The surfaces of two‐ and three‐dimensional φ29 connector crystals were imaged in buffer solution by atomic force microscopy (AFM). Both topographies show a rectangular unit cell with dimensions of 16.5 nm×16.5 nm. High resolution images of connectors from the two‐dimensional crystal surface show two connectors per unit cell confirming the p4212 symmetry. The height of the connector was estimated to be at least 7.6 nm, a value close to that found in previous studies using different techniques. The 12 subunits of the wide connector domain were clearly resolved and showed a right‐handed vorticity. The channel running along the connector had a diameter of 3.7 nm in the wide domain, while it was 1.7 nm in the narrow domain end, thus suggesting a tronco‐conical channel shape. Moreover, the narrow connector end appears to be rather flexible. When the force applied to the stylus was between 50 and 100 pN, the connector end was fully extended. At forces of ∼150 pN, these ends were pushed towards the crystal surface. The complementation of the AFM data with the three‐dimensional reconstruction obtained from electron microscopy not only confirmed the model proposed, but also offers new insights that may help to explain the role of the connector in DNA packing.
Bacteriophage assembly is a complex process that involves specific interactions to mature unassembled structural components into well‐defined intermediates of the viral particle. In this process, the proteins of the head–tail connecting region (connectors or portal proteins) are implicated in the first steps of prohead assembly and in the translocation of DNA inside the viral head. All the different connectors studied so far consist of an outer region with several domains, and an inner tubular region with a channel (reviewed by Valpuesta and Carrascosa, 1994). Connectors from the T4 phage (Driedonks et al., 1981), the T3 phage (Carazo et al., 1986b) and the λ phage (Kochan et al., 1984) exhibit a 12‐fold rotational symmetry. Intact necks extracted from phage φ29 particles which are built from the connector protein (p10) and an additional protein (p11) possess a 12‐fold rotational symmetry as well (Carrascosa et al., 1982). However, other connectors, such as those from the SPP1 phage, show 13 morphological subunits (Dube et al., 1993). Connectors of phage T7 (Kocsis et al., 1995; Cerritelli and Studier, 1996) and phage φ29 (Carrascosa et al., 1985; Carazo et al., 1986a; Tsuprun et al., 1994) assembled from overexpressed proteins using cloned genes exhibit polymorphic variants containing 12 and 13 subunits. The significance of this variation is unclear, although, at least for φ29, the physiologically relevant necks (Carrascosa et al., 1982; Carazo et al., 1985) and 9 Å resolution data from frozen–hydrated connector protein crystals (Valpuesta et al., 1994) display a 12‐fold symmetry.
The three‐dimensional reconstruction from electron microscopy data of the φ29 connector arranged in two‐dimensional crystals shows an open channel running along the H axis connectors (Carazo et al., 1986a). In contrast, the channel was found to be closed in necks of connectors extracted from viral particles (Carazo et al., 1985). The presence of two conformational states of the channel suggested an active role for the connector in DNA packaging (Carrascosa et al., 1990). DNA‐binding studies carried out with φ29 connectors in vitro have demonstrated that there is a clear affinity of the connector for DNA ends (Herranz et al., 1990) and also different binding behavior depending on the DNA structure (Turnquist et al., 1992; Valpuesta et al., 1992; Urbaneja et al., 1994). Furthermore, linear DNA has been proposed to interact with the connector channel (Valle et al., 1996).
Despite these extensive structural and functional analyses, the surface topography of the outer and inner domain of the connector are not well resolved. We have established conditions to reproducibly acquire topographs of membrane proteins at subnanometer resolution using the atomic force microscope (AFM) (Karrasch et al., 1994; Müller et al., 1995b; Schabert et al., 1995). Directly observed conformational changes of single bacteriorhodopsin loops (Müller et al., 1995a) induced by the interaction with the AFM stylus demonstrate the sensitivity of this technique. In addition, the observed time‐dependent opening and closing of single pores of an eubacterial surface layer (Müller et al., 1996) underline the potential of the AFM to monitor biological processes. In this work, we demonstrate that the 12 subunits of the outer domain of the φ29 connector can be resolved with the AFM. The new structural information provided by the the AFM is compared with the data from cryo‐transmission electron microscopy, and the flexibility of the narrow connector end is discussed.
