Primaquine-5,6-Orthoquinone Is Directly Hemolytic to Older G6PD Deficient RBCs in a Humanized Mouse Model

Abstract

Primaquine and Tafenoquine are the only approved drugs that can achieve a radical cure for Plasmodium vivax malaria but are contraindicated in patients who are deficient in glucose 6-phosphate dehydrogenase (G6PDd) due to risk of severe hemolysis from reactive oxygen species generated by redox cycling of drug metabolites. 5-hydroxyprimaquine and its quinoneimine cause robust redox cycling in red blood cells (RBCs) but are so labile as to not be detected in blood or urine. Rather, the quinoneimine is rapidly converted into primaquine-5,6-orthoquinone (5,6-POQ) that is then excreted in the urine. The extent to which 5,6-POQ contributes to hemolysis remains unclear, although some have suggested that it is a minor toxin that should be used predominantly as a surrogate to infer levels of 5-hydroxyprimaquine. In this report, we describe a novel humanized mouse model of the G6PD Mediterranean variant (hG6PD_Med-) that recapitulates the human biology of RBC age-dependent enzyme decay, as well as an isogenic matched control mouse with human nondeficient G6PD hG6PD_ND. In vitro challenge of RBCs with 5,6-POQ causes increased generation of superoxide and methemoglobin. Infusion of treated RBCs shows that 5,6-POQ selectively causes in vivo clearance of older hG6PD_Med- RBCs. These findings support the hypothesis that 5,6-POQ directly induces hemolysis and challenges the notion that 5,6-POQ is an inactive metabolic waste product. Indeed, given the extreme lability of 5-hydroxyprimaquine and the relative stability of 5,6-POQ, these data raise the possibility that 5,6-POQ is a major hemolytic primaquine metabolite in vivo.

SIGNIFICANCE STATEMENT

These findings demonstrate that 5,6-POQ, which has been considered an inert waste product of primaquine metabolism, directly induces ROS that cause clearance of older G6PDd RBCs. As 5,6-POQ is relatively stable compared with other active primaquine metabolites, these data support the hypothesis that 5,6-POQ is a major toxin in primaquine induced hemolysis. The findings herein also establish a new model of G6PDd and provide the first direct evidence, to our knowledge, that young G6PDd RBCs are resistant to primaquine-induced hemolysis.

Introduction

Glucose-6-phosphate dehydrogenase deficiency (G6PDd) is the most common human enzymopathy worldwide, affecting ∼1 in 20 individuals, or half a billion humans (Luzzatto et al., 2020). Missense mutations in G6PD cause enzyme instability, resulting in impaired NADPH generation by the pentose phosphate pathway (PPP) that decreases the ability to detoxify reactive oxygen species (ROS) (Minucci et al., 2012; Gómez-Manzo et al., 2016). G6PDd manifests most severely in red blood cells (RBCs) as they have neither ongoing gene expression to replace the unstable G6PD enzyme nor mitochondria as an alternate source of NADPH. Accordingly, G6PDd RBCs have decreased antioxidant capacity and G6PDd patients are at risk for life-threatening hemolytic anemia in response to certain drugs that generate ROS (Luzzatto and Seneca, 2014; Ryan and Tekwani, 2021).

8-aminoquinolines (8-AQs) (i.e., primaquine and tafenoquine), which are the only approved drugs that can cure liver phase Plasmodium vivax (P. vivax), are contraindicated in G6PDd patients due to risk of hemolysis from ROS generation in RBCs (Howes et al., 2013; Baird, 2019). Moreover, because G6PD testing is not widely available, use of 8-AQs is sometimes avoided altogether, limiting cure of P. vivax even in G6PD nondeficient patients. Thus, the hemolytic toxicity of 8-AQs in G6PDd patients presents a challenge for radical cure of P. vivax malaria (Thriemer et al., 2017).

8-AQs are prodrugs and both the therapeutic and hemolytic activities are due to metabolites generated in vivo rather than the parent compound (Brodie and Udenfriend, 1950; Fraser and Vesell, 1968; Fraser et al., 1971; Strother et al., 1981; Baird et al., 1986; Baird, 2019). Of the known primaquine metabolites (PMs), 5-hydroxy PMs have been shown to form potent redox cycling pairs with their quinoneimine derivatives and are posited to constitute the main hemolytic PMs (Baird et al., 1986; Vásquez-Vivar and Augusto, 1992) 5-hydroxyprimaquine, in particular, has been widely studied with regard to biology of RBC damage (Baird et al., 1986; Vásquez-Vivar and Augusto, 1992; Bowman et al., 2005a,b). However, despite its potent in vitro activity, 5-hydroxyprimaquine has not been reported in plasma of humans taking primaquine (presumably due to its instability) and its quinoneimine is detected only at very low concentrations in plasma and urine (Avula et al., 2018). Thus, it is unclear if 5-hydroxy PMs are present in sufficient quantities to be the main hemolytic PM, and if not, then the identity of primary hemolytic PMs is unclear.

