GAPDH cooperativity mediates drug resistance and metabolism in Plasmodium fal-ciparum malaria parasites

Efforts to control the global malaria health crisis are undermined by antimalarial resistance. Iden-tifying mechanisms of resistance will uncover the underlying biology of the Plasmodium falciparum malaria parasites that allow evasion of our most promising therapeutics and may reveal new drug targets. We utilized fosmidomycin (FSM) as a chemical inhibitor of plastidial isoprenoid biosynthesis through the methylerythritol phosphate (MEP) pathway. We have thus identified an unusual metabolic regulation scheme in the malaria parasite through the essential glycolytic enzyme, glyceraldehyde 3-phosphate dehydrogenase (GAPDH). Two parallel genetic screens converged on independent but functionally analogous resistance alleles in GAPDH. Metabolic profiling of FSM-resistant gapdh mutant parasites indicates that neither of these mutations disrupt overall glycolytic output. While FSM-resistant GAPDH variant proteins are catalytically active, they have reduced assembly into the homotetrameric state favored by wild-type GAPDH. Disrupted oligomerization of FSM-resistant GAPDH variant proteins is accompanied by altered enzymatic cooperativity and reduced susceptibility to inhibition by free heme. Together, our data identifies a new genetic biomarker of FSM-resistance and reveals the central role of GAPDH cooperativity in MEP pathway control and antimalarial sensitivity.

The phosphonate antibiotic fosmidomycin (FSM) is a highly specific, competitive inhibitor of DXR.

74
We have previously employed FSM as a chemical tool to identify novel modes of metabolic reg-  under aerobic fermentation, when glucose consumption is high in the presence of oxygen but the 93 absence of TCA cycle respiration (23, 24). This Warburg-like metabolic state is common in rapidly 94 proliferating cells like tumors but also in Plasmodium parasites, as glucose nearly acts as the sole 95 energy-producing carbon source with almost no aerobic respiration during the asexual blood-96 stage. In this state, GAPDH serves as a gate-keeper between upper (energy-consuming "prepar-97 atory" phase) and lower (energy-producing "payoff" phase) glycolysis. Beyond its glycolytic func-98 tion GAPDH functions in multiple other cellular processes including apoptosis, transcription, ve-99 sicular trafficking, and heme detoxification (25-30). Currently, the Plasmodium GAPDH has not 100 been implicated as a determinant of drug resistance, overall pathogenicity, or a regulator of ac-101 cessory cellular functions let alone controlling metabolic plasticity despite its primary role as a 102 glycolytic enzyme.

104
In our continued search of novel modes of FSM resistance we continue to uncover novel biological 105 processes important for parasite function. We present an innovative combination genetic screen-106 ing strategy that utilizes both enhanced FSM resistance selection and multi-round FSM re-

115
Enhanced FSM resistance in malaria parasites selects for GAPDH variant 116 FSM is a highly specific inhibitor of the MEP pathway, which produces the essential isoprenoid 117 precursors IPP and DMAPP. We have previously shown that P. falciparum malaria parasites with

138
In parallel to the "enhancer screen" described above, we also employed a second and independ-

165
To understand how these GAPDH alleles might impact protein function, we modeled the position 8 of the encoded variants on the published tertiary and quaternary structure of PfGAPDH. Both the 167 GAPDH I209T and GAPDH V238A variants are distant from the substrate-binding pocket and are not 168 predicted to directly impact catalysis ( Figure 3). Surprisingly, we found that both I209 and V238 169 are immediately physically adjacent in three-dimensional space. Both residues are present at the 170 base of a relatively disordered S-loop, which is normally stabilized by the oligomerization of the 171 GAPDH homo-tetramer (PDB: 2B4R) (31). Interestingly, we find that not only are I209 and V238 172 adjacent within a single monomer, they are predicted to directly interact at the hydrophobic inter-173 face of two of the homodimer subunits ( Figure 3). Classically, hydrophobic surfaces drive protein-174 protein interactions. Both FSM-resistant variants alter hydrophobic residues (isoleucine to threo-175 nine and valine to alanine, respectively). Therefore, we hypothesized that GAPDH oligomerization 176 was disrupted in our variants and that this was the structural mechanism of FSM resistance in our 177 strains.

180
To test this model, we sought to quantify oligomer formation in wild-type GAPDH and its FSM-

227
This creates a unique crossroads between GAPDH's ability to serve as a regulator during meta-228 bolic stress, heme toxicity, and drug resistance. While many non-glycolytic roles of GAPDH in-229 volve oligomer formation, not all of these non-glycolytic roles are independent from the glycolytic 230 function of GAPDH, especially as a heme chaperone. Therefore, we decided to evaluate whether    Meyerhof-Parnas (EMP) pathway of glycolysis, the oxidation of G3P to 1,3-bisphosphoglycerate

318
(1,3BPG), which also yields the reduced product NADH. In glycolysis, GAPDH both follows G3P 319 and proceeds PYR production, which are the two precursor substrates of the MEP pathway. This 320 14 provides a unique opportunity for GAPDH to maintain balanced substrate flow into the MEP path-321 way. Beyond its glycolytic metabolic function, GAPDH is involved in multiple other biological pro-322 cesses with roles as diverse as transcriptional regulation, vesicle trafficking, and heme trafficking.

