced by the conditioning regimen (also termed preparative regimen), the type and source of the donor graft, and the post-transplant immunosuppressive regimen, all of which are essential components of the HCT procedure. In HCT recipients, overall survival is improved in those conditioned with intravenous (IV) busulfan compared to total body irradiation (TBI); however, potential severe toxicity of busulfan-based conditioning persists.^3-5

Busulfan-based conditioning is the most commonly used high-dose conditioning regimen for allogeneic HCT. Busulfan is hepatically metabolized through glutathione (GSH) conjugation by glutathione S-transferase (GST) enzymes; this process depletes hepatocyte GSH stores in murine hepatocytes in vitro.⁶ Dysregulation of GSH and accumulation of cysteine, cystathione, and cysteinylglycine are associated with GVHD in experimental murine models of HCT (Supplemental Table 1, Table S1).⁷ Compared to syngeneic HCT mice and non-transplant control mice, allogeneic HCT mice had significant decreases in reduced glutathione and increases in oxidized glutathione, indicating early shifts in oxidative stress.⁷ Thus, we hypothesized that plasma endogenous metabolomic compounds (EMCs; i.e., not downstream metabolites of busulfan) are associated with GVHD in patients receiving busulfan-based conditioning regimens prior to allogeneic HCT.

Relapse is reduced by personalized dosing of busulfan to a target area under the plasma concentration-time curve (AUC), termed targeted (^TBU) or pharmacokinetic (PK)-directed dosing.⁸ However, neither targeted busulfan dosing nor the introduction of IV targeted busulfan have sufficiently reduced relapse. Based on preclinical data that GSH is important to busulfan toxicity, we recently conducted a clinical trial that reversed the order of administration – specifically giving cyclophosphamide (CY) first followed by IV targeted busulfan.⁹ Hepatic toxicity is low with both the traditional administration order of targeted busulfan followed by CY regimen and the newer regimen of giving CY first followed by IV targeted busulfan (i.e., CY/^TBU) regimens. However, for patients with acute myeloid leukemia or myelodysplastic syndrome, the incidence of relapse was higher with CY/^TBU compared to the traditional administration sequence of ^TBU followed by CY. This raised the hypothesis that GSH dysregulation may also influence relapse.

To test this hypothesis, we sought to identify biomarkers predictive of the efficacy and toxicity (i.e., acute GVHD) of targeted busulfan conditioning regimens in patients diagnosed with a hematologic malignancy. We applied a global pharmacometabonomics approach investigating 763 EMCs in plasma samples obtained longitudinally over the course of IV targeted busulfan administration in 84 allogeneic HCT recipients.

Methods

Study population.

Between December 2014 and November 2018, 84 patients participated in this prospective ancillary biomarker study. This study was approved by the Fred Hutchinson Cancer Research Center Institutional Review Board (December 2014-November 2018; clinicaltrials.gov protocol number NCT02291965) and the City of Hope Institutional Review Board (March 2017-November 2018). All participants provided written informed consent prior to study procedures. The conditioning regimen and post-graft immunosuppression were not affected by participation in this study.

As part of HCT conditioning, all 84 participants underwent pharmacokinetic-guided dosing of IV busulfan, also called targeted busulfan (^TBU), which personalized each patient’s busulfan dose to achieve the desired target busulfan exposure. The first busulfan dose was based on body weight or body surface area. Subsequent doses were personalized using the individual participant’s busulfan clearance as previously described;⁸ the target busulfan AUC was chosen by the treating physician. Because of the circadian variability of busulfan pharmacokinetics,¹⁰ all busulfan pharmacokinetic sampling occurred in the morning. Antiemetics, antibiotics, and antifungals were given per Institutional Standard Practice Guidelines.

Clinical outcomes.

The assessment of clinical outcomes was described previously.¹¹ Acute GVHD and chronic GVHD were graded according to established criteria.^12-14 We defined disease relapse or disease progression as disease recurrence following complete remission or progression of persistent disease. The primary endpoints were relapse and acute GVHD (grade 0-1 vs. 2-4). Each endpoint was treated as a binomial outcome and analyzed separately.

Global pharmacometabonomics sample collection.

