Structure and Mechanism of the Saposin-like Domain of a Plant Aspartic Protease*

Many plant aspartic proteases contain an additional sequence of ∼100 amino acids termed the plant-specific insert, which is involved in host defense and vacuolar targeting. Similar to all saposin-like proteins, the plant-specific insert functions via protein-membrane interactions; however, the structural basis for such interactions has not been studied, and the nature of plant-specific insert-mediated membrane disruption has not been characterized. In the present study, the crystal structure of the saposin-like domain of potato aspartic protease was resolved at a resolution of 1.9 Å, revealing an open V-shaped configuration similar to the open structure of human saposin C. Notably, vesicle disruption activity followed Michaelis-Menten-like kinetics, a finding not previously reported for saposin-like proteins including plant-specific inserts. Circular dichroism data suggested that secondary structure was pH-dependent in a fashion similar to influenza A hemagglutinin fusion peptide. Membrane effects characterized by atomic force microscopy and light scattering indicated bilayer solubilization as well as fusogenic activity. Taken together, the present study is the first report to elucidate the membrane interaction mechanism of plant saposin-like domains whereby pH-dependent membrane interactions resulted in bilayer fusogenic activity that probably arose from a viral type pH-dependent helix-kink-helix motif at the plant-specific insert N terminus.

Aspartic proteases (APs) 2 are characterized by a common bilobal tertiary structure containing two catalytic aspartic acid residues (Asp 32 and Asp 215 in pepsin) within an active site cleft (1,2). They are found in all higher organisms, and their respective roles are well established, although structural and functional characteristics of APs in plants are least understood. Of practical interest among plant APs are their roles in plant pathogen resistance (3) as well as in senescence and postharvest physiology (4,5). Plant APs share the common AP bilobal structure; however, some contain an additional sequence of ϳ100 residues inserted within the C-terminal primary structure. These additional amino acids unique to plant APs (6 -8) create an extra domain protruding from the canonical AP molecule (9 -11). This structural oddity among APs is called the plantspecific insert (PSI), also known as the plant-specific sequence, which belongs to the saposin-like protein (SAPLIP) family (12,13). Plant APs are found in either monomeric or heterodimeric forms (9,14); the latter result from post-translational proteolysis, which includes the removal of part or all of the PSI, whereas the PSI is retained in monomeric plant APs (6,8).
Fusion of cellular lipid membranes is an essential process in all forms of life (22), and the mechanism by which membrane fusion occurs, a process typically catalyzed by proteins, continues to be unraveled (23). Merely bringing membranes in proximity to one another is insufficient for fusion (23), and the nature of fusion peptide structures is critical to fusogenic function (24 -26). Disordering of bilayers by fusion proteins, thought to be a critical first step in the catalysis of bilayer fusion (26), results in an increased rate of energetically unfavorable hydrophobic lipid tail protrusion (23). The fusion transition state involves contact formation between lipid tails of opposite bilayers within the intervening hydrophilic region (23), resulting in stalk formation(s) between the two disordered bilayer patches (23,27). Dimerization of helical structures is part of the saposin-mediated bilayer fusion, transfer, and solubilizing mechanisms (28 -30), and these structural rearrangements take place after release from the parent molecule (prosaposin) (16).
The SAPLIP domains of plant APs display membrane permeabilizing activity independent of its "parent" protein (14,31), and they probably act independently (post-proteolytic processing) as a part of the plant defense mechanism against fungal pathogens (3,32). Like PSIs of heterodimeric plant APs, saposins are also expressed as a proprotein and are subsequently processed via proteolytic cleavage (15), resulting in distinct, active tertiary structures consisting of stable helical and coil secondary structures (15,28,(33)(34)(35).
Recently, recombinantly produced PSI of Solanum tuberosum aspartic proteinase (StAP) was shown to kill human pathogens as well as to inhibit fungal sporulation via interaction with and permeabilizaton of microbial plasma membranes (32). Understanding the structural basis for newly characterized antifungal activities is important in the development of novel therapeutic drugs for the treatment of fungal infections (36) in immunocompromised patients (37,38). Furthermore, we propose that understanding structure-function relationships involving PSI-membrane interactions may have relevance to non-plant membrane-bound APs (e.g. memapsins 1 and 2) implicated in Alzheimer's disease beyond the direct elucidation of SAPLIP primary functions. Using StAP PSI as a model system, the present study characterizes the structure of a plant AP PSI as it relates to membrane interactions. The observed saposin C-like tertiary structure and saposin B-like fusogenic activity and the apparent catalysis of energetically unfavorable membrane bilayer disruption and fusion via a pH-dependent helix fusion peptide motif at the PSI N terminus are discussed.

