Plasmodium falciparum Atg18 localizes to the food vacuole via interaction with the multi-drug resistance protein 1 and phosphatidylinositol 3-phosphate.

Autophagy, a lysosome-dependent degradative process, does not appear to be a major degradative process in malaria parasites and has a limited repertoire of genes. To better understand the autophagy process, we investigated Plasmodiumfalciparum Atg18 (PfAtg18), a PROPPIN family protein, whose members like S. cerevisiae Atg18 (ScAtg18) and human WIPI2 bind PI3P and play an essential role in autophagosome formation. Wild type and mutant PfAtg18 were expressed in P. falciparum and assessed for localization, the effect of various inhibitors and antimalarials on PfAtg18 localization, and identification of PfAtg18-interacting proteins. PfAtg18 is expressed in asexual erythrocytic stages and localized to the food vacuole, which was also observed with other Plasmodium Atg18 proteins, indicating that food vacuole localization is likely a shared feature. Interaction of PfAtg18 with the food vacuole-associated PI3P is essential for localization, as PfAtg18 mutants of PI3P-binding motifs neither bound PI3P nor localized to the food vacuole. Interestingly, wild type ScAtg18 interacted with PI3P, but its expression in P. falciparum showed complete cytoplasmic localization, indicating additional requirement for food vacuole localization. The food vacuole multi-drug resistance protein 1 (MDR1) was consistently identified in the immunoprecipitates of PfAtg18 and P. berghei Atg18, and also interacted with PfAtg18. In contrast to PfAtg18, ScAtg18 did not interact with MDR1, which, in addition to PI3P, could play a critical role in localization of PfAtg18. Chloroquine and amodiaquine caused cytoplasmic localization of PfAtg18, suggesting that these target PfAtg18 transport pathway. Thus, PI3P and MDR1 are critical mediators of PfAtg18 localization.


Introduction
Malaria parasites undergo a multi-stage development in diverse intracellular and extracellular environments. Several of these stages are specialized for invasion, degradation of host cellular contents to obtain nutrients, generation of trafficking systems for import and export, and extensive reorganization of intracellular organelles to meet stage-specific needs. Since autophagy performs both degradative and biosynthetic functions, and several autophagy proteins have key roles in endosomal transport and organelle reorganization, investigation of autophagy during parasite development is warranted.
Autophagy involves degradation of dispensable cellular contents, including large protein complexes, protein aggregates, organelles and lipids in the lysosomes [1][2][3]. The cargo to be degraded is generally randomly selected, but certain cargo like mitochondria, peroxisomes and protein aggregates, is selectively recruited for autophagy [4]. About 35 Atg proteins participate in different stages of the autophagy process in S. cerevisiae and mammalian cells. The autophagy process can be broadly divided as: initiation by the Atg1 complex at a site known as the phagophore assembly site (PAS), assembly and vesicle Downloaded from http://portlandpress.com/biochemj/article-pdf/doi/10.1042/BCJ20210001/907305/bcj-2021-0001.pdf by guest on 14 April 2021 nucleation by the Vps34 complex at the PAS to form a double membrane cup-like structure called the phagophore, expansion of the phagophore by the Atg8 and Atg12 conjugation systems into a double membrane vesicle called the autophagosome, fusion of the autophagosome with lysosome, cargo degradation in the lysosome lumen and the efflux of degradation products into the cytoplasm for reuse [2,4].
Atg8 is a popular marker for studying autophagy, and it is associated with membrane structures that are commonly referred as puncta [5]. Multiple independent studies focussing on the Atg8 protein indicate a limited autophagy repertoire in Plasmodium species [6][7][8][9][10][11]. It is not clear whether autophagy performs a degradative function in malaria parasites, as the majority of Atg8 puncta remain outside the food vacuole, a lysosome-like organelle wherein some of the essential and best studied biochemical processes, including haemoglobin degradation, occur [10]. Additionally, food vacuole is the site of chloroquine action, and mutations in the P. falciparum food vacuole membrane transporters chloroquine resistance transporter (PfCRT) and multi-drug resistance protein 1 (MDR1) have been shown to confer resistance to multiple antimalarials, including quinolines [12][13][14][15][16][17][18][19]. Autophagy may have a role in endosomal transport to the food vacuole, as some Atg8 puncta were observed near or within the food vacuole [8]. A role of autophagy in the biogenesis and/or the processes associated with apicoplast, a nonphotosynthetic plastid remnant that is essential for biosynthesis of isoprenoid precursors in Plasmodium, has been supported by colocalization of Atg8 with apicoplast, adverse effect on apicoplast proliferation upon overexpression of Atg8 during P.
Although mechanistic details of the association of Atg8/autophagy pathway with apicoplast remain to be uncovered, autophagy may mediate transport of biomolecules and membranes to the apicoplast. Knockdown of Atg7 and chemical inhibition of Atg3, two essential enzymes of the Atg8 conjugation system, has been shown to impair parasite development, supporting an essential role of autophagy in parasite development [23,24]. Autophagy pathway has also been associated with resistance to chloroquine and artemisinin [12,[25][26][27], but a link between autophagy and the associated drug resistance is not clear.
Notably, a recent report showed that Plasmodium Atg12 and Atg5 form a noncovalent complex that mediated lipidation of Atg8 [28], revealing a non-canonical feature of Plasmodium autophagy, as the complex formation also requires Atg10 and Atg16 in the majority of other organisms. Thus, multiple lines of data conclude to an atypical, but essential, autophagy pathway in malaria parasites, which could offer attractive targets for new chemotherapeutic interventions.
All Plasmodium species encode for a highly conserved Atg18 homolog [10]. An earlier study showed that P. falciparum Atg18 (PfAtg18) is expressed in asexual erythrocytic stages and localizes to vesicular structures in the cytoplasm, and also colocalized with the apicoplast and the food vacuole [45].
This study also showed that PI3P is critical for localization of PfAtg18 to vesicular structures, and knockdown of PfAtg18 caused reduced conjugation of Atg8 to membranes and the loss of apicoplast, thereby indicated a crucial role of PfAtg18 in the regulation of apicoplast and autophagy. Using colocalization and inhibition approaches, a recent study reported that PfAtg18 has a role in the food vacuole dynamics and autophagy pathway [19]. Both PI3P and PI3K/Vps34 have been reported in P. falciparum erythrocytic stages [46,47]. However, the homologs of several proteins that are essential for autophagosome formation, including the subunits of Vps34 complex (Atg6 and Atg14), Atg2, Atg10 and Atg16, are either absent or remain to be identified. Given that Atg8 puncta in Plasmodium species resemble autophagosomes [8], it is likely that Atg8 puncta are generated by a non-canonical mechanism in malaria parasites.
We hypothesized that investigation of PfAtg18 could provide insights into the formation of Atg8 puncta and the autophagy-associated drug resistance. Our data indicate that food vacuole localization is likely a conserved feature of Plasmodium Atg18, which is mediated via interaction with PI3P and MDR1, and altered upon treatment of parasites with chloroquine and amodiaquine.