Longitudinal blood samples (3 ml/sample) were scheduled to be collected in sodium heparin tubes at three time points during IV busulfan dosing (Table S2): up to two weeks prior to the first conditioning dose (2-week pre-busulfan sample), immediately before administration of IV busulfan dose 1 (pre-busulfan sample), and with the last busulfan pharmacokinetic sample after the last morning busulfan dose (last busulfan PK sample).

Because we sought to identify biomarkers predictive of efficacy and toxicity, samples were obtained before busulfan administration. The 2-week pre-busulfan sample collection time was the earliest feasible time within the final HCT workup (i.e., time period (typically up to 2-weeks) in which the patient undergoes final assessment if they can receive an HCT). If a predictive association were found and metabolomics-guided HCT were subsequently used clinically, this sample would allow for more time for metabolomics quantitation and data interpretation. The immediately pre-busulfan sample was the latest feasible time before busulfan administration. If a predictive association were found and metabolomics-guided HCT were subsequently used clinically, this sample necessitates a rapid quantitation and data interpretation. The last busulfan PK sample was obtained because busulfan is a glutathione S-transferase substrate which we hypothesized may cause glutathione dysregulation, which is associated with GVHD.⁷

A total of 230 metabolomic samples were obtained: 64 participants had samples available from all three time points; 18 participants had samples from two time points; and 2 participants had samples from one time point.

The 2-week pre-busulfan, pre-busulfan, and last busulfan PK samples were immediately refrigerated at 4°C, stored for up to 4 hours from time of collection, centrifuged to plasma, and immediately stored at −80°C. The samples underwent at most one freeze-thaw cycle before metabolomic analysis (i.e., the analysis was conducted after the first or the second thaw).

Global pharmacometabonomics analysis.

Metabolite profiling of plasma was completed by Metabolon (Durham, North Carolina, USA). The samples were shipped on dry ice to Metabolon’s facility and stored at −80°C upon receipt. Samples were divided into 5 aliquots; one was held in reserve while each of the others was analyzed by one of four different mass spectrometry methods. Raw data were extracted, peak-identified, and QC processed; then Metabolon’s proprietary software was used to confirm the consistency of peak identification across the various samples. Compounds were identified by comparison of the processed data to Metabolon’s library entries of purified standards or recurrent unknown entities. Library matches for each compound were checked for each sample and corrected if necessary. Using the criteria established by the Chemical Analysis Working Group as part of the Metabolomics Standards Initiative,¹⁵ most EMCs met the level 1 standards for metabolite identification with the remainder meeting the level 2 standards per the same criteria. ANCOVA contrasts were performed to examine differences in metabolite levels between the samples at each time point. See Supplemental Methods for more details about metabolite quantitation and quality control.

Statistical analysis.

All data transformations and analyses were carried out using R version 3.5.¹⁶

Of the 763 EMCs (Table S3) measured, 332 (44%) had a detectable signal in all samples; 666 (87%) had detectable signal in three quarters of the samples; and 741 (97%) had detectable signal in at least half the samples. Values were normalized to the sample volume extracted and missing values, if any, were imputed with the minimum observed value for each EMC. These results were transformed using the centered log-ratio (CLR)^17-18 to account for the relative nature of the abundance measures and to approximate a normal distribution prior to analysis. Analyzing the CLR-transformed data amounts to taking the log of all measures and then normalizing each sample by its mean (log) abundance. A permutation MANOVA (perMANOVA) test was performed to check for an association between the set of all EMCs and relapse and between the set of all EMCs and acute GVHD.

Principal component analysis (PCA) was used to visualize differences in metabolomics profiles. The profiles were somewhat affected by age (Supplemental Figure 1, Figure S1); thus, age was included as a covariate in subsequent analyses. The plots revealed a clear separation between the 2-week pre-busulfan and pre-busulfan metabolomics profiles, collectively, and the last busulfan PK metabolomics profiles (Figure S2).

Univariate analysis:

Each EMC in the set of pre-busulfan samples was individually tested for an association with relapse and with acute GVHD. Generalized linear models were fit for each EMC to test its association with both endpoints including age as a covariate. For acute GVHD, donor type (relation of donor to the recipient) and Human Leukocyte Antigen (HLA) category (which HLAs are matched between donor and recipient) were also included as covariates. The Benjamini-Hochberg (BH) procedure was used to control the false discovery rate for the large number of tests performed.¹⁹ In view of the large number of tests, we designated a significance level of 0.1 for BH-corrected p values.

