Plasmodium Geranylgeranyl Diphosphate Synthase (GGPPS) and Exhibit Potent Antimalarial Activity

Lipophilic analogs of zoledronate and risedronate inhibit Plasmodium geranylgeranyl diphosphate synthase (GGPPS) and exhibit potent antimalarial activity

Joo Hwan Noa,1,2, Fernando de Macedo Dossinb,1, Yonghui Zhangc,1, Yi-Liang Liua, Wei Zhua, Xinxin Fengc, Jinyoung Anny Yooc, Eunhae Leed, Ke Wangc, Raymond Huie, Lucio H. Freitas-Juniorb,3, and Eric Oldfielda,c,3

aCenter for Biophysics and Computational Biology, University of Illinois at Urbana-Champaign, Urbana, IL 61801; bCenter for Neglected Diseases Drug Discovery, Institut Pasteur Korea, Seongnam-si, Gyeonggi-do 463-400, South Korea; cDepartment of Chemistry, University of Illinois at Urbana-Champaign, Urbana, IL 61801; dDepartment of Biochemistry, University of Illinois at Urbana-Champaign, Urbana, IL 61801; and eStructural Genomics Consortium, University of Toronto, Toronto, ON, Canada M5G 1L7

Edited by J. Andrew McCammon, University of California at San Diego, La Jolla, CA, and approved January 9, 2012 (received for review November 5, 2011)

We report the results of an in vitro screening assay targeting nate had an IC50 of approximately 790 nM in the enzyme assay but the intraerythrocytic form of the malaria parasite Plasmodium an IC50 ¼ 120 μM in cells, and similar results were found with falciparum using a library of 560 prenyl-synthase inhibitors. Based several other systems with the average R2 value being 0.30 (17) for on “growth-rescue” and enzyme-inhibition experiments, geranyl- 10 diverse assays. This observation strongly suggested that zoledro- geranyl diphosphate synthase (GGPPS) is shown to be a major nate had poor permeability in the P. falciparum/red-cell assay due target for the most potent leads, BPH-703 and BPH-811, lipophilic to its highly polar nature (log P ¼ −2.9). However, when incorpor- analogs of the bone-resorption drugs zoledronate and risedronate. ating mathematical descriptors (such as logP) to begin to account We determined the crystal structures of these inhibitors bound to a for permeability, we found that good (R2 ¼ 0.7) correlations Plasmodium GGPPS finding that their head groups bind to the between cell and enzyme activity could be obtained (17). These ½ 2þ Mg 3 cluster in the active site in a similar manner to that found results indicated, at least for these types of compounds, that simply with their more hydrophilic parents, whereas their hydrophobic screening for good enzyme inhibitors might not be particularly tails occupy a long-hydrophobic tunnel spanning both molecules informative because many such inhibitors would be unable to get in the dimer. The results of isothermal-titration-calorimetric experi- into cells, and that cell assays would be much more desirable. ments show that both lipophilic bisphosphonates bind to GGPPS A second set of problems with the bisphosphonate class of −1 −1 with, on average, a ΔGof−9 kcal mol , only 0.5 kcal mol worse molecules is that they bind very tightly to bone mineral (18, 19), than the parent bisphosphonates, consistent with the observation resulting in their rapid removal from the bloodstream. This is, of that conversion to the lipophilic species has only a minor effect on course, a desirable feature of a bone drug but not of an antiin- enzyme activity. However, only the lipophilic species are active in fective-drug lead; and in recent work, we have been developing a cells. We also tested both compounds in mice, finding major de- class of compounds called “lipophilic bisphosphonates” (20, 21) creases in parasitemia and 100% survival. These results are of broad in which the 1-OH group on the bisphosphonate backbone, part general interest because they indicate that it may be possible to of the tridentate “bone-hook,” is removed and in which a variety overcome barriers to cell penetration of existing bisphosphonate of hydrophobic side chains are attached to the molecules to drugs in this and other systems by simple covalent modification increase logP values, typically from approximately −2 or −3 to to form lipophilic analogs that retain their enzyme-inhibition activity approximately 2 or 3. These lipophilic bisphosphonates have and are also effective in vitro and in vivo. far more potent activity both in vitro and in vivo than do conven- tional bisphosphonates in tumor cell-growth inhibition and γδ alaria, caused by Plasmodium spp., causes approximately T-cell-activation assays (20, 21) as well as against malaria para- M1 million deaths each year (1), and there are ever-present sites (14). In this work, we elected to screen our in-house library problems due to drug resistance (2). There is, therefore, a need for of 560 prenyl-synthase inhibitors, developed over the past decade new drugs and drug leads. In earlier work, we and others found as anticancer drug leads and as antibacterials (20–24), for their that the bisphosphonate class of drugs (3) used to treat bone- activity in P. falciparum growth inhibition inside red cells. We dis- related diseases—osteoporosis, Paget disease, and hypercalcemia covered two potent leads, BPH-703 and BPH-811 (Scheme 1), due to malignancy—also inhibited the growth of a range of para- lipophilic analogs of the commercial drugs zoledronate and rise- sitic protozoa, including Trypanosoma cruzi (4, 5), Trypanosoma dronate, and determined their crystal structures bound to P. vivax brucei (4, 6), Leishmania spp. (4, 7, 8), Toxoplasma gondii (4, 9), Cryptosporidium parvum (10, 11), Entamoeba histolytica (4, 12, 13), and Plasmodium spp. (4, 13–15). In the case of Plasmodium spp., Author contributions: J.H.N., F.d.M.D., Y.Z., Y.-L.L., L.H.F.-J., and E.O. designed research; the most potent inhibitors were not, however, the nitrogen- J.H.N., F.d.M.D., Y.Z., Y.-L.L., W.Z., X.F., J.A.Y., E.L., K.W., and E.O. performed research; containing bisphosphonates such as zoledronate or risedronate R.H. analyzed data; and L.H.F.-J. and E.O. wrote the paper. (Scheme 1) used to treat bone diseases, but more lipophilic n-alkyl The authors declare no conflict of interest. bisphosphonates (13). Their target in Plasmodium falciparum was *This Direct Submission article had a prearranged editor. not determined. However, more recently, a Plasmodium vivax Data deposition: The atomic coordinates and structure factors have been deposited in the geranylgeranyl diphosphate synthase (PvGGPPS) has been cloned, Research Collaboratory for Structural Bioinformatics Protein Data Bank, www.pdb.org for geranylgeranyl diphosphate synthase complexed with BPH-703 (PDB ID code 3RBM) and expressed, purified, and crystallized, and its three-dimensional for geranylgeranyl diphosphate synthase complexed with BPH-811 (PDB ID code 3RYW). structure determined (16). The enzyme is inhibited by bisphospho- 1J.H.N., F.d.M.D., and Y.Z. contributed equally to this work. nates (16), so it seemed possible that it might be a target for the 2Present address: Center for Neglected Diseases Drug Discovery, Institut Pasteur Korea, inhibitors discovered earlier. To investigate this possibility, we re- Seongnam-si, Gyeonggi-do 463-400, South Korea. cently determined the IC50 values for 25 bisphosphonates against 3To whom correspondence may be addressed. E-mail: [email protected] or eo@ PvGGPPS and compared the results for enzyme inhibition with chad.scs.uiuc.edu. ð¼ − Þ P. falciparum growth-inhibition pIC50 log10 IC50 values (17). This article contains supporting information online at www.pnas.org/lookup/suppl/ 2 The correlation was very poor: R ¼ 0.06. For example, zoledro- doi:10.1073/pnas.1118215109/-/DCSupplemental.

