Three-Dimensional Microarchitecture of the Plates (Primary, Secondary

Journal of Structural Biology Journal of Structural Biology 144 (2003) 282–300 www.elsevier.com/locate/yjsbi

Three-dimensional microarchitecture of the plates (primary, secondary, and carinar process) in the developing tooth of Lytechinus variegatus revealed by synchrotron X-ray absorption microtomography (microCT)

S.R. Stock,a,* K.I. Ignatiev,a T. Dahl,b A. Veis,b and F. De Carloc

a Institute for Bioengineering and Nanoscience in Advanced Medicine, Northwestern University, Chicago, IL 60611, USA b Department of Cell and Molecular Biology, Northwestern University Medical School, Chicago, IL 60611, USA c XOR, Advanced Photon Source, Argonne National Laboratory, Argonne, IL 60439, USA

Received 6 May 2003, and in revised form 5 September 2003

Abstract

This paper reports the first noninvasive, volumetric study of entire cross-sections of a sea urchin tooth in which the individual calcite structural elements could be resolved. Two cross-sectionally intact fragments of a Lytechinus variegatus tooth were studied with synchrotron microCT (microcomputed tomography) with 1.66 lm voxels (volume elements). These fragments were from the plumula, that is the tooth zone with rapidly increasing levels of mineral; one fragment was from a position aboral of where the keel developed and the second was from the zone where the keel was developing. The primary plates, secondary plates, carinar process plates, prisms, and elements of the lamellar–needle complex were resolved. Comparison of the microCT data with optical micrographs of stained thin sections confirmed the identifications and measured dimensions of the characteristic microarchitectural features. The interplay of reinforcing structures (plates and prisms) was more clearly revealed in the volumetric numerical data sets than in single or sequential slices. While it is well known that the primary plates and prisms in camarodont teeth are situated to improve resistance to bending (which can be termed primary bending), the data presented provide a new understanding of the mechanical role of the carinar process plates, that is, a geometry consistent with that required in the keel to resist lateral or transverse bending of the tooth about a second axis. The increase in robustness of teeth incorporating lateral keel reinforcement suggests that the relative development of carinar processes (toward a geometry similar to that of L. variegatus) is a character which can be used to infer which sea urchins among the stirodonts are most primitive and among the camarodonts which are more primitive. Ó 2003 Elsevier Inc. All rights reserved.

Keywords: Lytechinus variegatus; Microtomography (microCT); Sea urchin; Tooth; Synchrotron radiation; Biomineralization

1. Introduction architecture is developing and the mineral phase (high Mg calcite, Ca1xMgxCO3) is only partly complete, Modern sea urchins typically feed by scraping algae, oﬀer clear windows into the design space available to etc. from hard substrates. To resist abrasion and cata- echinoids for tailoring functional teeth. strophic fracture, the continuously growing teeth of Earlier studies of microarchitecture of sea urchin present-day sea urchins combine complex architectures teeth catalogued many aspects of the teeth and focused and multiple composite reinforcement strategies, and primarily on what could be learned from thin sections of various phylogenetic families (and perhaps even genera) the tissue (employing scanning and transmission elec- emphasize diﬀerent toughening or strengthening strate- tron microscopies, i.e., SEM and TEM, respectively, gies. The youngest portions of the teeth, where the and optical microscopy) or from the mineral remaining after the soft tissue was digested (Candia Carnevali et * Corresponding author. Fax: +312-503-2544. al., 1991; Chen and Lawrence, 1986; Kniprath, 1974; E-mail address: [email protected] (S.R. Stock). Maarkel,€ 1969a,b, 1970; Maarkel€ et al., 1969, 1973, 1976;

Wang et al., 1997). Recently, microfocus tube-based three suborders (see also Smith, 1981, 1984). The most microCT (microcomputed tomography or microto- primitive (in an evolutionary sense) are the Aulodonts mography) was employed to image sea urchin teeth (literally grooved tooth) characterized by grooved teeth (Lytechinus variegatus) noninvasively and to map min- and narrow epiphyses that do not bridge the shallow eral content variation as a function of 3-D position foramen magnum. Less primitive are the Stirodonts (Stock et al., 2002a,b). Low attenuation regions at/near (literally keeled tooth) which posses ‘‘T’’-shaped teeth the toothÕs stone part (spanning the flange between lat- and narrow epiphyses which do not meet over a deep eral corners, see the schematic in Fig. 9) and along the foramen magnum. Still less primitive are the Cama- carinar process—central prism boundary (both lateral rodonts (literally an object, any object, with an arched sides of the keel) were uncovered. The Mg fraction x in covering plus tooth) which have keeled teeth and wide the mineral phase (measured with synchrotron X-ray epiphyses meeting in a suture over the deep foramen diffraction with a small diameter beam in transmission) magnum. In the camarodont sea urchins, including L. could not account for all of the linear attenuation co- variegatus, the subject of the current report, weight efficient (l) decrease in the flange, and this suggested saved in the jaw structure (through a deep foramen that soft tissue is localized there (Stock et al., 2002b; magnum at the cost of decreased pyramid rigidity) al- Veis et al., 2002). Laboratory microCT was also used to lows more efficient grazing; the total metabolic cost of quantify the partition of stereom and stroma in pyra- forming the calcite of the jaw is also reduced. The mids and epiphyses of L. variegatus (Stock et al., 2003a), epiphyseal bridge restores rigidity at a minor cost of this despite the fact that the individual trabecula of added weight and energy to form calcite. stereom could not be resolved. Strengthening and toughening mechanisms in keeled Subsequent to the investigations cited above, sea urchin teeth center on the calcite mineral phase and synchrotron microCT was used to map the spatial include variation of reinforcement morphology (Candia distribution of mineral at the 1.3 lm level in a milli- Carnevali et al., 1991; Giesbrecht, 1880; Jensen, 1981; meter-sized fragment of a mature portion of the keel of Kniprath, 1974; Maarkel,€ 1969a,b, 1970; Maarkel€ et al., a L. variegatus tooth (Stock et al., 2003b). Two rows of 1969, 1973, 1976; Salter, 1861; Wang et al., 1997), high low absorption channels (i.e., primary channels) degree of crystallographic alignment of the discrete re- slightly less than 10 lm in diameter were found running inforcing elements (Donnay and Pawson, 1969; Maarkel,€ linearly from the flange to the base of the keel and 1969a,b; Nissen, 1969; Raup, 1959; Stock et al., 2002b, parallel to the two sides of the keel. MicroCT slices 2003b; Towe, 1967; Wang et al., 1997), alteration of revealed a planar secondary channel leading from each composition with position of the strengthening compo- primary channel to the side of the keel. The primary nent (Maarkel€ et al., 1971, 1976; Schroeder et al., 1969; and secondary channels were more or less coplanar and Stock et al., 2002b, 2003b; Wang et al., 1997), tailoring may correspond to the soft tissue between plates of the of the strength of the interface between adjacent primary carinar process. In other words, the low absorption calcite elements (Maarkel€ et al., 1976; Salter, 1861; Wang, zone between carinar process and central prism zone 1998; Wang et al., 1997), incorporation within the cal- seen in laboratory microCT slices was, in fact, the ar- cite of toughening inclusions comprised of globules of ray of low absorption channels which were too narrow macromolecules within the crystal elements (Berman in diameter and too closely spaced to be resolved with et al., 1988, 1990, 1993; Su et al., 2000) and employment the microfocus tube system. of noncrystalline CaCO3 as well as calcite (Beniash et al., The results on the L. variegatus keel fragment sug- 1997). As a consequence, indentation hardness varies gested that synchrotron microCT with volume elements significantly as a function of position (Maarkel€ and (voxels) on the order of 2 lm could be profitably em- Gorny, 1973; Wang et al., 1997) even though the mean ployed in noninvasive imaging of other portions of sea cutting edge hardnesses of exemplars from five disparate urchin teeth. The present study focuses on the relatively orders of regular echinoids showed no statistical differ- lightly mineralized portion of L. variegatus teeth, i.e., ences (Klinger and Lawrence, 1985) but quite large in- from plumula to midshaft, and is part of a larger intra-species variability. One wonders whether this last vestigation of tooth mineralization centered on the role observation is a consequence of unintentional position- of proteins in tailoring microstructure and on protein ing of the indents within zones of different hardness. A conservation across the wide evolutionary expanse be- second effect is that the continuously growing sea urchin tween echinoids, an advanced invertebrate type, and teeth are self-whetting: the abaxial tooth section consists mammals (viz, Veis et al., 2002, 1986). of an array of parallel primary plates which are weakly bonded to each other and which fracture preferentially 1.1. Background parallel to these interfaces and produce a new, sharp cutting edge. Jackson (1912), based upon lantern and tooth mor- The teeth of L. variegatus are ‘‘T’’-shaped and phology, divides the regular sea urchins (echinoids) into function as a ‘‘T’’-girder designed to resist bending 284 S.R. Stock et al. / Journal of Structural Biology 144 (2003) 282–300