Primaquine-5,6-orthoquinone (5,6-POQ), a downstream product of 5-hydoxyprimaquine, is detected in RBCs (Khan et al., 2021), with substantial amounts excreted in the urine (Spring et al., 2019; Luzzatto et al., 2020; Pookmanee et al., 2021; Khan et al., 2022; Vanachayangkul et al., 2022; Pookmanee et al., 2024). This has been interpreted as 5,6-POQ being a relatively stable metabolite, which is useful as a “surrogate” measure to infer the presence of the more hemolytic 5-hydroxy PMs (Camarda et al., 2019; Fasinu et al., 2019). However, this view presupposes that 5,6-POQ is not itself hemolytic, which to the best of our knowledge has not been directly tested with regards to clearance of G6PDd RBCs. In this report, we test the hypothesis that 5,6-POQ is directly hemolytic and has been overlooked as a major toxic metabolite.

Damaged G6PDd RBCs are mostly cleared by macrophage consumption in the spleen and liver. As such, the ability to circulate in vivo is the best metric of PM-induced hemolysis. Due to the lack of an experimental animal model that recapitulates the human enzymopathy (i.e., an RBC age-dependent decrease in G6PD), toxicology of PMs has not been studied with in vivo circulation as a metric. In the current report, we introduce a matched pair of novel mice that have mouse G6PD replaced with human G6PD — either a deficient variety of the human G6PD gene (Mediterranean variety) [hG6PD_Med- mouse] or a control strain with a nondeficient human G6PD [hG6PD_ND mouse].

As predicted, the hG6PD_Med- has decreased G6PD activity in RBCs that decays as a function of RBC age. We report that direct exposure to 5,6-POQ induces selective clearance from the circulation of older hG6PD_Med- RBCs, with no increased clearance of younger hG6PD_Med- RBCs or hG6PD_ND RBCs of any age. These findings support the hypothesis that 5,6-POQ directly induces hemolysis of G6PDd RBCs. Due to the extreme lability of 5-hydroxy PMs and the relative stability of 5,6-POQ, these findings raise the possibility that 5,6-POQ has been overlooked as a major hemolytic PM.

Materials and Methods

Additional Fine Details of below Methods Are Presented in Supplemental Methods

Mice.

Both the hG6PD_Med- and the hG6PD_ND mice were generated using ES cells from a B6 background (Bruce4) as described in the text. C57BL/6 J mice (cat # 000664) and B6-GFP mice (cat# 004353) were both purchased from Jackson Laboratory, USA. RBC-specific mCherry mice were generated as previously described by knocking an RBC-specific mCherry expression cassette into the ROSA26 locus as a safe harbor for expression of a targeted transgenic mouse (Hay et al., 2022). mCherry mice express mCherry specifically in RBCs and have a fluorescence emission of approximately 550–650 nm, allowing their visualization by flow cytometry. As G6PD is X-linked, it manifests predominantly in males that are hemizygous. Accordingly, all mice used in the study were male. Importantly, heterozygous G6PDd females have variable penetrance of enzymopathy due to Barr body inactivation and represent a neglected population with regards to research. The new model presented herein has the ability to test hypotheses in this context, but such studies are outside the scope of the current report. All “wild-type” B6 mice were purchased from Jackson Laboratories. All animal experiments were carried out under the approved Institutional Animal Care and Use protocols at the University of Virginia.

Measuring G6PD Expression and Activity.

G6PD mRNA, protein, and activity were tested as previously described (D’Alessandro et al., 2021). I kappa B kinase gamma mRNA was quantified by reverse transcription quantitative polymerase chain reaction (RT-qPCR) with the Mm00494927_m1 primer/probe set (ThermoFisher Cat# 4331182). Monoclonal primary antibodies used for Western blot included rabbit anti-human/mouse G6PD (Abcam, clone EPR20688, Cat# ab210702), mouse anti-human G6PD (SantaCruz, clone G-12, Cat# sc-373886), and anti-beta actin (Cell Signaling Technology, clone 13E5, cat# 4970 L). Secondary antibodies were horseradish peroxidase-conjugated polyclonal horse anti-mouse or goat anti-rabbit (Cell Signaling Technology, Cat# 7076S, 7074S) (D’Alessandro et al., 2021). Detection of horseradish peroxidase signal was carried out using ECL Prime Western blotting reagent (Cytiva, Cat# RPN2236) on an Amersham ImageQuant 800 biomolecular imager (Cytiva).

In Vivo Biotinylation and Physical Isolation of Younger RBCs.

RBC lifespan determination was carried out by the in vivo biotinylation method (Hoffmann-Fezer et al., 1991) and as described in detail in previous studies (Zimring et al., 2014). For isolation of younger RBCs, mice were exsanguinated at 6 days postbiotinylation, blood was washed three times, and RBCs older than 6 days were removed by incubation with streptavidin MagneSphere Paramagnetic Particles (Promega, Cat# Z5482) and depletion with a Permagen MSR6X15 magnetic separation rack.

5,6-POQ Challenge of RBCs and Calculating in Vivo Circulatory Lifespan.

RBC challenge with 5,6-POQ was carried out at the indicated concentration of 5,6-POQ and with previously described general methods using a different compound (Dziewulska et al., 2023). Specific details are in Supplemental Methods.

Electron and Light Microscopy.

Whole blood samples were prepared for scanning electron microscopy according to standard procedures as detailed in Supplemental Methods. Light microscopy was carried out using Wright-Giemsa stain (Sigma-Aldrich, Cat# WG32) according to the manufacturer’s protocol.