The overexpression of the gene coding for the protein comprising the φ29 connector (p10, Mr = 35 800) in Escherichia coli allowed large amounts of connectors to be produced and well‐ordered two‐dimensional crystals to be assembled (Valpuesta et al., 1994). Imaged in buffer solution with the AFM, these crystals were found to exhibit diameters of >8 μm (Figure 1a). The crystalline sheets adsorbed without fold or cracks onto the freshly cleaved mica surface. This is a prerequisite for high resolution imaging with the AFM (Müller et al., 1995b). The thickness of the sheets adsorbed to mica was 8.3 ± 0.6 nm (n = 49). When adsorbed onto a crystalline sheet, the thickness of the upper sheet was 8.5 ± 0.5 nm (n = 20). Images recorded at higher magnification showed single connectors assembled into a rectangular lattice (Figure 1b). The unit cell dimensions of 16.5 nm× 16.5 nm ± 0.5 nm were identical to those measured for two‐dimensional crystals using electron crystallography (16.5×16.5 nm; Carazo et al., 1986a). Depending on the purification batch used, three‐dimensional crystals could be grown (Valpuesta et al., 1994). The low magnification image showed that these three‐dimensional crystals were built up from sheet‐like structures (Figure 2). At higher magnification (Figure 2, inset), the rectangular arrangement of the connectors was distinct.
Submolecular resolution was achieved only with single layered two‐dimensional crystalline sheets (Figure 3). The topography of the connectors strongly depended on the force applied between the stylus and the crystal surface. At the top of the image, the applied force was between 50 and 100 pN (Figure 3; between a and b). The narrow domains of the connectors appeared in their extended conformation. When the force was increased to ∼150–200 pN, the connector ends were compressed and did not protrude from the crystal surface, although the central channel was clearly resolved (Figure 3; between b and c). When the force was reduced, the extended conformation of φ29 connectors was seen again (Figure 3; between c and d). This structural change was fully reversible and could be repeated several times during continuous scanning and imaging of the same area at high resolution. The narrow connector end was blurred in its extended form (i.e. when imaged at low force), whereas it was seen as a small rather sharp ring when pushed towards the substrate by the tip. The lower region in Figure 3 (below e) was imaged with a force of ∼250–300 pN. At this force, the crystal was disrupted by the interaction with the stylus, and the bare mica surface became visible.
To show the submolecular structures in more detail, the surface of the two‐dimensional crystal was imaged at high magnification (Figure 4). At minimum forces (≈50 pN), the narrow end of the connector protruded fully outwards from the surface. While the scanning stylus imaged the holes of the narrow ends of the connectors, it did not interact sufficiently with the wide connector domains to achieve submolecular resolution (Figure 4a). This was not the case in the area of lattice defects (white circle). Here, the subunits of the wide domain of the connector were directly imaged. At large defects, the stylus reached the mica surface, and the height difference from the core of the wide connector end to the mica was determined to be 7.6 ± 0.5 nm. If higher forces were applied to the stylus (Figure 4d), the narrow ends were pushed down towards the surface of the wide connector end. The latter was thus accessible to the tip, and all 12 subunits were often visible in the raw data. As in Figure 3, the narrow connector ends were less sharp in their extended form than when pushed down (compare Figure 4a and d). Occasionally, only half of these ends were imaged. The 12 subunits of the wide domains were distinct and exhibited a right‐handed vorticity in the unprocessed image. In none of the recorded images were 13 subunits observed. To estimate the inner and outer diameter of the narrow connector end, an average of the extended form was calculated from 180 unit cells (Figure 4b), and 4‐fold symmetrized (Figure 4c). The unsymmetrized average documents a lateral stretching of the protein in the fast scanning direction. The flexibility was probably responsible for the elliptical distortion and prevented the substructure of the narrow connector domain from becoming distinct in the average. The narrow end protruded by 2.3 ± 0.2 nm (n = 53) above the rosette of the wide connector end. The inner diameter was estimated to be 1.7 ± 0.3 nm, and the outer diameter 6.4 ± 0.4 nm, but both the geometry of the AFM tip (Schwarz et al., 1994) and the flexibility of the extended narrow end were likely to influence these values. The measured values of the outer and inner diameters thus represent the maximum and the minimum possible dimensions of the channel, respectively.
