《复合材料 Composites》课程教学资源（学习资料）第五章 陶瓷基复合材料_fibrous -2

Journal of Structural Biology 163 (2008)61-75 Contents lists available at ScienceDirect Journal of Structural Biology ELSEVIER journalhomepagewww.elsevier.com/locate/yjsbi Macromolecular structure of the organic framework of nacre in Haliotis rufescens Implications for growth and mechanical behavior Jiddu bezares a, Robert]. Asaro., Marilyn Hawley b of Structural Engineering. University of California, San Diego, Mail Code 0085, La jolla, CA 92093, US Materials Science and Technology Division, Los Alamos National laboratory, Los Alamos, NM 87545,USA ARTICLE IN FO A BSTRACT Article history. Ve have performed a macromolecular structural analysis of the interlamellar and intertabular parts of Received 8 January 2008 he organic framework of the nacreous part of the shell of Haliotis rufescens, including the identification Accepted 9 April 2008 of structural chitin. Using histochemical optical microscopy we have mapped the locations of carboxyl ates and sulfates of proteins and chitin on the surfaces and within the core of the interlamellar layers and Available online 25 April 2008 the intertabular matrix that together form the external organic matrix of site nacre. This extends the earlier work of Nudelmann et al. [ Nudelman, F. Gotliv, B.A. Addadi, L and Weiner, S. 2006. Mollusk Biomineralization tablet in nacre. J. Struct BioL. 153, 176-187] and Crenshaw and Ristedt [Crenshaw, MA, Ristedt, H Mollusk organic framework 1976. The histochemical localization of reactive groups in septal nacre from Nautilus pompilius. In: Omori, M, Watabe, N (Eds ) The Mechanisms of Biomineralization in Animals and Plants. Tokai University Press Toyko] on Nautilus pompilius. Our mapping identifies distinct regions, defined by the macromolecule groups, including what is proposed to be the sites of CacO, nucleation and that play a key role in nacre growth Using AFM scanning probe microscopy we have identified a fibrous core within the framework hat we as with chitin. The structural picture that is evolved is then used to develop a simple struc. ral model for the organic framework which is shown to be consistent with mechanical property mea- surements. The role of the intracrystalline matrix within the nacre tablets in mediating nacre's mechanical response is noted within the framework of our model e 2008 Elsevier Inc. All rights reserved. 1 Introduction macromolecular layout so that critical features such as the crystal nucleation site(s)and the chemical structural morphology that Biomineralization is a well regulated process within living control mechanical behavior may be understood in a more quanti- organisms, and involves control over, inter alia, the morphology fied manner. Possible extensions of such understanding include of mineral-biopolymer nano-composite structures, crystal nucle- biomimetic applications to, for example, nano-scale ceramic/poly ation within, and growth of, such structures, along with their poly- mer electronic devices and synthetic bone implants morph type(s) and crystallographic texture (e.g. Lowenstam and Crenshaw and Ristedt( 1976) took a unique approach to study Weiner, 1989: Simkiss and wilbur, 1989: Mann et al., 1989: Bae- ing the macromolecular structure in that they attempted a map- a wide array of other cases, mineralization appears to occur within microscopy. This approach has been recently pursued by nudel- a preformed 3-dimensional organic framework which acts as the man et al. (2006)who identify four different zones on the frame- template that provides the above mentioned control (e.g. Beve- work surface. This information was then shown to be of lander and Nakahara, 1969: Wada, 1972: Schaffer et al, 1997: immediate use for formulating more detailed and defensible mod- Nudelman et al., 2006 ). The organic framework thus mediates els for the biomineralization process. Crenshaw and ristedt's the growth, ie the"fabrication", of the mineralized nano-compos-(1976)and Nudelman et al. s(2006)work was performed on the e and, as it happens, is also key to what is seen to be a rather cephalopod Nautilus pompilius, whereas the latter performed com- cellent array of mechanical properties of the shell (e.g. Sarikaya parative study on the bivalve Atrina rigida. Here, we again et al., 1992: Evans et al, 2001). It is, accordingly, vital to under- the approach, but for the case of the gastropod Haliotis rufescens, stand the structure of the organic framework and in particular its with comparative study on N. pompilius so as to obtain a more comprehensive understanding of the commonality and variances among different members of the mollusk group. We confirm the mail address: asaro@ucsd. edu(r]. Asaro). general findings of Nudelman et al.(2006). now for H. rufescens, 047-8477s-see front matter o 2008 Elsevier Inc. All rights reserved. doi:10.1016/jsb.200804009

Macromolecular structure of the organic framework of nacre in Haliotis rufescens: Implications for growth and mechanical behavior Jiddu Bezares a , Robert J. Asaro a,*, Marilyn Hawley b aDepartment of Structural Engineering, University of California, San Diego, Mail Code 0085, La Jolla, CA 92093, USA b Materials Science and Technology Division, Los Alamos National laboratory, Los Alamos, NM 87545, USA article info Article history: Received 8 January 2008 Received in revised form 8 April 2008 Accepted 9 April 2008 Available online 25 April 2008 Keywords: Mollusk nacre Biomineralization Mollusk organic framework abstract We have performed a macromolecular structural analysis of the interlamellar and intertabular parts of the organic framework of the nacreous part of the shell of Haliotis rufescens, including the identification of structural chitin. Using histochemical optical microscopy we have mapped the locations of carboxyl￾ates and sulfates of proteins and chitin on the surfaces and within the core of the interlamellar layers and the intertabular matrix that together form the external organic matrix of composite nacre. This extends the earlier work of Nudelmann et al. [Nudelman, F., Gotliv, B.A., Addadi, L. and Weiner, S. 2006. Mollusk shell formation: mapping the distribution of organic matrix components underlying a single aragonite tablet in nacre. J. Struct. Biol. 153, 176–187] and Crenshaw and Ristedt [Crenshaw, M.A., Ristedt, H. 1976. The histochemical localization of reactive groups in septal nacre from Nautilus pompilius. In: Omori, M., Watabe, N. (Eds.) The Mechanisms of Biomineralization in Animals and Plants. Tokai University Press, Toyko] on Nautilus pompilius. Our mapping identifies distinct regions, defined by the macromolecular groups, including what is proposed to be the sites of CaCO3 nucleation and that play a key role in nacre growth. Using AFM scanning probe microscopy we have identified a fibrous core within the framework that we associate with chitin. The structural picture that is evolved is then used to develop a simple struc￾tural model for the organic framework which is shown to be consistent with mechanical property mea￾surements. The role of the intracrystalline matrix within the nacre tablets in mediating nacre’s mechanical response is noted within the framework of our model. 2008 Elsevier Inc. All rights reserved. 1. Introduction Biomineralization is a well regulated process within living organisms, and involves control over, inter alia, the morphology of mineral-biopolymer nano-composite structures, crystal nucle￾ation within, and growth of, such structures, along with their poly￾morph type(s) and crystallographic texture (e.g. Lowenstam and Weiner, 1989; Simkiss and Wilbur, 1989; Mann et al., 1989; Bae￾uerlein, 2000; Addadi and Weiner, 2001). In mollusk shells, among a wide array of other cases, mineralization appears to occur within a preformed 3-dimensional organic framework which acts as the template that provides the above mentioned control (e.g. Beve￾lander and Nakahara, 1969; Wada, 1972; Schaffer et al., 1997; Nudelman et al., 2006). The organic framework thus mediates the growth, i.e. the ‘‘fabrication”, of the mineralized nano-compos￾ite and, as it happens, is also key to what is seen to be a rather excellent array of mechanical properties of the shell (e.g. Sarikaya et al., 1992; Evans et al., 2001). It is, accordingly, vital to under￾stand the structure of the organic framework and in particular its macromolecular layout so that critical features such as the crystal nucleation site(s) and the chemical/structural morphology that control mechanical behavior may be understood in a more quanti- fied manner. Possible extensions of such understanding include biomimetic applications to, for example, nano-scale ceramic/poly￾mer electronic devices and synthetic bone implants. Crenshaw and Ristedt (1976) took a unique approach to study￾ing the macromolecular structure in that they attempted a map￾ping of the framework’s components using histochemical light microscopy. This approach has been recently pursued by Nudel￾man et al. (2006) who identify four different zones on the frame￾work surface. This information was then shown to be of immediate use for formulating more detailed and defensible mod￾els for the biomineralization process. Crenshaw and Ristedt’s (1976) and Nudelman et al.’s (2006) work was performed on the cephalopod Nautilus pompilius, whereas the latter performed com￾parative study on the bivalve Atrina rigida. Here, we again extend the approach, but for the case of the gastropod Haliotis rufescens, with comparative study on N. pompilius so as to obtain a more comprehensive understanding of the commonality and variances among different members of the mollusk group. We confirm the general findings of Nudelman et al. (2006), now for H. rufescens, 1047-8477/$ - see front matter 2008 Elsevier Inc. All rights reserved. doi:10.1016/j.jsb.2008.04.009 * Corresponding author. Fax: +1 858 534 6373. E-mail address: rasaro@ucsd.edu (R.J. Asaro). Journal of Structural Biology 163 (2008) 61–75 Contents lists available at ScienceDirect Journal of Structural Biology journal homepage: www.elsevier.com/locate/yjsbi

