BLVRB redox mutation defines heme degradation in a metabolic pathway of enhanced thrombopoiesis in humans

Megakaryocytopoiesis and proplatelet formation represent progressively linked stages of hematopoietic stem cell development that maintain the normal circulating pool of platelets,¹  cells critical to normal hemostasis, pathologic thrombosis, and host adaptive immunologic responses.^2,3  Platelet generation (∼1 × 10¹¹ cells daily) is largely controlled by the thrombopoietin (TPO)/cellular myeloproliferative leukemia virus (c-MPL) axis, derived by megakaryocyte (MK) commitment form common bipotent MK-erythrocyte progenitors (MEPs).⁴  Although MKs are reduced in MPL-deficient mice, animals still produce MKs and platelets, implying that hematopoietic stem cells (HSCs) maintain the capacity for lineage fate in the absence of MPL.⁵  Transcription factors including GATA-1, GATA-2, FOG1/ZFPM1, RUNX1, and NFE2 are important for MK development,²  but none exclusively specify MK fate.¹  Whereas human blood counts have a heritable component, known genetic loci account for ∼5% of platelet variability,^6,7  highlighting the considerable knowledge gap of genetic pathways regulating physiologic and pathologic thrombopoiesis.

In this study, we applied large-scale platelet transcriptomic sequencing to unmask heme degradation in a novel metabolic pathway controlling megakaryocytopoiesis and platelet production. Whereas nearly 80% of organismal heme is found in circulating erythrocytes, heme degradation is largely studied in the context of reticulendothelial cell-mediated detoxification, with no prior evidence that tetrapyrroles (or their processing enzymes) engage in regulatory functions governing hematopoiesis. Defective oxidation/reductase (redox) activity involving biliverdin IXβ reductase (BLVRB) provides novel insights into the accumulation of reactive oxygen species, with implications for metabolic and bioenergetic determinants of lineage fate. In principle, redox-selected inhibitors represent novel biochemical approaches retaining potential for modulating human platelet counts.

All subjects (myeloproliferative neoplasm [MPN; N = 36], reactive thrombocytosis [RT; N = 53], or healthy controls [N = 208]) were enrolled in an institutional review board–approved protocol conducted in accordance with the Declaration of Helsinki,⁸  applying well-established clinical criteria for phenotypic classification.^9,10  Informed consent was obtained from all subjects. Large-scale platelet RNA transcriptomic studies and single nucleotide variant (SNV) identification were completed using the Illumina HiSequation 2000 platform (100 ng RNA/sample); our strategy involved single-end reads and was restricted to nonsynonymous SNVs (nsSNVs) to the exclusion of alternative splicing defects and/or insertion/deletions (in/dels); nsSNVs identified in ≥2 essential thrombocythemia (ET) samples (tier 1) were confirmed by dideoxy sequence analysis prior to expanded genotypic studies, which were completed using Illumina human 610 or 660W single nucleotide polymorphism (SNP) arrays (analyzed using GenomeStudio V2010.2 software). Five distinct genetic models (genotypic, allelic, trend, dominant, and recessive) were applied for the association analyses of each nsSNV (χ² test, Fisher’s exact test, Cochran-Armigage trend test), comparing different case-control groups with genotypic data available from (1) the 1000 Genomes Project Consortium,¹¹  (2) an internal subset of matched healthy controls, or (3) a cohort subset with RT.⁸  Statistical comparisons were completed using analysis of variance or Kolmogorov-Smirnov tests, and all statistical analyses were performed using R version 3.1.2.

Lentiviruses expressing BLVRB (Lv/BLVRB^WT), BLVRB^S111L (Lv/BLVRB^S111L), or empty virus control (Lv/Control) driven by the cytomegalovirus promoter and containing the puromycin-resistant cassette were generated at the Stony Brook Stem Cell Viral Vector Core and used for transduction of induced pluripotent stem cells (iPSCs) derived from CD34⁺ human umbilical cords (NCRM1), or human CD34⁺ hematopoietic stem cell assays as previously described.¹²  Suspension cultures were maintained in puromycin-selected SFEM II expansion medium supplemented with 50 ng/mL thrombopoietin (MK cultures) or 50 ng/mL thrombopoietin and 2 U/mL erythropoietin (bilineage erythroid/MK expansion).^12,13  Cell surface marking and flow cytometry were completed as previously described,^12-14  modified for intracellular reactive oxygen species (ROS) accumulation using the cell-permeant fluorogenic probe CellROX Green. Cellular RNA quantitation was performed using fluorescence-based real-time polymerase chain reaction (PCR) technology¹⁴  (oligonucleotide primers are provided in supplemental Table 3, available on the Blood Web site).

