The molecular mechanism of cytoadherence to placental or tumor cells through VAR2CSA from Plasmodium falciparum

Dear Editor, Pregnancy-associated malaria (PAM) threatened more than one million women and their infants in endemic regions in 2019. This resulted in maternal anemia, stillbirth, and infant death. VAR2CSA encoded by a subfamily of var genes from Plasmodium falciparum (P. falciparum) named as var2csa, plays a vital role in the cytoadherence of infected erythrocytes to the placenta. Chondroitin sulfate A (CSA), which is displayed mostly on the surface of placental or tumor cells, has been recognized as a specific ligand for VAR2CSA. However, the molecular mechanism of cytoadherence to placental or tumor cells through VAR2CSA remains elusive. In this study, the VAR2CSA ectodomain from P. falciparum strain 3D7 (~306 kDa, containing six Duffybinding-like (DBL) domains, N-terminal sequence (NTS), and multiple inter-domains (IDs)) (Fig. 1a) was recombinantly expressed and purified using the Sf9 insect cell secretory system. The cryo-EM structures of VAR2CSA ectodomain and its complex with CSA were determined at a resolution of 3.6 Å and 3.4 Å, respectively (Fig. 1b, c; Supplementary Figs. S1–S5 and Table S1). In line with the previously proposed model, our structures showed that the core region is well-defined and covers NTS, DBL1X, DBL2X, ID2a, ID2b, DBL3X, ID3, and DBL4ε (Fig. 1b, c; Supplementary Fig. S6a). However, DBL5ε and DBL6ε formed a flexible wing region. Compared with the apo structure, the densities of the wing region in the complex were significantly improved, making it feasible for the flexible fitting of both DBL5ε and DBL6ε (Supplementary Fig. S4d). As for the core region, DBL2X and DBL4ε stacked closely with ID2a, ID2b, and ID3, and formed the most stable core center, which served as a base for anchoring DBL3X and DBL1X at the top or bottom sites, respectively (Fig. 1b, c). Intriguingly, a highly basic pocket formed by DBL2X, DBL1X, NTS, and DBL4ɛ has been identified to accommodate a dodecasaccharide with six sulfated disaccharide repeats from CSA, which fitted well into the density map. Moreover, 16 residues were identified to be responsible for the direct interaction with 10/12 monosaccharides, except for the 6th and 11th units of CSA dodecasaccharide. Among them, there are nine, four, and three residues derived from DBL2X, DBL4ɛ, and NTS, respectively. Furthermore, the nine residues of DBL2X were shown to interact directly with 7/ 10 monosaccharide units and thus may account for the major contribution to CSA binding (Fig. 1g; Supplementary Figs. S4e, f, S6b, c). Owing to its high resolution and potential high stability in the core center, DBL4ɛ was used as an immobilized reference for the structural alignment between the VAR2CSA-CSA and VAR2CSA ectodomain (Fig. 1d–f). Interestingly, a significant conformational change in the core region was observed (Supplementary Movie S1). Firstly, the DBL1X moves closer to DBL4ɛ at more than 3.2 Å (ranging from 3.2 Å to 4.4 Å) for multi-helixes to facilitate NTS interaction with CSA and close the pocket (Fig. 1f). Secondly, there was a 1.8 Å outward bend for the

, and R846E). The above proteins were expressed in Rosetta-gamiB (DE3) of E. coli. (WEIDI Biotechnology). Briefly, when the recombinant bacteria grow to the OD value around 0.6 at 37℃, the temperature was lowered to 16℃ for about half an hour. IPTG was then added to the culture at a final concentration of 0.2 mM. Cells were collected after 20-hour culture. Cell pellets were resuspended in wash buffer and lysed using Emulsiflex homogeniser (YongLian). The proteins were purified by sequential chromatography using Ni-NTA affinity, HiTrap S, gel filtration columns (GE Healthcare).
CSA extracted from bovine trachea (Sigma, C9819) was used as the ligand for VAR2CSA.
Various VAR2CSA proteins were incubated with CSA solution (10 mg/mL) overnight on ice and further separated by Superose6 increase 10/300 GL. The complex fractions were collected, pooled and concentrated to 3 mg/mL for the following experiments.
The samples were centrifuged at 38,000 rpm for 16 hours at 4°C using a Beckman SW41 Ti rotor.
Subsequently, 200 µL per fraction were harvested and the cross-linking reaction was terminated by adding quench buffer (1 M Tris pH 6.5, 250 mM NaCl) to a final concentration of 40 mM Tris.
Fractions containing cross-linked VAR2CSA monomer were pooled, concentrated, and dialyzed in buffer B. The cross-linked VAR2CSA monomer about 0.4 mg/mL was applied to the preparation of cryo-EM grids.
For negative staining EM analysis, carbon-coated copper grids were processed using a PELCO easiGlow (TED PELLA) cleaning system with a power of 30 W and a plasma current of 30 mA in the air for 30 s. Subsequently, samples (8 µL at a concentration of ~0.02 mg/mL) were applied onto the grids above and stained twice using 2% (w/v) uranyl acetate solution at room temperature for the following examination via 200kV TEM (TF20, FEI).
As for the preparation of cryo-EM grids, Amorphous Alloy Film Au300 R1.2/1.3 grids (CryoMatrix) were processed in the H2/O2 mixture for 30 s using a Gatan 950 Solarus plasma cleaning system with a power of 5 W. Then cross-linked VAR2CSA ectodomain (3 μL at a concentration of ~0.4 mg/mL) was applied to the grids for instant incubation under a relative humidity of 100% at 4 °C. Next, the grids were blotted for 2 s with a blot force of 1 in a Vitrobot Mark IV (Thermo Fisher) and plunge-frozen in liquid ethane. Similarly, VAR2CSA-CSA (3 μL at a concentration of ~0.5 mg/mL) was blotted for 3 s with a blot force of -2 and plunge-frozen in liquid ethane.

