Evaluating immune response and metabolic related biomarkers pre-allogenic hematopoietic stem cell transplant in acute myeloid leukemia

Although hematopoietic stem cell transplantation (HCT) is the only curative treatment for acute myeloid leukemia (AML), it is associated with significant treatment related morbidity and mortality. There is great need for predictive biomarkers associated with overall survival (OS) and clinical outcomes. We hypothesized that circulating metabolic, inflammatory, and immune molecules have potential as predictive biomarkers for AML patients who receive HCT treatment. This retrospective study was designed with an exploratory approach to comprehensively characterize immune, inflammatory, and metabolomic biomarkers. We identified patients with AML who underwent HCT and had existing baseline plasma samples. Using those samples (n = 34), we studied 65 blood based metabolomic and 61 immune/inflammatory related biomarkers, comparing patients with either long-term OS (≥ 3 years) or short-term OS (OS ≤ 1 years). We also compared the immune/inflammatory response and metabolomic biomarkers in younger vs. older AML patients (≤30 years vs. ≥ 55 years old). In addition, the biomarker profiles were analyzed for their association with clinical outcomes, namely OS, chronic graft versus host disease (cGVHD), acute graft versus host disease (aGVHD), infection and relapse. Several baseline biomarkers were elevated in older versus younger patients, and baseline levels were lower for three markers (IL13, SAA, CRP) in patients with OS ≥ 3 years. We also identified immune/inflammatory response markers associated with aGVHD (IL-9, Eotaxin-3), cGVHD (Flt-1), infection (D-dimer), or relapse (IL-17D, bFGF, Eotaxin-3). Evaluation of metabolic markers demonstrated higher baseline levels of medium- and long-chain acylcarnitines (AC) in older patients, association with aGVHD (lactate, long-chain AC), and cGVHD (medium-chain AC). These differentially expressed profiles merit further evaluation as predictive biomarkers.

Metabolic changes are part of the aging process. Aging related dysregulation of inflammatory/immune responses occurs in tandem with metabolomic dysregulation. However, the mechanisms are poorly understood [14]. It has been reported that changes in such circulating, energy-related metabolites as acylcarnitines, carbohydrates, and amino acids (AA), can be associated with age, BMI and insulin resistance. Metabolic markers such as adiponectin, glycine, nonessential AA, and relative proportions of circulating large, neutral AAs (LNAAs) and medium-chain acylcarnitines have been suggested as markers of metabolic health. For example, higher plasma concentrations of glycine have been reported to be associated with better metabolic health [14,[18][19][20].
We hypothesized that circulating metabolic, inflammatory, and immune molecules have potential as predictive biomarkers for AML patients who receive HCT treatment. We have studied those biomarkers in AML-HCT patients who have shorter vs longer OS (OS of less than one year (OS �1) or more than three years (OS �3). We compared the blood based biomarkers and metabolomics profiles in younger vs. older AML patients (�30 years vs. patients � 55 years). We also analyzed the biomarker and metabolomics profiles for their association with clinical outcomes, namely OS, chronic graft versus host disease (cGVHD), acute graft versus host disease (aGVHD), infection and relapse.

Patient population
We retrospectively identified patients with AML who underwent HCT, and had baseline (pre-HCT) EDTA plasma samples. The samples were stored according to our Duke Health Institutional Review Board (IRB)-approved protocols (IRB# Pro00006268 and Pro00100650). Our protocols involved obtaining witnessed informed consent for sample collection for future research purposes, as well as use of associated clinical data.
Fully anonymized samples were thawed, aliquoted, refrozen and stored at −80˚C until tested. All samples underwent one freeze-thaw cycle before ELISA analysis. In order to compare biomarkers in older vs. younger patients and those with good vs. poor OS outcomes, we selected samples based on patient age (�30 years vs. � 55 years old) and OS outcomes (OS � 1 years vs. OS� 3 years). ( Table 1). The age cut-offs were arbitrary selection and mostly based on what we had samples for. Though there is support that AML in patients >55years behaves differently than AML in patients <55 years [5].