Specific activity determination of BLVRB^WT and BLVRB^S111L were completed on bacterially expressed recombinant enzymes at 25°C using flavin mononucleotide or pooled biliverdin (BV) dimethyl esters synthesized by coupled oxidation of heme¹⁵ ; RIPA-solubilized cytoplasmic lysates served as the source for cellular BLVRB functional assays,¹⁶  and immunodetection was completed using sheep anti-human BLVRB (R&D Systems; 1:100 dilution) and anti-actin monoclonal antibody (mAb) (EMD Millipore; 1:1000).¹⁴  Detailed methods are provided in supplemental Methods.

We applied large-scale platelet transcriptome sequencing (RNASeq) to cohorts with ET,⁸  an MPN subtype causally associated with thrombohemorrhage and characterized by gain-of-function mutations involving Janus kinase 2 (JAK2^V617F),¹⁷  or CALR in nonmutated JAK2.^18,19  To restrict our analyses to candidate genetic defects regulating thrombopoiesis (to the exclusion of platelet functional abnormalities), our strategy included a validation step using subjects with RT, a pathophysiologic subset infrequently associated with thrombohemorrhagic complications (refer to schema; Figure 1A); none of the RT cohort contained CALR or JAK2^V617F mutations. RNASeq of highly purified platelets from 7 ET subjects (4 harbored the JAK2^V617F mutation and 3 were genotypically normal [G/G]) was completed in parallel using 5 healthy controls, followed by an iterative algorithm to identify nsSNVs as causally plausible candidate genes (Figure 1B). Of the ∼350 000 SNVs, 186 high-quality nsSNVs were identified, of which 33 are qualified as tier 1 based on a stringent filtering step designed to exclude private mutations (supplemental Table 1); 86% of evaluable gene/SNPs had minor allelic frequencies <2%, validating the strategy for identifying rare novel modifier genes using relatively small sample subsets from a clonally expanded hematopoietic disorder. Expectedly, the candidate nsSNV list included JAK2^V617F, although neither the initial screen nor targeted visualization of previously described MPN defects²⁰  including MPL or CALR^18,19  were identified. This is not unexpected given the relative infrequency of these mutations or the limitations of our SNV search in identifying small in/dels typical of CALR mutations (in our expanded cohort analysis [see “BLVRB gene/SNV is a general thrombocytosis risk allele”], 3 of 9 JAK2^V617F-negative patients were subsequently found to be CALR mutated based on sizing assays). Other than JAK2, none of the gene/SNVs have been described as MPN modifiers or overlapped with genetic loci modulating MK/platelet⁶  or erythroid⁷  parameters in humans.

We explored the gene expression patterns of the candidate genes using an atlas of 38 distinct hematopoietic cell types²¹  (supplemental Table 2); 29 of the genes were represented on the Affymetrix gene array (Figure 1C), and candidate transcripts were enriched on average in early-stage MEP and granulocyte/monocyte progenitor cells compared with noncandidate transcripts in the same cell lineage (Figure 1D). These aggregate lineage expression patterns became more restricted in terminally differentiating MKs and erythroblasts (both early and late stages), with less exaggerated (but present) lineage enrichment in myeloid subsets (colony forming unit [CFU]-granulocyte and early-stage neutrophilic metamyelocytes). Limited expression was observed in lymphoid cells (B cells, T cells, and natural killer cells), collectively highlighting that the candidate genes are enriched in all 3 hematopoietic lineages encompassing the clonal evolution of MPNs and underrepresented (or excluded) in lymphoid cells whose genetic composition is generally considered germline in origin.