Methemoglobin and Superoxide Measurements.

Hemoglobin (Hb) concentration was measured using an ABL90 FLEX blood gas analyzer (Radiometer America, Brea, CA, USA) at 4-minute intervals for the first 20 minutes of incubation and at 5-minute intervals for the last 40 minutes of incubation. The rates of methemoglobin (MetHb) generation were calculated as the ratio of MetHb produced between successive measurements to the time interval between measurements. Superoxide was measured by electron paramagnetic resonance (EPR) as previously described using 1-hydroxy-3-methoxycarbonyl-2,2,5,5-tetramethylpyrrolidine as a spin trap (Palha et al., 2022).

Metabolic Analysis of Glycolysis and PPP Flu.

High resolution metabolomics were carried out as previously described (Nemkov et al., 2019).

Instruments and Statistics.

All flow cytometry was carried out on an Attune cytometer and analyzed using FlowJo v10.8. Light microscopy used a 63x objective and digital images were recorded with a Leica DMC4500 color camera. All EPR spectra were recorded on a spectrometer operating at X-band (9.65 GHz; EMXnano, Bruker Corp., Billerica, MA). Statistical analysis was carried out using GraphPad Prism 9.

Results

Generation and Characterization of Humanized G6PD Deficient and Non-Deficient Mice.

A novel mouse model was engineered by replacing the murine G6PD gene with human genomic G6PD (Fig. 1A, targeting constructs and approach in Supplemental Fig. 1A). Two separate strains were generated using either a nondeficient G6PD(B+) variant (hG6PD_ND) or the Mediterranean S188F G6PD variant (hG6PD_Med-) (Gómez-Manzo et al., 2016). A complete sequence of targeting constructs is provided in the Supplemental Materials. The targeting constructs are related to a previous hG6PD_Med- mouse model we have reported (D’Alessandro et al., 2021; Dziewulska et al., 2023); however, the new mice are a direct knockin without going through a CRE-mediated conditional excision out of concern that the residual CRE scar may affect gene expression. Moreover, no control mouse expressing nondeficient human G6PD has previously been described.

Fig. 1.

Generation and characterization of a matched set of humanized mice expressing either nondeficient or the Mediterranean variant of human G6PD. (A) Schematic diagram of the genetic modification in each animal. Blue boxes and gray boxes indicate human and murine exon sequences, respectively. The hG6PD_ND mouse used the nondeficient B+ variant, while the hG6PD_Med- mouse encoded the S188F variant in exon 6 (exons numbered the same for mice as in humans) that constitutes the human Mediterranean-deficient variant. FRT sites (green) identify scars from excision of the FLP flanked neomycin cassette used to select ES cells. (B) mRNA levels by qPCR using two different primer/probe combinations specific for human G6PD mRNA spanning exons 3–4 (left panel) or exons 4–5 (right panel). mRNA expression is normalized to murine beta-actin expression and there was no significant (ns) difference in humanized mRNA expression in hG6PD_ND vs. hG6PD_Med- mice (Brown-Forsythe and Welch ANOVA with Dunnett’s T3 multiple comparisons were used as data did not meet the criteria for being normally distributed). Each RT-qPCR assay was repeated four times (representative experiment shown) and bar graphs represent means with error bars indicating standard deviation. For each experiment, between three and six independent biological samples were prepared for each genotype (biological replicates) and each mRNA was amplified in triplicate (technical replicates). Every sample had a control in which reverse transcriptase was omitted to test for any amplification from contaminating DNA. (C) Western blot measuring G6PD protein in RBC lysates using an antibody specific for human G6PD. The top and middle blots are the same blot with the middle blot being a long overexposure to reveal trace amounts of hG6PD protein in hG6PD_Med- RBCs. Three different mice from each strain (wild-type), hG6PD_ND, or hG6PD_Med-) were tested. A low amount of recombinant human G6PD (rG6PD, 10 ng/lane) was included to establish the limits of detection. Each sample was also tested for beta-actin to control for loading differences. (D) G6PD enzyme activity of RBCs from each strain (one-way ANOVA with Sidak’s multiple comparisons test, *P < 0.05, **** P < 0.001). (E) RBC circulatory lifespan of each strain. Mice underwent whole blood biotinylation and were repeatedly phlebotomized over the course of the experiment. Biotinylated [i.e., streptavidin reactive (StAv+)] RBCs were enumerated with flow cytometry at the indicated timepoints. (F) Light microscopy of peripheral blood smears stained with Wright-Giemsa stain (left panel) or a reticulocyte stain (right panel). In the absence of oxidative stress, RBCs had mostly normal morphology, with some indication of poikilocytosis (V and *). Consistent with flow cytometry, reticulocyte (arrows) numbers were similar between strains. Light microscopy used a 63x objective and digital images were recorded with a Leica DMC4500 color camera (size bar indicates 5 μM at 630x magnification). Each experiment was performed at least three times with similar results and representative experiments are shown. All data points shown are derived from samples taken from different mice (biological replicates) and do not represent technical replicates. In all cases, 3–5 mice were used per group, bar values represent means, and error bars represent standard deviation. Please note that in panel E, error bars are present but too small to be easily seen.