Biomarkers of metabolism
We evaluated 65 metabolic biomarkers including: amino acids (N = 15), acylcarnitines (N = 45), and conventional clinical analytes (N = 5). Amino acids and acylcarnitines were analyzed by flow injection electrospray-ionization tandem mass spectrometry and quantified by isotope or pseudo-isotope dilution using methods described previously [31,32]. Conventional analytes, including non-esterified fatty acids (NEFA), triglycerides, glycerol, 3-hydroxybutyrate, and lactate and were measured using a Beckman DxC 600 clinical analyzer (Brea, CA). Reagents for 3-hydroxybutyrate and NEFA were from Wako (Mountain View, CA), and those for lactate and triglycerides were from Beckman (Brea, CA). We measured free glycerol using the initial absorbance in the triglycerides assay, a signal that is normally blanked in the test procedure preceding the addition of lipase.

Statistical analysis for biomarker studies
The 54-plex immune/inflammatory response biomarkers, the additional plasma biomarkers, and metabolomic biomarkers were analyzed using Cox Proportional Hazard models with OS as the response, and age groups, baseline biomarkers, and their interactions as predictors. Corresponding to the Cox PH model, p-values were reported from Chi-square test with df = 2 [33]. Box plots were used to illustrate the variability of the markers for association with clinical outcomes. Boxes represent 25th (Q1) and 75th (Q3) percentiles; Horizontal lines indicate the medians; Upper whiskers indicate maximum; Lower whiskers indicate minimum; Points indicate any observations outside the whiskers. Expression levels were log2-transformed and analyzed as continuous measures (Please see S1 File for the data underlying the findings). OS groups with survival of equal and less than one year or more than three years (OS � 1 years or OS � 3 years) were used as the output in logistic regression models. OS time was defined as time from transplant to death/last follow-up. Baseline samples' biomarker levels in different age and OS groups were analyzed using a Wilcoxon rank-sum test [34]. Effect sizes of each biomarker for aGVHD, cGVHD, relapse and infection were assessed using logistic regression models. Odds ratios (OR), score test p-values, and 95% CI were reported for the logistic regression models assessing biomarker effects on clinical outcomes (aGVHD, cGVHD, relapse and infection). Multiple comparison was addressed within a framework of control of False Discovery Rate (FDR) using the method by Storey [35,36]. However, due to the limited sample size and number of markers analyzed, we used 0.1 as the significant q-value cutoff.
All analyses were performed using the R Statistical Environment [R], [6] and extension packages from CRAN [37,38]. The analyses were conducted with adherence to the principles of reproducible analysis using the knitr package for generation of dynamic reports [39].
One baseline plasma marker was significantly associated with cGVHD: fms-like tyrosine kinase 1 (Flt-1, also known as VEGFR-1, OR = 1.71, 95% CI (1.13, 2.78), p-value = 0.017). Patients who developed cGVHD post-HCT had higher baseline expression levels of Flt-1 (Fig 3E). The Wilcoxon rank-sum test was used to compare biomarker levels of baseline in different age groups (�30 years vs. � 55 years old). Data is presented as a boxplot. Of the n = 18 blood-based biomarkers with significant p-value, only IL23 was significantly higher in younger patients (age <30 years) compared to older patients (age >55 years). The rest of the markers had higher baseline levels in older patients (for the full list of the 18 markers please refer to Two baseline plasma markers were significantly associated with aGVHD: interleukin 9 (IL-9, OR = 0.37, 95% CI (0.16, 0.71), p-value = 0.009); and eotaxin-3 (also known as C-C motif chemokine ligand 26 (CCL26), OR = 0.35, 95% CI = (0.11, 0.84), p-value = 0.040). Patients who developed aGVHD post-HCT had lower baseline expression levels of both markers (Fig 3F and 3G).

Metabolic biomarkers
The aging process is a complex, characterized by physical, molecular, immune, and metabolic changes that can cause functional decline [24,28]. In addition to evaluating plasma biomarkers of immune and inflammatory response, we have 65 metabolites, including: amino acids (N = 15), acylcarnitines (N = 45), and conventional clinical analytes (N = 5). Markers were evaluated for association with age, OS and clinical outcomes (aGVHD, cGVHD, relapse and infection).
Association between baseline metabolic biomarkers and clinical outcomes (aGVHD, cGVHD, relapse and infection). No metabolic marker showed significant association with survival for AML/HCT patients.
Compared to younger patients, baseline levels of various medium-and long-chain acylcarnitines [42] (p<0.05) were higher in older HCT patients. (Fig 4 and metabolomics S2 File).

