Historical and biomechanical analysis of integration and ...R. Guralnick and K. Smith Historical patterns of molluscan feeding Page - 2 INTRODUCTION Joseph Needham ('33) used the wonderful

Historical and biomechanical analysis of integration and dissociation inmolluscan feeding, with special emphasis on the true limpets

(Patellogastropoda: Gastropoda)

by

Robert Guralnick 1 and Krister Smith2

1 Department of Integrative Biology and Museum of Paleontology, University of California, Berkeley, CA94720-3140 USA

email: [email protected]

2 Department of Geology and Geophysics, University of California, Berkeley, CA 94720 USA

ABSTRACTModifications of the molluscan feeding apparatus have long been recognized as a crucial feature in

molluscan diversification, related to the important process of gathering energy from the envirornment. Anecologically and evolutionarily significant dichotomy in molluscan feeding kinematics is whether radularteeth flex laterally (flexoglossate) or do not (stereoglossate). In this study, we use a combination ofphylogenetic inference and biomechanical modeling to understand the transformational and causal basis forflexure or lack thereof. We also determine whether structural subsystems making up the feeding systemare structurally, functionally, and evolutionary integrated or dissociated.

Regarding evolutionary dissociation, statistical analysis of state changes revealed by the phylogeneticanalysis shows that radular and cartilage subsystems evolved independently. Regarding kinematics, thephylogenetic analysis shows that flexure arose at the base of the Mollusca and lack of flexure is a derivedcondition in one gastropod clade, the Patellogastropoda. Significantly, radular morphology shows nochange at the node where kinematics become stereoglossate. However, acquisition of stereoglossy in thePatellogastropoda is correlated with the structural dissociation of the subradular membrane and underlyingcartilages. Correlation is not causality, so we present a biomechanical model explaining the structuralconditions necessary for the plesiomorphic kinematic state (flexoglossy). Our model suggests thatplesiomorphically the radular teeth must flex laterally as they pass over the bending plane as a result of themechanical restrictions in the flexible but inelastic subradular membrane and close association betweensubradular membrane and cartilages. Relating this model to the specific character states of the clades, weconclude that lack of flexure in patellogastropods is caused by the dissociation of the subradular membraneand cartilage supports.

preprint: to be published in Journal of Morphology

R. Guralnick and K. Smith Historical patterns of molluscan feeding

Page - 2

INTRODUCTIONJoseph Needham ('33) used the wonderful

analogy of a series of shafts and gears that may ormay not engage to describe the integrated butpotentially decouplable or dissociable elements of thedevelopmental machinery. Like developmentalprograms, anatomical parts and their kinematics arealso partially integrated and partially dissociable.Unlike developmental programs, the metaphoricalengaging shafts can become reality; biologicalstructures and kinematics sometimes very muchresemble gear or pulley systems. As T. H. Huxley(1853) first pointed out, and Herrick (‘06),Eigenbrodt (‘41), and Branch (‘81) also discussed,something close to a pulley system exists in thefunctioning (=kinematics sensu Lauder, ‘90) of thefeeding system of molluscs. In this paper we linkNeedham’s concept of integration and dissociabilitywith Huxley’s description of the kinematics of themolluscan feeding system, focusing on integrationand dissociability in structure and function of thesubparts that make up the feeding system inmolluscs. We discuss “dissociability” in threedifferent contexts: (1) dissociation or decoupling ofstructural elements, which may cause (2)dissociation of those parts during the use ofstructures, and (3) dissociation in evolution ofsubparts that make up systems (i.e., lack ofcoordinated evolution).

The molluscan buccal apparatus is composed ofmany discrete elements that operate in unison duringfeeding. These elements include the numerousmuscles that power the buccal machinery, pairs ofunderlying cartilage that serve as the pulley wheel,and the radular apparatus itself, which is drawn overthe cartilages during feeding (Fig. 1A,B). Thebuccal apparatus, in terms of its structure andfunction, is one of the most thoroughly studied ofmolluscan anatomical systems (Huxley, 1853;Geddes, 1879; Plate, 1897; Amaudrut,1898;Woodward,’01; Herrick, ‘06; Crofts, ‘29; Carriker,‘43; Starmühlner, ‘52; Hubendick, ‘56; Lemche andWingstrand, ‘59; Fretter and Graham, ‘62; Graham,‘64; Fretter, ’65; Morris and Hickman, ‘81;Wingstrand, ‘85; Hickman and Morris, ‘85).Unfortunately, much of this work focuses onindividual, or a few closely related, species.

Graham (‘73) compared coordinated change ofthe gastropod and polyplacophoran musculature anddentition. His goal of documenting coordinatedchanges was based on an a priori belief that thefeeding system of molluscs has been highly

integrated. For example, in the introduction to hispaper he stated that "[t]he evolution of the radularpattern has a parallel evolution in the anatomy of theodontophore, and, in particular, in the cartilageswhich support it and the muscles by which it ismanipulated" (Graham, ‘73: 318). Our approach isdifferent from Graham’s. Rather than looking forcorrelated characters, we examine the buccalapparatus using, in part, a historical approach,tracing character state transformations of structureand function and relating these transformations toone another. Thus, one outcome of our analysis is astatement of whether subparts of the molluscanfeeding system have evolved congruently or not.More importantly, we clarify ambiguities regardingmolluscan feeding function by integrating charactertransformation with a general biomechanical modelof one of the most important aspects of the feedingstroke.

The functional shift that we focus upon isperhaps the most highly recognized aspect of the useof the radula during feeding. Radular teeth caneither laterally flex outwards and then sweepinwards (the flexoglossate condition) or the teeth canremain fixed (the stereoglossate condition) duringthe feeding stroke (Salvini-Plawen, ‘88; Ponder andLindberg, ‘97). Figure 2A shows an example offlexoglossy, with the teeth laterally rotating at theanterior of the buccal system. In this study, we askthe following seven questions:

(1) Where in the evolution of the Mollusca havechanges from flexoglossy to stereoglossy (or viceversa) occured?(2) Do changes in the morphology of the feedingsystem subparts change in concert or independentlythrough evolution?(3) What changes in the radula and cartilages, if any,accompany (that is, occur at the same node as) shiftsfrom stereoglossy to flexoglossy or vice versa?(4) Which of these evolutionary structural changesare directly causally related to change in function,and which merely represent noise in the system?(5) Are these causal state changes related toassociations or dissociations of the many interactingparts involved in the feeding stroke?(6) If different parts of the feeding system canchange independently from one another, can similarradular morphologies be used in completely differentways?(7) Have dissociations led to functional flexibility orstereotypy in descendant taxa?


Page - 3

MATERIALS AND METHODSIn order to document structural changes, we

determine historical patterns of change for thefeeding system of Polyplacophora (chitons),Monoplacophora, and basal Gastropoda with aspecial emphasis on the Patellogastropoda, or truelimpets. In particular, we use a phylogenetichypothesis of Patellogastropoda with exemplars fromthe Orthogastropoda (the clade containing allgastropods except the Patellogastropoda, asdiscussed in Ponder and Lindberg, 1997),Polyplacophora, and Monoplacophora serving asoutgroups. Although our focus is on non-gastropodoutgroups and the more basal Gastropoda, we arecurrently sampling the more nested gastropod clades,the Caenogastropoda and Heterobranchia, tocompletely resolve evolutionary patterns across theGastropoda. Kinematic data for feeding function,largely culled from the literature, is used along withprimary morphological data as a character in theanalysis. Primary data were collected fromdissections, histological examinations, and three-dimensional reconstructions based on the histology.Mechanical models are based on structuralrelationships from our primary data and informationabout kinematics from the literature and our ownwork.

Phylogenetic hypothesisThis phylogenetic hypothesis is based on the

data-sets available in Lindberg and Hedegaard (‘96),Lindberg (‘98) and Sasaki (‘98) and includessampling of new cartilage and radular characters andcoding for five taxa not included in those originalanalyses. With the excpetion of Bathyacmaea, theincluded taxa all form polytomies with taxa alreadyin previous anaylses. Thus our re-analysis of theexisting dataset, despite additions of operationaltaxonomic units, has similar information content interms of branching patterns compared to the alreadypublished Patellogastropod trees. As well, the newcharacters and their distribution on the tree arepresented in this paper. The new taxa that have beensampled since Lindberg (‘98) are Bathyacmaea,Erginus, Paralepetopsis, Acmaea virginea, andRhodopetala. A total of eighteen ingroup and fouroutgroup taxa were analyzed. Sampling withinOTUs is shown in Table 1. Re-analysis of theoriginal dataset with new characters was performedin PAUP3.1 using tree-bisection and reconnectionwith ten replicates. The most parsimonious trees

were imported to MacClade3.06 to further analyzecharacter evolution.