Best averaging results of the wide connector domain were achieved after single particle averaging of selected particles (Figure 5). In the average of the wide connector end comprising 183 particles (Figure 5b), the right‐handed vorticity that was distinct in unprocessed single particle images (Figure 5a) was enhanced. Each of the 12 subunits was seen not only in the unsymmetrized average (Figure 5b), but could also be discerned in most of the unprocessed views of individual connectors (Figure 5a). The lateral resolution of these topographs was estimated according to the spectral signal‐to‐noise ratio to 1.8 nm (Unser et al., 1989). The height of the protrusions above the flat inner connector annulus was between 0.15 and 0.40 nm. The rotational power spectrum (Figure 6) of the average displayed in Figure 5b exhibited 2‐, 4‐ and 12‐fold components, supporting the 12‐fold rotational symmetrization applied to give the final result shown in Figure 5c. The outer diameter of the core was 13.3 ± 0.3 nm, a value that agrees with the projection data obtained by transmission electron microscopy of frozen–hydrated connector crystals at 9 Å resolution (Valpuesta et al., 1994), while the inner diameter was 3.7 ± 0.4 nm. As mentioned above, the geometry of the AFM tip (Schwarz et al., 1994) is likely to influence the measured value of the inner and outer diameter of the connector channel.
The perspective view of the crystal surface was calculated from 4‐fold‐symmetrized correlation averages of either the extended narrow connector end or the wide connector domain (Figure 7). In agreement with previous results (Carazo et al., 1986a), this view showed two connectors per unit cell oriented in opposite directions and perpendicular to the crystal, consistent with the space group p4212 described for these crystals by Valpuesta et al. (1994). The 12 protrusions of the wide connector domain were 2.7 ± 0.1 nm apart, and had a height of 0.3 ± 0.1 nm (n = 81). The narrow and the wide end of the connector exhibited different channel diameters, suggesting that the shape of the channel is not fully cylindrical, but rather tronco‐conical or bottleneck‐shaped.
We recorded topographs with molecular resolution of two‐ and three‐dimensional φ29 connector crystals using the AFM in buffer solution. The tetragonal unit cell exhibited a side length of 16.5 nm and showed two connectors in opposite orientations. Images with submolecular resolution were only obtained from single layered two‐dimensional crystal sheets. This indicates that shear forces were sufficient to distort the unstable three‐dimensional crystals. High resolution images demonstrate that the narrow end of the connector is rather flexible. At vertical forces of ≥150 pN it was pushed down onto the crystal surface. This either could result from a compression of the narrow end, or the connectors with the narrow end pointing towards the tip could have been pushed down onto the mica. While the latter would explain the rather sharp image of the thin ends at forces of ∼150 pN, the force difference of 100–150 pN required to induce the putative compression is remarkably close to the 200 pN reported to produce a conformational change of bacteriorhodopsin (Müller et al., 1995a). However, the binding forces between avidin and biotin (160 pN; Florin et al., 1994), streptavidin and biotin (340 pN; Lee et al., 1994) and between antigen and antibody complexes (60 pN; Dammer et al., 1996) are also within the same range. Therefore, the force does not permit a discrimination between the two postulated mechanisms, a compression (i.e. a conformational change) or a vertical displacement (i.e. a breakage of specific molecular interactions within the crystal). In any case, the deformation of the connectors by ∼2.3 nm corresponds to an energy of 55–82 kcal/mol. This low value is compatible with the reversibility of the process. It also implies that the extended narrow connector end was not stable, explaining its stretching along the fast scan direction as well as the low resolution achieved.