Experimental procedures Materials
All the biochemicals were from Sigma or Serva unless otherwise mentioned. The parasite culture reagents were from Lonza and Thermo Fisher Scientific. Restriction enzymes and DNA modifying enzymes were from New England Biolabs and Thermo Fisher Scientific. DNA isolation kits were from QIAGEN
berghei ANKA and infection was monitored regularly by observing Giemsa stained blood smears of the tail snips of infected mice. The infected mice were euthanized at 10-15% parasitemia, blood was collected in Alsever's solution (2.05% glucose, 0.8% sodium citrate, 0.055% citric acid and 0.42% sodium chloride) by cardiac puncture, parasites were purified by saponin lysis method, and the parasite pellets were processed for isolation of genomic DNA using the Puregene Blood Core kit according to the manufacturer's protocol.

Expression and purification of recombinant proteins
The PfAtg18 C-terminal coding region (PfAtg18ct: 571-1143 bps) was amplified from P.
falciparum cDNA using PfATG18-F2/Atg18expR primers (Table S2), and cloned into the pCR2.1 vector to obtain pCR2.1-PfAtg18ct. The insert was subcloned into pET32a at BamHI/HindIII sites to obtain pET32a-PfAtg18ct, which was transformed into BL21-CodonPlus(DE3)-RIL cells. pET32a-PfAtg18ct would express Thioredoxin/His-PfAtg18ct (Trx/His-PfAtg18ct) fusion protein, which was purified from IPTG induced cell under denaturing conditions. Briefly, the induced cell pellet was lysed in urea buffer (8M urea, 50 mM NaH 2 PO 4 , 500 mM NaCl, pH 8.0; 5 ml buffer/g pellet), sonicated using the SONICS Vibra Cell Ultrasonic Processor (9 secs pulses at 20% amplitude for 4 min), centrifuged, and the supernatant was separated. Imidazole (10 mM final), Triton-X 100 (0.5% final) and β-ME (5 mM final) were added to the supernatant, and incubated with Ni-NTA agarose resin (0.25 ml slurry/g weight of the initial pellet for 30 min at room temperature). The resin was washed with 50x column volume of wash buffer 1 (urea buffer with 30 mM imidazole, 0.5% Triton-X 100 and 5 mM β-ME) and wash buffer 2 (urea buffer with 50 mM imidazole). The bound proteins were eluted with elution buffer (250 mM imidazole in urea buffer) and was amplified from the genomic DNA of wild type BY4741 strain as two fragments using SCATG18EXPF/SCATG18MBR and SCATG18MBF/SCATG18EXPR primer sets. The two fragments were recombined using the primers SCATG18EXPF/SCATG18EXPR and cloned into the pGT-GFPbsc vector at BamHI/XhoI sites to obtain pGT-GFPScAtg18-Bm. The pGT-GFPbsc plasmid has been described previously [54]. For elimination of BglII site, pGT-GFPScAtg18-Bm was used as a template using primer sets SCATG18EXPF/SCATG18MBGR and SCATG18MBGF/SCATG18EXPR. The two fragments were recombined using SCATG18EXPF/SCATG18EXPR primers and cloned into the pGT-GFPbsc plasmid at BamHI/XhoI sites to obtain pGT-ScAtg18-BmBgm. The ScAtg18BmBgm insert was subcloned into the pGEX-6P-1 vector at BamHI/XhoI sites to generate pGEX/ScAtg18. All inserts were sequenced at the Automatic DNA Sequencing Facility of CCMB to ensure that they were free of undesired mutations. remove cross-reactive antibodies to E. coli proteins as has been described earlier [10]. The purified antibodies were stored in 50% glycerol with 0.01% sodium azide at -30°C.

Western blotting
For assessing reactivity and specificity of purified anti-Atg18 antibodies, the mock-transformed E.
coli lysate, the total lysate of induced PfAtg18ct-expressing E. coli, purified Trx/His-PfAtg18ct, lysates of wild type and recombinant D10 parasites, and RBC lysate were resolved on 12% SDS-PAGE gel. The proteins were transferred onto the Immobilon-P membrane, which was incubated with blocking buffer (3% skim milk in TBST), followed by incubation with rat anti-Atg18 (at 1/5000 dilution in blocking buffer) or mouse anti-β actin (1/500 dilution in blocking buffer) antibodies. The membrane was washed with blocking buffer, incubated with secondary antibodies (goat anti-Rat IgG-HRP or goat anti-Mouse IgG-HRP at 1/20,000 dilution in blocking buffer), washed again with TBST, and the signal was developed using the v/v β-ME and 0.25 M Tris-Cl, pH 6.8), centrifuged at 23755g for 20 min, the supernatants were separated and equal amounts of supernatant samples were processed for western blotting using anti-Atg18 or mouse anti-β-actin antibodies, followed by appropriate secondary antibodies as described above.
knowlesi Atg18 (PkAtg18, PlasmoDB gene identifier: PKNH_0812700) was amplified using PkA18-F/PkA18-R primers from the P. knowlesi genomic DNA, cloned into the pGT-GFPbsc vector at BamHI/XhoI sites to generate pGT-GFP/PkAtg18, which was further subcloned into pSTCII-GFP vector at BglII/XhoI sites to obtain pSTCII-GFP/PkAtg18 transfection plasmid. The PvAtg18 and PkAtg18 genes contain a single intron, which would be spliced from the transcript by the parasite splicing machinery.
PIPm and ALCAm mutants of PfAtg18 were generated by recombination PCR method using primers containing the desired mutations (FRRG to FAAG in PIPm and WLCL to ALCA in ALCAm) [56].

