Occurrence of Two Architectural Types of Hexagonal Bilayer

(S. Marco, personal communication) Hbs. In other words, type-II Hbs seems ..... ware: software of imagery and graphics for molecu- lar architecture. J. Struct. Biol.
5MB taille 0 téléchargements 236 vues
doi:10.1006/jmbi.2000.4344 available online at http://www.idealibrary.com on

J. Mol. Biol. (2001) 305, 757±771

Occurrence of Two Architectural Types of Hexagonal Bilayer Hemoglobin in Annelids: Comparison of 3D Reconstruction Volumes of Arenicola marina and Lumbricus terrestris Hemoglobins Ludovic Jouan1, Jean-Christophe Taveau1, Sergio Marco1 and FrancËois H. Lallier2 and Jean N. Lamy1* 1

Laboratoire des ProteÂines Complexes, Universite de Tours, 2 bis Boulevard TonnelleÂ, F-37032 Tours Cedex, France 2 Equipe Ecophysiologie UPMC, CNRS-INSU, Station Biologique, BP74, F-29682 Roscoff Cedex, France

Ê resolution of native hemoglobin of the polyA 3D reconstruction at 25 A chaete worm Arenicola marina was carried out from frozen-hydrated specimens examined in the electron microscope. The reconstruction volume of this large extracellular multimeric respiratory pigment appears as a hexagonal bilayer structure with eclipsed vertices in its upper and lower hexagonal layers. Conversely, in hemoglobins of oligochaetes, achaetes, and vestimentiferans and in chlorocruorins of the Sabellidae (polychaete) family, the vertices of the upper layer are 16  clockwise rotated with respect to those of the lower layer. The fact that two other polychaete hemoglobins (Alvinella pompejana and Tylorrhynchus heterochaetus) have the same architecture as Arenicola led us to de®ne two types of hexagonal bilayer hemoglobins/chlorocruorins: (i) type-I present in oligochaete, achaete, and vestimentiferan hemoglobins and in Sabellidae chlorocruorins; and (ii) type-II present in polychaete hemoglobins. A comparative study of the hemoglobins of Lumbricus terrestris (type-I) and Arenicola marina (type-II) showed that only two small differences located in the c4 and c5 linking units are responsible of the important architectural difference present in oligomers. A likely scheme proposed to explain the phylogenic distribution of the two types suggests that Clitellata, Sabellida (polychaete), and vestimentiferan hemoglobins and chlorocruorins derive from a type-I ancestral molecule, while Terebellida (Alvinella), Phyllodocida (Tylorrhynchus), and Scolecida (Arenicola) and possibly other polychaetes derive from an ancestor molecule with type-II hemoglobin. The architectures of the hollow globular substructures are highly similar in Arenicola and Lumbricus hemoglobins, with 12 globin chains and three linking units (c3a, c3b, and c4). The central piece of Arenicola hemoglobin is an ellipsoid while that of Lumbricus is a toroid. No phylogenic correlation could be found between the structure of the central pieces and the architecture type. # 2001 Academic Press

*Corresponding author

Keywords: hexagonal bilayer hemoglobin; annelids; cryoelectron microscopy; Arenicola marina: Lumbricus terrestris

Introduction Abbreviations used: 3D, three-dimensional; Chl, chlorocruorin; ESI-MS, electrospray ionization mass spectrometry; EMV, expected molecular volume; FSC, Fourier shell correlation; Hb, hemoglobin; HGS, hollow globular substructure; LMD, local maximum density; TMV, tobacco mosaic virus; SD, standard deviation. E-mail address of the corresponding author: [email protected] 0022-2836/01/040757±15 $35.00/0

Annelid and vestimentiferan hemoglobins (Hbs) are large extracellular multimeric proteins appearing in the electron microscope as hexagonal bilayer structures.1 ± 4 Their quaternary structures are still partially unknown, but different models have been proposed. The ``bracelet'' model proposed by the group of Vinogradov comprises 12 dodecamers of globin chains connected by 36-42 linker chains.5 # 2001 Academic Press

758 The second model, proposed by the group of Riggs, combines 12 hexadecameric subassemblies of globin chains and 24 linker chains.6 Beside these models designed on biochemical bases, in this and other laboratories several 3D reconstructions were carried out on frozen-hydrated specimens of various Hbs examined in the electron microscope. These works produced reconstruction volumes of oligochaete,7,8 achaete,9 polychaete,10 and vestimentiferan11 Hbs and of polychaete chlorocruorins (Chl)12,13 at resolutions ranging between 22 and Ê . All these volumes share a common architec50 A ture with a D6 point-group symmetry. The native molecule comprises two hexagonal layers made up each of six hollow globular subassemblies (HGSs) with local 3-fold symmetry linked to a central ellipsoidal or toroidal structure. Each HGS contains 12 globin chains disposed in four trimeric layers. Another character shared by all the reconstruction volumes except that of the polychaete worm Alvinella pompejana is the fact that the vertices of the upper hexagonal layers are 16  clockwise rotated compared to those of the lower layer (type-I). In Alvinella Hb, the vertices of the hexagonal layers are eclipsed (type-II). Since type-I Hbs and Chls were found in oligochaetes (Lumbricus terrestris), achaetes (Macrobdella decora), polychaetes (Eudistylia vancouverii, Sabella spallanzanii) and vestimentiferans (Riftia pachyptila), we ®rst thought that the type-II architecture observed in Alvinella Hb was induced by the adaptation to thermophilic habitat of this species.14 Recently, searching for structural particularities of annelid Hbs, we carried out a 3D reconstruction of the Hb of a mesophilic lugworm Arenicola marina (Am). This molecule was studied many years ago after negative staining15 and recently its subunit composition was determined in details by electrospray-ionization mass spectrometry (ESI-MS)16 leading to a surprising model with 12 globin chains located in the central piece of the molecule. Here, we describe the 3D reconstruction of AreniÊ resolution, we compare its cola marina Hb at 25 A architecture to that of Lumbricus terrestris Hb, and we draw general conclusions about the evolution of annelid and vestimentiferan hexagonal bilayer Hbs.

