Enzymatic and structural characterization of HAD5, an essential phosphomannomutase of malaria parasites

The malaria parasite Plasmodium falciparum is responsible for over 200 million infections and 400,000 deaths per year. At multiple stages during its complex life cycle, P. falciparum expresses several essential proteins tethered to its surface by glycosylphosphatidylinositol (GPI) anchors, which are critical for biological processes such as parasite egress and reinvasion of host red blood cells. Targeting this pathway therapeutically has the potential to broadly impact parasite development across several life stages. Here, we characterize an upstream component of GPI anchor biosynthesis, the putative phosphomannomutase (EC 5.4.2.8) of the parasites, HAD5 (PF3D7_1017400). We confirm the phosphomannomutase and phosphoglucomutase activity of purified recombinant HAD5. By regulating expression of HAD5 in transgenic parasites, we demonstrate that HAD5 is required for malaria parasite egress and erythrocyte reinvasion. Finally, we determine the three-dimensional crystal structure of HAD5 and identify a substrate analog that specifically inhibits HAD5, compared to orthologous human phosphomannomutases. These findings demonstrate that the GPI anchor biosynthesis pathway is exceptionally sensitive to inhibition, and that HAD5 has potential as a multi-stage antimalarial target.


INTRODUCTION
mannose metabolism is predicted to specifically inhibit GPI anchor synthesis. HAD5 is also 109 predicted to be essential (35), and transcriptomic studies show its expression during the blood 110 stage(36, 37) and sexual stages(38), making it a potential multi-stage antimalarial drug target. In 111 this study, we characterize the putative phosphomannomutase of P. falciparum, HAD5, 112 demonstrating its essentiality for parasite growth and its potential for specific targeting by future 113 antimalarial therapies.

141
Immunoblotting confirms substantial reduction in cellular abundance of HAD5 in the absence of 142 aTc (Fig. 2B). In HAD5 KD parasites, 0 nM aTc conditions led to an absence of growth, whereas 143 addition of aTc promoted growth in a dose-dependent manner (Fig. 2C), indicating that HAD5 is 144 essential for asexual growth of P. falciparum.

145
Because phosphosugar mutases often utilize more than one substrate, we examined 146 whether the phosphomannomutase activity of HAD5 was responsible for its essential function in 147 asexual parasites. Attempts to chemically rescue parasite growth with hexose phosphates such 148 as M6P and M1P were unsuccessful (Fig. S2A), as was expected due to the impermeability of 149 the erythrocyte and parasite membranes to such highly charged compounds. However, we found 150 that simple chemical supplementation of the media with D-mannose was sufficient to rescue 151 growth when HAD5 expression is reduced. This indicates that the primary mechanism of death in 152 these parasites is due to defects in mannose metabolism (Fig. 2D, S2B)

178
Mannose metabolism is linked to parasite egress through biosynthesis of GPI 179 anchors(56). In P. falciparum, GPI anchors are synthesized through addition of one glucosamine 180 (GlcN) and three to four mannose residues to a phosphatidylinositol backbone. These mannose 181 residues are derived from the product of phosphomannomutase, M1P, which is converted to 182 GDP-mannose and subsequently to dolichol-phosphate mannose, the direct mannose donor to 183 GPI anchors (Fig. 3A)(57). Several GPI-APs contribute to egress and invasion of parasites (16, 184 19, 21). We therefore hypothesized that reduced HAD5 expression leads to loss of 185 phosphomannomutase activity and causes parasite death by disruption of GPI anchor biosynthesis. To directly evaluate the effect of HAD5 knockdown on GPI anchor biosynthesis, we 187 labeled mid-to late-trophozoite parasites with [ 3 H]-GlcN and extracted GPI precursors as 188 previously described (18,56,58,59). HAD5 KD parasites grown in +aTc conditions had the 189 expected repertoire of GPI anchor precursors (Fig. 3B, 3C). A variety of precursors are observed, 190 with earlier, less polar species (with fewer mannose groups) migrating further than more polar, 191 highly-mannosylated species. When HAD5 expression is reduced, there is a relative accumulation 192 of the earlier precursors, as well as a reduced production of fully mature, highly mannosylated 193 precursors (Fig. 3B), indicating a defect in GPI anchor biosynthesis. In particular, there was a 194 significant reduction of the highly polar band 9 and significant buildup of less polar band 4 when 195 HAD5 expression was reduced (Fig. 3C). Intriguingly, despite the substantial knockdown of HAD5 196 and the complete loss of growth in these parasites, the abundance of many mannosylated 197 precursors is unchanged and highly mannosylated GPI precursors are still observed, suggesting 198 that this biosynthetic pathway is not completely ablated.