Praveen Balabaskaran Nina) using cDD-F/cDD-R primers and cloned into the pGT-GFPbsc plasmid at
KpnI/XhoI sites to obtain pGT-cDD HA plasmid. The cDD HA was excised from the pGT-cDD HA plasmid and subcloned into the HBA18KO plasmid at KpnI/XhoI sites to obtain HBA18cDD HA plasmid. The PbAtg18 coding region was amplified from P. berghei genomic DNA using PbA18Rep-F2/PbA18Rep-R primers and cloned into the HBA18cDD HA plasmid at NotI/KpnI sites, replacing the 5'UTR flank1, to obtain HBA18/cDD HA kd transfection plasmid. The HBA18KO, HBA18/GFPki and HBA18/cDD HA kd plasmids were digested with EcoRV to obtain linear transfection constructs, which were used for transfection of P.
The PfCRT (PlasmoDB gene identifier: PF3D7_0709000) coding region was amplified from P.
falciparum cDNA using PfCRT-Fepi/PfCRT-Repi primers and cloned into the pGEM-T easy vector to generate pGEM-PfCRT plasmid. The mCherry coding region was amplified from the pmCherry-N1 (Clontech) plasmid using mCher-F/mCher-R primers and cloned into the pGT-GFPbsc plasmid at BglII/XhoI sites, replacing the GFP region, to obtain pGT-mCherry. The PfCRT insert was excised from the pGEM-PfCRT plasmid with BglII/KpnI and subcloned into the similarly digested pGT-mCherry plasmid to generate pGT-PfCRT/mCherry. The HB-DJ1KO plasmid was modified to express PfCRT/mCherry plasmid with KpnI/XhoI and cloned into the similarly digested HB-Ddi plasmid to obtain HB-PfCRT/mCherry transfection plasmid.
For construction of a PfAtg18 knock-down plasmid, the PfAtg18 coding region (flank1) and 3'UTR (flank2) were amplified from P. falciparum genomic DNA. The internal KpnI site in flank1 was mutated by site-directed mutagenesis without changing the encoded amino acids, which involved amplifying it in two fragments (PFA18FL1-F/PFA18F1-RKMUT and PFA18F1-FKMUT/PFA18FL1-RKI primers), followed by recombination of the fragments using PFA18FL1-F/PFA18FL1-RKI primers. Flank 2 was amplified using PFA18FL2-F /PFA18FL2-R primers. The flank1 and flank2 were sequentially cloned into the HBA18/cDD HA kd plasmid at NotI/KpnI and AvrII/KasI sites, respectively, to generate HBPFA18/cDD HA plasmid. A synthetic GlmSAc fragment (GenScript) was cloned into the HBPFA18/cDD HA plasmid at XhoI/AgeI site to generate HBPFA18/cDD HA /GlmSAc plasmid.
The coding regions and flanks in all the transfection plasmids were sequenced to ensure that they were free of undesired mutations, and the presence of different regions was confirmed by digestion with region-specific restriction enzymes. The transfection constructs were purified using the NucleoBond® Xtra Midi plasmid DNA purification kit.
Recombinant parasites were usually maintained in the presence of blasticidin, which was withdrawn during experiments. For co-transfection, the pCENv3-GFP/PfAtg18-transfected P. falciparum D10 parasites were cultured in the absence of blasticidin for 2 cycles. The ring stage-infected RBCs were transfected with HB-PfCRT/mCherry plasmid, and subjected to selection (0.5 µg/ml blasticidin and 1 nM WR99210) from day 2-12, and thereafter maintained in the presence of 1 nM WR99210 till the emergence of resistant parasites.
For obtaining PfAtg18 knock-down parasites, the P. falciparum D10 ring-stage parasites were transfected with HBPFA18/cDD HA /GlmSAc plasmid and selected with 1 nM WR99210 till the emergence of resistant parasites, and thereafter maintained in the presence of 10 μM trimethoprim.

Transfection of P. berghei
P. berghei ANKA was maintained in BALB/c mice as described in the parasite culture section. The infected mice were euthanized at 6-8% parasitemia and the blood was collected in Alsever's solution by cardiac puncture. The trophozoite stage parasites were purified on a 65% Histodenz gradient at 360g with 0 deceleration using a swinging-bucket rotor, and cultured in RPMI 1640-FBS medium (supplemented with 2 g/litre sodium bicarbonate, 2 g/litre glucose, 25 µg/ml gentamicin, 300 mg/litre glutamine and 20% FBS) for 12-15 hrs at 35°C with shaking at 50 rpm. The culture was centrifuged at 1398g for 5 min and the parasite pellet, which mostly contained mature schizonts, was resuspended in 100 µl Nucleofector solution (Lonza) containing ~5 µg of desired circular (pSTCII-GFP/PfAtg18, pSTCII-GFP/PvAtg18, pSTCII-GFP/PkAtg18, pSTCII-GFP/PIPm, pSTCII-GFP/ALCAm) or linear (HBA18KO, HBA18/GFPki and HBA18/cDD HA kd) transfection plasmids. The content was electroporated using Amaxa Nucleofector device as has been described previously [60]. 120 µl of RPMI 1640-FBS medium was added to the cuvette, and the entire sample was injected intravenously into a naive mouse. The mouse was given pyrimethamine in drinking water (70 µg/ml) for 7 days for selection of transfected parasites. For transfections with HBA18/cDD HA kd, mice were given water with trimethoprim (400 µg/ml) for 2-3 hrs before the infection, and then given water containing trimethoprim+pyrimethamine unless otherwise mentioned. Pure clonal lines of HBA18/cDD HA kd transfected parasites were obtained by the dilution cloning method, which involved intravenous infection of 10 mice, each with 0.5 parasite. The parasites were harvested for various analyses as mentioned elsewhere.