Genotypic studies of the 33-member SNVs using an expanded ET cohort (N = 36)⁸  followed by statistical association analyses using genotypic frequencies of healthy controls from the 1000 Human Genomes Project¹¹  or an independently genotyped cohort (N = 208), established that 5 SNVs (excluding JAK2^V617F) were associated with the ET phenotype (Table 1). Single mutations involving BLVRB (odds ratio [OR], 31.67; P = .006) or KTN1 (OR, 17.32; P = .0018) represented the strongest thrombocytosis risk alleles. The TM7SF3 SNV was a moderate-risk allele (OR, 2.93; P = .005), whereas SNVs involving QSOX1 (OR, 1.96; P = .06) and FAM40B/STRP2 (OR, 3.02; P = .09) only approached marginal statistical significance. The 5 nsSNVs were distributed almost evenly across the JAK2 allelic spectrum, with no statistical evidence by zygosity analysis for any association with JAK2 allelic burden (Figure 2A). To extend these studies, we established the independent (ie, driver) nature of these SNVs by genotyping a new subject cohort with reactive thrombocytosis (N = 53), a nonclonal disorder of exaggerated platelet production due to interleukin-6 (IL-6)–induced thrombopoietin (TPO) release²² ; this approach restricted the gene/SNVs to the subset functioning as modifiers of platelet production independent of JAK2^V617F and/or dominant subclones harboring additional molecular abnormalities.²⁰  Only BLVRB^S111L retained its significance as a thrombocytosis risk allele in both ET and RT cohorts (Figure 2B), imputing a fundamentally important thrombopoietic function independent of etiology. Although we did not distinguish between somatic and acquired mutation status, its rare minor allelic frequency in the general population (minor allelic frequency ≤0.005; Table 1) is most consistent with a germline predisposition SNV that is phenotypically unmasked during a hematopoietic stimulus.

Substratification analysis designed to dissect relative genotypic contributions on blood cell counts confirmed the importance of JAK2^V617F allelic burden on hemoglobin concentration (P = .004),^23,24  with a modulating effect in the subset coexpressing BLVRB^S111L (P = .07; Figure 2C). These results contrasted to those focusing on platelet counts, where JAK2^V617F allelic burden displayed minimal effect,^23,24  with a trend for exaggerated platelet counts in the BLVRB^S111L cohort (Figure 2D), although firm conclusions are limited by the small subset doubly heterozygous for JAK2^V617F/BLVRB^S111L.

Despite the near-universal presence of hemoproteins across phylogenetic development and their enrichment in erythroid cells (hemoglobin, cytochromes, etc), there is a paucity of information on heme catabolism in either developing or mature hematopoietic cells.²⁵  The BLVRB gene product BV IXβ reductase and its structurally distant homolog BLVRA (BV IXα reductase) function downstream of heme oxygenase(s)-1 (inducible HMOX1) and -2 (constitutive HMOX2) within the heme degradation pathway to catalyze reduction of BV IXα (or IXβ) tetrapyrrole(s) to the potent antioxidants bilirubin (BR) IXα and IXβ.^16,26  We studied expression patterns of the 4 heme degradation pathway genes along erythroid/MK (E/Meg) lineage development using data extracted from platelet¹⁴  and genetic atlases.²¹  Initial in silico analyses demonstrated a striking ∼40-fold induction of BLVRB, most pronounced during the terminal phases of erythroid formation (Figure 3A), a pattern that sharply contrasted to that of BLVRA, which remained generally stable (or diminished) throughout the same differentiation period. These patterns were recapitulated in erythropoietin (Epo)-induced CD34⁺ cultures, again demonstrating exaggerated BLVRB induction that preceded modest induction of HMOX1, with no change in HMOX2 or BLVRA expression (Figure 3B). Comparison of these patterns using Tpo (MK)-directed cultures demonstrated low-level BLVRA, HMOX1, and HMOX2 expression, with essentially no changes in transcript abundance throughout the 18-day culture period (Figure 3C). In contrast, BLVRB maintained the highest transcript abundance throughout the culture, with a predominant peak at an early (day 3/4) time point. Although MK BLVRB peak transcript abundance remained considerably less (ie, ∼10%) than that in late-stage erythroid progenitors, peak levels were comparable to (but preceding) those of the MK commitment marker platelet factor 4 (PF4)¹⁴  and remained higher than those of MPL.

