Structure of keyhole limpet hemocyanin type 1 (KLH1) at 15 Å resolution by electron cryomicroscopy and angular reconstitution^†1

Abstract

A three-dimensional reconstruction of keyhole limpet hemocyanin type 1 (KLH1) has been obtained using electron cryomicroscopy at liquid helium temperatures and single particle image processing. The use of a high-contrast embedding medium, 1% (w/v) glucose and 2% (w/v) ammonium molybdate (pH 7.0), enables high-resolution electron micrographs to be recorded close to focus, i.e. with excellent transfer of high-resolution information, while maintaining enough image contrast to localise the individual macromolecules in the images. When low-pass filtered to ∼45 Å resolution, the new 15 Å resolution reconstruction is very similar to the earlier reconstructions of gastropodan hemocyanins of specimens embedded in vitreous ice. The map shows much detail and reveals many new symmetry elements in this very large cylindrical molluscan hemocyanin. The full KLH1 didecamer has D5 pointgroup symmetry, yet within the KLH1 decameric half-molecules local 2-fold axes have emerged that make the wall of the KLH1 decamer, in spite of its having an exact C5 symmetry only, resemble the D5-symmetric wall of the decameric cephalopod hemocyanins. In fact, the outside of each tier of this six-tiered gastropodan hemocyanin was found to have an approximate D5 symmetry. Local 2-fold axes also relate the “functional units” within the dimeric “morphological units” of the wall and the collar areas of the 8 MDa KLH1 molecule. Certain local-symmetry-related surface motifs may be present up to 60 times on the outside wall of this highly symmetric cylindrical hemocyanin. Keyhole limpet hemocyanin is used clinically as an immunostimulant. The very strong immune reaction elicited by this hemocyanin may be associated with its intricate hierarchy of local-symmetry components.

Introduction

The hemocyanins are large molecular mass, copper-containing, oxygen-transporting proteins found in the hemolymph of many molluscs and arthropods (for a review, see van Holde & Miller, 1995). Whereas the family of arthropod hemocyanins consists of assemblies containing between one and eight hexameric building blocks Markl and Decker 1992, van Heel and Dube 1994, the molluscan hemocyanins are organised as hollow cylindrical oligomers with external diameters of ∼370 Å. The polypeptide chain has a mass of 350 to 450 kDa and consists of a string of seven or eight functional substructures, sometimes termed functional domains. We prefer the term functional unit (FU), since the word domain possesses a more specific meaning when applied to the crystal structure of a protein. The hemocyanins of the cephalopod molluscs contain ten such polypeptide chains, always forming a single decameric cylinder of height ∼200 Å (mass 3.5 to 4 MDa), whereas gastropod molluscs contain hemocyanins built from two (or more) decamers. The “didecamer”, a cylinder with a total height of ∼400 Å and a mass of 7.5 to 8 MDa, is often the predominant hemocyanin species (e.g. KLH1), although longer multidecamers are sometimes encountered (e.g. KLH2). The hemocyanin from the giant keyhole limpet Megathura crenulata that we study here has undergone extensive investigation as a potent immunostimulant in mammals, including man. KLH is used, or is being investigated, as an immunotherapeutic agent (often as hapten carrier) for the treatment of certain cancers including melanoma (Helling et al., 1994), breast cancer (Longenecker et al., 1993), ovarian cancer (Yacyshyn et al., 1995), bladder cancer (Wishahi et al., 1995); KLH is used as a diagnostic tool for bilharziosis and as a hapten carrier for an AIDS vaccine (Naylor et al., 1991).

Molluscan hemocyanin chains generally have seven or eight functional units (i.e. oxygen-binding units), depending on the species (van Holde & Miller, 1995). The functional units are named a through g for the seven-FU species or a through h for the eight-FU species. The eight-FU gastropodan hemocyanins have an extra FU at the C-terminal side of the chain when compared with the cephalopod seven-FU subunits, such as that of Octopus dofleini. Cephalopod hemocyanin subunits with eight functional units, such as from Sepia officinalis, have an extra functional unit at the N-terminal side of the chain and their FU nomenclature is thus frame-shifted over a full functional unit with respect to the seven-FU cephalopod or the eight-FU gastropod hemocyanin polypeptides.

