The molluscs’ periodic table: Elemental analysis reveals distinct mineralization patterns in radular teeth


 The molluscan phylum is the second specious animal group inhabiting various habitats and feeding on a variety of food sources. The latter is enabled by the radula, a small chitinous membrane with embedded teeth, one important molluscan autapomorphy. Between species, radulae can vary in their morphology, mechanical, and chemical properties. With regard to radular chemical composition, some taxa (Polyplacophora and Patellogastropoda) were studied extensively in the past decades, due to their specificity to incorporate high proportions of iron, calcium, and silicon. There is, however, a huge lack of knowledge about radular chemical composition in other molluscan taxa. The work presented here aims at shedding light on the radular chemistry by performing energy-dispersive X-ray spectroscopy (EDX) analyses on overall 24 molluscan species, thereof two Polyplacophora, two Cephalopoda, and 20 Gastropoda. The elements, which are not part of chitin, and their proportions were documented for overall 1448 individual, mature teeth and hypotheses about potential biomineralizations types were proposed. The here presented work additionally comprises a detailed record on past studies about the chemical composition of molluscan teeth, which is an important basis for further investigation of the radular chemistry. The found disparity in elements detected, in their distribution and proportions highlights the diversity of evolutionary solutions, as it depicts multiple biomineralization types present within Mollusca.

III (Ca, P, Cl, F), IV (Si), OB (Na, S, K). In Cephalopoda, the types II, III, IV, V (Cu), and OB occur. In Vetigastropoda and Neritimorpha, the types II, III, IV, and OB were detected, in Vittina turrita -additionally the type I. In Caenogastropoda and Heterobranchia, the radular composition greatly varies between taxa. Overall, in Caenogastropoda, all composition types were found. However, in each species, the types II, III, and OB were always present, whereas Fe was only determined in Reymondia horei and Littorina littorea, Cu -in R. horei, and K -in Paramelania damoni, Cleopatra johnstoni, R. horei, Faunus ater, and L. littorea. C. johnstoni is the only species that seems to lack Ca. In Heterobranchia, the types II and OB were found in each species. The type III (apatite) is present in Onchidoris bilamellata, Aeolidia papillosa, Polycera quadrilineata, and Doris pseudoargus, but not in Cornu aspersum, as P, Cl, and F were not determined in this species. Si was only detected in C. aspersum and K in O. bilamellata.

Ingesta vs. radular morphology and elemental proportions
Morphology: The longest radulae were detected in species foraging on medium to solid ingesta, followed by solid-, medium-, soft-to-solid-, and nally softfeeders with the shortest radulae (see Table 2 and Supplementary Figures 28-30). The largest radular area was calculated for medium-, medium-to-solid-, soft-to-solid-, solid-, and nally for the soft-feeders with the smallest area. Species foraging on medium-to-solid ingesta possess the highest quantity of tooth rows, followed by solid-, soft-to-solid-, medium-, and nally with the least quantity of tooth rows the soft-feeders.
All elements: In general, we detect that radulae of species foraging on solid ingesta possess the highest proportions of all studied elements, followed by species foraging on medium, medium-to-solid, soft-to-solid, and nally species feeding on soft ingesta.
Composition-type I: The highest Fe-proportions (means) were detected in the exclusively solid-, followed by the medium-to-solid-feeders. No Fe was detected for all other ingesta types.
Composition-type II and III: The highest proportions of Mg were detected in species foraging on medium ingesta, followed by solid-, soft-, medium-to-solid-, and nally soft-to-solid-feeders. Ca was detected in the highest proportions in the medium-, followed by the soft-, solid-, soft-to-solid-, and medium-to-solid-feeders.
P was mainly found in species feeding on medium ingesta, followed by soft-to-solid-, solid-, medium-to-solid-, and nally soft-feeders. Cl was detected in the highest proportions in medium-feeders, followed by species feeding on soft-to-solid, soft, solid, and nally medium-to-solid ingesta. The highest proportions of F were found in radulae of medium-, followed by solid-, and nally medium-to-solid-feeders. In soft and soft-to-solid-feeders, no F was found.
Composition-type IV: The highest Si-content was detected in solid-, followed by medium-, medium-to-solid-, and nally soft-to-solid-feeders. No Si was found in species feeding on soft ingesta.
Composition-type V: The highest Cu proportions were detected in the solid-feeder Reymondia horei and less in the cephalopods foraging on soft-to-solid ingesta. All other radulae seem to lack Cu.
OB: S was detected in the highest proportions in species feeding on soft, followed by solid, soft-to-solid, medium-to-solid, and nally medium ingesta. The highest Na-proportions were detected in solid-, followed by the soft-, soft-to-solid-, medium-, and nally medium-to-solid-feeders. K was detected in the highest proportions in solid-, followed by medium-, medium-to-solid-, soft-to-solid-, and nally soft-feeders.