These patterns of relative transcript abundance were generally maintained in normal platelets (BLVRB > BLVRA, HMOX1 > HMOX2) as established by extracted reads per kilobase of transcript per million scores and confirmatory quantitative PCR (qPCR; Figure 3D-E). In contrast, ET platelets demonstrated significantly decreased transcript abundance of 3 heme degradation pathway genes (BLVRA, HMOX1, and HMOX2); only the BLVRB transcript level remained comparable to that of healthy controls (vide infra). These aggregate data collectively implicate heme degradation pathway genes in developmentally restricted functions of E/Meg lineage commitment; importantly, the dichotomous platelet expression patterns in ET, coupled with the temporally distinctive BLVRB lineage expression patterns in both erythroid and MK development, collectively suggest a BLVRB function distinct from that of other heme degradation pathway genes.

BLVRB was initially characterized as a flavin reductase²⁷  retaining physiologic relevance primarily as a redox coupler in the presence of methylene blue for treatment of acquired or congenital methemoglobinemia (CYB5A deficiency, Online Mendelian Inheritance in Man #250800). Its function(s) remain additionally enigmatic because it catalyzes formation of the primary rubin generated during fetal hematopoiesis (the IXβ isomer), with no substrate activity toward the predominant BV IXα isomer found in adults.²⁸  BLVRB was readily detectable in healthy control gel-filtered human platelets with no evidence for altered expression in either BLVRB^S111L ET or RT platelets (Figure 4A). Comprehensive BLVRB sequence analysis in both RT and ET cohorts identified no additional mutations, prompting more focused study of the BLVRB 462^C→T (S¹¹¹L) heterozygous mutation. Previously characterized crystal structure of the 206 amino acid BV IXβ reductase reveals a monomeric protein with a dinucleotide Rossmann binding fold that preferentially accommodates NAD(P)H as an electron donor with promiscuous binding of various linear tetrapyrroles as electron acceptors^27,28  (Figure 4B). Ser¹¹¹ embedded within the wide D⁸⁰–K¹²⁰ substrate binding pocket is structurally homologous to catalytic serines within uridine diphosphate–galactose epimerase (Ser¹²⁴)²⁹  and ferredoxin–NADP⁺ reductase (Ser⁴⁹),³⁰  and uniquely positioned for recognition and/or proton transfer between flavin isoalloxazine and nicotinamide rings.¹⁵  Bacterially expressed and purified recombinant BLVRB^WT (wild type) and BLVRB^S111L demonstrated disparate NAD(P)H-dependent redox coupling using both flavin- and BV-specific substrates (Figure 4C-E), the latter uniquely generated by coupled heme oxidation as verdin-restricted BLVRB activity probes retaining no BLVRA cross-reactivity.¹⁵  BLVRB^S111L enzymatic activity was defective using flavin mononucleotide (flavin reductase activity; P < .0001) and BV IXβ dimethyl esters (BV reductase [BVR] activity; P < .0001), establishing that the S¹¹¹L substitution represents a loss-of-function redox mutation with either substrate. These results were confirmed in HEK293 cells using lentivirus(Lv)-induced expression, where enhanced BVR activity was evident in HEK293/BLVRB^WT-infected cells with little to no activity in HEK293/BLVRB^S111L cells compared with HEK293/Control cells (Figure 4F); residual BVR activity in HEK293/Control and HEK293/BLVRB^S111L reflected endogenous BLVRB expression with no evidence for BLVRA cross-reactivity (Figure 4G).

Previous data have suggested that a BV/BR redox cycle provides a mechanism for neutralization of ROS and cytoprotection.^16,17  Accordingly, we hypothesized that defective BLVRB^S111L redox coupling could affect ROS accumulation, a requisite upstream signaling messenger of MK differentiation,^31,32  and stem cell quiescence during migration from hypoxic (low ROS) osteoblastic to oxygen-rich (high ROS) vascular niches.^33,34  Hematopoietic-derived (CD34⁺) iPSCs infected with individual lentiviruses established that iPSC/BLVRB^WT cells (expressing BLVRB approximately twofold greater than control) retained enhanced redox activity (P = .001) compared with both control and iPSC/BLVRB^S111L cells; redox coupling in iPSC/BLVRB^S111L paralleled that of control iPSCs (Figure 5A). Enhanced iPSC/BLVRB^WT-associated redox coupling was associated with statistically lower baseline ROS accumulation, whereas baseline ROS accumulation was highest in iPSC/BLVRB^S111L cells (Figure 5B). Incubation with the organic peroxide tert-butyl hydroperoxide (TBHP) as the oxidant stress source protected against ROS accumulation in iPSC/BLVRB^WT cells using a TBHP concentration causing ROS accumulation in ∼50% of either control or iPSC/BLVRB^S111L cells (EC₅₀ = 200 μM). The ROS-neutralizing characteristics of iPSC/BLVRB^WT cells contrasted sharply to those of iPSC/BLVRB^S111L, which demonstrated exaggerated TBHP-induced ROS accumulation when compared with either iPSC/BLVRB (P < .00001) or control iPSCs (P < .0002). Thus, the BLVRB-mediated ROS-neutralizing effect appears comparable to that of BLVRA,^16,35  whereas defective BLVRB^S111L activity enhances ROS at baseline with exaggerated ROS accumulation on oxidant stress. Importantly, as heterozygous models designed to phenocopy zygosity state in thrombocytosis cohorts, these collective data establish a BLVRB^S111L dominant inhibitory effect in cells expressing both native and mutant BLVRB.