Cryo-EM data collection and Image processing
The cryo-EM grids of cross-linked VAR2CSA ectodomain and VAR2CSA-CSA were firstly evaluated using a 200kV Talos Arctica microscope (Thermo Fisher). Cryo-EM datasets were collected on 300 kV Titan Krios microscope (Thermo Fisher) equipped with a Gatan K2 Summit direct electron detector and a 20-eV slit GIF Quantum energy filter (Gatan). The cryo-EM images were automatically recorded in the super-resolution counting mode using Serial-EM 2 software with a nominal magnification of 130,000 x (a super-resolution pixel size of 0.522 Å), and with a defocus ranging from -1.2 to -2.2 μm. Each micrograph stack was dose-fractionated into 36 frames with a total electron dose of ~ 57.6 e -/ Å 2 and a total exposure time of 7.2 s. Drift and beam-induced motion of the super-resolution movie stacks were corrected using MotionCor2 3 and binned twofold to a calibrated pixel size of 1.044 Å/pix, with both the dose weighted and non-dose weighted micrographs saved at the same time. The defocus values were estimated by Gctf 4 using the nondose weighted micrographs. Other procedures of cryo-EM data processing were performed using RELION v3.0 5 .
A total of 1,965 movies were recorded for the cross-linked VAR2CSA ectodomain. Among them, 1,800 micrographs were selected for further processing due to appropriate range of rlnDefocusU (5000-30000) and rlnCtfMaxResolution (2)(3)(4)(5)(6). Selected 2D class averages from 200kV cryo-EM data were low-pass filtered to 20 Å and used as references for auto-picking. Then an initial set of 984,517 particles were extracted for 2D classification. After several rounds of 2D classification, the relatively good classes with 960,039 particles were selected for three-dimensional reconstruction. Initially, one major class (~62% particles) was separated after 3D classification.
To further improve the resolution, a second cycle of 3D classification with mask was carried out with 304,160 particles selected for 3D auto-refinement and postprocessing (the B-factor automatically estimated). The final resolution was evaluated according the gold-standard Fourier shell correlation (threshold = 0.143) 6 . Although the overall resolution is up to 3.6 Å, the density of the wing region (DBL5ε and DBL6ε) is quite weak due to high flexibility. The local refinement strategy for focused classification and refinement was carried out for the whole structure divided into two half parts. Consequently, the resolution of the core region could be improved to 3.1 Å by merging two local-refined maps. However, the map quality of the wing region remained unimproved, which made it hardly interpreted by de novo model building. The local resolution was evaluated by ResMap 7 .
For VAR2CSA-CSA, 3,058 out of 3,356 micrographs were selected for further processing using similar procedures as above. Briefly, 1,163,625 auto-picked particles were extracted for 2D classification. After several rounds of 2D and 3D classification, 241,954 particles were selected for 3D auto-refinement and postprocessing. The final map was reconstructed at a resolution of 3.4 Å. Compared with apo structure, the density of the wing region has been significantly improved, which makes it feasible for the flexible fitting of both DBL5ε and DBL6ε. The resolution of the core region could be improved to 3.1 Å by merging local-refined maps.

Model building and refinement
The density map of VAR2CSA ectodomain was firstly interpreted using Phenix.map_to_model 8 to generate the initial model with the most helixes, which has been further improved by manual model building using COOT 9 . The structure of VAR2CSA-CSA was determined using the apo structure as a reference and manual model building in COOT with various refinement strategies.
The structures have been further validated by validation module in Phenix (Table S1). All structural figures were prepared using UCSF Chimera 10 , UCSF ChimeraX 11 or PyMOL 12 .

Interaction evaluated using Octet RED 96
CSA were biotinylated and loaded onto SA sensor (Pall corporation). VAR2CSA fragments and mutant were then added for real-time association and dissociation analysis using Octet RED 96 (Fortebio) at room temperature. Data Analysis Octet was used for data processing.

Confocal fluorescence microscopy
Cells were pre-seeded on a slide, washed with PBS and fixed using 4% PFA/PBS on ice. After fixation, cells were blocked in 2.5% FBS for 30 min on ice and incubated with recombinant VAR2CSA proteins diluted in PBS with 0.25% FBS for 1 hour on ice. After washing with PBS by three times, the specimens were incubated with anti-His-FITC antibody (1:500) for 1 hour at room temperature in the dark and washed as described before. DAPI was applied to stain and locate the cell nuclei. Slides were further analyzed using a confocal microscope. Negative controls (Mock) were prepared with the same procedures except the incubation of recombinant VAR2CSA fragments.

Data availability
The atomic coordinates and electron microscopy data have been deposited in the RCSB Protein      18 Movie S1.
The conformational changes of the core region of VAR2CSA ectodomain induced by recognizing and binding to CSA.