Discussion
A detailed list of previous studies aimed at determining the chemical composition of the molluscan radula is provided in Supplementary Table 2. Most of the previous research has been done on the Polyplacophora with the focus exclusively on the dominant lateral teeth (for reviews see Brooker & Shaw, 2012;Joester & Brooker, 2016;Kisailus & Nemoto, 2018;Moura & Unterlass, 2020), except for the study on Lepidochitona cinerea determining the elemental composition of all toot types (Krings et al., 2022). Many of these analyses focused on the Fe biomineralization and the phase transformations during maturation (e.g. Towe & Lowenstam, 1967;Evans et al., 1991;Macey & Brooker, 1996;Lee et al., 1998;Lee et al., 2003a;Shaw et al., 2009b;Weaver et al., 2010;Gordon & Joester, 2011;Emmanuel et al., 2014;Grunenfelder et al., 2014). Overall, in previous studies F, Na, Mg, Si, P, S, Cl, K, Ca, Fe, and Cu was detected in the dominant lateral teeth (=lateral teeth II) of Polyplacophora. For Lepidochitona cinerea, in our previous paper, we did not detect Cl, F, and Cu and in Acanthochitona fascicularis -no Si and Cu.
In Cephalopoda, only one study on the radular chemistry exists, to the best of our knowledge. In Octopoda, Jones et al. (1935) targeted, but did not detect Si and Fe. We here determined Na, S, Cu, Si, P, Cl, Ca, and Mg in the radula of Histioteuthis spec. and Loligo vulgaris with proportions <4%.
For the remaining gastropod taxa, only few analyses of the radular chemistry were conducted and usually the presence of elements, but not their proportions, could be determined. One of the earliest studies was done by Troschel (1863) depicting results from Bergh, who performed complex chemical analyses of ashing and dissolving radulae from the Caenogastropods Charonia lampas (detecting P, Ca, and Fe), Lamellaria perspicua (detecting no Si), and Gibberulus gibberulus (probably detecting none of these elements, this is not clear) in different acids. Additionally, Troschel (1863) presented his own results on the radulae of the Caenogastropod Tonna galea and the Heterobranch Helix nemoralis discovering P, Ca, and Fe in both by employing the same experiment. Sollas (1907) was the rst, who studied the radular chemistry in an elevated quantity of taxa and Jones et al. (1935) proceeded. Overall, their protocols are rather complex, involving analytical chemistry methods (ashing, staining, boiling, treating with acids, and using diffusion column) or physics (radula's refractive index). Sollas (1907) Sollas (1907) determined Si and P. She detected Si in specimens collected during winter and phosphoric acid (P) in specimens collected during spring. Krings et al. (2019b) performed EDX analyses on ve specimen of C. aspersum detecting Ca in all specimen and Si in one, even though specimens were also inventoried in spring depicting the inconsistency of elements embedded. For the Vetigastropoda (Haliotis tuberculata), Sollas (1907) detected Si, Ca, and Fe.
The following species were studied by Sollas (1907) and Jones et al. (1935), but for many species their results are contradictory. In the Caenogastropoda Littorina littorea, Sollas (1907)  inductively coupled plasma-optical emission spectrometry) in the limpet Megathura crenulata. Within Vetigastropoda, we detected Na, S, Cu, Si, P, Cl, Ca, and Mg; all of them in low proportions <2%. Cu and S were not documented before, whereas Fe was detected in previous studies (Mikovari et al., 2015;Ukmar-Godec et al., 2015). For the Neritimorpha, only one past study addresses the mineral content detecting S, Cl, K, Ca, Mg, Si, and Fe (Macey et al., 1997). We additionally detected Na and P in Vittina turrita; all elements detected are abundant in very low proportions (<4%). In the Caenogastropoda, we detected Fe, Mg, Ca, Cl, P, F, Si, Cu, S, Na, and K. Cu, F, Na, Si, S, and Cl were not determined before. In all species, proportions are <6%. For the Heterobranchia, we detected more elements (Mg, Ca, Cl, P, F, Si, S, Na, K) than described in past publications (Troschel, 1863;Sollas, 1907;Jones et al., 1935;Krings et al., 2019b). Mg, Cl, F, S, Na, and K were not detected before. All elements are abundant at proportions <15%.
Overall, the above data depicts that it is rather di cult to compare the percentages measured between studies, because in some weight percentages and in others atomic ratios were determined. Besides, methodology, sample preparation, and the analysed sample itself (whole radula or individual radular parts) differs. In addition, the presence and abundance of elements could potentially be in uenced by the food available (e.g. plants containing or lacking Si) or by the chemistry of the saliva. In some taxa, speci cally carnivorous gastropods, the saliva is acid (e.g. Barkalova et al, 2016;Ponte & Modica, 2017), so potentially the contact of the outermost radular teeth with the saliva leads to reduced proportions. Both ideas await further research.
Previous studies relate the radular length to the ingesta type. Herbivorous taxa were found to possess longer radulae than carnivorous ones (Meirelles & Matthews-Cascon, 2003). Littorinid species, feeding on algae covering rocks, were found to possess longer radulae than species feeding from plant surface (Peile, 1937;Marcus & Marcus, 1963;Reid 1986Reid , 1989Reid & Mak, 1999). For Patella species, it was determined that the radular length increases with increasing usage and wear (Fretter & Graham, 1994) and, when algae are less abundant and the radula must thus be used more frequently to obtain the food necessary, its length increases (Cabral, 2007).
In general, we detected a similar pattern for the species studied here as the longest radulae with the highest quantity of tooth rows were found in species foraging on harder ingesta types (medium-to-solid, solid, medium) and the shortest ones in soft-substrate feeders. We, however, could not directly relate herbivory with longer radulae and carnivorous feeding with shorter ones. We additionally detected some relationship between radular length and proportions of elements (e.g. in Patella vulgata), so potentially more mineralized radulae are longer, because their maturation and mineralization requires more time and a longer contact to the overlain epithelia in the radular sack and mineralization zone. However, this does not seem to be the case for every species, as Lepidochitona cinerea and Acanthochitona fascicularis have relatively short heavily mineralized radulae. Thus, in these polyplacophoran species, the overlain epithelia can presumably incorporate more minerals at the same time or the radular replacement rate is faster in P. vulgata in contrast to the one in the Polyplacophora. Unfortunately, the radular replacement rate is known for few taxa: for Polyplacophorans (Acanthopleura, Plaxiphora, Patelloida, Mopalia), a rate of 0.36-0.80 rows per day was determined (Nesson, 1969;Shaw et al., 2002Shaw et al., , 2008 and for P. vulgata, a rate of 1.5 rows/day was described (Isarankura & Runham, 1968). In Caenogastropoda, for Lacuna (Littorinidae), the rate of 3 rows/day (Padilla et al., 1996), for three Littorina species (Littorinidae) -5-6 rows/day depending on the temperature (Runham & Isarankura, 1966;Isarankura & Runham, 1968), and for Pomatias elegans -5.02 rows/day (Runham & Isarankura, 1966) was determined. For Heterobranchia, the rate of 2.9 rows/day in Lymnaea stagnalis (Runham, 1962), 5.02 rows/day in Agriolimax reticulatus (Runham & Isarankura, 1966), 3.6 rows/day in adult Helix aspersa (= Cornu aspersum; Runham & Isarankura, 1966) was detected. For Cepaea nemoralis, the whole radula was found to be renewed within 30-35 days (Mackenstedt & Märkel, 1987). Thus, in general, a higher degree of mineralization is inversely related to the higher replacement rate (teeth that possess larger proportions of minerals are replaced slower). However, radular replacement seems to depend on many factors, such as water temperature, metabolic rates, or age of animals (Isarankura & Runham, 1968;Mischor & Märkel, 1984;Fujioka, 1985;Padilla et al., 1996). Further studies on these questions are required.
In general, we detected that radulae of species, foraging on the solid ingesta, possess heavy mineralized teeth and species feeding on the soft ingesta show the smallest proportions. In biological materials, heterogeneities can have their origin in geometry, chemistry, and/or structure (for a review see Liu et al., 2017). In the dominant lateral teeth of chitons and limpets they have their origin in the distribution of the inorganic components and in the architecture of organic components (Weaver et al., 2010;Wang et al., 2013;Grunenfelder et al., 2014;Herrera et al., 2015;Ukmar-Godec, 2016;Pohl et al., 2020;Stegbauer et al., 2021). We have previously correlated the hardness and elasticity values in Lepidochitona cinerea with the iron and the calcium proportions (Krings et al., 2022), which was previously also described for limpet teeth (Runham et al., 1969;Vincent, 1980;van der Wal, 1989;Barber et al., 2015) and for other chitons (Weaver et al., 2010;Wang et al., 2013;Grunenfelder et al., 2014;Herrera et al., 2015). For the paludomid gastropods, we previously measured elasticity modulus values ranging from 2 GPa at the tooth basis to 8 GPa in the cusp in solid substrate feeders, whereas soft substrate feeders possessed signi cantly softer teeth (4.6 GPa) (Krings et al., 2019a(Krings et al., , 2021cGorb & Krings, 2021). In these species, we here detected inorganic elements in rather small proportions. We thus propose that speci c cross-linking conditions of the chitin due to tanning (Runham, 1963), bre arrangement, and density (Evans et al., 1990(Evans et al., , 1994Wealthall et al, 2005;Gordon & Joester, 2011;Lu & Barber, 2012;Grunenfelder et al., 2014;Ukmar-Godec, 2016;Ukmar-Godec et al., 2017;Stegbauer et al., 2021) rather cause the heterogeneities in mechanical properties. We previously also detected that the capability of wet teeth to rely on one another and to redistribute the mechanical stress increases the radula's resistance to structural failure in paludomid gastropods (Krings et al., 2021d. This altogether probably enables the feeding on harder ingesta types. Whether these mechanisms are also applicable for the other molluscan species, await further investigations.

