Purification, crystallization and characterization of the Pseudomonas outer membrane protein FapF, a functional amyloid transporter

FapF from Pseudomonas facilitates the secretion of the amyloid-forming polypeptide FapC across the bacterial outer membrane. A protocol for forming reproducible well diffracting crystals is described in which the switch to tetraethylene glycol monooctyl ether detergent during purification proved to be critical and represents a handy tip for β-barrel protein crystallography.


Introduction
Amyloids, fibrillar proteinaceous aggregates with a cross--strand structure, are commonly associated with human disease states such as Alzheimer's and Parkinson's. However, certain bacteria have evolved to exploit such fibres to their advantage by secreting monomeric subunits in a highly controlled manner and allowing their self-assembly as part of a biofilm matrix. This matrix assists them in attachment to their host (animal or surface), persistence in infection and antibiotic resistance. Understanding the molecular basis for controlled amyloid formation is of great interest, not only for new strategies to counter bacterial infection, but also to inspire the design of therapeutic approaches to pathogenic amyloidogenesis in humans. The model systems that have been best characterized to date are curli from Escherichia coli and ISSN 2053-230X functional amyloid protein from Pseudomonas (Fap). These two systems form morphologically similar fibres despite being genetically distinct (Dueholm et al., 2010).
The curli system has been extensively studied, with structures available for all but one of the encoded proteins [CsgA (Tian et al., 2015), CsgC (Taylor et al., 2011), CsgE (Shu et al., 2016) and CsgG (Cao et al., 2014;Goyal et al., 2013Goyal et al., , 2014]. In contrast, no structural information exists for the more recently discovered Fap system. Both of these systems require the polypeptide amyloid subunit to be transported across the complex bacterial outer membrane (OM). In curli, the membrane component CsgG was found to be a nonameric -barrel in which each monomer contributes four -strands to a central 36-residue OM channel. The integral membrane component from the fap operon, FapF, is similarly predicted to be a -barrel membrane protein; however, bioinformatics analyses suggest that its structure is likely to be more akin to the fatty-acid transporter -barrels [e.g. FadL (van den Berg et al., 2004) and TodX (Hearn et al., 2008)]. This suggests that the two membrane components are distinct, raising the intriguing prospect that the mechanism of amyloid secretion is completely independent. Structural predictions using PSIPRED (Bryson et al., 2005) indicate that FapF comprises an $30-residue helical region at the N-terminus, followed by an $60-residue disordered region and then an integral membrane -barrel component.
We set out to solve the structure of the FapF outer membrane protein. As our initial efforts to crystallize fulllength FapF proved challenging, N-terminally truncated constructs were designed based on molecular-weight information from limited proteolysis (Fig. 1). These were expressed and purified from the outer membrane fraction of E. coli and used for crystallization trials. A complete detergent exchange into tetraethylene glycol monooctyl ether (C8E4) during chromatographic purification was essential for forming well diffracting crystals and suggests that this could be a useful step worth considering for membrane -barrel proteins that remain challenging to work with . Reproducible crystals that diffract to 2.4 Å resolution were grown for a construct comprising residues 106-430 of FapF, which represents a major step towards providing the first structural insight into the Fap machinery.

Crystallization
Conditions for crystallization of FapF 106-430 were initially screened by the sitting-drop vapour-diffusion method at 20 C using sparse-matrix crystallization kits from Hampton Research and Molecular Dimensions in MRC 96-well optimization plates (Molecular Dimensions), with 100 nl protein solution and 100 nl reservoir solution, using a Mosquito nanolitre high-throughput robot (TTP Labtech). Protein crystals were obtained in a reproducible manner from 100 mM sodium citrate, 20-30%(w/v) PEG 400, 100 mM NaCl. These were manually optimized, screening over sodium citrate in the pH range 5.5-6.5 in one dimension and an NaCl concentration gradient of 50-100 mM in the second dimension using MRC 96-well plates with up to 400 nl protein solution and 400 nl reservoir solution. See Table 2 for crystallization information.

Data collection and processing
Crystals were mounted in a MicroLoop (MiTeGen) and immediately flash-cooled in liquid nitrogen. Diffraction data from a single native crystal were collected on beamline I03 of the Diamond Light Source (DLS), England. The data were processed with XDS (Kabsch, 2010) and scaled using SCALA (Evans, 2006) within the xia2 package (Winter et al., 2013). Data-collection statistics are shown in Table 3. The content of the unit cell was analyzed using the Matthews coefficient (Matthews, 1968). Molecular-replacement (MR) attempts were carried out using complete structures and polyalanine models with and without loop truncations of available -barrel  Table 1 Macromolecule-production information.  crystal structures in the Protein Data Bank (http:// www.rscb.org) as well as using idealized models of 8-16stranded -barrels using C traces of various tilt angles generated using in-house scripts. MR attempts were performed using MOLREP (Vagin & Teplyakov, 2010) and Phaser (McCoy et al., 2007). High-resolution data were used between 2.5 and 6 Å .

Results and discussion
To date, no structural information is available for any of the components of the fap operon. Refolding attempts using fulllength and N-terminally truncated FapF yielded microcrystals that could not be further optimized for crystallography despite exhaustive attempts. We therefore isolated and purified the FapF 106-430 construct from the physiologically relevant environment of the outer membrane (Fig. 1) and subsequently obtained suitably sized crystals (Fig. 2).
After the screening of various detergents for both extraction efficiency and appearance in gel-filtration profiles, we found that the crucial step in obtaining reproducible crystals was to reduce the concentration of LDAO in buffers to zero during nickel purification and gradually replace it with C8E4. Mixtures of LDAO and C8E4 have proven to be successful in several recent membrane-protein crystallography studies (see,
for example, Goyal et al., 2014;van den Berg et al., 2015). Under these purification conditions crystals grew readily to up to 200 mm 3 over the course of 5-7 d.
Diffraction data were collected to 2.36 Å resolution (Fig. 3) and indexed in space group C121. Data were finally scaled at 2.5 Å resolution. Analysis of the crystal content indicated that there are three to five molecules in the asymmetric unit, with a Matthews coefficient in the range 3.5-2.1 Å 3 Da À1 and a corresponding solvent content in the range 65-41%; however, self-rotation analysis indicated threefold symmetry. This, combined with the fact that membrane-protein crystals tend to have a higher solvent content than their soluble counterparts owing to the presence of detergent species, leads us to believe that it is almost certainly a trimer that is present with 65% solvent content. SFCHECK (Vaguine et al., 1999) suggests that twinning is not present (hE 4 i/hI 2 i = 2.35). Data-collection and processing statistics are listed in Table 3.
Our work involving the extensive optimization of extraction and purification protocols demonstrated that a reduction in LDAO during the early nickel-affinity purification steps was essential to obtain well diffracting crystals. It appears that gel filtration into C8E4 alone is not sufficient to remove the comparatively 'sticky' detergent LDAO (Newstead et al., 2008). Presumably, the zwitterionic nature of LDAO is more disruptive to protein packing interactions within crystals than the milder, uncharged C8E4 (le Maire et al., 2000). LDAO and C8E4 make up the majority of detergents used for outer membrane protein crystal structures. We suggest that extraction by LDAO, followed by buffer exchange into C8E4 as the sole detergent for downstream crystallization, is a worthwhile combination together with limited proteolysis when considering -barrel membrane-protein preparations for crystal trials.