Although we focus on changes in major clades,our phylogenetic analysis also allows us to determinehow variability within subclades does or does notaffect function. Thus, state changes at allhierarchical levels in the phylogeny are examined.The radular and cartilage characters that we discussare shown in Table 2. One functional character,whether or not the radular teeth flex laterally as theradula is pulled over the bending plane, was alsoincluded (labeled with FS for stereoglossy or FF forflexoglossy in Fig. 3B) in the analysis, although thesampling for this character is not as good as for thestructural characters. This character was codedlargely from the literature. Data about theflexoglossate condition have been based on the workof Ankel (‘36a, b, ‘38), who studied Helcion,Eigenbrodt (‘41) who worked with Patella, andPadilla (‘85) who examined Acmaea mitra and thelottiid Notoacmaea. Ankel (‘36a, b, ‘38),Eigenbrodt (‘41), Runham (‘69), Morris andHickman (‘81) and Hawkins et al. (‘89) studied taxafrom the other major gastropod clades, theVetigastropoda, Neritopsina, Caenogastropoda, andHeterobranchia (see Ponder and Lindberg, ‘97), andall these taxa were found to show flexure, althoughthe amount of flexure depends on the group. In thenon-gastropod outgroups, Salvini-Plawen (‘88) andmany subsequent authors have assumed stereoglossybut have not directly studied kinematics. However,Jüch and Boekschoten (‘80, see especially Fig. 10)examined the kinematics and clearly showed that inthe polyplacophoran Lepidochitona the main lateralteeth flex during the feeding stroke. Graham (‘73)also noticed that the teeth tend to flex in Chitonia.This is also confirmed by D. Padilla (pers. comm.)working on Katharina tunicata. Radular kinematicsin Monoplacophora have not been directly examined.

The distribution of character states is shown onaccompanying cladograms (Fig. 3A for radular andFig. 3B for functional and cartilage characters),which display the strict consensus of the mostparsimonious trees. Numbering and lettering onFigure 3 is isomorphic with numbering and letteringin Table 2 for the radular and cartilage characters. Ifa state is ambiguous (optimized differently underACCTRAN and DELTRAN), a question markfollows the number as opposed to a letter.

Three-dimensional reconstructionsThree-dimensional reconstructions were used in

determining some morphological character statesand to help visualize major differences in relative


Page - 4

shape and position of structures. Reconstructionswere generated from transverse sections for seventaxa included in the phylogenetic analysis: Mopaliamucosa, Nerita polita, Fissurella volcano, Cellanatransomerica, Bathyacmaea sp., Collisella scabra,and Acmaea mitra. Sections were captured from adigital camera, either hooked to a polarizing lightmicroscope or a dissecting microscope, and stored ona Macintosh PowerPC. Using NIH image, the digitalimages of sections were converted to a stack. Due tomemory limitations, the stack size was kept at orbelow thirty images. Thus, the spacing between eachtransverse section varied depending on specimensize. Although a consistent spacing betweentransverse sections was generally kept for eachsample, we did vary this if the section located at theexact interval was inadequate.

Once a complete stack was generated, each imagein the stack was hand registered to the other images.We tried to minimize the amount of overall changein position of any one morphological feature duringregistration; such features included the cartilagemidpoints, radula, radular diverticulum, esophagus,ganglia, salivary glands, and mantle. Onceregistered, each image in the stack was edited so thatonly subradular membrane and cartilagesremained— all the other morphology was erasedfrom the images and the radula and cartilage werefilled with different gray scale values.

We used the nearest-point projection function ofNIH image to build a three-dimensionalreconstruction, rotating around the x- or y-axis.Twenty degree increments were used for ourrotations. The lower and upper thresholds were setsuch that strongly white and strongly black pixelsare considered background, and everything between,foreground. Colorized snapshots from the 3-Dreconstructions are shown in Figure 4 for six of thesampled taxa. Collisella scabra and Fissurellavolcano reconstructions were built twice, once by theauthors and once by an independent research team.We did this to ensure that reconstructions wereconsistent when done by independent researchers,since hand registration of sections may lead tobiases. Independent reconstructions of the sametaxon showed a high degree of consistency.

RESULTS AND DISCUSSIONCharacter State Distribution

Kinematic character evolutionAll aspects of the kinematics of the buccal

apparatus involve the muscles, radular apparatus,and odontophoral cartilages over which the radula is

drawn, not the radular teeth or muscles alone. Thefeeding stroke involves two separate but coordinatedmovements: (1) odontophoral protraction andretraction, which moves the buccal mass forward anddownward as a unit towards the substrate, and (2)radular protraction and retraction, which involvesmovement of the radula relative to the underlyingcartilages (Graham, ‘73). During the feeding stroke,radular protractor muscles (Fig. 1) pull the radularapparatus over and around the anterior end of thepaired supporting structures; the point where theradula goes over the most anterior portion (that is,where the teeth are shifted to point down instead ofup) is defined as the bending plane. Thus, whenprotracted the anterior-most teeth on the ribbon arelocated ventral to the bending plane, along the lowersurface of the cartilages (Fig. 1). Since whether ornot the radular teeth flex is determined only duringradular protraction and retraction, we focus on thesetwo events and not protraction and retraction of theodontophore.

Patellogastropods have a stereoglossate feedingstroke while all other gastropod and non-gastropodoutgroups are flexoglossate. Lack of rotation can beseen in the parallel feeding traces left on thesubstrate (Fig. 2B). Besides lack of flexure,patellogastropod feeding shows additionaldifferences compared to other intertidal molluscangrazers (Fretter and Graham, ‘62). For example, thepatellogastropod radula strikes the substrate duringodontophoral protraction, not retraction as in othergroups (but see Hawkins et al., ‘89). As well,numerous tooth-rows simultaneously contact withthe substrate anterior to the bending plane, unlikesingle-row contact seen in other groups. Finally, thebuccal apparatus of patellogastropods forcefullystrikes the substrate on which they feed, collectingnot only plant material but rock gouged from thesubstrate (Hawkins et. al., ‘89).

Although our sampling of kinematics is not asgood as our sampling of structural features, whenoptimized along with the rest of the characters usingparsimony the distribution of states is not ambiguous(Fig. 3, states FF and FS). The flexoglossatecondition is primitive given our sampling and thestereoglossate feeding stroke has arisen only once asa synapomorphy for the Patellogastropoda.

Structural character evolutionIn order to better clarify the distribution of

radular character states (Fig. 3A), we briefly reviewthe various elements composing the radula. Theradular subsystem is composed of distinct elements:


Page - 5

the radular and subradular membranes and theradular teeth (see Fig. 5 for tooth morphology). Theradular teeth and membrane are secreted as organicstructures in the posteriorly located radular sac andare then moved progressively anteriorly towards thebending plane. While moving anteriorly duringontogeny, the teeth are often hardened by thedeposition of minerals; it is important to distinguishbetween the bases and cusps of the teeth, which maydifferentially mineralize. The subradular membraneis similar in composition to the radular membranebut is produced along the sides of the pharynx, faranterior to the radular sac. The subradularmembrane underlies and affixes to the radula, andalso extends to either side of the radular teeth andmembrane.

Radular and subradular membranes aresynapomorphic for either the Mollusca or the cladecontaining all Mollusca besides the Aplacophora,depedent on the position of that groupphylogenetically, which is still in doubt. Thesestructures have also been lost in some groups (thebivalves, for example). Unlike the subradularmembrane, which is generally a flat sheet ofchitinous material, the radula is complex in itscomposition and morphology, rendering it character-rich. The radulae of polyplacophorans andpatellogastropods have been considered very similar,reflected by the traditional unification of thesegroups as the “Docoglossa” (Fig. 5C, D forgeneralized views of these radulae). These taxa havefew plate-like marginal teeth, the number of whichhas been subsequently reduced, from six inpolyplacophorans to three in monoplacophorans, toeven fewer in patellogastropods— the patellids havetwo or three, and acmaeoideans (Fig. 3A fordefinition of Acmaeoidea clade) have between zeroand two. The marginals in these “docoglossan”groups are the most weakly developed teeth; on theother hand, in Vetigastropoda and Neritopsina themarginals are numerous and highly plicate, formingbrushlike structures (Fig. 5A, B).

In our character coding we distinguish betweeninner lateral (IL) and outer lateral (OL) tooth fields(following Lindberg, ‘88; McLean, ‘90) and countthe number of cusps for those fields.Monoplacophorans, polyplacophorans, and somepatellogastropods have only one IL cusp, althoughthe state at the patellogastropod node is equivocal.Inner laterals have been lost in Bathyacmaea andtwo pairs are present in the patellid and the lepetidclades (see Fig. 3A for clade names). In theVetigastropoda and Neritopsina, the number of ILcusps increases; these groups have three or more.

The number of OL cusps is plesiomorphically two,but varies in gastropod outgroups from one in Neritato multiple in Fissurella. Most Patellogastropodahave two OL cusps; however, the cusps have beengreatly multiplied in Pectinodonta, while the patellidclade usually has three to four cusps. A rachidian isplesiomorphically reduced in patellogastropods, butthis median tooth is as large as the inner laterals inthe Lepetid clade (although fused to the innerlaterals in lepetids according to Sasaki, ‘98) and insome patellids.