EXPERIMENTAL PROCEDURES
Materials-A PSI synthetic gene optimized for expression in Escherichia coli was purchased from Mr. Gene GmbH (Regensburg, Germany). Plasmids pET19b(ϩ) and pET32b(ϩ), E. coli Rosetta-gami B (DE3)pLysS, and u-MAC TM columns were obtained from EMD Biosciences (San Diego, CA). E. coli TOP10FЈ was from Invitrogen. The GenElute TM plasmid miniprep kit was obtained from Sigma-Aldrich. The QIAquick PCR purification kit and QIAquick gel extraction kit were from Qiagen (Germantown, MD). Restriction enzymes, T4 DNA ligase, and Pfu DNA polymerase were obtained from Fermentas Life Sciences (Burlington, Canada). Primers were synthesized by Sigma Genosys (Oakville, Canada), and thrombin was purchased from Fisher. The RPC column was from GE Healthcare. Phospholipids were from Avanti Polar Lipids (Alabaster, AL).
Construction of Expression Vector pET32z-PSI-Two constructs, a 6.0-kb construct named pET19b-PSI and a 6.2-kb construct named pET32z-PSI, were made for the expression of PSI and thioredoxin-PSI fusion protein (Trx-PSI), respectively, in E. coli. For pET19b-PSI, the PSI insert was amplified using primers FwdPSINdeI (5Ј-CATATGATTGTAAGCATGGAG-TGTAAAACC) and RevPSIXhoI (5Ј-ATCTCGAGTTACGG-GATTTTTTCACACAGTTG), followed by ligation between the NdeI and XhoI restriction sites. The pET32z-PSI construct was made using the PSI insert amplified using primers FwdPSINcoI (5Ј-ATCCATGGCGATTGTAAGCATGGAGT-GTAAAACC) and RevPSIXhoI, followed by ligation between the NcoI and XhoI restriction sites of pET32z, a modified ver-sion of pET32b that contains a deletion between the thrombin cut site and the end of the enterokinase cut site. Each construct was transformed into E. coli TOP10FЈ using the method of Hanahan (39).
Protein Expression-Overnight cultures of E. coli BL21 (DE3)pLysS or Rosetta-gami B (DE3)pLysS transformed with either pET19b-PSI or pET32z-PSI were used to express PSI as per the manufacturer's instructions. Cells were harvested by centrifugation at 4,500 ϫ g for 10 min at 4°C and stored at Ϫ20°C until further use. Frozen cells were thawed at room temperature and resuspended in 20 ml of 20 mM Tris-Cl, pH 7.5. Suspensions were incubated at room temperature for 1 h with gentle shaking, and the resulting cell lysates were centrifuged at 21,000 ϫ g for 30 min at 4°C to remove insoluble matter.
Protein Purification-The following applies only to Trx-PSI fusion protein purification because pET19b-derived PSI was expressed at far lower concentrations and thus was not pursued to purity. Protein purification was performed using an AKTA TM FPLC system (GE Healthcare). Cell lysate soluble fractions were applied to five 1-ml u-MAC columns in series (EMD Biosciences, San Diego, CA) equilibrated with 300 mM NaCl, 20 mM imidazole in 50 mM sodium phosphate, pH 7.4 (binding buffer), followed by washing with the same buffer until a steady base line was obtained. Samples were eluted with 300 mM NaCl, 250 mM imidazole in 50 mM sodium phosphate, pH 7.4, and then dialyzed in 20 mM Tris-Cl, pH 7.4. Thrombin was added to the dialysates at a 1:2000 mass ratio for incubation at room temperature for at least 12 h followed by reapplication of samples to u-MAC in binding buffer three times consecutively at 2 ml/min to remove the Trx fusion tag. Flow-through was collected, dialyzed as above, and then applied to a 1-ml MonoQ column (GE Healthcare) and separated using a 0 -500 mM NaCl gradient in 10 mM Tris-Cl, pH 7.4. Eluent sample was further purified and desalted on a 3-ml RPC column (GE Healthcare); washed with 2% acetonitrile, 0.065% TFA; and eluted with a 90-ml gradient (80% acetonitrile, 0.05% TFA elution buffer). The PSI peak, verified by SDS-PAGE and amino acid analysis (Advanced Protein Analysis Center, Hospital for Sick Children, Toronto, Canada), was collected and placed under vacuum in a Centrivap (Labconco Corp., Kansas City, MO) for 1 h at room temperature to remove the majority of the acetonitrile followed by dialysis against 4 ϫ 1 liter of 5 mM Tris-Cl, pH 7.4, using 1000 Dalton molecular mass cut-off dialysis tubing.