Pathway analysis:

To evaluate whether groups of EMCs were associated with relapse or acute GVHD, pathway analyses integrating pathway enrichment analysis and pathway topology analysis were carried out with MetaboAnalyst 4.0 using CLR-transformed EMC measures from the pre-busulfan samples.^20-21 Relapse (yes or no) and acute GVHD (grade 0-1 or 2-4) were evaluated as discrete outcomes, and only pre-busulfan samples were considered in the pathway analyses. 270 EMCs out of our total 763 EMCs were not present in the MetaboAnalyst compound library and were therefore not included in the pathway analysis. We tested pathway-defined sets of EMCs for their association with relapse and acute GVHD using the Global test²² within MetaboAnalyst. We also applied MetaboAnalyst to perform pathway enrichment analysis to exploit potential information in pathway topology. For this, we used relative betweenness centrality (number of shortest paths passing through a node), based on EMC centrality in a given metabolic network, to calculate EMC importance.²³ Pathway impact was calculated as the sum of the importance measures of the pathway-specific EMCs, normalized by the sum of the importance measures of all EMCs in each pathway.²⁴ For pathway-level testing, we only used pathways for which our measured EMCs represent at least 5% of the total number of pathway metabolites (60 pathways).

Results

Patient characteristics & clinical outcomes

The pre-transplant characteristics of the 84 participants are given in Table 1. The median age was 53 y (range 1.7 – 66.2), and slightly more participants were male (62%). Seventy-six (90%) participants received an HLA-identical graft, and 57 (68%) participants received a graft from an unrelated donor. Of the patient characteristics, only age impacted the plasma metabolome (Figure S1). PCA of metabolite abundances revealed a moderate separation of metabolic profiles according to age. Overall, the separation was proportional to the difference in age. Thus, age was included as a covariate in subsequent analyses.

Table 1.

Participant characteristics

Characteristic	Na
N	84
Age (y)	53.0 (1.7-66.2)
Male sex	52 (62%)
HLA match
Matched donor (HLA-identical)	76 (90%)
Mismatched donor	8 (10%)
Donor type
Unrelated	57 (68%)
Related – matched sibling	26 (31%)
Related – father	1 (1%)
HCT Conditioningb
CY/TBU	51 (61%)
TBU/CY	20 (24%)
FLU/TBU	10 (12%)
FLU/CY/TBU	2 (2%)
TBU/CY/TBI	1(1%)
TBU dosing frequency
Every 24 hours	77 (92%)
Every 6 hours	7 (8%)
Diagnosis
Myelodysplastic syndrome	52 (62%)
Acute myeloid leukemia	20 (24%)
Chronic myeloid leukemia	10 (12%)
Otherc	2 (2%)

^aData presented as: number (%) or median (range); percentages may not add up to 100 due to rounding

^bListed in administration order; all participants received targeted busulfan (^TBU), in which the IV busulfan dose was personalized based on the patient’s busulfan clearance to a target plasma exposure; CY

^cOther diagnoses include; chronic leukemia NOS n=1; and eosinophilic leukemia n=1.

Abbreviations. ^TBU: Targeted busulfan; CY: cyclophosphamide; FLU: fludarabine; TBI: total body irradiation

All patients received targeted busulfan over 4 days; the majority of the participants (n=77, 92%) received daily (every 24 hour) administration of busulfan. For targeted busulfan, an initial dose of busulfan based on body weight or body surface area was chosen (see FAQ6 of Palmer & McCune et al⁸ ) and administered on the first day of targeted busulfan. Next, sequential pharmacokinetic samples were drawn before the subsequent busulfan dose in order to estimate a patient’s busulfan exposure. These pharmacokinetic samples must be drawn over an acceptable time period that accounts for the half-life of busulfan (2-3 hours), the dosing frequency, and the need to obtain samples quickly enough to personalize subsequent doses of busulfan. Pharmacokinetic sampling was typically completed within 4 hours for a 2-hour busulfan infusion and every 6 hour (Q6H) dosing and within 8 hours for a 3-hour busulfan infusion and every 24 hour (daily) dosing. The doses of targeted busulfangiven on the second, third, and fourth days of targeted busulfan are adjusted based on the patient’s busulfan exposure, as estimated from the pharmacokinetic samples. The busulfan clearance is calculated from the administered busulfan dose and the resulting busulfan exposure (AUC). The majority (n=51, 61%) of participants received CY (60 mg/kg/day on each of two sequential days) followed by targeted busulfan (on each of four sequential days). The remaining participants (n=21, 25%) received targeted busulfan followed by CY (60 mg/kg/day × 2 days) and total body irradiation; and fludarabine with targeted busulfan ± CY (n=12, 14%). Prophylaxis of busulfan-induced seizures consisted of phenytoin (n=75), unknown (n=7), or levetiracetam (n=2). In a Center for International Blood and Marrow Transplant analysis, no differences were found in relapse-free survival or increased risks of relapse or acute GVHD with the use of alternative anti-epileptic medications as compared to phenytoin.²⁵