323
Unique to Plasmodium spp. is a link between GAPDH's heme chaperone role and its glycolytic 324 function. The glycolytic function of PfGAPDH is inhibited by heme binding while the mammalian 325 GAPDH is not. This closely ties heme toxicity to metabolic regulation. This is in addition to the 326 regulation of GAPDH function through the formation of tetrameric or dimeric states and the dis-327 sociation into is monomeric form. While the ability to perform "alternative" functions may require 328 oligomerization, the glycolytic function does not require oligomer formation.

330
The most compelling aspect of our discovering novel metabolic control through GAPDH is that 331 two independent screens with the same phenotype of FSM resistance converged on the same 332 genetic locus. Importantly, these mutations were not genetically identical but functionally the 333 same. Our separate non-synonymous mutations generated protein alterations at the same part

362
The one-gene one-enzyme hypothesis certainly does not apply to one-function. We aim to under-

385
Cultures that were positive for parasite growth were isolated and verified for FSM resistance as 386 compared to strain E1-C12. Enhanced FSM-resistant strain 59 was cloned by limiting dilution.

387
Two clones 59a and 59b were isolated, confirmed for FSM resistance, and submitted for whole 388 genome sequencing.

390
Generating FSM resistance strains in FSM resistance suppressed P. falciparum Strains E2 and E2-S1 were created as previously published and described (21, 22). Strain E2-S1 392 was first cloned by limiting dilution. The newly obtained clone was deemed strain E5. From this 393 clone independent selections were performed in separated wells. Parasites were continually cul-394 tured as described with the addition of 0.5 µM FSM. Cultures that were positive for parasite growth 395 were isolated and verified for FSM resistance as compared to strain E5. Four independent FSM 396 resistant isolates were selected and named E5-1, E5-2, E5-3, and E5-4. Each isolate was cloned 397 by limiting dilution and two subclones were isolated for each, confirmed for FSM resistance, and 398 submitted for whole genome sequencing.

459
Conversion of NAD+ to NADH was determined using the extinction coefficient of 6,220 M -1 cm -1 .

20
A fresh stock of hemin (2.5mM) was prepared by dissolving hemin in 100mM NaOH with 5% 470 DMSO. The stock hemin solution was diluted further to 1mM in 100mM NaOH with 5% DMSO.

471
Hemin was diluted 1:5 in GAPDH reaction buffer as described above. Serial dilutions of hemin 472 were added to recombinant enzyme and allowed to incubate for 10 minutes at room temperature 473 before being added to the standard GAPDH assay reactions described above.

492
The upper layer was transferred to a new tube and dried using a speed-vac. The pellets were re-493 dissolved in 100 µL of 50% acetonitrile.

21
For LC separation of the MEP intermediates, a Luna-NH2 column (3 um, 150 x 2 mm, Phenom-496 enex) was used flowing at 0.4 mL/min. The gradient of the mobile phases A (20 mM ammonium 497 acetate, pH 9.8, 5% ACN) and B (100% acetonitrile) was as follows: 60% B for 1 min, to 6% B in 498 3 min, hold at 6% B for 5 min, then back to 60% B in 0.5 min. For LC separation of the TCA/Gly-499 colysis/PPP intermediates, an InfinityLab Poroshell 120 HILIC (2.7 um, 150 x 2.1 mm, Agilent) 500 was used flowing at 0.5 mL/min. The gradient of the mobile phases A (20 mM ammonium acetate, 501 pH 9.8, 5% ACN) and B (100% acetonitrile) was as follows: 85% B for 1 min, to 40% B in 9 min, 502 hold at 40% B for 2 min, then back to 85% B in 0.5 min. The LC system was interfaced with a 503 Sciex QTRAP 6500+ mass spectrometer equipped with a TurboIonSpray (TIS) electrospray ion 504 source. Analyst software (version 1.6.3) was used to control sample acquisition and data analysis.

505
The QTRAP 6500+ mass spectrometer was tuned and calibrated according to the manufacturer's 506 recommendations. Metabolites were detected using MRM transitions that were previously opti-507 mized using standards. The instrument was set-up to acquire in negative mode. For quantification, 508 an external standard curve was prepared using a series of standard samples containing different Metabolite abundances were log-transformed with pareto scaling to generate normally distributed 519 data. Metabolites with abundances below the limit of detection were imputed at half the value of 520 the lowest measured abundance for that metabolite. Samples were collapsed into two groups, 521 22 those with parent genotypes and those with mutations in GAPDH. A type III two-way analysis of 522 variance (ANOVA) was performed on each metabolite to test for significant differences across 523 genotype, drug treatment, and/or interaction. A Bonferroni correction was applied to adjust for 524 multiple comparisons. All analysis were performed using the R package of MetaboAnalyst.