Atomic Resolution Structure of the HFBII Hydrophobin, a Self-assembling Amphiphile*

Hydrophobins are proteins specific to filamentous fungi. Hydrophobins have several important roles in fungal physiology, for example, adhesion, formation of protective surface coatings, and the reduction of the surface tension of water, which allows growth of aerial structures. Hydrophobins show remarkable biophysical properties, for example, they are the most powerful surface-active proteins known. To this point the molecular basis of the function of this group of proteins has been largely unknown. We have now determined the crystal structure of the hydrophobin HFBII from Trichoderma reesei at 1.0 Å resolution. HFBII has a novel, compact single domain structure containing one α-helix and four antiparallel β-strands that completely envelop two disulfide bridges. The protein surface is mainly hydrophilic, but two β-hairpin loops contain several conserved aliphatic side chains that form a flat hydrophobic patch that makes the molecule amphiphilic. The amphiphilicity of the HFBII molecule is expected to be a source for surface activity, and we suggest that the behavior of this surfactant is greatly enhanced by the self-assembly that is favored by the combination of size and rigidity. This mechanism of function is supported by atomic force micrographs that show highly ordered arrays of HFBII at the air water interface. The data presented show that much of the current views on structure function relations in hydrophobins must be re-evaluated.

Fungi are found all around us in nature in decaying plant litter, as mushrooms that grow in the forests, or as molds that spoil foods. They have a remarkable ability to adapt to a wide variety of environmental conditions and ecosystems (1). They grow by spreading their hyphae, which can penetrate substrates upon which they grow, they can grow upwards into the air to spread spores, or they can attach to solid supports. One group of proteins that have been found to have an important role for fungal growth and development are the hydrophobins. They seem to be unique to fungi and have not been found in any other organisms. A common property in hydrophobins is that they seem to have a wide range of functions that are all apparently related to surface activity in one way or the other (2). Here, surface activity means the tendency to adsorb at interfaces and surfaces. The term interface means the boundary between, for example, air and liquid, or liquid and solid. Another word for a surface-active compound is "surfactant" (3). When a surfactant compound migrates to the air-water interface, it lowers the surface tension, i.e. the cohesive forces between water molecules at the surface.
Hydrophobins are, for example, involved in the formation of aerial structures in fungi (4 -6). Here, the hydrophobins seem to serve a dual role by first lowering the water surface tension, which allows the fungus to penetrate this barrier, and then by forming a protective coating on the aerial structures and spores. Fungi also rely on hydrophobins for attachment to surfaces such as plant leaves or insect cuticles. This property has been demonstrated to be essential for the infectivity of some pathogenic fungi, which cause Dutch elm disease, rice blast, and chestnut blight (7). It is quite possible that the properties of hydrophobins have played an important role during evolution in enabling filamentous fungi to occupy their ecological niches.
Hydrophobins are among the most surface-active biomolecules known (2). They have a size of about 100 amino acids, and are also remarkably stable and can withstand temperatures near the boiling point of water. Consequently, they have potential for several applications. Hydrophobins can be used as adhesion domains for the immobilization of proteins to solid supports (8,9) and as tags in fusion proteins for affinity purification (10). The latter method relies on the very high partitioning of hydrophobins in surfactant-water two phase systems (11). Hydrophobins are also highly efficient foam-forming agents. This property can be illustrated by a practical example. It has been found that microgram amounts of hydrophobin in a bottle of beer causes an explosive gush of foam upon opening the bottle. The contaminating hydrophobins probably originated from fungal growth during the malting or earlier stages (12).