Only exons 3–13 of G6PD were targeted to avoid disrupting coding regions of the gamma chain of IKKγ on the opposite strand (Supplemental Fig. 1B). Homologous recombination and lack of random integration in ES cells was confirmed by qPCR (data not shown). RT-qPCR on bone marrow-derived RNA confirmed that IKKγ mRNA expression was not significantly different in either hG6PD mouse compared with wild-type animals (Supplemental Fig. 1C). The final G6PD primary sequence from hG6PD_ND and hG6PD_Med- is fully humanized except for two murine amino acids (Y21F and A29S) encoded by exon 2, which was left intact to avoid disruption of IKKγ expression; Y21F and A29S are both in an unstructured region of the N-terminus, which is not predicted to be involved in G6PD enzymatic function by X-ray crystallography (Au et al., 2000).

Human G6PD mRNA was detected in bone marrow of both hG6PD_ND and hG6PD_Med- but not wild-type mice by qPCR using two different primer/probe sets that spanned exons to avoid any inadvertent amplification of contaminating genomic DNA (exons 3-4, or exon 4-5) (Fig. 1B). A qPCR primer/probe set for exons 1–2 (common to both mouse and human G6PD mRNA) demonstrated that mRNA in both hG6PD_ND and hG6PD_Med- mice was lower than wild-type mice, although the difference bordered on statistical significance (Supplemental Fig. 1C). No difference was observed in hG6PD mRNA between hG6PD_ND and hG6PD_Med- mice, supporting the hG6PD_ND mouse as an appropriate control for the hG6PD_Med- mouse. As predicted, a primer/probe set specific for a region spanning murine mRNA (exons 12–13) did not detect any murine G6PD mRNA in either hG6PD_ND or hG6PD_Med- mice (Supplemental Fig. 1C).

Western blots on RBC lysates from multiple animals using human-specific anti-hG6PD demonstrated a strong signal in hG6PD_ND RBCs but only a very faint band from hG6PD_Med- RBCs, which required overexposure to observe (Fig. 1C). hG6PD_Med- RBCs had <5% G6PD activity of hG6PD_ND RBCs (Fig. 1D), similar to observed G6PD activity in RBCs from humans carrying the Med variant (Mattè et al., 2020). hG6PD_ND RBCs had enzymatic activity slightly lower than wild-type murine G6PD (Fig. 1D) (raw unnormalized activity in Supplemental Fig. 1D). Whole blood pulse chase biotinylation studies demonstrated that hG6PD_Med- RBCs had a normal circulatory lifespan at baseline (Fig. 1E), generally consistent with G6PD_Med humans (Bernini et al., 1964). hG6PD_ND and hG6PD_Med- RBCs had similar morphology on peripheral smear (Fig. 1F); no significant difference in hematocrit, hemoglobin, or reticulocyte counts (thiazole orange positive cells) were observed among hG6PD_ND or hG6PD_Med- mice compared with wild-type mice (Supplemental Fig. 1, D and E).

Young hG6PD_Med- RBCs Have Higher G6PD Levels.

G6PDd has been ascribed to missense mutations that cause enzyme instability (Kahn et al., 1974). Accordingly, it is generally held that G6PD activity declines with RBC cell age, a conclusion supported by higher G6DP activity in density gradient isolated younger RBCs from G6PDd individuals (Kahn et al., 1974; Morelli et al., 1978). However, it has also been reported that decreased G6DP activity in humans carrying the G6PD_Med variant is entirely due to altered specific activity, that the enzyme has a normal stability, and that previous reports of younger RBCs having higher G6PD activity was due to leukocyte contamination (Morelli et al., 1984). To help address this disagreement in the literature, a biotinylation “pulse chase, depletion” approach was used to isolate younger RBCs (Fig. 2A). On day zero post in vivo biotinylation, 100% of RBCs were streptavidin reactive (including thiazole orange positive reticulocytes) (Fig. 2B). By 6 days postbiotinylation, “young” RBCs had emerged from the bone marrow that were not biotinylated (streptavidin nonreactive). As predicted, reticulocytes were now predominantly observed in the young RBC population (Fig. 2B). The biotinylation procedure itself did not significantly change the reticulocyte count or the percentage of very young reticulocytes (CD71+).

Fig. 2.

Young RBCs fromhG6PD_Med-- mice have increased G6PD protein and enzymatic-activity. Wild-type, hG6PD_ND, and hG6PD_Med-- mice were subjected to in vivo biotinylation and blood was harvested 6 days postbiotinylation. (A) Experimental schematic highlights the in vivo biotinylation and streptavidin-based depletion of RBCs older than 6 days of age. (B) All RBCs were stained with thiazole orange to visualize reticulocytes, combined with either streptavidin [StAv (top row) or anti-CD71 (bottom row)]. Flow cytometric analysis of thiazole orange x streptavidin staining demonstrates 99% biotinylation on day zero postbiotinylation that decreases to 85% at day 6 post biotinylation as new (and unbiotinylated) RBCs emerge from the bone marrow. As predicted, reticulocyte (thiazole orange +) events are contained in the young (StAv negative) population on day 6 postbiotinylation. Of the thiazole orange + reticulocytes, 76% are CD71-positive, indicating young reticulocytes. (C) Flow cytometric analysis of biotin pulse-chased and control (mock-chased) samples undergoing streptavidin depletion demonstrates enrichment in streptavidin nonreactive (young) reticulocytes (thiazole orange and CD71+). (D) Western blot analysis with an antibody reactive with both human and mouse G6PD in lysates from the mock-depleted (all ages) and streptavidin-depleted (young) RBCs shows increased G6PD protein in young hG6PD_Med- lysates compared with all age lysates. Molecular weights are noted to the right of the blot (in kDa) and are estimated based on molecular marker standard (not shown). Raw images are contained in supplemental information. (E) Change in G6PD enzymatic activity of young vs. all age RBCS for each strain. Activity in young RBCs vs. all age RBCs is normalized to baseline activity in hG6PD_ND RBCs. Each point represents the values of blood pooled from 3–5 animals, which is required to obtain sufficient volume of young RBCs to perform the assay — no statistical test was performed as individual mouse data are not available (due to pooling), but differences far exceed two standard deviations [shown as error bars (on the graph)], which by standard distribution theory represents a significant finding. The height of the bar represents the mean. Each experiment in this figure was performed at least three times with similar results, and a representative experiment is shown. Error bars represent standard deviation.