keel. More details of the tooth architecture of L. variegatus and Sphaerechinus granularis, both members of the family Toxopneustidae, appear elsewhere (Chen and Lawrence, 1986; Maarkel€ and Gorny, 1973, respectively); details of keeled tooth architecture of other echinoid families, are covered in other studies (Candia Carnevali et al., 1991; Jensen, 1981; Kniprath, 1974; Maarkel,€ 1969b,b, 1970; Maarkel€ et al., 1976; Wang et al., 1997) with emphasis not only on the fully developed portions of the teeth but also the areas of rapidly increasing mineralization, the plumula. As a sea urchin tooth grows (1.5 mm/week growth rate and 10–11 weeks for complete tooth renewal in Strongylocentrotus purpuratus (Holland, 1965; Maarkel,€ 1969a,b)), ‘‘new odontoblasts must continuously differentiate from a pool of preodontoblasts at the terminus of the plumula’’ (Cavey and Maarkel,€ 1994). The sheets of odontoblasts which emerge from the toothÕs apical tip form syncytia in which the calcite primary plates nu- cleate and grow. After some plates have grown, a nucleus-free zone forms adaxial from the primary plates and abaxial from a thick highly nucleated peripheral cell Fig. 1. MicroCT-derived 3-D rendering of a L. variegatus pyramid layer. Approximately one week into tooth growth produced by treating all voxels with values greater than a given (about 1.5 mm from the aboral tip), a row of columnar threshold as solid and all those with lower values as empty space. The cells diverges abaxially from the peripheral cell layer and length of the tooth ‘‘t’’ is slightly less than 20 mm, and the plumula, occupies a position near the center of the nucleus-free keel, and flange are indicated by ‘‘P,’’ ‘‘K,’’ and ‘‘F,’’ respectively. One of the pair of demipyramids is labeled by ‘‘dp’’ and one of the zone (Chen and Lawrence, 1986; Holland, 1965; Mac- epiphyses by ’’e.’’ The fragmentsÕ rough positions are indicated by Gregor et al., 1956; Maarkel,€ 1969a,b; Maarkel€ et al., ‘‘a2’’ and ‘‘c.’’ The deep foramen magnum, bridged by the epiphyses, 1986). Maarkel€ (1969a,b) terms the columnar cells lies behind the tooth (i.e., is partly visible between ‘‘e’’ and ‘‘c’’) in this ‘‘lithoblasts’’ in order to emphasize their role in forming rendering. The x–y–z coordinate axes are used in the text and are the toothÕs stone part. shown on the upper left, and the arrows and first three letters of adaxial, abaxial, adoral, and aboral are used to indicate these directions. The studies cited above showed only single longitudinal sections or isolated transverse sections of the mineralizing plumula, no doubt because of the difficulty moments (Fig. 1). The primary plates form a large and labor required to produce serial sections of even portion of the abaxial flange of L. variegatus teeth partially calcified tissue. Techniques such as X-ray mi- (i.e., the bar of the ‘‘T’’), and backing the primary croCT which are largely insensitive to different soft tis- plates is the stone part of the tooth, a narrow band of sue types can contribute to the understanding of highly aligned needles. At the adoral end of the tooth, ultrastructure development in the plumula of sea urchin the stone part protrudes beyond the rest of the tooth teeth, not in the least because the distribution of mineral and, as it is the hardest part of the tooth (Maarkel€ and over large volumes is recorded automatically (without Gorny, 1973; Wang et al., 1997), forms the cutting appreciable sample preparation) in digital form for edge. Secondary plates occupy the adaxial portion of subsequent numerical analysis. The earlier lab microCT the flange, and the volume between the stone part and studies of L. variegatus plumula (Stock et al., 2002a,b; the secondary plates is filled with prisms. In cross- Veis et al., 2002), where spatial resolution was on the sections normal to the toothÕs axis, prism size appears order of 10 lm, documented large changes in lateral and to increase continuously as one moves adaxially away longitudinal mineral distribution over the 2 cm long from the stone part. In reality, needles and prisms are teeth. Of particular interest was a notch in the miner- different parts of the same 3-D structural element: the alized material on the abaxial median portion of the developing tooth first forms the primary plates and tooth which extended some millimeters along the plu- needles grow from the plate and thicken into prisms. mula. Second, an unmineralized area adaxial from the The distinction between needles and prisms, therefore, primary plates of the flange was observed to begin in the arises solely from where the sectioning plane intersects plumula and run most of the length of the tooth; this a particular reinforcing element. Relatively coarse feature may correspond to the portion of the nucleus- prisms fill the interior of the keel while the plates of free zone noted by others between the primary plates the carinar process form the exterior envelope of the and needles (Maarkel,€ 1969a,b; Veis et al., 1986). Spatial S.R. Stock et al. / Journal of Structural Biology 144 (2003) 282–300 285 resolution and contrast in the lab microCT studies, deionized water and dried. Transmission optical mi- however, were insufficient to allow individual calcite crographs were then recorded. elements to be resolved even in the plumula. This study Histograms containing 2000 or more voxels were aimed to improve contrast/spatial resolution and reveal computed for different areas of fragments a2 and c using new features in L. variegatus teeth. IDL (Interactive Data Language, 2003). The number of voxels in a slice which could be used for histograms of a particular microstructural zone was limited by that 2. Materials and methods zoneÕs size. Multiple slices could have been used to increase the number of voxels sampled, but, as will be Sections of the soft portion of a tooth of the short- discussed below, the spread in the histograms was an spined sea urchin L. variegatus were examined with intrinsic characteristic of features in the object sampled synchrotron microCT in this study. The originally in- with voxels that were only slightly smaller in size. The tact plumula was previously removed from the fixing mean linear attenuation coefficient hliQ and standard solution and used for laboratory microCT (Veis et al., deviation rQ were determined for each microstructural 2002). Handling prior to synchrotron microCT data region Q evident in the reconstructions, where Q might collection broke the plumula into several sections, and denote CP (carinar process), P (prisms), PP (primary a number of the sections were imaged with synchrotron plates, lateral areas of the flange), PP0 (primary plates, microCT. Only two of the two sections, a2 and c, were medial area) or LA (low absorption band running across free enough from damage and informative enough to the width of the flange). For comparison with the ex- warrant reporting; the approximate positions of these pected linear attenuation coefficients of pure calcite and two sections are indicated by a2 and c in the schematic calcite with various x, hliexperimental would normally be of Fig. 1. The term ‘‘depth’’ is reserved for dimensions used for a dense solid. Because a large number of the along the abaxial–adaxial axis (i.e., the y-axis in Fig. 1), voxels in every slice were only partially occupied by and ‘‘width’’ for dimensions along the line between the mineral, comparing hliQ with expected values was in- lateral corners of the flange (i.e., x-axis in Fig. 1) appropriate. In situations where partial volumes of shown by ‘‘de’’ and ‘‘w,’’ respectively, in slice 300 of empty space/soft tissue are prominent, one can some- Fig. 2). Depending on context, the term ‘‘length’’ may times use the maximum observed values of the linear denote dimensions along the adoral–aboral axis of the attenuation coefficient lmax for comparisons, but, as is tooth (i.e., along the z-axis) or to the largest extent of a discussed below, the presence of a small fraction of feature in a given slice. physically unrealistic values (i.e., voxels with l < 0) X-ray microCT was performed on station 2-BM of suggested that lmax might be similarly affected. Even in APS (Advanced Photon Source). A monochromatic the absence of such an artefact, the presence of noise beam (photon energy of 14 keV) and a 1K 1K element makes it undesirable to base comparisons on voxel CCD camera coupled (via a Zeiss AXIOPLAN 4 values tending toward the extremes of the distribution. neofluar lens) to a single-crystal CdWO4 scintillator Therefore, the value of l80;Q the linear attenuation co- were used. Views were recorded every 0.25° from 0° to efficient value below which 80% of the voxels in region Q 180° and were normalized for detector and beam non- fall was adopted to provide a robust alternative for uniformities; the sample was reconstructed on a comparison with expected values. 1024 1024 grid of isotropic 1.66 lm voxels. Data for Longitudinal numerical cuts parallel to the x–z plane reconstruction of fragments a2 and c were recorded with were obtained from the reconstructed volumes using unchanged microCT instrumental settings and within IDL. The stacks of slices were imported into the soft- little more than 12 h intervening between the start of ware used with a lab microCT system (Scanco Mi- data collection for a2 and end of data collection for croCT-40, 2003), and two types of 3-D image fragment c. More details on the apparatus appear else- interrogation were performed to develop understanding where (Wang et al., 2001). of the complex tooth geometry, particularly of the car- The soft part of a second tooth (for comparison with inar process plates. The first used the Scanco software to the synchrotron microCT data) was divided into 15 assign a threshold (binarizing the data into mineral and equal length portions starting from the aboral end of the empty space) and construct a 3-D rendering of the piece, and each portion was mounted in epoxy (Veis mineral microstructure. The second used the Scanco et al., 2002). Sections, 1 lm thick, were cut on a Reic- software to display three orthogonal sectioning planes in hert-Jung Ultracut E microtome using a Diatome his- the correct relative orientations, much like the unit cube tology diamond knife and floated from a water bath used to illustrate processing-related anisotropic micro- onto a glass slide. The sections were rinsed briefly with structure (e.g., Fig. 1 of Venkateswara Rao et al., 1988). 25% methanol and then stained with Toluidine blue. These will be termed isometric gray-scale images to After staining, the sections were then cleaned in 10% differentiate them from 3-D renderings based on the acetic acid in 50% methanol solution, washed with binarized data. 286 S.R. Stock et al. / Journal of Structural Biology 144 (2003) 282–300