Localization of wild type and mutant Atg18 proteins in parasites
For localization of PfAtg18 and its mutants, GFP/PfAtg18-, GFP/PIPm-and GFP/ALCAm- The localization of GFP/PfAtg18, GFP/PIPm, GFP/ALCAm, GFP/PvAtg18 and GFP/PkAtg18 was assessed in P. berghei trophozoites and schizonts. For localization in trophozoites, 10-20 µl blood was collected in Alsever's solution from the tail snips of infected-mice at 4-5% parasitemia, washed with PBS and processed for live cell fluorescence microscopy as described for GFP/PfAtg18-expressing P.
falciparum parasites. For localization in schizonts, 50-100 µl blood was collected by retro-orbital bleeding and cultured in 10 ml RPMI 1640-FBS medium to obtain schizont stage as described in the transfection of P. berghei section. The culture was processed for live cell fluorescence microscopy as described for GFP/PfAtg18-expressing P. falciparum parasites. Asynchronous cultures of GFP/PfAtg18-, GFP/PIPm-, GFP/ALCAm-, GFP/PvAtg18-, and GFP/PkAtg18-expressing P. falciparum and P. berghei were harvested at 10-15% parasitemia by saponin lysis. The pellets were processed for western blot using mouse anti-GFP (at 1/500 dilution in blocking buffer) or mouse anti-β-actin antibodies, followed by detection with appropriate secondary antibodies as described in the Western blotting section.
GFP/ScAtg18-expressing P. falciparum parasites were processed for live cell microscopy and western blotting as described for GFP/PfAtg18-expressing P. falciparum parasites.

Colocalization of PfAtg18
For colocalization with lysotracker, GFP/PfAtg18-expressing P. falciparum trophozoites were cuvettes were used for far-UV and near-UV CD measurements, respectively. The instrument was set on wavelength spectrum scan mode and the spectra were recorded from 190 nm to 250 nm and 250 nm to 350 nm for far-UV and near-UV CD measurements, respectively. Five spectra were averaged to increase the signal-to-noise ratio, and readings were plotted using the Origin 2020b software.

Inhibition experiments
A variety of inhibitors and antimalarials were assessed for their effects on PfAtg18 localization.
The concentrations of compounds were normalized to their IC 50 concentrations, which were either determined in this study or taken from previous reports (  For comparison of growth rates, 6 months old BALB/c mice were divided into three groups and infected intraperitoneally with wild type or PbAtg18-KD parasites (10 6 parasites/mouse). One group was infected with wild type P. berghei ANKA, 2 nd group was infected with PbAtg18-KD and not given To determine the effect of PbAtg18 knock-down on expression, 6 BALB/c mice were infected intraperitoneally with PbAtg18-KD parasites and given trimethoprim from 2-3 hours before the infection till the parasitemia reached to 10-15% parasitemia. Three mice were euthanized while they were still under trimethoprim and the blood was collected for isolation of parasites (+TMP parasites). Trimethoprim was withdrawn from the remaining 3 mice, and they were euthanized 24 hours later, blood was collected for isolation of parasites (-TMP parasites). The parasite pellets were processed for western blotting using rabbit anti-HA or mouse anti-β-actin antibodies, followed by appropriate secondary antibodies as described in the western blotting section.

Immunoprecipitation of PbAtg18 and PfAtg18
Parasites were isolated from asynchronous cultures of GFP-expressing and GFP/PfAtg18expressing P. falciparum at 10-15% parasitemia as described in the parasite culture section. Wild type P.
berghei ANKA and PbAtg18-KI parasites were maintained in 6-10 weeks old naïve BALB/c mice and parasites were isolated as described in the parasite culture section. The parasite pellet was resuspended in 5x pellet volume of the lysis buffer (10 mM Tris, 150 mM NaCl, 0.5 mM EDTA, 0.5% NP-40, pH 7.5, protease inhibitor cocktail), subjected to 5 cycles of freeze-thaw and 5 passages through a 26.5 G needle.
The sample was incubated in ice for 30 min, centrifuged at 25000g for 30 min at 4 o C, and the supernatant was transferred into a fresh tube. The pellet was re-extracted with 3x pellet volume of the lysis buffer as shaking. The beads were washed with wash buffer 1 (10 mM Tris, 150 mm NaCl, 0.5 mM EDTA, pH 7.5, protease inhibitor cocktail), followed by with wash buffer 2 (10 mM Tris, 300 mm NaCl, 0.5 mM EDTA, pH 7.5, protease inhibitor cocktail). The bound proteins were eluted by boiling the beads in 100 µl 2x SDS PAGE sample buffer for 15 min, and the eluate was processed for western blotting and mass spectrometry.
Aliquots of the eluates were assessed for the presence of GFP/PfAtg18 or PbAtg18/GFP along with appropriate control samples (input, pellet post-extraction, flow through, washes, and beads after elution) by western blotting using rabbit anti-GFP antibodies as described above in the western blotting section.

Mass spectrometry of immunoprecipitates
The immunoprecipitate eluates were run on a 12% SDS-PAGE gel till the pre-stained protein ladder completely entered into the resolving gel. The gel was stained with coomassie blue, destained, and the gel were compared with those in the GFP sample. Common proteins were excluded, and unique proteins common to 3 biological replicates with a minimum of 1 unique peptide were considered.