Ferrous oxide minerals (e.g., goethite) aredeposited in the tooth cusps and bases of severaltaxa. The lateral teeth of Polyplacophora and mostPatellogastropoda are mineralized; exceptionsamong basal patellogastropods include Eulepetopsis(no mineralization on ILs or OLs) and Lepetidae (noOL mineralization). Deposition of iron oxides isalso lacking in some Monoplacophora, mostVetigastropoda, and Neritopsina.

Compared to the outgroups, patellogastropodsubclades are apomorphic in terms of the integrationof the radular apparatus. In the outgroups, the basesof the OL teeth attach at one point on the membrane,leaving the rest of the base free above the radularmembrane. The OL bases in all patellogastropodsexcept the basal Bathyacmaea and lepetids are notfree but are attached along their length to themembrane. However, the state at the base of thePatellogastropoda is equivocal. The attachment oftooth bases to the radular membrane is only one waythat the patellogastropods have integrated theirradular apparatus. For example, the IL and OL teethwithin a row are apomorphically fused in theAcmaeoidea (Guralnick and Lindberg, 1999);plesiomorphically the outer laterals are not fusedwith the inners within the row but instead extendinto the row behind (Guralnick and Lindberg, 1999).Teeth do not always rest directly on the membrane,and clear “plate structures” lying between teeth andmembrane have arisen homoplastically. Plates canbe numerous (as in Cellana and Nacella) or singular(as in the Acmaeoidea and Bathyacmaea).

The similarity between the Patellogastropodaand non-gastropod outgroups has long beenrecognized, and our analysis confirms that thepatellogastropod radula retains plesiomorphic states.No unequivocal radular apomorphies diagnose thePatellogastropod clade. By contrast,patellogastropods have many diagnostic cartilageapomorphies (Fig. 3B). Buccal cartilages inmolluscs are more or less discrete pieces of tissuecomposed of what appears morphologically to behyaline cartilage. Although buccal cartilages are


Page - 6

found in all Mollusca, the homologies of differentcartilages have not been well established (but seeSasaki, ‘98). We code not only number of cartilages,as in Ponder and Lindberg (‘97) and most molluscanworkers, but also determine putative homologiesbased on position and shape. Use of shape andposition suffers from one potential drawback; theposition of the structures varies. In some of ourspecimens, the apparatus is protracted, while inothers it is in a resting state. Although the positionof structures does vary from one specimen to thenext, the shifts are not dramatic and patterns are stilldiscernible among and between taxa. As well, wecan use these differences to document position andshape change of elements during the feeding stroke.

The presence of a medial cartilage pair isplesiomorphic for Mollusca. These cartilages areusually long rods that run just lateral to the midlineof the buccal mass. Although their presence isplesiomorphic, medials have different shapes anddetailed morphology in different groups. In thePatellogastropoda, they tend to becomedorsoventrally elongated near the anterior of thebuccal apparatus. By contrast, the medials in theother groups show no change in height from anteriorto posterior (compare Fig. 6B,C to Fig. 6E). Also,the medials in patellogastropods are ventrally fusedor closely approximated at their anterior ends. In theoutgroups, the medials are more separated at thebending plane.

Like medial cartilages, dorsolateral cartilages arealso probably plesiomorphic for Mollusca, althoughthis again depends on the placement ofAplacophorans (which lack dorsolaterals) within theMollusca. Sasaki (‘98) considered the dorsolaterals(anterolaterals in his nomenclature) as de novostructures in patellogastropods and not related to thelaterally placed cartilages in poly- ormonoplacophorans. However, based on position,shape, and composition, we argue that chitons andmonoplacophorans and patellogastropoddorsolaterals are putative homologs. In the“placophorans” the medials and dorsolaterals areattached by a connective tissue sheath, the spacebetween constituting the hollow vesicles (Figs. 7C,8A). The dorsolaterals are also more lateral thandorsal in the non-gastropod outgroups. This positionis largely maintained in patellids, with limited dorsalshifting. In acmaeoideans this pair is furtherreduced, especially in its anteroposterior length, andtends to be located even more dorsally, often lyingdirectly above the medial pair (Figs. 7D, 8C).Dorsolateral cartilages are absent in all othergastropods.

At first glance the medials appear to be the mainsupports of the radular apparatus. In all gastropodclades besides Patellogastropoda, the medials arevery closely associated with the subradularmembrane, the membrane conforming to the shapeof the groove between the medials (see Figs. 7A,B,Fig. 8B and especially Fig. 4B). However, inPolyplacophora and Patellogastropoda the situationis different. In Polyplacophora a small, flatteneddorsal cartilage pair lies just above the vesicle; formost of its length the radula rests on the hollowvesicles, but it is lifted slightly above the vesicles tolie on the dorsal cartilages where they occur (Figs.7C, 8A and especially Fig. 4A). Inmonoplacophorans, the dorsal cartilage is notpresent, and the radula is associated with thevesicles.

In patellogastropods, the subradular membrane israised far above the medial cartilages, though thishas been accomplished in different ways (Fig. 6Band especially Figs. 7D and 8C). In the primitivepatellogastropod condition, the anterior portion ofthe subradular membrane rests on a dorsal cartilagepair located above the dorsolateral pair (Figs. 7D,8C), although in some taxa this situation has beensubsequently modified. More derived conditionsinclude those of: Patella, in which the dorsal pair isreplaced by a connective tissue pad; Patelloida, inwhich the subradular membrane rests on thedorsolateral pair; and lottiids and Rhodopetala, inwhich the dorsal pair is absent and the subradularmembrane rests instead on the hypertrophiedmuscles of the buccal mass. The dorsal cartilagesare only found in the anterior portion of the buccalmass (Fig. 6C). Posterior to the dorsal cartilages,however, the radula still remains dissociated fromthe medials and rests on muscle (Acmaeoidea) or apair of muscle rich dorsomedial cartilages(patellids).

Posterior cartilages show great variability inpresence and absence throughout the Mollusca, andstate assessment is equivocal at the base of theMollusca. They are present at least in some chitons(Kathrina and Mopalia), although they are quitesmall and contained within the connective tissuemaking up the vesicle. Similarly positioned butlarger cartilages are found in the patellid lineages.The posteriors are absent in Monoplacophora andacmaoidean patellogastropods. Inpolyplacophoranand the patellid clade, there are twoposterior cartilages; a dorsal and a ventral pair. Asingle posterior pair is found in other gastropodgroups, potentially a fusion of the two pairs (basedon its large dorsoventral extent). Although absent in


Page - 7

Fissurella, they are present in Haliotis (Crofts, ‘29);thus the character is polymorphic for theVetigastropoda. One posterior cartilage pair ispresent in Nerita.

Some aspects of these radula and cartilagecharacters are best visualized in the three-dimensional reconstructions. One of the morestriking aspects of patellogastropod buccalmorphology is the change in slope of the radulaalong the dorsal surface of the odontophore. Movingfrom posterodorsal to anteroventral, the radula has along region of zero slope (segment 1 in Fig. 6B, D;see also Figs. 4C,D,E) before bending sharplydownwards at an angle usually greater than thirtydegrees (segment 2 in Fig. 6B). This distinctivedownturn occurs toward the anterior end of theradula, more than three quarters along its lengthfrom the radular sac. Importantly, the slope angle ismuch higher in those patellogastropod specimenswhere thebuccal mass is in a semi-protracted state.After this downward sloping, the radula passes overthe anterior portion of the medial cartilages (thebending plane) before moving back posteroventrally(segment 3 in Fig. 6B). In all other groups, theradula usually remains at zero slope from the radularsac posteriorly up to the bending plane anteriorly(segment 1 in Fig. 6D).

Given the distribution of cartilage and radularcharacter states, we want to test whether the radularand cartilage suites show strong or weak patterns ofcorrelated evolution. Testing this question withcharacter suites involves determining if the numberof inferred changes along branches for each suite iscorrelated or independent. We use the samestatistical approach as Smith et al. (‘95): the non-parametric Spearman’s rank correlation. Here weexclude states that are equivocal depending uponoptimization method. Across the twenty-fourbranches in our phylogenetic analysis, the Rs valueis 0.015. This analysis therefore strongly supportsthe claim that radular morphology and cartilagemorphology do not evolve in concert but areuncorrelated. Also, at the node where functionchanges from flexoglossate to stereoglossate, theradula shows no change while the cartilages showmajor changes in shape and position. This suggeststhe “docoglossan” radula may either be flexoglossateor stereoglossate in function.

