in the Doushantuo fossils [for example, opalinidsare multinuclear (32)]. Only volvocalean em-bryos show so many rounds of palintomy, but theresulting blastomeres are connected by a systemof cytoplasmic bridges (35) that are not presentin the fossils. The combination of palintomy with-in a multilayered cyst wall and peanut-shapedgermination stages as seen in the fossils conformsto the pattern seen in nonmetazoan holozoans;nonetheless, there are no discrete characters in theDoushantuo fossils that are uniquely holozoan.The “animal embryos” likely represent nonmeta-zoan holozoans or possibly even more distanteukaryote branches.

References and Notes1. S. Xiao, Y. Zhang, A. Knoll, Nature 391, 553

(1998).2. S. Xiao, Paleobiology 28, 244 (2002).3. C. Yin, S. Bengtson, Z. Yue, Acta Palaeontol. Pol. 49,

1 (2004).4. J.-Y. Chen et al., Science 312, 1644 (2006).5. J. W. Hagadorn et al., Science 314, 291 (2006).6. L. Yin et al., Nature 446, 661 (2007).7. P. A. Cohen, A. H. Knoll, R. B. Kodner, Proc. Natl. Acad.

Sci. U.S.A. 106, 6519 (2009).8. J.-Y. Chen et al., Proc. Natl. Acad. Sci. U.S.A. 106, 19056

(2009).9. J. V. Bailey, S. B. Joye, K. M. Kalanetra, B. E. Flood,

F. A. Corsetti, Nature 445, 198 (2007).10. E. C. Raff et al., Proc. Natl. Acad. Sci. U.S.A. 105, 19360

(2008).11. R. J. Horodyski, J. Bauld, J. H. Lipps, C. V. Mendelson,

in The Proterozoic Biosphere: A Multidisciplinary Study,

J. W. Schopf, C. Klein, Eds. (Cambridge Univ. Press,Cambridge, 1992), pp. 185–193.

12. J. D. Schiffbauer, S. Xiao, K. S. Sharma, G. Wang,Geology, 10.1130/G32546.1 (2011).

13. Materials and methods are available as supportingmaterial on Science online.

14. S. Xiao, C. Zhou, X. Yuan, Nature 446, E9, discussion E10(2007).

15. Y.-S. Xue, T.-F. Tang, C.-L. Yu, C.-M. Zhou, Acta Palaeont.Sin. 34, 688 (1995).

16. P. C. J. Donoghue, Nature 445, 155 (2007).17. J. V. Bailey, S. B. Joye, K. M. Kalanetra, B. E. Flood,

F. A. Corsetti, Nature 446, E10 (2007) Reply.18. S. Xiao, J. W. Hagadorn, C. Zhou, X. Yuan, Geology 35,

115 (2007).19. P.-J. Liu, C.-Y. Yin, S.-M. Chen, F. Tang, L.-Z. Gao,

Acta Geosci. Sin. 30, 457 (2009).20. Z. Yin et al., Precambr. Res., published online 9 September

2011 (10.1016/j.precamres.2011.08.011).21. A. Rose, in Cell Division Control in Plants, D. P. S. Verma,

Z. Hong, Eds. (Springer, Heidelberg, 2007),pp. 207–230.

22. L. Mendoza, J. W. Taylor, L. Ajello, Annu. Rev. Microbiol.56, 315 (2002).

23. K. V. Mikhailov et al., Bioessays 31, 758 (2009).24. W. L. Marshall, M. L. Berbee, Protist 162, 33 (2011).25. M. Pekkarinen, K. Lotman, J. Nat. Hist. 37, 1155 (2003).26. L. Mendoza, R. A. Herr, S. N. Arseculeratne, L. Ajello,

Mycopathologia 148, 9 (1999).27. A. Franco-Sierra, P. Alvarez-Pellitero, Parasitol. Res. 85,

562 (1999).28. S. Raghu-Kumar, Bot. Mar. 30, 83 (1987).29. B. S. Leander, J. Eukaryot. Microbiol. 55, 59 (2008).30. J. T. Bonner, Integr. Biol. 1, 27 (1998).31. M. Elbrächter, Helgol. Meersunters. 42, 593 (1988).32. K. Hanamura, H. Endoh, Zoolog. Sci. 18, 381 (2001).33. D. P. Molloy, D. H. Lynn, L. Giamberini, Dis. Aquat. Organ.