J. Bezares et al /Joumal of structural Biology 163 (2008)61-75 but extend the structural characterization by focusing as well on et al, 2001) but until our assays, to be reported herein, were com the location of chitin within the framework. Our findings concern- pleted no truly definitive and quantitative evidence for chitin had ng chitin provide an important, and previously not understood existed in H rufescens: reasons for this stem from, in particular, dif picture of the organic framework that has specific implications to ficulties in properly hydrolyzing the complete framework prior to the mechanical and structural performance of the shell. analysis. We have, however, recently succeeded in performing an Several mollusk shells, such as that of H. rufescens, contain an accurate analysis with the interesting result that for the as ex- outer prismatic(calcitic) layer(see e.g. Simkiss and wilbur, 1989 tracted insoluble framework we find 6. 4 wt% chitin, whereas in or Zaremba et al., 1996)and an inner nacre. The nacreous portion the extracted framework subjected to digestion with trypsin the is a biomineralized structure in which quasi-hexagonal CaCO3 (ara- content is higher at 6.9 wt%. This is a remarkably high chitin con- gonite) tiles are layered between, and bound to. a thin biopolymer tent even when compared to that found in fungal cell walls protein framework secreted by epithelial cells (eg. Simkiss and (Ruiz-Herrera, 1992: San-Blas and Calderone, 2004), and more spe- Wilbur, 1989; Belcher and Gooch, 2000: Addadi et al, 2006; Rous- cifically as compared with the much lower contents reported by seau et al., 2005). The structure and crystallography of the ceramic Poulicek(1983)in gastropods. Our assay is, however, consistent tiles mollusk nacre have been studied via X-ray diffraction, SEM, with the chitin contents reported by Goffinet and Jeuniaux, 1979 TEM and AFM(see e.g. Wada, 1961, 1968: Wise, 1970: Crenshaw for other mollusk species. There are soluble proteins that are re- and Ristedt, 1976: Mutvei, 1979: Weiner et al., 1983: Sarikaya moved during demineralization and some have been et al, 1995; and Manne et al, 1994). Tiles are crystallographically (see e.g. Belcher et al, 1996). They too are rich in Asp and textured aragonite with a thickness in the range 300-500 nm and Ca, and can play a vital role in mineralization( Falini et al diameters in the range 4-10 um Nacre tiles are arranged in nearly Belcher et al., 1996; Addadi et al., 2001; Gotliv et al, 2003) parallel lamellae separated by thin interlamellar layers of biopoly Of particular interest is the structure of the CaCO3 nucleation mer framework with a thickness in the shell of approximately sites. Crenshaw and Ristedt(1976). following Crenshaw(1972), used mella the tiles grow laterally and meet at what become polygonal the outlines of the tiles, and mapped the location of sulfates, carbox boundaries separated by an intertabular matrix; examples are ylates, and calcium binding sites within them. Wada (1980)later shown below. Nacre tiles often respond, via diffraction, as single confirmed the presence of high concentrations of sulfur in the cen- crystalline tablets but are known to contain an organic intracrys- tral region, thus suggesting that the nucleation site was located talline matrix. Recently, for example, Rousseau et al. (2005)have there. Here, we use AFM methods combined with histochemical performed AFM imaging, in tapping mode, and tEM dark field fluorescent microscopy to map such structure in H rufescens and in- imaging of nacre tablets in the oyster Pinctada maxima and pro- clude chitin within our maps. Our mapping is then used to confirm vided evidence for a continuous intracrystalline matrix surround- and extend models for nacre growth and for developing an approach g coherent nanograins that comprise individual tablets. Their to modeling the shell's mechanical properties. This provides the path ults suggest, among other things, a pathway for modeling the to biomimetics and bio-duplication of synthetic materials. mechanical response of nacre that we use below in suggesting a reliminary model for the interlamellar layers. Likewise, Oaki 2 Materials and methods and Imai (2005)describe a hierarchical structure of nacre in the pearl oyster Pinctada fucata in which individual tiles are seen to 2.1. Materials be composed of nano-scale"building blocks"(ie. nano-crystals) surrounded by an organic matrix(ie the intracrystalline matrix). Fresh shells of both H. rufescens and N. pompilius were obtained The AFM images shown by Rousseau et al. (2005)can be used to from The Abalone Farm Inc. of Monterey, CA and were stored dry at demonstrate that the stiffness of individual tiles should be less 4C. The nacreous portion of the shells were removed by slowly than that of monolithic CaCO3 and, in fact, just on the basis of a grinding off the prismatic (outer) part and then washing the nacre simple rule of mixtures should be on the order of at least 10% less. sections in deionized water(Di). Sections varied in size, but often This estimate is based on the apparent thickness of the intracrys- were up to 20 x 20 mm in cross section and up to 2 mm thick. talline matrix as seen, for example, in the phase image of Fig 3b For the tensile specimens described in Section 4, the sections were f rousseau et al. (2005). At the same time, the intracrystalline ma- larger and up to 40 mm long. Fifteen live H rufescens were also pro- trix would impart increased toughness to the structure via the en- vided by the abalone Farm Inc and were kept in large tanks with ergy absorptive capability of a visco-elastic matrix. continuously running sea water at the Scripps Institution of Ocean- The biopolymer framework has previously been studied primar- ography at UCSD. These were used for the making of"flat pearl ily through optical microscopy (e.g. Crenshaw and Ristedt, 1976: inter alia, as shown in Fig. Ic Flat pearls were prepared by inserting Gregoire, 1957, 1972)and biochemical studies of amino acid com- 5 mm glass slides under the mantle of live H rufescens and allow- position(see e.g. Wheeler and Sykes, 1984: Cariolou and Morse, ing approximately three weeks for nacre to form. The slides were 1988: Simkiss and Wilbur. 1989: Addadi and Weiner. 1985 and then removed and washed in Dl. Aizenberg et al, 1999). Quite limited study using either TEM(Wei- ner et al, 1983; Levi-Kalisman et al, 2001)or AFM( Schaffer et al., 2.2. Demineralization 1997: Manne et al., 1994: Rousseau et al, 2005) has been per formed to date; only the latter was performed on H. rufescens After washing in Dl, shell sections were demineralized either in As is well established, the matrix associated with nacre tissue is Ethylene Diamine Tetraacetic Acid (EDtA, Sigma), or by cation-ex approximately 75-80 wt% protein and balance carbohydrate with change resin( Dowex 50W 50-100 mesh, Sigma). Demineral- me glycoproteins(see e.g. Addadi and Weiner, 1985). Our own ization was carried out at 20C for times that depended on the assays are consistent with this and a typical amino acid analysis method and on the size and thickness of the shell fragments. Decal shows that our H. rufescens insoluble nacre framework contains cification by ion exchange resin was adapted from Gotliv et al residues that are rich in Asp and Glu and are acidic as reported(2003). Nacre fragments were placed in sections of dialysis tubing by others. approximately 20% of the residues are Asp and 4% Glu. (MwCO 3500, 19 mm flat width, Fisher). Unfixed samples were There have been reports of chitin in the framework of several mol- placed in dialysis membranes filled with DI water while fixed sam- ner et al.1983; Poulicek, 1983: Goffinet and Jeuniaux, 1979 zep i- ples were placed in dialysis membranes pc) solution. 10 mm long lusks other than H. rufescer (Weiner and Traub, 1980, 1984; W taining a 4% formalde- hyde, 0.5% cetylpyridinium chloride(CI

but extend the structural characterization by focusing as well on the location of chitin within the framework. Our findings concern￾ing chitin provide an important, and previously not understood picture of the organic framework that has specific implications to the mechanical and structural performance of the shell. Several mollusk shells, such as that of H. rufescens, contain an outer prismatic (calcitic) layer (see e.g. Simkiss and Wilbur, 1989 or Zaremba et al., 1996) and an inner nacre. The nacreous portion is a biomineralized structure in which quasi-hexagonal CaCO3 (ara￾gonite) tiles are layered between, and bound to, a thin biopolymer protein framework secreted by epithelial cells (e.g. Simkiss and Wilbur, 1989; Belcher and Gooch, 2000; Addadi et al., 2006; Rous￾seau et al., 2005). The structure and crystallography of the ceramic tiles mollusk nacre have been studied via X-ray diffraction, SEM, TEM and AFM (see e.g. Wada, 1961, 1968; Wise, 1970; Crenshaw and Ristedt, 1976; Mutvei, 1979; Weiner et al., 1983; Sarikaya et al., 1995; and Manne et al., 1994). Tiles are crystallographically textured aragonite with a thickness in the range 300–500 nm and diameters in the range 4–10 lm. Nacre tiles are arranged in nearly parallel lamellae separated by thin interlamellar layers of biopoly￾mer framework with a thickness in the shell of approximately 30 nm (e.g. Checa and Rodriguez-Navarro, 2005) . Within each la￾mella the tiles grow laterally and meet at what become polygonal boundaries separated by an intertabular matrix; examples are shown below. Nacre tiles often respond, via diffraction, as single crystalline tablets but are known to contain an organic intracrys￾talline matrix. Recently, for example, Rousseau et al. (2005) have performed AFM imaging, in tapping mode, and TEM dark field imaging of nacre tablets in the oyster Pinctada maxima and pro￾vided evidence for a continuous intracrystalline matrix surround￾ing coherent nanograins that comprise individual tablets. Their results suggest, among other things, a pathway for modeling the mechanical response of nacre that we use below in suggesting a preliminary model for the interlamellar layers. Likewise, Oaki and Imai (2005) describe a hierarchical structure of nacre in the pearl oyster Pinctada fucata in which individual tiles are seen to be composed of nano-scale ‘‘building blocks” (i.e. nano-crystals) surrounded by an organic matrix (i.e. the intracrystalline matrix). The AFM images shown by Rousseau et al. (2005) can be used to demonstrate that the stiffness of individual tiles should be less than that of monolithic CaCO3 and, in fact, just on the basis of a simple rule of mixtures should be on the order of at least 10% less. This estimate is based on the apparent thickness of the intracrys￾talline matrix as seen, for example, in the phase image of Fig. 3b of Rousseau et al. (2005). At the same time, the intracrystalline ma￾trix would impart increased toughness to the structure via the en￾ergy absorptive capability of a visco-elastic matrix. The biopolymer framework has previously been studied primar￾ily through optical microscopy (e.g. Crenshaw and Ristedt, 1976; Gregoire, 1957, 1972) and biochemical studies of amino acid com￾position (see e.g. Wheeler and Sykes, 1984; Cariolou and Morse, 1988; Simkiss and Wilbur, 1989; Addadi and Weiner, 1985 and Aizenberg et al., 1999). Quite limited study using either TEM (Wei￾ner et al., 1983; Levi-Kalisman et al., 2001) or AFM (Schaffer et al., 1997; Manne et al., 1994; Rousseau et al., 2005) has been per￾formed to date; only the latter was performed on H. rufescens . As is well established, the matrix associated with nacre tissue is approximately 75–80 wt% protein and balance carbohydrate with some glycoproteins (see e.g. Addadi and Weiner, 1985). Our own assays are consistent with this and a typical amino acid analysis shows that our H. rufescens insoluble nacre framework contains residues that are rich in Asp and Glu and are acidic as reported by others. Approximately 20% of the residues are Asp and 4% Glu. There have been reports of chitin in the framework of several mol￾lusks other than H. rufescens (Weiner and Traub, 1980, 1984; Wei￾ner et al., 1983; Poulicek, 1983; Goffinet and Jeuniaux, 1979; Zentz et al., 2001) but until our assays, to be reported herein, were com￾pleted no truly definitive and quantitative evidence for chitin had existed in H. rufescens; reasons for this stem from, in particular, dif- ficulties in properly hydrolyzing the complete framework prior to analysis. We have, however, recently succeeded in performing an accurate analysis with the interesting result that for the as ex￾tracted insoluble framework we find 6.4 wt% chitin, whereas in the extracted framework subjected to digestion with trypsin the content is higher at 6.9 wt%. This is a remarkably high chitin con￾tent even when compared to that found in fungal cell walls (Ruiz-Herrera, 1992; San-Blas and Calderone, 2004), and more spe￾cifically as compared with the much lower contents reported by Poulicek (1983) in gastropods. Our assay is, however, consistent with the chitin contents reported by Goffinet and Jeuniaux, 1979 for other mollusk species. There are soluble proteins that are re￾moved during demineralization and some have been isolated, (see e.g. Belcher et al., 1996). They too are rich in Asp and Glu, bind Ca++, and can play a vital role in mineralization (Falini et al., 1996; Belcher et al., 1996; Addadi et al., 2001; Gotliv et al., 2003). Of particular interest is the structure of the CaCO3 nucleation sites.Crenshaw and Ristedt (1976), followingCrenshaw (1972), used the fact that once demineralized, the interlamellar matrix reveals the outlines of the tiles, and mapped the location of sulfates, carbox￾ylates, and calcium binding sites within them. Wada (1980) later confirmed the presence of high concentrations of sulfur in the cen￾tral region, thus suggesting that the nucleation site was located there. Here, we use AFM methods combined with histochemical fluorescent microscopy to map such structure in H. rufescens and in￾clude chitin within our maps. Our mapping is then used to confirm and extend models for nacre growth and for developing an approach tomodeling the shell’smechanical properties. This provides the path to biomimetics and bio-duplication of synthetic materials. 2. Materials and methods 2.1. Materials Fresh shells of both H. rufescens and N. pompilius were obtained from The Abalone Farm Inc. of Monterey, CA and were stored dry at 4 C. The nacreous portion of the shells were removed by slowly grinding off the prismatic (outer) part and then washing the nacre sections in deionized water (DI). Sections varied in size, but often were up to 20 20 mm in cross section and up to 2 mm thick. For the tensile specimens described in Section 4, the sections were larger and up to 40 mm long. Fifteen live H. rufescens were also pro￾vided by the Abalone Farm Inc. and were kept in large tanks with continuously running sea water at the Scripps Institution of Ocean￾ography at UCSD. These were used for the making of ‘‘flat pearls”, inter alia, as shown in Fig. 1c. Flat pearls were prepared by inserting 5 mm glass slides under the mantle of live H. rufescens and allow￾ing approximately three weeks for nacre to form. The slides were then removed and washed in DI. 2.2. Demineralization After washing in DI, shell sections were demineralized either in Ethylene Diamine Tetraacetic Acid (EDTA, Sigma), or by cation-ex￾change resin (Dowex 50 W 8 50–100 mesh, Sigma). Demineral￾ization was carried out at 20 C for times that depended on the method and on the size and thickness of the shell fragments. Decal￾cification by ion exchange resin was adapted from Gotliv et al. (2003). Nacre fragments were placed in sections of dialysis tubing (MWCO 3500, 19 mm flat width, Fisher). Unfixed samples were placed in dialysis membranes filled with DI water while fixed sam￾ples were placed in dialysis membranes containing a 4% formalde￾hyde, 0.5% cetylpyridinium chloride (CPC) solution. 10 mm long 62 J. Bezares et al. / Journal of Structural Biology 163 (2008) 61–75