Specimen studied and dissection
Molluscs were obtained from various sources (see Tab. 1 for details): individuals of Littorina littorea and Lepidochitona cinerea were collected at the North Sea in summer 2019 and those of Patella vulgata in autumn 2020. The gastropods Cornu aspersum, Rochia conus, Haliotis tuberculata, Vittina turrita, Faunus ater, and Anentome helena were bought from online pet shops. All of them were shortly boiled and preserved in 70% ethanol. Individuals of Aeolidia papillosa, Onchidoris bilamellata, Polycera quadrilineata, and Doris pseudoargus were received from the Biologische Anstalt Helgoland in May 2021, kept in aquarium for 2 weeks in Hamburg, before the gastropods naturally died and then were preserved in 70% ethanol. Frozen specimens of Loligo vulgaris were bought from Fische Schmidt (store specialized on edible sh, Eppendorfer Baum 18, 20249 Hamburg) for the dissection course of the Universität Hamburg, radulae were directly extracted from defrozen squids and preserved in 70% ethanol. Samples of Lavigeria grandis, L. nassa, Paramelania damoni, Cleopatra johnstoni, Reymondia horei, Spekia zonata, Paludomus siamensis, Buccinum undatum, Acanthochitona fascicularis, Histioteuthis sp. were extracted from already preserved (70% ethanol) specimens, some of them already inventoried in museum collections (Zoologisches Museum Hamburg, ZMH; Museum für Naturkunde Berlin, ZMB).
Species identi cation was reviewed by employing the relevant literature, the nomenclature and systematic position were checked using molluscabase.org. Not previously inventoried specimens were incorporated in the malacological collection of the ZMH.
Overall, data from 72 adult specimens were analysed for this study. For each species, three adult specimens were selected, except for Histioteuthis spec. with two. We have chosen specimens of similar size per species, since the relationship between specimens' length and radular size is puzzling. Some previous studies relate both parameters (Radwin & Wells, 1968) and others rather see a loose relationship or could not relate them for every species (Meirelles & Matthews-Cascon, 2003;Grünbaum & Padilla, 2014). Additionally, the specimens chosen for each species were collected at the same time since seasonal dependencies in radular length were previously reported (Ito et al., 2002). All data presented here is new, except for the elemental composition and radular morphology of Lepidochitona cinerea, which was taken from Krings et al. (2022). In this previous study, we analysed the ontogeny of the elemental composition and the mechanical parameters hardness and elasticity in three specimens of L. cinerea. For the present study, we included only the data from the working zone (the mature part) for the purpose of comparison between species.
Habitus images were either taken employing the Keyence Digital Microscope VHX-5000 (KEYENCE, Neu-Isenburg, Germany) or by using an iPad Pro (11 zoll; Apple Inc., Cupertino, USA) equipped with a 12-megapixel wide angle lens. Each specimen was dissected, the radula was carefully extracted by tweezers and then manually freed from surrounding tissue.