Fig. 2. MicroCT slices of fragment a2, top to bottom, slices 180, 300, 400, and 500, and a schematic showing the plates which were especially well- resolved in slice 500. In these and all other slices, the lighter the pixel, the higher the attenuation coefficient of the voxel. The images are at the same magnification, and the horizontal field of view in the slice 500 image is 1075 lm. The low absorption region, secondary plates, primary plates, modeling clay and umbo are labeled ‘‘LA’’, ‘‘SP,’’ ‘‘PP,’’ ‘‘M,’’ and ‘‘U,’’ respectively, and the flange width ‘‘w’’ and depth ‘‘de’’ are also indicated.

3. Results ing voxels. Separations between slices can be computed by calculating the diﬀerence in slice numbers Fig. 2 shows four slices from fragment a2 (Fig. 1) and multiplying by the voxel size (1.66 lm). The tooth with the higher slice number indicating more aboral width and depth in slice 180 are 880 and 205 lm, slices. The lighter the pixels in these (and all other) respectively, and in slice 500 they are 1075 and slices, the greater the absorptivity of the correspond- 260 lm, respectively. S.R. Stock et al. / Journal of Structural Biology 144 (2003) 282–300 287

In the slices of Fig. 2, the low absorption region of the visible. ‘‘Two rows of calcareous tooth elements [pri- flange LA appears more distinctly in some slices (180 mary plus secondary plates], similar to the halves of and 400) than in others (300 and 500) and consists of bottom-up pyramids (or twin-bladed snow plows) two distinct zones running across the width of the tooth. stacked into each other’’ are seen, and ‘‘the elements of The depth of LA is between 40 and 45 lm in slice 180. both rows are symmetrical, though in a staggered order, The primary plates PP are well defined over significant and more or less interlocking,’’ at least over a small sections of each slice, and, within fragment a2, the sec- fraction of their length near the center of the flange ondary plates SP are just beginning to be formed. abaxial of the low absorption zone (Maarkel€ et al., 1976). Modeling clay ‘‘M’’ (used to affix the sample to the The schematic indicates the stagger (near ‘‘U’’ in Fig. 3) microCT apparatus) is seen on the abaxial side of the seen between several rows from either side of the flange. tooth in slice 180 (i.e., the lower surface in Fig. 2). Fig. 4 shows enlargements of the right sides (as ori- The schematic at the bottom of Fig. 2 traces the ented in Fig. 2) of slices 180, 280, and 380 (fragment a2) plates which are particularly distinct in slice 500. Fig. 3 in which well-defined primary and secondary plates shows an enlargement of the area surrounding the appear. The schematic to the right of the reconstructed toothÕs plane of mirror symmetry and a schematic slices illustrates the sharp change of plate orientation tracing of those primary plates which are particularly where the primary plate becomes the secondary plate, i.e., the corner of the inverted pyramids cited above. The calcite reinforcing elements bend sharply although the elements appear continuous, and this is clearly visible wherever the low absorption spaces can be seen between plates. In slice 180, eleven adjacent plates can be resolved within 79 lm, corresponding to a mean primary plate, center-to-center spacing of 7.9 lm. In slice 380, seven plates could be seen distinctly within 47 lm, yielding a 6.7 lm mean spacing. In slice 500, nine plates were seen in 63 lm, and the mean spacing was 7.0 lm. These spacings are in the plane of the slices, and cor- rection for the 45° tilt with respect to the slice plane yields spacings between 4.7 and 5.6 lm. The space between plates (the dark voxels, i.e., the soft tissue) was only a small fraction of each period in all of the slices. A60 60 voxel area within the image of slice 380 in Fig. 4 was carefully chosen to include only primary plates. The mean voxel value was 15 cm1, the standard Fig. 3. Enlargement of the center of slice 500 (fragment a2) and the 1 schematic of the central area of the tooth. The vertical field of view is deviation was 8.5 cm and there was only one distinct 390 lm in the microCT image. peak in the voxel value distribution, Fig. 5. Histograms

Fig. 4. Enlargements of the right sides of slices 180, 280, and 380 of fragment a2 showing the parallel plates PP transition to secondary plates SP. The vertical ﬁeld of view is 400 lm in the microCT images. 288 S.R. Stock et al. / Journal of Structural Biology 144 (2003) 282–300

by a thick layer of cell nuclei which are in turn bordered by cell layers with a density of nuclei comparable to that abaxial from the calcite plates. The arrows in Fig. 6 (and arrowheads in Fig. 7) mark rows of cells covered in Section 4. In the nucleus-free zone, needles (end on to the sectioning plane) are just beginning to appear in T741 of Fig. 6 (between and to the right of the pair of arrows and on the corresponding mirror position on the right side of the tooth). By T891 the needles cover the entire width of the flange; note that the needles on the right side of the flange in T891 appear more devel- Fig. 5. Histogram of linear attenuation coefficient values (cm1) for the oped due to a slight tilt in the sectioning plane relative to parallel plate area of slice 380 (Fig. 4). The arrows labeled ‘‘a,’’ ‘‘b,’’ the tooth axis. and ‘‘c’’ show positions on the histogram corresponding to the linear Fig. 7 shows enlargement of the area to the left of and attenuation coefficients for Ca1xMgxCO3 with x ¼ 0:3, 0.12, and 0, respectively. below ‘‘a’’ in thin sections T401 through T891. In Fig. 8, images T401b and T591b are enlargements from a position immediately above and to the left of ‘‘b’’ in Fig. 6; (not shown) of slices 180 and 280 were essentially images T401c and T591c are enlargements of an area to identical to Fig. 5. Often, well-defined peaks in histo- the right of ‘‘c’’ in Fig. 6. The light blue lines in Fig. 8, grams provide a robust partition of the volume into low which parallel the calcite plates and which lie abaxial attenuation (e.g., empty space or soft tissue) and high from them, are membranes defining the syncytial space attenuation (e.g., mineral) sub-volumes independent of in which the calcite plates grow. their spatial distribution; in cases such as the present one The width of the calcite-containing portion of the where the objects within the image (primary plates) have flange varies almost linearly from 680 lm in T401 to dimensions and separations only slightly larger than the about 970 lm in T1101. The width and depth of the voxel size in the reconstruction, partial volume effects mineralized part of T891 is 880 and 170 lm, respectively, will dominate and obscure the presence of the different while the depth in T1101 is 210 lm. The periodicity of populations (mineral voxels in the primary plates and parallel plates in the plane of the thin sections were soft tissue or empty space between plates). In this case measured on the lateral flanks of flange (i.e., in areas one is forced into ad hoc sampling of individual voxels such as shown in Fig. 8) and near the plane of symmetry appearing to suffer least from partial volume effects. of the tooth; the number of plates across which the total Within primary and secondary plates, l ranged between separation was measured varied between 5 and 15, de- 23 and 38 cm1. In the spaces between the plates, linear pending on the local quality of the thin section. The attenuation coefficients were between 2.1 and 3.9 cm1. mean center-to-center plate separation in the optical The low absorption area LA in Fig. 2 had a signifi- micrographs was between 4.4 and 7.4 lm, and no pat- cantly lower mean absorptivity than the parallel platesÕ tern in the variation could be ascertained. portion of the slice. For a 20 150 area of LA in slice Fig. 9 shows slice 650 of fragment c and a schematic 380, the mean linear attenuation coefficient hliLA was of the different microstructural zones within the tooth. A 1 about 12 cm and the standard deviation rLA was low absorption band LA, also seen in fragment a2, runs 8.4 cm1, and the histogram of area LA differs little, across the width the flange and separates the primary aside from the shift of the mean, from that shown in plate zone from the secondary plates, needles and Fig. 5. Once again, the histograms for slices 180 and 280 prisms. There may be two zones within the portion of differed little from that of slice 380. the flange containing the primary plates, PP lateral from Five transmission optical thin section (1 lm thick) the central zone PP0; there appears to be a sharp low micrographs of a second tooth are shown in Fig. 6 for attenuation demarcation between PP and PP0. The comparison with the microCT slices of fragment a2. plates of the carinar process ‘‘CP’’ appear well separated Thin section 401 of the tooth is designated ‘‘T401’’ in this slice and, based on this single slice, the plates (lower left of upper image). Thus, T741 is from a posi- would be interpreted to extend into block-like prisms tion approximately 0.340 mm from that of T401. The running perpendicular to the slice (That this is not the small dark blue disks are stained cell nuclei, and the case is shown in the slices of Fig. 10 and the rendering in calcite mineral appears as dark brown. These sections Fig. 11). show stacks of calcite plates surrounded by layers of That PP and PP0 represent distinct zones is seen from cells on the abaxial and lateral sides of the tooth. Im- the data in Table 1. The mean linear attenuation coef- 0 1 mediately adaxial from the calcite plates, T401 shows a ficient for PP, hliPP exceeds that of PP by 12 cm ,a nucleus-free zone (labeled ‘‘nfz’’) extending the entire difference much larger than the standard deviations for 0 1 1 width of the tooth and bordered on its adaxial margin either PP or PP (rPP 7:7cm and rPP0 8:7cm , S.R. Stock et al. / Journal of Structural Biology 144 (2003) 282–300 289