Plasmodium Atg18 proteins localize to the food vacuole
PfAtg18 has been shown to be expressed in erythrocytic stages and localized to vesicular structures throughout the parasite, some of which also colocalized with the food vacuole and apicoplast [19,26,45].
To determine the expression of PfAtg18 in asexual erythrocytic stages, we produced the C-terminal region of PfAtg18 as a Trx/His-PfAtg18ct fusion protein and generated anti-PfAtg18 antibodies ( Figure S1A). PfAtg18 shares about 91% and 97% sequence identities with the Atg18 proteins of mouse and monkey/human malaria parasites, respectively [10]. However, food vacuole has two major morphologies in Plasmodium species: P. falciparum exhibits a single large food vacuole in all asexual erythrocytic stages, whereas other Plasmodium species, including P. vivax, P. knowlesi and P. berghei, exhibit multiple vesicles most of which fuse into a single large vesicle in the schizont stage [61]. Hence, we investigated the localization of selected Plasmodium Atg18 proteins in P. falciparum and P. berghei, which represent the two food vacuole morphologies. For localization in P. berghei, GFP/PfAtg18 was expressed using an episomally maintained plasmid, and its expression was confirmed by western blot using anti-GFP antibodies ( Figure S4A). These parasites showed GFP/PfAtg18 signal at multiple loci, which also contained  To investigate if PfAtg18 follows endosomal transport to the food vacuole, we evaluated its localization in GFP/PfAtg18-expressing P. falciparum parasites upon treatment with the inhibitors of vesicle trafficking (brefeldin A and bafilomycin) and fusion (bafilomycin and NH 4 Cl). None of the inhibitors significantly altered PfAtg18 localization ( Figure S7), suggesting that it does not take endosomal trafficking route, which is in line with the absence of a signal/transmembrane sequence in PfAtg18. About 30% of the NH 4 Cl-treated parasites showed PfAtg18 around an enlarged food vacuole ( Figure S7). NH 4 Cl has been shown to inhibit lysosomal degradation activity by alkalinizing the compartment [62], hence, the food vacuole enlargement could be due to accumulation of undegraded haemoglobin, as has been observed in parasites treated with E64 and pepstatin A [63][64][65][66][67]. Only a small fraction of NH 4 Cl-treated parasites, particularly late ring and early trophozoite stages, contained multiple PfAtg18 foci near the parasite periphery or in the cytoplasm ( Figure S7), which may be due to failure of vesicles to fuse to form the food vacuole, as has also been reported earlier [68]. The PfAtg18 vesicles in NH 4 Cl-treated parasites may be part of the haemoglobin trafficking pathway, as a recent report proposed that PfAtg18 transport to the food vacuole uses haemoglobin trafficking pathway [19].

FRRG and WLCL motifs are essential for PI3P-dependent food vacuole localization of PfAtg18
PfAtg18 showed maximum sequence identity (34.7%) with the Pichia angusta Atg18 (PaAtg18) at amino acid level. Sequence analysis and homology modelling of PfAtg18 on the PaAtg18 structure (PDB id: 5LTD) predicted seven WD40 repeats in PfAtg18, which form a seven bladed β-propeller structure with two PI3P-binding sites, indicating that PfAtg18 is a PROPPIN family protein [34,39], as are ScAtg18 and WIPI2 ( Figure S8). The PI3P-binding sites are formed by the FRRG motif and amino acids from the 23 rd β strand in blade 6 ( Figure S8). PfAtg18 and other Plasmodium Atg18 proteins also share a positionally conserved "WLCL" motif that resembles the LC3-interacting region/Atg8-interacting motif (LIR/AIM) "WxxL" in a variety of proteins, which interact with Atg8. The LIR/AIM motif is mostly preceded by negatively charged amino acids [49], whereas the "WLCL" motif is not. Nonetheless, the "WLCL" motif is conserved at "WL" positions in the majority of Atg18 proteins ( Figure S8) and forms the major part of 23 rd β strand in blade 6 that contains the 2 nd PI3P-binding site in S. cerevisiae Hsv2 [34,39] shown to localize to punctate structures close or around the food vacuole; based on co-localization of PfAtg18 with anti-PI3P antibodies and inhibition studies using the PI3K inhibitor wortmannin, this localization was suggested to be mediated by PI3P [19]. However, Atg18 localization was restricted to the food vacuole in our studies whether it was expressed in a homologous or a heterologous system ( Figure 1C and Figure 2). Furthermore, both episomally and endogenously expressed Atg18 showed similar localization ( Figure 2). Since GFP-2xFYVE, a routinely used marker for PI3P localization studies [70], indicated abundant PI3P in the food vacuole membrane of P. falciparum [46], we investigated if the food vacuole localization of PfAtg18 is mediated through its interaction with PI3P on the food vacuole membrane. We first treated GFP/PfAtg18-expressing P. falciparum parasites with the PI3K inhibitor LY294002 and scored the parasites for PfAtg18 localization. The treatment resulted in a dose-dependent diffuse localization of PfAtg18 throughout the parasite in a significantly larger number of parasites (18.4% ± 2.8 at 1IC 50 concentration to 59.8% ± 0.9 at 5IC 50 concentration) than in control parasites (1.6% ± 1.1) ( Figure 4A), suggesting that PI3P is critical for PfAtg18 localization to the food vacuole. We next mutated the FRRG motif to FAAG, and the mutant (PIPm) was expressed as a GFP-fusion protein (GFP/PIPm) in P. falciparum and P. berghei ( Figure S4A and S4C). The GFP/PIPm was present outside the food vacuole in the cytoplasm of all P. falciparum ( Figure 4A) and P. berghei parasites ( Figure S9), indicating that FRRG motif is essential for localization of PfAtg18 to the food vacuole. We also mutated the "WLCL" motif to "ALCA" and expressed the mutant (ALCAm) as a GFP-fusion protein (GFP/ALCAm) in both P.
falciparum and P. berghei ( Figure S4A and S4C). The GFP/ALCAm was localised outside the food vacuole in the cytoplasm of P. falciparum ( Figure 4A) and P. berghei ( Figure S9), indicating that "WLCL" motif is essential for localization of PfAtg18 to the food vacuole. The localization of both GFP/PIPm and GFP/ALCAm was similar to that of a cytoplasmic GFP construct (GFPcyt) ( Figure 4A), confirming that FRRG and WLCL motifs are essential for the food vacuole localization of PfAtg18.
To further investigate whether the loss of food vacuole localization of PfAtg18 mutants was due to their inability to interact with PI3P, recombinant GST/PfAtg18, GST/PIPm, GST/ALCAm, GST/ScAtg18 and GST were produced in BL21(DE3) cells ( Figure S10), and assessed for binding with PI3P.
GST/PfAtg18 bound with PI3P just like recombinant GST/ScAtg18 ( Figure 4B), whereas GST/PIPm, GST/ALCAm and GST did not bind, confirming that FRRG and WLCL motifs are critical for PfAtg18 interaction with PI3P, and this interaction is essential for food vacuole localization. Since "WLCL" motif does not directly contribute to the 2 nd PI3P-binding site, rather, it forms the major part of 23 rd β strand in blade 6 that contains the 2 nd PI3P-binding site in S. cerevisiae Hsv2, we compared CD spectra of GST/PfAtg18 and GST/ALCAm proteins. The far-UV CD spectrum of GST/ALCAm was different from that of GST/PfAtg18, whereas the near-UV CD spectra were similar, suggesting a local structural perturbation in GST/ALCAm resulting in the loss of binding to PI3P ( Figure S11).