Potential consequences of differential BLVRB redox coupling and ROS handling on lineage commitment were evaluated using primary CD34⁺ HSCs transduced with Lv/BLVRB^WT, Lv/BLVRB^S111L, and Lv/Control. Suspension cultures prior to terminal differentiation (day 0) demonstrated no differences in cellular viability (Figure 6A), excluding potential ROS-damaging effects in CD34⁺/BLVRB^S111L cells that contained >70% mutant T alleles as established by pyrosequencing. Furthermore, there was general expansion of CD34⁺/BLVRB^WT and CD34⁺/BLVRB^S111L progenitor cells compared with CD34⁺/Control, with results that were evident both by cell number and by flow cytometric quantification of gated, live (7-amino-actinomycin D [7-AAD] negative) CD34⁺ cells (Figure 6B). Comparative effects of BLVRB-associated redox coupling/ROS handling on proliferative potential and lineage commitment were established using primary CD34⁺ methylcellulose multipotential progenitor cultures, results that confirmed exaggerated aggregate colony formation in Lv/BLVRB^WT- and Lv/BLVRB^S111L-transduced cells compared with Lv/Control (Figure 6C). Conversely, there was disproportionate expansion of primitive CFU-granulocytes/erythrocytes/monocytes/MKs in CD34⁺/BLVRB^S111L cells (P = .001), and an absolute increase of burst forming units, erythroid colonies in CD34⁺/BLVRB^WT cells (P = .001). These results confirmed that the proliferative effect occurred independently from an effect on lineage commitment. Collagen-based cultures designed to specifically quantify MK progenitor potential at a single-cell level demonstrated a statistically significant increase of CD41⁺ CFU-MKs in CD34⁺/BLVRB^S111L cells (P < .01) with no increased CFU-MKs in CD34⁺/BLVRB^WT cells (Figure 6D), confirming disparate effects on MK lineage commitment and a preserved proliferative function distinct from its redox capacity.

The Tpo differentiation suspension culture across the genotypes confirmed that >98% of live cells were ROS^high and CD41⁺ (αIIBβ3) acquisition occurred almost exclusively within ROS^high subsets (Figure 6E). These data are consistent with a requisite developmental ROS signal during terminal megakaryocytopoiesis^25,26  and provided a surrogate marker for BLVRB redox activity that remained below threshold sensitivity throughout the 10-day culture period.³⁶  We tracked the distribution of ROS accumulation within ROS^high subsets and noted ROS-attenuating effects of CD34⁺/BLVRB^WT that contrasted with exaggerated ROS accumulation of CD34⁺/BLVRB^S111L; these differences were most pronounced at day 4, but were still evident at day 10 of terminal differentiation (P < .05; Figure 6F). Exaggerated ROS accumulation in MK/BLVRB^S111L cells at Day 4 corresponded to temporally earlier CD41 induction (P < .05); in contrast, ROS-attenuating effects of day 4 MK/BLVRB^WT cells were not associated with diminished CD41 acquisition at day 4, likely explained by a threshold ROS accumulation sufficient to maintain its “priming” effect on MK development.³³  Bilineage (Tpo/Epo) culture conditions designed to address putative effects on MEP balance^12,13  demonstrated no differences among the genotypes on MK partitioning (CD41⁺/glycophorin A⁻ [GlyA⁻] cells), suggesting that BLVRB^S111L MK effect was restricted to postcommitment expansion and not MEP partitioning; interestingly, there was a general expansion in both BLVRB^WT (P = .03) and BLVRB^S111L (P = .06) erythroid (CD41⁻/GlyA⁺) compartments compared with Lv/Control (Figure 6G). Finally, we completed UV-visible spectroscopy of late-stage bilineage cultures (Figure 6H) (in which CD41⁻/GlyA⁺ erythroid fraction accounts for >95% of cells¹³ ), with no evidence for methemoglobin accumulation in BLVRB^S111L erythroid cells.^27,37 