The structure of the large molecular mass molluscan hemocyanins has been studied by transmission electron microscopy since the introduction of the negative staining technique van Bruggen et al 1960, Fernandez-Moran et al 1966. The basic oligomeric molluscan hemocyanin assembly, including KLH, is a cylindrical three-tiered decamer Mellema and Klug 1972, Siezen and van Bruggen 1974, Miller et al 1990, Harris et al 1992, Lamy et al 1993, Lambert et al 1994. The three-tiered decamer in a cephalopod is thought to have a D5 (also known as 52) pointgroup symmetry (cf. Lambert et al., 1994), the most likely pointgroup symmetry for a homodecamer. In the case of gastropod hemocyanins, the top and bottom of the cylindrical homodecamer are not identical: the pointgroup symmetry of the decamer is thus C5 rather than D5. The gastropod hemocyanin decamers form a stable dimer, the didecamer, of mass ∼8 MDa with a D5 pointgroup symmetry (Mellema & Klug, 1972). At each end of the cylindrical didecamer a “collar” region has been defined within the inner wall of the outer tiers of the structure.

It was recently shown that the giant keyhole limpet M. crenulata contains two types of hemocyanin: an eight-FU KLH1 and a seven-FU KLH2 molecule Markl et al 1991, Gebauer et al 1994, Sohngen et al 1995, which can be separated biochemically Harris et al 1995a, Swerdlow et al 1996. Structural studies can now be pursued using purified KLH1 rather than using a mixture of the two types of hemocyanin Harris et al 1992, Harris et al 1993, Harris et al 1997, Dube et al 1995a. Biochemical and immunological studies on KLH1 and KLH2 Markl et al 1991, Gebauer et al 1994, Sohngen et al 1995, Sohngen et al 1997 have provided insights into the functional unit organisation of the polypeptides. The polypeptides have a molecular mass of ∼400 kDa for KLH1 and ∼350 kDa for KLH2, with a C-terminal h FU missing in KLH2. The data obtained correlate well with the immunomicroscopical studies on KLH2 using monoclonal antibodies of defined functional unit specificity (Harris et al., 1993). The collar of the KLH1 didecamer structure is thought to contain the C-terminal FUs (g and/or h) of the chains. In the case of Helix pomatia hemocyanin, the collar can be removed by limited trypsinolysis, leaving multi-FU fragments that spontaneously form hollow cylindrical tubes van Breemen et al 1975, van Breemen et al 1979 of diameter similar to that of the didecamers.

Recently, efforts have been made to produce three-dimensional (3D) reconstructions from unstained molluscan hemocyanin molecules embedded in vitreous ice Lambert et al 1994, Lambert et al 1995a, Lambert et al 1995b, Dube et al 1995a. At the 43 to 45 Å level of resolution obtained in these 3D reconstructions one can recognise the overall structural organisation of the hemocyanin and its main structural components: the wall, the collar and the arc. The internal arc structure is attached to the central tier of the decamers, but from these reconstructions Lambert et al 1995a, Lambert et al 1995b, Dube et al 1995a it is not yet clear whether, or to what extent, the arc and collar structures are directly connected in gastropodan hemocyanins. The reconstructions by Lambert et al 1995a, Lambert et al 1995bwere primarily calculated using the random conical tilt (RCT) technique, also known as the SECReT technique, developed by Radermacher et al. (1988). This technique, however, has some inherent limitations that make it difficult to achieve high-resolution levels van Heel et al 1992b, Schatz et al 1995.

The angular reconstitution approach van Heel 1987b, van Heel et al 1992b has recently been brought to the stage of almost-routine, applicable to individual macromolecules with arbitrary pointgroup symmetry (C4, Serysheva et al 1995, Orlova et al 1996a; D5, Dube et al., 1995a; D6, Schatz et al., 1995; C13, Tavares et al 1995, van Heel et al 1996). The extremes are the entirely asymmetric 70 S Escherichia coli ribosome Stark et al 1995, Stark et al 1997, on the one hand, and viruses with icosahedral symmetry (unpublished results) on the other. With the angular reconstitution approach, thousands of molecular images from particles randomly oriented in an embedding matrix are sorted and classified into “characteristic views” (van Heel & Stöffler-Meilicke, 1985) by multi-reference alignment techniques and multivariate statistical classification. The characteristic views are class averages into which a number of similar molecular images are averaged in order to reduce the noise inherent in low-dose electron micrographs. One can then assign Euler angle orientations to the individual characteristic views and calculate a 3D reconstruction. A first 3D reconstruction of the keyhole limpet hemocyanin embedded in vitreous ice was recently presented (Dube et al., 1995a) but the samples contained a mixture of KLH1 and KLH2 (although probably considerably more KLH1 than KLH2). Moreover, the sample yielded only a small number of useable molecular images (only ∼125 “side” views were processed), thus limiting the attainable resolution. The purpose of the current study is to improve the resolution of the KLH reconstruction, using only the purified KLH1 hemolymph component.