Structural and functional model of molluscanfeeding

The evolution of characters recognized fromphylogenetic analysis is an important first steptoward determining how changes in structure and

function inter-relate. The next step is to decidewhich character state changes are directly, causallyrelated to function. Models attempting to tietogether structure and function in gastropods havebeen proposed since the work of Huxley (1853).Previous descriptions of morphology and functionhave been exhaustive and excellent, but as Morrisand Hickman (‘81) point out, most models havesimplified the feeding action and thus do not takeinto account the complexity of structure or function.Analysis of all of the components of the buccalapparatus is daunting, particularly without robustphylogenies to detail the sequence of character statechanges, and workers have tended to focus on onemorphological component to the exclusion of theothers. For example, the functional-morphologicalstudies of Vera Fretter and Alastair Graham (Fretterand Graham, ‘62; Fretter, ’65; Graham, ’64, '73)attempted to explain major differences in function byinferring the actions of the numerous musclesinvolved in the feeding process.

Morris and Hickman’s (‘81) now well-established “slit-cylinder” model instead focused onthe morphology of the radula. They conceptualizedthe radula as rolled up on both sides until thebending plane when it is pulled outwards, openingup the cylinder and forming a semicircular crease atthe point where the radula contacts the substrate.This model does make explicit the importance ofradular shape change during feeding. However,Morris and Hickman (‘81) were primarily concernedwith describing the kinematics, in one particulartaxon (Trochidae), of the radula itself and onlyallude to the importance of the cartilage andmuscles. They correctly point out that “protractions[in] species with fundamentally different radularmorphologies and odontophore shapes will revealsomewhat different functional configurations”(Morris and Hickman, ‘81:89). It is precisely thelimited structural and taxonomic focus of theirmodel that makes it inadequate for examiningbroader form-function relationships in the Mollusca.Without reference to all the structures necessary toset the model in motion, it is difficult to interpretwhich structures are necessary and sufficient toexplain function. As well, without knowing whichstates are related to the functional mode of thesystem, it is difficult to apply the Morris-Hickmanmodel to other taxa with differing morphologies.We utilize a biomechanical model of function andcharacter state transformations as reciprocallyilluminating types of explanation that can be appliedto all radulate molluscan groups.


Page - 8

A biomechanical model of radular flexing inMollusca

Because the flexoglossate condition isplesiomorphic given our phylogenetic hypothesis, weuse this condition as our baseline in modelgeneration. We suggest four morphologicalparameters that must be fulfilled in order for theradula to function in a flexoglossate manner. If eachcondition is met, then feeding mode will beflexoglossate. If any of these conditions is not metthen the feeding mode will be stereoglossate.Although our conditions clearly relate tosynapomorphies, we wait until presentation of themodel to fit character state transformations into theframework of the model.

(1) Some sturdy anterior structure must bepresent that will support the membrane as it ispulled over the bending plane. Without structuralsupport, the teeth can neither be flexed outwardalong the edge of the support nor can they impartforces generated by the muscles to the substrate.

(2,3) The radula must be infolded, about alongitudinal axis, and the subradular membraneclosely associated with the underlying structuralsupports, just prior to reaching the anterior end ofthe primary supports (the location of the bendingplane). If the teeth are not folded inward, theycannot flex out as they pass over the bending plane.If the subradular membrane is not closely associatedwith the support structure prior to the bending plane,the teeth cannot unfold along the curved edge of thesupporting structure as discussed below. Folding ofthe radula occurs because a groove is presentbetween the supporting structures in which theradula sits (especially clear in Fig. 7B; see also Fig.8B). This groove must be narrower than the widthof the radula, such that the entire width of theradular ribbon may not fit into the groove withoutfolding.(4) The radular apparatus must be partially or fullyflattened as it passes over the bending plane. Wesuggest this is largely the result of a mechanicalrestriction imposed on the radula as it is protractedand retracted. Consider a transverse section ofradula (one tooth row, for example) as it approachesthe bending plane (Fig. 9A,B), assuming all otherparameters are met and the anterior portion of theodontophore is approximately hemispherical.

Two different measurements of motion can bemade concerning change in position of teeth at thebending plane during protraction and retraction (Fig.9C). The first is circular: the teeth are rotatedaround a fixed point in the odontophore and

consequently move a certain number of degreesabout this point (P); this is the angular excursion, θ.The angular excursion for any point along the toothrow, regardless of whether it lies on the midline(points B in Fig. 9A,B) or lateral margin of theradular ribbon (points A in Fig. 9A,B), will be thesame. Second, the radular teeth also move someabsolute distance through space, which is related tothe arc described by the motion of any particulartooth around the pivot (P). For a point located onthe midline of the radular ribbon (distance from P =radius Rb; Fig. 9C), the absolute distance traveled isgiven by:

Sb = (2π Rb) θwhere θ is in radians; similarly a point on the

lateral margin of the ribbon (radius Ra; Fig. 9C) willtravel a distance given by:

Sa = (2π Ra) θ.If the radula is folded prior to reaching the

bending plane, teeth along the lateral margin will befurther from the pivot point (as shown in Fig. 9C)than those lying along the midline (i.e., Ra > Rb),and hence the absolute distance traveled will also begreater for the lateral-lying teeth (Sa > Sb). This canbe clearly seen when two rows of teeth are viewedtogether as they pass over the bending plane (Fig.9D). As the absolute distance traveled by lateralteeth is greater than that traveled by teeth along themidline, the lateral teeth in successive rows becomespread apart from one another. For this to occur, theradular and subradular membranes must stretch.However, though these membranes are flexible, theyare relatively inelastic due to tanning and depositionof calcium salts (Rinkevich, ‘93). Significanttensional strain is thus implausible, and distancebetween teeth (along the membrane) may not changeduring protraction and retraction: the ribbon mustmaintain its structural integrity. Therefore, theradula must flatten as it begins to rotate around theanterior end of the odontophore. The radular teeth,which are more or less fixed in relation to thesubradular membrane, then passively changeorientation by rotating laterally, or “flexing out.” Inthis way, distances between teeth are not alteredduring the feeding stroke. Though the simpleequations used in this analysis are applicable only toa circular rotary path, a similar argument could beapplied to any surface with curvature.

Beside the flattening caused passively byprotraction of the subradular membrane, two activemechanisms promote greater flattening. Pull frommuscles orginating lateral to the radular apparatusand attaching to the subradular membrane can exert


Page - 9

a lateral component of force on the radula whencontracted. This has the effect of pulling thesubradular membrane (and associated radula) taut asit passes over the bending plane, aiding to flatten thegroove. Flattening can also occur due to dorsaldivarication of supporting structures if the supportsand subradular membrane are closely associated.Dorsal divarication at the bending plane forces thegroove to be flattened by increasing its width.

Biomechanical model and character statetransformations

Polyplacophorans conform to all three structuralparameters of the model, and despite statements tothe contrary (e.g., Salvini-Plawen, ‘88; Ponder andLindberg, ‘97), kinematic studies (Jüch andBoekschoten, ‘80) have shown them to beflexoglossate. In chitons a combination of cartilage,connective tissue and turgidity from vesicular fluidfurnishes the supporting structure at the bendingplane. These elements are highly integrated and thevesicle as a whole acts as the medials alone in theorthogastropods (see below). The space between thetwo vesicles anteriorly is the groove in which theradula sits as it is pulled over the bending plane.The radular apparatus is not situated deep within thegroove but is found more dorsally, lying atop thespace between the medial cartilages (Fig. 8A).

Vetigastropods, Neritopsines and as far as weaware all other Orthogastropoda (all gastropodsexcluding the Patellogastropoda as defined byPonder and Lindberg, ‘97) also fulfill each of theconditions enumerated above, and they have beenclearly demonstrated to be flexoglossate feeders(Ankel, ‘36a, b, ‘38; Eigenbrodt, ‘41; Morris andHickman, ‘81; Hawkins et al., ‘89). The radularteeth and membrane also usually sit low in thegroove between the medials deeper than in Poly- orMonoplacophorans, especially in the neritids (Figs.4B, 7B). During the feeding stroke dorsaldivarication of cartilages and lateral force exerted byslips of the radular protractor muscles have bothbeen shown to be important in increasing flattening.Dorsal divarication was emphasized by Starmühlner(‘52), who showed that ventral approximatormuscles shorten the width at the ventral part of thecartilage and thus divaricate the dorsal part.Redundant systems for flattening may be a way tobuild flexibility in the amount of flattening, and suchflexibility has been shown by Hickman and Morris(‘85) in trochids. A certain basic amount offlattening occurs passively due to protraction andretraction, while greater flattening (and hence

greater flexure) involves a combination of lateralprotractors and cartilage divarication.

The Patellogastropoda fulfill some but not all ofthe conditions described above. The robustcartilages provide ample structural support at thevery anterior end of the odontophore. There is agroove between the supporting cartilages along thelength of the radula that is narrower than the widthof the radular ribbon. Based on the hypothesis ofcharacter evolution, patellogastropods have arelatively plesiomorphic radular configuration, andother taxa with similar radulae (i.e.,polyplacophorans) certainly infold their radula priorto the bending plane so nothing inherent in radularmorphology may be construed as preventinginfolding. To reiterate— based on our sections it isnot that chitons or Orthrogastropoda have foldedradulae and the Patellogastropoda do not. In fact, inour sections all are clearly folded posteriorly andanteriorly. As well, the protractor muscles of thesubradular membrane include both lateral andventral slips (a plesiomorphic condition), ensuringthat some lateral force is exerted during radularprotraction. Thus nothing apparent in the grossmorphology of the buccal mass prevents folding ofthe radula prior to reaching the bending plane.