Fig. 6. Transmission optical micrographs of thin sections of a second L. variegatus tooth spanning approximately the same positions as the slices shown in Figs. 2–4. The arrows and symbols are described in the text, and the arrow labeled ‘‘aba’’ to the right of thin section T591 shows the abaxial direction. respectively). At the 80% level (i.e., l80;PP), a similar values in the corresponding area of primary plates in 1 0 diﬀerence is seen (11 cm ) between PP and PP .In fragment a2 while hliPP0 is roughly comparable to that 1 fragment c, hliPP is larger by about 10–11 cm than measured in fragment a2. 290 S.R. Stock et al. / Journal of Structural Biology 144 (2003) 282–300

Fig. 7. Enlargements of area ‘‘a’’ in Fig. 6. The vertical ﬁeld of view is 380 lm in each image.

Fig. 8. Enlargements of areas ‘‘b’’ and ‘‘c’’ in Fig. 6. The vertical ﬁeld of view is 415 lm in each image.

The mean linear attenuation coefficient within zone base of keel to the edge of the flange, the depth in slice LA was about 9 cm1 with a standard deviation of 650 of fragment c is 550 lm (This tooth dimension is 7.5 cm1, both being slightly smaller than was seen in given only for purposes of completeness and will not be fragment a2 for zone LA. In the prism zone P of the discussed further nor compared with optical thin sec- 1 1 1 keel, hliP 24 cm , l80;P 39 cm ,andrP 19 cm tions: valid comparisons could not be made because the 1 while in the needle zone N of the flange hliN 23 cm , keel is so rapidly growing in the portion of the tooth 1 1 l80;N 29 cm , and rP 9cm . In areas of the car- covered by fragment c and because the keel has not inar process encompassing plates and inter-plate spaces, reached its full length in fragment c). The low absorption 1 1 1 hliCP 19 cm , l80;CP 41 cm , and rP 26 cm . zone LA has a depth of about 24 lm, roughly one-half of In fragment c, the width of the flange is about 1180 lm that measured for LA in slice 500 of fragment a2. Except and is nominally the same as the width in slice 500 of at the external edges of the secondary plate zone, where fragment a2. The distance from the abaxial surface of the the plates appear to thicken and flare apart, the only flange to the adaxial edge of zone LA is between 160 and plates which can be resolved are in zone PP0 and these 165 lm; this is analogous to depth de defined in Fig. 2 have a center-to-center spacing of roughly 5–6 lm. and is somewhat smaller than that in fragment a2 The keel width in slice 650 of fragment c is 260– (205 lm for slice 180 and 260 lm for slice 500). From the 270 lm at the position marked by the arrow in Fig. 9. S.R. Stock et al. / Journal of Structural Biology 144 (2003) 282–300 291

The long, laterally inclined plates seen near slice 200 persist to positions at least as adoral as slice 650, primarily near the keel and on the exterior, lateral surfaces. In addition to the wing-shaped sections of the plates (‘‘W’’ in slice 650 of Fig. 10), the plates end in what appear to be block-like sections (‘‘B’’ in Fig. 10, slice 650). Inspection of the individual slices 400–650 of Fig. 10 might convince one that the carinar process plates terminate in coarsely spaced blocks running more-or-less along the length of the keel. The carinar process plates are, in fact, curved and intersect the slice planes at very different angles depending on the nearness of the flange and the secondary plates from which they grow. Fig. 11 shows a 3-D rendering of the keel of fragment c. In this rendering, the low absorption voxels are eliminated (i.e., shown as empty space), and the fragment is viewed from the base of the keel K. The trace of the secondary plates (SPt, dashed white line) appears on the right-hand vertical side of the rendering, and the toothÕs axis is indicated by the arrow labeled ‘‘TA.’’ The carinar process plates CP clearly bend to an orientation Fig. 9. MicroCT slice 650 of fragment c; its width (flange from tip to which seems nearly perpendicular to the toothÕs axis, tip) is 1.18 mm. The white box superimposed on the slice indicates the quite contrary to the picture which might be inferred area used to produce Fig. 13 and Fig. 16. The schematic below the from single slices or from illustrations in the literature reconstructed slice contains dotted and dash-dotted lines marking transition between different zones visible in the slice. Letters ‘‘LA, PP0, (see Section 4). PP, N, P, CP, and SP’’ denote the low absorption, central primary Along an adaxial–abaxial line across the block-like- plate, lateral primary plate, needle, prism, carinar process plate, and appearing carinar process plates (Figs. 9 and 10), the secondary plate regions, respectively. linear attenuation coefficients increase gradually to a maximum at the center of each plate. This is another The central prism zone occupies 155–160 lm of this clear indication that the plate is tilted with respect to the width with the carinar processes on either side of the plane of reconstruction. The amount of tilt is deter- prisms each covering about 55 lm width. mined explicitly in Section 4 and is used to correct the Near slice 200 of fragment c, the keel is shorter than measured carinar process plate spacings given in the in the larger-numbered slices: therefore, the lower the following paragraph. slice number in fragment c, the more aboral the location In order to determine carinar process plate thick- (see Fig. 10).1 The adaxial–abaxial extent of the carinar nesses and their center-to-center spacing in fragment c, process plates is large and indicates that this plate axis is the variation of linear attenuation coefficient at position nearly coplanar with the slice plane. Examination of the 585, 460 was measured for 1K 1K slices 645 through sequence of slices around slice 200 shows significant 666 (Fig. 12). This position is at the center of the carinar lateral tilt of the carinar process plates, i.e., the plates process plate identified by the arrow in Fig. 9. In this are rotated about the adaxial–abaxial line out of the tooth position, the center-to-center plate spacing is 1 slice plane. There is a smooth transition from the tilt of about 11 voxels or 17–18 lm. If a value of 10 cm is the secondary plates with respect to the z-axis (and the taken as marking the boundary of each plate, the plates axis of the tooth) to the carinar process plate orientation are 5.5–6.5 voxels thick (9–11 lm), and the empty space/ in the longer portions of the keel in fragment c (see soft tissue between plates is slightly smaller than the Fig. 11). Unfortunately, fragment c ends shortly after plate thicknesses. These thicknesses are measured with slice 650, and characterization of more mature sections respect to the z-axis and not perpendicular to the plate of the keel must await examination of new specimens. surfaces which would be more accurate (see Section 4). Fig. 10 also illustrates variation and spatial distribution of prisms within the keel. In some areas of most slices 1 This is the exact opposite of the slice numbering for fragment a2 the prisms appear arrayed in lateral rows, but this is and is a consequence of fragment c being inverted during data probably an artefact of aliasing in the images. Aside from collection relative to a2. While this may be somewhat confusing to the reader, this preserves the relationship of the position on the detector of the obvious adaxial–abaxial gradient in prism size (in- a given slice from each sample in case as-of-yet-unsuspected systematic creasing diameter with increasing y), the prisms are not so errors are uncovered. well-defined as to allowed detailed characterization 292 S.R. Stock et al. / Journal of Structural Biology 144 (2003) 282–300

Fig. 10. Development of the keel (fragment c) illustrated in a series of slices (the slice number is given on each image). The horizontal ﬁeld of view in each image is 250 voxels, and ‘‘B’’ and ‘‘W’’ (slice 650) are discusssed in the text.

Table 1 Values of linear attenuation coeﬃcients within areas indicated in Fig. 9 for fragment c slice 600 Position Linear attenuation coeﬃcient (cm1)

Mean SD 80% Max Min Nvox CP 18.9 26.0 41 90 )34 2365 P 24.4 18.7 39 75 )39 3719 N 23.0 9.1 29 49 )11 3599 LA 8.9 7.5 14 35 )18 2889 PP0 13.3 8.7 20 40 )24 3599 PP 25.6 7.7 31 45 )18 2065 Mean and standard deviation have their normal meanings. The 80% voxel value is the linear attenuation coeﬃcient below which 80% of the observed voxel values are found (in the area in question), and max and min are the greatest and smallest observed values.