Young RBCs (1-6 days old) were enriched to 90% purity by removing older biotinylated RBCs using streptavidin-coated magnetic beads (Fig. 2C). To control for any changes in RBCs due to the enrichment procedure, control RBCs consisted of a “mock chase” in which “all age” biotinylated and nonbiotinylated RBCs were mixed in the same ratio observed in day 6 postbiotinylation animals (Supplemental Fig. 2A and Fig. 2C). Young hG6PD_Med- RBCs had a 2.5-fold increased G6PD activity and higher protein levels by Western blot compared with “all age” hG6PD_Med- RBCs (Fig. 2, D–E). The observed findings were not an artifact of the RBC isolation procedure, as G6PD activity of mock chase RBCs did not change due to the enrichment procedure. The same number of RBCs were lysed in each sample and the same amount of protein extract was loaded in each lane. The equal protein loadings combined with Western blot findings indicate that young RBCs have higher actin content. Together, these data demonstrate that younger murine RBCs have higher G6PD activity and protein, confirming that the hG6PD_Med- variant has decreased stability in murine RBCs.

Young hG6PD_Med- RBCs Are Resistant to 5,6-POQ Induced Hemolysis.

To test the sensitivity of young versus older hG6PD_ND and hG6PD_Med- RBCs to 5,6-POQ challenge, 6-day postbiotinylation blood was treated with a titration of 5,6-POQ for 1 hour at 37°C (Fig. 3). Control RBCs were treated with vehicle only (veh) under the same conditions. RBCs were then washed and a fixed amount of untreated mCherry+ tracer RBCs was added to each tube (Fig. 3A). In vitro hemolysis was tested by enumerating treated RBCs relative to mCherry RBCs as well as assessing relative amounts of young, treated RBCs (mCherry-, streptavidin nonreactive) versus older treated RBCs (mCherry-, streptavidin reactive). In vivo hemolysis was tested by transfusing the mixture into GFP transgenic recipients (Fig. 3A). RBC survival in vivo was determined by enumerating test RBCs (nonfluorescent) in peripheral blood. Recipient RBCs were gated out by GFP (Fig. 3B). Counts of test RBCs were normalized to mCherry tracer RBCs to control for any differences in infusion or phlebotomy. The ratio of (test:tracer) RBCs in experimental animals (Fig. 3B, right panel) was divided by the pretransfusion (test:tracer) ratio (Fig. 3B, left panel) to allow determination of a zero-time point. (formula in Supplemental Fig. 3D). Amounts of young versus older RBCs in the test population was determined by streptavidin reactivity (Fig. 3B) of GFP negative RBCs.

Fig. 3.

YounghG6PD_Med-- RBCs are resistant to 5,6-POQ induced hemolysis. (A) Blood at day 6 postbiotinylation was collected and treated with the indicated concentrations of 5,6-POQ for 1 hour at 37°C and then washed. A fixed number of mCherry+ RBCs were added to each tube as a tracer control and then transfused into GFP transgenic recipient mice. (B) Gating strategy to separate test vs. control RBCs (GFP- vs. GFP+, respectively) and to enumerate tracer [StAv-, mCherry+ (pink gate)], younger [StAv-, mCherry-(blue gate)], and older [StAv+, mCherry-(violet gate)]. (C) In vitro hemolysis was assayed by enumerating young and older test RBCs vs. mCherry RBCs in the input mix by flow cytometry. As such, individual values are not shown as the bar value represents determination from thousands of events counted by flow cytometry. (D) Circulation of young (filled-in symbols, solid line) and older (empty symbols, dashed line) RBCs was assayed over time and graphed as ratio of test:tracer RBCs of indicated time point to input test:tracer ratio, and normalized to respective vehicle control. Representative data are shown for one of two independent experiments, with mixed blood from 3–5 donor mice for each condition. Please see figure legend for Fig. 2 for statistical considerations. In vivo graph represents recipient (n = 3) means ± S.D.

One hundred fifty micromolar 5,6-POQ caused in vitro hemolysis during treatment (prior to infusion), as evidenced both by visual inspection (Supplemental Fig. 3B) and flow cytometry (Fig. 3C). The in vitro hemolysis was specific to older hG6PD_Med- RBCs (i.e., biotinylated), which decreased from approximately 75% (pretreatment) to 20% (post treatment). No significant in vitro hemolysis was observed with 15 μM 5,6-POQ (Fig. 3C) or with 30 μM 5,6-POQ (data not shown).