Results Nomenclature of the building blocks observed in annelid Hbs Annelid Hbs contain as many as 144 globin chains and 36-42 linker chains present as various polypeptide types. Thus, in Lumbricus terrestris Hb four globin chain types occur as disul®de-linked trimer (b-a-c) and monomer (d) and four linker chains (L1 through L4) occur as monomers. In Arenicola marina six globin chains (b1, b2, b3, c, d1, d2) are present as disul®de-bonded trimers (b, c, d) and two (a1 and a2) as monomers. The two linker chains (L1 and L2) occur as disul®de-bonded

Structural Types of Hb in Annelids

homodimer (L1-L1) and heterodimer (L1-L2). Besides this important complexity based on subunit diversity, we identi®ed in the 3D reconstruction volumes a number of different building blocks. Their nomenclature shortly described hereafter will be used for the description of the architectural differences between Arenicola and Lumbricus Hbs. The whole Hb can be arbitrarily decomposed into two parts: the HGS and the central piece. The HGS comprises 12 globular globin chains disposed in four parallel planes perpendicular to the local 3fold axis. Looking from outside along the 3-fold axis, one crosses a ®rst plane containing the centers of three globin subunits designated as 2A, 3A, 5A. The second plane contains globin chains 2B, 3B, 5B, the third plane 1A, 4A, 6A, and the fourth plane 1B, 4B, 6B. In projection along the 3-fold axis, subunits 2A and 2B, 3A and 3B, and 5A and 5B project radially, while subunits 1A and 1B, 4A and 4B, and 6A and 6B project on a circle. At resolÊ the two corresponding utions worse than 25 A subunits forming (i) the ®rst and second and (ii) the third and fourth layers are not resolved so that six elongated structures are observed instead of 12 globular masses. For example, globin chains 2A and 2B form a single elongated mass termed ``mass 2``. A detailed description of the HGS can be found in the work by Taveau et al.8 The cohesion of native and reassembled Hbs results from the presence of several types of links occurring between HGSs or between HGSs and the central piece. Initially, all these links were designated as ``connections''. However, increasing the resolution allowed recognition that certain connections are simple contact zones while others involve particular building blocks. Therefore, in this paper we improve our nomenclature and use the more precise terms of ``contact zones'' and ``linking units'' instead of the general term ``connection''. To prevent confusions, since identi®cation of the various linking units to the four linker chains is still highly speculative, we reserve the term linking unit to the description of the building blocks in the reconstruction volume and the term ``linker chain'' to the designation of the L1-L4 subunits identi®ed by biochemists. With these conventions, the Hb (Chl) molecule can be described as follows. First, each HGS is linked to its two neighbors of the same hexagonal layer (i) by c1 and c2 contact zones located on the external face of the molecule and (ii) by an inner bracelet of c3 linking units located on the luminal face of the Hb molecule. The interHGS c1 and c2 contact zones are located between globin chains (i) 2B and 3B and (ii) 4B and 6A, respectively. The c3 bracelet is clearly heterogeneous and composed of two elongated masses, the c3a and c3b linking units, bound to globin chains 1B and 4B, respectively. Secondly, each HGS of one hexagonal layer is linked to one HGS of the opposite layer by a third elongated structure bound to globin chain 6B, called c4 linking unit. Therefore, the three globin chains of the fourth

759

Structural Types of Hb in Annelids

plane (1B, 4B, 6B) of each HGS, all bind a linking unit (c3a, c3b, c4) that clearly corresponds to a distinct linker chain type or portion of linker chain type. Thirdly, the c3a, c3b, and c4 linking units of each HGS are linked in their middle to a rodlike structure, called ``c5 body'', that we interpret as an expansion of the central piece. The structure comprising the c5 body and the central piece very likely correspond to a second type of linker chain or portion of linker chain. This nomenclature is valuable for all the annelid and vestimentiferan Hbs so far described. We use it throughout the paper and in all the Figures. Electron microscopy In the electron micrograph ®eld, shown in Figure 1, displaying a frozen-hydrated specimen of Arenicola marina Hb, the molecules appear randomly oriented. Occasionally the two characteristic views previously observed with other annelid and vestimentiferan Hbs are found. These views are designated as ``top'' or ``hexagonal'' (white arrow), ``side'' or ``rectangular'' (black arrows), and intermediate (broken arrow). Due to the low contrast resulting from the use of ice as contrasting agent and the low defocus, structural details are hardly observed in these images. However, in the average images shown in the insets of Figure 1(b) to (d) obtained after image classi®cation, the well-known features are easily observed. Thus, in the hexagonal (Figure 1(b)) and rectangular (Figure 1(c)) views the molecule is seen along its 6-fold and 2fold axes, respectively. The view shown in Figure 1(d) is in an intermediate orientation corresponding approximately to a 45  rotation around one horizontal axis of the molecule oriented as in Figure 1(b). The long rods visible on the left side

of the picture are tobacco mosaic virus (TMV) molecules used for calibrating the magni®cation. 3D reconstruction A ®rst series of 24 negatives containing 1513 pairs of tilted and untilted-specimen images was digitized, windowed, contrast inverted, and normalized. After two-dimensional (2D) alignment of the untilted-specimen images, homogeneous image classes were sorted according to different classi®cation methods: (i) correspondence analysis and hierarchical ascendant classi®cation; and (ii) Kohonen self-organizing map neuronal network. In the particular case of Arenicola marina Hb, the two methods gave approximately the same results and correspondence analysis was used for the reconstruction volume described here. Three homogeneous classes of untilted-specimen images of rectangular views (containing 74, 49, and 96 images) and one class of hexagonal view (37 images) were selected and the corresponding tilted-specimen images were submitted to 3D reconstruction according to the random conical tilt series method17,18 with an enforced D6 point-group symmetry. The four resulting primary volumes were aligned in real space in a common orientation and a merged volume was calculated and re®ned by two cycles of 3D projection alignment. Then, the whole set of 1513 tilted-specimen images was added to the reconstruction process and four new re®nement cycles were carried out eventually producing the reference volume used thereafter. In a second step, Eulerian angles were assigned to a second series of 2530 untilted-specimen images collected at low defocus ( 0.8 mm) by a single cycle of 3D projection alignment using the reference volume obtained above. As judged by the

Figure 1. A frozen-hydrated sample of Arenicola marina Hb observed at 0  tilt angle. (a) A ®eld showing the three main characteristic orientations of the Hb. Black, white, and broken arrows point to rectangular, hexagonal, and intermediate views, respectively. (b)-(d) Average images of homogeneous clusters of hexagonal (top), rectangular (side), and intermediate views, respectively. The scale bars represent 60 nm.

