Regulation of kynurenine metabolism by blood donor genetics and biology impacts red cell hemolysis in vitro and in vivo

Abstract

In the field of transfusion medicine, the clinical relevance of the metabolic markers of the red blood cell (RBC) storage lesion is incompletely understood. Here, we performed metabolomics of RBC units from 643 donors enrolled in the Recipient Epidemiology and Donor Evaluation Study, REDS RBC Omics. These units were tested on storage days 10, 23, and 42 for a total of 1929 samples and also characterized for end-of-storage hemolytic propensity after oxidative and osmotic insults. Our results indicate that the metabolic markers of the storage lesion poorly correlated with hemolytic propensity. In contrast, kynurenine was not affected by storage duration and was identified as the top predictor of osmotic fragility. RBC kynurenine levels were affected by donor age and body mass index and were reproducible within the same donor across multiple donations from 2 to 12 months apart. To delve into the genetic underpinnings of kynurenine levels in stored RBCs, we thus tested kynurenine levels in stored RBCs on day 42 from 13 091 donors from the REDS RBC Omics study, a population that was also genotyped for 879 000 single nucleotide polymorphisms. Through a metabolite quantitative trait loci analysis, we identified polymorphisms in SLC7A5, ATXN2, and a series of rate-limiting enzymes (eg, kynurenine monooxygenase, indoleamine 2,3-dioxygenase, and tryptophan dioxygenase) in the kynurenine pathway as critical factors affecting RBC kynurenine levels. By interrogating a donor-recipient linkage vein-to-vein database, we then report that SLC7A5 polymorphisms are also associated with changes in hemoglobin and bilirubin levels, suggestive of in vivo hemolysis in 4470 individuals who were critically ill and receiving single-unit transfusions.

Graphical abstract

Figure 1.

Metabolomics of the REDS RBC Omics recalled donor population. An initial screening for hemolytic propensity of RBCs at the end of storage (day 42) was performed on 13 403 index donors who were enrolled at 4 different American blood centers. Donors with the highest and lowest hemolytic parameters (5th and 95th percentile) were invited to donate a second unit of blood, and metabolomics analyses were performed on RBC units from 643 recalled donors on storage days 10, 23, and 42 (A). Unsupervised (principal component analysis [PCA]) (B), partially supervised (partial least square–discriminant analysis [PLS-DA]), (C) and hierarchical clustering analysis. Top 50 metabolites by time series analysis of variance in (D) showed a significant impact of storage duration and storage additive solutions on the metabolic variance across the 1929 samples tested in this study. Previously reported metabolic markers of the storage lesion were confirmed among the top variables increasing and decreasing over storage (E). uMAP confirmed a strong impact of storage duration and additives, though an additional subgroup was identified; 3-dimensional [3D] and 2D views (F-G). Top discriminants across these groups were AS components citrate (highest in AS-3) and mannitol (highest in AS-1 in this study) (H). ATP, adenosine triphosphate.

Figure 1.

Metabolomics of the REDS RBC Omics recalled donor population. An initial screening for hemolytic propensity of RBCs at the end of storage (day 42) was performed on 13 403 index donors who were enrolled at 4 different American blood centers. Donors with the highest and lowest hemolytic parameters (5th and 95th percentile) were invited to donate a second unit of blood, and metabolomics analyses were performed on RBC units from 643 recalled donors on storage days 10, 23, and 42 (A). Unsupervised (principal component analysis [PCA]) (B), partially supervised (partial least square–discriminant analysis [PLS-DA]), (C) and hierarchical clustering analysis. Top 50 metabolites by time series analysis of variance in (D) showed a significant impact of storage duration and storage additive solutions on the metabolic variance across the 1929 samples tested in this study. Previously reported metabolic markers of the storage lesion were confirmed among the top variables increasing and decreasing over storage (E). uMAP confirmed a strong impact of storage duration and additives, though an additional subgroup was identified; 3-dimensional [3D] and 2D views (F-G). Top discriminants across these groups were AS components citrate (highest in AS-3) and mannitol (highest in AS-1 in this study) (H). ATP, adenosine triphosphate.

Figure 2.

Metabolic markers of the storage lesion are poor correlates to hemolytic propensity in the REDS RBC Omics recalled donor cohort. (A-C) 3D uMAPs are color coded as a function of storage (spontaneous), oxidative, or osmotic hemolysis in the REDS RBC Omics recalled donor cohort (z-axis represent storage day, with uMAP1 and 2 calculated via uMAP). (D-F) Volcano plots indicate metabolite associations to the same hemolysis parameters, with the x-axis indicating the Spearman-determined correlation between hemolysis and metabolite levels and the y-axis indicating the −log10 of the false discovery rate–corrected P values for such associations). (G) Line plots indicate the donors in the 5th and 95th percentile (red and blue lines) for a selected metabolic markers of the storage lesion across the 3 time points (storage day 10, 23, and 42) tested for each donor (n = 643).

Figure 3.