199
To confirm the role of HAD5 in GPI biosynthesis, we deployed two established chemical

209
Reduced GPI anchor biosynthesis in malaria parasites is expected to impact the 210 localization and function of a number of essential GPI-anchored parasite proteins. While several 211 GPI-anchored proteins have been characterized in P. falciparum intraerythrocytic stages, the most abundant is MSP1(16,17). MSP1 must be targeted and anchored through GPIs and 213 proteolytically processed in order for schizont-stage parasites to egress from the erythrocyte, and 214 the MSP1 complex is also critical for binding and reinvading new red blood cells (19,62). For this 215 reason, we investigated whether HAD5-dependent GPI anchor synthesis is required for 216 localization and anchoring of MSP1. We expected that, when the pathway is intact, MSP1 is 217 successfully anchored to the parasite plasma membrane. In contrast, when HAD5 expression is 218 knocked down, GPI anchors will fail to fully incorporate mannose and GPI-anchored proteins, 219 including MSP1, will remain untethered to the membrane (Fig. 3A). To evaluate this effect, we 220 used immunofluorescence to detect the localization of MSP1. When schizonts grown in ±aTc 221 conditions were mechanically lysed and the resultant merozoites were imaged, there was a 222 modest but significant decrease in MSP1 signal surrounding the daughter merozoites when HAD5 223 expression is reduced (Fig. 4A, 4B), indicating that MSP1 membrane attachment is diminished, 224 causing it to diffuse away from the cell.

225
To independently confirm this finding, we partitioned lysate from early schizont-stage  266 with a half-maximal inhibitory concentration of 79 ± 2.6 µM, several-fold more potently than the 267 inhibition of either PMM1 or PMM2 (Fig. 5A). Moreover, we find dramatic time-dependent effects 268 on the ability of D9 to inhibit HAD5, as pre-incubating HAD5 with D9 prior to assaying activity 269 substantially increased D9 potency, such that a 60-minute pre-incubation yielded HAD5 activity 270 of only 4.5% of a vehicle-treated control (Fig. 5B). This effect was not seen for HsPMM1, 271 demonstrating that the potential to specifically inhibit HAD5 may be even greater under ideal 272 binding conditions. As expected given its poor drug-like characteristics (and likely inadequate

277
To uncover the structural basis for HAD5-specific inhibition, we solved the crystal structure  Table S1). There were no large structural differences between the human and parasite 281 enzymes, so we next evaluated the substrate-binding pocket by computationally docking D9 onto 282 the HAD5 crystal structure (Fig. S9B). Although the overall binding pockets were similar and the 283 amino acid sequence within the pocket are highly conserved, subtle differences in side-chain

292
We report here the lethal knockdown and biochemical characterization of the 293 phosphomannomutase in P. falciparum, HAD5. Loss of HAD5 leads to growth arrest in asexually 294 replicating parasites, marked by defects in egress and reinvasion. This growth defect can be 295 rescued by media supplementation with D-mannose, indicating that disruption of mannose 296 metabolism is the primary mechanism of death in HAD5 KD parasites; however, a physiologically 297 relevant concentration of 50 µM D-mannose is unable to significantly rescue growth, bolstering 298 the case for this pathway as a therapeutic target. We further report the specific inhibition of HAD5 299 enzymatic activity compared to orthologous human phosphomannomutases by the hexose-300 phosphate analogue, compound D9, highlighting the potential for specific therapeutics to be 301 developed to HAD5.