Ten patients experienced relapse. Regarding acute GVHD, 30 participants had grade 0 (n=25) or grade 1 (n=5) acute GVHD which were grouped together. The remaining participants were grouped together, with grade 2 (n=46), grade 3 (n=5), and grade 4 (n=3) acute GVHD.

Busulfan administration alters the plasma EMCs

Principal component analysis of all samples revealed a clear separation in metabolic profiles between the 2-week pre-busulfan and pre-busulfan samples, collectively, and the last busulfan PK sample (Figure S2). This separation indicates that the metabolic profiles of HCT patients are altered following busulfan administration. The PCA plot failed to show any visible difference between the 2-week pre-busulfan and the pre-busulfan time points, suggesting that the primary changes in the plasma metabolome are associated with busulfan administration rather than with merely the passage of time. Changes in a large number of EMCs were observed between pre-busulfan and last busulfan PK samples (by ANCOVA contrast, 542 EMCs exhibited an unadjusted P<0.05).

Pharmacometabonomics

In addition to the association with busulfan administration, perMANOVA testing found that metabolite levels in all (i.e., 2-week pre-busulfan, pre-busulfan, and last busulfan PK) samples collectively were associated with relapse (p =.005) and with acute GVHD (p =.001). Samples from the same participant within the combined 2-week pre-busulfan and pre-busulfan group tended to be more similar to each other than to samples from other individuals in the combined pre-busulfan group. These results suggest high inter-participant variability in the plasma metabolome.

Although our statistical analyses showed limited results for an association between the difference in EMC profiles and relapse (Figure 1) or acute GVHD (Figure 2), Partial Least Squares-Discriminant Analysis (PLS-DA) plots provide some evidence that a supervised selection of components can partially distinguish profiles based on these two outcomes (Figure 3 and 4, respectively).

Figure 1. Box plots of pre-busulfan plasma EMCs associated with relapse.

Box plots represent mean (interquartile range) of centered-log ratio (CLR)-transformed EMC abundances in relapse and non-relapse groups. EMCs shown had a FDR-adjusted p value of <.5. * indicates a compound identified at a lower level of confidence, as described in Results.

Abbreviations. EMC: endogenous metabolomic compounds

Figure 2. Box plots of pre-busulfan plasma EMCs associated with acute GVHD.

Box plots represent mean (interquartile range) of centered-log ratio (CLR)-transformed EMC abundances in grade 0–1 acute GVHD event and grade 2–4 acute GVHD event groups. EMCs shown had a FDR-adjusted p value of <.2. * indicates a compound identified at a lower level of confidence, as described in Results. Abbreviations. EMC: endogenous metabolomic compound; GVHD: graft-versus-host disease

Figure 3: Partial Least Squares-Discriminant Analysis (PLS-DA) of relapse.

A) 2D scores plot of the first two components calculated with PLS-DA for predicting relapse using pre-busulfan samples. The variance explained by each component is given in parentheses. B), Parameters of the PLS-DA model with 1, 2, 3, 4, and 5 components. PLS-DA was performed using MetaboAnalyst version 4.0. EMC abundance data were centered log-ratio (CLR) transformed prior to PLS-DA. Abbreviations. EMC: endogenous metabolomic compound

Figure 4: Partial Least Squares-Discriminant Analysis (PLS-DA) of acute GVHD.