SDS-PAGE-Tris/glycine-buffered SDS-PAGE was conducted according to the method of Laemmli (40) in a Mini-Protean III electrophoresis cell (Bio-Rad). Gels were stained with GelCode Blue (Pierce) and were analyzed for band size and relative intensities using a ChemiGenius II system (PerkinElmer Life Sciences).
Crystallization-The purified StAP PSI protein sample was crystallized using the sitting-drop vapor diffusion method at 293 K using Qiagen PEG Suite screen solutions. The best crystals appeared in the drop containing 0.4 l of protein solution, 0.2 l of reservoir solution, equilibrated against 75 l of reservoir solution (0.2 M lithium sulfate monohydrate, 20% polyethylene glycol 3350).

Diffraction Data Collection, Structure Solution, and
Refinement-X-ray diffraction data for StAP PSI crystals were collected to 1.9 Å resolution using a Rigaku MicroMax 007HF rotating anode and a MAR345dtb system at a wavelength of 1.5418 Å. A data set was collected at 100 K using 25% (v/v) glycerol added to the reservoir solution as cryoprotectant. All data sets were indexed and integrated using the program XDS (41). Integrated intensities were converted to structure factors with modules F2MTZ and CAD of CCP4 (42). BUCCANEER (43) was used for initial automated model building. The structure was refined with REFMAC5 (44), rebuilt with COOT (45), and analyzed using PROCHECK (46) and COOT. Structural superpositions were performed using SSM (47) and ALIGN (48). Figures were generated using PyMOL (49) and UCSF Chimera (50).
Circular Dichroism (CD) Spectropolarimetry-CD analysis of PSI secondary structure was carried out using a Jasco J-810 spectropolarimeter (Jasco Inc., Easton, MD). 200 l of 200 g/ml PSI was loaded into a 1-mm path length quartz cell and scanned over 180 -260 nm at 100 nm/min, 0.5-s response, standard sensitivity, and room temperature. Buffer 140 mM NaCl, 10 mM Tris-Cl, pH 7.4, or 140 mM NaCl, 20 mM MES, pH 4.5, was degassed under vacuum. For reducing condition effect determinations, DTT was added to final concentrations of 1.0, 2.5, and 5.0 mM, and heating was at 95°C in a standard heating block for 5 min in a fume hood, followed by a 30-min cooling period on the bench top.
Preparation of Large Unilamellar Vesicles (LUVs)-LUVs were made of equimolar phosphatidylcholine (PC), phosphatidylethanolamine (PE) and/or phosphatidylserine (PS). To obtain 4 mM phospholipid (PL) suspensions, aliquots of 12.5 mg/ml PL stocks were mixed in a tube and dried under N 2 flush for at least 30 min and then suspended in 500 l of 80 mM calcein, 140 mM NaCl, 10 mM HEPES, pH 7.4, by incubation at 37°C with periodic sonication and vortexing over a minimum of 30 min. LUVs were prepared using a standard miniextruder (Avanti Polar Lipids, Alabaster, AL) containing a 100-nm pore membrane. The LUV prep was then desalted to remove untrapped calcein by gel filtration using a 5-ml HiTrap TM desalting column (GE Healthcare) and visual detection of free calcein in column. To quantify PL postdesalting, the micro-Bartlett phosphorous assay (51, 52) was used to determine the concentration of PLs based on inorganic phosphate content (53). Vesicle concentrations were calculated using PL concentrations, average vesicle diameter (140 nm), and the previously reported areas per lipid molecule: 59.7 Å 2 for PC (54), 57.4 Å 2 for PS, and 59.2 Å 2 for PE (55).
LUV Disruption Assays-PSI-caused perturbation of LUVs was measured by calcein leakage (56, 57) as detected using a Victor2 1420 Multilabel Counter (PerkinElmer Life Sciences) at 25°C. 200-l reactions were set up in 96-well microplates with varying concentrations of LUVs, 500 nM PSI, and either 140 mM NaCl, 10 mM HEPES, pH 7.4, or 140 mM NaCl, 20 mM MES, pH 4.5. Leakage was detected using excitation at 385 nm and emission at 435 nm with 3 s of shaking between readings. End points were measured by incubating LUVs in 0.5% Triton for each condition. Non-linear regression analyses were done using GraphPad Prism 4 (GraphPad Software Inc., La Jolla, CA).