4058–4063 ∣ PNAS ∣ March 13, 2012 ∣ vol. 109 ∣ no. 11 www.pnas.org/cgi/doi/10.1073/pnas.1118215109 Downloaded by guest on September 27, 2021 The screening library consisted primarily of bisphosphonates that might inhibit Plasmodium GGPPS. In addition, we included phosphonosulfonates and related systems known to inhibit head-to-head prenyl synthases (22–24) such as dehydrosqualene synthase and squalene synthase (31), which could also inhibit Plasmodium phytoene synthase, and cationic species (such as Ro 48-8071 and quinuclidines), some of which are known to inhibit Plasmodium cell growth (28, 32). The structures of all 560 compounds are shown in Fig. S1. Scheme 1. Chemical structures of selected bisphosphonates. We first screened all compounds at 10 μM to find possible hits, ¼ 20 GGPPS, in addition to testing them in vivo, finding potent activity. using artemisinin (IC50 nM) as a positive control. There ≥70% μ This finding opens up the possibility that other commercial bispho- were 78 hits (defined as giving inhibition at 10 M) as shown in Fig. 2A, and the Z0 factor for the control wells was 0.72 sphonates, inactive themselves against Plasmodium spp. as well Z0 as other organisms, may be converted to species that are highly as shown in Fig. 2B. The factor is defined (33) as 0 þ − þ − active both in vitro and in vivo, via simple chemical modification. Z ¼ 1–3ðσc þ σc Þ∕jμc − μc j Results and Discussion þ − where σc ∕σc are the standard deviations of the positive (arte- þ High-Throughput Screening (HTS). In Plasmodium spp., the initial misinin)/negative (phosphate-buffered saline) controls and μc ∕ − 0 steps in isoprenoid biosynthesis are carried out by the so-called μc are the corresponding mean values (33), with Z ¼ 0.72 methylerythritol phosphate (MEP) pathway, which produces indicating an excellent assay. Full-assay details are given in SI isopentenyl diphosphate (IPP) and dimethylallyl diphosphate Methods. We then performed dose-response assays for these hits (DMAPP) from pyruvate and glyceraldehyde-3-phosphate (25) (structures shown in Fig. S2) to determine their IC50 values; dose- (Fig. 1). DMAPP then condenses sequentially with three mole- response curves are shown in Fig. S3. Todetermine which of these cules of IPP to form geranyl diphosphate (GPP), farnesyl dipho- compounds might be worth further investigation, we next deter- sphate (FPP), and geranylgeranyl diphosphate (GGPP), which is mined their IC50 values for inhibiting the growth of three human then used to prenylate proteins (26). In addition, in P. falciparum, cell lines, to assess toxicity. Typical dose-response curves are GGPP is converted via prephytoene diphosphate to phytoene and shown in Fig. S4. We then used a therapeutic-index plot, Fig. 2C, then to carotenoids (27); plus, the longer-chain diphosphates are in which we plotted the therapeutic index (T.I.), defined as converted to quinones such as Men-4 (28) as well as dolichols T:I: ¼ IC50ðhuman cell assayÞ∕IC50ðP: falciparum assayÞ (29). In Plasmodium spp., GGPPS appears to be bifunctional, making both FPP as well as GGPP; but based on its sequence and versus the P. falciparum pIC50 values to find those compounds X-ray crystallographic structure with bound inhibitors (16), the en- among the 78 hits with the best activity versus P. falciparum that zyme is clearly more similar to other farnesyl diphosphate synthase also had the lowest toxicity to human cells. Most compounds (FPPS) than GGPPS enzymes—as evidenced, for example, by the were toxic but two: BPH-703 and BPH-811 had relatively low CHEMISTRY presence of a third Asp in the second conserved DDXXD-domain toxicity in these assays (IC50 values of 133 and 28 μM, respec- (16), inhibition by zoledronate and risedronate (unlike human and tively) and good activity in the P. falciparum growth-inhibition 2þ Saccharomyces cerevisae GGPPS), and the presence of the 3Mg assay (IC50 ¼ 690 nM and 1 μM for BPH-703 and BPH-811, 2þ seen in these structures, compared to the 2Mg typically found in respectively). Both compounds also have favorable logP values S. cerevisae GGPPS (30). However, the enzyme is a more effective (logP approximately 2; Table S1) in addition to good cell activity, producer of GGPP than of FPP (16). as illustrated in Fig. 2D. Remarkably, of the 560 compounds

