The ESCRT-III machinery participates in the production of extracellular vesicles and protein export during Plasmodium falciparum infection

Infection with Plasmodium falciparum enhances extracellular vesicle (EV) production in parasitized red blood cells (pRBCs), an important mechanism for parasite-to-parasite communication during the asexual intraerythrocytic life cycle. The endosomal sorting complex required for transport (ESCRT), and in particular the ESCRT-III sub-complex, participates in the formation of EVs in higher eukaryotes. However, RBCs have lost the majority of their organelles through the maturation process, including an important reduction in their vesicular network. Therefore, the mechanism of EV production in P. falciparum-infected RBCs remains to be elucidated. Here we demonstrate that P. falciparum possesses a functional ESCRT-III machinery activated by an alternative recruitment pathway involving the action of PfBro1 and PfVps32/PfVps60 proteins. Additionally, multivesicular body formation and membrane shedding, both reported mechanisms of EV production, were reconstituted in the membrane model of giant unilamellar vesicles using the purified recombinant proteins. Moreover, the presence of PfVps32, PfVps60 and PfBro1 in EVs purified from a pRBC culture was confirmed by super-resolution microscopy and dot blot assays. Finally, disruption of the PfVps60 gene led to a reduction in the number of the produced EVs in the KO strain and affected the distribution of other ESCRT-III components. Overall, our results increase the knowledge on the underlying molecular mechanisms during malaria pathogenesis and demonstrate that ESCRT-III P. falciparum proteins participate in EV production.


Supplemental Materials and Methods
Unless otherwise indicated, reagents were purchased from Sigma-Aldrich (St. Louis, MO, USA).

In silico analysis of PfBro1
The Bro1 protein domain was searched in the P. falciparum genome database (http//www.plasmodb.org). Comparison between putative PfBro1 and human and yeast orthologs was determined using the Expert Protein Analysis Systems (ExPASy) Proteomics Server by the NCBI BLAST service program. Sequence alignments were generated using the Clustal Omega program [1] and edited in Jalview [2].

Cloning and expression of PfVps32, PfVps60, PfBro1 and PfBro1t
The proteins PfVps32, PfVps60 and PfBro1 (PlasmoDB accession numbers PF3D7_1243500, PF3D7_1441800 and PF3D7_1224200, respectively) were expressed from a codon-optimized synthetic gene (Genscript, Leiden, The Netherlands). PfVps32 and PfVps60 were inserted into pGEX-6P-1 (GE Healthcare, Freiburg, Germany) as a fusion protein with a Glutathione-S-transferase (GST) tag linked to the N terminus of the proteins via a preScission protease site. PfBro1 was inserted into pET20b(+) as a fusion protein with histidine tag (6×, C-terminal). To generate the truncated PfBro1t, the first 397 residues from PfBro1 were PCRamplified using plasmid cDNA as template and the specific primers

Subcellular protein extraction
For the analysis of PfVps32, PfVps60 and PfBro1 throughout the whole

Simulations
All simulations were executed on the 'hot' computer cluster of the Max Planck Institute of Colloids and Interfaces. The Rosetta software suite version 3.6 was used for structure refining and protein docking simulations. Visualizations of protein structures were produced using pyMOL with custom Python scripts. For all data analyses and curvature calculations custom Python scripts were written.

Structure refining
Rough structure predictions of PfBro1 and PfVps32 were acquired from the Phyre2 server [7] (http://www.sbg.bio.ic.ac.uk/phyre2/html/page.cgi?id=index) and subjected to structure refining to prepare both proteins for docking and adapt them to the Rosetta force field. A total of 11,652 PfBro1 and 27,520 PfVps32 structures were generated with a standard FastRelax protocol. Given the rigidity and size of the PfBro1 protein (819 residues), a further refinement in a backrub simulation was performed where 200 structures were generated, of which the best 20 structures from the relax ensemble were chosen based on their total score.

Protein docking
Potential orientations for protein docking of PfBro1 and PfVps32 were previously acquired using the ClusPro web server (https://cluspro.org) [8], omitting the need for global docking simulations. As input we used two PfVps32 structures in 'open' state and a good scoring PfBro1 structure. For docking simulations these structures were aligned with a prominent prediction from ClusPro using pyMOL. To remove any steric hindrances that could be present due to our manual construction of the proteinprotein system, a quick restrained relaxation was performed with 96 structures each, the best of which were chosen for the following docking simulations. After performing a prepacking protocol (also with the 96 best scoring structures), local docking simulations were performed generating 6,144 structures for each system. Out of these, the 1,000 lowest scoring structures were filtered using a low-pass filter on an interface score of minus 5 REU. The ca. 20 structures per system that remained after filtering were then further refined by performing additional backrubbing simulations, once again generating 200 structures for each filtered structure.