Interaction with PI3P is not the sole requirement for food vacuole localization of PfAtg18
The "FRRG" motif in ScAtg18 and WIPI2 is essential for interaction with PI3P and targetting to the PI3P-rich sites, and the loss of autophagic function of "FTTG" mutant of ScAtg18 can be fully restored by fusing it with the 2×FYVE domain, which specifically binds PI3P [69]. Hence, we investigated if the "FRRG" motif of ScAtg18 is sufficient to recruit it to the food vacuole of P. falciparum. GFP/ScAtg18 was expressed in P. falciparum ( Figure S4D), and the parasites were assessed for localization of GFP/ScAtg18.
To our surprise, GFP/ScAtg18 did not localize to the food vacuole, rather, it remained in the cytoplasm just like GFPcyt (Figure 5), indicating that interaction with PI3P alone is not sufficient and additional factors could be involved in recruitment of PfAtg18 to the food vacuole.

MDR1 mediates PfAtg18 localization to the food vacuole
To identify Plasmodium Atg18-associated proteins, we immunoprecipitated PbAtg18/GFP and GFP/PfAtg18 ( Figure S12), and the samples were subjected to mass spectrophotometry. 10 unique proteins were identified in the GFP/PfAtg18 immunoprecipitate (Table 1) and 37 unique proteins were identified in the PbAtg18/GFP immunoprecipitate (Table S1). MDR1 and DnaJ proteins were common in both the samples, suggesting association of Plasmodium Atg18 with MDR1 and DnaJ. The GFP/PfAtg18 immunoprecipitate also contained HSP40, a Plasmodium exported protein, PHISTb, three ribosomal proteins, and a proteasome regulatory subunit. MDR1 was selected for further investigation as: 1) excluding the bait protein Atg18, it had the 3 rd highest score in GFP/PfAtg18 immunoprecipitate (Table 1) and the highest score in the GFP/PbAtg18 immunoprecipitate (Table S1), 2) it is localized on the food vacuole membrane and associated with resistance to multiple antimalarials [13][14][15][16]71]. MDR1 is a multi-pass transmembrane protein belonging to the ABC family of transporters; it is predicted to have 11 transmembrane helices, an inside domain and a cytoplasmic domain ( Figure 6A). Hence, to investigate whether PfAtg18 directly interacts with MDR1, we produced recombinant GST-tagged inside (GST/MDRID) and cytoplasmic (GST/MDRCD) domains ( Figure 6B and C), and tested both the domains for interaction with GST/PfAtg18, GST/ScAtg18 and GST using dot blot protein overlay assay.
GST/MDRCD, but not GST/MDRID, interacted with GST/PfAtg18 ( Figure 6D). Both the recombinant MDR domains did not interact with GST and GST/ScAtg18, indicating that PfAtg18-MDR1 interaction is direct and specific. We also compared interaction of wild type PfAtg18 and its mutants with GST/MDRCD by ELISA. The PIPm and ALCAm mutants showed significantly weaker interaction with GST/MDRCD than wild type Atg18 ( Figure S13), which further substantiates PfAtg8-MDR1 interaction. Hence, MDR1, in addition to PI3P, plays a critical role in localization of PfAtg18 to the food vacuole.

Food vacuole-localization of PfAtg18 is sensitive to chloroquine and amodiaquine
The food vacuole of malaria parasites is the site of some of the essential and best studied biochemical processes, including haemoglobin degradation and haemozoin formation. Mutations in PfCRT and MDR1, which are also present in the food vacuole membrane, confer resistance to multiple antimalarials, including the quinolines [12][13][14][15][16][17][18][19]. Altered distribution of Atg8 puncta and Thr38Ile mutation of PfAtg18 have been associated with resistance to chloroquine and artemisinin, respectively [12,[25][26][27].
However, a direct link between autophagy and drug resistance is yet to be identified. Hence, we treated GFP/PfAtg18-expressing P. falciparum parasites with quinolines (chloroquine, amodiaquine, mefloquine, halofantrine and quinine) and evaluated for localization of PfAtg18. Chloroquine and amodiaquine caused diffuse localization of PfAtg18 throughout the parasite in a significant number of parasites, which became more pronounced with increasing drug concentrations, whereas mefloquine, halofantrine and quinine did not affect the localization (Figures 7 and 8), indicating that the effect of chloroquine and amodiaquine on PfAtg18 localization is specific. The immunoblot of quinoline-treated parasites did not show any noticeable change in PfAtg18 levels compared to that in control parasites ( Figure S14A). The treatment of GFP/PfAtg18-expressing P. falciparum parasites with artemisinin neither affected PfAtg18 localization ( Figure 7) nor the levels ( Figure S14A). A recent study reported that PfAtg18 participates in food vacuole dynamics. Hence, altered food vacuole localization of PfAtg18 could affect multiple food vacuoleassociated processes like haemoglobin degradation, which could subsequently affect the action of these drugs, as this degradation produces heme for quinoline action and iron for artemisinin activation.