Fig. 11. 3-D rendering of the keel (fragment c) derived from the mi- croCT data. The arrow ‘‘TA’’ points along the toothÕs axis toward the aboral end of the tooth. The other symbols are deﬁned elsewhere. of their distribution of diameters, of their cross-sectional aspect ratios, of the length of empty space/soft tissue, etc. However, characteristic dimensions can be estimated: in slice 650 the prism widths appear to be as large as 10 lm and their depth (i.e., dimension along the y-axis) appear to be about 50% larger. About 4 voxels of empty space Fig. 12. Linear attenuation coeﬃcients at position 585, 460 for slices separate adjacent prisms, and seven prisms comprise the 645 through 666; this illustrates the center-to-center plate spacing and bottom ‘‘row’’ of the keel. At the position of the arrow in plate thicknesses. S.R. Stock et al. / Journal of Structural Biology 144 (2003) 282–300 293

Fig. 9, ten or perhaps eleven prisms are seen in the width tions of the primary plates in region LA (reflected with of the prism zone, and the characteristic prism width and respect to the primary plates across the border of LA). space between prisms combined with this number pro- These continuations are similar to what Maarkel€ duce a prism zone width consistent with that reported (1969a,b, Fig. 8) reports in transverse polished sections above from direct measurement. (observed under polarized light) through the highly mineralized sections of teeth of Echinus acutus, Strong- ylocentrotus droebachiensis and Sphaerechinus granu- 4. Discussion laris; in this case rows of laterally adjacent needles with identical orientations separated by rows of touching, The width of fragment a2 in microCT slice 180 out-of-contrast needles give an appearance similar to (880 lm) matches that in T891 (880 lm), and it should what is seen in slice 500. At the stage of mineralization be noted that the needles in the nucleus-free zone, sep- present in slice 500, it is unlikely that needles, if present, arated from the primary plates by a mineral free zone are so highly packed as to produce part of the bands 40 lm deep, would be expected to produce contrast seen by Maarkel.€ While the orientation and spacing like that observed in slice 180 (Fig. 2). The depth of the of the features in slice 500 are similar to those of the mineralized portion of the tooth in T891 is only 170 lm secondary plates, their origin and development (compared to 200 lm in slice 180) and in T1101 is differ markedly from secondary plates, ruling out this 210 lm (vs. 205 lm in slice 280 and 260 lm in slice 500). identification. Taken together, the data for fragment a2 indicate that The features in slice 500 (Fig. 2) may, however, be the depth of the calcite-containing portion of the flange small stacks of the lamellae linking the primary plates remains constant for several hundred micrometers while and needles/prisms and which are often highly ordered the width increases at a nearly constant rate; subse- near their origin on the primary plate (Maarkel,€ 1969a; quently the depth of the calcified region also increases see also Fig. 35 D of Cavey and Maarkel,€ 1994). These toward the oral end of the tooth (Note that this is prior components of the lamellar–needle complex have re- to the formation of a visible keel). ceived considerable attention previously (Giesbrecht, The primary plate spacings measured in the optical 1880; Jensen, 1981; Maarkel,€ 1969a; Maarkel,€ 1970; thin sections (between 4.4 and 7.4 lm) are in good Maarkel€ and Titschack, 1969; Maarkel€ et al., 1976; Salter, agreement with those measured in the microCT slices of 1861). The lamellae appear before needles, and, in cross- fragment a2 (between 6.5 and 8 lm). The slight differ- sections, the eventual disorder of lamellae and the ences can be ascribed to differences in inclination of the growth of needles would appear at positions removed two sampling planes relative to the tooth axis, changes adaxially from the lamella-primary plate border. in inclination which are impossible to avoid in a curving A second observation supports the conclusion that tooth sectioned parallel to a single plane. Wang et al. the features in LA are different from those in the pri- (1997) give the primary plate thicknesses for Paracen- mary and secondary plate zones. The mean voxel values trotus lividus as 1 lm adjacent to the stone part and 5 lm from the plate-containing regions of the tooth are con- near the outer tooth surface, and the observed separa- stant within experimental uncertainty for slices 180, 280, tion between plates was only a fraction of the platesÕ and 380 and about 25% higher than means from region thicknesses. Also for Paracentrotus lividus, a similar LA from the same slices (also constant). How significant geometry and 4 lm plate spacings away from the stone this difference is cannot be determined at present but part can be measured in Fig. 35d of Cavey and Maarkel€ must await further synchrotron microCT of better pre- (1994). Maarkel€ (1969a,b) reports primary plate spacings served samples. Assuming that the difference is signifi- for 14 species (10 families) with keeled teeth and these cant, a change in composition x might be responsible for values range between 16 and 33 lm for the positions the difference, but, given the gradual and relatively small where the thicknesses were measured in the present variation in composition x observed by Wang et al. study; Maarkel€ Õs values for Tripneustes gratilla and (1997) along the length of needle/prisms and the thinness Sphaerechinus granularis, both members of the same and separation of lamellae, the probable origin is partial family of regular echinoids as L. variegatus, are 20 lm, volume effects (i.e., inclusions of more soft tissue with and this suggests that there is something systematically the mineralized lamellae than is the case with the pri- different with his definition of the separation of tooth mary plates). elements, i.e., ‘‘Zahnelementabstand’’ (it is not the effect In the Section 2 it was noted that the number of of overlapping of plates in the toothÕs central region voxels which could be included in a histogram of a given because Maarkel€ explicitly notes that ‘‘in the (central) microstructural zone was limited by that zoneÕs area region of overlap the tooth elements lie twice as densely’’ within the slice in question. A glance at the large stan- [authorsÕ translation]). dard deviations in Table 1 might lead one to conclude Careful inspection of adaxial portion of slice 500 of that the spread in voxel values could be improved by fragment a2 (Fig. 2) suggests that there are continua- using data from many slices. The data show otherwise: 294 S.R. Stock et al. / Journal of Structural Biology 144 (2003) 282–300

portion in the latter rendering. The white dotted lines are used to indicate the traces of the primary plates (PPt), and the low absorption band running across the width of the flange is present in the volumes of the left- hand and center images. As one would expect from the geometry of the primary plates (the curved portion of the inverted snowplow blade between the sharp corners), both curved and linear traces are observed. Maarkel€ (1969a, Fig. 9) showed longitudinal polished sections through the plumula of Temnopleurus hardwickii and Colocentrotus pedifer and show the inclination between primary plates and their bordering central lamel- Fig. 13. Histogram of the carinar process area of slice 650 of fragment c. lae and needles/prisms. Wang et al. (1997) also showed longitudinal sections through teeth of Paracentrotus liv- the spread in voxel values is due primarily to partial idus adjacent to the cutting edge. Fig. 15 shows a longi- volumes. Only the histogram from the area encom- tudinal numerical section through fragment a2 passing plates many voxels wide and equally wide int- approximately at the medial plane. The horizontal field erplate spaces (the carinar process) showed, albeit of view in Fig. 15 is 230 voxels, and this y–z plane extends indistinctly, separation into low absorption and high from slice 100 (bottom) to 630 (top). Over the 0.8 mm absorption peaks (Fig. 13). length of the fragment, considerable curvature is present. Fig. 14 shows three isometric gray scale images (same Taking the adaxial edge of the tooth as an indicator, the gray scale as in the slices) of fragment a2 sectioned on tilt varies from 10° clockwise from the z-axis at slice 100 different planes. The abaxial and adaxial sides of the to 15° counter-clockwise at slice 600. The traces of tooth are labeled ‘‘aba’’ and ‘‘ada,’’ respectively, the several primary plates (and their extensions into lamellae positive z-axis points toward the toothÕs oral end, and in LA) are indicated by dotted lines, and there is some right-hand face of each image (the y–z plane) is near the curvature present. Within this section, the more abaxial a medial plane (plane of approximate mirror symmetry) of portion of the plate, the smaller its angle with the z-axis, the tooth. Note that the x–z faces of the center and right- and the greatest angles are seen adjacent to LA. There hand gray scale renderings are the same numerical appears to be a sharp bend in the ‘‘plates’’ within LA, sectioning plane; but the abaxial portion of the tooth and this probably reflects the lamella–needle complex volume is shown in the former rendering and the adaxial formation described above and shown quite graphically