The height difference between the surface of the crystalline sheet and the mica surface was 8.3 ± 0.6 nm, while in sheets that formed stacks the height difference was 8.5 ± 0.5 nm. The height of a single connector measured at lattice defects imaged at high magnification (Figure 4a) was 7.6 ± 0.5 nm. Since the narrow connector end protruded 2.3 nm above the wide end, the height of the crystal is calculated to be 9.9 nm, ∼1.5 nm more than the heights measured from the low magnification topographs. The pixel size in the latter case, 20 nm, was larger than the unit cell size, 16.5 nm. This implies that an average thickness is measured rather than the total one. Moreover, the force applied by the stylus is different in each case: the force used to measure the crystal thickness was ≥300 pN, whereas forces between 50 and 100 pN were used to obtain high resolution images. It cannot be excluded that the thin connector ends were slightly displaced or deformed even during low force imaging. The measured connector length, 7.6 ± 0.5 nm, is therefore a minimum value. This value agrees with results from other techniques. Three‐dimensional reconstruction of a negatively stained specimen (Carazo et al., 1986) gave a height of 7.5 nm, whereas scanning tunneling images of metal replicas of freeze‐dried crystals gave a height estimate of 7.0 ± 0.5 nm (Vélez et al., 1996). The thickness of the wide domains of the φ29 connectors can be derived from the height differences recorded with the AFM. Assuming that the narrow ends protrude fully from the crystal surface on both sides, the thickness of the wide domain is calculated to be 2.7 ± 0.6 nm. Thus, the length of the narrow connector end is 5.0 ± 0.8 nm. These values agree within the experimental errors with results from a negatively stained specimen (Carazo et al., 1986a) and from STM measurements (Vélez et al., 1996). However, attachment of biomolecules can result in a flattening of the surface facing the solid support, a phenomenon which is enhanced further by air‐drying effects (Kellenberger et al., 1982; Baumeister et al., 1986). Therefore, the results obtained from either negatively stained or freeze‐dried samples represent underestimates of the connector dimensions, too.
A distinct structural feature evident in the topographs of the connector crystal recorded with the AFM is the right‐handed orientation vorticity of the wide connector end. This feature has been observed less clearly in negatively stained samples (Carazo et al., 1985; Carrascosa et al., 1990), and it may be relevant for understanding the DNA packaging mechanism. In fact, several models proposed before (reviewed in Valpuesta and Carrascosa, 1994) involved some sort of rotational movement for the connector during DNA packaging, either actively coupled to the DNA movement or as a strain‐relieving mechanism. In this context, the vorticity of the 12‐folded wide connector domain which interacts with the head vortex imposes a defined rotational direction and provides structural support for models relying on an active movement of the connector during DNA translocation (Hendrix, 1978).
Another interesting morphological feature obtained by AFM is the characteristic protrusion of each subunit from the outer surface of the wide domain. This detailed surface information completes the three‐dimensional structure of the wide connector end which interacts with the prohead structural proteins. In previous three‐dimensional reconstructions from negatively stained samples, this surface information was particularly poor as a result of the limitation imposed by the missing cone inherent to tilt series data (Carazo et al., 1986a) and by alteration of the structure during adsorption (Kellenberger et al., 1982; Baumeister et al., 1986). A comparison of the connector model determined by electron microscopy (Figure 7b) and the model generated with the additional data from AFM topographs illustrates the progress (Figure 7c).
The AFM images have also provided new information showing a difference in the diameter of the channel on the two sides of the connector: while it is 3.7 nm in the wide domain, it is 1.7 nm in the narrow domain. Although these values are likely to be underestimates as a result of tip geometry (Schwarz et al., 1994), it is clear that the shape of the channel is not fully cylindrical, but rather tronco‐conical or bottleneck‐shaped. The precise definition of the channel structure is important, because the closing of the channel may be linked to the final step of DNA packaging (Carrascosa et al., 1990). A narrow channel end close to the location where the necks extracted from viral particles appear to be plugged (Carazo et al., 1985) supports a model involving small concerted movements of the subunits for the opening/closing mechanism of the channel. As reported for bacteriorhodopsin and the hexagonally packed intermediate layer (HPI), conformational changes can be monitored directly with the AFM (Müller et al., 1995a, 1996). Therefore, predicted conformational changes of the narrow domains (Valpuesta and Carrascosa, 1994) may be observed in situ with the AFM. Studies to monitor this process are now in progress.