All doses of 5,6-POQ caused decreased circulation of treated hG6PD_Med- RBCs compared with hG6PD_ND RBCs (Fig. 3D and Supplemental Fig. 3) and the clearance was specific to older hG6PD_Med- RBCs with little to no clearance of young hG6PD_Med- RBCs. No clearance was observed in hG6PD_ND RBCs regardless of age. Thus, 5,6-POQ causes selective destruction of older hG6PD_Med- RBCs both by cellular rupture (at high dose) and due to increased RBC clearance at all doses tested.

Scanning electron microscopy demonstrated that untreated hG6PD_Med- RBCs had a largely normal morphology, with only a slight increase in echinocytes compared with hG6PD_ND RBCs (Fig. 4, A and C). Overall, 5,6-POQ treatment induced echinocytes and sphero-echinocytes in both populations, but approximately 25% of the total population lost normal discocytic morphology in hG6PD_Med- compared with ∼10% of hG6PD_ND RBCs (Fig. 4, B and D). 5,6-POQ-treated hG6PD_Med- RBCs also developed spheroechinocytes and stomatocytes as well as dacrocytes that were not observed in the corresponding hG6PD_ND RBCs.

Fig. 4.

Electron microscopy of hG6PD_ND and hG6PD_Med- RBCs in the absence or presence of 5,6-POQ. Arrows indicate damaged RBCs with altered morphology in each group.

5,6-POQ Treatment Increases MetHb, Superoxide, and PPP Flux.

Treatment of RBCs with 5,6-POQ increased superoxide in both hG6PD_ND and hG6PD_Med- RBCs as measured by EPR using 1-hydroxy-3-methoxycarbonyl-2,2,5,5-tetramethylpyrrolidine as a spin trap (Palha et al., 2022) (Fig. 5, A and B). However, superoxide was spin-trapped at a significantly faster rate in the hG6PD_Med- compared with hG6PD_ND RBCs. The rates implied steady-state superoxide concentrations of 12.2 ± 0.9 nM and 28.1 ± 0.9 nM in 5,6-POQ-treated hG6PD_ND and hG6PD_Med- RBCs, respectively, compared with 5.8 ± 0.5 nM and 5.3 ± 0.5 nM for vehicle treatment (see Supplemental Methods for determination of steady-state concentration) Early generation of MetHb was similar between hG6PD_Med- and hG6PD_ND RBCs (Fig. 5, C and D). However, both rate of generation and concentration of MetHb was increased in hG6PD_Med- RBCs after 30 minutes. Together, these data show that hG6PD_Med- RBCs have persistently higher concentrations of ROS and MetHb compared with hG6PD_ND RBCs.

Fig. 5.

hG6PD_Med- RBCs have increased superoxide and MetHb levels in response to 5,6-POQ exposure (A+B) Superoxide was assayed by EPR through measurement of the unique spectrum generated by CM• formation as a result of superoxide reaction with the 1-hydroxy-3-methoxycarbonyl-2,2,5,5-tetramethylpyrrolidine spin probe. Treatment with 150 μM 5,6-POQ causes CM• concentration to rise more steeply in hG6PD_Med- RBCs (panel B) than in hG6PD_ND RBCs (panel A), indicating higher superoxide concentration in the former. Quadruplicate data sets are plotted with different symbols (○▽◊▵) with corresponding least-squares fitted lines (dashed, solid, dotted, short-dashed). Each bold solid line is the average of all least-squares lines in the data set. Red: 5,6-POQ treatment; blue: vehicle control; gray, buffer control (cell-free PBS-G). (C+D) Methemoglobin (MetHb) concentration (panel C) and production rate (panel D) were measured over the course of 1 hour incubation with 150 μM 5,6-POQ using a ABL90 FLEX blood gas analyzer. This experiment was repeated three times with similar results. Each point shows mean and error bars represent ± S.D.

Metabolic flux analysis was performed on hG6PD_ND and hG6PD_Med- RBCs treated with either low dose (15 μM) or high dose (150 μM) 5,6-POQ in the presence of [1,2,3-¹³C₃]glucose (Fig. 6). Glycolytic metabolites maintain all three labeled carbon atoms; in contrast, the C1 carbon is lost as CO₂ in the PPP. Reentry of PPP intermediates into glycolysis yields ¹³C₂ late glycolytic metabolites that are easily distinguished from the glycolysis generated ¹³C₃ forms (red and white circles in Fig. 6A). As predicted, hG6PD_Med- RBCs had significantly decreased PPP metabolites downstream of G6PD (Fig. 6A). 6-Phosphogluconate (total and ¹³C₃-enriched), ribose phosphate, and sedoheptulose phosphate were each decreased at baseline in hG6PD_Med- RBCs. 5,6-POQ treatment caused a sharp influx through the PPP in hG6PD_ND RBCs as evidenced by increased levels of enriched PPP metabolites. In contrast, little to no PPP increase was seen in hG6PD_Med- RBCs treated with 5,6-POQ (Fig. 6A). Lower ¹³C₂/¹³C₃ ratios were observed in lactate in hG6PD_Med- RBCs compared with in hG6PD_ND RBCs, consistent with decreased glucose flux through the PPP (Fig. 6B). NADPH generated by the PPP is the cofactor for glutathione reductase and is thus required to maintain the pool of reduced glutathione (GSH). Consistent with a defective PPP, 5,6-POQ caused a dose-dependent decrease in GSH levels in hG6PD_Med- RBCs, while GSH levels were maintained in hG6PD_ND (Fig. 6C).