The first electron images of unstained hemocyanins (Stanley & Anderson, 1942; images reproduced by van Bruggen, 1986) revealed the average diameter of the assemblies. It was only with the introduction of the negative stain technique Brenner and Horne 1959, van Bruggen et al 1960, Fernandez-Moran et al 1966 that individual macromolecules could be visualised in the electron microscope with some structural detail. In its first decennium, then, the improvements of the specimen preparation techniques were aimed at improved visualisation of small details in the macromolecules. For the conventional uranyl acetate negative stain technique, optimising the conditions for direct visual interpretation of the images appears to have stimulated some unfavourable habits in specimen preparation: (a) the enhancement of strong preferential attachment of the protein on the carbon support film produces recognisable views but may deform the structure; (b) partial-depth, one-sided staining produces finer recognisable details in the molecular images than does a complete embedding into a matrix; and (c) partial dissociation of the macromolecules by the low pH stain and the strong interactions at carbon-film interface further enhance “fine details” in the structure. The properties of negative staining techniques have been investigated by several workers Kellenberger et al 1982, Cejka et al 1991, Cejka et al 1992, Bremer et al 1992, Harris and Horne 1994.

With the advent of image processing and 3D reconstructions in electron microscopy Crowther et al 1970a, DeRosier and Moore 1970, Mellema and Klug 1972 the structural preservation of the specimen has become a more important issue. In order to be able to apply a 3D reconstruction algorithm, which invariably relies on some form of the central section theorem Bracewell 1956, Bracewell 1960, DeRosier and Klug 1968, it was important to have the molecules fully and homogeneously surrounded by the embedding material. The introduction of glucose embedding Unwin and Henderson 1975, Henderson and Unwin 1975 and the vitreous-ice embedding technique (Dubochet et al., 1988), led to impressive 3D results. These homogeneous embedding media proved superior to the conventional negative staining techniques. Other carbohydrates such as trehalose (Jap & Walian, 1990) and tannin Akey and Edelstein 1983, Kuhlbrandt and Wang 1991 were also used successfully to image biological structures at high levels of resolution.

Nevertheless, the negative staining technique has advantages in terms of ease of specimen preparation with respect to the vitreous-ice embedding techniques, so it has continued to be used extensively, even if just for checking whether a sample is good enough for the vitreous-ice approach (Bremer et al., 1992). High-contrast embedding media have also been promoted as homogeneous environments for preserving 3D structures Stoops et al 1992, Ris and Allen 1975, Malecki and Ris 1993. It has been our philosophy for some years now Dube et al 1995b, Boettcher et al 1996, Harris et al 1995b, Harris et al 1997, that a high-contrast embedding medium facilitates the data processing as long as it does not interfere with the macromolecular structure and does indeed envelop the molecules completely (in “deep stain”). The idea is that the high contrast between protein and medium helps in finding the particle and is instrumental in attaining a good inter-particle alignment. The relatively high contrast between protein and surrounding, however, should preferably not contribute to the internal contrast within the particle. The internal density variations within the protein are expected to be mainly caused by, for example, the secondary structure of the protein. Such high-resolution information remains accessible only by averaging over a sufficiently large number of individual molecular images. A preliminary 3D reconstruction of the Lumbricus terrestris giant hemoglobin structure embedded in glucose/ammonium molybdate was calculated at the 15 Å resolution level (Dube et al., 1995b). We here apply these ideas for the first time to elucidate the 3D structure of a molluscan hemocyanin, KLH1. A preliminary report on this work has been presented (Orlova et al., 1996b).