J. Bezares et al/ Journal of Structural Biology 163(2008)61-75 a b d Fig. 1.(a and b)sEM images of fractured nacre from H rufescens illustrating tiles on nearly parallel lamella. The" terrace consisting of one interlamellar layer of nacre is arls grown on a glass slid image of a cross section of H. rufescens organic matrix, demineralized in EDTA, illustrating individual and apparently porous interlamellar layer p of each stack(d)SEM inserted into the mantle of a live red abalone(described belo sections of tubing containing unfixed specimens were placed in )keer in an Eppendorf centrifuge. The supernatant was removed and re- DI containing 750 ml of pre-washed resin. A stir bar was used to placed by 1 ml of Hepes buffer containing 200 ug/mL proteinase-K the resin in constant suspension at room tempe Sigma). The pellets were sonicated in this solution for 10 min and was replaced once a day. Decalcification was verified by infrared left to incubate for 2 h at room temperature. The samples were then spectroscopy. Fixed samples placed in sections of dialysis tubing centrifuged for 10 min at 12,000 rpm. The supernatant was removed were allowed to demineralize in 50 ml conical tubes filled with and the pellets were washed with Hepes buffer. 15 ml of resin and topped off with the formalyn/CPC solution. The tubes were placed on a tilt platform such that the resin remained 2.3.3. Collagenase in suspension during decalcification. The demineralization was As described in Schaffer et al. (1997). sheets of demineralized complete after 1-4 weeks For starting fragments with a thickness tissue were incubated for 2 h in 1 mL of 200 ug/ mL collagen of approximately 0.5 mm, typically 1 week-10 days was sufficient. combined with 200 HM CaCl2 in 5 mM Hepes buffer, pH 7.5. Sam- Thicker fragments, e.g. those with thickness up to 1 mm, required ples were centrifuged for 10 min at 12,000 rpm and sonicated in DI the longer times. Following Crenshaw and Ristedt (1976), decalcifi- for 10 min cation using EDta was performed by placing nacre fragments in 50 ml conical tubes filled with 1 M EDTA, pH8, at room temperature 2.3. 4 N-Glycosidase F under gentle shaking. Samples were also decalcified in EDTA con- N-Glycosidase F(PNGase F, New England Bio Labs, P0704S) is an taining 4% formaldehyde and 0.5% CPC (Williams and Jackson, amidase that cleaves between the innermost glucosamine and 1956. After demineralization, all samples were extensively washed asparagine residues of numerous N-linked glycoproteins. One cen- in Di water ove any remaining EDTA, formaldehyde or CPC. timetre square sections of demineralized tissue were separated Interlamellar sheets were separated using fine tweezers under an into approximately 1 um thick, interlamellar sheets. The sheets ptical microscope. Demineralization times ranged from 1 week to were incubated in 1 uL of 10 x glycoprotein denaturing buffer 10 days depending on the thickness of the starting fragment. and Di water to make a total reaction volume of 750 HL. The sam- ples and buffer were combined in 1.5 mL centrifuge tubes. The 2.3. Enzym tubes containing the samples were placed in water heated to 100C for 15 min. A total reaction volume of 20 HL was prepared 2.3.1. Trypsin by adding 2 HL of 10 x G7 reaction buffer, 2 uL of 10% NP40, 2 HL Ten microgram of wet interlamellar sheets were incubated for N-Glycosidase F, and 14 HL of DI water. The samples were incu- 24 h at room temperature in 0.1 M ammonium bicarbonate buffer, bated at room temperature in this solution for 2 h. pH 8.0, containing 1 mg/mL trypsin(Invitrogen). 2.4. Staining methods 23.2 Proteinase-K Following Schaffer et al. (1997). interlamellar sheets of deminer- 2. 4. 1. Calcofluor white alized shell were sonicated for 10 min in 1 mLof 5 mM Hepes buffer, Calcofluor white(CW)is a fluorescent brightener that binds to pH7.5.The samples were then centrifuged at 12,000 rpm for 10 min B-1-3 and B-1-4 polysaccharides as are found in cellulose and

sections of tubing containing unfixed specimens were placed in 4 l of DI containing 750 ml of pre-washed resin. A stir bar was used to keep the resin in constant suspension at room temperature and the water was replaced once a day. Decalcification was verified by infrared spectroscopy. Fixed samples placed in sections of dialysis tubing were allowed to demineralize in 50 ml conical tubes filled with 15 ml of resin and topped off with the formalyn/CPC solution. The tubes were placed on a tilt platform such that the resin remained in suspension during decalcification. The demineralization was complete after 1–4 weeks. For starting fragments with a thickness of approximately 0.5 mm, typically 1 week–10 days was sufficient. Thicker fragments, e.g. those with thickness up to 1 mm, required the longer times. Following Crenshaw and Ristedt (1976), decalcifi- cation using EDTA was performed by placing nacre fragments in 50 ml conical tubes filled with 1 M EDTA, pH 8, at room temperature under gentle shaking. Samples were also decalcified in EDTA con￾taining 4% formaldehyde and 0.5% CPC (Williams and Jackson, 1956). After demineralization, all samples were extensively washed in DI water to remove any remaining EDTA, formaldehyde or CPC. Interlamellar sheets were separated using fine tweezers under an optical microscope. Demineralization times ranged from 1 week to 10 days depending on the thickness of the starting fragment. 2.3. Enzymatic digestion 2.3.1. Trypsin Ten microgram of wet interlamellar sheets were incubated for 24 h at room temperature in 0.1 M ammonium bicarbonate buffer, pH 8.0, containing 1 mg/mL trypsin (Invitrogen). 2.3.2. Proteinase-K Following Schaffer et al. (1997), interlamellar sheets of deminer￾alized shell were sonicated for 10 min in 1 mL of 5 mM Hepes buffer, pH 7.5. The samples were then centrifuged at 12,000 rpm for 10 min in an Eppendorf centrifuge. The supernatant was removed and re￾placed by 1 mL of Hepes buffer containing 200 lg/mL proteinase-K (Sigma). The pellets were sonicated in this solution for 10 min and left to incubate for 2 h at room temperature. The samples were then centrifuged for 10 min at 12,000 rpm. The supernatant was removed and the pellets were washed with Hepes buffer. 2.3.3. Collagenase As described in Schaffer et al. (1997), sheets of demineralized tissue were incubated for 2 h in 1 mL of 200 lg/ mL collagenase combined with 200 lM CaCl2 in 5 mM Hepes buffer, pH 7.5. Sam￾ples were centrifuged for 10 min at 12,000 rpm and sonicated in DI for 10 min. 2.3.4. N-Glycosidase F N-Glycosidase F (PNGase F, New England Bio Labs, P0704S) is an amidase that cleaves between the innermost glucosamine and asparagine residues of numerous N-linked glycoproteins. One cen￾timetre square sections of demineralized tissue were separated into approximately 1 lm thick, interlamellar sheets. The sheets were incubated in 1 lL of 10 glycoprotein denaturing buffer and DI water to make a total reaction volume of 750 lL. The sam￾ples and buffer were combined in 1.5 mL centrifuge tubes. The tubes containing the samples were placed in water heated to 100 C for 15 min. A total reaction volume of 20 lL was prepared by adding 2 lL of 10 G7 reaction buffer, 2 lL of 10% NP40, 2 lL N-Glycosidase F, and 14 lL of DI water. The samples were incu￾bated at room temperature in this solution for 2 h. 2.4. Staining methods 2.4.1. Calcofluor white Calcofluor white (CW) is a fluorescent brightener that binds to b-1-3 and b-1-4 polysaccharides as are found in cellulose and Fig. 1. (a and b) SEM images of fractured nacre from H. rufescens illustrating tiles on nearly parallel lamella. The ‘‘terrace” consisting of one interlamellar layer of nacre is shown at higher magnification in (b), where the black arrow points to a central region discussed below and referred to in Mutvei (1979). (c) Flat pearls grown on a glass slide inserted into the mantle of a live red abalone (described below). Note the ‘‘stack of coins” arrangement with a smaller tile (or tiles) nucleated at the top of each stack. (d) SEM image of a cross section of H. rufescens organic matrix, demineralized in EDTA, illustrating individual and apparently porous interlamellar layers. J. Bezares et al. / Journal of Structural Biology 163 (2008) 61–75 63