A large number of genes for hydrophobins have been identified, and they all share a characteristic pattern of Cys residues in their primary sequence ( Fig. 1) (13). Hydropathy patterns can be used to divide them into classes I and II (14). The proteins in the two classes seem to differ in the solubility of the aggregates that they form. The sequence similarity of hydrophobins is in general weak, making it sometimes difficult to define them based on primary sequence. However, the characteristic and unique pattern of Cys residues is conserved in all hydrophobins. An essential feature of hydrophobins is that they form different supramolecular assemblies. It has been shown by atomic force microscopy that the class II hydropho-bins HFBI and HFBII from Trichoderma reesei form highly ordered monolayer films (15,16). The same proteins also form crystalline fibrils (17). On the other hand, the class I hydrophobin SC3 from Schizophyllum commune forms a "rodlet coating" where nanometer size rods are formed on interfaces by self-assembly (18). These and other class I hydrophobin assemblages are remarkable because of their insolubility. They are insoluble even in hot solutions of sodium dodecyl sulfate and can only be dissolved in some strong acids, such as trifluoroacetic acid. After evaporation of the acid, it is possible to again solubilize the protein in water and to make new coatings that are indistinguishable from the original ones. For some insoluble class I members, it has been shown that aggregates formed have similarities to amyloid fibrils based on reactions with the stain Congo Red (19,20).
To this point, the properties of hydrophobins such as aggregation, adhesion, and surface activity have hampered structural studies. For example, as is shown in the current work, the widely used data on disulfide pairing (21) were incorrect. An NMR study of the EAS hydrophobin from Neurospora crassa conidia (class I) suggested that the protein is largely unstructured in solution, showing only a small region of ␤-sheet structure (19). Studies on the SC3 hydrophobin have suggested that the protein has different conformational states and different secondary structure contents when in solution or at a hydrophobic-water interface (20,22). It has also been proposed that the protein refolds at hydrophobic-hydrophilic interfaces (23).
In this work, we present the first three-dimensional structure of a hydrophobin. We also present new atomic force microscopy (AFM) 1 data showing that this hydrophobin forms highly ordered self-assembled layers at the air/water interface. Together, these data suggest a mechanism for how this protein functions.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-HFBII was produced using the T. reesei strain (VTT number D-99745) and grown as described in Ref. 24. To one liter of the centrifuged (4000 ϫ g, 20 min) culture supernatant 2 g Berol 325 (Akzo Nobel) was added. The solution was allowed to settle in a separation funnel after mixing and the Berol phase was collected. The Berol was extracted with a 100 ml acetate buffer (50 mM, pH 5.0) and 15 ml of isobutanol with a subsequent centrifugation (3220 ϫ g, 15 min). The protein was further purified by preparative reversed phase chromatography using a Vydac C4 (1 ϫ 20 cm) column and a gradient elution from 0.1% trifluoroacetic acid to 100% acetonitrile containing 0.1% trifluoroacetic acid. Peak fractions were pooled and lyophilized. The protein identity was confirmed using mass spectroscopy and antibodies (11,25).
Crystallization and Structure Determination-The protein crystals grew in hanging drops using 25% polyethylene glycol 2000, 0.2 M lithium sulfate, 10 mM MnCl 2 and 0.1 M Na-HEPES (pH 7.5). Crystallization drops were prepared by mixing equal amounts of protein solution (8 mg/ml) and crystallization agent. The data set was collected at 1.5 Å resolution using a copper rotating anode as an x-ray source (wavelength 1.54 Å) and at 1.0 Å resolution at the EMBL X11 beamline at the DORIS storage ring, DESY (wavelength 0.8126 Å) and processed with DENZO (26). The anomalous difference Patterson map calculated with the data measured at the home laboratory showed a clear single peak at Harker sections, suggesting that one manganese ion would have bound to the protein. The coordinates of this position were used in the program ACORN (27) as a starting point for the calculation of phases. The refined phases were further used in the program ARP/WARP (28), which automatically built the preliminary model for two protein molecules in the asymmetric unit. The structure was refined by iterative cycles of manual refitting with O (29) and positional refinement with REFMAC. The final model contains two protein molecules in an asymmetric unit. The refined molecules contain residues 1-70. The last residue Phe-71 was not visible and not included in the model. The figures were drawn with Setor (30) and PyMOL. The details of data collection and structure refinement are provided in Table I. Coordinates have been deposited in the Protein Data Bank with the code 1R2M.