302
Unexpectedly, despite the dramatic decrease in HAD5 protein levels and the compelling 303 loss of growth upon HAD5 knockdown, GPI anchor synthesis and MSP1 anchoring to merozoite 304 surfaces are only modestly impacted. Residual HAD5 activity may be present after knockdown 305 and sufficient for generating detectable GPI precursors and subsequent tethering of some GPI-

306
APs, including MSP1. Alternatively, P. falciparum expresses two annotated 307 phosphoglucomutases (PGMs) (PF3D7_1012500 and PF3D7_0413500)(75) that could 308 potentially catalyze phosphomannomutase activity as well, indicating some functional redundancy 309 to HAD5. Residual HAD5 or functional redundancy by PGMs could also explain our observation 310 that D-mannose rescues HAD5 knockdown. Intracellular mannose is likely converted to M6P by 311 hexokinase, and thus some phosphomannomutase activity would still be required for successful 312 rescue of the GPI-anchor pathway. In either case, while residual phosphomannomutase activity 313 present in HAD5 KD may not eliminate GPI synthesis entirely, it is insufficient to sustain parasite 314 growth. This speaks to an exquisite sensitivity of malaria parasite cells to disruption in this pathway, whereby minor perturbations in GPI anchor synthesis nonetheless completely interrupt 316 parasite growth, highlighting the promise of this pathway as a therapeutic target.

317
In addition, we find that HAD5 KD parasites have a complete cell cycle arrest, although 318 others have found that untethering MSP1 from the membrane by removing its GPI-anchoring C-319 terminus still allows for minimal parasite growth (19). We expect that this discrepancy is due to the 320 role of HAD5 in function of all GPI-anchored parasite proteins, not solely MSP1. These other GPI-

321
APs include related MSPs, RAMA, and 6-cysteine proteins, many of which are refractory to 322 deletion and likely essential (16,21,35,76). With the incomplete loss of GPI synthesis, it may be 323 that each one of these GPI-APs, including MSP1, are only relatively de-anchored, but the modest 324 reduction in this post-translational modification across multiple cellular proteins works in concert 325 to cause parasite growth arrest.

326
We therefore propose that HAD5, as an upstream member of the GPI biosynthesis, has 327 great potential as an antimalarial target. We expect that HAD5 inhibition will have broad 328 downstream effects on parasite biology across several life cycle stages. That compound D9 has 329 markedly improved potency against malaria HAD5 compared to orthologous human enzymes 330 provides key proof-of-concept for ongoing development of specific HAD5-directed antimalarial 331 therapeutics. While compound D9 has limited antimalarial efficacy, this is likely due to its charged 332 phosphonate, expected to have poor cellular penetration. This liability may be improved through 333 a variety of medicinal chemistry strategies, including the addition of pro-drug moieties to mask 334 this charge, a strategy that has been highly effective for other phosphonate antimalarials in 335 development (77-79). Finally, the crystal structure of HAD5 is likely to be valuable to ongoing 336 efforts to develop more potent and specific HAD5 inhibitors as antimalarials.

337
This study also adds to the growing literature on HAD-like proteins in P. falciparum. Three 338 related HAD proteins each independently modulate parasite sensitivity to the isoprenoid 339 biosynthesis inhibitor FSM, which prompts the question of whether this effect would be similarly 340 seen with other HAD proteins(67-69). Alternatively, the non-mevalonate isoprenoid biosynthesis pathway may be particularly sensitive to cellular metabolic perturbations. We find that HAD5 342 serves as an interesting counterexample. HAD5 knockdown yields no changes to FSM sensitivity, 343 providing evidence of a HAD protein and a metabolic perturbation that does not impact the 344 sensitivity of parasites to inhibition of isoprenoid metabolism.