760 topology sphere method, the directions of projection used in the reconstruction process were evenly distributed in the three dimensions of space, thus preventing noticeable deformation of the re®ned 3D reconstruction volume. After six additional re®nement cycles, the resolution of the volume was Ê according to the FSC method estimated to 24.8 A Ê according to modi®ed by Orlova et al.19 and 25.3 A the FSC method thresholded at 0.5 according to Penczek.20 Finally, the volume was subjected to a Ê limit. low-pass Fermi ®ltration down to the 25 A Cladistic exploration A cladistic analysis of the 3D reconstruction volume of Arenicola marina Hb was calculated by analyzing in a tree called cladogram the variations of the local maximal densities (LMDs) as a function of the surface representation threshold. The highest and lowest threshold values, rescaled to 1 and 0, display the LMD with the highest density of the reconstruction volume and the expected molecular volume (EMV), respectively. As shown in Figure 2(a), the cladogram comprises 160 leaves corresponding each to one LMD. Cutting the tree at different thresholds produces different surface representations. For example, the broken line (level 1) indicates the threshold corresponding to Figure 2(c) and (d) and the dotted line the threshold corresponding to Figure 3(b) and (d). At this stage, the cladogram is a very complex structure. However, starting from the root and following the large branches, one easily identi®es a small group (clade I), with eight densities located in the left part of the diagram and six very similar complex groups (clades IIa through IIf) of densities forming the rest of the structure (152 LMDs). This simple analysis allowed the design of the simpli®ed version of the cladogram shown in Figure 2(b). In the rest of the investigation, clades I and IIa through IIf were visualized and analyzed. For example, thresholding the cladogram at level 1 (broken line in Figure 2(a)) allowed an automatic separation of the two clades corresponding to the central piece (Figure 2(c)) and the rest of the molecule (Figure 2(d)). In addition, since the procedure can be used recurrently, in the surface shown in Figure 2(d) clades IIa through IIc were colored in dark gray and clades IId through IIf in light gray. Figure 3(a) shows an enlarged representation of the branch corresponding to clade I. This substructure comprises eight densities, six of which with a high value and two with a much lower value. When the volume is visualized at the level-2 threshold (dotted line in Figure 2(a)), one obtains the surface representation shown in Figure 3(b). The identi®cation of the eight densities derived from the eight LMDs is simple. The six high densities correspond to the six structures labeled ea radiating in the equatorial plane of the molecule. The two small densities labeled eb are centered on the 6-fold axis below and above the equatorial plane. At the level-1 threshold (broken line in

Structural Types of Hb in Annelids

Figure 2(a)), the eight densities are included in a single ellipsoidal structure (Figure 2(c)). When examining in detail the six clades labeled IIa through IIf in Figure 2, one observes that they each comprise 25 to 26 LMDs. As shown in Figure 3(c), based on the value of their highest densities, the 26 branches can be separated into two groups. In the ®rst group, six LMDs with low-density values correspond to two c3a, two c3b, and two c4 linking units. Of these six densities, only the two c4 linking units are visible in the center of the surface representation of Figure 3(d). Indeed, the c3a and c3b are hidden behind mass 4. The 20 LMDs with high-density values in Figure 3(c), i.e. above the broken line, all correspond to globin chains. They are labeled as indicated in the abovedescribed nomenclature. As shown in the cladogram, the LMD pairs corresponding to globin chains 1A and 1B, 3A and 3B, 4A and 4B, and 5A and 5B fuse at threshold slightly smaller than their apparition threshold into masses 1, 3, 4, and 5, respectively. Conversely, at the resolution reached Ê ), masses 2 and 6 are in this reconstruction (25 A not resolved into two globin chains and the locations of the LMDs 2 and 6 are exactly those of globin chains 2A and 6A in Lumbricus Hb. The absence of LMDs for globin chains 2B and 6B only means that at the threshold increment used in this experiment the LMD and at least one of its 26 neighboring voxels disappear at the same time, so that the LMD is never surrounded by 26 voxels with smaller densities. The last interesting aspect of Figure 3(d) is the distribution of the 20 globin chains into the various HGSs. Indeed, at the threshold used for the surface representation, because of the 2-fold symmetry ten globin chains forming six masses are located in each layer. However, in each layer, the six masses do not belong to the same HGS. As shown in Figure 3(d) masses 1 and 2 belong to one HGS (broken circle) and masses 3, 4, 5, and 6 to the neighboring HGS (full circle). This particular pattern in the cladogram means that a closer contact occurs between mass 2 (actually 2B) of one HGS and mass 3 (actually 3B) of the neighboring HGS, i.e. in the c1 contact zone, than between masses 2 and 3 of a given HGS. Comparison of the reconstruction volumes of Arenicola marina and Lumbricus terrestris Surface representation of the native molecules A rapid examination of the 3D volume of Arenicola marina Hb showed important differences with the recently published volume of Lumbricus terrestris Hb.8 In this section, rather than to describe in detail the volume of Arenicola Hb, a task that would be in large part redundant with our previous description of Lumbricus Hb, we prefer to present a comparative study of the two volumes that lead us to propose two different architectural types for annelid Hbs. The surface representations

Structural Types of Hb in Annelids

761

Figure 2. Cladistic analysis of the 3D reconstruction volume of Arenicola marina Hb. (a) The complete and (b) simpli®ed versions of the cladogram. (c) and (d) The two clades obtained by thresholding the 3D volume at the level indicated in (a) by the broken line. The dotted line in (a) is the threshold corresponding to the clades shown in Figure 3.

displaying the EMV of Arenicola and Lumbricus Hbs are shown in Figures 4 and 5. In Figure 4(a) and (b) the molecules observed along their 6-fold axis are displayed in such a way that the upper hexagonal layers have exactly the same orientations with the broken line joining opposite vertices vertically oriented. In this disposition the 2fold axis of type 1 (black ellipse) is 0.5  clockwise rotated with respect to the vertical broken line in Arenicola Hb, and 8  anticlockwise rotated in Lumbricus Hb. As a result of the 2-fold symmetry, the vertices of the lower layer are 1  clockwise and

16  anticlockwise rotated relatively to those of the upper layer in Arenicola and Lumbricus Hbs, respectively. We de®ne this difference in the positions of the vertices of the two hexagonal layers as type-I (Lumbricus) and type-II (Arenicola) architecture. Starting from the molecules oriented as in Figure 4(a) and (b), 45 rotations around horizontal axes produced Figure 4(c) and (d) called 45  views. Notice the relative locations of the HGSs in the two layers resulting from the different positions of the 2-fold axes of type 1 described above. Another important feature is the shape of the c5

762

Structural Types of Hb in Annelids

Figure 3. Detailed cladistic analyses ((a) and (c)) and surface representations ((b) and (d)) of the main two subgroups of LMDs present in the cladogram. (a), (b) Clade I; (c), (d) clade IIf. The surface representation of the central piece shown in (b) has been obtained at the level-2 threshold indicated by the dotted line in Figure 2(a). The surface representation of the globin chains shown in (d) corresponds to clade IIf of Figure 2(b). In (d) the hatched zone in the inset shows the intramolecular location of clade IIf. Full and broken line circles circumscribe subunits belonging to different HGSs. In (a) and (b) ea and eb designate the densities located in the periphery and in the center of the central piece. In (c) c3a, c3b, and c4 designate the linking units formerly called connections. Masses 1A, 1B, 2A, 2B, etc. designate the positions of the 12 globin chains located in the HGS. The broken line separates the densities of the globin chains from those of the linking units. In (d) c4 refers to the c4 linking unit and 1 through 6 to globin masses. For example mass 1 corresponds to globin chains 1A ‡ 1B.