Scanning electron microscopy (SEM)
For images of the whole radula or the radular working zone, radulae (three per species, except for Histioteuthis spec. with two) were cleaned in an ultrasonic bath for 2-20 seconds and afterwards arranged on scanning electron microscopy (SEM) sample holders (see Supplementary Figures 1-24). All radulae were rst documented with the Keyence Digital Microscope VHX-5000 or VHX-7000 (KEYENCE, Neu-Isenburg, Germany). Here the length and width of each radula were measured and the quantity of tooth rows counted. From the length and width, the radular area was calculated. Two radulae per species were then visualized uncoated employing the Tabletop Microscope TM 4000 Plus (Hitachi, Tokyo, Japan) for more detailed images and one radula per species was coated and documented with the Zeiss LEO 1525 (One Zeiss Drive, Thornwood, USA) to receive images with a very high resolution (except for images of L. cinerea, they were taken from Scheel et al., 2020, and of H. spec. , as their radulae were documented uncoated and afterwards used for the EDX). Based on the morphology and arrangement of teeth, which were also categorized (e.g. central tooth, lateral tooth I, lateral tooth II, marginal tooth, inner teeth, outer teeth, etc.) according to their shape, size, and position on the membrane, radulae were assigned to different radular types (e.g. docogloss, isodont, rhipidogloss, etc.), if a suitable category could be determined from literature (e.g. Gray, 1853;Hyman, 1967;Steneck & Watling, 1982;Nixon, 1995;Haszprunar & Götting, 2007).
Then, the radulae, which were previously documented uncoated (two per species), were rewetted with 70% ethanol and loosened from the SEM sample holder and used for elemental analysis.

