ATG Ubiquitination Is Required for Circumsporozoite Protein to Subvert Host Innate Immunity Against Rodent Malaria Liver Stage

Although exo-erythrocytic forms (EEFs) of liver stage malaria parasite in the parasitophorous vacuole (PV) are encountered with robust host innate immunity, EEFs can still survive and successfully complete the infection of hepatocytes, and the underlying mechanism is largely unknown. Here, we showed that sporozoite circumsporozoite protein (CSP) translocated from the parasitophorous vacuole into the hepatocyte cytoplasm significantly mediated the resistance to the killing of EEFs by interferon-gamma (IFN-γ). Attenuation of IFN-γ-mediated killing of EEFs by CSP was dependent on its ability to reduce the levels of autophagy-related genes (ATGs) in hepatocytes. The ATGs downregulation occurred through its enhanced ubiquitination mediated by E3 ligase NEDD4, an enzyme that was upregulated by CSP when it translocated from the cytoplasm into the nucleus of hepatocytes via its nuclear localization signal (NLS) domain. Thus, we have revealed an unrecognized role of CSP in subverting host innate immunity and shed new light for a prophylaxis strategy against liver-stage infection.


INTRODUCTION
Malaria is still one of the most devastating diseases worldwide and is caused by infection with the genus Plasmodium that is initiated in the mammalian host by inoculation of sporozoites into the skin by Anopheles mosquitoes. From the skin, sporozoites travel to the liver via blood circulation, where they infect hepatocytes.
In the liver, sporozoites bind to highly sulfated heparin sulfate proteoglycans (HSPGs) on the hepatocyte surface by circumsporozoite protein (CSP), which triggers their invasion into hepatocytes (1). After invading hepatocytes, sporozoites inside a parasitophorous vacuole (PV) and transform into exo-erythrocytic forms (EEFs). To avoid fusion of the PV with lysosomes and to shield itself from the cytosol of the host, the PV membrane is modified by parasite-derived proteins, such as upregulated in infective sporozoite gene 4 (UIS4) and exported protein 1 (EXP-1) (2, 3). However, EEFs in the PV can be sensed by the Melanoma differentiation-associated gene 5 (MD-5) and elicit hepatocytes to generate a type-I IFN response (4). This recruits innate immune cells, such as natural killer (NK) and NK T cells, and eliminates the EEFs in the infected hepatocytes through the secretion of IFN-g (5). The IFN-g-mediated killing of EEFs was previously suggested to involve the production of nitric oxide (NO) (6,7); it was recently reported to be regulated in an NOindependent manner (8). In addition, the infection of sporozoites could also induce Plasmodium-associated autophagy-related responses (PAAR) of hepatocytes to limit EEF development (9,10). However, EEFs can still survive and successfully complete the infection of hepatocytes, and the underlying mechanism of parasite survival in this hostile environment remains largely unknown.
A previous study has shown that CSP, the major surface protein of sporozoites, translocating from the PV into the cytoplasm of hepatocytes, could promote liver stage development, which was thought to be associated with the reduced NF-kB activation in hepatocytes by CSP (11). However, in the present study we found that CSP enhanced the liver stage development through its resistance to the killing effect of IFN-g on EEFs. Therefore, we uncovered a novel immune escape strategy for EEFs. Moreover, further study showed that the resistance mediated by CSP to the IFN-g-killing of EEFs was dependent on its ability to downregulate autophagy-related genes (ATGs) through enhanced ubiquitination mediated by the E3 ligase NEDD4.

Parasites and Mice
Plasmodium berghei ANKA (P.b ANKA) was maintained in our laboratory, and the P.b ANKA CSP mut parasite was constructed by mutation of pexel I-II of CSP using CRISPR-Cas9. All parasites were maintained by passage between Kunming mice (Swiss Webster strain) and Anopheles stephensi. ATG5 fl/fl mice (B6.129S-Atg5<tm1Myok>) were obtained from the RIKEN BioResource Center (Ibaraki, Japan), and Alb-Cre mice (B6.Cg-Speer6-ps1Tg(Alb-cre)21Mgn/J) were purchased from Jackson Laboratory (Bar Harbor, ME, USA). ATG5 fl/fl mice were crossed with Alb-Cre mice to conditionally delete the ATG5 cassette in hepatocytes. ATG5 fl/fl mice were used as controls. IFN-gR1 knockout mice were gifts from Dr. Bo Guo (Army Medical University, Chongqing, China). Kunming mice were purchased from the Experimental Animal Center of the Army Medical University (Chongqing, China). Animals were kept in a specific pathogen-free laboratory at the Institute of Immunology of Army Medical University. All methods were carried out in accordance with the approved Guide for the Care and Use of Laboratory Animals of the Army Medical University (approval number AMUWEC20171828).

Mosquito Rearing and Infection
Anopheles stephensi (Hor strain) was maintained at 27°C and 70%-80% relative humidity and fed with a 5% sugar solution. For infection with the parasite, 3-to 5-day-old female adults were kept at 20-21°C and 70%-80% relative humidity and fed on P.b ANKA, P.b ANKA-RFP, or P.b ANKA-CSP mut -infected Kunming mice with gametocytemia ≥ 0.5%. Seven days after infection, the mosquitoes were dissected, and the oocysts on the midguts were examined under a light microscope.