Fig. 1. Isoprenoid biosynthesis and potential targets. Bisphosphonates inhibit FPPS/GGPPS as well as the long-chain prenyl synthases involved in quinone biosynthesis, protein prenylation, and dolichol formation; phosphonosulfonates and related species inhibit head-to-tail prenyl transferases such as squalene and dehydrosqualene synthase and may target phytoene synthase in Plasmodium spp.; cationic nitrogen-containing species such as Ro48-8071 inhibit diverse prenyl transferases. DXR, deoxyxylulose-5-phosphate reductoisomerase; FTI, farnesyl transferase inhibitor; PSY, phytoene synthase; CtrM, dehydrosqualene synthase; OPPS, octaprenyl diphosphate synthase; MEP, methylerythritol phosphate; and FTase, protein farnesyl transferase.

No et al. PNAS ∣ March 13, 2012 ∣ vol. 109 ∣ no. 11 ∣ 4059 Downloaded by guest on September 27, 2021 Target Identification. The results presented above are of consider- able interest because it can be seen that the most potent Plasmo- dium growth inhibitors with low toxicity to human cells are “lipophilic” analogs of the well-known bisphosphonate drugs zoledronate (Zometa®/Reclast®) and risedronate (Actonel®/ Atelvia®), both of which target FPPS (3). This discovery strongly suggested the possibility that both of these lipophilic bisphosphonates target one or more prenyl synthases. In the case of BPH- 703, as illustrated in Fig. S5A, it can be seen that this molecule can be thought of as being derived from a “fusion” of the key structural features found in zoledronate (the bisphosphonate and imidazole group) with the hydrophilic diphosphate and lipophilic features found in FPP or GGPP. In previous work, we found that the lipophilic pyridinium bisphosphonate BPH-715 (Scheme 1) was a potent inhibitor of both liver-stage as well as blood-stage malaria parasites (14), but BPH-703 is far less active than is BPH-715 against three human cell lines (approximately 133 μM vs. approximately 168 nM for BPH-715; Table S1). In the case of BPH-811, it can again be seen (Fig. S5A) that this compound is simply a more lipophilic analog of risedronate in which there is a decyloxy side chain addition, and the 1-OH is replaced by H. The logP values for risedronate and zoledronate are −2.9 and −3.5, whereas those for the lipophilic bisphosphonates are Fig. 2. Selection of lead compounds against intraerythrocytic P. falciparum. much more positive (Table S1): BPH-703 ¼ 2.5; BPH-811 ¼ (A) Representative screening results. Black, negative control (PBS, N ¼ 399 1.5, which can be expected to result in enhanced cell activity (17). wells); blue, positive control (artemisinin, N ¼ 181); green, test compounds To determine whether Plasmodium GGPPS is, in fact, a major (N ¼ 564 compounds, 10 μM). (B) Positive and negative controls alone, 0 target for the lipophilic bisphosphonates, we expressed P. vivax Z ¼ 0.72.(C) Therapeutic index versus IC50 plot showing BPH-703, BPH-811 GGPPS (whose X-ray structure is known and which has 73% hits. (D)pIC50 versus logP plot for hits. identity and 89% similarity to the P. falciparum protein) and de- assayed, the two most promising leads were both found to be termined dose-response curves for both inhibitors from which we simple derivatives of the commercial drugs risedronate and zole- find IC50 values of 1.2 μM (BPH-703) and 2.5 μM (BPH-811), K dronate in which the 1-OH group is replaced by an H, and the aryl corresponding to i values of 270 and 570 nM, consistent with group contains a C10 or C12, hydrophobic substituent that im- a GGPPS target in P. falciparum. To further test this idea, we car- proves logP and is expected to reduce bone binding, as shown ried out cell-growth inhibition “rescue” experiments, as reported previously with BPH-715 (Scheme 1) (19). previously for T. brucei (34) and human tumor cell lines (20),