body that goes straight from the central piece to the HGS in Arenicola (arrow) and is bent to the right in Lumbricus. This curvature is also an aspect of the respective rotations of the two hexagonal layers. Removing the front halves allows observation of the internal structures. In Figure 4(e) and (f) two important differences are visible. First, a hole is present in the central piece in Lumbricus Hb and absent from Arenicola Hb. Second, in the cutting plane only the section of c4 linking unit is visible in Arenicola, while in Lumbricus the lower part of globin chains 6B appears on both sides of the c4 linking unit, producing a larger section. This difference in the section sizes (hatched zones in Figure 4(e) and (f)) means that the distance between the hexagonal layers is slightly larger in Arenicola than in Lumbricus. The side views of Figure 5 show the same differences, but in perpendicular directions. Due to the rotation of the hexa-

gonal layers, in Figure 5(a) and (b) the line passing through the black ellipses enclosing the positions of globin masses 5 (5A ‡ 5B globin chains) are inclined in Lumbricus Hb and vertical in Arenicola Hb. Another consequence of this rotation is the different shape of the holes (bordered by a black line in Figure 5(c) and (d)) through which the 2fold axes of type 2 (white ellipses) pass. Because of the eclipsed disposition of the two layers, the hole is large in Arenicola Hb (Figure 5(c)) while the 16  rotation strongly reduces its size in Lumbricus Hb (Figure 5(d)). Finally, removing the front halves of the reconstruction volumes allows observation of the central piece: an ellipsoid in Arenicola Hb (Figure 5(e)), a toroid in Lumbricus Hb (Figure 5(f)). The distribution of the high densities in clade I (Figure 3(a) and (b)) explains satisfactorily the ellipsoidal shape of the central piece in Arenicola Hb.

Structural Types of Hb in Annelids

763

Figure 4. Surface representation at a threshold displaying the EMV of the native Hbs of Arenicola marina ((a), (c) and (e)) and Lumbricus terrestris ((b), (d) and (f)). In (a) and (b) the molecule is viewed along the 6-fold axis. In (c) and (d) the molecules initially oriented as in (a) and (b) have been rotated 45  around a horizontal axis. In (e) and (f) the front halves of the molecules oriented as in (a) and (b) have been removed. The black hexagon, triangle, ellipse and the white ellipse indicate the 6-fold, 3-fold, type-1 and type-2 2-fold axes. For the nomenclature of the globin chains (1A, 1B through 6A, 6B), contact zones (c1 and c2) and linking units (c3a, c3b, c4, c5) see the ®rst part of Results.

HGS architecture Both species exhibit exactly the same HGS architecture. The only minor difference due to the lower resolution is that the globin chains composing the six masses are not completely resolved in Arenicola Hb. Figure 6(a) and (b) shows one HGS of each species viewed perpendicularly to its 3-fold axes at two different thresholds. Obviously, globin masses have the same locations in both species. Figure 6(a) to (d) also shows that the respective positions of the c4 linking units and c5 body are perfectly similar. In addition, as shown in Figure 6(c) and (d), in which HGSs are viewed along their 3-fold axis from the central piece, the c3a, c3b, and c4 linking units bind masses 1B, 4B, and 6B exactly in the same locations in the two species. As shown in

Figure 6(e) through (h), the positions of the globin chains in the four planes perpendicular to the 3fold axes are also remarkably similar. The small spheres (dark for Arenicola and gray for Lumbricus) materializing the LMDs almost perfectly overlap. The only cases where the overlapping is imperfect Ê resolution are for masses 2 and 6 because at 25 A these masses are not resolved into 2A and 2B or 6A and 6B in Arenicola Hb. It results that in Arenicola the LMDs corresponding to masses 2 and 6 are located almost exactly where globin chains 2A and 6A are located in Lumbricus Hb. However, their elongated shapes clearly indicate that each of them contains two globin chains. Thus, despite the difference in the resolution of the two volumes, one can conclude that HGSs have the same architecture in both species and the reader is referred to

764

Structural Types of Hb in Annelids

Figure 5. Surface representation at a threshold displaying the EMV of the native Hbs of Arenicola marina ((a), (c) and (e)) and Lumbricus terrestris ((b), (d) and (f)). In (a) and (c) the molecules are viewed along the type-1 2-fold axis. In (c) and (d) the molecules initially oriented as in (a) and (b) have been rotated 30  around vertical axes and are now viewed along their type-2 2fold axis. In (e) and (f) the front halves of the molecules oriented as in (a) and (b) have been removed. For the nomenclature of the building blocks, see the ®rst part of Results. The black and white ellipses indicate the type-1 and type-2 2-fold axes.

our recent paper on the structure of Lumbricus Hb for a more detailed description of HGS structure.8 Structures of the c4 linking units and central pieces Since type-I (Lumbricus) and type-II (Arenicola) Hbs occur in molecules having similar HGSs and hexagonal layers, the origin of the architectural types must be searched in the interlayer contacts between c4 linking units and in the links between the HGSs and the central piece. Figure 7(a) and (b) shows masses 6 (globin chains 6A and 6B) and the c4 linking units of the two hexagonal layers at a threshold displaying only a fraction of EMV. In Arenicola Hb (Figure 7(a)), the two c4 linking units are almost superposed and come into contact in their middle. In Lumbricus Hb (Figure 7(b)), the c4 bodies are more distant from each other and their closest distance occurs near at their ends. Consequently, the distance between the centers of the

two globin chains 6B is 1.6 times greater in Lumbricus than in Arenicola. As discussed below, this difference in the contact between the c4 linking units may induce the rotational difference between hexagonal layers producing types I and II. The c5 bodies also are different in the type-I and type-II Hbs. As shown in Figure 7(e) and (f), the c5 body is rod-shaped in Arenicola (Figure 7(e)) but strongly incurved in Lumbricus Hb (Figure 7(f)). The central pieces of Arenicola and Lumbricus Hbs exhibit differences and resemblances. The major difference, obvious in Figures 4(e) and (f) and 5(e) and (f), is the presence of an ellipsoid in Arenicola and a toroid in Lumbricus. The major resemblance refers to the positions and orientations of the high densities located in the central piece and at the beginning of the c5 bodies (Figure 7(c) and (d)). Our current view of the architecture of both Lumbricus and Arenicola Hbs drawn from

Structural Types of Hb in Annelids

765

Figure 6. Comparison of HGS architecture in Arenicola ((a) and (c)) and Lumbricus ((b) and (d)) Hbs. One HGS of each species extracted from the 3D volume with the help of the cladistic analysis is viewed perpendicularly to its 3fold axis in (a) and (b). In (c) and (d) the HGSs are viewed from the center of the volume along the 3fold axis. In (a) to (d), the surface in wireframe displays the EMV and the ®lled bodies a smaller portion of the molecular volume. Isocontour lines shown in (e) through (g) correspond to four one-voxel thick slices cut through the HGS of Arenicola (large lines) and Lumbricus Hb (narrow lines) perpendicularly to the 3-fold axis. The slices shown in (e) and (h) correspond to the top and bottom of the HGS, respectively. The small dark (Arenicola) and gray (Lumbricus) spheres indicate the positions of the various LMDs. For the nomenclature of the building blocks see the ®rst part of Results.

cryoelectron microscopy 3D reconstructions is summarized in Figure 8.