J. Bezares et al /Joumal of structural Biology 163 (2008)61-75 chitin, respectively(Maeda and Ishida, 1967: Peters and Latka, 2.5.3. Silver intensification 1986: Hayat, 1993: Harrington and Hageage, 2003). Staining of Aurion SE-EM silver enhancement reagent(Aurion) was pre- demineralized tissue was performed as follows. A 10% potassium pared immediately before use. Samples treated with WGA-gold hydroxide reagent was prepared by dissolving 10 g of KOH in complexes were rinsed in DI water three times and placed in the 90 mL of DI water to which 10 mL of glycerin were added. a second enhancing reagent for 4 min after which they were washed in five CW reagent was prepared by dissolving 0. 1 g of fluorescent bright- changes of DI water. ener 28(Sigma)in 100 mL of DI water under gentle heating. Inter lamellar sheets, approximately 1-2 um thick were mounted on 2.5.4. Cytochemical controls slides. Two drops of each reagent were added to the samples. After The specificity to chitin of WGA-gold complexes was deter 4 min the samples were rinsed in water, air dried and mounted mined by pre-incubating the complexes in N, N, N-triacetylchitotri- Entellan mounting medium(Merck). ose. Another set of samples was incubated in unlabeled WGA (0.5 mg/mL). Specificity to chitin was further determined by a 2. 4.2. Colloidal iron omparative analysis using FITC-WGA(Molecular Probes)and cw Colloidal iron staining was performed following methods out- which binds specifically to B-1-3 and B-1-4 glucosides ie cellulose lined in Pearse(1968). Stock colloidal iron solution was prepared and chitin, (Maeda and Ishida, 1967; Peters and Latka, 1986; Lukes by stirring in 29% ferric chloride to boiling water. Once the solution et al., 1993). Of these two polysaccharides, chitin is the only one rned dark red it was allowed to cool to 20C. The solution was which is labeled by both WGa and CV then dialyzed three times for 24 h against a volume of DI 5x that of the stock solution. The reagent was stored at 20C. The working 2.5.5. Fluorescein-WGA colloidal iron solution, pH 1.8, had a shelf life of 24 h. This solution Following El Gueddari et al.(2002), interlamellar sheets were consisted of Dl, glacial acetic acid, and stock colloidal iron solution incubated with 2% w/v bovine serum albumin(BSa)in PBS for combined at a ratio of 18: 12: 10. As in Nudelman et al. (2006), the 30 min at 20C. Samples were tissue samples were submerged in the working solution for 1 h, Tween20 and incubated in FITC-WGA(O 1 mg/mL in PBS, 1% w/v rinsed thoroughly with 12% acetic acid, incubated with 5% potas- BSA)for 1 h Samples were washed three times with PBS/Tween20 sium ferrocyanide/5% HCl(1: 1)for 20 min, and washed with water. washed once in DI water, air dried on glass cover slips, and Samples were mounted as described above. mounted in Entellan 2. 43. Aminoacridine 2.6. Scanning electron microscopy Following Nudelman et al. (2006), thin layers of interlamellar tissue were incubated in 1% N-(3 dimethylaminopropyl )-N-ethyl- Samples were dehydrated in an ethanol series and critical point carbodiimide hydrochloride(EDC)in a 20 mM phosphate buffer dried with liquid carbon dioxide. Samples were mounted with ca at pH 4.5. The samples were washed three times with 0.2 M borate bon tabs and sputter coated with 15 nm of chromium. Samples buffer, pH 8.5, and incubated for 12 h in aminoacridone(1 mg/ were immediately examined using a Philips XL30 ESEM operated 100 ml)in the borate buffer. The samples were rinsed with water and mounted as described above 2.7 Immunohistoch .5. Wheat germ agglut Gotliv et al.(2003) produced polyclonal antibodies raised .5.1. Probe preparata against soluble aragonite-nucleating proteins from the bivalve A. Wheat germ agglutinin (WGA)is a lectin that binds to se- rigida. the proteins were earlier found to nucleate aragonite by quences of three B-1-4 linked N-acetyl-D-glucosamine residues as Falini et al. (1996). Nudelman et al. (2006)found positive labeling well as to sialic acid residues. As a result WGa has a high affini with these antibodies in both A rigida and N. pompilius and conse- chitin and to a lesser extent to some glycoproteins. Colloidal quently to explore the possibility of similar effects in H. rufescens gold can be complexed to WGA for the purpose of locating chitin we used similar immunohistochemical labeling procedures. Poly on the biopolymer layer remaining after shell demineralization. clonal antibodies were generously supplied to us by s. Weiner Colloidal gold particles typically used for high resolution TEM stud -(2007). ies can be enhanced using silver enhancement reagents. This al- Interlamellar sheets from H rufescens were incubated in serum lows for the imaging of these particles using standard SEM and containing polyclonal antibodies for 1 h. The serum was diluted adapted from King et al.(1987)and Geoghegan and Ackerman ing The samples were washed twice for 5 min with PBS containing (1997). WGA-gold complexes were prepared as follows. 0.75 mg Tween20(0.05% w/). Samples were then incubated for 40 min in WGA in 1 HL of distilled water were combined with 25 mL of col- the secondary antibody, rhodamine conjugated goat-anti-rab- loidal gold, pH 9.9. The gold particles were of a nominal size of bit(ackson ImmunoResearch, diluted 1: 100 in PBS). Before imag 10 nm(Sigma). The complexes were stirred for 3 min after which ing the specimens were washed three times for 5 min with PBS 1% polyethylene glycol (PEG)was added. After 5 min the reagent and Tween20(0.05% w/v) to remove unbound antibodies. After was centrifuged at 60,000g in a Beckman ultracentrifuge for 1 h rinsing the samples in DI water, the samples were air dried on cov at 4 C The supernatant was removed and the sedimented WGA- er slips and mounted in Entellan. As a control, samples incubated gold complexes were resuspended to 5 mL with phosphate buf- in a pre-immune serum were prepared as above. fered saline(PBS)pH 8, containing 0.2 mg/mL of PEG 2.8. Optical microscopy 2.5.2. Labeling procedure Fixed and unfixed interlamellar tissue was sectioned it Samples were observed using a Nikon Eclipse 80i optical approximately 1 mm thick Samples were incubated in scope equipped with a Photometrics CoolSNAPez digital gold complexes for 30 min. The WGA-gold complexes Calcofluor white and aminoacridone-stained samples wer aI mIcro- t a dilution of 1: 100 in PBS, pH 7.0. After labeling th using a 11003 V3 filter set, while samples labeled with FITC and were washed in six changes of PBS, pH 7.2 rhodamine where viewed using a 41001 filter set(Chroma)

chitin, respectively (Maeda and Ishida, 1967; Peters and Latka, 1986; Hayat, 1993; Harrington and Hageage, 2003). Staining of demineralized tissue was performed as follows. A 10% potassium hydroxide reagent was prepared by dissolving 10 g of KOH in 90 mL of DI water to which 10 mL of glycerin were added. A second CW reagent was prepared by dissolving 0.1 g of fluorescent bright￾ener 28 (Sigma) in 100 mL of DI water under gentle heating. Inter￾lamellar sheets, approximately 1–2 lm thick were mounted on slides. Two drops of each reagent were added to the samples. After 4 min the samples were rinsed in water, air dried and mounted in Entellan mounting medium(Merck). 2.4.2. Colloidal iron Colloidal iron staining was performed following methods out￾lined in Pearse (1968). Stock colloidal iron solution was prepared by stirring in 29% ferric chloride to boiling water. Once the solution turned dark red it was allowed to cool to 20 C. The solution was then dialyzed three times for 24 h against a volume of DI 5 that of the stock solution. The reagent was stored at 20 C. The working colloidal iron solution, pH 1.8, had a shelf life of 24 h. This solution consisted of DI, glacial acetic acid, and stock colloidal iron solution combined at a ratio of 18:12:10. As in Nudelman et al. (2006), the tissue samples were submerged in the working solution for 1 h, rinsed thoroughly with 12% acetic acid, incubated with 5% potas￾sium ferrocyanide/5% HCl (1:1) for 20 min, and washed with water. Samples were mounted as described above. 2.4.3. Aminoacridone Following Nudelman et al. (2006), thin layers of interlamellar tissue were incubated in 1% N-(3 dimethylaminopropyl)-N0 -ethyl￾carbodiimide hydrochloride (EDC) in a 20 mM phosphate buffer at pH 4.5. The samples were washed three times with 0.2 M borate buffer, pH 8.5, and incubated for 12 h in aminoacridone (1 mg/ 100 ml) in the borate buffer. The samples were rinsed with water and mounted as described above. 2.5. Wheat germ agglutinin–colloidal gold (WGA–gold) 2.5.1. Probe preparation Wheat germ agglutinin (WGA) is a lectin that binds to se￾quences of three b-1-4 linked N-acetyl-D-glucosamine residues as well as to sialic acid residues. As a result WGA has a high affinity to chitin and to a lesser extent to some glycoproteins. Colloidal gold can be complexed to WGA for the purpose of locating chitin on the biopolymer layer remaining after shell demineralization. Colloidal gold particles typically used for high resolution TEM stud￾ies can be enhanced using silver enhancement reagents. This al￾lows for the imaging of these particles using standard SEM and optical microscopic techniques. Our procedures were closely adapted from King et al. (1987) and Geoghegan and Ackerman (1997). WGA–gold complexes were prepared as follows. 0.75 mg WGA in 1 lL of distilled water were combined with 25 mL of col￾loidal gold, pH 9.9. The gold particles were of a nominal size of 10 nm(Sigma). The complexes were stirred for 3 min after which 1% polyethylene glycol (PEG) was added. After 5 min the reagent was centrifuged at 60,000g in a Beckman ultracentrifuge for 1 h at 4 C. The supernatant was removed and the sedimented WGA– gold complexes were resuspended to 5 mL with phosphate buf￾fered saline (PBS) pH 8, containing 0.2 mg/mL of PEG. 2.5.2. Labeling procedure Fixed and unfixed interlamellar tissue was sectioned into 1 cm2 approximately 1 mm thick. Samples were incubated in the WGA– gold complexes for 30 min. The WGA–gold complexes were used at a dilution of 1:100 in PBS, pH 7.0. After labeling the samples were washed in six changes of PBS, pH 7.2. 2.5.3. Silver intensification Aurion SE-EM silver enhancement reagent(Aurion) was pre￾pared immediately before use. Samples treated with WGA–gold complexes were rinsed in DI water three times and placed in the enhancing reagent for 4 min after which they were washed in five changes of DI water. 2.5.4. Cytochemical controls The specificity to chitin of WGA–gold complexes was deter￾mined by pre-incubating the complexes in N,N0 ,N00-triacetylchitotri￾ose. Another set of samples was incubated in unlabeled WGA (0.5 mg/mL). Specificity to chitin was further determined by a comparative analysis using FITC-WGA(Molecular Probes) and CW which binds specifically to b-1-3 and b-1-4 glucosides i.e. cellulose and chitin, (Maeda and Ishida, 1967; Peters and Latka, 1986; Lukes et al., 1993). Of these two polysaccharides, chitin is the only one which is labeled by both WGA and CW. 2.5.5. Fluorescein-WGA Following El Gueddari et al. (2002), interlamellar sheets were incubated with 2% w/v bovine serum albumin (BSA) in PBS for 30 min at 20 C. Samples were washed three times with PBS/ Tween20 and incubated in FITC-WGA(0.1 mg/mL in PBS, 1% w/v BSA) for 1 h. Samples were washed three times with PBS/Tween20, washed once in DI water, air dried on glass cover slips, and mounted in Entellan. 2.6. Scanning electron microscopy Samples were dehydrated in an ethanol series and critical point dried with liquid carbon dioxide. Samples were mounted with car￾bon tabs and sputter coated with 15 nm of chromium. Samples were immediately examined using a Philips XL30 ESEM operated at 10 kV. 2.7. Immunohistochemistry Gotliv et al. (2003) produced polyclonal antibodies raised against soluble aragonite-nucleating proteins from the bivalve A. rigida. The proteins were earlier found to nucleate aragonite by Falini et al. (1996). Nudelman et al. (2006) found positive labeling with these antibodies in both A. rigida and N. pompilius and conse￾quently to explore the possibility of similar effects in H. rufescens we used similar immunohistochemical labeling procedures. Poly￾clonal antibodies were generously supplied to us by S. Weiner (2007). Interlamellar sheets from H. rufescens were incubated in serum containing polyclonal antibodies for 1 h. The serum was diluted 1:25 in PBS containing 0.25% w/v BSA to block non-specific bind￾ing. The samples were washed twice for 5 min with PBS containing Tween20(0.05% w/v). Samples were then incubated for 40 min in the secondary antibody, rhodamine conjugated goat-anti-rab￾bit(Jackson ImmunoResearch, diluted 1:100 in PBS). Before imag￾ing the specimens were washed three times for 5 min with PBS and Tween20(0.05% w/v) to remove unbound antibodies. After rinsing the samples in DI water, the samples were air dried on cov￾er slips and mounted in Entellan. As a control, samples incubated in a pre-immune serum were prepared as above. 2.8. Optical microscopy Samples were observed using a Nikon Eclipse 80i optical micro￾scope equipped with a Photometrics CoolSNAPez digital camera. Calcofluor white and aminoacridone-stained samples were viewed using a 11003 V3 filter set, while samples labeled with FITC and rhodamine where viewed using a 41001 filter set(Chroma). 64 J. Bezares et al. / Journal of Structural Biology 163 (2008) 61–75