65, 237 (2005).

34. M. D. Herron, A. G. Desnitskiy, R. E. Michod, J. Phycol.46, 316 (2010).

35. K. J. Green, D. L. Kirk, J. Cell Biol. 91, 743 (1981).

Acknowledgments: We thank S. Xiao, T. Cavalier-Smith, andB. Landfald for discussion; T. Hode and Z. Yue for field-workcollaboration; A. Groso for assistance with the srXTM work;and P. Varvarigos and D. Elliott for the use of fig. S7. The workwas supported by the Swedish Research Council, NaturalEnvironment Research Council, Ministry of Science andTechnology of China, National Natural Science Foundation ofChina, EU FP7, and the Paul Scherrer Institute. Figured ormeasured specimens are deposited at the Swedish Museum ofNatural History and the Museum of Earth Science, ChineseAcademy of Geological Sciences. The srXTM investigations wereconducted at the X04SA and X02DA (TOMCAT) beamlines ofthe Swiss Light Source. The data were visualized and analyzedby using Avizo software. Data are available in the SOM. S.B.and P.C.J.D. designed the research and wrote the paper; T.H.found the nucleic structures, prepared the correspondingvisualizations, and wrote the specimen descriptions in theSOM; J.A.C. found the propagule-like structures and performedtaphonomic analyses and volumetric measurements; C.Y.and S.B. did the field work; C.Y. provided the additional datafrom Hubei; and M.S., F.M., S.B. and P.C.J.D. designed thesrXTM experiments.

Supporting Online Materialwww.sciencemag.org/cgi/content/full/334/6063/1696/DC1Materials and MethodsSOM TextFigs. S1 to S7Table S1References (36–67)Movies S1 to S5

8 June 2011; accepted 16 November 201110.1126/science.1209537

From Flat Foot to Fat Foot: Structure,Ontogeny, Function, and Evolutionof Elephant “Sixth Toes”John R. Hutchinson,1 Cyrille Delmer,2 Charlotte E. Miller,1 Thomas Hildebrandt,3

Andrew A. Pitsillides,1 Alan Boyde4

Several groups of tetrapods have expanded sesamoid (small, tendon-anchoring) bones intodigit-like structures (“predigits”), such as pandas’ “thumbs.” Elephants similarly have expandedstructures in the fat pads of their fore- and hindfeet, but for three centuries these have beenoverlooked as mere cartilaginous curiosities. We show that these are indeed massive sesamoidsthat employ a patchy mode of ossification of a massive cartilaginous precursor and that thepredigits act functionally like digits. Further, we reveal clear osteological correlates of predigit jointarticulation with the carpals/tarsals that are visible in fossils. Our survey shows that basalproboscideans were relatively “flat-footed” (plantigrade), whereas early elephantiforms evolved themore derived “tip-toed” (subunguligrade) morphology, including the predigits and fat pad, ofextant elephants. Thus, elephants co-opted sesamoid bones into a role as false digits and usedthem for support as they changed their foot posture.

The enlarged radial sesamoid bones of giantpanda forefeet (1, 2) are classic examplesof evolutionary exaptation (3, 4): co-option

of old structures for new functions. It is lesswidely recognized that such “sixth toes” or “falsethumbs” have evolved convergently in numeroustetrapods, such as moles and frogs (5, 6). Theyexist in numerous mammals in a less enlargedstate, variably called the prepollex/prehallux (here

called predigits), radial/tibial sesamoids, or otherterms (such as falciform, accessory scaphoid, ornavicular). Whether these sesamoids are ances-trally or convergently evolved in various tetra-pod clades remains to be determined. The latterseems likely, given the absence of similar sesa-moids in most fossil outgroups, yet a cartilag-inous nodular precursor cannot be excluded.Regardless, enlarged sesamoids are quite prom-

inent in both the manus (forefeet) and the pedes(hindfeet) of elephants, where they have beenmistaken for sixth digits or otherwise presumedto play a role in foot support (7–9). Indeed, therecent discovery that moles have developmen-tally switched their radial sesamoid (prepollex)to a digit-like identity (10) intimates that ele-phants and other species may have done the same.Here, we report a multidisciplinary anatomical, his-tological, functional, and phylogenetic analysis (11)of the predigits in elephant feet. We hoped thiswould illuminate how elephants evolved their char-acteristic subunguligrade (nearly “tip-toed,” withonly distal toes contacting the ground) foot postureand function, as compared with the plesiomorphicplantigrade (“flat-footed,” with wrists/ankles con-tacting the ground) foot posture in many othertetrapods.