Kynurenine is a marker of osmotic fragility of stored RBCs and is affected by donor age, BMI, and sex. Machine-learning approaches were used to determine metabolic predictors of osmotic fragility of stored RBCs. Results identified kynurenine as the top predictor (A), which was confirmed via a Spearman correlation analyses (B). (C) Scatterplot of osmotic fragility vs kynurenine: x-axis and y-axis, respectively. (D-I) Volcano plots, scatter plots, or box and whisker plots showing the strong positive association between kynurenine levels and donor age, sex, and BMI. AC, acyl-carnitine; AU, arbitrary unit.

Figure 3.

Kynurenine is a marker of osmotic fragility of stored RBCs and is affected by donor age, BMI, and sex. Machine-learning approaches were used to determine metabolic predictors of osmotic fragility of stored RBCs. Results identified kynurenine as the top predictor (A), which was confirmed via a Spearman correlation analyses (B). (C) Scatterplot of osmotic fragility vs kynurenine: x-axis and y-axis, respectively. (D-I) Volcano plots, scatter plots, or box and whisker plots showing the strong positive association between kynurenine levels and donor age, sex, and BMI. AC, acyl-carnitine; AU, arbitrary unit.

Figure 4.

Protein and metabolite correlates to kynurenine levels in the REDS RBC Omics recalled donor cohort. (A) For the 643 recalled donor units (storage day 10, 23, and 42), proteomics and metabolomics data were generated to identify omics correlates to kynurenine levels. (B) Results indicate an enrichment between kynurenine levels and complement and fibrinogen components during a depletion in the levels of RBC structural and functional proteins. (C) At the metabolic level, increases in kynurenine were accompanied by elevation of all intracellular amino acids, especially aromatic amino acids. (D) Pathway analysis of combined metabolomics and proteomics data indicates an enrichment in components involved in viral responses, complement, and coagulation. CFD, complement factor D; CFH, complement factor H; FGA, fibrogen alpha chain; FGB, fibrinogen beta chain; FGG, fibrogen gamma chain; HPX, hemopexin.

Figure 5.

Measurement of kynurenine levels in 13 091 donors from the REDS RBC Omics index cohort. (A) Kynurenine measurements were performed on 13 091 end-of-storage (day 42) pRBC units from the REDS RBC Omics index donor cohort. (B) Kynurenine levels were found to be nonnormally distributed in this population. (C) 3D uMAP of the 13 091 donors, based on metabolomics analyses (informing uMAP1 and uMAP2) and kynurenine levels (z-axis and color scheme). This analysis highlights a subset of donors with significant elevation in kynurenine levels compared with the rest of the population. (D) Line plots show kynurenine levels (y-axis) as a function of donor age (x-axis), either alone (top) or as a function of donor sex (middle) or BMI (bottom). (E) Kynurenine levels rank among the top correlates to osmotic fragility in the index donor cohort. (F) Correlation analyses were performed for end-of-storage (day 42) kynurenine levels of the 643 donors who were screened both at index and recalled donation (2 independent blood units). Results indicated a significant level of reproducibility for kynurenine within the same donor across multiple donations in panel F. Kyn, kynurenine.

Figure 6.

Interdonor heterogeneity in kynurenine levels is partly explained by genetic regulation. Kynurenine measurements in identical monozygotic twins vs nonidentical dizygotic twins (A) shows potential heritability of kynurenine levels, with stronger twin-twin correlations in the former than in the latter group (B-C). (D) mQTL analysis was performed to determine the genetic polymorphisms associated with interdonor heterogeneity in the end-of-storage kynurenine levels in the REDS RBC Omics index donor cohort (n = 13 091). (E) The 3D uMAP overlaps homozygous state for the dominant vs recessive SNP (SLC7A5, rs8052118, and intronic) associated with significant elevation in RBC kynurenine levels. (F) Volcano plot shows log2 fold-changes between the 2 groups. (G-H) Volcano plot of the Spearman association of metabolite levels to the dosage of the most significant SNP associated with kynurenine levels. (I) The lowest kynurenine levels were observed in donors who were homozygous dominant for SLC7A5 and recessive for ATXN2. FA, fatty acid; HD, homozygous dominant; HR, homozygous recessive; KMO, kynurenine monooxygenase; TDO, tryptophan dioxygenase.

Figure 7.

Elevated kynurenine levels and associated genetic polymorphisms result in increased extravascular hemolysis and lower hemoglobin increments in murine models of storage and PTR and in patients who receive transfusion in the clinics. Storage and PTR studies were performed in 350 mice with extreme diverse genetics (JAX Diversity Outbred mice). (A) Results show a significant (P < .001) negative association between kynurenine levels and posttransfusion recoveries in mice. (B) Leveraging a vein-to-vein database, we linked kynurenine levels and associated SNP for SLC7A5 to hemoglobin and bilirubin increments in 4470 single-unit transfusion events in patients requiring transfusion from products donated by REDS RBC Omics index donors. Results indicate significantly lower hemoglobin increments (B-C) and higher bilirubin levels (D) in patients receiving transfusion with donors carrying the SLC7A5 rs8052118 SNP in homozygosity. CI, confidence interval; GFP, green fluorescence protein; HoD, hen egg lysozyme-ovalbumin-Duffy; Ref, reference.