Infection of P. berghei ANKA Sporozoites
At 19 days after infection with P.b ANKA, P.b ANKA-RFP, or P.b ANKA-CSP mut , infected female mosquitoes were extensively washed in sterile PBS, and the salivary glands were dissected and collected in RPMI 1640 medium containing 2.5 mg/mL amphotericin B (Sangon Biotech, Shanghai, China), 100 units/ mL penicillin, and 100 mg/mL streptomycin (Beyotime, Beijing, China). The sporozoites were released from salivary glands and counted, and then incubated with HepG2 cells at a ratio of 1:3. Three hours after infection, the supernatant was discarded and replaced with fresh culture medium with the three antibiotics as described above. For the in vivo assay, ATG5 fl/fl and ATG5 fl/fl -Alb-Cre were challenged with an intravenous injection of 1,000 sporozoites. Control and IFN-g R1 knockout mice were challenged with an intravenous injection of 10,000 sporozoites. seamless cloning according to the instructions of the In-fusion ® HD Cloning Kit (Clontech, Palo Alto, CA, USA). NLS-HA was inserted into pcDNA3.1 using the High-Efficient Ligation Reagent Ligation High (TOYOBO). The pcDNA3.1 vector was linearized using the restriction enzymes BamHI and XholI (TAKARA, Otsu, Japan) for ATG5 and LC3B, and with KpnI and NotI for ATG7. The pFLAG-cmv8 vector was linearized using the restriction enzymes HindIII and EcoRI for ATG5 and ATG7. Human LC3B, ATG5, and ATG7 were inserted into the pcDNA3.1 or pFLAG-cmv8 vector, respectively, by seamless cloning according to the instructions of the In-fusion ® HD Cloning Kit (Clontech). The pGL3-NEDD4 promoter reporter plasmid was constructed by cloning the NEDD4 promoter sequence (2000bp upstream form 1st exon) into the pGL3 vector through KpnI/HindIII. All recombinant plasmids were verified by DNA sequencing.

Construction of CSP-Stably Transfected Cell Lines
The CSP coding sequence of P.b ANKA was cloned downstream of the CMV-7 promoter in the Lenti-pCDH plasmid (a gift from Professor Ying Wan, Army Medical University, China). HEK293T cells were transfected with Lenti-pCDH-CSP or Lenti-pCDH control plasmids with psPAX2 and pMD2.G packaging plasmids at a ratio of 20:15:5 using Lipofectamine 3000 (Invitrogen) to obtain the lentivirus. Seventy-two hours after transfection, the supernatant was filtered through a 0.45mm filter (Millipore), and the lentivirus was collected by ultracentrifugation (20,000 rpm) at 4°C for 2 h. HepG2 cells were infected with Lenti-pCDH-CSP or Lenti-pCDH control lentivirus at an MOI of 15 for 48 h. Then, the positively infected cells were screened using 5 mg/mL puromycin (Gibco). CSPstably transfected and control cells were cloned by a limiteddilution method in a 96-well plate, and 3 mg/mL puromycin was used to maintain the resistance of stably transfected cells. CSP expression was confirmed using immunofluorescence.
The peptide was conjugated to bovine serum albumin (BSA). The resulting conjugated peptide was emulsified with Freund's adjuvant (Sigma Aldrich) and immunized in a specific pathogenfree-grade rabbit at 0, 2, 4, and 6 weeks. Seven days after the final immunization, rabbit serum was collected, and the titer of the antibody against the epitope was detected using an enzymelinked immunosorbent assay (cat. no. 44-2404-21, Nunc MaxiSorp flat bottom, Nalge Nunc International, Penfield, NY, USA). Protein G Sepharose (GE Healthcare) was used to purify the CSP antibody.

Immunofluorescence Assay
HepG2 cells (1 × 10 5 ) were placed on a 14-mm-diameter slide in a 24-well plate. To detect the distribution of CSP, HepG2 cells were infected with CSP wt or CSP mut sporozoites. Three hours after infection, the cells were washed with PBS to remove noninvaded sporozoites and incubated with fresh DMEM (Hyclone) containing 10% FBS (Gibco) and three antibiotics. Twenty-four hours later, the slides were immobilized with 4% paraformaldehyde (Sangon Biotech) for 10 min, permeabilized with 0.1% Triton-X 100 (Amersco, Albany, NY, USA) for 15 min, and then blocked with PBS containing 5% BSA and 0.02% Tween-20 for 1 h at room temperature. The cells were labeled with 1:500 goat anti-UIS4 (Sicgen, Cantanhede, Portugal) overnight at 4°C and 1:100 green donkey anti-goat secondary antibody (Abbkine, Wuhan, China) for 1 h at room temperature. After washing with PBS three times, the cells were labeled with 1:500 rabbit anti-CSP overnight at 4°C and 1:100 red donkey anti-rabbit secondary antibody (Abbkine) for 1 h at room temperature. Nuclei were counterstained with DAPI (Beyotime) for 5 min at room temperature.
For invasion capacity comparison, UIS4 staining was used to identify the CSP wt and CSP mut parasites in HepG2 cells, and images were obtained at 6 h after infection. To evaluate the colocalization of EEFs and LC3, control and CSP-stably transfected HepG2 cells were transfected with LC3B-RFP plasmid and infected with CSP wt or CSP mut sporozoites, and then labeled with goat anti-UIS4 and the secondary antibody as described above. To observe the effect of CSP on the IFN-g-mediated killing of EEFs, control and CSP-stably transfected HepG2 cells were pre-treated with 1 U/mL recombinant human IFN-g (Peprotech, Rocky Hill, NJ, USA) or equivoluminal medium for 6 h. The cells were then infected with CSP wt or CSP mut as stated above and co-treated with IFN-g for 46 h.
For autophagy flux observation, the control and CSP-stably transfected HepG2 cells were pre-treated with 8 mg/mL polybrene for 2 h, and then infected with adenovirus containing tandem GFP-RFP-LC3 (Hanbio, Shanghai, China). Since the cell culture was changed, the cells were treated with 0.25 mg/mL rapamycin (Cell Signaling Technology) or equivoluminal medium for 24 h.
For nuclear translocation of CSP observation, HepG2 cells were transfected with pcDNA3.1-CSP-HA and pcDNA3.1-CSPDNLS-HA plasmids for 24 h. The cells were then labeled with rabbit anti-HA and the secondary antibody, as stated above.
The cells were immobilized, penetrated, and stained with DAPI. After washing with PBS, the slides were mounted on a microscope slide using DAKO Fluorescence Mounting Medium (Agilent). LSM 780 NLO microscope systems and ZEN Imaging Software (Zeiss) were used for image acquisition and export. Image J software (NIH) was used to analyze the number and size of the EEFs.