Fig 2. Association between OS and baseline biomarker.
We have evaluated the association between components of the blood based biomarkers' baseline levels and OS. Data is presented as a boxplot. OS groups (OS � 1 years or OS � 3 years) were used as the output in logistic regression models. Logistic regression with OS groups as response, age groups, marker and their interaction as predictors was performed. The function was used to provide p-values of testing whether marker has significant effect on OS groups in either age group. Three markers (CRP, SAA and IL13) showed significant association with survival. Baseline levels are lower in all the three markers in patients with longer OS (OS� 3) years. https://doi.org/10.1371/journal.pone.0268963.g002

Discussion
Age related inflammation, also termed "inflammaging" is related to the activation/dysregulation of both innate and adaptive immune systems and is considered a significant risk factor in many age-related diseases [13,16]. The exact mechanism of inflammaging and its contribution to adverse health outcomes are mostly unknown [16]. However, numerous studies have shown that several pro inflammatory cytokines, including IL-6 and TNFα increase with age in healthy individuals and in the absence of infection [43][44][45].
Of the 61 immune and inflammatory response related biomarkers we investigated, 17 were increased in older patients and only one marker (IL-23) was increased in younger patients pre-HCT compared to older patients. Unsurprisingly, most of the biomarkers elevated in elderly patients were inflammatory cytokines, such as IL16, IL17, MIP1a, TNF-α, and TNF receptors TNFRI and TNFRII. In addition, TNF-α an essential signaling protein in the innate and adaptive immune systems, is considered a biomarker of HCT treatment toxicity and a key cytokine in the effector phase of GVHD. TNF inhibitors have shown efficacy in clinical and experimental models of GVHD [46,47]. Surprisingly, IL-23 levels were increased in younger patients compared to older patients. The differing expression levels of IL-23 and IL-17 were unexpected, since IL-23 and IL-17 are closely intertwined. IL-23 is known to drive promotion of T helper type 17 (Th17) cells that secrete IL-17 [48]. This signaling pathway is critical in autoimmune conditions such as psoriasis and rheumatoid arthritis, and therapeutics targeting IL-17 or IL-23 are under clinical investigation [49]. In addition, ustekinumab, an antibody targeting IL-23 was reported to be effective in glucocorticoid-refractory aGVHD [50]. Other potential targets for aGVHD treatment involving the IL-23/IL-17 pathway, include the bromodomain and extra-terminal domain (BET) proteins. In vitro and in vivo assays [51], have demonstrated the potent anti-inflammatory effects of BET inhibition, and its effect on impacting IL-23R/IL-17 immune axis (decreasing the expression levels of IL23-R and IL17). BET inhibitors (Plexxikon-51107 and -2853 (PLX51107 and PLX2853)) have demonstrated that they can significantly improves survival and reduces aGVHD progression. PLX51107 will be studied in a phase Ib/II clinical trial for its effects on in treating Steroid-Refractory aGVHD (NCT04910152) [51].
Baseline levels of three markers (CRP, SAA and IL13) showed significant association with survival. Basal levels were lower in all the three markers in HCT patients with longer OS (OS� 3 years). IL-13 is involved in Th2 inflammation and higher pre transplant levels of IL13 have been reported as a strong predictor of developing aGVHD [52]. IL13 and its receptors have been evaluated as a possible therapeutic target in different diseases including various types of cancer [52][53][54][55]. Clinically, CRP is frequently examined broad-scale in a variety of disease states. CRP, C-reactive protein, is often measured as a broad-scale marker of inflammation in infection and autoimmune/rheumatologic conditions [56]. SAA or serum-amyloid A is a group of apolipoproteins associated with high density lipoprotein (HDL) that is both expressed constitutively and in response to inflammation [57]. Additionally, SAA has been implicated in the suppressive effects of tumor cells on immune cells in both melanoma and glioblastoma models [58,59].
Similar to CRP, D-dimer is an established biomarker of a number of disease states. Ddimer, a product of blood clot degradation, is most often measured when there is suspicion of pulmonary embolism or deep-vein thrombosis, but it can also be elevated in infection, inflammation, pregnancy, trauma, and malignancy [60]. The association between elevated D-dimer and infection post-HCT observed here might suggest a chronic inflammatory state, which combined with the immunosuppression after bone marrow transplant might lead to increased inability to fight off infection. CRP, D-dimer and SAA are notable for being highly upregulated during the "acute phase response"-the body's physiological reaction to stresses such as infection, inflammation, and trauma [61].