Our data provide the first evidence linking redox activity within the heme degradation pathway to MK lineage fate and physiologically relevant platelet expansion in humans. Support for this conclusion stems from (1) genetic association studies in both clonal and nonclonal disorders of thrombocytosis, (2) biochemical analyses using complementary NADPH-dependent enzymatic determinations of both flavin and BV substrates, (3) documentation that defective redox activity is associated with ROS mishandling as a metabolic cellular consequence in both iPSCs and primary hematopoietic cells, and (4) evidence for disparate effects on MK lineage fate in suspension and collagen-based culture systems. Although BLVRB^S111L is clearly associated with ROS mishandling as a putative explanation for accelerated differentiation (or “priming”)³³  during a developmentally regulated stage of megakaryocytopoiesis, a direct metabolic effect related to tetrapyrrole coupling (unrelated to the surrogate ROS accumulation) remains plausible and is currently under investigation; similarly, we cannot exclude an additive effect of enhanced ROS-associated proplatelet formation during late-stage MK development.³⁸ 

Interestingly, there is minimal evidence implicating any of the heme degradation pathway components in hematopoietic development, which is an unexpected finding in our study. Despite evidence for cytoprotective effects of HMOX and BLVRB,^16,39  heme degradation is generally considered a catabolic pathway, predominantly active in cells of the reticulo-endothelial system to process pro-oxidant (free) heme generated during erythroid senescence or turnover of cytochrome p450 enzymes. Within the subset of heme degradation pathway genes, BLVRB expression (to the exclusion of the other 3 genes HMOX1, HMOX2, and BLVRA) demonstrated a striking induction during terminal erythropoiesis and a restricted temporal pattern of expression during megakaryocytopoiesis. Thus, our data are consistent with a recent report highlighting the importance of HMOX1 in murine erythroid development²⁵  and extend this observation to a lineage-restricted effect on megakaryocytopoiesis. Importantly, the comparatively disparate expression patterns also suggest that BLVRB maintains functional effects distinct from those of other heme degradation components.

To date, it remains unestablished if BLVRB megakaryocytic effects are restricted to its activity as a flavin reductase, in partnered electron exchange with an unidentified protein, or as a verdin-regulated redox coupler. The latter mechanism implies a developmentally coordinated hematopoietic function(s) regulated by heme degradation and isomer-restricted BV IXβ (or IXδ, IXγ) generation. Indeed, this conclusion seems paradoxical given our current understanding of BV/BR isomeric patterns in adults. All natural BVs are formed by ring cleavage of preassembled heme and not by de novo assembly from smaller units⁴⁰ ; although cleavage could occur at any of the 4 meso bridge carbons, there is striking regioselectivity of HMOX for the α-meso (IXα) carbon bridge with limited generation of BLVRB-specific IXβ, IXδ, and IXγ isomers.²⁸  Interestingly, BV isomer generation appears distinct in the fetus where BV IXβ may predominate,⁴¹  although the mechanism for BV IXβ generation remains enigmatic. Nonetheless, BLVRB binds promiscuously (and nonproductively) to other tetrapyrroles such as protohemin²⁷  or BV IXα,²⁸  suggesting that its redox activity could be modulated by endogenous cellular heme byproducts with resultant effects on redox coupling and lineage commitment (ie, function as a metabolically regulated cellular rheostat³⁴ ).