Total RNA Extraction and Real-Time PCR
For parasite burden detection, the livers of mice were dissected at 46 h after infection with CSP wt or CSP mut sporozoites, and homogenized in 1.5 mL Trizol (Invitrogen), and HepG2 cells were collected 46 h after infection with CSP wt or CSP mut sporozoites and lysed with 1 mL Trizol (Invitrogen). Liver homogenate (100 mL) or 1 mL of cell lysis were used for total RNA extraction using TRIzol in accordance with the manufacturer's instructions. cDNA was synthesized from equivalent total RNA using the PrimeScript ™ RT reagent Kit with gDNA Eraser (TAKARA) in accordance with the manufacturer's instructions. The parasite load in the livers and HepG2 cells was evaluated by Taqman PCR with primers and probes for 18S rRNA and GAPDH following the manufacturer's instructions of Premix Ex Taq ™ (Probe qPCR) (TAKARA). The mRNA levels of IFN-g, iNOS, LC3B, ATG3, ATG5, ATG7, STUB1, SYVN1, CBL, SMURF1, PAFAH1B1, and NEDD4 were evaluated using TB Green ™ Premix Ex Taq ™ II (Tli RNaseH Plus) (TAKARA). CFX96 Touch ™ Real-Time PCR Detection System and CFX ManagerTM Software (Bio-Rad, Hercules, CA, USA) were used for RT-PCR data collection and analysis. The annealing temperature used for the above two RT-qPCR experiments was 60°C.

Western Blot and Immunoprecipitation
To investigate the influence of CSP overexpression on the protein levels of ATGs, 1 × 10 6 control and CSP-stably transfected HepG2 cells were incubated in 6-well plates and treated with 0.25 mg/mL rapamycin for 0, 6, 12, and 24 h. To observe the influence of CSP overexpression on the half-life of ATG proteins, CSP-stably transfected HepG2 and control cells were treated with 20 mM cycloheximide (CHX, Cell Signaling Technology) or 10 mM MG132 (Cell Signaling Technology) for 0, 6, 12, and 24 h. To test whether the degeneration of ATGs was dependent on the ubiquitin-proteasome system, CSP-stably transfected HepG2 and control cells were treated with IFN-g, 1mM TAK-243 or both for 24 h. Uniformly, the cells were lysed with 300 mL RIPA lysis buffer containing a protease and phosphatase inhibitor cocktail (Thermo Fisher) at various time points and protein concentrations were detected using the BCA Protein Assay Kit (Sangon Biotech). Total proteins (20 mg equivalent) were then separated by 10% or 12% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE; Bio-Rad), transferred using a 0.22-mm polyvinylidene fluoride (Amersco) and incubated with 1:20000 goat anti-rabbit or antimouse IgG-HRP secondary antibody (Santa Cruz) for 1 h at room temperature.
T h e u b i q u i t i n a t i o n o f A T G s w a s d e t e c t e d b y immunoprecipitation, as previously described (12). To detect the influence of CSP overexpression on the ubiquitination of ATGs, 2 × 10 6 CSP-stably transfected HepG2 and control cells in two 6-wells were transfected with 2mg His-UB, and 6 mg pFLAG-cmv8 or pFLAG-cmv8-human ATG5/7 plasmids, respectively. To confirm CSP-induced ubiquitination of ATGs via upregulation of NEDD4, 2 × 106 293FT cells in two 6-wells were co-transfected with 2 mg His-UB, 4 mg pFLAG-cmv8 or pFLAG-cmv8-human ATG5/7, 2 mg shNC or shNEDD4, and 2 mg pcDNA3.1 or pcDNA3.1-CSP-HA plasmids, respectively. To test the influence of the NLS domain on the CSP-induced ubiquitination of ATGs, 2 × 106 293FT cells in two 6-wells were co-transfected with 2mg His-UB plasmids, 4 mg pFLAG-cmv8-human ATG5/7, 4 mg pcDNA3.1, pcDNA3.1-CSP-HA, and pcDNA3.1-CSPDNLS-HA, or pcDNA3.1-NLS-HA, respectively. Briefly, the cells were treated with 10 mM MG-132 for 24 h and lysed with 1 mL of western and IP cell lysis buffer. The protein concentrations of each group were determined and equalized as described above. Fusion FLAG-hATG5 and hATG7 proteins were captured using anti-FLAG ® M2 Magnetic Beads (Sigma-Aldrich) according to the manufacturer's instructions. After washing five times with equilibrium buffer (50 mM Tris HCl, 150 mM NaCl, pH 7.4), the beads were boiled with SDS-PAGE loading buffer for 10 min. An equal volume of total cell lysates (input) was loaded and separated by 10% SDS-PAGE, as described above. The PVDF filter was incubated with 1:500 mouse anti-FLAG tag (Invitrogen), 1:1000 rabbit monoclonal anti-His-Tag (Cell Signaling Technology), 1:2000 mouse anti-bactin (Sigma-Aldrich), or 1:500 rabbit anti-NEDD4 overnight at 4°C and incubated with secondary antibody (Santa Cruz) for 1 h at room temperature. The protein bands were visualized using the Western BLoT Hyper HRP Substrate (TAKARA) and exposed using a chemiluminescence imaging system (Fusion Solo S, Vilber, France). Image J software (NIH) was used for grey value analysis.