Atomic Force Microscopy-A Veeco Picoforce multimode scanning probe microscope was used with a Nanoscope IV controller to image PE/PS membranes on the native oxide layer on silicon substrates. Images were collected in contact mode before and after the addition of PSI protein in situ. Suspensions of 3 mg/ml PE/PS vesicles were incubated on substrates for 60 min; rinsed with 100 l of 140 mM NaCl, 20 mM MES, pH 4.5, to remove unfused material; and inserted into the fluid cell under 50 l of buffer. Soft triangular cantilevers were used with spring constants between 0.02 and 0.03 newton/m, and the force applied during each scan was 1.5-2.0 nanonewtons. Scans of 5 ϫ 5 m were collected at a rate of 1.5 Hz, and 10 ϫ 10-m scans were collected at a rate of 0.75 Hz, corresponding to a tip velocity of 15.2 m/s. After scanning the same region of substrate repeatedly over 30 min, 20 l of 25 M protein solution was injected directly into the buffer in the fluid cell, resulting in 7 M PSI. Successive images were generated for a single region for time lapse data and for unscanned regions at the end of incubation, to assess changes to the membrane that were caused by repeated scanning of the AFM tip.
Particle Size Determination by Light Scattering-LUVs (100 M) at pH 4.5 were incubated with PSI at room temperature and subjected to light scattering in a Malvern Zetasizer Nano-S (Malvern Instruments, Malvern, UK). A standard 1-ml cuvette was used containing 0.6 ml of sample that was allowed to equilibrate for a minimum of 15 min. Three consecutive measurements of five 30-s runs each were averaged using the refractive index for polystyrene, yielding the calculated average sizes and polydispersity indices.

Structure Solution and Refinement
Recombinant StAP PSI was expressed and purified to Ͼ98% purity with a typical yield of 5 mg/liter of culture, and its identity was verified by N-terminal sequencing (Advanced Protein Analysis Centre, Toronto, Canada). Diffraction data collection statistics are presented in Table 1. Crystals were hexagonal in space group P3 2 21 with unit cell parameters of a ϭ b ϭ 56.47, c ϭ 55.34 Å. The Matthews coefficient (58) for the crystals was 2.24 Å 3 Da Ϫ1 , assuming the presence of one molecule in the asymmetric unit. StAP PSI exhibits a high level of sequence identity (53%) with the SAPLIP domain of prophytepsin PSI (Protein Data Bank code 1QDM; residues 4S-102S), which was used as a model for molecular replacement automated search by PHASER (59). The starting model consisted of a compact molecule and produced a weak solution. Analysis of the initial map showed that PHASER had placed only half of the initial model in a proper orientation despite good quality of the resulting electron density; however, there was sufficient density to accommodate the properly oriented half. At the next step, BUCCANEER (43) was used for automated model building, thereby producing the model of the StAP PSI structure with proper side chains for residues 59 -100 and assigning the other residues as polyalanine. Iterative refinement of the partial model using REFMAC5 (44) and rebuilding in the electron density maps using COOT (45) produced the next model, which corresponded to an elongated, boomerang-shaped molecule.
When all of the residues visible in the electron density were built, translation-libration-screw parameters were introduced during the refinement. The overall anisotropy was modeled with translation-libration-screw parameters by dividing the molecule into three groups comprising residues 0 -26, 27-82, and 83-103. The final model lacks residues 40 -63, which could not be built due to insufficient electron density in this region of the crystal. The refinement statistics for the refined structure are presented in Table 1.

Tertiary and Quaternary Structures of StAP PSI
The overall fold of StAP PSI has a boomerang shape with an extended, open conformation (Fig. 1A) composed of four helices labeled H1-H4 (residues 1-24, 27-34, 66 -82, and 85-99, respectively). H1 is connected to H4 via a disulfide linkage formed between Cys 6 and Cys 99 , H3 is cross-linked to H2 via a disulfide bond between Cys 31 and Cys 71 , and Cys 37 forms a disulfide bond with Cys 68 . The tertiary structure is organized in such a way that one side (top) of the molecule is enriched with polar residues, and the other side (bottom) is enriched with hydrophobic residues (Fig. 1, B and C). Due to the crystallographic symmetry, two molecules form a very tight dimer (Fig.  1, B and C) with a buried surface area of 1746 Å 2 , with the residues involved in the formation of the dimer interface being predominantly hydrophobic.

pH Dependence of Secondary Structure
Because PSI-induced membrane disruption requires acidic conditions (14), the secondary structures of StAP PSI at neutral and acidic pH were compared. CD scans were done in the same buffers used for all other experiments (i.e. 140 mM NaCl buffered by either 10 mM Tris-Cl (pH 7.4) or 20 mM Na-MES (pH 4.5)). Data below 195 nm were noisy, so secondary structure content could not be quantified. Qualitatively, the scans revealed spectra typical for high helix proteins (60); distinct negative absorption peaks occurred in the 220 and 208 nm spectral regions (Fig. 2). Helix content was higher overall at pH 4.5, similar to the pH dependence of influenza A hemagglutinin fusion peptide (61).