The main way in which patellogastropods differfrom polyplacophorans and more derived gastropodsis their lack of close association of the radularapparatus and underlying structural supports nearthe bending plane. This dissociation is most clearand pronounced in animals whose radulae are in thesemi-protracted state suggesting that dissociation isachieved both structurally and functionally.Dissociating the subradular membrane from thesupporting structure for most of the length of thebuccal mass means that the radular teeth inpatellogastropods are never in position to rotatearound the edge of the cartilage prior to the bendingplane. Further, the teeth are never forced into thegroove between the cartilages and thus are also notinfolded prior to the bending plane. Thus thepatellogastropod buccal apparatus does not fulfillparameters two and three of our model, and wehypothesize that the structural dissociation ofsubradular membrane from medial cartilages is thecausal change in structure involved in the shift infunction.

Dissociation of radula from cartilage isstructurally accomplished by dissociations inunderlying support structures. The cartilages thatwere plesiomorphically integrated into the vesicles inthe Mono- and Polyplacophora have becomeseparated in Patellogastopoda. In the latter, the


Page - 10

dorsolaterals and medials remain, but the connectivetissue linking the two is lost and only the medialssupport the radula at the bending plane, as inOrthogastropoda. Unlike the orthrogastropods, thesubradular membrane does not become associatedwith the medials posterior to the bending plane.Instead, the subradular membrane rests on dorsalcartilages or musclature. The dorsals and dorsalmuscles (and hence radula) appear to moveanteriorly and even further dorsally relative to themedials during the feeding stroke (based on semi-protracted specimens), causing even greaterdissociation from the medials and increasing thedorsoventral distance the radula travels over thebending plane. Thus the structural changes cascadeinto function and lead to greater dissociation.

Generation of hypotheses from the modelBecause the model attempts to explain the

structural parameters for radula function, we can usethe model to generate explicit hypotheses concerningfunction based on structure for other molluscanclades. Just as importantly, we can also turn theproblem on its head and hypothesize structure basedon function. Hypotheses based on the model are justthat; we can recognize that the model needs to bemodified (at best) or discarded if our hypotheses turnout to be inaccurate . One set of hypotheses we canmake relates to the inference of function in othermolluscan clades. Recent analyses (Wingstrand,‘85; Salvini-Plawen, ‘90; Lindberg and Ponder, ‘96)support the position of polyplacophorans andaplacophorans as basal molluscan clades and sistertaxa to the Conchifera. The basal conchiferan cladeis the Monoplacophora, which itself is the sistertaxon to the scaphopods and bivalves on one branchand the cephalopods and gastropods on the other.Using this broader phylogenetic context, we cansample from these other major clades to determinestructure and kinematics. Monoplacophoransprovide one excellent test case. Themonoplacophoran radula most closely resembles thatof lepetid patellogastropods (Wingstrand, ‘85) but itssupporting apparatus is like that of chitons(Wingstrand, ‘85; Haszprunar, ‘88). Both havefluid-filled vesicles and a radular apparatus restingin the groove formed by the vesicles, resulting infolding of the radula prior to the bending plane.Therefore, we predict [as have Lemche andWingstrand (’59) and Wingstrand (‘85)] thatmonoplacophorans use their radula in a flexoglossatemanner. Literature concerning the feedingmorphology or kinematics of single species in the

other extant molluscan clades (includingScaphopoda, Cephalopoda and Aplacophora) isavailable. Heath (‘05), for example, carefullydescribed both the structure and function of theradular apparatus in the aplacophoran caudofoveateLimifossor talpoideus. In Limifossor, “from thepoint of entrance into the pharynx to their extremeanterior end the two rows of teeth separate along themid line. As the rows of teeth diverge and their longbases shift to the outer rounded surface of thesupports their tips become more inclined outward”(Heath, ‘05: 714). ThusLimifossor appears flexoglossate. As well, theradula in Limifossor is kept deep within the grooveformed by the supporting structure anterodorsally(Heath, ’05: Fig. 4). Thus feeding in Limifossorsupports our model and also bolsters the hypothesisthat flexoglossy is plesiomorphic in Mollusca giventhat the basal Aplacophora and Polyplacophora haveflexoglossate feeding strokes.

Structural and functional information is alsoavailable in the more derived conchiferan cladesScaphopoda and Cephalopoda. The literature oncephalopod feeding morphology and kinematics isscattered. In Loligo pealii, Williams (‘10) states thatalthough the teeth usually point backwards, as theyare pulled over the bending plane they point firstupwards and then forward, suggesting rotationduring feeding. In Octopus, Wells (‘78) observedthat the lateral teeth fan outward during protractionand then move back inward during retraction, againsuggesting flexoglossate kinematics. Finally, Griffin(1900) thought it probable that the long lateral teethin Nautilus are erected so that they no longer covermore central teeth. These different groups havedifferent supporting structures. Loligo does havemedial cartilages, but these appear to be replaced bymuscle and connective tissue in Nautilus andOctopus. Boucaud-Camou and Boucher-Rodoni’s(‘83) cross-sections of Sepia show teeth foldedanteriorly into a groove formed by a combination ofmuscle and cartilage, although muscle appears topredominate at the bending plane. In Scaphopoda,the radular apparatus is folded between medialcartilages, most resembling the situation inVetigastropoda or Neritopsina (Salvini-Plawen, ‘88:Fig. 29), and Lacaze-Duthier (1856: Pl. 10, Fig. 2)diagrams the anterior portion of the radularapparatus flexing outward.

Although our model appears to be consistent withthe data in the literature, we should state animportant caveat. It is not entirely clear in the workof Wells (‘78) on Octopus and Griffin (1900) onNautilus, whether their descriptions of feeding


Page - 11

kinematics are based on primary observations offunction or inferences based on buccal morphologyor “artificial radular protraction.” More rigorousanalysis and sampling of radular kinematics andstructure in these clades are needed to fully test ourmodel.

Another way to test and extend this model is toexamine feeding kinematics and buccal morphologyin more Patellogastropoda. Within thePatellogastropoda, McLean (‘90) thought thatEulepetopsis showed an intermediate conditionbetween stereoglossy and flexoglossy based onexamination of partially extended radulae in fixedmaterial. If McLean (‘90) is correct andneolepetopsids do show flexure, then our model as itstands would be falsified since they would not fulfillthe conditions of the model and would still be atleast partially flexoglossate. As well, lepetids andBathyacmaea have OL teeth that are neither fused toILs nor have overlapping bases as in patellids andAcmaeoidea, and these taxa have not been sampledfor kinematics. It is possible that a highlyintegrated, inflexible radula is also correlated withstereoglossy.

Yet another way to test this model is to insteadinfer structure given function. In Haliotis, juvenilesare stereoglossate feeders (Garland et al. ‘85) whileadults are presumed flexoglossate. We wouldhypothesize that differences in buccal morphologyexist between juvenile and adult abalone and thatthese relate to the fulfillment of our modelparameters. We predict that the radula in juvenileHaliotis is either not in the groove between themedials as it passes over the bending plane or that agroove is missing at the bending plane (or both).These predictions can be tested by examining buccalmorphology in juvenile and adult animals.

Finally, we can predict based on our model thatthe amount of flexing relates to the amount offolding of the subradular membrane (parameter 2 ofour model) and amount of spreading (parameter 3).The amount of flexing varies between taxa and weshould be able to predict how much flexing occursbased purely on morphological relationships.Feeding traces clearly show that teeth farther fromthe pivot point flex more than closer ones, a factconsistent with the model (see Fig. 2C).

Ecological and Evolutionary Implications ofFeeding Patterns

One important variable usually related to thedistinction between stereoglossate and flexoglossatefeeding is force production during the feeding stroke

and thus also utilization of the substrate. Forexample, the Patellogastropoda can penetrate andeven excavate hard substrates while other taxa canonly penetrate softer substrates or brush the surfaceof the substrate (Hawkins et al., ‘89). Thedetermination of a model of force production iscomplex, but flexure or lack thereof is probably oneimportant parameter. We might expect that flexingteeth exert force laterally as well as downward, andthus do not have as strong a net downward forcecompared to the stereoglossate true limpets. It isimportant, however, to clarify that lack of flexure isone of many state changes that might be responsiblefor higher force production. Other potential changesleading to higher force production can be determinedby examining all state changes at the base of thePatellogastropoda.