Fig. 6.

5,6-POQ treatment induces a specific metabolic lesion that is exacerbated in hG6PD_Med- RBCs. RBCs treated with the indicated concentration of 5,6-POQ, in the presence of [1,2,3-¹³C₃]glucose, were subjected to metabolomic analysis. (A) Glycolysis and the PPP were analyzed for steady state metabolite levels and flux was inferred by different ¹³C species abundance. The isotope at each carbon is indicated by red and white circles and ¹²C vs. ¹³C at each position is indicated by different shades in the bar graph to show relative amounts of labeled vs. endogenous unlabeled metabolites. (B) Lactate ¹³C₂/¹³C₃ levels indicate shunting of PPP back into distal glycolysis. (C) Free (GSH) and oxidized glutathione were measured. These data are the combined measurements from three different experiments.

Discussion

Primaquine is simultaneously metabolized by two competing pathways. (Overall Model in Fig. 7) (Avula et al., 2018) MAO-A generates primaquine aldehyde that is then converted to carboxyprimaquine (the most abundant primaquine metabolite in plasma) by aldehyde dehydrogenase. At the same time, CYPs (1A2, 2B6, 3A4, 2D6, and 2E1) generate phenolic metabolites of primaquine. Of these, 5-hydroxyprimaquine that is generated mainly by CYP2D6 is the most potent cellular oxidant reported, forming a redox cycling pair with its quinoneimine. As such, 5-hydroxyprimaquine has been speculated to be central to primaquine induced hemolysis of G6PDd RBCs (Vásquez-Vivar and Augusto, 1990, 1992, 1994; Bowman et al., 2005a) through generation of ROS. Formation of adducts by 5-hydroxyprimaquine has also been posited to play a role in hemolysis, although this concept has been challenged as the 1,4-reductive addition is blocked (Augusto et al., 1988). Importantly, 5-hydroxyprimaquine is highly unstable and is rapidly converted into 5,6-POQ, which is stable and excreted in the urine. 5,6-POQ has been shown to have some redox activity (Link et al., 1985; Agarwal et al., 1988, 1991; Fletcher et al., 1988; Vásquez-Vivar and Augusto, 1994; Bowman et al., 2004); however, the direct hemolytic potential of 5,6-POQ on G6PDd RBCs has not been reported. Thus, 5,6-POQ has been considered more as a stable waste product than an active hemolytic metabolite.

Fig. 7.

Model of primaquine metabolism with 5,6-POQ as a major hemolytic metabolite. Primaquine is a prodrug that is neither antimalarial nor hemolytic as a parent compound but undergoes metabolism by two main pathways. Primaquine is metabolized into carboxy primaquine through successive actions of monoamine oxidase A (MAO-A) and aldehyde dehydrogenase (ALDH). Alternatively, cytochrome p450s (CYP) convert primaquine to a series of monohydroxy forms (in this case 5-hydroxyprimaquine) The reactive 5-hydroxyprimaquine metabolite undergoes redox cycling if an appropriate enzymatic reductase (also called diaphorase) is present as indicated in red (Vásquez-Vivar and Augusto, 1992). Enzymatic reduction of the 5-hydroxyprimaquine produces a more stable quinone-imine, which has been detected both in plasma and urine (Avula et al., 2018) (Gray box). Although the 5-hydroxyprimaquine/5-quinone-imine can generate robust ROS through redox cycling (gray box), the quinone-imine is rapidly converted to the more stable 5,6-POQ, detected in RBCs and urine (Avula et al., 2018; Khan et al., 2021). The results of the present study confirm the hemotoxic property of 5,6-POQ through further redox cycling (blue box).

In this report, we demonstrate that 5,6-POQ actively causes ROS generation in G6PDd RBCs and results in hemolysis as measured by removal of RBCs from circulation in vivo, in a mouse model. These findings challenge the previous notion that 5,6-POQ is relatively inert (albeit stable) and should be used as a surrogate of the more hemolytically active 5-hydroxyprimaquine as opposed to being considered a hemolytic metabolite in its own right. The current findings do not challenge that 5-hydroxyprimaquine is a potent hemolytic PM (Fasinu et al., 2019; Khan et al., 2021; Vanachayangkul et al., 2022). However, as 5-hydroxyprimaquine is so labile as to not be detected in humans taking primaquine, and 5,6-POQ is easily detected (Avula et al., 2018), the current findings raise the possibility that 5,6-POQ is a more physiologically important hemolytic PM.