Transcriptome Sequencing
CSP-stably transfected or control cells (1 × 10 6 ) were lysed with 1 mL Trizol (Invitrogen), and total RNA was extracted as described above. A total of 2 mg RNA per sample was used as the input material for the RNA sample preparations. mRNA was purified from total RNA using poly T oligo-attached magnetic beads. Fragmentation was carried out using divalent cations under elevated temperatures in the VAHTSTM First Strand Synthesis Reaction Buffer (5X). First-strand cDNA was synthesized using a random hexamer primer and M-MuLV reverse transcriptase (RNase H). Second-strand cDNA synthesis was subsequently performed using DNA polymerase I and RNase H. Remaining overhangs were converted into blunt ends via exonuclease/polymerase activities. After adenylation of the 3' ends of the DNA fragments, an adaptor was ligated for library preparation. To select cDNA fragments of preferentially 150-200 bp in length, the library fragments were purified with the AMPure XP system (Beckman Coulter, Beverly, USA). Then, 3 mL USER Enzyme (NEB, Ipswich, MA, UK) was incubated with size-selected, adaptor-ligated cDNA at 37°C for 15 min followed by 5 min at 95°C before PCR. PCR was performed with Phusion High-Fidelity DNA polymerase, Universal PCR primers, and Index (X) Primer. Finally, PCR products were purified (AMPure XP system) and library quality was assessed using the Agilent Bioanalyzer 2100 system. Libraries were quantified and pooled. Paired-end sequencing of the library was performed using HiSeq XTen sequencers (Illumina, San Diego, CA). RAW data were submitted to the GEO database, and the series records were GSE129323. A heatmap was generated using R language version 4.0.2, and the pheatmap package v1.0.12.

Dual Luciferase Reporter Assay
Promoter activity assay was conducted by co-transfecting 100 ng pcDNA3.1 vector, pcDNA3.  Figures  S1A-C). As no mutations were found in any of the potential offtarget regions screened by genome-wide comparison ( Supplementary Table S1), the potential of genomic off-target mutations generated by the CRISPR-Cas9 strategy could be excluded. The growth of the blood-stage mutant parasite was normal as compared to that of the WT parasite (CSP wt ), and the salivary gland sporozoites of CSP mut could be successfully generated after the mutant parasite-infected mosquitoes (Supplementary Figure S1D). Meanwhile, no difference of invasion ability into hepatocytes was found between CSP mut and CSP wt parasites (4.3 ± 0.2% vs 4.0 ± 0.1%) at 6 h post infection in vitro. Immunofluorescence assay with anti-UIS4, a parasitophorous vacuole membrane (PVM) specific marker and anti-CSP demonstrated that the CSP of mutant sporozoites was significantly inhibited to translocate from the PV into the hepatocyte cytoplasm when infecting HepG2 cells ( Figure 1A). Consistent with a previous study (11), we also found that the burden of CSP mut in the liver was greatly reduced compared to that of the CSP wt after the mice were infected ( Figure 1B). However, transient transfection with full-length CSP plasmid did not enhance the development of EEFs in vitro, even when up to 800 ng of plasmids were transfected ( Figure 1C). There was also no difference in parasite load in HepG2 cells infected with CSP wt or CSP mut parasites in vitro ( Figure 1D). As sporozoite infection could trigger mice to generate immune responses against parasites (4), the different results between our in vivo and in vitro assays might be interpreted by the in vivo immune response that was not considered in our in vitro assay. Thus, we postulated that CSP might indirectly promote EEF development by suppressing host immune responses.
Plasmodium in the liver can trigger a type I interferon (IFN) response and recruit CD1d-restricted NKT cells (5). IFN-g, predominately secreted by recruited NKT cells, is considered to be not only the major innate immune effector to suppress the development of the liver stage in vivo (5,13), but also the critical effector to inhibit the EEFs in the PV of hepatocytes in vitro (14,15). Altogether, these findings strongly suggest that IFN-g should be included in our investigation of the effects of CSP on EEF development both in vitro and in vivo. Strikingly, we found that overexpression of CSP greatly inhibited the IFN-g-mediated killing of EEFs, as both EEF number and parasite load in CSPstably transfected HepG2 cells were much higher than those in control cells, when treated with IFN-g ( Figure 1E and Supplementary Figure S2). In addition, although the mutant parasite in HepG2 cells was susceptible to IFN-g-mediated killing, overexpression of CSP remarkably reversed the killing effect of IFN-g ( Figure 1F). Subsequently, this finding was further confirmed in a physiologically relevant context, as no substantial difference in the liver parasite burden was found between the CSP wt and CSP mut -infected mice when IFN-g was depleted ( Figure 1G). Similar results were obtained when two parasite strains were infected with IFN-gR1 knockout mice ( Figure 1H). The reduced parasite burden in CSP mut infected mouse livers was not due to the capacity of mutant parasites to induce the host to produce a much higher level of IFN-g because a comparable mRNA level of IFN-g was found in the livers of mice infected with CSP wt or CSP mut sporozoites at different doses (Supplementary Figure S3). Thus, our data demonstrates that (without lentivirus infection), control (infected with Lenti-pCDH control lentivirus and selected by puromycin) and CSP-stably transfected HepG2 cells were pre-treated with or without 1 U/mL IFN-g, and then incubated with 4 × 10 4 CSP wt sporozoites for 46 h. EEF number per 100 HepG2 cells was determined and statistically analyzed. Each dot represents one visual field (n=66, magnification, 630×). (F) 1.2 × 10 5 control and CSP-stably transfected HepG2 cells were treated with 1 U/mL IFN-g, and then infected with CSP wt or CSP mut parasite as above; 46 h later, the parasite burden in hepatocytes was determined and compared as described in (C).
Three individual experiments have been performed. (G) Mice pre-treated with anti-IFN-g or isotype antibody, were infected with 1,000 CSP wt or CSP mut sporozoites. Then, the liver burden of parasite at 46h post infection was determined as described in (B). Each dot represents one mouse (n=15). (H) IFN-gR1 knockout and WT mice were infected with 1,000 CSP wt or CSP mut sporozoites, and liver burden was determined at 46 h after infection as described as above. Each dot represents one mouse (n=15). The pooled data of three repeated experiments was presented, data are represented as mean ± SD and analyzed by a student's t-test or One-Way ANOVA. ns, not significant; *P < 0.05; ***P < 0.001.