Discussion Phylogenic distribution of type-I and type-II Hbs in annelids and vestimentiferans The identi®cation of a type-I Hb molecule relies on the presence of a clockwise 16  rotation angle of the upper hexagonal layer vertices with respect to those of the lower layer. To date such structures have been identi®ed without ambiguity in the Hbs of Lumbricus terrestris,21 Macrobdella decora,9 and

Riftia pachyptila11 and in the Chls of Eudistylia vancouverii12 and Sabella spallanzanii.13 Type-II has been observed in Alvinella pompejana,10 Arenicola marina (this paper), and Tylorrhynchus heterochaetus (S. Marco, personal communication) Hbs. In other words, type-II Hbs seems circumscribed to the polychaete class, while type-I is present in oligochaete, achaete, and vestimentiferan Hbs and in polychaete Chls. The phylogenic relationship between the various families of annelids is still a matter of controversy. The latest attempt to classify Annelida as a whole is based on cladistic analysis of a large number of larval and adult anatomical characters.22 In this

766

Structural Types of Hb in Annelids

Figure 7. Comparison of the c4 linking unit, c5 body, and central piece in Arenicola ((a), (c) and (e)) and Lumbricus ((b), (d) and (f)) Hbs. (a) and (b) Surface representations of two clades corresponding to the globin masses 6 (globin chains 6A ‡ 6B) and c4 linking units present in the two opposite hexagonal layers and viewed along the 2-fold axis of type 2. (c) and (d) The central piece viewed along the 6-fold axis. Notice the difference in the orientations of the boxed c5 bodies starting from each vertex. (e) and (f) Orientation of the two c5 bodies issued from a vertex of the central piece and viewed along the 2-fold axis.

paper, compared to the ®rmly established monophyly of the Clitellata (Lumbricus, Macrobdella, HBL Hb type-I), that of the Polychaeta is questioned. Interestingly, Rouse & Fauchald22 consider Vestimentifera (and their sister group Pogonophora) to be part of a monophyletic taxon of polychaetes, the Sabellida, comprising the family Siboglinidae (Riftia, type-I) close to the Sabellidae (Eudistylia, Sabella, type-I). Other polychaetes families are grouped in different taxa, e.g. Terebellida (Alvinella, type-II), Phyllodocida (Tylorrhynchus, type-II), and Scolecida (Arenicola, type-II). A very recent study, based on mitochondrial genome analysis, con®rms the inclusion of Pogonophora (s.l.) within Annelida, but closer to Clitellata than to polychaetes.23 However, their analysis is based on a single Nereididae species (Polycheta, Phyllodocida), a clearly insuf®cient sample considering Rouse & Fauchald22 phylogeny. A likely scheme proposed to explain the phylogenic distribution of the two types suggests that Clitellata, Sabellida (polychaete), and vestimentiferan Hbs and Chls derive from a type-I

ancestral molecule, while Terebellida (Alvinella), Phyllodocida (Tylorrhynchus), and Scolecida (Arenicola) and possibly other polychaetes derive from an ancestor molecule with type-II Hb. To sustain this proposition, further data on other Sabellida, and on Spionida and Eunicida are needed. In addition, knowledge of the architectural type of Ampharetidae (Terebellida), Flabelligeridae (Sabellida), and Serpulidae (Sabellida) Chls would be particularly interesting (Table 1). The distribution of ellipsoidal and toroidal central pieces seems to follow a different rational. Indeed, on the one hand, to date toroids have been reported in native molecules in the main three annelid classes with Lumbricus terrestris,7,21 Macrobdella decora,9 Alvinella pompejana,10 Ophelia vestimentiferans Riftia bicornis24,25 and in pachyptila,11 while ellipsoids only occur in polychaete Hbs with Arenicola marina (this paper), Tylorrhynchus heterochaetus (S. Marco, unpublished results) and in Eudistylia vancouverii12 and Sabella spallanzanii13 Chl. Thus, the fact that ellipsoids

767

Structural Types of Hb in Annelids

native structure has a toroid. Third, the presence of toroids and ellipsoids seems inconstant for a given Hb at least in negatively stained specimens of Lumbricus terrestris,3 Eophila tellinii,26 and Ophelia bicornis.27 Fourth, as visible in early work of Van Bruggen & Weber,28 negatively stained Arenicola Hb does not contain the central mass inducing the ellipsoidal structure. These arguments indicate that toroids and ellipsoids are not distributed in Hbs and Chls on a simple phylogenic basis. If evolution is involved, conformational change of one structure into the other of unknown origin may seriously complicate the problem. Agreement of the 3D reconstruction volume of Arenicola Hb with previous model Figure 8. Schematic representation of annelid hexagonal bilayer Hbs. The whole molecule is composed of six copies of the unit displayed comprising two HGSs. Each HGS comprises 12 globin chains and three linking units (c3a, c3b, and c4). The connection between the linking units and the trimers of globin chain are established (1B-c3a, 4B-c3b, 6B-c4). The c4-c4 contact can occur in two different positions: side-to-side leading to type-II Hbs or end-to-end leading to type-I Hbs/Chls. The three linking units of each HGS are connected to the c5 linking unit.

have been found in native Hbs and Chls exclusively in polychaetes would normally suggest that this structure appeared in the polychaete group. There are, however, several dif®culties to such a generalization. First, Hb of the hyperthermophilic polychaete Alvinella pompejana possesses a toroid. Second, the defective Hb reassembled from puri®ed trimers of globin chains and the complete set of linker chains puri®ed from Lumbricus Hb (HBL[T ‡ L]) possesses an ellipsoid21 while the

Zal et al.16 determined the polypeptide chains composition of Am Hb from ESI-MS studies. They found eight globin chain types called a1, a2, b1, b2, b3, c, d1, and d2 of Mm ranging between 15921 and 17033 Da and two linker chains (L1: 25174 Da and L2: 26830 Da). These authors found that globin chains b, c, and d are arranged in trimers and that linker chains occur as homodimer (L1-L1) and heterodimer (L1-L2). They proposed a model comprising 156 globin and 42 linker chains arranged in 12 protomers containing each nine monomers and one trimer of globin chains (a13,a26,bcd). The 156 144 ˆ 12 remaining globin chains were assigned to the central piece. This model disagrees with the present 3D reconstruction volume. Indeed, if the six elongated masses present in HGSs merely account for 12 globin chains, the accommodation of 12 globin chains in the central piece is in con¯ict with the small available volume and the density distribution in the central piece. At the resolution reached by this reconstruction, we cannot con®rm or disagree with the model of Zal et al.16 about the distribution of the trimers and monomers of globin chains within the HGS, but our results do not contradict the model. About lin-

Table 1. Classi®cation and architectural type of hemoglobin (chlorocruorin) in annelids Classification Clitellata Polychaeta

Oligochaeta Hirudinea Sabellida

Terebellida Spionida Phyllodocida Eunicida Scolecida

Sabellidae

Species

Transporter

Type

Lumbricus Macrobdella Sabella, Eudistylia

Hb Hb Chl

I I I

Chl Hb Chl Chl

I I ND ND

Hb Chl/Hb Hb Hb Hb Hb

II ND ND II ND II

Siboglinidae Flabelligeridae Serpulidae

Riftia

Alvinellidae Ampharetidae

Alvinella

Nereididae

Tylorrhynchus

Arenicolidae

Arenicola

The broken line separates the type-I and type-II Hbs/Chls. ND, not determined.