Food vacuole metabolism inhibitors affect PfAtg18 localization
Haemoglobin degradation by asexual erythrocytic stage parasites provides amino acids for parasite protein synthesis and has also been proposed to keep the infected erythrocyte osmotically stable [72,73].
Although not yet reported in Plasmodium, lysosome is the primary site of lipid catabolism that involves transport of lipid droplets to lysosomes by lipophagy and subsequent degradation of lipid droplets by lysosomal lipases. Lipid droplets have been shown to be associated with the Plasmodium food vacuole [75], but specific food vacuole lipases are yet to be characterized according to our knowledge.
Hence, we treated GFP/PfAtg18-expressing P. falciparum parasites with several inhibitors of the food vacuole-associated metabolic processes and evaluated for PfAtg18 localization. E64 and pepstatin A caused a profound ring-like localization of GFP/PfAtg18 around the food vacuole, suggesting that it is associated with the food vacuole membrane (Figure 9). The ring-like localization pattern was seen in a much larger number of parasites in case of E64 and pepstatin A than control parasites. The dot-like pattern might be of the PfAtg18 in the food vacuole lumen or due to processing of PfAtg18 on the luminal side by the food vacuole-resident proteases. A significant increase in the number of parasites with ring-like pattern upon treatment with E64 and pepstatin A supports this point. We also observed intense signal at a single site in the peri-food vacuole region, which was particularly evident in parasites treated with E64 and pepstatin A. Orlistat, an inhibitor of pancreatic lipases and fatty acid synthase [76,77], has been shown to block the development of malaria parasites by inhibition of triacylglycerol hydrolysis [78]. A significant number of orlistat-treated parasites showed diffuse PfAtg18 localization all over the parasite, which could be due to decreased levels of fatty acids for membrane synthesis and phosphatidylinositide precursors ( Figure 9). Localization of GFP/PfAtg18 in parasites treated with pristimerin (inhibitor of monoacylglycerol lipase and proteasome) and epoxomicin (proteasome inhibitor) was nearly identical to that in control parasites ( Figure 9). The immunoblot of inhibitor treated-parasites and control parasites had similar PfAtg18 levels ( Figure S14B), which ruled out any effect of treatment on PfAtg18 expression level.
To address whether food vacuole integrity was intact during inhibitor treatment experiments, we treated the parasites expressing PfCRTmCherry, a food vacuole marker, with E64 and chloroquine, which inhibit haemoglobin degradation and haemozoin formation in the food vacuole, respectively. In both the cases, the number of parasites showing food vacuole-associated PfCRTmCherry was comparable to the DMSO control ( Figure S15), indicating that food vacuole integrity was not compromised in our experiments. As observed in the case of PfAtg18, the E64-treated parasites also showed an enlarged ring of PfCRTmCherry due to enlargement of the food vacuole.

Atg18 is crucial for parasite development
To determine the role of Atg18 during parasite development, we attempted to knock-out and knockdown PbAtg18. Multiple attempts to generate knock-out parasites were not successful, whereas knockdown parasites (PbAtg18-KD) were readily obtained by replacing the wildtype PbAtg18 coding sequence with PbAtg18/cDD HA coding sequence ( Figure 10A and B). cDD is a mutant of E. coli DHFR, and cDDfusion proteins undergo degradation in the absence of trimethoprim (TMP) but not in the presence of TMP, thereby producing a knock-down phenotype at the protein level. The PbAtg18/cDD HA fusion protein was localized to haemozoin-containing vesicles in PbAtg18-KD parasites ( Figure 10C), which also showed slightly lower Atg18 level in the absence of TMP than the parasites grown in the presence of TMP ( Figure   10D), indicating a partial knock-down effect. The effect of knock-down on development of PbAtg18-KD parasites was assessed in BALB/c and C57/BL/6J mice. Without TMP, PbAtg18-KD parasites showed drastically reduced growth and were eventually cleared, whereas wild type and PbAtg18-KD (with TMP) parasites grew similarly and the infected mice had to be euthanized or succumbed to high parasitemia ( Figure 10E and F), indicating indispensability of PbAtg18 for parasite development.