Of the 65 metabolomic biomarkers analyzed, we noted that lower mean levels of lactate were associated with aGVHD. A growing body of clinical evidence indicts dysregulation of lactate-pyruvate fuel metabolism in poor clinical outcomes after HCT [62,63]. Lactate is the product of glycolysis and substrate for mitochondrial respiration. It has a variety of roles, including its importance as an energy source, a precursor for gluconeogenesis, and an essential signaling molecule [64]. Critically, T cells rely upon glycolysis and lactate to perform their effector functions. In fact, 13C-pyruvate MRI has been used to detect aGVHD in a mouse model of HCT. In that study, plasma levels of lactate were significantly elevated at day 7 post-HCT after allogenic but not syngeneic transplants [65]. This timing of lactate elevation could explain why we observed that a lower baseline level of lactate was associated with aGVHD.
Acylcarnitines also play an important role in energy metabolism by participating in the transfer of fatty acids into mitochondrial membrane [42,66]. Acylcarnitines, esters of L-carnitines and fatty acids, are important intermediates in metabolism. They are responsible for transporting long chain fatty acids across the mitochondria for beta-oxidation. They have been used as biomarkers for inherited metabolic disorders since dysfunction in fatty acid or amino acid metabolism can lead to changes in plasma acylcarnitine concentration [67]. Our previous studies have shown that long-chain acylcarnitines, and carnitine esters of dicarboxylic acids are associated with cardiovascular disease, including heart failure [68][69][70]. Acylcarnitines can mediate inflammation [71], and have been proposed as biomarkers for hepatocellular carcinoma and liver dysfunction since the liver is the most active organ for acylcarnitine synthesis and metabolism [72].
Our results suggest that baseline levels of medium-and long-chain AC markers were higher in older HCT patients compared to younger HCT patients. In patients with aGVHD compared to patients who did not develop aGVHD, the baseline levels of certain long-chain acylcarnitines were lower. Baseline levels of the short chain AC marker, glutaryly carnitine, C5-DC [42], a marker of omega oxidation and oxidative stress, demonstrated association with cGVHD: patients with cGVHD had higher levels of C5-DC. Accumulation of medium-and long-chain acylcarnitines in older patients suggests incomplete beta-oxidation of fatty acids, resulting in lower flux of fatty hydrocarbon toward acetylcarnitine (AC C2) and, hence, acetyl CoA. To maintain overall energy balance, glycolytic pyruvate metabolism might shift away from lactic acid and toward production of acetyl CoA and Krebs cycling [73,74].
Our data show a significant decrease in lactate with aGVHD, as they also do decreases in various ACs. This could potentially signify a pathway emphasis towards Krebs (and oxidative phosphorylation) and away from lactate or fatty acid synthesis, or alternatively a drive in the opposite direction through pyruvate towards gluconeogenesis. In order to have the data and conclusion, to support such claims further in-depth studies are needed.
Strengths of this exploratory study include the large number of biomarkers (>100 total) analyzed from patient samples, the long-term survival data, and similar pre-HCT and peri-HCT conditioning treatment. The major limitation of this study is the small cohort size. The small sample size and the large number of the markers will not allow for the adjusted p-value analysis. Other limitations include disparate cohort sizes, lack of matched healthy control data and reliance on single institution data. In addition, we have only measured the baseline levels of the markers, which did not evaluate the longitudinal changes of these markers (including known GVHD markers REG3 and ST2, that did not show significant association in our baseline analysis). As previously stated this retrospective study was designed with an exploratory and hypothesis-generating approach to comprehensively characterize immune, inflammatory, and metabolomic biomarkers. Thus, biomarkers highlighted here should be subjected to further analyses.
In conclusion, we have identified several inflammatory and metabolomic biomarkers that are differentiate young vs old HCT patients and are associated with survival and clinical outcomes in HCT. Many of these biomarkers have been previously associated with malignancy or immune responses. Given the generally poor outcomes for older patients with AML, additional investigation is warranted. A combination of biomarkers could guide interventions, and personalized immune-modulatory therapeutics post-HCT could help to improve clinical outcomes in AML.