Fig. 14. Three isometric gray scale images of fragment a2 cut along three diﬀerent orthogonal planes. The traces of the primary plates on the diﬀerent numerical sectioning planes are marked with dotted lines and ‘‘PPt.’’ The other symbols have the same meaning as before. S.R. Stock et al. / Journal of Structural Biology 144 (2003) 282–300 295

of tooth. The present data for fragment c (the plate structure on the lateral sides of the keel) clearly correspond with the secondary channel structure seen earlier, but because of the maturity difference, the scale of the structures cannot be directly compared. The end of the carinar process plates next to the prism zone is where the primary channels will be located in the more mature sections of the tooth. In fragment c there is so much open space that it is difficult to identify the early stages of the primary channels. As noted in Section 3, the shape of the carinar process plates changes considerably in the zone linking flange and keel. Even in the main portion of the keel (i.e., adaxial and abaxial of the arrow in Fig. 9), the carinar process plates are not flat but are rather curved (Fig. 11). As noted above, the plates are also inclined with respect to the z-axis along which their separation was measured. Following changes in position of the plate identified by an arrow in Fig. 9 allows estimation of this plateÕs inclination with respect to the z-axis. The data are consistent with the plate normal being inclined 50–60° from the z-axis (tooth axis). The apparent carinar process plate spacing (center-to-center) and thickness measured along the z-axis (17–18 and 9–11 lm, respectively) must be corrected to 9 and 5 lm, respectively. It is interesting to note that the thicknesses are on the same order as those in the primary plate zone. This picture is quite different from that most readers would infer from published diagrams of camarodont carinar processes (e.g. Maarkel,€ 1969a); to be fair, Maarkel€ appears to only reproduce what is seen in polished sections and does not discuss the carinar process geometry Fig. 15. Longitudinal numerical section through the center of fragment within the keel except as it pertains to growth from the a2 at the medial plane. The dotted lines mark the traces of the primary secondary plates. Instead of plates running more-or-less plates (and their continuations into area LA. The horizontal field of along the keelÕs axis (i.e., along line TA in Fig. 11), the view is 380 lm. plates can be better described as being inclined midway between parallel to and perpendicular to the keelÕs axis, in Fig. 9 of Maarkel€ and Titschack (1969), albeit for an and this alignment serves the important mechanical aulodont tooth (which in this respect differs little from function described below. One wonders whether the camarodont and stirodont teeth). carinar processes in other camarodonts are similarly The true separation between centers of adjacent pri- oriented, and the authors believe that future studies will mary plates depends on the inclination of the stack of show this to be the case. plates with respect to the rotation axis of the microCT Interpolating to 14 keV X-ray energy using the values apparatus. The macroscopic curvature of the tooth is one of the mass attenuation coefficients tabulated by NIST source of inclination (variable with respect to slice (2001) and using the density q ¼ 2:72 g/cm3 for pure number); as noted above, the ‘‘inverted snow plow calcite (x ¼ 0), linear attenuation coefficients for 1 1 blade’’ geometry of the stack of plates is a second source Ca1xMgxCO3 are 41 cm for x ¼ 0; 39 cm for of inclination. The plate spacing in fragment a2 (center- x ¼ 0:07, 38 cm1 for x ¼ 0:12 and 33 cm1 for x ¼ 0:3 to-center, corrected for tilt relative to the slice plane) was (see Fig. 5). Strictly speaking, the density should be that earlier given as between 4.7 and 5.6 lm. These values are for the Mg composition of the plates, but, because quite comparable with those in Paracentrotus lividus, density changes very little with Mg fraction x (Cards 86– another camarodont sea urchin (Wang et al., 1997). 2335, 86–2336, 86–2343, and 86–2348 of the Powder An earlier synchrotron microCT study of a small keel Diffraction File, 1999), the value for x ¼ 0 can be used fragment of L. variegatus showed what were identified as without introducing appreciable error. For water at primary and secondary channels (Stock et al., 2003b), 14 keV, the NIST compilation gives l ¼ 2cm1, and this but this fragment was from a much more mature section should be a good approximation for non-mineralized 296 S.R. Stock et al. / Journal of Structural Biology 144 (2003) 282–300 tissue between the primary plates. The values of linear the high linear attenuation coefficients. Further, the as- attenuation coefficients measured in fragment a2 for the sociation of the high values only with and within larger primary plates (23–38 cm1) and those for the space calcite structural elements argues against contamination between plates (2.1–3.9 cm1) are in good agreement as a source of this effect. with the calculated values when one considers the large After the anomalously high voxel values were noted partial volume effects arising from features no greater in the carinar process areas of fragment c, different than a few voxels wide and spaced a voxel or less apart. reconstruction algorithms and filtering methods were Note that spread of linear attenuation coefficients down- applied to rule out the possibility that the anomaly ward from the expected calcite value and upward from was an artefact of the Fourier transform reconstruc- the expected soft tissue value are exactly what one would tion algorithm (Gridrec, Dowd et al., 1999) used for expect if partial volumes were important. reconstructing the data. The same area of the carinar The histogram for slice 380 of a2 (Fig. 5), taken within processes of slice 600 were reconstructed with different the parallel plate region and away from the exterior methods (Gridrec without filter, with Shepp-Logan surface of the tooth, shows 4.5% of the voxels with linear filter, with Hamming filter; filtered back projection; attenuation coefficients less than zero (a physically un- simple backprojection), and this did not affect the realistic result) and no voxels above 45 cm1, a result resulting histograms appreciably. The individual ra- typical of the data set for fragment a2. The reconstruc- diographs were inspected for anomalies and none were tion errors at either end of the peak in the histogram) are noted. directly attributable to a combination of the finite dy- Synchrotron X-rays from sources such as APS pro- namic range of the detection system, the opacity of the duce prominent features in microradiographs related to tooth at 14 keV, scattering artefacts (Kak and Slaney, refraction at surfaces between domains of different 2001) and the inability of the reconstruction algorithm to electron density. It is also well known that, with this accurately track changes of attenuation in the resulting radiation, refraction acts as an analog sharpening filter data. This minor inaccuracy is not uncommon in mi- for features in the interior of objects imaged with ab- croCT data but is rarely commented upon. In addition, as sorption microCT. To be sure, the refractive index of is discussed below for fragment c, X-ray refraction effects 14 keV photons differs from unity by only a few parts also contribute to the very high and very low values. per million for materials such as calcite, but this is en- Partial volume effects are minimized in the large ough to produce ‘‘hot edges’’ (i.e., complementary high blocks of the carinar process of fragment c. Within the intensity/low intensity fringes visible in microradio- plates of the carinar process in slice 200, voxels values of graphs at the outer margins of a sample. This outer 60 cm1 were not uncommon with the range 35–45 cm1 margin effect in the individual views used to reconstruct being more common; in slice 600 voxels had values as slices translates into unphysical values of l at the edges high as 90 cm1 and as low as )34 cm1 (Table 1) with of the reconstructed sample, but it is difficult to see how the normal range being somewhat higher than that in outer margin effects will make large alterations in l tens slice 200. This sample was imaged previously with lower of micrometers below the sampleÕs surface. Local ge- resolution lab microCT, and such local high absorptivity ometry within the sample can, however, lead to large was not observed (Veis et al., 2002). Because the frag- changes in l. ments were kept sealed in a clean plastic tube through- Consider the small sections of slices from the volume out and between lab and synchrotron microCT imaging, containing the plates of one of the carinar processes it is unlikely that contaminants could be the source of (Fig. 16). The box in Fig. 9 shows the approximate