J Bezares et al/ Journal of Structural Biology 163(2008)61-75 2.9. Chitin assays 2.11. Mechanical properties and mode Our method was based on the evaluation of glucosamine con tent found in demineralized tissue samples. 1 mg of tissue(dry ished flat sections of the nacreous parts of whole shells. Steel t/- Specimens were"dog bone "type and were machined from weight) was resuspended in 1 ml of 6 M HCl and hydrolyzed at plates were first made of the specimens to ensure dimensional 100C for 17 h, together with N-acetylglucosamine(GlcNAc) as uniformity across the gauge section and the machining was per a standard. Samples were then dried and resuspended in 1 m formed via slow grinding to minimize damage to the edges. The DI water. The quantity of glucosamine released by the hydrolysis gauge sections 14 mm in length, 2 mm wide, and 0.5 mm of 100 HL of this material was determined as follows. An equal thick Axial strain was measured using an extensometer with a res- volume of 4%(v/v)acetylacetone in 1.5 M Na, CO3 was added, olution in strain of approximately 5 x 10-6. Testing was performed and the preparation heated at 100C for 20 min. Samples were in an Instron 5565 material test system, in stroke(ie. displace- then diluted with 700 HL 96% ethanol and incubated for 1 hr at ment) control; this system has a mechanically controlled cross 23C, after 100 HL Ehrlich's reagent(1.6 g/ml p-dimethylamin head and thus is extremely stable. Axial strain rates ranged from benzaldehyde, 6M HCl. 50% ethanol) was added. Samples were 3 to-10-4. Modeling was performed for the purpose incubated for 15 min at 65C before the absorbance was read extracting estimates for the shear stiffness of the interlamellar lay at 520 nm ers using a"shear lag "analytical model that has been verified by finite element analysis 2.10. AFM methods Atomic force microscopy(AFM)techniques were used to 3. Results characterize the protein network structure of decalcified abalone shells.Intermittent contact( tapping)mode of operation was used 3. 1 SEM observations to obtain both surface topographic and phase information In tap- ing mode the"diving board"shaped, low force-constant cantile- The following SEM observations are briefly described most ver, which has a very sharp silicon tip located at the end on the ticularly to provide additional perspective on the structure of nacre bottom side, is oscillated at its resonance frequency by a piezo- in our H. rufescens. Fig. la and b shows SEM images of the tile electric element located on the cantilever holder. which is cross-section of a fractured shell that illustrate the optimal inter mounted on a piezoelectric tube scanner. This intermittent con- digitated nacre brick-wall structure. Additional description of the tact mode of operation and low force-constant cantilevers were structure, including the interlamellar and intertabular layers may riginally developed specifically to minimize damage to soft, eas- be seen in the images and are described in the caption. Interlamel- ily damaged samples. A laser beam, focused onto the backside of regions are identified by the horizontal white arrows in Fig. 1b the cantilever end, is reflected via an adjustable mirror onto an the intertabular regions are identified by the vertical white arrows. ptical sensor. As the cantilever tip is rastered back and forth The black arrow indicates a central region observed in all tiles over the sample surface, variations in sample height result deflection of the cantilever from its rest position. These deflec- an image of individual demineralized interlamellar layers of the tions cause variations in the optical sensor signal that are con- external organic matrix. Such images were obtained from critically verted to a voltage that is applied to the z electrode of the point dried matrix tissue and show thickness in the range 80- scanner, causing it to expand or contract, and thus raise or lower 100 nm; in the shell the interlamellar layers are 30 nm as just the cantilever back to its undeflected position. Since the chang noted. The layers appear porous in Fig. 1d and this will be further in the scanner dimensions are calibrated in nanometers per volt, elucidated when our AFM results are shown later. Fig. 1c shows the the changes in applied voltage are mapped as local variations in growth of"flat pearls"grown on the surface of glass slides inserted sample height superimposed on an x-y grid (scan size)set by under the mantle of live H rufescens; such observations will be re- the user the sharp tip allows one to obtain nanometer resolution ferred to in our discussion of nacre growth. A noteworthy feature of surface fea of the flat pearls is the appearance of more than one tile nucleated Typical resonance frequencies for the commercial AFM cantile- atop a tile growing laterally on a layer just below. An example of vers(Nano Devices, approximately 125 um long x 45 um such multiple tile nucleation is indicated by the black arrow in idex 4 um thick) used in this study were around 300 kHz. In Fig 1c. dition to monitoring changes in the cantilever deflection posi tion, simultaneously phase imaging, which captures the phase 3.2. Optical histochemical microscopy lag angle between the drive signal and actual cantilever oscillation was used to obtain maps of variations in local properties such as 3. 1. Calcofluor white staining stiffness Therefore there is a one-to-one correlation between the Calcofluor white is a fluorophore that binds to glycans and opographic data and the corresponding phase information. fibrillar polysaccharides such as chitin(e.g. Herth, 1980: Ramasw The aFM is equipped with a special cantilever wet cell holder, amy et al. 1997). Organic matrices from H rufescens were stained which was used to image samples under in vitro conditions. Sam- and imaged after demineralization in EDTA and by cation-ex les were mounted on steel pucks that are held in place on the change resin. In both cases strong binding was detected as shown AFM sample stage by a magnet buried in the base. Since the hy in Fig. 2c and d. H perspective, however, we first show SEM were cut large enough to be held in place using the wet cell o-ring plane, Fig. 2a and b. These images may be compared to those re- seal. Only thin continuous sections were used for imaging to pre- ported by Mutvei (1979, 1977), where similarities and differences vent small sections from coming lose during imaging and interfer- may be noted ing with the imaging process. Ambient imaging was used for dried As in Mutvei's(1977)SEM images, the intertabular matrix, su and stained samples. In all cases, tapping mode imaging was used. rounding the tiles, is clearly revealed as is the appearance of an evi- We used Veeco Metrology Nanoscope llla controllers with a D3000 dently organic ring-like structure organized around what was the microscope to image the dried samples and a Multimode micro- tiles'centers. Etching, in the form of"pits", occurs throughout scope for the wet cell imaging the surface of the interlamellar layers, but we do not observe the

2.9. Chitin assays Our method was based on the evaluation of glucosamine con￾tent found in demineralized tissue samples. 1 mg of tissue (dry weight) was resuspended in 1 ml of 6 M HCl and hydrolyzed at 100 C for 17 h, together with N-acetylglucosamine (GlcNAc) as a standard. Samples were then dried and resuspended in 1 ml DI water. The quantity of glucosamine released by the hydrolysis of 100 lL of this material was determined as follows. An equal volume of 4% (v/v) acetylacetone in 1.5 M Na2CO3 was added, and the preparation heated at 100 C for 20 min. Samples were then diluted with 700 lL 96% ethanol and incubated for 1 hr at 23 C, after 100 lL Ehrlich’s reagent (1.6 g/ml p-dimethylamino￾benzaldehyde, 6 M HCl, 50% ethanol) was added. Samples were incubated for 15 min at 65 C before the absorbance was read at 520 nm. 2.10. AFM methods Atomic force microscopy (AFM) techniques were used to characterize the protein network structure of decalcified abalone shells. Intermittent contact (tapping) mode of operation was used to obtain both surface topographic and phase information. In tap￾ping mode, the ‘‘diving board” shaped, low force-constant cantile￾ver, which has a very sharp silicon tip located at the end on the bottom side, is oscillated at its resonance frequency by a piezo￾electric element located on the cantilever holder, which is mounted on a piezoelectric tube scanner. This intermittent con￾tact mode of operation and low force-constant cantilevers were originally developed specifically to minimize damage to soft, eas￾ily damaged samples. A laser beam, focused onto the backside of the cantilever end, is reflected via an adjustable mirror onto an optical sensor. As the cantilever tip is rastered back and forth over the sample surface, variations in sample height result in deflection of the cantilever from its rest position. These deflec￾tions cause variations in the optical sensor signal that are con￾verted to a voltage that is applied to the z electrode of the scanner, causing it to expand or contract, and thus raise or lower the cantilever back to its undeflected position. Since the changes in the scanner dimensions are calibrated in nanometers per volt, the changes in applied voltage are mapped as local variations in sample height superimposed on an x-y grid (scan size) set by the user. The sharp tip allows one to obtain nanometer resolution of surface features. Typical resonance frequencies for the commercial AFM cantile￾vers (NanoDevices, approximately 125 lm long 45 lm wide 4 lm thick) used in this study were around 300 kHz. In addition to monitoring changes in the cantilever deflection posi￾tion, simultaneously phase imaging, which captures the phase lag angle between the drive signal and actual cantilever oscillation, was used to obtain maps of variations in local properties such as stiffness. Therefore there is a one-to-one correlation between the topographic data and the corresponding phase information. The AFM is equipped with a special cantilever wet cell holder, which was used to image samples under in vitro conditions. Sam￾ples were mounted on steel pucks that are held in place on the AFM sample stage by a magnet buried in the base. Since the hy￾drated samples could not be fixed to the sample by tape, samples were cut large enough to be held in place using the wet cell o-ring seal. Only thin continuous sections were used for imaging to pre￾vent small sections from coming lose during imaging and interfer￾ing with the imaging process. Ambient imaging was used for dried and stained samples. In all cases, tapping mode imaging was used. We used Veeco Metrology Nanoscope IIIa controllers with a D3000 microscope to image the dried samples and a Multimode micro￾scope for the wet cell imaging. 2.11. Mechanical properties and modeling Specimens were ‘‘dog bone” type and were machined from pol￾ished flat sections of the nacreous parts of whole shells. Steel tem￾plates were first made of the specimens to ensure dimensional uniformity across the gauge section and the machining was per￾formed via slow grinding to minimize damage to the edges. The gauge sections were 14 mm in length, 2 mm wide, and 0.5 mm thick. Axial strain was measured using an extensometer with a res￾olution in strain of approximately 5 106 . Testing was performed in an Instron 5565 material test system, in stroke (i.e. displace￾ment) control; this system has a mechanically controlled cross head and thus is extremely stable. Axial strain rates ranged from 103 to 104 . Modeling was performed for the purpose of extracting estimates for the shear stiffness of the interlamellar lay￾ers using a ‘‘shear lag” analytical model that has been verified by finite element analysis. 3. Results 3.1. SEM observations The following SEM observations are briefly described most par￾ticularly to provide additional perspective on the structure of nacre in our H. rufescens. Fig. 1a and b shows SEM images of the tile cross-section of a fractured shell that illustrate the optimal inter￾digitated nacre brick-wall structure. Additional description of the structure, including the interlamellar and intertabular layers may be seen in the images and are described in the caption. Interlamel￾lar regions are identified by the horizontal white arrows in Fig. 1b; the intertabular regions are identified by the vertical white arrows. The black arrow indicates a central region observed in all tiles which had been referred to earlier by Mutvei (1979). Fig. 1d shows an image of individual demineralized interlamellar layers of the external organic matrix. Such images were obtained from critically point dried matrix tissue and show thickness in the range 80– 100 nm; in the shell the interlamellar layers are 30 nm as just noted. The layers appear porous in Fig. 1d and this will be further elucidated when our AFM results are shown later. Fig. 1c shows the growth of ‘‘flat pearls” grown on the surface of glass slides inserted under the mantle of live H. rufescens; such observations will be re￾ferred to in our discussion of nacre growth. A noteworthy feature of the flat pearls is the appearance of more than one tile nucleated atop a tile growing laterally on a layer just below. An example of such multiple tile nucleation is indicated by the black arrow in Fig. 1c. 3.2. Optical histochemical microscopy 3.2.1. Calcofluor white staining Calcofluor white is a fluorophore that binds to glycans and fibrillar polysaccharides such as chitin (e.g. Herth, 1980; Ramasw￾amy et al., 1997). Organic matrices from H. rufescens were stained and imaged after demineralization in EDTA and by cation-ex￾change resin. In both cases strong binding was detected as shown in Fig. 2c and d. For perspective, however, we first show SEM images of demineralized nacre taken normal to the interlamellar plane, Fig. 2a and b. These images may be compared to those re￾ported by Mutvei (1979, 1977), where similarities and differences may be noted. As in Mutvei’s (1977) SEM images, the intertabular matrix, sur￾rounding the tiles, is clearly revealed as is the appearance of an evi￾dently organic ring-like structure organized around what was the tiles’ centers. Etching, in the form of ‘‘pits”, occurs throughout the surface of the interlamellar layers, but we do not observe the J. Bezares et al. / Journal of Structural Biology 163 (2008) 61–75 65