The loss-of-function redox mutation causes defective ROS handling in the background of endogenous BLVRB expression, establishing a dominant inhibitory mechanism of action. This is somewhat unexpected because the crystal structure provides no convincing evidence for dimerization sequences²⁸  (either homo- or heterodimers) as one putative explanation, although an alternative model of enzymatically locked substrate (and/or cofactor) binding remains plausible. Indeed, a compulsory ordered kinetic mechanism has been proposed in which NADPH binding followed by substrate results in sequential release of product and oxidized cofactor.²⁸  A definitive explanation will require crystallization and structure determination of BLVRB^S111L; indeed, this issue may be relevant for previous suggestions imputing the existence of a BV/BR redox cycle and progressively amplified BR antioxidant functions.¹⁶  Finally, the dominant inhibitory mechanism predicts that low-level BLVRB 462^C→T allelic expression could have exaggerated cellular effects in the background of wild-type C alleles.

Low intracellular ROS levels evident in quiescent HSCs are regulated by the FoxO family of transcription factors,⁴²  and the primary non-mitochondrial sources for ROS production are NADPH oxidases. In addition, the thiol balance also contributes to hematopoietic progenitor cell mobilization, as cellular redox status can be regulated by free or protein-incorporated thiols. Previous data have established that p22^hox-dependent NADPH oxidase activity is responsible for ROS production and required for MK differentiation.³²  Despite the importance of ROS signaling in cell quiescence and lineage fate,³³  there are no identified ROS-regulating mutations that modulate human platelet counts, and we would speculate that the temporally restricted nature of BLVRB-associated ROS priming provides the developmentally regulated signal for accelerated MK differentiation. Using a bilineage model of erythro/megakaryocytopoiesis, we saw no evidence for MK lineage partitioning, most consistent with a model of postcommitment megakaryocytopoiesis and not altered E/Meg lineage balance.¹³  Nonetheless, it is intriguing that our limited cohort analysis demonstrated reciprocal effects on hemoglobin and platelet counts in BLVRB^S111L subsets, observations that need to be confirmed in expanded MPN subtypes including polycythemia vera. Expectedly, we saw no evidence that BLVRB^S111L caused methemoglobinemia in the setting of functionally intact cytochrome b5 reductase,²⁷  although statistically significant erythroid expansion was evident in BLVRB^WT multipotential progenitor and bilineage cultures. The latter observation remains under investigation but suggests an erythroid function in redox-regulated bioenergetic metabolism, with possible effects on stage-specific erythropoiesis.^43,44 

BLVRB (and the BLVRA homolog) are unusual, possibly unique, in their ability to use either NADPH or NADH as cofactors with a different pH optimum for each cofactor.⁴⁵  Nonetheless, BLVRB accommodates NADPH more favorably,²⁸  and cofactor binding is required for apoprotein stability; furthermore, broad substrate discrimination occurs by steric constraints for chemical groups adjacent to the electrophilic C10 methene carbon. When integrated with our data, this model suggests that development of BLVRB-selective redox inhibitors with thermodynamically distinct chemical modifications represents logical approaches to selectively alter a regulatory pathway controlling MK lineage expansion and human platelet counts.

The authors thank M. Golightly (Stony Brook University) for assistance with flow cytometric analyses, I. Carrico and E. Boon (Stony Brook University) for assistance with heme oxidation studies, M. Seeliger and R. Rizzo (Stony Brook University) for assistance with computational modeling, and P. Pereira and S. Macedo-Ribeiro (University of Porto) for helpful discussions.

This work was supported by National Institutes of Health (NIH)/National Heart Lung, and Blood Institute (NHLBI) grant HL091939, NIH/NHLBI DNA Resequencing and Genotyping Service (subcontract G175), the New York State Stem Cell Foundation (grant C026716), and the Burroughs Wellcome Fund (W.F.B.); grants from the Stony Brook Targeted Research Opportunities (S.W.); and NIH/NHLBI grant HL12945 (N.M.N.).

Contribution: W.F.B. and S.W. conceptualized the study and formulated the hypothesis; W.F.B. and S.W. designed the research; S.W., Z.L, D.V.G., B.Z., L.Z., L.E.M., N.M., T.J.M., N.M.N., and W.F.B. performed research; W.F.B. and S.W. wrote the paper; and W.F.B. directed the overall research.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Wadie F. Bahou, Department of Medicine and the Stony Brook Stem Cell Facility, Health Sciences Center T15-040, Stony Brook University School of Medicine, Stony Brook, NY 11794-8151; e-mail: wadie.bahou@sbumed.org.