A) 2D scores plot of the first two components calculated with PLS-DA for predicting acute GVHD grade using pre-busulfan samples. The variance explained by each component is given in parentheses. B), Parameters of the PLS-DA model with 1, 2, 3, 4, and 5 components. PLS-DA was performed using MetaboAnalyst version 4.0. EMC abundance data were centered log-ratio (CLR) transformed prior to PLS-DA. Abbreviations. EMC: endogenous metabolomic compound

In the univariate analysis for relapse, 31 EMCs in the pre-busulfan samples exhibited an unadjusted p value less than .05 (12 in the positive and 19 in the negative direction; Figure 3; Table S4, Figure S3); however, none of these EMCs were significant after correcting for FDR. In similar analyses for acute GVHD, 53 EMCs in the pre-busulfan samples exhibited an unadjusted P value less than .05 (16 positively and 37 negatively; Figure 4, Table S5 and Figure S4).

The pathway enrichment analysis was performed on the pre-busulfan samples considering the EMCs in each pathway together. The top two pathways exhibited FDR-adjusted P =.500 (unadjusted P <.05) for an association with relapse, and both of them had a pathway impact factor > 0.5 (Table 2, Figure 5). Fifteen EMCs from our analysis were included in the top pathway, cysteine and methionine metabolism (Figure 6). This result is consistent with the fact that cysteine, methionine, and several related EMCs underwent significant changes in abundance as a result of busulfan administration. Two pathways, the arginine and ornithine metabolism pathway and the arginine and proline metabolism pathway, showed some evidence of an association with acute GVHD (each exhibited pathway impact ≥.5 and raw P<.05) but neither met our criteria for FDR-adjusted significance (Table 2, Figure 7, Figure 8).

Table 2. Pathway enrichment analysis of relapse and acute GVHD:

Top pathways, significance, and impact from pathway enrichment analyses, sorted by increasing P-values. Only those pathways with a p-value <.05 that have over 5% of the pathway matched within our dataset are shown. Data also shown in Figures 5 and 7.

Outcome	Number of outcome eventsa	Pathway Name	Total EMCsb	Matched EMCsc	P- value	−log(P)d	FDR Pe	Impactf
relapse	10	Cysteine and methionine metabolism	56	15	0.0186	3.98	0.500	0.54
relapse	10	Glycine, serine and threonine metabolism	48	17	0.0434	3.14	0.500	0.53
acute GVHD	54	D-Arginine and D-ornithine metabolism	8	3	0.0278	3.58	0.907	0.50
acute GVHD	54	Arginine and proline metabolism	77	20	0.0497	3.00	0.907	0.52

^aOutcome events are relapse and grade 2-4 acute GVHD

^bTotal number of EMCs in the pathway

^cNumber of matched EMCs, explained in Statistical Analysis section

^d−log(P) is the negative natural log of the P value for each pathway shown in Figures 5 and 7

^eFalse Discovery Rate (Benjamini-Hochberg)-adjusted P-value

^fImpact is the pathway impact value on relapse calculated from pathway topology analysis

Figure 5. Pathway enrichment analysis of relapse.

All dots represent matched pathways from pathway topology analysis. Pathways are colored according to their P values from pathway enrichment analysis, with gradations from yellow – having the largest P – to red – having the lowest P (exact P values are given in Tables 2 and S3). Pathways above the horizontal red line correspond to p<.05. Pathway impact is indicated on the x-axis. All pathways with p<.05 and a pathway impact value >0 are labeled. None of these pathways included EMCs with missing values in this dataset.

Figure 6. Cysteine and Methionine Metabolism pathway, the pathway with the strongest association with relapse.