AFM-A 10 l droplet of 10 g/ml solution of HFBII in double distilled water was dried on highly oriented pyrolytic graphite (HOPG) in a vacuum desiccator. Images were acquired under ambient conditions in tapping mode using a NanoScopeIIIa Multimode AFM (Digital Instruments) equipped with an "E" scanner. The tapping mode was used with scan rates 0.6 -1 Hz and as low a force as possible. Noncontact silicon cantilevers (NCH, NanoSensors) with the nominal resonance frequency of around 300 kHz and tip radius ca. 10 nm were used. A scanning probe image processor (SPIP, Image Metrology, Denmark) was used in image processing, which only included flattening in order to remove possible tilt in the image data, and in image analysis. 1 The abbreviation used is: AFM, atomic force microscopy.  Imaging of the HFBII surface was reproducible. The topography and phase contrast images were captured simultaneously. The phase contrast image shows the phase difference between the oscillations of the cantilever driving piezo and the detected oscillations. It is thought that image contrast is derived from surface properties such as stiffness and viscoelasticity (hard tapping) or hydrophilicity/hydrophobicity (light tapping), but it also shows enhanced edge structures.

RESULTS
Structure of HFBII-We have crystallized a member of the class II hydrophobins, HFBII from T. reesei, and determined its three-dimensional structure at 1.0 Å resolution. HFBII is a single domain protein with the approximate dimensions of 24 ϫ 27 ϫ 30 Å. The structure represents a fold not previously described. It consists of one ␤-hairpin motif (strands S1 and S2) linked to an ␣-helix that is linked to the second ␤-hairpin motif (strands S3 and S4). These two motifs are arranged together in an unusual way (31). They form a barrel consisting of four antiparallel ␤-strands in the order S4-S2-S1-S3. The two ␤-hairpins interlock together in much the same way as the two leather pieces of a baseball are sewn together (Fig. 2).
Disulfide Bridges-A distinctive feature of hydrophobins is that they contain eight Cys residues that form four disulfide bridges (Fig. 3). The current structure shows that bridge 14 -26 connects the strands of the first ␤-hairpin together and that bridge 53-64 connects the strands of the second ␤-hairpin. The four sulfur atoms of these two bridges are completely inside the ␤-barrel. The remaining two bridges are outside the ␤-barrel. The bridge 3-52 connects the N-terminal loop to ␤-strand S3, and the bridge 13-43 connects the ␤-strand S1 and the helix together. The disulfide bridges are symmetrically located in almost the same plane in the structure (Fig. 3). They form a disulfide bridge network that efficiently spans the entire molecule thus creating a very stable and compact structure.
Hydrophobic Surface Patch-Hydrophobins are highly surface-active and interact strongly with hydrophobic surfaces. Surface features of the protein are expected to be important for this function. The amino acid sequence comparison of class II hydrophobins revealed a number of conserved and exposed hydrophobic residues (Figs. 1 and 4). Interestingly, these residues are all aliphatic and are located in two ␤-hairpins. The HFBII residues Val-18, Leu-19, Leu-21, Ile-22, and Val-24 are located in the first hairpin (loop 1). Residues Val-54, Val-57, Ala-58, Ala-61, and Leu-63 are located in the second hairpin (loop 2). A hydrophobic surface patch is thus formed and is completed by Leu-7 in the N-terminal loop. As we examine the overall shape of the protein, we note that the surface at the hydrophobic patch is relatively flat. It also seems that loop 1 has a pronounced role in the formation of this hydrophobic patch. The total surface area of a HFBII monomer is about 3200 Å 2 , and the estimated area of the conserved hydrophobic patch is about 400 Å 2 (12% of the total).