345
Finally, we note that HAD5 and the GPI anchor biosynthesis pathway are expressed

396
All immunofluorescence and bright field images were taken using a Zeiss AxioObserver  The coding sequence of HAD5 was cloned from cDNA of 3D7 parasites using primers 417 P1 and P2 (Table S2) and cloned into the BG1861 vector(91), which introduces an N-terminal 418 6xHis-tag, by ligation independent cloning (LIC). This coding sequence was subsequently cut-419 and-pasted into a pET28a vector with NdeI and BamHI-HF (NEB), followed by ligation with NEB 420 Quick Ligase using manufacturer's protocols. The coding sequence did not match published 421 reference sequence of PF3D7_1017400, as an adenosine-to-guanosine mutation yielded an 422 Asn-to-Ser substitution at residue 100. To revert this sequence to the reference sequence, 423 primer P3 (Table S2)

425
The HAD5 D11A allele was generated from the WT plasmid, again using the QuikChange 426 multisite-directed mutagenesis kit and primer P4 (Table S2). Homo sapiens PMM1 and PMM2 427 coding sequences were identified from UniProt, codon-optimized for E. coli, and synthesized by 428 Integrated DNA Technologies (Table S3). These gene blocks were amplified and extended 429 using primers P5+P6 (PMM1) and P7+P6 (PMM2) and PrimeSTAR GXL DNA polymerase 430 (Takara)( Table S2). These sequences were then cloned into the pET28a plasmid by Gibson 431 assembly at NdeI and BamHI-HF cut sites, and plasmids were transformed into XL10 Gold 432 ultracompetent E. coli cells. phosphomannomutase assays. The coding sequence of EcManC was identified from UniProt 435 and a gene block of the sequence was ordered from Integrated DNA Technologies (Table S3).

436
This sequence was cloned by LIC into a BG1861 vector that had been modified with a starting 437 KFS sequence downstream of the 6xHis-tag to enhance protein expression (92)

462
EcManC for enzyme activity assays was expressed by cloning the BG1861:ManC vector 463 into BL21(DE3) pLysS E. coli cells (Life Technologies). Cells were grown in LB broth to an 464 OD600 of 0.7-0.8 and induced with 1mM IPTG for 2hr at 37˚C. Cells were harvested and protein 465 was purified as described for other proteins above. However, as this construct lacked the 466 thrombin-cleavable site from the pET28a vector, the elution from nickel beads was directly run 467 over the size exclusion column, then pooled and concentrated ( Figure S7B).

486
All reactions took place at room temperature in clear CoStar 96-well half-area plates and

493
with the exception of compound D9, whose synthesis is described, and characterization data 494 included, in the attached supporting information. To assess inhibition of HAD5, compounds D1-495 D11 were suspended in water and added to the phosphoglucomutase assay of HAD5 activity 496 (the phosphomannomutase assay was not used, as cross inhibition was seen with downstream 497 components of that assay, but not with the phosphoglucomutase assay; Fig S8D). 5 µL of water 498 volume in the assay was replaced with 5 µL serial dilutions of analogs, with final concentrations 499 ranging from 0 µM -1 mM. The rate of product formation in each condition was used to 500 determine the half maximal inhibitory concentration (IC50) for each inhibitor. In addition to HAD5, 501 these assays were performed with recombinantly purified human PMM1 and PMM2. For these

547
Resultant glycolipids were run on TLC Silica gel 60 F254 plates (Millipore Sigma) using 548 chloroform:methanol:water (10:10:3) as a solvent. TLC plates were exposed to autoradiography 549 films (MidSci), which were developed after 1 week of exposure. Films were imaged on a BIO-550 RAD ChemiDoc MP imaging system, and signal was quantified by the ImageLab software from precondensed Triton X-114 was added to lyse parasites, and lysates were incubated on ice for 559 15 min. Lysates were centrifuged to remove debris. Then followed a series of 5 extractions with 560 cold TBS+PI and precondensed TritonX-114, followed by warming to 37˚C and centrifugation to 561 separate phases. The lysate, detergent-enriched phase, and the soluble phase were analyzed by SDS-PAGE and Western blotted as described above using mouse anti-plasmepsin V 563 1:20(93) and rabbit anti-HAD1 1:10 (67)