Nitric Oxide (NO) Is Not Involved in the Suppression of the IFN-g-Killing of EEFs Mediated by CSP
Previous studies have shown that IFN-g prevents Plasmodium liver-stage development by inducing the expression of inducible nitric oxide synthase (iNOS), an enzyme required for the production of NO (6,16,17). Hence, we sought to investigate whether inhibition of the IFN-g-mediated killing of EEFs by CSP was dependent on the downregulation of iNOS and NO. We found that transfection with CSP did not reduce the mRNA level of iNOS or the concentration of NO in the infected-HepG2 cells treated with IFN-g (Supplementary Figure S4A). Similarly, both the levels of iNOS and NO were comparable between CSP wt and CSP mut parasite-infected HepG2 cells treated with IFN-g (Supplementary Figure S4B). In addition, the resistance to the IFN-g-mediated suppression of EEFs and the production of NO in CSP-transfected HepG2 cells treated with IFN-g were not affected by treatment with two iNOS inhibitors, aminoguanidine (AG) and L-NAME. Although they could greatly inhibit the production of NO in lipopolysaccharide (LPS)-stimulated macrophages (Supplementary Figure S4C). Therefore, NO is not involved in the suppression of IFN-g-mediated killing of EEFs by CSP.

Autophagy and Its Related Proteins Are Essential for the IFN-g-Mediated Killing of EEFs
Evidence has shown that autophagy and autophagy-related genes (ATGs) are essential for the IFN-g-mediated killing of Toxoplasma gondii in macrophages (18)(19)(20)(21), an apicomplexan parasite closely related to Plasmodium. Thus, we next sought to investigate whether autophagy was also involved in the regulation of IFN-g-mediated killing of EEFs in HepG2 cells. Consistent with our previous study (22), autophagy alone, either enhanced by rapamycin or inhibited by LY294002, had no effect on liver stage development in HepG2 cells (Figures 2A, B). However, rapamycin-enhanced autophagy significantly reduced the EEF number, parasite load, and EEF size in HepG2 cells in the presence of 0.2 U/mL IFN-g (Figure 2A and Supplementary Figure S5A). We found that rapamycin could not further reduce the parasite number and load, as well as the size, when treated with 0.5 U/mL or 1 U/mL IFN-g (Figure 2A and Supplementary Figure S5A), possibly because most of the EEFs were already killed by IFN-g at these higher concentrations (14), and enhanced autophagy could not further increase the effect of IFN-g. Meanwhile, the autophagy inhibitor LY294002 considerably reverted the killing effect of 0.5 U/mL and 1 U/ mL IFN-g on EEFs ( Figure 2B and Supplementary Figure S5B), and a similar result was also obtained using another autophagy inhibitor, wortmannin ( Figure 2C). No toxicity effect on cells was found for any of the above three drugs and IFN-g at the working concentrations (Supplementary Figure S5C), excluding the possible effect of these drugs on EEF development in HepG2 cells. Furthermore, knockdown of either ATG5 (key component of the ATG12-ATG5 conjugate) or ATG7(E1-like enzyme), by specific small hairpin RNA (shRNA) could significantly inhibit the IFN-g-mediated suppression of EEF development ( Figures 2D, E).
In addition, the depletion of IFN-g greatly increased the liver parasite load in control ATG5 fl/fl mice but had no effect on parasite burden in ATG5 fl/fl -Alb-Cre mice, in which ATG5 in hepatocytes was specifically knocked out ( Figure 2F), indicating the essential role of ATG5 in mediating the IFN-g-killing of EEFs. However, we also noted that the level of liver burden in ATG5 fl/fl -Alb-Cre mice was even lower than that in control ATG5 fl/fl mice in the presence of IFN-g ( Figure 2F). The reduced liver burden in ATG5 fl/fl -Alb-Cre mice may be caused by completely abolishing of canonical autophagy after ATG5 knockout. As canonical autophagy could promote parasite development by supplying essential nutrients (23), and the lack of canonical autophagy greatly limited the replication of EEF in hepatocyte, even in the absence of IFN-g. Thus, our findings show that both autophagy and ATGs are pivotal for the killing and suppression of EEF development in hepatocytes by IFN-g, which is consistent with the report that an ATGs-mediated LAPlike process was also involved in the IFN-g-mediated killing of the malaria liver stage (8).

CSP Inhibits the IFN-g-Mediated Suppression of EEFs by the Downregulation of ATGs
We then investigated whether CSP could inhibit autophagy and the expression of ATGs. Immunofluorescence assay showed that rapamycin-induced autophagy was significantly inhibited in CSP-stably transfected HepG2 cells as either the fluorescence intensity of the key autophagy marker, microtubule-associated protein 1 light chain 3 (LC3)-RFP or LC3-GFP puncta, induced by rapamycin, was reduced by more than 3 folds in CSP-stably transfected HepG2 cells as compared to that in control cells ( Figure 3A). In addition, the intensity of LC3 fluorescence around the CSP mut EEFs was much stronger than that of CSP wt EEFs in HepG2 cells, and the level of LC3 fluorescence intensity around CSP mut EEFs was comparable to that around CSP wt EEFs in CSP-overexpressing HepG2 cells ( Figure 3B). Next, the effect of CSP on the expression of ATGs was investigated. LC3 levels were significantly reduced in CSPstably transfected HepG2 cells treated with rapamycin at all test time points when compared with control group, and the levels of autophagy-related genes, including Beclin-1, ATG3, ATG5, and ATG7, were also significantly decreased at 24 h. In contrast, the autophagy adaptor protein SQSTM1 (p62) significantly accumulated in CSP-stably transfected HepG2 cells, indicating the suppression of autophagy ( Figure 3C and Supplementary Figure S6). Overall, these findings demonstrate that CSP could inhibit autophagy and the expression of ATGs.
As CSP could significantly suppress autophagy and its related proteins, we next investigated whether the resistance of CSP to IFN-g-mediated suppression of EEFs was closely associated with its ability to suppress ATGs expression. Although the CSP-stably transfected HepG2 cells had a significant resistance to the IFN-gmediated suppression of EEFs, the overexpression of either ATG5 or ATG7 significantly reduced this inhibitory effect ( Figure 3D). Similarly, the CSP wt and CSP mut parasite burdens in ATG5-knockdown HepG2 cells were comparable when the cells were treated with IFN-g ( Figure 3E). To further confirm this finding, both control (ATG5 fl/fl ) and ATG5 fl/fl -Alb-Cre mice were infected with either CSP wt or CSP mut sporozoites, and as expected, there was no difference in the liver burdens in ATG5 fl/fl -Alb-Cre mice between the two parasites ( Figure 3F). Taken together, these findings demonstrate the essential role of ATGs in the resistance to IFN-g-mediated suppression by CSP. (E) HepG2 cells were transiently transfected with Scramble or ATG5 shRNA plasmids, and then treated with 1 U/mL IFN-g and incubated with CSP wt or CSP mut sporozoites as above. 46 h later, the parasite load in the HepaG2 cells was determined and compared. The experiment has been performed for three times. (F) Both control (ATG5 fl/fl ) and ATG5 fl/fl -Alb-Cre mice were infected by intravenous injection with 1,000 CSP wt or CSP mut sporozoites, 46 h later, the liver burden of the two parasites in control and ATG5 fl/fl -Alb-Cre mice were determined and compared. Each dot represents one mouse, n=15. The pooled data of three repeated experiments was presented, data are represented as mean ± SD, and analyzed by a student's t-test, Mann-Whitney U test or One-Way ANOVA; ns, not significant; **P < 0.01; ***P < 0.001. The CSP-Mediated Resistance of IFN-g-