Fig. 3. Crystallographic structure of the lipophilic zoledronate analog BPH-703 bound to PvGGPPS. (A) Crystallographic structures of IPP, FPP, GGPP, and a bisphosphonate bound to human, yeast, or P. vivax GGPPS, after figure 4 in ref. 36. (B) BPH-703 [cyan, Protein Data Bank (PDB) ID code 3RBM] bound to PvGGPPS superimposed on zoledronate (ZOL) structure (yellow, PDB ID code 3LDW ). Conserved Asp in red; Arg/Lys in blue. Note Val-153(B), Val-156(B), and Leu-157(B) are from the second GGPPS molecule in the dimer. (C) Two BPH-703 bound to the dimer with Leu-157 (orange) below the inhibitor. (D) GGPP bound to PvGGPPS showing Leu-157 inhibits tunnel formation. (E) Bound zoledronate showing absence of any ligand-binding tunnel. Leu-157 in orange.

4060 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1118215109 No et al. Downloaded by guest on September 27, 2021 in which we added either farnesol (FOH), geranylgeraniol In the case of the lipophilic-risedronate analog, BPH-811, we 2þ (GGOH), or solanesol (C45) to the growth medium. The results again find that the bisphosphonate head group binds to a ½Mg 3 (Fig. S5B) clearly indicated that only GGOH (at 20 μM) fully res- cluster, and the local structure is very similar to that seen with cued cell-growth inhibition caused by BPH-703, supporting a risedronate bound to T. cruzi FPPS (a 0.82 Å rmsd, protein align- GGPPS target. There was only a partial rescue with FOH, and ment; 0.50 Å rmsd, ligand∕Mg2þ local alignment, as described no effect with solanesol (Fig. S5C); similar results were seen with above), as shown in Fig. 4A. Moreover, the BPH-811 structure BPH-811 (Fig. S5 D and E). These results are of interest because is very similar to that seen with BPH-703, as shown in Fig. 4B.This the extremely potent Plasmodium GGPPS inhibitors risedronate very similar chain length and organization results in similar tunnel and zoledronate (16) have little activity in P. falciparum growth formation with BPH-811 as found with BPH-703, as shown in assays (4) and were not found to be hits in our initial screen, even Fig. 4C. Overall then, both PvGGPPS structures containing lipo- 2þ though they have GGPPS IC50 values of approximately 500 nM philic bisphosphonates are characterized by binding to a ½Mg 3 (Ta b l e S 1 , 16) due, we believe, to their poor cell permeability. cluster; close register of the risedronate and zoledronate rings in We next sought to determine how BPH-703 and BPH-811 bound the lipophilic bisphosphonates with their more hydrophilic parent to Plasmodium vivax GGPPS. compounds (in GGPPS and/or FPPS); close register between the hydrophobic side chains in both PvGGPPS inhibitors (a 1.19 Å Crystallographic Structures of PvGGPPS-Inhibitor Complexes. As rmsd for 10 carbons); and formation of a continuous hydrophobic we and others reported previously (30, 35, 36), there are four tunnel between the two molecules in the dimer into which the main ligand (substrate or inhibitor) binding sites in human and inhibitors bind. This tunnel formation is likely an important con- S. cerevisae GGPPS, illustrated in Fig. 3A. Two are polar-binding tributor to efficacy (in vitro and in vivo) because it may permit re- sites (a, c) where the diphosphates in the substrates (IPP and latively large structural changes to the inhibitor, facilitating cell FPP) bind, and two (b, d) are hydrophobic sites where the side penetration, without abolishing enzyme-inhibition activity. chain in the C15 (FPP) substrate or C20 (GGPP) product can bind. Bisphosphonate inhibitors (such as BPH-629, Scheme 1) can Isothermal Titration Calorimetry. Next, we investigated the thermo- bind with their phosphonate groups in either polar-binding site. dynamics of binding of BPH-703, BPH-811, as well as zoledronate There are thus four likely binding modes for the lipophilic bis- and risedronate, to PvGGPPS by using isothermal titration calori- phosphonates: a, b; a, d; b, c;orc, d (Fig. 3A). We next obtained metry (ITC). The head groups in the lipophilic bisphosphonates the crystal structures of BPH-703 and BPH-811 bound to PvGGPPS, using cocrystallization. Full crystallographic and structure refinement details are given in Table S2, crystallization details were as described previously (16), and electron-density results are in Fig. S6 A and B. With PvGGPPS (Protein Data Bank ID code 3LDW), zoledronate (shown in yellow in Fig. 3B) binds to the FPP diphosphate site and, unlike the situation found with human and S. cerevisae