Elemental analysis (EDX)
Wet radulae were arranged on glass object slides (Carl Roth, Karlsruhe, Germany) with double-sided adhesive tape. They were positioned along their longitudinal axis so that the outermost teeth of one side were directly attached to the slide. The adjacent and more inner teeth were located above, followed by the central teeth, the inner teeth from the other side, and nally, on top, outer teeth again. Each radula was then dried for three days under ambient temperature and afterwards surrounded with a small, metallic ring ensuring an almost parallel sample surface. Epoxy resin (RECKLIEPOXIWST, RECKLI GmbH, Herne, Germany) was lled into the metallic ring and left polymerizing at room temperature for three days. This speci c epoxy was chosen, since it does not in ltrate the teeth. Object slide and tape were then removed and, to receive longitudinal sections of each tooth, the embedded radulae were polished until the outer teeth were on display (controlled by examining the samples in the light microscope) using sandpapers of different roughness. Then samples were smoothed with aluminium oxide polishing powder suspension of 0.3 μm grainsize (PRESI GmbH, Hagen, Germany) on a polishing machine (Minitech 233/333, PRESI GmbH, Hagen, Germany). After polishing, the samples were cleaned from the polishing powder by an ultrasonic bath lasting ve minutes. Samples were then coated with platinum (5 nm layer) and the elemental compositions of speci c areas of the embedded teeth were examined employing the SEM Zeiss LEO 1525 (One Zeiss Drive, Thornwood, New York, USA) equipped with an Octane Silicon Drift Detector (SDD) (micro analyses system TEAM, EDAX Inc., New Jersey, USA) always using an acceleration voltage of 20 keV and the same settings (e.g. lens opening, working distance, etc.). Before measuring a sample the detector was always calibrated using cupper. We performed elemental mappings for test purposes, but for elements that are present in rather lower proportions, this method is not sensitive enough. We thus focused on the elemental analysis of small areas (10-200 μm 2 , depending on the tooth) trying to analyse the largest possible area.
The elements H (hydrogen), C (carbon), N (nitrogen), O (oxygen), Pt (platinum), Al (aluminium), Ca (calcium), Na (sodium), Mg (magnesium), Si (silicon), P (phosphorus), S (sulphur), Cl (chlorine), K (potassium), F ( uorine), Cu (copper), and Fe (iron) were detected and their proportions measured. We used the data of atomic ratio (atomic %) for this study. These values were received with two positions after the decimal point, lower proportions were not detectable with this method and therefore they were given as 0.00. We did not analyse and discuss the following elements, as they are either the elemental basis of chitin (H, C, N, O), the coating (Pt), or the polishing powder (Al, O).
After analysing the outer teeth, each sample was again polished and smoothened until the next tooth type was on display; cleaning procedure and EDX analyses were again performed. These steps were repeated until all teeth were analysed. In this study, we present the results of the radular working zone, which is not covered by epithelia and mature. Overall, we use data of 1448 analysed teeth from 49 specimens.