Killing of EEFs Is Dependent on Its Ability to Enhance ATGs Ubiquitination
To explore the underlying mechanism by which CSP downregulates the expression of ATGs, mRNAs from CSPstably transfected and control HepG2 cells treated with rapamycin were sequenced and compared. Unexpectedly, there was no significant downregulation in the mRNA levels of most ATGs in CSP-stably transfected HepG2 cells as compared to that of the control ( Figure 4A), which was further confirmed by realtime PCR ( Figure 4B). As CSP downregulated the protein levels of ATGs, this finding suggests that CSP does not regulate the expression of ATGs at the transcriptional level. Eukaryotic cells use the autophagy-lysosome and ubiquitinproteasome system as their major intracellular protein degradation pathways (24). However, autophagy level was significantly reduced in the CSP-stably transfected HepG2 cells ( Figure 3A), indicating that the autophagy-lysosome system may not be responsible for the downregulation of the ATGs protein levels. Therefore, we next investigated whether CSP could downregulate the expression of ATGs at the post-translational level through the ubiquitin-proteasome pathway, which is a major intracellular degradation pathway and a widespread post-translational modification (25). The half-lives of LC3, ATG3, ATG5, and ATG7 in the CSP-stably transfected HepG2 cells were significantly reduced compared to those of the control, when cells were treated with the protein synthesis inhibitor cycloheximide (CHX) ( Figure 4C). In addition, the levels of LC3, ATG3, ATG5, and ATG7 in CSP-stably transfected and control HepG2 cells were comparable when the cells were treated with the specific proteasome inhibitor MG132 ( Figure 4D and Supplementary Figure S7). These findings indicate that CSP could regulate the degradation of ATGs in a proteasomedependent manner. To investigate whether the degradation of ATGs in the proteasome was dependent on ubiquitination, the CSP-stably transfected and control cells were transfected with His-UB, FLAG-hATG5 (human ATG5), or FLAG-hATG7 (human ATG7) plasmids, and co-immunoprecipitation was performed using anti-FLAG beads to detect ubiquitin bound to ATGs. The ubiquitination assay showed that CSP promoted the ubiquitination of both ATG5 and ATG7 in HepG2 cells ( Figure 4E).
To verify that the resistance of CSP to IFN-g-mediated suppression of EEF development was due to the enhancement of ATGs ubiquitination, a ubiquitination-specific inhibitor TAK-243 (26), was used at non-toxic concentrations in vitro and in vivo (Supplementary Figure S8) to suppress ubiquitination. We found that treatment with TAK-243 significantly elevated the protein levels of ATGs ( Figure 4F) and enhanced the IFN-g-mediated killing of EEFs in CSP-stably transfected HepG2 cells, as compared to those in control cells ( Figure 4G). Furthermore, the administration of TAK-243 greatly reduced the liver parasite burden in WT mice but had no significant effect on liver parasite burden in IFN-g R1 knockout mice ( Figure 4H). Thus, we demonstrate that CSP mediated resistance to IFN-g-killing of EEFs through the upregulation of specific ubiquitination of ATGs.
CSP Upregulates the E3 Ubiquitin Ligase NEDD4 to Promote the Ubiquitination of ATGs E3 ubiquitin ligase, which transfers the ubiquitin of the E2 enzyme to its attached substrate, is the key enzyme in the ubiquitination process (25). Therefore, we hypothesized that E3 ubiquitin ligases may be involved in the ubiquitination of ATGs (27). Six candidate E3 ligases, including STIP1 homology and U box-containing protein 1 (STUB1), Synovial apoptosis inhibitor 1 (SYVN1), Cbl proto-oncogene (CBL), SMAD specific E3 ubiquitin protein ligase 1 (SMURF1), Platelet activating factor acetylhydrolase 1b regulatory subunit 1 (PAFAH1B1), and neural precursor cell expressed, developmentally downregulated 4 (NEDD4), were predicted to regulate the ubiquitination of at least two ATGs using an online prediction tool (Supplementary Figure S9 and Table S2) (27). However, the RNA-seq data showed that among these ligases, only NEDD4 was significantly upregulated in CSP-stably transfected HepG2 cells ( Figure 5A), and the increase in both the mRNA and protein levels of NEDD4 was confirmed by real-time PCR and western blotting, respectively ( Figures 5B, C). Furthermore, infection with the CSP wt parasite, but not the CSP mut parasite, significantly elevated the mRNA level of NEDD4 in HepG2 cells ( Figure 5D), strongly suggesting that the translocation of CSP from the PV into the cytoplasm could upregulate the expression of E3 ubiquitin ligase NEDD4. Consistently, the upregulation of NEDD4 in sporozoite-infected hepatocyte cell lines was also found in the P. b ANKA infection model (data from GSE78931 and GSE72049) (28) ( Figure 5E). Moreover, knockdown of NEDD4 by shRNA significantly reduced CSP-mediated ubiquitination of ATG5 and ATG7 ( Figure 5F). In addition, the resistance of CSP-overexpression to IFN-g-mediated killing of CSP wt parasites was significantly reduced after NEDD4 was silenced ( Figure 5G). This finding strongly suggests that NEDD4 was essential for CSP-mediated resistance to IFN-g suppression of EEFs.