768 ker chains, on the one hand the experimental values given by Zal et al.16 are: 26.5(8.5) copies of L1 and 4.6(2.05) copies of L2. On the other hand, Royer et al. (see note added in proof) recently demonstrated by X-ray crystallography that the total number of linker chains is 3  12 ˆ 36. In addition, we know (i) from the 3D reconstructions carried out from both cryoelectron microscopy and X-ray crystallography that the copy numbers of linker chains must be multiple of 6 and (ii) from Zal et al.16 and that all the linker chains must occur either as disul®de-bonded homodimers (L1-L1) or heterodimers (L1-L2). If we admit that all these conditions must be applied to each molecule, then there is only one solution compatible with ESI-MS experimental data: 12 L1-L1 homodimers and 6 L1-L2 heterodimers, making a total of 30 L1 and 6 L2 linker chains, i.e. 36 linker chains.

Conclusion The 3D reconstruction of the Arenicola marina Hb allows the de®nition of a second and new type of architecture for annelid and vestimentiferan Hbs and Chls. The ®rst type (type-I), in which the vertices of the upper hexagonal layer are 16  clockwise rotated with respect to those of the lower layer, occurs in oligochaete, achaete, and vestimentiferan Hbs and in the Chl of the Sabellidae (polychaete) family. The second type (type-II), with two eclipsed hexagonal layers, has been found to date exclusively in polychaete Hbs. Two small structural differences located in the c4 and c5 connections of the linker complex seem to cause the 16  rotation characteristic of type-I Hbs and Chls. The HGSs corresponding to types I and II look highly Ê ressimilar and no difference is perceptible at 25 A olution. The distribution of types I and II in annelid and vestimentiferan Hbs and Chls is clearly dependent on evolution. The facts that Chls of the Sabellidae family belong to type-I like oligochaete, achaete, and vestimentiferan Hbs and not to typeII like other polychaete Hbs suggest that type-II Hbs appeared early in polychaetes and that Chl evolved from a preserve type-I Hb. The central pieces are different in Lumbricus (toroid) and Arenicola (ellipsoid) Hbs. However, this difference does not seem linked to evolution.

Materials and Methods Hemoglobin The worms were captured at low tide on a beach near Roscoff (France) and kept in local running water for 24 hours. Blood samples were withdrawn directly from the ventral or dorsal vessel in the presence of phenylmethylsulfonyl ¯uoride (PMSF) 1 mM. Then, the blood was freed from insoluble material by centrifugation at 10,000 g for ten minutes at 4  C. After puri®cation by gel-®ltration on a Superose 6TM column (Amersham Pharmacia Biotechnology, Uppsala, Sweden) as previously described,16 Hb was frozen and stored in liquid nitrogen. The ®nal protein concentration, estimated by

Structural Types of Hb in Annelids spectrophotometry using the following formula: Protein concentration …g=l† ˆ 1:55 A280

0:76 A260

with A280 and A260 corresponding to the absorbance at 280 and 260 nm, respectively, was 6.55 g/l. The expected molecular volume (EMV) was calculated from the Mm value of 3.648 MDa determined in ESIMS16 and the partial speci®c volume value of 0.73 cm3 g 1. This latter value was estimated for Lumbricus from the partial speci®c volumes of the globin (0.738 cm3 g 1) and linker (0.722 cm3 g 1) chains assuming a weight ratio of 70/30.29 The EMV value calculated with these Ê 3. In our 3D reconstruction assumptions was 8.29  106 A volume of Arenicola marina Hb, rescaled between 0 and 1, it corresponds to a threshold value of 0.66. Electron microscopy A sample of Arenicola marina Hb (0.9825 mg/ml) was prepared by dilution in Tris-HCl buffer (pH 7.2), 10 mM CaCl2, 10 mM MgCl2 and applied on 400mesh copper grids coated with a holey carbon ®lm and thin carbon ®lm. After blotting the excess solution, the grid was quickly plunged into liquid ethane.30 ± 32 The samples were loaded using the Gatan 626 cryotransfer system into a Philips CM12 electron microscope equipped with a Gatan CCD camera and a Gatan 651N anticontaminator. Frozen-hydrated specimens maintained at 173  C were observed under low-dose conditions at accelerating voltage of 100 kV, 0.8 mm defocus, and magni®cations in the range of 43,478 to 43,820 nm. The condenser and objective apertures were 200 nm and 100 nm, respectively. The images recorded on SO163 Kodak ®lms were developed as described for 12 minutes in full-strength Kodak D19 developer. During the observation, the grid was tilted by a 45  angle or untilted. Magni®Ê layer lines of cations were calculated from the 23 A tobacco mosaic virus (TMV) included in the sample.33 Image processing Micrographs were digitized using an Optronics P1000 microdensitometer, with a spot size and a scanning step of 25 mm corresponding to pixel sizes in the range 5.70 Ê . Images of single particles, with a background to 5.75 A free from contamination artifacts and overlapping molecules were selected and windowed on a Silicon Graphics workstation, using an interactive selection program. Then, they were contrast inverted (to render the protein material white and the background dark) and normalized so that their noise statistics would match a reference distribution corresponding to a portion of micrograph free of particles and artifacts. All programs used for image processing are part of SPIDER34 and SIGMA35 software. Three-dimensional reconstruction Alignment and classification The 3D reconstruction was carried out according to the method of the random conical tilt series17,18 adapted for cryoelectron microscopy.36,37 All the images were subjected to a reference-free bidimensional alignment. Then, the complete set of untilted-specimen images was sorted by correspondence analysis38 followed by hierarchical ascendant classi®cation38,39 using Ward's