Self-assembled Layers-AFM images show polycrystalline structures consisting of two-dimensional single crystalline domains in the surface layers of HFBII. In Fig. 5, an image of 100 ϫ 100 nm 2 in size is shown. The crystallinity of the protein layer was confirmed and the lattice constants were determined by Fourier transformation of the image data. An oblique structure with lattice constants a ϭ 67.1 Å, b ϭ 44.8 Å, ␥ ϭ 139.2 degree was obtained. The lattice constants equal qualitatively to the lattice constants found in Langmuir-Blodgett films of HFBII (15). The slight distortion observed in the AFM images, and hence in the transformed images also, is due to scanner hysteresis, creep, and drift in the microscope. These problems are important to consider because of the relatively slow scan speeds and because images of the first full scan were captured in order to obtain images of the soft protein surfaces with a non-contaminated tip. The way in which the sample was prepared and the similarity to the previously studied Langmuir-Blodgett films suggest that the image most likely represents the air-water interface that is deposited on the solid support as the water has evaporated. DISCUSSION The presented structure of HFBII offers new perspectives for understanding and exploiting the biophysical properties of hydrophobins, such as their strongly amphiphilic nature, selfassembly, and aggregation. HFBII has a globular structure that is stabilized by a network of disulfide bonds. On the protein surface, there is distinct patch of aliphatic side chains that are conserved in homologous proteins. We suggest that this patch makes the hydrophobin molecule amphiphilic and is therefore highly important to the function of the protein. One side of the molecule is thus easily solvated by water, whereas the other tries to escape it. However, the exposed hydrophobic side chains must be prevented from turning to the inside of the protein hydrophobic core to escape water. Therefore, the rigidity of the HFBII molecule may assist in keeping the side chains of the hydrophobic patch exposed to solvent.
In addition to the atomic resolution structure of HFBII, we also present here atomic force micrographs of HFBII that show molecular films with highly regular patterns. We previously have shown that ordered films can be made using the Langmuir-Blodgett technique (15), but by refining the technique we were now able to show that such ordered interface layers are formed spontaneously at the surface of water droplets. We thus have two views, i.e. the structure of the individual molecule, and the structure of an assembly that they form. Together, these data can be used to understand how this protein functions.
As surfactants (surface-active agents), the hydrophobins are undoubtedly interesting molecules. HFBII is able to reduce the surface tension of water from 72 mJ/m 2 to 28 mJ/m 2 at a concentration of 20 g/ml. 2 One major difference, compared with most other highly surface-active molecules, is that the relatively large size of HFBII is combined with rigidity in the molecule, i.e. it is a shape-persistent molecule. The question arises how shape persistency affects amphiphilicity compared with more small or flexible surface-active molecules. A possible effect is that by reducing the conformational degrees of freedom, the mixing entropy in solutions can be drastically reduced, which promotes aggregation tendency and reduces solubility (see Ref. 32). This may be important in surface attachment in order to guarantee that the hydrophobins adhere well and are not lost due to redissolvation. That is, the highly ordered surface layer is formed because the individual molecules are rigid, amphiphilic, and large enough to make favorable intermolecular contacts. Therefore, favorable lateral contacts can enhance the surface activity of hydrophobins. The role of these factors on the origin of the high surface activity of 2 S. Askolin, manuscript in preparation.

FIG. 4. The hydrophobic patch of HFBII in two orientations.
The amino acid residues that probably participate in interactions with a hydrophobic substrate are shown. The first ␤-hairpin is in red and the second in purple. A) and topography (B)). For comparison of dimensions, a black dot representing HFBII in the same scale has been drawn between the images. The sample was prepared by drying down a drop of HFBII solution on a solid support. Therefore, the film most likely represents the air-water interface of the drop. The data suggest that formation of self-assembled films is part of the mechanism through which HFBII functions. The image size is 100 ϫ 100 nm 2 , and the height scale in the topography image is 2 nm. The z-scale in the phase contrast image is 35 degrees. hydrophobins could be tested by mutating amino acids, such as the charged residues, putatively forming lateral interactions.

FIG. 5. An atomic force micrograph of an ordered film of HF-BII (phase contrast (
Previously, it has been shown that hydrophobins can have a complex multimerization behavior in solutions and at interfaces (15). HFBII probably forms tetramers at high protein concentrations, but dimers or monomers occur at lower concentrations (17). The wild-type HFBI hydrophobin behaved similarly, but formed aggregates approximately the size of decamers when it was genetically fused to a 40-kDa cellulase enzyme (9). In the crystal structure of HFBII, the asymmetric unit contains two protein molecules that form a dimer (Fig. 6A). The molecules pack with their hydrophobic patches toward each other, with the contact face being roughly half of the hydrophobic patch, about 230 Å 2 . Almost all residues on the dimer interface belong to the first hairpin: Val-18, Leu-19, Leu-21, Ile-22, and Val-24. Only Ala-61 from the second hairpin participates in the crystal contact, leaving the majority of the hydrophobic residues from the second hairpin exposed to solvent. Thus, it is evident that by burying hydrophobic surfaces in the oligomer interfaces, hydrophobins become more soluble.