A previous study showed that CSP contained a nuclear location signal (NLS) domain, which could import the CSP into the nucleus (11), and we also confirmed that the transfection of CSP, but not CSPDNLS plasmid (CSP without NLS domain), could import into the nucleus of HepG2 cells (Supplementary Figure S10). To better elucidate the mechanism of the upregulation of NEDD4 by CSP, the effects of CSP, CSPDNLS, and NLS (only NLS domain) expression on the transcription of NEDD4 in HepG2 cells were determined. Although the transfection of CSP upregulated NEDD4 expression, neither the transfection of CSPDNLS nor NLS influenced the transcription of NEDD4 in HepG2 cells ( Figure 5H). The dual luciferase reporter assay confirmed that only CSP significantly activated the NEDD4 promoter ( Figure 5I). These data indicate that translocation into the nucleus was essential for CSP to modulate the transcription of NEDD4, but the modulatory motif might not exist in the sequence of CSPDNLS. In addition, co-immunoprecipitation results showed that more NEDD4 proteins could directly bind to both ATG5 and ATG7 and increased the ubiquitination of (E) Ubiquitination assay showed that CSP promoted the ubiquitination of ATG5 (left) and ATG7 (right). Control and CSP-stably transfected HepG2 cells were transfected with His-UB, and FLAG-hATG5 or FLAG-hATG7 plasmids, and treated with MG-132 for 24 h. Then, cells were collected for immunoprecipitation of FLAG-hATG5 and FLAG-hATG7 followed by immunoblot analyses of ubiquitin (His-UB). Cells transfected with FLAG vector was used as negative controls. Total cell lysates (input) were also immunoblotted with antibodies against FLAG and b-actin. The experiment has been repeated for three times. (F) The levels of protein LC3, ATG5 and ATG7 were determined by western blotting in both control and CSP-stably transfected HepG2 treated with or without 1mM TAK (TAK-243) and 1 U/mL IFN-g for 24 h. (G) 1.2 × 10 5 CSP-stably transfected HepG2 or control cells were treated with or without 1mM TAK (TAK-243), and then were treated with 1 U/mL IFN-g and infected with 4 × 10 4 CSP wt sporozoites for 46 h. The EEF number (left, each dot represents one visual field, n=57-73, magnification, 630×) and parasite load (right) were determined and compared as above. Three individual experiments have been performed (H) IFN-g R1 knockout and WT mice were per-treated with or without 20mg/kg TAK-243 and infected with 1,000 CSP wt sporozoites in the next day, then liver parasite burden was determined at 46 h after infection as described as above. Each dot represents one mouse (n=15). The pooled data of three repeated experiments was presented, data are represented as mean ± SD and analyzed by a student's t-test or One-way ANOVA; ns, not significant; *P < 0.05; **P < 0.01; ***P < 0.001. Three individual experiments have been performed. (D) After HepG2 cells were infected with CSP wt or CSP mut sporozoites as described as before for 24 h, the mRNA level of NEDD4 was determined by real-time PCR, uninfected HepG2 cells were used as control. The experiment has been repeated for three times. (E) Relative level of NEDD4 mRNA in P.b ANKA sporozoite infected Huh7 and HepG2 cells were analyzed based on data extracted from GSE78931 and GSE72049. (F) Ubiquitination assay showed that shNEDD4 deceased CSP-induced hATG5 (left) and hATG7 (right) ubiquitination in 293FT cells. 293FT cells were transfected with His-UB, pcDNA3.1 vector or pcDNA3.1-CSP, shNC or shNEDD4, and FLAG vector, FLAG-hATG5 or FLAG-hATG7 plasmids, and treated with MG-132 for 24 h. Then, cells were collected for immunoprecipitation of FLAG-hATG5 and FLAG-hATG7 followed by immunoblot analyses of ubiquitin (His-UB). Cells transfected with FLAG vector was used as negative controls. Total cell lysates (input) were also immunoblotted with antibodies against FLAG and b-actin. Three individual experiments have been performed. (G) 1.2 × 10 5 control and CSP-stably transfected HepG2 cells were transfected with shNC or shNEDD4 and treated with 1 U/mL IFN-g or not, then infected with 4 × 10 4 CSP wt sporozoites, for 46 h. The parasite load in HepG2 cells was determined as above. The experiment has been performed for three times. (H) HepG2 cells were transfected with vector (Vec), CSP, CSPDNLS, or NLS plasmids for 24 h, and the transcription of NEDD4 was determined by real-time PCR. Three individual experiments have been performed. (I) HEK293T cells were transfected with RL-TK, NEDD4 promoter reporter plasmid, and with or without Vec, CSP, CSPDNLS, or NLS plasmids for 24 h, and the ratio of firefly luciferase to Renilla luciferase were determined using Dual Luciferase Assay kit. Three individual experiments have been performed. Med, medium. (J) The ubiquitination assay showed that deletion of NLS domain (DNLS) deceased CSP-induced hATG5 (left) and hATG7 (right) ubiquitination in 293FT cells. 293FT cells were transfected His-UB, and Vec, CSP, CSPDNLS or NLS, and FLAG-hATG5 or FLAG-hATG7 plasmids, and treated with MG-132 for 24 h. Then, cells were collected for immunoprecipitation of FLAG-hATG5 and FLAG-hATG7 followed by immunoblot analyses of ubiquitin (His-UB). Cells transfected with pcDNA3.1 vector was used as negative controls. Total cell lysates (input) were also immunoblotted with antibodies against FLAG and b-actin. The experiment has been repeated for three times. (K) 1.2 × 10 5 HepG2 cells were transfected with Vec, CSP, CSPDNLS, or NLS plasmids and treated with or without 1 U/mL IFN-g, followed by the infection with 4 × 10 4 CSP wt sporozoites for 46 h. The parasite load in HepG2 cells was determined as above. Three individual experiments have been performed. The pooled data of three repeated experiments was presented (except E), data are represented as mean ± SD and analyzed by Mann-Whitney U test or One-way ANOVA. UB, ubiquitin; ns, not significant; *P < 0.05; **P < 0.01, ***P < 0.001. these two proteins in CSP, but not in the vector, CSPDNLS or NLS-transfected 293FT cells ( Figure 5J), and only the transfection of CSP remarkably inhibited the IFN-g-mediated suppression of EEFs in HepG2 cells ( Figure 5K). Overall, these data suggest that NEDD4 was an E3 ubiquitin ligase that specifically mediated the ubiquitination of ATG5 and ATG7, and that CSP enhanced the ubiquitination of ATGs through the upregulation of NEDD4.