Fig. 16. Small sections (55 lm horizontal field of view, defined by the box in Fig. 9) containing images of the carinar process plates, from left to right, of the odd numbered slices 591–609. In slice 595, ‘‘F’’ and ‘‘K’’ indicate the part of the structure closest to the flange and keel, respectively, and the dashed lines and arrowheads are positioned to aid the eye in following one plate from slice to slice. The white areas are the plates, and all the voxels shown as dark gray (within the interior of the plates) have values higher than 95% of the voxels in the same area of slice 590 (not shown). S.R. Stock et al. / Journal of Structural Biology 144 (2003) 282–300 297 position of the sub-volume, and every other slice be- teeth could resist substantially higher tensile stresses tween 591 and 609 is shown in Fig. 16. One plate is compared to the flange (at least 1.8 the tensile stress followed through the volume (arrowheads), and four for Sphaerechinus granularis, 2.1 for Paracentrotus other plates can also be seen in one or more of the slices. lividus and 2.3 for Arbacia lixula, with the first two The high absorption plates (white) are clearly delin- being camarodonts and the last a stirodont). These ex- eated, and all of the voxels with the highest five percent perimental values of the asymmetry of tensile strength in of values of l are shown as a uniform dark gray. Note flange and keel reflect loading along a single axis, a that the top five percent voxels are generally grouped situation which does not reflect the complex environ- together in large areas within the plate interiors. ment under which the teeth function. The center of the carinar process plate labeled in The teeth of large specimens of L. variegatus can ex- Fig. 16 moves 170 lm over 30 lm height, indicating ceed 20 mm in length, and the primary bending moment that the plate is 15° tilted from parallel to the slice encountered during feeding (i.e., radial bending within a plane. Fig. 12 showed the plate thickness to be 10 lm plane parallel to and intersecting the lanternÕs axis of with low absorption areas spanning somewhat shorter symmetry, shown upper left, Fig. 17) is not the only distances. This means that large (tens of micrometers on bending force. Indeed, it is well known that the intra- a side), fairly flat objects often take orientations where a pyramidal and other muscles allow each pyramid some significant refracted intensity can be built up. Therefore, freedom to align to scrape an uneven or skewed substrate the plate thicknesses and spacings, the plate inclinations (e.g., Andrietti et al., 1993). As is shown in Fig. 17, upper and lateral dimensions and refraction redistributing in- right), lateral or transverse (secondary bending may oc- tensity at a fraction of a degree from the incident beam cur when one side of the flange is loaded in compression direction combine to produce the unphysically high (i.e., is in contact with the substrate). The entire structure absorptivities within the plates and the large, negative l will bend about this second axis, and in the absence of between the plates; while this geometrical, refraction- constraint, a fatal crack could form and propagate be- related situation may be uncommon, it might occur in tween primary plates in the portion of the flange loaded other samples of unknown composition l and led to in tension. This fracture would not necessarily be be- erroneous interpretation of l derived from these syn- tween the outermost primary plates (as in the self-whet- chrotron microCT reconstructions. ting of the tooth) but might be much deeper in the tooth. Earlier, lab microCT revealed the presence two re- The array of parallel, weakly bonded prisms could not gions in the primary plate portion of a L. variegatus provide the required constraint since their resistance to tooth flange (Stock et al., 2002b); the geometry of the shear along their axes is minimal. An array of parallel, two zones is essentially identical for positions of equal closely spaced plates occupying the lateral portions of maturity for both synchrotron and for lab microCT. The each side of the keel can constrain transverse bending. relatively sharp demarcation between PP and PP0 seen in For the flange loading shown in Fig. 17 (upper left), the synchrotron reconstructions was not apparent in the displacements might occur across the keel as pictured. lab results, which is not surprising given the considerably The carinar process plates on the compression side of better spatial resolution and contrast sensitivity of the the keel could be displaced only a very small amount former. The line of low absorption between PP and PP0 before further structural deflection would require com- may be a real feature or an artefact of drying, and this pression of the entire crystal element (i.e., loading over can be settled only be obtaining data on other samples. the entire face). Because the plates are symmetric on Fig. 17 (lower right) shows a rendering of fragment c either side of the keel, transverse bending in either di- with some of the secondary and carinar process platesÕ rection would be resisted equally well. traces on the sides of the rendering shown by white The presence of the carinar process plates is probably dotted lines. One plate on the top of the rendering is as important an improvement in the design of ‘‘T’’- highlighted in white. The lower left schematic of Fig. 17 shaped sea urchin teeth as was the introduction of the shows the relative orientations of the reinforcing ele- keel over preceding forms. That this is the case can be ments in this portion of the tooth. As noted previously seen from the schematics of Fig. 17 and the discussion in (Maarkel€ and Gorny, 1973; Maarkel€ et al., 1976), the the previous paragraphs. The mechanical entity of pri- shape (i.e., architecture) of a T-shaped girder is equally mary plate plus outgrown secondary plate plus attached well suited to resisting bending a) with compression on carinar process is, in isolation, a fairly delicate 3-D the flange and tension on the web and b) with the object. A stack of these complete shapes can be very loading reversed, i.e., tension on the flange and com- robust. The microCT data presented above show no pression on the web. The schematic in the upper left of evidence of anything other than these plates being con- Fig. 17 illustrates situation a). In point of fact, the pri- tinuous, and microbeam diffraction mapping (Stock mary plates are loaded in compression when the sea et al., 2002b) showed these elements were crystallo- urchin rasps its teeth against rocks while eating. Maarkel€ graphically continuous. The continuity of these three et al. (1976) noted that the keel of T-shaped sea urchin types of plates from the far side of the flange to the 298 S.R. Stock et al. / Journal of Structural Biology 144 (2003) 282–300

Fig. 17. Function of the carinar process plates in resisting transverse tooth bending. In all drawings and the rendering, the cutting edge of the tooth is up, and symbols have the same meaning as in the other figures. Upper left) Primary bending to which the tooth is exposed (compression in the flange and tension in the keel). The stone part ST (center of the flange), the primary plates PP and secondary plates SP are shown schematically. Upper right) Possible displacements in the keel in response to uneven flange loading during scraping on an uneven surface, i.e., transverse bending about a second axis. Lower left) Geometry of the reinforcing elements of the structure. Lower right) Rendering of fragment c from the microCT data in the same orientation as the schematics and illustrating the relative orientations and locations of plates and prisms. margins of the keel encourages the speculation that load primary plates. This allows significant transport to the transfer occurs from primary plate to secondary plate to interior of the densifying tooth. As the disks extend carinar process plate. If such load re-distribution were laterally and replace more and more of the organic occurring, then weaker features such as the high Mg material, the primary skeletal elements (i.e., the bor- disks between primary plates could be shielded. The dering primary plates) become better cemented. Small tooth could then be loaded to a level which, in the ab- channels of organic material apparently remain even sence of such load re-distribution, would cause fracture after the highest mineralization levels are achieved. through the high Mg ‘‘interfacial’’ layer. While the transport role is most likely no longer rele- It is also interesting to note from othersÕ observations vant, the organic channels can interrupt interface crack (Maarkel€ et al. 1973, 1976; Wang et al., 1997) that the propagation, acting as crack arrestors in the weak in- Mg-rich disks of the secondary mineral system may also terface zone between primary plates, and increase the be ‘‘designed’’ for fracture resistance, as a secondary energy required to split two primary plates. consequence of their mode of formation. Initially the As was discussed by others (e.g., Jackson, 1912; disks are well-separated with substantial soft tissue be- Maarkel,€ 1969a; Smith, 1984) the absence of a closed tween nearest-neighbor disks within the space between foramen magnus (i.e., joined epiphyses and support of S.R. Stock et al. / Journal of Structural Biology 144 (2003) 282–300 299 the mineralizing tooth at the epiphyseal crest) indicates Sylocidaris affinis with the typical camarodont lantern. J. Zool. that stirodonts are a more primitive form than cama- (Lond.) 231, 595–610. rodonts (whose lanterns possess these features). Both Beniash, E., Aizenberg, J., Addadi, L., Weiner, S., 1997. Amorphous calcium carbonate transforms into calcite during sea urchin larval types possess ‘‘T’’-shaped teeth with carinar processes. spicule growth. Proc. Roy. Soc. (Lond.) B 264, 461–465. The tooth cross-sections (stirodonts and camarodonts) Berman, A., Addadi, L., Weiner, S., 1988. Interactions of sea-urchin figured by Maarkel€ (1969a) show that the carinar pro- skeleton macromolecules with growing calcite crystals—a study of intracrystalline proteins. Nature 331, 546–548. cesses of the stirodonts are more like prisms than like the plates of the camarodonts, which leads one to infer that Berman, A., Addadi, L., Kvick, AA., Leiserowitz, L., Nelson, M., Weiner, S., 1990. Intercalation of sea urchin proteins in calcite: a the more prism-like the elements of the carinar process, study of a crystalline composite material. Science 250, 664–667. the more primitive the sea urchin. This character (more Berman, A., Hanson, J., Leiserowitz, L., Koetzle, T.F., Weiner, S., prism-like vs. more plate-like carinar process elements) Addadi, L., 1993. Biological control of crystal texture: a wide- is, therefore, a feature which should not be neglected in spread strategy for adapting crystal properties to function. Science cladistic analysis; to the best of the authorÕs knowledge 259, 776–779. Candia Carnevali, D.M., Bonasoro, F., Melone, G., 1991. Micro- this character has not be commented upon previously, at structure and mechanical design in the lantern ossicles of the least with respect to characterization of more/less regular sea-urchin Paracentrotus lividus: a scanning electron primitive teeth in echinoids. microscope study. Bull. Zool. 58, 1–42. If the hypothesis that more prism-like carinar process Cavey, M.J., Maarkel,€ K., 1994. Echinoidea. In: Harrison, F.W., Chia, elements indicate more primitive teeth, then Maarkel€ Õs F.S. (Eds.), Microscopic Anatomy of Invertebrates. Echinoder- mata, vol. 14. Wiley-Liss, New York, pp. 345–400. (1969a) diagrams of Stomopneustes variolaris and Ar- Chen, C.P., Lawrence, J.M., 1986. The ultrastructure of the plumula of bacia lixula teeth (two different orders) might suggest the tooth of Lytechnius variegatus (Echinodermata: Echinoidea). which is more primitive. The carinar process elements of Acta Zool. (Stockh) 67, 33–41. Stomopneustes variolaris are shown laterally elongated Donnay, G., Pawson, D.L., 1969. X-ray diffraction studies of with respect to the equiaxed prisms as are the carinar echinoderm plates. Science 166, 1147–1150. Dowd, B.A., Campbell, G.H., Marr, R.B., Nagarkar, V., Tipnis, S., process elements of Arbacia lixula with respect to its Axe, L., Siddons, D.P., 1999. Developments in synchrotron X-ray prisms, but the latter appear more plate-like and those computed microtomography at the National Synchrotron Light of the former appear more block-like. This admittedly Source. SPIE Proc. 3772, 224–236. sparse data can only suggest that, based upon the car- Giesbrecht, W., 1880. Der feinere Bau der Seeigelzaahne.€ Morph. Jahrb. inar process morphology, Stomopneustes variolaris is 6, 79–105, Tf II-V. Holland, N.D., 1965. An autoradiographic investigation of tooth marginally more primitive than Arbacia lixula. Jensen renewal in the purple sea urchin (Strongylocentotus purpuratus). J. (1981) comes to a similar conclusion based upon anal- Exp. Zool. 158, 275–282. ysis of ambulacral plates, and Smith (1984) agrees with Interactive Data Language, 2003. Version 5.3 Research Systems. JensenÕs assignment. Checking whether the carinar Available from www.rsinc.com. process elements are in fact as indicated by Maarkel€ International Centre for Diffraction Data, Powder Diffraction File, Release 1999. (1969a) would be a useful application of microCT or Jackson, R.T., 1912. Phylogeny of the Echini with a revision of optical microscopy of polished sections. Palaeozoic species. Memoirs Boston Soc. Nat. History 7, 1–491, plates 1–72. Jensen, M., 1981. Morphology and classification of Euechinoidea Bronn, 1860—A cladistic analysis. Vidensk Meddr dansk naturh Acknowledgments Foren 143, 7–99. Kak, A.C., Slaney, M., 2001. Principles of Computerized Tomo- Research on sea urchin teeth was supported by graphic Imaging. Soc. Industrial Appl. Math., Philadelphia. NIDCR Grants DE07201 and DE01374. X-ray diffrac- Klinger, T.S., Lawrence, J.M., 1985. The hardness of the teeth of five species of echinoids (Echinodermata). J. Nat. Hist. 19, 917– tion and synchrotron microCT were performed at the 920. facilities of SRI-CAT of APS, supported by the DOE Kniprath, E., 1974. Ultrastructure and growth of the sea urchin tooth. under contract W-31-109-Eng-38. An unknown re- Calcif. Tiss. Res. 14, 211–228. viewerÕs suggestion, a reminder of something well- MacGregor, A.B., MacInnes, D.G., Marsland, E.A., 1956. The teeth known to the authors (refraction of synchrotron X-rays of the Echinoidea. Proc. Zool. Soc. London 127, 573–577, pl. 1–3. Maarkel,€ K., 1969a. Morphologie der Seeigelzaahne:€ II. Die gekielten redirects intensity and produces bands of unphysically Zaahne€ der Echinacea (Echinodermata, Echinoidea). Z. Morph. high and low l at sample edges) prompted another look Tiere. 66, 1–50. at the geometry of fragment c and is gratefully Maarkel,€ K., 1969b. Morphologie der Seeigelzaahne:€ III. Die Zaahne€ der acknowledged. Diadematoida und Echinothuroida (Echinodermata, Echinoidea). Z. Morph. Tiere. 66, 189–211. Maarkel,€ K., 1970. The tooth skeleton of Echinometra mathaei (Blainville) (Echinodermata, Echinoidea). Annot. Zool. Jap. 43, References 188–199. Maarkel,€ K., Titschack, H., 1969. Morphologie der Seeigelzaahne:€ I. Der Andrietti, F., Candia Carnevali, M.D., Wilkie, I.C., 1993. A biome- Zahn von Sylocidaris affinis (Phil.) (Echinodermata, Echinoidea). chanical comparison of the lantern of the cidarid sea-urchin Z. Morph. Tiere. 64, 179–200. 300 S.R. Stock et al. / Journal of Structural Biology 144 (2003) 282–300