Two such changes that have been discussed inthe literature are hypertrophied muscles and use ofthe radula during protraction of the odontophore inPatellogastropoda (Fretter and Graham, ‘62;Graham, ‘64). Given our analysis, we suggest thatany model of force production also include the shapeof the medial cartilages and the radular apparatus.Only Patellogastropoda have medials that changedramatically in shape anteroposteriorly. At thebending plane, the cartilage is usually a small nubthat gets taller dorsally toward the posterior beforeleveling out. Perhaps correlated with the shapechange in the medial cartilages, the radularapparatus itself changes slope dramatically at theanterior portion before also leveling out (Fig. 5B),especially in a semi-protracted state. These changesmay or may not affect some aspect of feedingfunction, like force production, but should beincluded in any biomechanical simulation.

Although our analysis focuses on intrinsicaspects of form and function, differences in truelimpet feeding kinematics have long been linked todifferential substrate use and different selectiveenvironments. The ability to penetrate the hardsubstrate has often been related to the ecologicalsuccess of the Patellogastropoda in the rocky shoreintertidal (Branch, ‘81). Although feedingkinematic shifts in the Patellogastropoda are possiblyan adaptation given the selective environment in therocky shore intertidal, such shifts are alsopotentially limiting in both a kinematic andevolutionary sense. In particular, flexoglossy appearsflexible or labile in regards to diet while stereoglossyis not.

As we have shown above, all molluscan groupsbesides Patellogastropoda are likely flexoglossate.These groups are quite diverse in their diet. Chitons,


Page - 12

for example, have carnivorous (Placiphorella) andherbivorous (Lepidochitonia) members. Members ofthe Aplacophora either feed on foraminifera(caudofoveates) or are carnivores living oncoelenterate fluids and tissues (solenogastres).Cephalopods are active carnivores or scavengers,and scaphopods are microcarnivores tomicroomnivores (Salvini-Plawen, ‘88). Members ofthe Orthogastropoda show a wide range of dietaryhabits, including carnivory, herbivory anddetritivory, and these different dietary types havearisen independently numerous times at manyphylogenetic levels (Ponder and Lindberg, ‘97).Thus, flexoglossy is found associated with manydifferent independently derived dietary types,perhaps because this type of feeding mode allows theanimal to: (1) grasp and hold items immobile whilepulling food towards the midline of the esophagus,and (2) keep ingested food from moving backtowards the mouth (Runham, ‘72). Grasping andholding onto food items either directly or afteringestion may be essential for carnivores seekingactive prey, and give extra flexibility in food choiceand method of obtaining food for herbivores. Forexample, herbivorous polyplacophorans can not onlyscrape the microbial film but also snip off andconsume algal blades (Himmelman and Carefoot,‘75). Snipping blades with the radula requires thatthe teeth cut the material like scissors. This is onlypossible for flexoglossate feeders.

The stereoglossate condition, in contrast, doesnot appear as labile. Animals with such a feedingstroke cannot use their radular teeth to grab, hold,cut or bite items, but only to rasp and collect thedislodged food. Based on the interconnected data wedo have concerning structure, function, ecology, andevolution in the patellogastropods, we suggest thefollowing dialectic: (1) Specific novel modificationsin the buccal apparatus have led to novel functionswhich allow the Patellogastropoda to efficientlycollect food in hard substrate habitats (Branch, ‘81),and (2) these same modifications of structure andfunction may also have disallowed thePatellogastropoda some flexibility in feeding modeand food choice, perhaps confining them moreclosely to a particular niche (small particle grazingon hard substrates) than other clades.

ConclusionsWe have tried to emphasize the importance of

phylogenetic analysis, biomechanical models, andindependent kinematic and structural data whenexplaining how form and function change in a

lineage. With our phylogenetic hypothesis, we showthat change in function from flexoglossy tostereoglossy occurs at the base of Patellogastropoda.We also use a simple statistical test to document onekind of dissociation, the discordance of characterstate transformations in systems thought to beintegrated. Contra Graham (‘73), the cartilages andradula do not have to change in concert duringevolution. Thus, even though the patellogastropodradula is relatively plesiomorphic, the cartilages arenot, and the buccal apparatus is used in a differentmanner during the feeding stroke.

We then determine which of the state changes inour phylogenetic analysis directly relate to change infunction. We do this by first building abiomechanical model for the plesiomorphic functionand then determining which parameters must bealtered to produce change in function. Our historicalanalysis and biomechanical model are reciprocallyillumunating in that the model is built on knowledgeof character evolution and changes in modelparameters and in character states are both necessaryin understanding pattern and mechanism of feedingsystem change. For example, patellogastropodradular teeth lack flexure because the radularapparatus is neither folded in the groove norassociated with main supporting structures prior tothe bending plane, an assertion based on the model.The phylogeny shows the state changes that have ledto these conditions not being fulfulled. InPatellogastropoda, the subradular membrane is notassociated with medial cartilages like in othergastropods but instead becomes associated withplesiomorphic dorsolateral and dorsal cartilages andthus is raised far above the medials except at thepoint of the bending plane.

Modification of the feeding systems has been anobvious feature in molluscan diversification, relatedto the important process of gathering energy fromthe environment. Unfortunately, the feeding systemhas not been utilized to its fullest in analyticalstudies. For example, few workers have directlymeasured or modeled force production given thecomponents that make up the buccal apparatus (butsee Padilla, ‘85), and no one has tied thesedifferences to some kind of optimality (or other)criterion to understand potential payoffs fordifferences in feeding mode. Also, other keycomponents involved in kinematics of the buccalsystem have been unstudied, especially the nervoussystem control on muscle activation patterns.Although much remains to be done, the molluscanfeeding system clearly has the potential to be amodel system for understanding and synthesizing the


Page - 13

underlying phylogenetic, developmental, andecological controls on diversity and diversification ofform and function.

ACKNOWLEGMENTS

We thank Winston Ponder, Dianna Padilla andDavid Lindberg for suggestions on earlier versions ofthe manuscript and all the people who helped withcollecting material and providing museumspecimens. Gerhard Haszprunar provided access tomonoplacophoran sections and Dianna Padillaallowed use of unpublished data from videotapes ofchiton and limpet feeding. JohnnyChen and Ajay Bautista produced the three-dimensional reconstructions of the radula apparatusfrom the same target taxa independent fromreconstructions performed by the authors. We wouldespecially like to thank two anonymous reviewers fortheir comments and criticisms and Fredrick Harrisonand the Scientific Visualization Center for providingthe means to make our 3D reconstructionspolychromatic. Animated grey-scale 3Dreconstructions are available for viewing on theWWW(http://www.ucmp.berkeley.edu/people/rpg/cart3D/sfpapsup.html).

LITERATURE CITED

Amaudrut MA 1898. La partie antérieure du tubedigestif de la torsion chez les mollusquesgastéropodes. Ann. Sci. Nat., Zool. Biol. Anim. (7)8:1-291.

Ankel WE. 1936a. Die Freβspuren von Helcion undLittorina und die Funktion der Radula.Verhandlungen der Deutschen ZoologischenGesellschaft. (Zoologischer AnzeigerSupplementband) 9: 174-182.

Ankel WE. 1936b. Prosobranchia. In: Grimpe G.and E. Wagler, editors. Die Tierwelt der Nord- undOstsee, IX.bI. Leipzig: AkademischeVerlagsgesellschaft. 240pp.

Ankel WE. 1938. Erwerb und Aufnahme derNahrung bei den Gastropoden. Verh. Dtsch. Zool.Ges. 40: 223-295.

Boucaud-Camou, R. and R. Boucher-Rodoni. 1983.Feeding and digestion in cephalopods. In: ASMSaleuddin and KM Wilbur, editors. The Mollusca,

Vol. 5/2 – Physiology. New York: Academic Press,pp. 149-187.

Branch GM. 1981. The biology of limpets: Physicalfactors, energy flow, and ecological interactions.Oceanogr. Mar. Biol. 19:235-380.

Carriker MR. 1943. On the structure and functionof the proboscis in the common oyster drill,Urosalpinx cinerea Say. J Morphol. 73: 441-506.

Crofts DR. 1929. Haliotis. Liverpool Marine BiologyCommittee Memoir 29. London: Williams &Norgate. 174 p. 8 pls.

Crofts DR. 1937. The development of Haliotistuberculata, with special reference to theorganogenesis during torsion. Philos. Trans. R. Soc.Lond., Ser. B 228: 219-268.

Eigenbrodt H. 1941. Untersuchungen über dieFunktion der Radula einiger Schnecken. Zeitschriftfur Morphologie und Ökologie der Tiere 37: 735-791.

Fretter V, Graham A. 1962. British ProsobranchMolluscs. London: Ray Society. 755 p.

Fretter, V. 1965. Functional studies of the anatomyof some neritid prosobranchs. J Zool. 147: 46-74.

Garland CD, Cooke SL, Grant JF, McMeekin TA.1985. Ingestion of the Bacteria on the cuticle ofcrustose (non-articulated) coralline algae by post-larval and juvenile abalone (Haliotis ruber Leach)from Tasmanian waters. J Exp. Mar. Biol. Ecol. 91:137-149.

Geddes P. 1879. On the mechanism of theodontophore in certain Molluscs. Trans. Zool. Soc.London 10: 485-491.