DISCUSSION
Considerable progress has been made in identifying the molecules of sporozoites involved in their selective attachment to hepatocytes and the formation of the PV after their invasion (1,(29)(30)(31); however, the underlying mechanism of the survival of parasites in the PV has thus far remained unknown. Here, we found that nuclear translocation of cytoplasmic CSP could mediate the resistance to IFN-g-killing of EEFs and facilitate the survival of parasites in the PV. CSP is a multifunctional protein of the malaria parasite that is not only critical for the invasion of sporozoites into hepatocytes but is also essential for sporozoite development in mosquitoes (32)(33)(34)(35)(36)(37)(38). Therefore, CSP is regarded as a leading candidate protective antigen (39), which is included in several subunit vaccines to induce CSPspecific antibody and/or CD8 + T cell responses against sporozoites (40)(41)(42)(43)(44)(45)(46). Our finding of this new aspect of CSP to subvert the host innate immune response not only supports the important role of CSP as the dominant protective antigen but also sheds new light on a novel prophylactic strategy against the liver stage.
Recent studies have shown that parasites can shed host autophagic proteins from the PVM to escape the deleterious effects of the PAAR response (47) and UIS3 sequestered LC3 to suppress autophagy in EEFs (48). These findings support the existence of subversive strategies for EEFs against hepatocyte autonomous immunity. In this study, we found that CSP translocated from the PV into the cytoplasm of hepatocytes inhibited the IFN-g-mediated killing of EEFs by downregulating the expression of ATGs by ubiquitination, revealing a novel strategy for the liver stage of malaria to subvert host innate immunity. However, a previous study suggested that liver-stage development enhanced by CSP was associated with its ability to suppress NF-kB activation in infected hepatocytes. Although the real reason is definitively unknown for the different mechanisms of CSP to promote EEF development between our study and previous studies, the possible interpretation might be the missed consideration of IFN-g in the natural malaria infection in a previous study.
It is well known that IFN-g-mediated killing of intracellular T. gondii is regulated by IFN-g-inducible GTPase Irg6 and guanylate-binding protein (GBP)1 (49)(50)(51), whereas Irg6 is not involved in the IFN-g-mediated killing of EEFs (49). Indeed, we found that the knockout of GBP-1 also had no effect on EEF development in HepG2 cells treated with IFN-g (Supplementary Figure S11). Thus, the roles of GTPase and GBPs in the IFN-gmediated killing of EEFs might differ from those in T. gondii and should be investigated in further detail.
Recent studies have demonstrated that infection with sporozoites could induce both canonical autophagy and the PAAR response of hepatocytes to EEFs (9,22). Although canonical autophagy promotes parasite development by supplying essential nutrients (23), the role of PAAR response in liver stage development is conflicting (9,52). One possible reason is that conditional ATG knockout of non-permissive mouse embryonic fibroblasts (MEFs) has been used in previous studies (9,52). In the present study, we found that knockdown of either ATG5 or ATG7 in HepG2 cells, the permissive cells of sporozoites, had no significant effect on EEF development ( Figures 2D, E). In contrast, the induction of autophagy by rapamycin enhanced the killing effect of IFN-g (Figure 2A), and autophagy inhibitors or knockdown of ATGs attenuated the killing of EEFs by IFN-g ( Figures 2B-E). This demonstrates the essential role of autophagy in the regulation of the IFN-g-mediated killing of EEFs. Therefore, we showed that autophagy alone has no significant effect on EEF development in vitro, but it plays a pivotal role in the regulation of IFN-g-mediated killing of EEFs.
Taken together, we found that CSP translocated from the PV into the cytoplasm of hepatocytes could be imported into the nucleus and upregulated the transcription of E3 ligase NEDD4, depending on its NLS domain. The E3 ligase NEDD4 enhances the ubiquitination of ATGs, leading to a decrease in the protein level of ATGs in hepatocytes. As ATGs are essential for IFN-gmediated killing of EEFs, CSP resists the IFN-g-mediated killing of EEFs and facilitates the survival of EEFs in hepatocytes ( Figure 6). Thus, our findings explain why CSP translocated from the PV into the hepatocyte cytoplasm could promote liver stage development, and provide novel prophylaxis strategies to eliminate liver stage infection by chemically targeting CSP pexel, NLS domain, or ubiquitination in hepatocytes. FIGURE 6 | A mechanistic pathway of CSP-mediated resistance to IFN-gkilling of EEFs in hepatocytes. ①After sporozoite invaded into hepatocytes, CSP on its surface would be translocated from PV into the cytoplasm of hepatocytes and is then introduced into nucleus dependent on its NLS domain. As a result, CSP activates the promoter of ligase3 NEDD4 and enhances its transcription. The upregulated NEDD4 promotes the ubiquitination of ATGs, leading to the degradation of ATGs. As ATGs are essential for IFN-g-mediated killing of EEFs, the CSP-mediated resistance to IFN-g, which is dependent on its ability to downregulate ATGs through enhancing NEDD4-mediated ubiquitination, facilitates the survival of EEFs. ②Otherwise, EEFs are destroyed by IFN-g.