GGPPS, is a potent inhibitor (IC50 510 nM, versus approximately CHEMISTRY 100 μM) of Plasmodium GGPPS (Table S1) due to the fact that the zoledronate bisphosphonate group binds to 3 Mg2þ in the PvGGPPS protein, just as in human FPPS (37, 38). With the zoledronate analog BPH-703 (cyan in Fig. 3B), it can be seen that the bisphosphonate head group also binds to 3 Mg2þ in the active site, just as the parent molecule zoledronate does. In addition, the imidazole group is situated in essentially the same position as found with zoledronate bound to PvGGPPS: a 0.79 Å rmsd based on a protein-structure alignment; 0.37 Å for the imidazole, Mg2þ and bisphosphonate P, O atoms, based on a ligand∕Mg2þ alignment. The polar head group in this lipophilic bisphosphonate thus binds in a very similar manner as does its zoledronate parent. However, the alkyl chain present in the lipophilic species BPH- 703 occupies the FPP substrate-hydrophobic site (b) rather than the GGPP product (or inhibitor) site (d) seen in other GGPPS structures (30, 35, 36). There are several reasons for this. First, the GGPP product site found in the yeast and human enzymes is blocked by R50 (Fig. S6C). Second, the site that houses the “tail” of the FPP- analog inhibitor, S-thiolo-FPP, is blocked by F88 (Fig. S6C). As a result, the BPH-703 side chain can now be seen to occupy a long hydrophobic pocket that spans both molecules in the dimer (Fig. 3 B and C). In order to accommodate the extended C10 side chain in BPH-703, both Leu-157 in the dimer are located below the chain termini of the inhibitors, as can be seen in Fig. 3C. In the PvGGPPS GGPP structure (Fig. 3D), the GGPP chain extends from site C (the PPi-binding site) to site b, but in this case the Fig. 4. Binding of BPH-703, BPH-811 to PvGGPPS, and comparison with rise- tunnel connecting the two molecules in the dimer is closed by dronate bound to T. cruzi FPPS [Protein Data Bank (PDB) ID code 1YHL]. (A) Superimposition of the lipophilic-risedronate analog BPH-811 bound to Leu-157 (orange), the bent ligand chain perhaps being a conse- PvGGPPS on the risedronate FPPS structure (cyan). (B) Comparison between quence of the presence of the double bond in the side chain. BPH-703 (cyan) and BPH-811 (yelllow) bound to PvGGPPS. (C) BPH-811 occu- Finally, in the zoledronate structure, there is no tunnel because pies the same hydrophobic tunnel in the dimer as does BPH-703. Leu-157 there is no extended side chain (Fig. 3E). in orange.