J. Bezares et al /Joumal of structural Biology 163 (2008)61-75 b Fig. 2.(a and b)are SEM images of demineralized organic framework in H rufescens. Note that the center of each tile imprint is more deeply etched; the apparen ccumulation of organic material at the center. (c) An epi-fluorescent micrograph of H rufescens stained with Cw following demineralization in EDTA The insert is of N. ompilius demineralized and stained in the same manner. (d)Same as(c)except with demineralization with ion exchange. The lower contrast in(d)is due to a low exposure radial by Mutvei(1977)in his study as is most evident in Fig. 3c(see the white arrow ) lamellar layers at locations of ses the"fibers"of such structures are seen to emanate indicating an organic/ mineral he intertabular matrix as indicated by the arrows in accumulation there as reported by Mutvei (1979, 1977). The cen- Fig 3b. The diameters of the fibers was only approximately deter rs of the lamellar layers appear to contain significant amounts mined and appear to lie in the range of 30 nm. Fluorescein stain of mineral that is removed rapidly by both EDta and ion exchange ing, shown in Fig. 3d, reveals again the strong labeling of the along with organic material that is exposed by the demineraliza- intertabular matrix and the pattern of projection of structure ema- tion process. As mineral is removed, and organic material is ex- nating from it. Later we present evidence of a chitin core and sug- posed, the appearance of the latter is that of aring-like" gest that the lighter staining with Cw at the center is due to structure. Fig 2c and d shows the effect of Cw staining on H rufes- exposure, and not an excess accumulation, of chitin beneath the cens:the insert in Fig. 2c is an additional image obtained from N. protein layers of the framework Referring back to Fig 2c and d pompilius showing a similar pattern of staining as for H rufescens. we now suggest that the lighter staining with CW at the center de- trong binding of CW is observed at the intertabular matrix in both scribed there was due to the exposure and not an excess accumu- cases and, at a lesser intensity, at the very center of the tiles. The lation, of chitin beneath the protein layers of the framework. center illumination is not in the form of a ring-like structure but rather appears as a quite localized area of Cw binding. This ev 3. 2.3. Aminoacridone staining dence, combined with supporting results using WGA-gold and Aminoacridone is a fluorescent compound that binds to car- FITC-WGA conjugates, indicates that these locations contain signif- boxyl groups in the presence of carbodiimide(edc). H. rufescens tissue, demineralized in EDTA, was imaged in both the fixed and note that the staining at the intertabular matrix often ap- unfixed states as shown in Fig 4a and b, respectively. As the sam- pears in the form of a"double-layer"as, for example, in Fig. 2c. This ples are observed in epifluorescence, labeling appears white in the might suggest the possibility of a two layered structure recently images. The unfixed samples showed labeling in the intertabular described by Marin et al. (2007)in the bivalve Pinna nobilis. We be matrix, but little or no labeling was observed within the tile im lieve, however, that in this case the structure is caused by the prints and in particular, at the center of the tiles, Fig. 4. This is physical topography of the intertabular matrix as shown and de- in contrast with results reported by Nudelman et al.(2006)for N scribed below pompilius who reported definite labeling at the center region for unfixed samples. On the other hand for fixed samples, Fig. 4. 3. 2. 2. WGA-gold staining labeling was observed albeit faint(see insert and the white arrow Fig. 3a-c shows optical of H rufescens obtained after head) as also noted by Nudelman et al. (2006)for N pompilius. The demineralization with EDTA and labeled with WGA-gold; Fig 3d central regions typically appeared dark, and in fact black in the un- is enhanced with fluorescein ted to WGA after demineral- fixed samples. We suspect from this that we were not suffering PNGase F Distinct staining is from possible artifacts caused by residual EDTA as noted by Nu evident on the intertabular matrix with heavy accumulations of man et al. (2006)and by, for example Wheeler et al. (1987). Very colloidal gold seen outlining, in many cases, complete intertabular low levels of fluorescence were detected in regions between the boundaries. Labeling of fibrous structures is also seen within the center and the intertabular regions. For further confirmation of

radial organic membranes reported by Mutvei (1977) in his study of the gastropod Gibbula. The interlamellar layers at locations of the tile centers typically etch deeply indicating an organic/mineral accumulation there as reported by Mutvei (1979, 1977). The cen￾ters of the lamellar layers appear to contain significant amounts of mineral that is removed rapidly by both EDTA and ion exchange along with organic material that is exposed by the demineraliza￾tion process. As mineral is removed, and organic material is ex￾posed, the appearance of the latter is that of a ‘‘ring-like” structure. Fig. 2c and d shows the effect of CW staining on H. rufes￾cens; the insert in Fig. 2c is an additional image obtained from N. pompilius showing a similar pattern of staining as for H. rufescens. Strong binding of CW is observed at the intertabular matrix in both cases and, at a lesser intensity, at the very center of the tiles. The center illumination is not in the form of a ring-like structure, but rather appears as a quite localized area of CW binding. This evi￾dence, combined with supporting results using WGA–gold and FITC-WGA conjugates, indicates that these locations contain signif￾icant amounts of chitin. We note that the staining at the intertabular matrix often ap￾pears in the form of a ‘‘double-layer” as, for example, in Fig. 2c. This might suggest the possibility of a two layered structure recently described by Marin et al. (2007) in the bivalve Pinna nobilis. We be￾lieve, however, that in this case the structure is caused by the physical topography of the intertabular matrix as shown and de￾scribed below. 3.2.2. WGA–gold staining Fig. 3a–c shows optical images of H. rufescens obtained after demineralization with EDTA and labeled with WGA–gold; Fig. 3d is enhanced with fluorescein conjugated to WGA after demineral￾ization in EDTA and digestion with PNGase F. Distinct staining is evident on the intertabular matrix with heavy accumulations of colloidal gold seen outlining, in many cases, complete intertabular boundaries. Labeling of fibrous structures is also seen within the tile imprints as is most evident in Fig. 3c (see the white arrow). In many cases the ‘‘fibers” of such structures are seen to emanate out from the intertabular matrix as indicated by the arrows in Fig. 3b. The diameters of the fibers was only approximately deter￾mined and appear to lie in the range of 30 nm. Fluorescein stain￾ing, shown in Fig. 3d, reveals again the strong labeling of the intertabular matrix and the pattern of projection of structure ema￾nating from it. Later we present evidence of a chitin core and sug￾gest that the lighter staining with CW at the center is due to exposure, and not an excess accumulation, of chitin beneath the protein layers of the framework. Referring back to Fig. 2c and d we now suggest that the lighter staining with CW at the center de￾scribed there was due to the exposure, and not an excess accumu￾lation, of chitin beneath the protein layers of the framework. 3.2.3. Aminoacridone staining Aminoacridone is a fluorescent compound that binds to car￾boxyl groups in the presence of carbodiimide (EDC). H. rufescens tissue, demineralized in EDTA, was imaged in both the fixed and unfixed states as shown in Fig. 4a and b, respectively. As the sam￾ples are observed in epifluorescence, labeling appears white in the images. The unfixed samples showed labeling in the intertabular matrix, but little or no labeling was observed within the tile im￾prints and, in particular, at the center of the tiles, Fig. 4. This is in contrast with results reported by Nudelman et al. (2006) for N. pompilius who reported definite labeling at the center region for unfixed samples. On the other hand for fixed samples, Fig. 4a, labeling was observed albeit faint (see insert and the white arrow head) as also noted by Nudelman et al. (2006) for N. pompilius. The central regions typically appeared dark, and in fact black in the un- fixed samples. We suspect from this that we were not suffering from possible artifacts caused by residual EDTA as noted by Nudel￾man et al. (2006) and by, for example Wheeler et al. (1987). Very low levels of fluorescence were detected in regions between the center and the intertabular regions. For further confirmation of Fig. 2. (a and b) are SEM images of demineralized organic framework in H. rufescens. Note that the center of each tile imprint is more deeply etched; the apparent accumulation of organic material at the center. (c) An epi-fluorescent micrograph of H. rufescens stained with CW following demineralization in EDTA. The insert is of N. pompilius demineralized and stained in the same manner. (d) Same as (c) except with demineralization with ion exchange. The lower contrast in (d) is due to a low exposure time of 300 ms. 66 J. Bezares et al. / Journal of Structural Biology 163 (2008) 61–75

J. Bezares et al/ Journal of Structural Biology 163(2008)61-75 b c mineralization with EDTA, without fixing, and labeled with WGA-gold. Strong labeling of the intertabular matrix is evident as is the appearance of a fibrous network(small arrow heads in(b)and (c)) Gold labeling appears white in(a-c(d) Epifluorese i EDTA demineralized sample, digested, with PNGase F, and stained with FITC-WGA b 10 0 d 0 Fig 4. (a)Epifluorescent images of H rufescens demineralized using EDTA, fixed, and stained with amir one.(b) Unfixed H rufescens imaged after staining. Note the faint but distinct labeling at the center in the fixed sample, shown more clearly in the insert to(a) (c)Epifluorescenct images of H rufescens demineralized using ion exchange, fixed, and stained with aminoacridone. The white arrows indicate staining at the center of tile imprints. (d) Bright field image of H rufescens demineralized using ion exchange. our images being artifact free, we stained and imaged sar sert and arrows). For reference we show a bright field image in demineralized using ion exchange; a typical example for a Fig 4d. We had used such images to compliment our infrared spe pecimen is shown in Fig. 4c. The images are quite similar to troscopy to assess the degree of demineralization and to view the obtained with fixed samples demineralized in EDTA Staining was general topography of the tissue samples. We judged the sample as observed at the intertabular matrix and at the tiles centers (see in- shown in Fig. 4d to be demineralized but we note that additional

our images being artifact free, we stained and imaged samples demineralized using ion exchange; a typical example for a fixed specimen is shown in Fig. 4c. The images are quite similar to those obtained with fixed samples demineralized in EDTA. Staining was observed at the intertabular matrix and at the tiles’ centers (see in￾sert and arrows). For reference we show a bright field image in Fig. 4d. We had used such images to compliment our infrared spec￾troscopy to assess the degree of demineralization and to view the general topography of the tissue samples. We judged the sample as shown in Fig. 4d to be demineralized, but we note that additional Fig. 3. (a–c) are images obtained after demineralization with EDTA, without fixing, and labeled with WGA–gold. Strong labeling of the intertabular matrix is evident as is the appearance of a fibrous network (small arrow heads in (b) and (c)). Gold labeling appears white in (a–c). (d) Epifluorescence of an EDTA demineralized sample, digested, with PNGase F, and stained with FITC-WGA. Fig. 4. (a) Epifluorescenct images of H. rufescens demineralized using EDTA, fixed, and stained with aminoacridone. (b) Unfixed H. rufescens imaged after staining. Note the faint but distinct labeling at the center in the fixed sample, shown more clearly in the insert to (a). (c) Epifluorescenct images of H. rufescens demineralized using ion exchange, fixed, and stained with aminoacridone. The white arrows indicate staining at the center of tile imprints. (d) Bright field image of H. rufescens demineralized using ion exchange. J. Bezares et al. / Journal of Structural Biology 163 (2008) 61–75 67