Secondary Structure Dependence on Disulfide Bonds
The structure of StAP PSI contains three disulfide bonds within its relatively small 12-kDa tertiary structure, and it was recently suggested that these cross-links are critical to PSI antimicrobial function (32). Hence, the dependence of PSI secondary structure on the presence of cystines was investigated. Fig. 3 contains the spectra of PSI at three concentrations of the reducing agent DTT, each with and without heating. No changes were observed for PSI heated at 95°C under non-reducing conditions, indicating that native PSI secondary structural elements were heat-stable. DTT resulted in high interference at wavelengths below 200 nm, so its levels were limited to 5 mM or lower. Qualitatively from Fig. 3, a relatively minor loss of helix structure (220 and 208 nm negative peaks) occurred with the presence of reducing agent in an apparent dose-dependent manner. By contrast, when PSI was heated under reducing conditions, a more pronounced loss of secondary structure was observed, resulting in an approximate two-state structure change such that heating in 2.5 mM DTT resulted in a more dramatic loss of secondary structure. Further doubling of reducing agent concentration did not cause an equivalent effect, suggesting differential susceptibility of the respective disulfide bonds; the more robust cystine(s) were apparently critical to secondary structure stability of this predominantly helical structure.

Membrane Disruption Activity
Vesicle Leakage-LUVs containing self-quenching fluorophore (80 mM calcein) were used as substrate for the characterization of StAP PSI phospholipid bilayer disruption activity.
Four combinations of PLs were tested at varying concentrations. Neutral vesicles made of 1:1 PC/PE were not affected by StAP PSI at any concentration tested, and none of the PL combinations were disrupted at neutral pH. By contrast, equimolar preparations of PS combined with PC and/or PE resulted in readily detectable activity, and leakage rates were calculated for 0.5 M PSI over the PL concentration range 20 -500 M. Accurate LUV disruption rate determinations above 500 M were prevented by excessive fluorescence signal, and the quality of determinations below 20 M was limited by excessive background noise levels due to untrapped calcein. Rate determinations were calculated relative to PL concentrations in units of M/min (in terms of both PL and LUV concentration) and yielded initial rates with low relative errors. PSI-induced lipid bilayer disruption varied non-linearly with PL/LUV concentration, and the order of leakage rates was PE/PS Ͼ PC/PS Ͼ PE/PC/PS within the concentration range tested (Fig. 4). Non-linear regression analyses for both one-phase exponential association and the Michaelis-Menten equation were compared by F-tests using GraphPad Prism 4, and the independent data sets for all three PL combinations fit better to the Michaelis-Menten model. Table 2 summarizes the kinetic results, including goodness of fit parameters for the regression analyses. Because the mechanism of action has not been characterized (see "Discussion"), apparent Michaelis constants (K m ) were not reported because their meaning in terms of substrate affinity would be undefined in the absence of understanding the relative dissociation rates of PSI and PL in original bilayer (reactant; k Ϫ1 ) and displaced PL in new environment (product; k 2 ). A lack of data at higher lipid concentrations (which produced excessive fluorescence signals) resulted in a relatively large turnover number S.E. value for PC/PS. LUV concentrations above 500 M would allow for more precise kinetic parameter determinations and will require further characterization. Nevertheless, the overall fit to the model was good for each data set, including that for PC/PS (R 2 ϭ 0.99).
Atomic Force Microscopy-AFM height images of PE/PS bilayers were collected using contact mode as described under "Experimental Procedures." LUVs were incubated on the surface of the native oxide layer on a silicon wafer in the same buffer used for LUV disruption assays for 1 h, resulting in  AUGUST 12, 2011 • VOLUME 286 • NUMBER 32 bilayer fusion to the substrate surface. The fused membrane was organized as 5-nm-high islands that were ϳ100 nm wide, dispersed across the substrate (Fig. 5A, top). The bilayer height was within the expected range (35,62), and repeated scans in contact mode over a single region did not change the bilayer morphology significantly. Upon PSI injection, membrane patches fused to form larger regions of uniform membrane separated by larger membrane-free areas (Fig. 5A, bottom). Additionally, 20 -50-nm-high lipid islands formed in regions unperturbed by the AFM tip (Fig. 5B, right); thus, PSI appeared to induce fusion of membrane patches as evidenced by bilayer rearrangement in areas that were not repeatedly scanned. Such formations did not occur on regions subjected to successive scanning; on these regions, considerable smoothing of the membrane occurred. PSI apparently softened or lubricated bilayers, allowing them to be displaced by the AFM tip upon repeated scanning. Fig. 5B (right) shows the distinct effects of PSI both with (center square region) and without (surrounding region) repeated AFM tip scanning. Large regions of 5-nm-high bilayer were observed after eight scans on a single region, whereas repeated scanning in the absence of PSI did not result in similar bilayer rearrangement.