No et al. PNAS ∣ March 13, 2012 ∣ vol. 109 ∣ no. 11 ∣ 4061 Downloaded by guest on September 27, 2021 bind to PvGGPPS in a similar manner to that seen with risedronate parent has only very low activity (an IC50 ¼ 167 μM; ref. 13): and zoledronate, but how does alkyl substitution affect the thermo- Both species inhibit the GGPPS target, but only the lipophilic dynamics of binding? species have cell activity. The same trend is seen with BPH-811 In earlier work, we investigated the thermodynamics of binding and its parent, risedronate (Fig. S7 A and B). With BPH-811, of six bisphosphonates to FPPS from T. brucei (39), and related ΔG ¼ −9.3 kcal mol−1, ΔH ¼ −2.9 kcal mol−1, and ΔS ¼ 21 cal results have been reported for bisphosphonates binding to human deg−1 mol−1 (Table S1) whereas for the risedronate parent, FPPS (37, 38). With bisphosphonates having cationic side chains, ΔG ¼ −9.3 kcal mol−1, ΔH ¼ −2.5 kcal mol−1 and ΔS ¼ 25 cal binding was overwhelmingly entropy driven (39), due to the hy- deg−1 mol−1. So, in both cases, conversion of the parent bispho- drophobic effect. However, with neutral side chain containing sphonates to the more lipophilic analogs, expected to have en- species, binding was enthalpy driven. With PvGGPPS, binding is hanced cell/tissue permeability (17) and poor bone-binding neither overwhelmingly entropy nor enthalpy driven. For BPH- capacity (18), has little or no effect on ligand binding and only 703, the zoledronate analog, we find (Table S1) ΔG ¼ −8.7 kcal a small effect on enzyme inhibition—even though there are major mol−1 with ΔH ¼ −5 kcal mol−1 and ΔS ¼ 12 cal deg−1 mol−1 effects in cell-based assays. (Fig. 5A), to be compared with ΔG ¼ −9.5 kcal mol−1, ΔH ¼ It is also of interest to see how these results with GGPPS −4.8 kcal mol−1, and ΔS ¼ 16 cal deg−1 mol−1 for zoledronate compare with those we reported previously, with FPPS (39). As (Fig. 5B). So alkyl substitution (and loss of the 1-OH group) re- can be seen in Fig. 5C, the four ΔH, ΔS values fall on the line sults in only a small overall change in ΔG of approximately discussed previously for six bisphosphonates (at 2 pH values) 0.8 kcal mol−1. This result is encouraging because it indicates that binding to FPPS from T. brucei (39). What is surprising about making a major structural change in the inhibitor—adding a C12 the results is that the zoledronate/BPH-703 ΔS∕ΔH results are side chain—has only a relatively minor effect on GGPPS binding, very close, as are the risedronate/BPH-811 results (Fig. 5C). consistent with the enzyme-inhibition results (Table S1, where of Moreover, the two sets of results are themselves quite close, with course Mg2þ and IPP-binding are also involved). This observation neither the extreme entropy or enthalpy driven behavior seen with is in sharp contrast to the finding that BPH-703 is a potent inhi- FPPS (39), suggesting either that the charge center does not con- bitor of P. falciparum growth in vitro, whereas the zoledronate tribute to binding in GGPPS, or that the presence of the large neutral side chain simply reduces the effects of the cation center. The former view is supported by the observation in S. cerevisae GGPPS that the cation center is not essential for GGPPS inhibition (36), unlike the situation with FPPS (3).

In Vivo Compound Activity. To assess the in vivo activity of the lipophilic bisphosphonates, we used a P.chabaudi suppressive test. We first injected mice with 106 parasites followed by 10 mg∕kg of each bisphosphonate, i.p., for 4 d. With BPH-703, there was a 100% reduction in parasitaemia (Fig. 6A) and a 100% survival at day 14 (Fig. 6B). There was weight loss during treatment, but this was regained after treatment. In the case of BPH-811, there

Fig. 5. Isothermal titration calorimetry results. (A) Zoledronate binding to PvGGPPS. (B) BPH-703 binding to PvGGPPS. (C) ΔH∕ΔS plot for zoledronate, BPH-703, risedronate, and BPH-811 binding to PvGGPPS (in red) superposed on risedronate, zoledronate, and other bisphosphonates binding to T. brucei Fig. 6. Effect of BPH-703, BPH-811, and chloroquine (CQ) on mouse survival FPPS (the latter based on results in ref. 39). There is no evidence for the dom- and parasitemia. (A) Survival at 10 mg∕kg. (B) Parasitemia at 10 mg∕kg. (C) inance of entropy driven binding seen with cationic species binding to FPPS, Survival at 3 mg∕kg. (D) Parasitemia at 3 mg∕kg. Black circles, PBS control; consistent with the observation that the presence of a cationic species is not red triangles, BPH-703; blue triangles, BPH-811; and green triangles, chloro- important for GGPPS inhibition. quine-positive control.