J. Bezares et al /Joumal of structural Biology 163 (2008)61-75 and N pompilius using polyclonal antibodies raised against these 3. 4. Colloidal iron staining molecules. Here we apply the same immunohistochemical proce taining with colloidal iron is commonly to label dure to label the demineralized interlamellar/intertabular matrix acidic polysaccharides via the binding of iron acidic of H rufescens to explore the possibility that these mineral nucle- groups, including carboxylates and sulfates. To distinguish the ating macromolecules may be present, and thereby aide in locating labeling of sulfates from carboxylates, staining is done at a ph potential aragonite nucleating sites. Fig. 6a and b shows that in H. rufescens labeling is quite sharp performed at a pH 1.8, so that only sulfates were labeled. Strong and highly localized as found by Nudelman et al. (2006)in both labeling was observed within ring-like structures in the center of A rigida and N. pompilius. Labeling is distinct at the intertabular the tile imprints, as clearly seen in both Fig 5a and b. In fact, the matrix, and perhaps sharpest at the"vertex junctions"of the tile stained ring-like structures are most pronounced in H. rufescens imprints. Labeling is also seen at the imprint centers as is espe- demineralized in EDTA and unfixed contrary to other reports, viz. cially clear in Fig. 6b. In fact, labeling of multiple sites within the Crenshaw and Ristedt, 1976: Nudelman et al, 2006, concerned interior of a tile imprint is often observed this feature will be dis with N pompilius, that note the lack of labeling in unfixed samples. cussed in view of the already noted observation of multiple tile Still another feature that contrasts with the later's findings is the nucleation as observed in flat pearl nucleation and growth(see lear absence of labeling in the intertabular matrix. The ring-like Fig. 1c). Fig. 6d shows a typical image from a sample treated with structure of those regions believed to contain acidic sulfates is, only pre-immune serum where no staining is detected at all. Our however, in agreement with the results and conclusions of Nudel- results are consistent with those of Nudelman et al. (2006), with man et al, 2006, concerning similar structure in N. pompilius. the exception that they observe slight labeling of the intertabular Samples were also demineralized and digested with trypsin matrix in N pompilius with pre-immune serum Fig 6c is an image proteinase K, stained and imaged Fig 5c and d shows results fol- obtained with aminoacridone staining as shown earlier in Fig. 4. lowing staining with colloidal iron and Cw, respectively. labeling that is shown here for perspective. The patterns of staining are by colloidal iron was essentially eradicated after digestion indicat quite similar suggesting co-location of molecules rich in carboxyl- ing that the macromolecules responsible for colloidal iron binding ates and aragonite nucleating macromolecules. We note again the were removed. On the other hand, labeling of the intertabular re- double-layer like structure(e.g. Fig. 6a and c)at the intertabular gions and in the center regions by Cw was hardly affected. matrix that we attribute to the physical topography of this matrix. 3. 2.5. Immunohistochemical staining 3.3. AFM analysis Acidic proteins containing high concentrations of aspartic and/ or glutamic acids are an integral part of the organic matrix as al Both high and low resolution scans, in tapping mode, were per eady noted. Acidic polysaccharides, including some that are sul- formed over areas ranging from 500 x 500 nm up to 30 x 30 um. fated have also been identified. Falini et al.(1996) and the latter covering the range of our optical fluorescent views. Scans et al.(2003)have isolated such acidic macromolecules the that they were made on specimens fixed and unfixed and with various de- showed helped nucleate aragonite when introduced n grees of digestion and or staining. The images reveal fibrous struc- in vitro assembly of (Loligo B)chitin and(Bombyx mori) silk fibroin. tures characterized by at least two distinct size scales, the smaller a b 10μm-10m置 d Fig. 5. H. rufescens stained with colloidal iron in the range 1.8& pH< 1.9. (a)Demineralization performed using ion exchange and fixation.(b) Demineralization performed using EDTA and unfixed. (c)H rut demineralized in EDTA, unfixed, and digested in Trypsin+ proteinase K(d)Same as (c) but stained with Cw

care was required to assess the completeness of demineralization via ion exchange. 3.2.4. Colloidal iron staining Staining with colloidal iron is commonly performed to label acidic polysaccharides via the binding of iron to charged acidic groups, including carboxylates and sulfates. To distinguish the labeling of sulfates from carboxylates, staining is done at a pH low enough to protonate the carboxylates. Thus our staining was performed at a pH 1.8, so that only sulfates were labeled. Strong labeling was observed within ring-like structures in the center of the tile imprints, as clearly seen in both Fig. 5a and b. In fact, the stained ring-like structures are most pronounced in H. rufescens demineralized in EDTA and unfixed contrary to other reports, viz. Crenshaw and Ristedt, 1976; Nudelman et al., 2006, concerned with N. pompilius, that note the lack of labeling in unfixed samples. Still another feature that contrasts with the later’s findings is the near absence of labeling in the intertabular matrix. The ring-like structure of those regions believed to contain acidic sulfates is, however, in agreement with the results and conclusions of Nudel￾man et al., 2006, concerning similar structure in N. pompilius. Samples were also demineralized and digested with trypsin + proteinase K, stained and imaged. Fig. 5c and d shows results fol￾lowing staining with colloidal iron and CW, respectively. Labeling by colloidal iron was essentially eradicated after digestion indicat￾ing that the macromolecules responsible for colloidal iron binding were removed. On the other hand, labeling of the intertabular re￾gions and in the center regions by CW was hardly affected. 3.2.5. Immunohistochemical staining Acidic proteins containing high concentrations of aspartic and/ or glutamic acids are an integral part of the organic matrix as al￾ready noted. Acidic polysaccharides, including some that are sul￾fated have also been identified. Falini et al. (1996) and Gotliv et al. (2003) have isolated such acidic macromolecules that they showed helped nucleate aragonite when introduced into an in vitro assembly of (Loligo b) chitin and (Bombyx mori) silk fibroin. Gotliv et al. (2003) isolated their matrix molecules from A. rigida and Nudelman et al. (2006) showed positive results on both Atrina and N. pompilius using polyclonal antibodies raised against these molecules. Here we apply the same immunohistochemical proce￾dure to label the demineralized interlamellar/intertabular matrix of H. rufescens to explore the possibility that these mineral nucle￾ating macromolecules may be present, and thereby aide in locating potential aragonite nucleating sites. Fig. 6a and b shows that in H. rufescens labeling is quite sharp and highly localized as found by Nudelman et al. (2006) in both A. rigida and N. pompilius. Labeling is distinct at the intertabular matrix, and perhaps sharpest at the ‘‘vertex junctions” of the tile imprints. Labeling is also seen at the imprint centers as is espe￾cially clear in Fig. 6b. In fact, labeling of multiple sites within the interior of a tile imprint is often observed; this feature will be dis￾cussed in view of the already noted observation of multiple tile nucleation as observed in flat pearl nucleation and growth (see Fig. 1c). Fig. 6d shows a typical image from a sample treated with only pre-immune serum where no staining is detected at all. Our results are consistent with those of Nudelman et al. (2006), with the exception that they observe slight labeling of the intertabular matrix in N. pompilius with pre-immune serum. Fig. 6c is an image obtained with aminoacridone staining as shown earlier in Fig. 4, that is shown here for perspective. The patterns of staining are quite similar suggesting co-location of molecules rich in carboxyl￾ates and aragonite nucleating macromolecules. We note again the double-layer like structure (e.g. Fig. 6a and c) at the intertabular matrix that we attribute to the physical topography of this matrix. 3.3. AFM analysis Both high and low resolution scans, in tapping mode, were per￾formed over areas ranging from 500 500 nm up to 30 30 lm, the latter covering the range of our optical fluorescent views. Scans were made on specimens fixed and unfixed and with various de￾grees of digestion and/or staining. The images reveal fibrous struc￾tures characterized by at least two distinct size scales, the smaller Fig. 5. H. rufescens stained with colloidal iron at pH in the range 1.8 6 pH 6 1.9. (a) Demineralization performed using ion exchange and fixation. (b) Demineralization performed using EDTA and unfixed. (c) H. rufescens demineralized in EDTA, unfixed, and digested in Trypsin + proteinase K. (d) Same as (c) but stained with CW. 68 J. Bezares et al. / Journal of Structural Biology 163 (2008) 61–75

J. Bezares et al/ Journal of Structural Biology 163(2008)61-75 10 um b 10 Hm 10μmd Fig. 6.(a and b)Images obtained from unfixed H. ns, demineralized with EDtA, and labeled with polyclonal antibodies. (c) An image obtained on a sample identical to that used in (a and b) using aminoacridone staining shown for perspective. (d)An image of unfixed H rufescens, demineralize EDTA, and labeled with pre-immune characterized by fibers, or fiber bundles, and the larger by an open interlamellar layers are not"porous"per se but are composed of network that creates pore-like openings a fibrous core or network, that is embedded in and sandwiched be- Fig. 7 shows two typical scans over a larger area for an unfixed tween less organized organic material. Individual fibers appear to and undigested specimen of H. rufescens stained with calcofluor have diameters in the range 5-10 nm and run for lengths of at least white. Fig. 7b shows a topograph image where Fig. 7a shows an 1 Hm, although to date their length distribution has not been quan elevation profile of that region. As in the histochemical images titatively determined. Fig 8c shows a higher resolution scan which shown above, these large scans reveal the outlines of tiles on the provides additional evidence of the fibrous network and suggests urface of the organic interlamellar layers. Where the intertabular hat the fibers may indeed be organized in nearly parallel bundles. matrix has"collapsed", and where calcofluor white has evidently In both fixed and unfixed samples there are examples where the fi- bound the elevations can be as high as 250 nm above the locations bers appear to be so organized and examples that suggest a more f what were the tile/interlamellar layer interfaces. As we have random orientation. Our analysis of fiber orientation via 2D Fourier determined that the tile diameter in our H. rufescens is spectra. however, did not reveal long range order but did indicate 7+0.5 um, the images leave little doubt as to representing tile out- short range order over distances of an imprint diameter. This lines. The central regions are depressed relative to the surrounding would suggest that beneath a tile the fibers are ordered possibly tile imprint as noted by the two central arrows. This depressed re- as suggested by Weiner and Traub(1980). Fiber organization as ob gion is collocated with the central regions described in Fig. 2a and c served here is also consistent with the recent observations of Rous- where it is believed that mineral is preferentially removed during seau et al.(2005) studying P. maxima who suggest that both the demineralization. What is also notable from the height profile of intracrystalline and intertabular matrices are highly organized In Fig 7a is that the top edges of the intertabular layers are not flat the case of the intracrystalline matrix, they propose that it"reacts but contain height differences of the order of 50 nm. This itself as a single crystal"; in the case of the interlamellar matrix, they may account for the appearance of a double layer"effect as dis- conclude that it is"highly crystalized". Our observations such as Issed earlier in connection with Fig 2c and d As also noted above, shown in Fig. 8c and also Fig. 8d for H. rufescens are supportive the tile imprint centers are depressed, and the scans show these of this proposal. Complete resolution, however, of the orientation depressions are nearly 100 nm. The structure at the imprint cen- distribution awaits further study ters is discussed below Fig 8d shows a phase image from a scan taken in a completely Fig. 8 are phase images of H rufescens all obtained in tapping different orientation. Here a section of EDta demineralized, uI node under fully hydrated conditions in a wet cell. Fig 8a-c shows fixed and undigested, framework tissue was first mounted, edge- a progression of higher resolution scans of unfixed specimens di- on, on a histology slide and sliced in a microtome so that approx gested with trypsin or trypsin+ proteinase K. The lowest resolution imately 1-2 um of tissue stood nearly upright, ie nearly orthogo scan of Fig. Sa, of a specimen that was digested in trypsin only. nal to the glass slide; the orientation was much like that shown in shows what appears to be a"porous"structure. Closer inspection, the sEm image of Fig 1d. The sections were, though, inclined to the owever, reveals embedded fibers(e. g. at the white arrow )and the vertical by about 30-40. also consistent with the scenario of fact that the layers contain a core of fibers. The fibrous core is more Fig. 1d. Specimens of this type were fully hydrated and scanned clearly revealed in Fig. 8b which is of a specimen digested in tryp- in tapping mode in a wet cell. The individual layers are evident sin+ proteinase K Individual fibers are visible, and that they are and are hydrated so that their thickness is estimated(correcting embedded in less organized organic material is evident. Thus the for the angle of inclination) at 80-120 nm. In the shell, the inter-