Structure and Function of a Plant-specific Insert
Light Scattering-LUVs (100 M) at pH 4.5 were incubated with PSI at room temperature, and particle size was measured by light scattering in a Malvern Zetasizer Nano-S (Malvern Instruments); results are summarized in Table 3. After 60 min at a PSI concentration equivalent to LUV disruption assays (0.5 M), average particle size slightly increased, and a concomitant increase in the polydispersity index (mass distribution) was more pronounced (Table 3). By 100 min, average particle size had increased by nearly two-thirds (65%), and the polydispersity index had more than tripled. Higher PSI concentration (4 M) was assayed to further investigate the peak area tendency toward higher vesicle size. After just 15 min equilibration time post-PSI addition, average particle size and polydispersity index had not only increased more than for the entire 0.5 M PSI assay, but a newly formed, larger vesicle size (1510 nm) accounting for one-third of the total peak area was observed. Together with the above AFM results, vesicle size measurements indicated that PSI activity resulted in lipid bilayer fusion.

DISCUSSION
Structural Comparison-The overall fold of StAP PSI is similar to the open, extended form of human saposin C (Fig. 6A) as observed in the tetragonal crystal structure (63). The structure of StAP PSI has been compared with the PSI domain of plant phytepsin (9), revealing that the r.m.s. deviation for 38 aligned C ␣ atoms is 0.9 Å. Overall, superposition of a pseudomonomer StAP PSI structure (constructed from its dimer crystal structure) onto the PSI of phytepsin produced an r.m.s. deviation of 1.4 Å (71 C ␣ pairs). Helices H1 and H4 of StAP PSI were superimposable with the equivalent helices of phytepsin PSI (Fig.  6A), whereas helices H2 and H3 from the second StAP PSI molecule in the crystallographic dimer superimpose onto helices H2 and H3 of phytepsin PSI. This result clearly shows that the crystallographic dimer structure of StAP PSI is composed of two extended, domain-swapped monomers. The StAP PSI structure has also been compared with three different crystal structures of human saposin C (Fig. 6A). The C ␣ atoms of res-  idues 6 -19 from saposin C were superimposed (LSQ) to the C ␣ atoms of the first 13 residues of the StAP PSI structure, yielding an r.m.s. deviation of 0.6 Å (tetragonal crystal form, Protein Data Bank code 2Z9A), 0.6 Å (orthorhombic crystal form, Protein Data Bank code 2QYP) and 0.9 Å (hexagonal crystal form, Protein Data Bank code 2GTG). This "open" conformation for SAPLIPs has also been observed for saposin B (Protein Data Bank code 1N69), and all three structures are shown side by side in Fig. 6B.
Saposin-like Activity-StAP PSI-liposome disruption rates varied with PL/LUV concentration in a non-linear fashion, and, to our knowledge, the present study is the first to characterize the kinetic activity of a SAPLIP for varying PL/LUV concentrations of multiple PL compositions, revealing Michaelis-Menten-like kinetics. In attempting to elucidate the nature of PSI activity, AFM results suggested that StAP PSI caused pronounced rearrangement of acidic PL bilayers under the same conditions used for liposome disruption. The type of lipid structures that formed were similar to those previously characterized for saposin C, which showed that saposin C interactions with PL bilayers (64) resulted in rearrangement of membrane patches of increased thickness and membrane destabilization (35,64). Saposin C has also been shown to cause vesicle fusion (65), which involves insertion of the terminal helices into the membrane where saposin-membrane and saposin-saposin interactions carry out a "clip-on" mechanism at neutral and acidic pH. Saposin C causes fusion when present at as little as 0.05 M and produced fused vesicles up to 3000 nm (29).
By contrast, saposin B displays fusogenic activity only against anionic vesicles and exclusively at acidic pH and induces only minor vesicle size increases at protein concentrations of 1 M or below. However, at 2 M, saposin B induces more significant vesicle fusion, yielding an average product ϳ1800 nm in diameter (29). The fusogenic results for StAP PSI in the present study were strikingly similar in that 0.5 M PSI induced only minor average diameter increases, whereas 4 M PSI induced dramatic changes, resulting in new lipid structures averaging 1510 nm (Table 3). Also similar to StAP PSI disruption and AFM observations, saposins B and D disrupt anionic membranes in a pH-dependent process where they solubilize (30,66) and mobilize (30) lipids. Furthermore, saposin B binds and transfers PLs of anionic membranes such that it has a preference for PC transfer (28). Although StAP PSI-mediated vesicle disruption rates were all higher for PE/PS within the concentration range used in the present study, kinetic analysis suggested a higher maximum velocity for PC/PS disruption (p ϭ 0.06), indicating a PL preference similar to saposin B. Perhaps the bulkier choline substituent, relative to ethanolamine, results in more favorable PSI interactions with the bilayer surface and/or results in different bilayer packing density. A study of the nature of SAPLIP PL preference is under way.