Materials and methods
φ29 connectors were obtained from E.coli carrying a recombinant plasmid containing the gene encoding protein p10 (Mr = 35 800). The overexpressed p10 was purified by two‐step ion exchange chromatography (Ibañez et al., 1984). Large two‐dimensional crystals were obtained after incubating a solution of the purified connectors (3–4 mg/ml) at 4°C and increasing the ionic strength of the solution to 2 M NaCl over a few weeks. The crystal‐containing stock solution (∼10 mg/ml in 2 M NaCl) was diluted 300‐fold in 1 M KCl and 20 mM HEPES at pH 7.2. A 25 μl drop of this solution was deposited on freshly cleaved muscovite mica discs (10 mm in diameter) that had been glued onto 25 mm Teflon discs. After 15 min, the sample was rinsed gently with the imaging buffer (0.5 M KCl, 20 mM Tris–HCl, pH 8.4 for two‐dimensional crystals) to remove sheets that were not firmly attached.
Atomic force microscopy
After the crystals were adsorbed to mica, the sample was mounted on the piezo scanner of the atomic force microscope (Nanoscope III, Digital Instruments, Santa Barbara, 93117, USA) which was equipped with a liquid cell. Cantilevers purchased from Digital Instruments were 200 μm long and had oxide‐sharpened Si3N4 tips. The force constant of the cantilevers used to calibrate the applied force was calculated to be 0.15 N/m according to Butt et al. (1993) from the cantilever dimensions measured with a scanning electron microscope. After thermal relaxation, the drift of the cantilever deflection angle was at a minimum. Initial engagement of the tip was then performed by setting the scan size to zero to minimize specimen deformation or tip contamination. Prior to scanning the sample, the operating point of the servo system was set to forces below 0.5 nN.
At low magnification (frame size ∼600 nm), images were recorded in the error signal mode, acquiring the deflection and height signals simultaneously. The deflection signal was minimized by optimizing gains and scan speed. At high magnification, the deformation of the sample (Weisenhorn et al., 1993) was monitored and minimized by comparing the height profiles acquired in the trace and retrace direction, and at different scan angles. The applied force was corrected manually to compensate for the thermal drift of the microscope.
The ‘J’‐piezoscanner used had a scan range of 120 μm. Calibration of the scanners was carried out using layered crystals such as mica (Bailey, 1984) and transition metal dichalcogenides (Wilson and Yoffe, 1969) as substrate references. The lateral calibration was carried out by comparing the lattice constants of the layered surface and, for calibration in the z‐direction, defects (large holes or steps) were used (Jungblut et al., 1992).
Image processing was carried out with the SEMPER image processing system (Saxton et al., 1979). Images (512×512 pixels) were flattened line by line prior to calculating correlation averages using a reference unit cell selected from the raw data. Single particle averages were obtained as described (Schabert et al., 1994). Angular harmonics were determined by interpolation of the wide connector end topography to radial coordinates, and integration of the radial interval containing the most prominent corrugation. The rotational origin has been taken as the center of the channel of the averaged topography. The three‐dimensional view of the averaged surface profile (Figure 7) was calculated with the Imagic software (version 0.9d65.0, share ware). Averages from single views of the wide domain were obtained using a variant of the reference‐free algorithm described by Penczek et al. (1992), using a pyramidal system for particle pre‐alignment. The resolution of averages was calculated using the spectral signal‐to‐noise ratio method (Unser et al., 1989), and the final averages were filtered to that resolution (in this case, 1.8 nm).
D.J.M., who was supported in part by grant SFB 189 from the Deutsche Forschungsgemeinschaft, wants to thank G.Büldt for his continuos encouragment. We are grateful to José María Valpuesta who helped with the image processing and the crystallization. This work was supported by the Swiss National Foundation for Scientific Research (grant 31‐42435.94 to A.E.), the Maurice E.Müller Foundation of Switzerland and the Spanish Ministry of Science and Education Grants PB91‐0109 (to J.L.C.) and PB93‐0278 from the Direccíon General de Política Científica.
- Copyright © 1997 European Molecular Biology Organization