The resistance of younger RBCs to 5,6-POQ-induced hemolysis is highly relevant to development of dose-escalation and short-course high-dose regimens of primaquine to induce low level early hemolysis, which induces reticulocytosis, resulting in an increase in young RBCs that can better tolerate higher subsequent primaquine doses (Moore et al., 2023). Using an in vivo radiolabel pulse chase approach in a single human subject, Beutler et al. showed that soon after the pulse, young RBCs continued to circulate despite a primaquine challenge, but were cleared by a subsequent primaquine challenge when the same RBCs were older (Beutler et al., 1954). Because the older RBCs were exposed to two primaquine challenges while younger RBCs were only exposed to one, this approach could not separate an age-dependent sensitivity of RBCs to primaquine from an “area under the curve” aggregate effect of primaquine exposure. The results presented herein, using in vivo methods not permissible in humans, support that young G6PDd RBCs are resistant to primaquine-induced hemolysis.

As with all studies, the current findings have limitations. There is always the risk that findings in rodents may not translate into human biology. Moreover, while ex vivo challenge with 5,6-POQ linked to an in vivo assay for RBC clearance allows the isolation of important variables (i.e., exposure to a single compound), it also is clearly artificial in several regards. It is not necessarily the case that a short exposure (1 hour) to a higher concentration of 5,6-POQ (15 μM) recapitulates the biology of a longer exposure (several days) to lower concentrations of 5,6-POQ seen in human serum (∼40 nM) (Khan et al., 2022). In addition, serum levels of 5,6-POQ differs in humans based upon their genetics of drug metabolism (Vanachayangkul et al., 2022). However, it has been argued that hemolysis of G6PDd RBCs by redox cycling drugs is an “area under the curve” function (Grossman and Jollow, 1988), and if such is the case, then a shorter exposure at higher does may be a reasonable model for in vivo exposure. Nevertheless, in the context of such limitations the current approach allows demonstration that 5,6-POQ has direct redox cycling activity, generates oxidative stress, causes accumulation of MetHb, and leads to accelerated in vivo clearance of young G6PDd RBCs.

The increased oxidative stress in G6PDd RBCs treated with 5,6-POQ (compared with nondeficient RBCs) may be due to increased generation of ROS, decreased detoxification of ROS, or both. However, based upon known G6PDd RBC biology (Ryan and Tekwani, 2021), we speculated a more prominent contribution of the latter. Nevertheless, because reduction of MetHb depends upon NADH and not NADPH, the decreased ability of G6PDd RBCs to reduce MetHb levels over time suggests increased ROS generation. However, there is an alternate MetHb pathway that uses NADPH, which is expected to be limited in G6PDd RBCs due to a defective PPP (Yubisui et al., 1977). Thus, the MetHb data may suggest a greater contribution of the NADPH MetHb reductase pathway than previously appreciated. Future studies specifically focused on ROS generation and MetHb mechanisms will be required to assess these competing explanations.

In aggregate, the current report introduces a novel humanized G6PDd mouse model that recapitulates the enzyme instability known to be present in human G6PDd, including a control animal with nondeficient human G6PD. Using this model, we report that 5,6-POQ both generates ROS and results in selective hemolysis of older G6PDd RBCs. These data support a reconsideration of the role of 5,6-POQ as an active hemolytic PM rather than simply a nonreactive metabolic product used as a surrogate of more active PM metabolites. Given the relative stability of 5,6-POQ and also its generation directly in RBCs, these findings raise the possibility that 5,6-POQ is a major hemolytic metabolite of primaquine (Fig. 7). Redox cycling by 5,6-POQ requires an enzymatic reductase, which could either drive a single electron reduction (generating the semiquinone that would generate ROS while oxidizing back into 5,6-POQ) or a double electron reduction that generates the hydroquinone that could also oxidize back to the semiquinone or 5,6-POQ. Future studies focusing on specific reductases in RBCs will be important to elucidate such pathways.

Abbreviations

8-AQ

8-aminoquinoline

EPR

electron paramagnetic resonance

G6PD

glucose 6-phosphate dehydrogenase

G6PDd

G6PD deficiency

GSH

glutathione

Hb

hemoglobin

hG6PD_Med-

humanized G6PD Mediterranean mouse

hG6PD_ND

humanized G6PD non-deficient mouse

MetHb

methemoglobin

ns

not significant

5,6-POQ

primaquine-5,6-orthoquinone

PM

primaquine metabolite

PPP

pentose phosphate pathway

P. vivax

Plasmodium vivax

RBC

red blood cell

ROS

reactive oxygen species

RT-qPCR

reverse transcription quantitative polymerase chain reaction

StAv

streptavidin

Authorship Contributions

Participated in research design: Dziewulska-Cronk, Reisz, Hay, Nemkov, Cendali, Issaian, Lamb, Palha, Legenzov, Kao, Walker, Tekwani, Buehler, D’Alessandro, Zimring.

Conducted experiments: Dziewulska-Cronk, Hay, Cendali, Issaian, Lamb, Legenzov.

Contributed new reagents or analytic tools: Dziewulska, Zimring.

Performed data analysis: Dziewulska-Cronk, Reisz, Hay, Nemkov, Cendali, Issaian, Lamb, Palha, Legenzov, Kao, Walker, Tekwani, Buehler, D’Alessandro, Zimring.

Wrote or contributed to the writing of the manuscript: Dziewulska-Cronk, Reisz, Hay, Nemkov, Cendali, Issaian, Lamb, Palha, Legenzov, Kao, Walker, Tekwani, Buehler, D’Alessandro, Zimring.