The crystal structure of saposin B with bound lipid indicated that PLs interact with the dimeric form of the protein. Dimerization occurs via clasping together two V-shaped protein monomers, thereby forming a shell-like monolayer of ␣-helices with a long interface that buries a relatively large hydrophobic cavity (33). A similar V-shape and quaternary structural arrangement was observed in the present study for the StAP PSI crystal structure, although PSI and open saposin C contain a more obtuse angle in their boomerang shape relative to saposin B (see Fig. 6B), according to structural alignments. Furthermore, PSI membrane effects as detected by AFM were in partial agreement with saposin B in that new lipid structures formed that were higher than the surrounding bilayers; structures observed for saposin B were smaller in size, however (35). These newly formed "granules" could be dislodged by the scanning AFM tip, indicating that they were loosely bound upon saposin B action (35). Similarly, raised lipid bilayers acted upon by StAP PSI became more fluid in regions repeatedly scanned by the AFM tip and were spread out into smooth continuous lipid regions (Fig. 5B), indicating solubilizing and/or mobilizing activity.
Plant AP PSI Activity-Recombinant StAP PSI is toxic to plant (Phytophthora infestans and Fusarium solani) and human (Staphylococcus aureus, Bacillus cereus, and E. coli) pathogens, and it permeabilizes their plasma membranes (32). In both procardosin A and StAP, the PSI has been confirmed to be responsible for membrane interactions (14,32). The interaction of procardosin A PSI with membrane vesicles is pH-dependent and varies with lipid composition (14), in agreement with the present study. Interestingly, the only other crystal structure for a plant AP PSI, prophytepsin, was shown to have a tertiary structure consisting of the "closed" saposin fold (9). That crystal structure dealt with a PSI attached to its parent zymogen molecule and contrasts with the "open" saposin fold observed in the present study for a PSI independent of its zymogen source. By definition, fusogenic activity via the "clip-on" mechanism (29)  The cytotoxicity and plasma membrane interactions of StAP PSI were previously shown to be dependent on its secondary and/or tertiary structure, as evidenced by loss of activity upon DTT treatment (32). In the present study, CD scans at 1 mM DTT resulted in no apparent spectral change, and only minor changes were indicated at DTT concentrations up to 5 mM (without heating). This suggested that disulfide bonds are not critical to PSI secondary structure under normal temperature conditions. When synthetic peptides equivalent to the individual helices of saposin C (i.e. no native tertiary structure) were studied, bilayer fusogenic activity was not observed (29), sug- gesting a critical role of tertiary structure in SAPLIP-catalyzed bilayer fusion. Because cystines are critical to antimicrobial function (32), their role must be to maintain tertiary structure required for fusogenic activity.
Additionally, the DTT titration CD experiment indicated that the disulfide bonds conferred stability to PSI secondary structures, as evidenced by a much more pronounced CD spectral change for heat-treated PSI. The latter was superimposable with that for the non-reducing, unheated sample (Fig. 3), indicating that the PSI disulfide bonds apparently protected the individual helices from heat denaturation. Since both open and closed forms of SAPLIPs contain the same disulfide bonds (15), the primary role of PSI cystines is probably not to maintain the overall fold but perhaps to confer rigidity. Fusion of adjacent membranes consists of the displacement of lipid from its stable, energetically favorable bilayer environment to an aqueous, high energy intermediate state, and such a transition requires enzymatic action (67). Perhaps the energy required to catalyze this event is related to multimer formation (68) and conformation changes (69,70) related to PSI rigidity/stability dependent on disulfide bonds.
Fusogenic Mechanism-There is a correlation between the ability of a fusion peptide to adopt a helical configuration and its ability to promote membrane fusion (71). In addition to saposins, the formation of dimers via complexing of stable, predominantly helical structures is a protein structure scheme common to yeast SNARE-mediated membrane fusion (72,73) as well as viral fusases that are derived from larger precursors that require proteolytic processing to potentiate their fusion activity (71). These associations of ␣-helices contain one hydrophobic face, an arrangement similar to StAP PSI with its five helices, hydrophobic inner cavity, and N terminus. An important fusase fusion peptide for virus-cell fusion is the N-terminal portion of influenza A hemagglutinin, which adopts an ␣-helical conformation in lipid bilayers (74), constitutes an autonomous folding unit in the membrane, and catalyzes lipid exchange between juxtaposed membranes (61).