Statistical analyses
All statistical analyses (mean, standard deviations) and visualizations with boxplots, pie charts, or trend lines were performed with JMP Pro, Version 14 (SAS Institute Inc., Cary, NC, 1989-2007.

Composition-and biomineralization-types
With EDX analysis, the proportions of the individual elements, present in a de ned area, can be documented, whereas the speci c bonding and structure of molecules cannot be analysed. However, from the percental occurrence, in comparison with past studies on the radular chemical composition involving, we propose that the elements, detected here, are potentially part of the following molecules or minerals. These were assigned to different composition-or biomineralization-types: Category Type 1: Characterized by the presence of Fe. Potentially present in the form of magnetite as documented in polyplacophorans or goethite in limpets (e.g. Lowenstam, 1962;Kirschvink & Lowenstam, 1979;Lowenstam & Weiner, 1989;Huang et al., 1992;Han et al., 2011;Wang et al., 2013;Ukmar-Godec, T., 2016, Nemoto et al., 2019McCoey et al., 2020).
Category Type 2: Characterized by the presence of Mg and Ca. Elements are potentially involved in the protein packing, an increase in density of chitin bres and in material stiffness as documented in limpet teeth (Ukmar-Godec et al., 2017).
Category Type 4: Characterized by the presence of Si. Potentially present in the form of silica as documented in limpet teeth (e.g. Hua & Li, 2007;.
Category Type 5: Characterized by the presence of Cu.
Category OB (organic bonds): The presence of Na, K, and S is often related to the protein bonding (e.g. Creighton, 1997;Harding, 2002).
For this study, we de ned ingesta categories that are rather broad, because (a) either the species' food preference has only been described anecdotally and (b) even if the speci c food type was known, its precise mechanical properties and its structural resistance to feeding is di cult to determine (see Padilla, 1985Padilla, , 1989 and that is why these properties remain unknown. We collated the data on ingesta from the literature, if available, and assigned it to the following ingesta categories (please, see Table 1 for food types per species and corresponding references): Soft: Algae from sand or mud, sea anemones.
Solid: Algae/plants from rocks and/or corals.
Medium to solid: Algae, eshy macroalgae, also from rocks.

Declarations
Authors' contribution WK and SG together initiated the project, designed the study, and discussed the data. In the course of his master thesis, JOB performed EDX analyses of six species and collated the literature. WK together with JOB wrote the rst draft of the manuscript. WK executed the remaining EDX analyses and created the gures. All authors contributed to and approved the nal version of the manuscript for publication.
Polycera, and Doris. We are grateful for the access to Ilka Sötje's highly equipped lab at the UHH, which enabled us to keep and feed the nudibranch molluscs, and her expertise on the preservation of marine invertebrates, the lecturers of the students' dissection course from the UHH and LIB for reserving Loligo's radulae and beaks, Heinz Büscher from Basel, Switzerland, for collecting specimens of the Paludomidae at Lake Tanganyika, and Bernhard Hausdorf from the LIB for providing access to Histioteuthis specimens. We are grateful for the technical support of Esther Appel and Alexander Kovalev from the Department "Functional Morphology and Biomechanics" (CAU) at the beginning of the study, and the helpful comments of the anonymous reviewers. Tables   Table 1 is available in the Supplementary Files section.  Figure 1 Elemental proportions of the species' radular teeth, summarized for taxa (Patellogastropoda, Polyplacophora, Vetigastropoda, Caenogastropoda, Cephalopoda, Heterobranchia, Neritimorpha) to ease comparison in the radular mineral content.

Figure 2
Elements detected, in mean atomic percent (represented by different colours), and assigned ingesta category (soft, medium, solid, soft to solid, medium to solid) for each species, sorted to the major molluscan groups (Polyplacophora, Cephalopoda, Gastropoda: Patellogastropoda, Vetigastropoda, Caenogastropoda, Neritimorpha, Heterobranchia). Elements are sorted to the de ned composition-types (I, II, III, IV, V, OB).

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.