Graham A. 1932. On the structure and function ofthe alimentary system of the limpet. Trans. R. Soc.Edinburgh 57: 287-308.

Graham A. 1964. The functional anatomy of thebuccal mass of the limpet (Patella vulgata). Proc.Zool. Soc. Lond. 143: 301-329.

Graham A. 1973. The anatomical basis of functionin the buccal mass of prosobranch and amphineuranmolluscs. J Zool. 169: 317-348.


Page - 14

Griffin L. E. 1900. The anatomy of Nautiluspompilus. Memoirs of the National Academy ofSciences Volume 8. Pp. 109-197, 17 pls.

Guralnick, R. P. and D. R. Lindberg. 1999.Integrating developmental evolutionary patterns andmechanisms: A case study using the Gastropodradula. Evolution 53(2): 447-459.

Haszprunar G. 1988. On the orgin and evolution ofmajor gastropod groups, with special reference to theStreptoneura (Mollusca). J Molluscan Stud. 54: 367-441.

Hawkins SJ, Watson, DC, Hill SS, Harding SP,Kyriakides MA, Hutchinson S, Norton TA. 1989. Acomparison of feeding mechanisms in microphagousherbivorous intertidal prosobranchs in relation toresource partitioning. J Molluscan Stud. 55(2): 151-166.

Heath H. 1905. The morphology of a Solenogastre.Zoologische Jahrbücher 21:703-734

Herrick JC. 1906. Mechanism of the odontophoralapparatus in Sycotypus canaliculatus. Am. Nat.40:707-737.

Hickman CS and TE Morris 1985. Gastropodfeeding tracks as a source of data in analysis of thefunctional morphology of radulae. The Veliger27(4): 357-365.

Himmelman JH and TH Carefoot 1975. Seasonalchanges in calorific value of three Pacific Coastseaweeds and their significance to some marineinvertebrate herbivores. J Exp. Mar. Biol. Ecol.18:139-151.

Hubendick B. 1956. The eating function inLymnaea stagnalis. Ark. Zool. 10(15): 511-521.

Huxley TH. 1853. On the morphology of thecephalous Mollusca, as illustrated by the anatomy ofcertain Heteropoda and Pteropoda collected duringthe voyage of the H.M.S. ‘Rattlesnake’ in 1846-1850. Philos. Trans. R. Soc. Lond., Ser. B 143: 29-65.

Jüch PJW and GJ Boekschoten. 1980. Trace fossilsand grazing traces produced by Littorina andLepidochitona. Dutch Wadden Sea. Geol. Mijnb. 59:33-42

Lacaze-Duthiers H. 1856. Historie de l’organisationet du développement du dentale. Ann. Sci. Nat.,Zool. Biol. Anim. 6: 225-281.

Lauder GV. 1990. Functional morphology andsystematics: Studying functional patterns in ahistorical context. Ann. Rev. Ecolog. Syst. 21: 317-340.

Lemche H and KG Wingstrand. 1959. The anatomyof Neopilina galatheae Lemche, 1957. GalatheaReports 3: 9-71.

Lindberg, DR. 1988. The Patellogastropoda.Malacological Review supp. 4:35-63.

Lindberg, DR. 1998. Order Patellogasropoda. In: P.L. Beesley, G. J. B. Ross, and A. Wells, editors.Mollusca: The Southern Synthesis. Part B. Fauna ofAustralia Vol. 5. Melbourne Australia: CSIROPublishing. p. 639-652.

Lindberg, DR and Hedegaard, C. 1996. A deepwater patellogastropod from Oligocene water-loggedwood of Washington State, USA (Acmaeoidea:Pectinodonta). J Molluscan Stud. 62(3): 299-314.

Lindberg DR and WF Ponder 1996. An evolutionarytree for the Mollusca: Branches or roots? In: J.Taylor, editor. Origins and Evolutionary Radiationsof the Mollusca. Oxford, London: Oxford UniversityPress. p. 67-75.

McLean JH. 1990. Neolepetopsidae, a newdocoglossate limpet family from hydrothermal ventsand its relevance to patellogastropod evolution. JZool. 222(3): 485-528.

Morris TE and CS Hickman 1981. A method forartificially protruding the gastropod radulae and anew model for radula function. The Veliger 24(2):85-90.

Needham J. 1933. On the dissociability of thefundamental processes in ontogenesis. Biol. Rev. 8:180-223.

Padilla DK. 1985. Structural resistance of algae toherbivores: A biomechanical approach. Mar. Biol.90: 103-109.


Page - 15

Plate L. 1897. Die Anatomie und Phylogenie derChitonen. Zoologische Jahrbücher, supp. 4 (Faunachilensis 1): 1-227.

Ponder WF and DR Lindberg. 1997. Towards aphylogeny of gastropod mollusc: An analysis usingmorphological characters. Zool. J Linn. Soc. 119(2):83-265.

Rinkevich B. 1993. Major primary stages ofbiomineralization in radular teeth of the limpetLottia gigantea. Mar. Biol. 117(2): 269-277.

Rudman WB. 1972. Structure and functioning ofthe gut in the Bullomorpha (Opisthobranchia). J Nat.Hist. 6:311-324

Runham NW. 1969. The use of the scanningelectron microscope in the study of the gastropodradula: The radulae of Agriolimax reticulatus andNucella lapillus. Malacologia 9(1): 179-185.

Sasaki, T 1998. Comparative anatomy andphylogeny of the recent Archaeogastropoda(Mollusca: Gastropoda). The University of TokyoBulletin No. 38. 224 p.

Salvini-Plawen LV. 1988. The structure andfunction of molluscan digestive systems. In:Trueman ER and Clarke MR, editors. TheMollusca. Form and Function, 11. Orlando:Academic Press. p. 301-379.

Salvini-Plawen LV. 1990. Origin, phylogeny andclassification of the phylum Mollusca. Iberus 9:1-33.

Starmühlner F. 1952. Zur Anatomie, Histologieund Biologie einheimischer Prosobranchier.Österreichische Zoologische Zeitschrift 3:546-590.

Smith AB, DTJ Littlewood and GA Wray. 1995.Comparing patterns of evolution: larval and adultlife history stages and ribosomal RNA of post-Palaeozoic echinoids. Philos. Trans. R. Soc. Lond.,Ser. B 349:11-18.

Wells, MJ. 1978. Octopus: Physiology andBehaviours of an Advanced Invertebrate. London:Chapman and Hall. 417 p, 3 pls.

Williams, LW. 1910. The Anatomy of theCommon Squid Loligo Pealii Lesueur. Leiden,Holland: E. J. Brill. p. 1-92.

Wingstrand KG. 1985. On the anatomy andrelationships of Recent Monoplacophora. GalatheaReport 16: 7-94.

Woodward MF. 1901. The anatomy ofPleurotomaria beyrichii Hilg. Quarterly Journal ofMicroscopical Science 44:215-268.


Page - 16

TABLESTable 1. Sampling within operational taxonomic units.

Species name OTU N Source Procedure StainingCymbula compressa Cymbula 2 H.C. D N/ACymbula safiana Cymbula 1 H.C. Hi H+EScutellastra longicosta Scutellastra 1 H.C. Hi H+E“Patella” granularis Scutellastra 1 H.C. Hi H+EPatella vulgata Patella 6 UCMP 10635 Hi, D H+E; MNacella concinna Nacella 1 UCMP 10232 Hi H+E

Nacella magellanica Nacella 1 H.C. Hi H+ECellana ardosiaea Cellana 1 H.C. Hi H+ECellana tramoserica Cellana 2 H.C. Hi, 3D H+E;Paralepetopsis terreguvora Paralepetopsis 2 SMNH Uncat. Hi, D H+E; H+GTEulepetopsis vitrea Eulepetopsis 2 USNM860497,

LACM146402Hi, D H+E; H+GT

Iothia fulva Lepeta 2 H.C. Hi WH+E, H+GTLepeta caeca Lepeta 1 H.C. Hi WH+ELepeta concentrica Lepeta 2 H.C. Hi, D H+E; PASPropilidium sp. Propilidium 1 H.C. Hi H+EBathyacmaea sp. Bathyacmaea 3 MNHN uncat. Hi, D, 3D H+E; HH+E, GTPectinodonta aupouria Pectinodonta 1 USNM859397 Hi H+EPectinodonta sp. Pectinodonta 2 H.C. Hi H+EAcmaea mitra Acmaea mitra 3 H.C. Hi, D, 3D H+EAcmaea virginea Acmaea virginea 1 H.C. Hi H+ERhodopatella rosea Rhodopetala 1 H.C. Hi H+EErginus apicina Erginus 1 H.C. Hi H+ENotoacmaea scutum Lottia 1 H.C. D N/APatelloida profunda Patelloida 1 H.C. Hi H+ECollisella scabra Lottia 4 H.C. Hi, 3D H+E; WH+VG;

MPatelloida pustulata Patelloida 1 H.C. Hi H+E

Nerita polita Neritopsinaa 1 CAS021197 Hi, 3D H+EFissurella volcano Vetigastropoda 1 CAS001023 Hi, 3D H+EKathrina tunicata Polyplacophora 1 H.C. Hi MMopalia sp. Polyplacophora 2 H.C. Hi, 3D H+ENeopilina galathea Monoplacophora 1 N/A Hi* H+E

* Histological sections made by Lemche and Wingstrand (‘59) were examined.Abbreviations for Source: H.C., hand collected; BNHM, British Natural History Museum; LACM; Los AngelesCounty Museum; MNHN, Muséum National d'Histoire Naturelle; SMNH; Swedish Museum of Natural History;UCMP, University of California Museum of Paleontology; USNM, United States National Museum. ProcedureAbbreviations: Hi, histology; D, dissection; 3D, three-dimensional reconstruction. Staining Abbreviations: GT,Gomori’s trichrome; H+E, haematoxylin and eosin, HH+E, Heidenhain’s haematoxylin and eosin; M, Mallory’strichrome; PAS, periodic acid Schiff; WH+E, Weigert’s haematoxylin and eosin; VG, Van Geison’s.