769

Structural Types of Hb in Annelids merging criterion.40 Alternatively, Kohonen's self organization map, adapted for cryoelectron microscopy by Marabini & Carazo,41 was used for the classi®cation step. 3D reconstruction After determining the Eulerian angles corresponding to each image, primary 3D reconstructions using the images of homogeneous classes corresponding to the hexagonal (top) and rectangular (side) views were carried out according to the simultaneous iterative reconstruction technique (SIRT) algorithm modi®ed for electron microscopy images36 with an enforced D6 point-group symmetry. The corresponding primary volumes were aligned in common orientation and a ``merged'' volume was calculated and re®ned during six cycles of 3D-projection alignment.37 Then, a series of untilted-specimen images, whose Eulerian angles were determined by 3D-projection alignment of the re®ned merged reconstruction volume, was introduced in the reconstruction process. Eventually, the resulting reconstruction volume was re®ned with six cycles of 3D projection alignment. Resolution For the resolution limit estimation of the 3D reconstruction volumes, two subsets were randomly drawn from the whole image set. The corresponding 3D volumes were then compared in reciprocal space on spherical shells with increasing radii using the Fourier shell correlation (FSC) criterion42,43 according to the method described by Orlova et al.19 In this method, the number of symmetries for the calculation of the noise curve is taken into account, and the resolution is calculated from the value of the crossing point between the FSC curve and the random noise curve at a threshold of 3.0 SD. The resolution may also be estimated by thresholding the experimental FSC curve at a value of 0.5 corresponding to signal-to-noise ratio of 1.20 Computer graphics The exploration of the 3D reconstruction volume (surface representation and slicing) was done with the SIGMA software.35 The analysis of the densities composing the 3D reconstruction volume of Arenicola marina Hb was also carried out with a recently devised method based on a tree-like representation of the volume, called ``cladistic representation''.8 The method describes the evolution of LMDs, (i.e. voxels entirely surrounded by voxels with lower densities), when the threshold decreases from the value displaying the highest density of the volume to the value displaying the EMV. The program draws a tree in which (i) leaves correspond to LMDs, (ii) branches to the growth of density when the threshold decreases, (iii) nodes to the fusion of the regions derived from different LMDs and (iv) root to the whole reconstruction volume. The tree, or ``cladogram'' is drawn with the package PHYLIP44 and a 3D graphic tool allows interactive calculation and visualization of the surface corresponding to any branch. Note added in proof During the evaluation of this paper, a 3D reconÊ resolution struction of Lumbricus terrestris Hb at 5.5 A

by W. E. Royer Jr et al. (Proc. Natl Acad. Sci. USA, 97, 7107-7111) was published. Despite that the crystals were available for more than ten years, the phase problem remained unsolved until 1997. At that time, attending a conference in Asilomar (CA, USA), where one of us (J.L.) presented a 3D reconstruction volume Ê resolution of Lumbricus terrestris Hb computed at 22 A from cryoelectron microscopy images, W. E. Royer had the idea to use a similar volume previously computed (by Shatz et al. (1995). J. Struct. Biol. 114, 28-40) to solve the phases (W.E. Royer Jr, personal communication). The experiment was perfectly successful and con®rmed all our previous observations, in particular the disposition of the globin chains in HGSs and the linker nature of our c3a, c3b and c4 linking units. The only information, that was never suspected from cryoelectron microscopy volumes, is that the c5 connection and the central piece are actually built from coiled-coils comprised of the three linker chains. One could have expected that the resolution of the 3D reconstruction volumes calculated from randomly oriented particles examined in cryoelectron microscopy that had strongly improved in the 1990s would have reached atomic resolutions in the coming years. In fact, unexpectedly, the excellent results produced by this method helped X-ray crystallography to solve one of its last challenges. Therefore, it is highly probable that in the future this phenomenon will repeat.

References 1. Antonini, M. & Chiancone, M. (1977). Assembly of multisubunit respiratory proteins. Annu. Rev. Biophys. Bioeng. 6, 239-271. 2. Vinogradov, S. N., Kapp, O. H. & Ohtsuki, L. (1982). The extracellular hemoglobins and chlorocruorins of annelids. In Electron Microscopy of Proteins (Harris, J., ed.), vol. 3, pp. 135-163, Academic Press, New York. 3. Boekema, E. J. & Van Heel, M. (1989). Molecular shape of Lumbricus terrestris erythrocruorin studied by electron microscopy and image analysis. Biochim. Biophys. Acta, 957, 370-379. 4. Lamy, J. N., Green, B. N., Toulmond, A., Wall, J. S., Weber, R. E. & Vinogradov, S. N. (1996). Giant hexagonal bilayer hemoglobins. Chem. Rev. 96, 31133124. 5. Vinogradov, S. N., Lugo, S. D., Mainwaring, M. G., Kapp, O. H. & Crewe, A. V. (1986). Bracelet protein: a quaternary structure proposed for the giant extracellular hemoglobin of Lumbricus terrestris. Proc. Natl Acad. Sci. USA, 83, 8034-8038. 6. Ownby, D. W., Zhu, H., Schneider, K., Beavis, B. T., Chait, B. T. & Riggs, A. F. (1993). The extracellular hemoglobin of the earthworm Lumbricus terrestris Determination of subunit stoichiometry. J. Biol. Chem. 268, 13539-13547. 7. Schatz, M., Orlova, E. V., Dube, P., JaÈger, J. & Van Heel, M. (1995). Structure of Lumbricus terrestris Ê resolution determined using hemoglobin at 30 A angular reconstitution. J. Struct. Biol. 114, 28-40. 8. Taveau, J.-C., Boisset, N., Vinogradov, S. N. & Lamy, J. N. (1999). Three-dimensional reconstruction Ê resolution: of Lumbricus terrestris hemoglobin at 22 A intramolecular localization of the globin and linker chains. J. Mol. Biol. 289, 1343-1359. 9. de Haas, F., Boisset, N., Taveau, J.-C., Lambert, O., Vinogradov, S. N. & Lamy, J. N. (1996). Three-

770

10.

11.

12.

13.

14.

15.

16.

17.

18.

19.

20. 21.

22.