Each box represents an EMC in the KEGG pathway. Colored EMCs were significant in the pathway analysis with P < .05. Red boxes indicate EMCs that were elevated in the relapse group; blue boxes indicate EMCs that were decreased in the relapse group. C00019: S-adenosylmethionine; C00021: S-adenosylhomocysteine; C00022: pyruvic acid; C00041: L-alanine; C00049: L-aspartic acid; C00051: glutathione; C00059: sulfate; C00065: L-serine; C00073: L-methionine; C00094: sulfite; C00097: L-cysteine; C00109: 2-ketobutyric acid; C00155: L-homocysteine; C00170: 5’-methylthioadenosine; C00263: L-homoserine; C00283: hydrogen sulfide; C00320: thiosulfate; C00409: Methanethiol; C00441: L- aspartate-semialdehyde; C00606: 3-sulfinoalanine; C00793: D-cysteine; C00957: 3-mercaptopyruvic acid; C00979: O-acetylserine; C01005: phosphoserine; C01077: O-acetyl-L-homoserine; C01118: O-succinyl-L-homoserine; C01137: S-adenosylmethioninamine; C01180: 2-oxo-4-methylthiobutanoic acid; C01234: 1-aminocyclopropane-1-carboxylate; C01817: DL-homocystine; C01962: thiocysteine; C02218: 2-aminoacrylic acid; C02291: L-cystathionine; C03082: L-aspartyl-4-phosphate; C03089: 5-methylthioribose; C03145: N-formyl-L-methionine; C03539: S-ribosyl-L-homocysteine; C04188: 5-methylthioribose 1 phosphate; C04582: 5-methylthioribulose 1 phosphate; C05524: aminoacyl-L-methionine; C05526: S- glutathionyl-L-cysteine; C05527: 3-sulfinylpyruvic acid; C05528: 3-sulfopyruvic acid; C05823: 3-mercaptolactic acid; C05824: cysteine-S-sulfate; C06547: ethylene; C08276: 3-methylthiopropionic acid; C09306: sulfur dioxide; C11481: hydrogen sulfite; C15650: 2,3-diketo-5-methylthiopentyl-1-phosphate; C15651: 2-hydroxy-3-keto-5-methylthiopentenyl-1-phosphate; C15606: 1,2-dihydroxy-3-keto-5-methylthiopentene; C16069: 3-sulfolactate.

Figure 7. Pathway enrichment analysis of acute GVHD, comparing grade 0–1 and grade 2–4 acute GVHD groups.

All dots represent matched pathways from pathway topology analysis. Pathways are colored according to their P values from pathway enrichment analysis, with gradations from yellow – having the largest P – to red – having the lowest P (exact P values are given in Tables 2 and S4). Pathways above the horizontal red line correspond to p<.05. Pathway impact is indicated on the x-axis. All pathways with p<.05 and a pathway impact value >0 are labeled. This pathway did not include EMCs with missing values in this dataset.

Figure 8: D-arginine and D-Ornithine metabolism pathway, the pathway with strongest association with acute GVHD.

Each box represents an EMC in the KEGG pathway. Colored EMCs were significant in the pathway analysis with P <.05. Red boxes indicate EMCs that were elevated in the grade 2-4 acute GVHD group; blue boxes indicate EMCs that were decreased in the grade 2-4 acute GVHD group. C00062: L-arginine; C00077: ornithine; C00515: D-ornithine; C00792: D-arginine; C01110: 5-amino-2-oxopentanoic acid; C03341: 2-amino-4-oxopentanoic acid; C03943: (2R,4S)-2,4-diaminopentanoate

Discussion

The key findings of this analysis are, of the 60 pathways with sufficient EMCs for analysis: 1), the cysteine/methionine pathway and the glycine, serine, and threonine metabolism pathway exhibited the strongest association with relapse; 2) the D-arginine and D-ornithine metabolism pathway and the arginine and proline metabolism pathway exhibited the strongest association with acute GVHD. Although these pathways did not exhibit statistical significance, our analysis suggests they deserve further investigation. In this study, we took a first step towards identifying plasma EMCs associated with clinical outcomes with the long-range goal of personalizing the choice of the HCT conditioning regimen, IV busulfan doses, or GVHD prophylaxis using biomarkers identified via pharmacometabonomics.

“Pharmacometabonomics” is the concept of personalized drug treatment using pre-dose metabolite profiling to predict drug response in individuals.^26-27 We chose to focus on the alkylating agent busulfan because of its frequent use in HCT conditioning and its narrow therapeutic index. Recent discoveries demonstrate that metabolomics is an important piece of the puzzle of personalized medicine and that EMCs influence organ function, immune function, nutrient sensing, and gut physiology.²⁸ In 75 HCT recipients,²⁹ altered pre-transplant levels of several immunoregulatory EMCs, including BCAA and tyrosine derivatives, were found among those who later developed GVHD. This led Reikvam et al to hypothesize that these EMCs may be involved in developing GVHD and suggests that HCT recipients with high levels of these EMCs may benefit from a stronger immunosuppressive regimen. These studies, as well as our own evaluating the association of the plasma metabolome with IV busulfan clearance,^30-31 suggest pharmacometabonomics may improve clinical outcomes in HCT recipients. Collectively, this work could lead to system-wide perspective of the allogeneic HCT biology wherein EMCs, proteins, and genes are understood to interact synergistically to modify the functions within the allogeneic HCT recipient. .