Discussion
Plasmodium Atg8 is present as punctate structures throughout the parasite, which also co-localize with the food vacuole and apicoplast [19,26,45]. Altered distribution of Atg8 puncta and mutation in PfAtg18 have been observed in chloroquine and artemisinin resistant P. falciparum strains, respectively.
We investigated the Plasmodium Atg18 to understand the mechanism of its localization and association with drug resistance. Our study reveals that Plasmodium Atg18 proteins localize to the food vacuole, which is mediated by interaction with PI3P and MDR1 on the food vacuole membrane. Chloroquine and amodiaquine altered PfAtg18 localization in a significant number of cells.
PfAtg18 and its orthologs in P. vivax, P. knowlesi and P. berghei localized to the food vacuole, indicating that food vacuole localization is likely a common feature of Plasmodium Atg18 proteins.
Interaction of PfAtg18 with the food vacuole-associated PI3P was critical for food vacuole localization, as PfAtg18 mutants of PI3P-binding sites (PIPm and ALCAm) had complete cytosolic localization. Based on the PfAtg18 homology model, "WLCL" motif forms part of the 23 rd β strand in blade 6, which contains the 2 nd PI3P-binding site and the residues contributing to the PI3P-binding site are highly conserved in PROPPINs [34,39], suggesting that "WLCL" motif is a critical structural requirement for the formation of 2 nd PI3P-binding site in PfAtg18. The failure of ALCAm to bind PI3P and localize to the food vacuole could be due to disruption of the 2 nd PI3P-binding site. Hence, the "WLCL" motif of PfAtg18 and the corresponding sequence in other PROPPINs represent a new sequence requirement for Atg18 functions.
The PI3P-mediated localization of PfAtg18 in our study is consistent with previous reports of FRRGmediated interaction of PROPPINs with PI3P, and the essentiality of this interaction for recruitment to the membrane structures [19, 34-39, 45, 69].
The Atg18 localization was restricted to food vacuole in our study regardless of the host used for expression, whereas previous studies have reported Atg18 localization as punctate structures all over the parasite, including colocalization with the apicoplast and some signal close or around the food vacuole [19,26,45]. This discrepancy in PfAtg18 localization between our and the previous studies may be due to differences in the procedures used for localization: live-cell imaging in our study and IFA of fixed parasites in the previous reports. The punctate Atg18 localization close or around the food vacuole could be due to the loss of cellular morphology during fixation and the subsequent immunofluorescence assay procedures, whereas better maintenance of cellular morphology and short duration of live-cell microscopy would have preserved the complete food vacuole localization in our study. In fact, PfAtg18 localization in our study is nearly identical to that in the live cell microscopy images of the previous study [19]. Nonetheless, multiple lines of data support food vacuole localization of PfAtg18 in our study. 1) P. falciparum has been shown to have abundant PI3P in erythrocytic stages and localization studies using the PI3P probe GFP-2xFYVE indicated that PI3P is abundantly associated with the food vacuole membrane [46], which is consistent with PI3P-dependent localization of PfAtg18 to the food vacuole.
2) The localization was rendered all over the parasite upon treatment of parasites with the PI3K inhibitor LY294002, which might be due to decreased production of PI3P, hence, less PI3P on the food vacuole membrane. 3) PIPm and ALCAm neither bound PI3P nor localized to the food vacuole. 4) In P. berghei that contains multiple food vacuoles, the episomally expressed GFP/PfAtg18 and endogenously expressed PbAtg18/GFP showed identical localization with the food vacuoles, which ruled out any artefact of episomal expression and GFP fusion. 5) GFP fusions of ScAtg18 and WIPI have been extensively used in autophagy-related studies without any report of adverse effect on localization and functions of these proteins [34,42,69]. A previous study of the purified P.
falciparum food vacuole proteome also contained PfAtg18 together with other food vacuole proteases and transporters, which further corroborates our result [79].
ScAtg18 has been shown to localize to endosomes, vacuole and autophagic membranes, and perivacuole membrane compartments in a PI3P-dependent manner [38,69,80,81]. However, unlike The altered PfAtg18 localization upon chloroquine and amodiaquine treatment is of particular significance because quinolines, particularly chloroquine, has been proposed to block heme polymerization to hemozoin in the food vacuole and alter Atg8 distribution in chloroquine sensitive strains as compared to that in resistant strains of P. falciparum [27,82,83]. In mammalian cells, quinolines have been shown to inhibit autophagy by alkalinizing the lysosomes, thereby inhibiting the activities of lysosomal hydrolases that are required for degradation of the autophagy cargo [84,85]. However, the concentration required for alkalinization is 1-2 orders of magnitude higher than the concentrations used in our study [83]. Also, treatment of parasites with other quinolines did not have any effect. Furthermore, E64 and chloroquine did not affect the food vacuole localization of PfCRTmCherry, a food vacuole membrane transport protein, indicating that food vacuole integrity was not compromised in our experiments. Taken together, the diffuse localization of PfAtg18 in parasites treated with chloroquine and amodiaquine is unlikely due to alkalinization or disruption of the food vacuole and these two drugs likely target Atg18 transport to the food vacuole, which may contribute to the overall antimalarial effect of these compounds. PfAtg18 has also been shown to participate in the food vacuole dynamics and it is transported via the haemoglobin trafficking pathway [19]. Quinolines have also been shown to have different effects on haemoglobin endocytosis and vesicle trafficking; in particular, inhibition of the transport of haemoglobin-containing vesicles to the food vacuole by chloroquine and amodiaquine [86]. It is possible that inhibition of the transport of haemoglobincontaining vesicles by amodiaquine and chloroquine resulted in the altered food vacuole localization of PfAtg18. However, the canonical vesicle trafficking inhibitors brefeldin A and bafilomycin did not affect PfAtg18 localization. This suggests for a different mode of inhibition of PfAtg18 localization by amodiaquine and chloroquine that might affect food vacuole-associated processes like haemoglobin degradation, which could subsequently affect the action of these drugs, as this degradation produces heme for quinoline action and iron for artemisinin activation. Further investigation is necessary to find out how chloroquine and amodiaquine affect the PfAtg18 localization.
The Plasmodium Atg18 gene is essential for parasite development [87,88], and knock-down of PfAtg18 has been shown to cause the loss of apicoplast [45]. We generated PbAtg18-KD parasites, which did not show any noticeable change in the localization of PbAtg18, most likely because there was only a partial reduction in the PbAtg18 protein level. A more efficient knock-down approach will be required to define the role of Plasmodium Atg18 in autophagy and the functional importance of its association with the food vacuole.
Atg18 proteins of S. cerevisiae and P. pastoris are also required for selective degradation of peroxisomes by autophagy, which is known as pexophagy. The proteins are concentrated at one or more spots that resemble protuberances of the vacuole membrane [89]. Elegant microscopic images indicated colocalization of the protuberance with the peroxisome cluster [89,90]. PfAtg18 also showed an intense spot on or near the food vacuole membrane, which tempts us to speculate if PfAtg18 functions in the engulfment of cellular contents directly by the food vacuole membrane. Although peroxisomes appear to be absent in Plasmodium [91], direct engulfment of cellular contents by the Plasmodium food vacuole membrane may occur. One such process could be the fusion of haemoglobin-containing vesicles with the food vacuole [92,93]. Similar PfAtg18-enriched dot-like structure in the food vacuole periphery was also observed in a recent study, which the authors explained as a result of the food vacuole fission [19]. ScAtg18 has also been shown to be present as an intense spot in a peri-vacuole membrane compartment during starvation wherein it colocalizes with a number of autophagy proteins, and this compartment has been proposed to be a site for vesicle formation or membrane source [94]. The intense PfAtg18 spot in the peri-food vacuole region may be a platform where autophagy and some other pathways overlap. Further studies are required to gain insights into the nature and function of this peri-food vacuole compartment.
Thus, our data demonstrate that Plasmodium Atg18 proteins localize to the food vacuole, which is mediated by interaction with PI3P and MDR1. Additional study is needed to investigate the physiological significance of MDR1-PfAtg18 interaction.