Page - 17

Table 2. Radular and cartilage characters sampled in this analysis. Characters are numbered and stateslettered. The numbering and lettering are isomorphic with Figure 3.

Char#

RADULAR CHARACTERS CARTILAGE CHARACTERS

1 Number of inner lateral tooth cusps: a, zeropairs; b, one pair; c, two pairs; d, three or morepairs.

Dorsolateral cartilage: a, present; b, absent.

2 Outer lateral tooth cusps: a, one pair; b, twopairs; c, three pairs; d, four pairs e, more thansix pairs (multiple).

Medials and dorsolaterals: a, linked byconnective tissue to form hollow vesicles; b,not linked.

3 Outer lateral bases attached to the membrane:a, fully attached; b, free.

Shape of medials: a, elongate anteriorly; b,same height from anterior to posterior.

4 Ferrous oxide in inner laterals: a, present; b,absent.

Dorsal cartilage: a, present; b, absent.

5 Ferrous oxide in outer laterals: a, present or b,absent.

Posterior ventral cartilage: a, present; b,absent.

6 Number of marginal teeth (uncini): a, zeropairs; b, two pairs; c, three pairs; d, six pairs;e, ten or more pairs (multiple).

Posterior dorsal cartilage: a, present; b,absent.

7 Outer laterals overlap with inner laterals withina row of radular teeth: a, absent; b, partial; c,complete.

Subradular membrane support: a,connective tissue; b, hollow vesicle; c,strong dorsal cartilage; d, dorsolateralcartilage; e, medial cartilage; f, muscle; g,weak dorsal cartilage.

8 Outer lateral bases extending into the next row:a, absent; b, present.

Medials cartilages: a, fused; b, closelyappositioned; c, far apart anteriorly.

9 Basal plates lying between tooth bases andradular membrane: a, present as a single largeunit; b, present as many small units; c, absent.

Dorsolaterals: a, dorsal of the medials; b,flush with the medials in the dorsoventralplane.

10 Middle (rachidian) tooth: a, present; b, presentbut reduced; c, absent.

Dorsolaterals: a, widely separated from themedials; b, overlapping the medials in themediolateral plane.


Page - 18

Fig. 1. Generalized view of the main structural components involved in the feeding stroke in the (A)normal, retracted position and in (B) protracted position. B.p., bending plane; d.l., dorsal lip; dl.c.,dorsolateral cartilage; m, medial cartilage; o.pr., odontophore protractor; r., radula; r.pr., radularprotractor; r.r., radular retractor; v.l., ventral lip.


Page - 19

Fig. 2. The flexoglossate condition showing teeth rotating around the cartilage and feeding traces forflexoglossy and stereoglossy. A: Dorsal view of Emarginula (Vetigastropoda) showing teeth rotating andflexing outwards from folded to unfolded position near the bending plane. Redrawn from Eigenbrodt (‘41).B: Feeding traces left on the substrate from a stereoglossate feeder (based on Hawkins et al., ‘89) C:Feeding traces left on the substrate from a flexoglossate feeder. Redrawn from Jüch and Boekschoten(‘80). M.c., medial cartilages; r., radula.


Page - 20

Fig. 3. Character state evolution based on the phylogenetic analysis. Characters are numbered, and statesare lettered for the structural characters. For the one functional character, the flexoglossate state is labeledFF and stereoglossate is labelled FS. Refer to the text for a description of characters and states. Ifcharacter evolution along a branch was ambiguous, we placed the character number and a question mark toreflect that ambiguity. A: Radula characters and clade names for Patellogastropoda subclades. B:Cartilage characters and functional character.


Page - 21

Fig. 4. Colorized snapshots from the animated three-dimensional reconstructions. Note labelling on image forcolor-scheme showing various elements. A-C are outgroups and D-F are ingroups (Patellogastropoda) in thephylogenetic analysis. A: Reconstruction of Mopalia mucosa. (Polyplacophora) rotated around the vertical axis.B: Nerita polita (Neritopsina) rotated around the horizontal axis. C: Fissurella volcano (Vetigastropoda) rotatedaround the horizontal axis. D: Collisella scabra, horizontal axis. E: Bathyacmaea sp., horizontal axis. F: Cellanatransomerica, horizontal axis. Arrows denote the change in the position of the radula from anterior to posterior.


Page - 22

Fig. 5. Diagrams of the radula for the outgroups (excluding monoplacophorans) and ingroup flat mounted(as opposed to the position occupied in life) to show the morphology of the individual units (i.e., tooth“rows”). Marginals, laterals and rachidian are labelled for each diagram. A: Fissurellid. B: Neritid. C:Polyplacophoran. D: Patellogastropod. All redrawn from Troschel (1866-1893). Lat., lateral teeth; marg.,marginal teeth; rach., rachidian teeth.


Page - 23

Fig. 6. Schematics of sagittal sections of the radula. The position of the radular sac differs in each groupfrom that shown here but has been generalized: The main elements are labelled. A: Generalized “textbook”version showing the position of radula and cartilages. B: Just off midline cut of the patellogastropodbuccal mass. Medial cartilage remains fused posteriorly and the radula is not associated with cartilageuntil the bending plane. Segments 1, 2, and 3 clarify areas where the slope of the radula changes along itslength. C: More lateral slice of the patellogastropod buccal mass: The teeth are not present here, with onlythe subradular membrane resting on dorsal cartilages. (Vertical line shows place of transverse section forFigure 7C.) D: Just off midline cut of the neritid buccal mass. Segment 1 represents almost zero-slopefrom anterior to posterior. E: More lateral cut of the neritid radula. (Vertical line shows position oftransverse section in Figure 7B.) F: Lateral cut of the polyplacophoran radula through the dorsolateral anddorsal cartilage with the vesicle extending posteriorly past the dorsolaterals. (Vertical line shows place oftransverse cut shown in Figure 7A.). D.c., dorsal cartilages; dl.c, dorsal lateral cartilages; m.c., medialcartilages; p.d.c., posterior dorsal cartilage; p.v.c., posterior ventral cartilage; r, radula; r.m., radularmembrane; r.ves., radular vesicle; s.r.m, sub-radular membrane.


Page - 24

Fig. 7. Histological transverse sections showing the morphology of the anterior portion of the buccal mass.Radula, subradular membrane and cartilages are labelled. Note the position of the subradular membraneand cartilages in each diagram. A: Fissurella volcano. B: Nerita polita. C: Mopalia mucosa. D:Erginus apicina.


Page - 25

Fig. 8. Schematic version of Figure 6 showing the positions and shapes of the cartilage and radulasystems. Important components are labelled on the figure directly. A: Polyplacophoran B: Vetigastropodand Neritopsina C: Patellogastropod. D. cart, dorsal cartilages; d. lat, dorsal lateral cartilages.


Page - 26

Fig. 9. A generalized model for flexure. A: Top view of the radula and cartilage showing many rows ofteeth lying in the groove formed by the medial cartilages (or vesicle). Points A1 and B1 are lateral andmedial points, respectively, within one row, and A2 and B2 the analogous points in the next most posteriorrow. Note groove and medial cartilages (m.c.). B: A cross-sectional view of one tooth row from the sameradula showing the location of points A and B. C: Semi-sagittal view near bending plane. Pivot point (P)is located at center of the hemisphere formed by the horns of the cartilages. The distance (R) to the pivotpoint is greater for the more laterally placed point A than for medial point B. The absolute distance (S)travelled is also greater given that angular excursion stays the same. D: The position of lateral (A) andmedial (B) points in two rows before the bending plane (time 1) and as the teeth cross the bending plane(time 2). The lateral teeth of successive rows (A1 and A2) are spreading apart due the greater absolutedistance they must travel as they are rotated around the pivot.

Documents

Historical and biomechanical analysis of integration and ...R. Guralnick and K. Smith Historical patterns of molluscan feeding Page - 2 INTRODUCTION Joseph Needham ('33) used the wonderful