dimensional reconstruction of Macrobdella decora (leech) hemoglobin by cryoelectron microscopy. Biophys. J. 70, 1973-1984. de Haas, F., Zal, F., You, V., Lallier, F., Toulmond, A. & Lamy, J. N. (1996). Three-dimensional reconstruction by cryoelectron microscopy of the giant hemoglobin of the polychaete worm Alvinella pompejana. J. Mol. Biol. 264, 111-120. de Haas, F., Zal, F., Lallier, F., Toulmond, A. & Lamy, J. N. (1996). Three-dimensional reconstruction of the hexagonal bilayer hemoglobin of the hydrothermal vent tube worm Riftia pachyptila by cryoelectron microscopy. Proteins: Struct. Funct. Genet. 26, 241-256. de Haas, F., Boisset, N., Taveau, J-C., Lambert, O., Vinogradov, S. N. & Lamy, J. N. (1996). Threedimensional reconstruction of the chlorocruorin of the polychaete hemoglobin Eudistylia vancouverii. J. Mol. Biol, 255, 140-163. Lanzavecchia, S., Wade, R. H., Ghiretti Magaldi, A., Tognon, G. & Bellon, P. L. (1999). A two-exposure technique for ice-embedded samples successfully reconstructs the chlorocruorin pigment of Sabella spallanzanii at 2.1 nm resolution. J. Struct. Biol. 127, 53-63. DesbruyeÁres, D., Chevaldonne, P., Alayse, A. M., Jollivet, D., Lallier, F. H. & Jouin Toulmond, C. et al. (1998). Biology and ecology of the ``Pompeii worm'' (Alvinella pompejana DesbruyeÁres and Laubier), a normal dweller of an extreme deep-sea environment: a synthesis of current knowledge and recent developments. Deep Sea Res. II, 45, 383-422. Breton-Gorius, J. (1963). Etude au microscope eÂlectronique des cellules chloragogeÁnes d'Arenicola marina L. Leur roÃle dans la syntheÁse de l'heÂmoglobine, TheÁse de Doctorat es-Sciences Naturelles SeÂrie A n 4129, Paris. Zal, F., Green, N., Lallier, F. H., Vinogradov, S. N. & Toulmond, A. (1997). Quaternary structure of the extracellular haemoglobin of the lugworm Arenicola marina. A multiple-angle-laser-light-scattering and electrospray-ionization-mass-spectrometry analysis. Eur. J. Biochem. 243, 85-92. Radermacher, M., Wagenknecht, T., Verschoor, A. & Frank, J. (1987). Three-dimensional reconstruction from a single-exposure, random conical tilt series applied to the 50 S ribosomal subunit of Escherichia coli. J. Microsc. 146, 113-136. Radermacher, M., Wagenknecht, T., Verschoor, A. & Frank, J. (1987). Three-dimensional structure of the large ribosomal subunit from Escherichia coli. EMBO J, 6, 1107-1114. Orlova, E. V., Dube, P., Harris, J. R., Beckman, E., Zemlin, E., Markl, J. & Van Heel, M. (1997). Structure of keyhole limpet hemocyanin type 1 Ê resolution by electron cryomicro(KLH1) at 15 A scopy and angular reconstuction. J. Mol. Biol. 280, 115-116. Penczek, P. (1998). Appendix: measures of resolution using Fourier shell correlation. J. Mol. Biol. 280, 115116. de Haas, F., Kushumov, A., Taveau, J.-C., Boisset, N., Vinogradov, S. N. & Lamy, J. N. (1997). Three-dimensional reconstruction of native and reassembled Lumbricus terrestris extracellular hemoglobin. Localization of the monomeric globin chains. Biochemistry, 36, 7330-7338. Rouse, G. W. & Fauchald, K. (1997). Cladistics and polychaetes. Zool. Scripta, 262, 139-204.

Structural Types of Hb in Annelids 23. Boore, J. L. & Brown, W. M. (2000). Mitochondrial genomes of Galathealinum, Helobdella, and Platynereis: sequence and gene arrangement comparisons indicate that Pogonophora is not a phylum and Annelida and Arthropoda are not sister taxa. Mol. Biol. Evol. 17, 87-106. 24. Cejka, Z., Santini, C., Tognon, G. & Ghiretti Magaldi, A. (1991). The molecular architecture of the extracellular hemoglobin of Ophelia bicornis: analysis of two-dimensional crystalline arrays. J. Struct. Biol. 107, 259-267. 25. Cejka, Z., Kleinz, J., Santini, C., Hegerl, R. & Ghiretti Magaldi, A. (1992). The molecular architecture of the extracellular hemoglobin of Ophelia bicornis: analysis of individual molecules. J. Struct. Biol. 109, 52-60. 26. Cejka, Z., Ghiretti Magaldi, A., Tognon, G., Zanotti, G. & Baumeister, W. (1989). The molecular architecture of extracellular hemoglobin of Eophila tellinii. J. Ultrastruct. Mol. Struct. Res. 102, 7181. 27. Mezzasalma, V., Di Stefano, L., Piazzese, S., Zagra, M., Salvato, B., Tognon, G. & Ghiretti Magaldi, A. (1985). Physicochemical and structural properties of the extracellular haemoglobin of Ophelia bicornis. Biochim. Biophys. Acta, 829, 135-143. 28. Van Bruggen, E. F. J. & Weber, R. (1974). Erythrocruorin with anomalous quaternary structure from the polychaete Oenone fulgida. Biochim. Biophys. Acta, 359, 210-214. 29. Martin, P. D., Kuchumov, A. R., Green, B. N., Oliver, R. W. A., Braswell, E. H., Wall, J. S. & Vinogradov, S. N. (1996). Mass spectrometry composition and molecular mass of Lumbricus terrestris hemoglobin: a re®ned model of its quaternary structure. J. Mol. Biol. 255, 144-169. 30. Adrian, M., Dubochet, J., Lepault, J. & McDowall, A. W. (1984). Cryoelectron microscopy of viruses. Nature, 308, 32-36. 31. Dubochet, J., Adrian, M., Chang, J.-J., Homo, J.-C., Lepault, J., McDowall, A. W. & Schultz, P. (1988). Cryoelectron microscopy of vitri®ed specimens. Quart. Rev. Biophys. 21, 129-228. 32. Milligan, R. A., Brisson, A. & Unwin, P. N. T. (1984). Molecular structure determination of crystalline specimens in frozen aqueous solutions. Ultramicroscopy, 13, 1-10. 33. Henderson, R. (1992). Image contrast in high-resolution electron microscopy of biological macromolecules. TMV in ice. Ultramicroscopy, 46, 110. 34. Frank, J., Radermacher, M., Penczek, P., Zhu, J., Li, Y., Ladjadj, M. & Leith, A. (1996). Spider and web: processing and visualization of images in 3D electron microscopy and related ®elds. J. Struct. Biol. 116, 190-199. 35. Taveau, J.-C. (1996). Presentation of the SIGMA software: software of imagery and graphics for molecular architecture. J. Struct. Biol. 116, 223-229. 36. Penczek, P., Radermacher, M. & Franck, J. (1992). Three-dimensional reconstruction of single particles embedded in ice. Ultramicroscopy, 40, 33-53. 37. Penczek, P. A., Grassuci, R. A. & Franck, J. (1994). The ribosome at improved resolution/new technique for merging and orientation re®nement in 3D cryoelectron microscopy of biological particles. Ultramicroscopy, 53, 251-270.

Structural Types of Hb in Annelids 38. Lebart, L., Morineau, A. & FeÂnelon, J.-P. (1982). Traitement des donneÂes statistiques meÂthodes et programmes, Dunod. 39. Van Heel, M. (1984). Multivariate statistical classi®cation of noisy images (randomly oriented biological macromolecules). Ultramicroscopy, 21, 164-184. 40. Ward, J. H., Jr (1963). Hierarchical grouping to optimize an objective function. J. Am. Statis. Assoc. 58, 236-244. 41. Marabini, R. & Carazo, J. M. (1994). Pattern recognition and classi®cation of images of biological

771 macromolecules using arti®cial neural networks. Biophys. J. 66, 1804-1814. 42. Saxton, W. O. & Baumeister, W. (1982). The correlation averaging of a regularly arranged bacterial cell envelope protein. J. Microsc. 127, 127-128. 43. Harauz, G. & Van Heel, M. (1986). Exact ®lters for general geometry three-dimensional reconstruction. Optik, 73, 146-156. 44. Felsenstein, J. (1989). PHYLIP-phylogeny inference package (version 3.2). Cladisties, 5, 164-166.

Edited by M. F. Moody (Received 12 June 2000; received in revised form 18 November 2000; accepted 18 November 2000)