The small monomer-monomer contact area in the crystal suggests a connection between multimer formation and surface adhesion. If steric factors in the dimer interactions leave a significant part of the hydrophobic surface to be solvated by water, this dimer state can be considered a high-energy state. Larger hydrophobic interfaces such as the air water interface, could more easily provide a larger water-free environment, and thus produce a low-energy state. HFBII molecules would then preferentially expose their hydrophobic faces to larger interfaces than to each other (Fig. 6B).
Class II hydrophobins have structurally critical residues that are highly conserved within the family. For example, Phe-39, which is in the ␣-helix, is completely conserved and packed against the ␤-barrel. This kind of conservation would indicate that the structures of other class II hydrophobins would be relatively similar to HFBII and that the fold does not allow larger changes in the amino acid sequence. However, the sequence similarity between classes I and II is much weaker. Because the disulfide pattern and spacing between the hydrophobin families is the most conserved part, it would also suggest that the pairing would be the same in the two families. So far this has not been experimentally verified. However, it may be a very difficult task to demonstrate disulfide connections using biochemical methods because the C and N termini are connected by one of the disulfide bridges, Cys-3-Cys-52. As previously noted in the literature, one main difference between class I and II hydrophobins is the length of the segment between the third and fourth Cys residues (18,33). Interestingly, this segment coincides with the first hairpin, which has an essential role in forming the hydrophobic patch. In class I hydrophobins, this segment also contains several conserved aliphatic residues, which would suggest that the loop has similar roles in both classes.
In view of published data on some class I hydrophobins, it is a surprising finding that the structure of HFBII is very compact and has a disulfide pairing that does not allow much flexibility in the chain. The previously suggested (and now proved incorrect) disulfide arrangement allowed for hypothesizing on significantly more conformational freedom (18). Structural studies on class I hydrophobins such as SC3 and EAS (18,19,34) have suggested that an disorder-order transition is involved in the formation of aggregates and films. It has also been proposed that SC3 goes through several conformational stages with varying contents of ␣-helix and ␤-sheet that depend on interface and external conditions when it aggregates. The structure does not provide any clues as to how this could occur, except that for reasons discussed above it is likely that any such conformational changes would be restricted to movements in loops.
It is interesting that filamentous fungi have, during evolution, produced a surfactant protein that at least structurally, and most likely also in its mechanism, differs from typical surfactants. One consequence seems to be that hydrophobins show more diverse properties, such as tight adhesion, and formation of protection coatings and can, therefore, perform several diverse functions that are important for growth and adaptation to the environment. It is also interesting that fungi often have several genes for hydrophobins and that these are expressed differentially (13,25,35). It is clear that these different hydrophobins differ in their physical properties, but so far little progress has been made toward understanding how the fungi utilize the differences in different stages of their life cycle and in response to changes in external conditions.
Additionally, it is exciting that the current structure reveals a highly atypical surfactant that can add new knowledge to our understanding of the physical chemistry of surfactants in general. Currently, self-assembly in materials are receiving great attention because of possible use in nanotechnology (36). This also motivates further study of this protein, because it can clearly be regarded as a self-assembling nanosystem that has acquired extraordinary properties as a material due to specific interactions and self-organization at the nanoscale.
FIG. 6. Multimerization and assembly of HFBII. A, the observed structure of a HFBII dimer. Different HFBII monomers are colored in blue and red. The hydrophobic patches of both molecules are shown in yellow and orange. B, a model for the assembly of HFBII. There is equilibrium between the monomeric, oligomeric, and monolayer state. Amphiphilicity drives a monomeric hydrophobin to oligomers or monolayers. The monolayer has a highly crystalline structure (see Fig. 5), indicating favorable lateral interactions that further stabilize it.