characterized by fibers, or fiber bundles, and the larger by an open network that creates pore-like openings. Fig. 7 shows two typical scans over a larger area for an unfixed and undigested specimen of H. rufescens stained with calcofluor white. Fig. 7b shows a topograph image where Fig. 7a shows an elevation profile of that region. As in the histochemical images shown above, these large scans reveal the outlines of tiles on the surface of the organic interlamellar layers. Where the intertabular matrix has ‘‘collapsed”, and where calcofluor white has evidently bound, the elevations can be as high as 250 nm above the locations of what were the tile/interlamellar layer interfaces. As we have determined that the tile diameter in our H. rufescens is 7 ± 0.5 lm, the images leave little doubt as to representing tile out￾lines. The central regions are depressed relative to the surrounding tile imprint as noted by the two central arrows. This depressed re￾gion is collocated with the central regions described in Fig. 2a and c where it is believed that mineral is preferentially removed during demineralization. What is also notable from the height profile of Fig. 7a is that the top edges of the intertabular layers are not flat but contain height differences of the order of 50 nm. This itself may account for the appearance of a ‘‘double layer” effect as dis￾cussed earlier in connection with Fig. 2c and d. As also noted above, the tile imprint centers are depressed, and the scans show these depressions are nearly 100 nm. The structure at the imprint cen￾ters is discussed below. Fig. 8 are phase images of H. rufescens all obtained in tapping mode under fully hydrated conditions in a wet cell. Fig. 8a–c shows a progression of higher resolution scans of unfixed specimens di￾gested with trypsin or trypsin + proteinase K. The lowest resolution scan of Fig. 8a, of a specimen that was digested in trypsin only, shows what appears to be a ‘‘porous” structure. Closer inspection, however, reveals embedded fibers (e.g. at the white arrow) and the fact that the layers contain a core of fibers. The fibrous core is more clearly revealed in Fig. 8b which is of a specimen digested in tryp￾sin + proteinase K. Individual fibers are visible, and that they are embedded in less organized organic material is evident. Thus the interlamellar layers are not ‘‘porous” per se but are composed of a fibrous core, or network, that is embedded in and sandwiched be￾tween less organized organic material. Individual fibers appear to have diameters in the range 5–10 nm and run for lengths of at least 1 lm, although to date their length distribution has not been quan￾titatively determined. Fig. 8c shows a higher resolution scan which provides additional evidence of the fibrous network and suggests that the fibers may indeed be organized in nearly parallel bundles. In both fixed and unfixed samples there are examples where the fi- bers appear to be so organized, and examples that suggest a more random orientation. Our analysis of fiber orientation via 2D Fourier spectra, however, did not reveal long range order but did indicate short range order over distances of an imprint diameter. This would suggest that beneath a tile the fibers are ordered possibly as suggested by Weiner and Traub (1980). Fiber organization as ob￾served here is also consistent with the recent observations of Rous￾seau et al. (2005) studying P. maxima who suggest that both the intracrystalline and intertabular matrices are highly organized. In the case of the intracrystalline matrix, they propose that it ‘‘reacts as a single crystal”; in the case of the interlamellar matrix, they conclude that it is ‘‘highly crystalized”. Our observations such as shown in Fig. 8c and also Fig. 8d for H. rufescens are supportive of this proposal. Complete resolution, however, of the orientation distribution awaits further study. Fig. 8d shows a phase image from a scan taken in a completely different orientation. Here a section of EDTA demineralized, un- fixed and undigested, framework tissue was first mounted, edge￾on, on a histology slide and sliced in a microtome so that approx￾imately 1–2 lm of tissue stood nearly upright, i.e. nearly orthogo￾nal to the glass slide; the orientation was much like that shown in the SEM image of Fig. 1d. The sections were, though, inclined to the vertical by about 30–40, also consistent with the scenario of Fig. 1d. Specimens of this type were fully hydrated and scanned in tapping mode in a wet cell. The individual layers are evident and are hydrated so that their thickness is estimated (correcting for the angle of inclination) at 80–120 nm. In the shell, the inter￾Fig. 6. (a and b) Images obtained from unfixed H. rufescens, demineralized with EDTA, and labeled with polyclonal antibodies. (c) An image obtained on a sample identical to that used in (a and b) using aminoacridone staining shown for perspective. (d) An image of unfixed H. rufescens, demineralized with EDTA, and labeled with pre-immune serum. J. Bezares et al. / Journal of Structural Biology 163 (2008) 61–75 69

J Bezares et aL /Joumal of Structural Biology 163(2008)61-75 20um rufescens, demineralized in EDTA, undigested, and unfixed Scan formed in tapping mode, in water. (b)Topograph nage accompanying (a) showing the elevations at the intertabular matrix(c)Topograph similar to(b) from a different area. lamellar layers appear to be in the range -30 nm in thickness, but demineralized tissue in water maintains a thickness of roughly that of the intact shell, ie. it hydrates and swells by factors of approximately 3-4 Fibers are visible even in such undigested sec- tions, and in fact on sections such as these appear to be arranged in an organized network. With reference to the staining of chitin with WGA as shown in Fig 3 we suggest that the finer resolution of the fibers in the AFM images is due the residual amounts of organic matter surrounding the chitin fibers without, or with partial, diges- tion as used In addition, the SEM specimens were Cr coated and this would also contribute to their appearing somewhat thicker 1.1pm Fig 9a and b shows results from a high resolution scan. The profile 634nm scan of Fig 9a clearly illustrates the physical topography of the lay- ers, including the intertabular matrix. Fig. 7a, associated with much larger scan including at least two tiles, more clearly reveals the varying height profile across an intertabular zone. This sort of height difference, we believe, contributes to the patterns of stain- ing described as"double-layers"earlier in images such as in Figs. 2a or 6a, for example. Fig 9b is a phase image again showing the fibrous core that reveals "pores"about 40-80 nm in diameter that in our scans appeared to be about 10-20 nm deep. Fig. 9a is a height profile along the line shown in Fig. 9b: note the correspond- ng green and red markers in Fig. 9a and b Profiles such as these indicate that variations in fiber height were in the range of less than a nanometer to up to 10 nm suggesting that they are not in Fig 8.(a-c)are phase images of intermellar sheets obtai mode from the same plane. Moreover, there were instances where fibers ap- H rufescens demineralized in EDTA, unfixed, and digested w sin, (b and c) peared to run under and over other fibers suggesting the possibility vith trypsin+ proteinase K (d) phase image obtained in demineralized H. rufescens, unfixed, but mounted on a histology slide in a ne and 9b, the fibers appear arranged more as a two dimensional scaf- dge-on orientation. fold analogous to the three dimensional chitin scaffold recently de- scribed by Miserez et al. (2008 )within beak of the Humboldt squid 15.0 ding to(a). The height profile was ers appear to be areas. The softer areas of the imprint centers correspond to the center annuluses stained by colloidal iron as in Fig 5a and b

lamellar layers appear to be in the range 30 nm in thickness, but demineralized tissue in water maintains a thickness of roughly that of the intact shell, i.e. it hydrates and swells by factors of approximately 3–4. Fibers are visible even in such undigested sec￾tions, and in fact on sections such as these appear to be arranged in an organized network. With reference to the staining of chitin with WGA as shown in Fig. 3 we suggest that the finer resolution of the fibers in the AFM images is due the residual amounts of organic matter surrounding the chitin fibers without, or with partial, diges￾tion as used. In addition, the SEM specimens were Cr coated and this would also contribute to their appearing somewhat thicker. Fig. 9a and b shows results from a high resolution scan. The profile scan of Fig. 9a clearly illustrates the physical topography of the lay￾ers, including the intertabular matrix. Fig. 7a, associated with a much larger scan including at least two tiles, more clearly reveals the varying height profile across an intertabular zone. This sort of height difference, we believe, contributes to the patterns of stain￾ing described as ‘‘double-layers” earlier in images such as in Figs. 2a or 6a, for example. Fig. 9b is a phase image again showing the fibrous core that reveals ‘‘pores” about 40–80 nm in diameter that in our scans appeared to be about 10–20 nm deep. Fig. 9a is a height profile along the line shown in Fig. 9b; note the correspond￾ing green and red markers in Fig. 9a and b. Profiles such as these indicate that variations in fiber height were in the range of less than a nanometer to up to 10 nm suggesting that they are not in the same plane. Moreover, there were instances where fibers ap￾peared to run under and over other fibers suggesting the possibility of a weave like pattern. In some cases, such as shown in Figs. 8b and 9b, the fibers appear arranged more as a two dimensional scaf￾fold analogous to the three dimensional chitin scaffold recently de￾scribed by Miserez et al. (2008) within beak of the Humboldt squid Fig. 8. (a–c) are phase images of intermellar sheets obtained in tapping mode from H. rufescens demineralized in EDTA, unfixed, and digested with (a) trypsin, (b and c) with trypsin + proteinase K. (d) phase image obtained in tapping mode of EDTA demineralized H. rufescens, unfixed, but mounted on a histology slide in a near edge-on orientation. Fig. 7. (a) Elevation profile obtained from a large scan of H. rufescens, demineralized in EDTA, undigested, and unfixed. Scan formed in tapping mode, in water. (b) Topograph image accompanying (a) showing the elevations at the intertabular matrix. (c) Topograph similar to (b) from a different area. Fig. 9. (a) Height profile of the imprint surface of H. rufescens demineralized in EDTA and fixed. (b) A phase image of the area corresponding to (a). The height profile was made along the indicated line. (c) Phase image of a low resolution scan of the tissue used in (a and b). Note that the imprint centers appear to be less stiff than the surrounding areas. The softer areas of the imprint centers correspond to the center annuluses stained by colloidal iron as in Fig. 5a and b. 70 J. Bezares et al. / Journal of Structural Biology 163 (2008) 61–75