Maarkel,€ K., Gorny, P., 1973. Zur functionellen Anatomie der Stock, S.R., Barss, J., Dahl, T., Veis, A., Almer, J.D., 2002b. X-ray Seeigelzaahne€ (Echinodermata, Echinoidea). Z. Morph. Tiere. 75, absorption microtomography (microCT) and small beam diffrac- 223–242. tion mapping of sea urchin teeth. J. Struct. Biol. 139, 1–12. Maarkel,€ K., Kubanek, F., Willgallis, A., 1971. Polykristalliner Calcit Stock, S.R., Nagaraja, S., Barss, J., Dahl, T., Veis, A., 2003a. X-ray bei Seeigeln (Echinodermata, Echinoidea). Z. Zellforsch. 119, 355– MicroCT study of pyramids of the sea urchin Lytechinus varieg- 377. atus. J. Struct. Biol. 141, 9–21. Maarkel,€ K., Gorny, P., Abraham, K., 1976. Microarchitecture of sea Stock, S.R., Barss, J., Dahl, T., Veis, A., Almer, J.D., De Carlo, F., urchin teeth. Fort. Zool. 24, 103–114. 2003b. Synchrotron X-ray studies of the keel of the short-spined Maarkel,€ K., Rooser,€ U., Mackenstedt, U., Klostermann, M., 1986. sea Urchin Lytechinus variegatus: absorption microtomography Ultrastructural investigation of matrix-mediated biomineralization (microCT) and small beam diffraction mapping. Calcif. Tiss. Int. in echinoids (Echinodermata, Echinoida). Zoomorph. 106, 232– 72, 555–566. 243. Su, X., Kamat, S., Heuer, A.H., 2000. The structure of sea urchin Nissen, H.U., 1969. Crystal orientation and plate structure in echinoid spines, large biogenic single crystals of calcite. J. Mater. Sci. 35, skeletal units. Science 166, 1150–1152. 5545–5551. NIST, Tables of X-ray Mass Attenuation Coefficients and Mass Towe, K., 1967. Echinoderm calcite: single crystal or polycrystalline Energy Absorption Coefficients from 1 keV to 20 MeV for Ele- aggregate. Science 157, 148–150. ments Z ¼ 1 to 92 and 48 Additional Substances of Dosimetric Veis, A., Barss, J., Rahima, M., Stock, S., 2002. Mineral related Interest, NISTIR 5632, July 2001. proteins of the sea urchin teeth: Lytechinus variegatus. Microsc. Raup, D.M., 1959. Crystallography of echinoid calcite. J. Geol. 67, Res. Tech. 59, 342–351. 661–674. Veis, D.J., Alberger, T.M., Clohisy, J., Rahima, M., Sabsay, B., Veis, Salter, S.J.S., 1861. On the Structure and Growth of the Tooth of A., 1986. Matrix proteins of the teeth of the sea urchin Lytechinus Echinus. Phil. Trans. Roy. Soc. (Lond.) 151, 387–407, pl. VI– variegatus. J. Exp. Zool. 240, 35–46. VIII. Venkateswara Rao, K.T., Yu, W., Ritchie, R.O., 1988. Fatigue crack Scanco Medical AG, 2003. Available from www.scanco.ch. propagation in aluminum-lithium alloy 2090: Part I. Long crack Schroeder, J.H., Dwornik, E.J., Papike, J.J., 1969. Primary protodo- behavior. Metall. Trans. A 19A, 549–560. lomite in echinoid skeletons. Bull. Geol. Soc. Am. 80, 1613– Wang, R.Z., 1998. Fracture toughness and interfacial design of a 1616. biological fiber-matrix ceramic composite in sea urchin teeth. J. Smith, A.B., 1981. Implications of lantern morphology for phylogeny Am. Ceram Soc. 81, 1037–1040. of post-palaeozoic echinoids. Palaeontology 24, 779–801. Wang, R.Z., Addadi, L., Weiner, S., 1997. Design strategies of sea Smith, A.B., 1984. Echinoid Palaeobiology. Allen and Unwin, Lon- urchin teeth: structure, composition and micromechanical relations don. p. 190. to function. Phil. Trans. Roy. Soc. (Lond.) B 352, 469–480. Stock, S.R., Dahl, T., Barss, J., Veis, A., Fezzaa, K., Lee, W.K., 2002a. Wang, Y., De Carlo, F., Mancini, D., McNulty, I., Tieman, B., Mineral phase microstructure in teeth of the short spined sea Bresnahan, J., Foster, I., Insley, J., Lane, P., von Laszewski, G., urchin (Lytechinus variegatus) studied with X-ray phase contrast Kesselman, C., Su, M.-H., Thiebaux, M., 2001. High-throughput imaging and with absorption microtomography. Adv. X-ray Anal. X-ray microtomography system at the Advanced Photon Source. 45, 133–138. Rev. Sci. Instrum. 72, 2062–2068.