In the present study, we analyzed 763 EMCs representing over 60 pathways. We sought a global metabolomic assay that had sufficient representation of EMCs within glutathione and related pathways. For acute GVHD, pathway enrichment analysis revealed that D-arginine and D-ornithine metabolism and arginine and proline metabolism were the top pathways associated with grade 2 – 4 acute GVHD (Table 2, Figure 7 and Table S5). This contrasts the data in mouse models of HCT showing that early GVHD is associated with accumulation of cysteine, cystathione, and cysteinylglycine .⁷ For relapse, pathway enrichment analysis revealed that cysteine and methionine metabolism and glycine, serine and threonine metabolism were the top two pathways associated with post-transplant relapse (Table 2). Although cysteine and methionine metabolism is amongst 10 pathways suggested for further investigation in FMS-like tyrosine kinase 3-internal tandem duplication acute myeloid leukemia, no statistically significant associations between relapse and the cysteine and methionine metabolism pathway were found.³² No statistically significant associations between the glycine, serine, and threonine metabolism and relapse were found in pathway enrichment analyses. Glycine is an important component of glutathione, and glutathione is involved in busulfan metabolism.³³ Boxplots of plasma EMCs associated with relapse are shown in Figure 1, without a clear delineation in the EMC abundance between those who did or did not relapse.

Strengths of this work include the large population of over 75 HCT participants, a contemporary patient population receiving targeted busulfan, and the global panel providing high accuracy of EMC identification and relative abundances within relevant pathways. However, there are limitations worth noting. Importantly, there were few relapse events (N=10) and few patients with grade 0-1 GVHD (n=24). Future studies with larger sample sizes (and thus, more relapse and GVHD events) are needed, as they would allow for inclusion of risk factors for relapse (e.g., cytogenetics) or GVHD (e.g., HLA). These results show the feasibility of conducting metabolomics studies in allogeneic HCT, with the hope of gaining mechanistic insight into the pathophysiology of relapse and/or GVHD while improving clinical outcomes.

Conclusion

This work suggests that pre-busulfan plasma levels of EMCs in the cysteine and methionine metabolism pathway and the glycine, serine and threonine metabolism pathway may be associated with relapse in HCT patients receiving IV busulfan-based conditioning while EMCs in the D-arginine and D-ornithine metabolism pathway and the arginine and proline metabolism pathway may be associated with acute GVHD in HCT patients receiving IV targeted busulfan -based conditioning. Further studies, including those that subsequently interrogate the glutathione pathway in larger patient populations, are needed to substantiate these results which may improve the prospect of personalizing the HCT conditioning regimen and potentially improve clinical outcomes.

Supplementary Material

Supplemental Material

Supplemental methods: Pharmacometabonomics analysis.

Supplemental Figure 1 (Figure S1): Change in plasma endogenous metabolomic compounds (EMCs) with age.

Figure S2: Change in plasma EMCs from pre- busulfan administration (purple and teal dots) to last busulfan PK sample (yellow dots).

Figure S3: Volcano plot for relapse.

Figure S4: Volcano plot for acute graft versus host disease (GVHD).

Supplemental Table 1 (Table S1): Metabolites previously found to be associated with transplant outcomes.

Table S2: Timing of longitudinal pharmacometabonomic sample collection.

Table S3: Full list of measured EMCs.

Table S4: Univariate results with raw P <0.05 for relapse, sorted by increasing P-values.

Table S5: Univariate results with raw P < 0.05 for acute GVHD, sorted by increasing P-values

Abbreviations:

AUC

area under the plasma concentration-time curve

BH

Benjamini-Hochberg

^Tbusulfan

targeted (pharmacokinetic-guided) intravenous busulfan dosing

CLR

centered log-ratio

CY

cyclophosphamide

EMC

endogenous metabolomic compound

FDR

false discovery rate

FLU

fludarabine monophosphate

GSH

glutathione

GST

glutathione-S-transferase

GVHD

graft-versus-host disease

HCT

hematopoietic cell transplant

HLA

human leukocyte antigen

IV

intravenous

PCA

principal component analysis

perMANOVA

permutation ANOVA

PK

pharmacokinetics

PLS-DA

partial least squares-discriminant analysis

TBI

total body irradiation