In some plant APs, the release of PSI occurs via proteolytic cleavage upon acid-induced autoactivation of the precursor protein and subsequent processing, albeit via self-cleavage (6,8,75). Interestingly, influenza A hemagglutinin fusion peptide is inaccessible to membranes at neutral pH; however, a drop of the pH inside the endosome below a critical threshold induces a large conformational change in the parent protein and is subsequently activated by a protease (plasmin) that cleaves the precursor polypeptide (76) into two disulfide-linked polypeptides and a fusion peptide (77). The hemagglutinin fusion peptide has a slightly higher helix content at pH 5 than at pH 7.4, as revealed by comparison of CD spectra (61,74), which are remarkably similar to those for StAP PSI in the present study (Fig. 2) in terms of overall shape (dominant helix content) as well as their x intercepts and superimposed relative spectra (gain of helical structure upon acidification). In addition, the fold of the hemagglutinin fusion peptide (61) is similar to the StAP PSI structure reported here as well as to saposins (78); structural alignment of the N terminus of hemagglutinin fusion peptide (Protein Data Bank code 1IBN) (61) with the N terminus of StAP PSI revealed similar (r.m.s. deviation 1.43 Å for 9 C ␣ carbons) helix-kink-helix folds (Fig. 6C). The tryptophan within the hemagglutinin fusion peptide has been shown to induce its characteristic boomerang shape (26), which is critical to the fusogenic and membrane-disrupting activities of this fold (79). Alignment of StAP PSI with hemagglutinin fusion peptide suggested a similar role for the critical tryptophan residue (Fig. 6C). In this context, the likely protein structural reason for the acidic pH requirement of AP PSI-membrane interactions is the existence of acid-induced helical structure critical for membrane interaction.
Fusase within a Protease-The apparent sharing of a common hydrophobic region fold that is subject to similar pH-dependent secondary structure changes required for membrane FIGURE 6. A, structural superposition of StAP PSI (green) with the plant-specific insert domain of prophytepsin (cyan), tetragonal (yellow), hexagonal (gray), and orthorhombic (red) crystal forms of saposin C. B, side by side comparison of the "open" saposin boomerang fold of saposin B (Protein Data Bank code 1N69), saposin C (Protein Data Bank code 2QYP), and StAP PSI (Protein Data Bank code 3RFI). C, pairwise structural alignment of influenza A hemagglutinin fusion peptide (1IBN; purple) superimposed over StAP PSI (3RFI; green) using UCSF Chimera version 1.5.2 implementing the Needleman-Wunsch algorithm. The overall r.m.s. deviation was 2.013 Å for 19 residues that aligned, ignoring gaps. (65% Ϯ 1 of total area) 15 (peak 2) 1510 Ϯ 300 (34% Ϯ 1 of total area) fusion suggests a common mode of action. Perhaps this helixkink-helix fold is universal (i.e. saposins in animals, hemagglutinin in viruses, and APs in plants) in its membrane fusogenic nature, extending functionality across various species and kingdoms. Collectively, the findings that a plant AP domain displays fusase-like activity (liposome disruption, bilayer solubilization/ lubrication, and bilayer fusion) as well as fusase-like structurefunction character (inter-and intramolecular helix oligomer association, hemagglutinin fusion peptide-like fold, hydrophobic helix end region, and pH dependence of secondary structure-activity) lead to the conclusion that they are indeed fusaselike proteins, acting as discrete entities.
Considering the myriad proteins that have more than one function, the idea of one gene, one protein, one function is insufficient in the study of proteins (80). Recently, a bifunctional AP has been engineered (81); however, no reports have characterized cases of AP "moonlighting" (82) (i.e. serving an additional function beyond the main enzymatic reaction) (83), and only one moonlighting plant peptidase (mitochondrial processing peptidase) has been reported (82,84). In this case, the multiple functions arise from a singular structural fold (85), whereas PSIs are structurally unrelated to their AP "hosts" (i.e. the PSI domain of a plant AP has apparent enzymatic activity and is independent of its "parent" proenzyme, which has its own distinct class of enzymatic activity). Thus, the saposin-like domains of plant APs present a unique case: a distinct, functionally unrelated domain within the primary structure of another domain (the C-terminal domain) of its enclosing protein. Fusase activity from within a protease sequence presents, to our knowledge, the first confirmation and characterization of an independently acting "enzyme within an enzyme."