4062 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1118215109 No et al. Downloaded by guest on September 27, 2021 was an approximately 80% decrease in parasitaemia and an 80% for GGPPS binding. It thus seems likely that these and related survival at day 14. We then carried out a second series of experi- lipophilic bisphosphonates will have improved activity over that ments using a reduced (3 mg∕kg, i.p.) dosing protocol. As with already reported with more hydrophilic bisphosphonates (4–15) the 10 mg∕kg dose, the best results were obtained with BPH- against other parasitic protozoa, such as T. brucei, T. cruzi, Leish- 703 (Fig. 6 C and D) where there was a 100% survival (Fig. 6C) mania spp., T. gondii, C. parvum, and E. histolytica. and only approximately 8% parasitaemia (Fig. 6D) 14 d after dosing. As can be seen in Table S1, BPH-703 has good in vitro IC50 Methods and logP values and little effect on human cell lines, as well as High-throughput screening intracellular cell-growth inhibition assays for good in vivo activity, and thus represents a potentially attractive P. falciparum were carried out using parasite lactate dehydrogenase and lead for further development. SybrGreen assays; full details are given in SI Methods. PvGGPPS crystals were obtained as described previously (16) with some modifications and were co- Conclusions crystallized with BPH-703 or BPH-811 to obtain the complex structures. Full Overall, the results presented here are of interest because they details are given in SI Methods. Structure determinations and refinements show that GGPPS is a promising target for antimalarials and that were carried out basically as described previously (16) with full details given the lipophilic analogs of the bone-resorption drugs zoledronate in SI Methods. PvGGPPS/bisphosphonate ITC experiments were carried out as described previously, for FPPS (39). and risedronate have activity both in vitro as well as in vivo. BPH-703 is a lipophilic analog of zoledronate in which the 1-OH group is removed (to prevent bone binding) and the ring is alky- ACKNOWLEDGMENTS. We thank Fabio T.M. Costa (Universidade Estadual de — Campinas, Sao Paulo, Brazil) for providing P. chabaudi stocks and protocols lated to enhance lipophilicity to increase cell permeability, de- for in vivo compound tests, Yigui Gao from University of Illinois at Urbana- crease bone binding, and enhance tissue distribution. BPH-811 is Champaign Macromolecular Crystallization Lab for providing crystallization an analog of risedronate in which, again, the 1-OH group is re- facilities, and Minje Ku and Minjung Ma for helping with parasite culture and moved and a lipophilic tail added. Both compounds have potent handling of infected animals. This work was supported by the United States Public Health Science, National Institutes of Health (Grant GM 65307) and activity in vitro and in vivo, whereas the parent drugs, zoledronate by a National Research Foundation of Korea grant funded by the Korean and risedronate, have only modest activity against P. falciparum government, Ministry of Education, Science, and Technology (2010-00395), in intraerythrocytic assays; and risedronate, although reducing and Korea Institute of Science and Technology Information. Use of the parasitaemia, has no effect on survival (15). Also of interest is Advanced Photon Source was supported by the US Department of Energy, the observation that the lipophilic bisphosphonate inhibitors in Office of Science, Office of Basic Energy Sciences, under Contract DE-AC02- 06CH11357. Use of the Life Science Collaborative Access Team Sector 21 was the Plasmodium GGPPS structures adopt very similar binding supported by the Michigan Economic Development Corporation and the poses to those of their parent compounds bound to both GGPPS Michigan Technology Tri-Corridor for the support of this research program as well as FPPS, and have similar enzyme activity and ΔG values, (Grant 085P1000817).

1. WHO (2010) World Malaria Report 2010. (WHO, Geneva, Switzerland), http://www. 20. Zhang Y, et al. (2009) Lipophilic bisphosphonates as dual farnesyl/geranylgeranyl who.int/malaria/world_malaria_report_2010/en/index.html. diphosphate synthase inhibitors: An X-ray and NMR investigation. J Am Chem Soc 2. Dondorp AM, et al. (2010) Artemisinin resistance: Current status and scenarios for con- 131:5153–5162. tainment. Nat Rev Microbiol 8:272–280. 21. Zhang Y, et al. (2010) Lipophilic pyridinium bisphosphonates: Potent gammadelta T 3. Ebetino FH, et al. (2011) The relationship between the chemistry and biological activity cell stimulators. Angew Chem Int Ed Engl 49:1136–1138. of the bisphosphonates. Bone 49:20–33. 22. Liu CI, et al. (2008) A cholesterol biosynthesis inhibitor blocks Staphylococcus aureus CHEMISTRY 4. Martin MB, et al. (2001) Bisphosphonates inhibit the growth of Trypanosoma brucei, virulence. Science 319:1391–1394. Trypanosoma cruzi, Leishmania donovani, Toxoplasma gondii, and Plasmodium falci- 23. Song Y, et al. (2009) Phosphonosulfonates are potent, selective inhibitors of dehydros- parum: A potential route to chemotherapy. J Med Chem 44:909–916. qualene synthase and staphyloxanthin biosynthesis in Staphylococcus aureus. J Med 5. Garzoni LR, et al. (2004) Selective in vitro effects of the farnesyl pyrophosphate synthase Chem 52:976–988. – inhibitor risedronate on Trypanosoma cruzi. Int J Antimicrob Agents 23:273 285. 24. Song Y, et al. (2009) Inhibition of staphyloxanthin virulence factor biosynthesis in Staphy- 6. Montalvetti A, et al. (2003) Farnesyl pyrophosphate synthase is an essential enzyme in lococcus aureus: In vitro, in vivo, and crystallographic results. JMedChem52:3869–3880. Trypanosoma brucei. In vitro RNA interference and in vivo inhibition studies. J Biol 25. Jomaa H, et al. (1999) Inhibitors of the nonmevalonate pathway of isoprenoid – Chem 278:17075 17083. biosynthesis as antimalarial drugs. Science 285:1573–1576. 7. Rodriguez N, et al. (2002) Radical cure of experimental cutaneous leishmaniasis by the 26. Chakrabarti D, et al. (1998) Protein prenyl transferase activities of Plasmodium falci- – bisphosphonate pamidronate. J Infect Dis 186:138 140. parum. Mol Biochem Parasitol 94:175–184. 8. Yardley V, et al. (2002) In vivo activities of farnesyl pyrophosphate synthase inhibitors 27. Tonhosolo R, et al. (2009) Carotenoid biosynthesis in intraerythrocytic stages of Plas- against Leishmania donovani and Toxoplasma gondii. Antimicrob Agents Chemother modium falciparum. J Biol Chem 284:9974–9985. 46:929–931. 28. Tonhosolo R, et al. (2010) Intraerythrocytic stages of Plasmodium falciparum bio- 9. Ling Y, et al. (2005) Bisphosphonate inhibitors of Toxoplasma gondi growth: In vitro, synthesize menaquinone. FEBS Lett 584:4761–4768. QSAR, and in vivo investigations. J Med Chem 48:3130–3140. 29. D’Alexandri FL, Gozzo FC, Eberlin MN, Katzin AM (2006) Electrospray ionization mass 10. Moreno B, et al. (2001) (31)P NMR of apicomplexans and the effects of risedronate on spectrometry analysis of polyisoprenoid alcohols via Liþ cationization. Anal Biochem Cryptosporidium parvum growth. Biochem Biophys Res Commun 284:632–637. 355:189–200. 11. Artz JD, et al. (2008) Targeting a uniquely nonspecific prenyl synthase with bispho- 30. Chen CK-M, et al. (2008) Inhibition of geranylgeranyl diphosphate synthase by bisphospho- sphonates to combat cryptosporidiosis. Chem Biol 15:1296–1306. nates: A crystallographic and computational investigation. J Med Chem 51:5594–5607. 12. Bruchhaus I, Jacobs T, Denart M, Tannich E (1996) Pyrophosphate-dependent phospho- 31. Lin FY, et al. (2010) Mechanism of action and inhibition of dehydrosqualene synthase. fructokinase of Entamoeba histolytica: Molecular cloning, recombinant expression Proc Natl Acad Sci USA 107:21337–21342. and inhibition by pyrophosphate analogues. Biochem J 316:57–63. 32. Cammerer SB, et al. (2007) Quinuclidine derivatives as potential antiparasitics. Anti- 13. Ghosh S, et al. (2004) Effects of bisphosphonates on the growth of Entamoeba histo- – lytica and Plasmodium species in vitro and in vivo. J Med Chem 47:175–187. microb Agents Chemother 51:4049 4061. 14. Singh AP, et al. (2010) Lipophilic bisphosphonates are potent inhibitors of Plasmodium 33. Zhang JH, Chung TD, Oldenburg KR (1999) A simple statistical parameter for use in eva- – liver-stage growth. Antimicrob Agents Chemother 54:2987–2993. luation and validation of high throughput screening assays. JBiomolScreen4:67 73. 15. Jordao FM, et al. (2011) In vitro and in vivo antiplasmodial activities of risedronate and 34. Martin MB, et al. (2002) Activity of bisphosphonates against Trypanosoma brucei rho- – its interference with protein prenylation in Plasmodium falciparum. Antimicrob desiense. J Med Chem 45:2904 2914. Agents Chemother 55:2026–2031. 35. Kavanagh KL, Dunford JE, Bunkoczi G, Russell RG, Oppermann U (2006) The crystal 16. Artz JD, et al. (2011) Molecular characterization of a novel geranylgeranyl pyropho- structure of human geranylgeranyl pyrophosphate synthase reveals a novel hexameric sphate synthase from Plasmodium parasites. J Biol Chem 286:3315–3322. arrangement and inhibitory product binding. J Biol Chem 281:22004–22012. 17. Mukkamala D, No JH, Cass LM, Chang TK, Oldfield E (2008) Bisphosphonate inhibition 36. Guo RT, et al. (2007) Bisphosphonates target multiple sites in both cis-and trans-pre- of a Plasmodium farnesyl diphosphate synthase and a general method for predicting nyltransferases. Proc Natl Acad Sci USA 104:10022–10027. cell-based activity from enzyme data. J Med Chem 51:7827–7833. 37. Rondeau JM, et al. (2006) Structural basis for the exceptional in vivo efficacy of bispho- 18. Mukherjee S, Song Y, Oldfield E (2008) NMR investigations of the static and dynamic sphonate drugs. Chem Med Chem 1:267–273. structures of bisphosphonates on human bone: A molecular model. J Am Chem Soc 38. Kavanagh KL, et al. (2006) The molecular mechanism of nitrogen-containing bispho- 130:1264–1273. sphonates as antiosteoporosis drugs. Proc Natl Acad Sci USA 103:7829–7834. 19. Mukherjee S, Huang C, Guerra F, Wang K, Oldfield E (2009) Thermodynamics of 39. Yin F, Cao R, Goddard A, Zhang Y, Oldfield E (2006) Enthalpy versus entropy-driven bisphosphonates binding to human bone: A two-site model. J Am Chem Soc binding of bisphosphonates to farnesyl diphosphate synthase. J Am Chem Soc 131:8374–8375. 128:3524–3525.

No et al. PNAS ∣ March 13, 2012 ∣ vol. 109 ∣ no. 11 ∣ 4063 Downloaded by guest on September 27, 2021