In 1710, Blair (7) provided the first detailedosteological description of elephants, conclud-ing that they have six toes. The “sixth toes”(medialmost position; corresponding to digit zero)were later identified as the enigmatic prepollex

1Department of Veterinary Basic Sciences and Structure andMotion Laboratory, The Royal Veterinary College, Hatfield AL97TA and London NW1 0TU, UK. 2Department of Palaeontology,The Natural History Museum, Cromwell Road, London SW75BD, UK. 3Leibniz Institute for Zoo and Wildlife Research,im Forschungsverbund Berlin e.V., Postfach 601103, BerlinD-10252, Germany. 4Dental Physical Sciences, Barts and TheLondon School of Medicine and Dentistry, Queen Mary Uni-versity of London, Mile End Road, London E1 4NS, UK.

www.sciencemag.org SCIENCE VOL 334 23 DECEMBER 2011 1699

REPORTS

on

Dec

embe

r 23

, 201

1w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

and prehallux (8, 9, 12) (figs. S1 to S4). Threecenturies of sporadic discussion about the iden-tities of predigits in tetrapods have ensued (8, 9),sometimes returning to the question of whetherthey are actually atavistic digits (9, 13). Consid-ering the characteristic variability (8) and apparentmineralization late in the ontogeny of sesamoids(14), as well as their articulations (15) with meta-carpal I/tarsal I and metatarsal I (Fig. 1 and movieS1), the prepollex/prehallux of elephants mustcorrespond to the radial/tibial sesamoids of othertetrapods. The late mineralization and confoundedscientific history of predigits have tended to pre-vent their preservation, discovery, scholarly descrip-tion, and even museum exhibition. The vexingissue of the homology of elephant predigits re-mains unresolved, complicated by the specializa-tion of paenungulate outgroups (such as Sireniaand Hyracoidea).

Despite early studies, it remains unclearwhether elephant predigits are never more thancartilaginous rods, as current literature assumes(8, 9, 12), or whether they become true bones atsome point in ontogeny. We used a combination(11) of dissection, computed x-ray tomography(CT) scans, histology, and backscattered elec-tron scanning electron microscopy (BSE SEM)to address this question (Fig. 2 and figs. S5 toS13). Using this combination of methods, wefound that elephant predigits initially form asmassive, purely cartilaginous rods and that thesecan become further stiffened through a slow con-version to bone (that is, forming endochondrally)by an unusual ossification mechanism. Histolog-ical examination showed that this initial hyalinecartilage element lacks a preferential orientationof chondrocytes or growth-plate–like stratification(fig. S13). Imaging with BSE SEM and CT inaddition to histology revealed that patches of thiscartilage calcify and are resorbed and replacedby bone that subsequently models to a foam- orhoneycomb-like cancellous (spongy bone) struc-ture. The advancing mineralizing fronts and thethickness of the calcified cartilage layers resem-ble those seen in mature articular cartilage.

Together, our analyses not only show that thecartilaginous predigits are slowly replaced bybone during late ontogeny, but that this bone isunusual in its development [Fig. 2, supportingonline material (SOM) text, and figs. S5 to S13].Ossification typically begins years after other ses-amoids have become well mineralized (for ex-ample, the proximal digital sesamoids, at ~3 to7 years of age), and it occurs in a large cartilagestructure surrounded by a fat pad rather than bytendon or ligament. Such ossification can remainincomplete [in 10 out of 37 (10/37) feet exam-ined] or even uninitiated (11/37 feet) in some adult(~20+ years old) individuals (figs. S4 to S6). Thissingular mode of ossification is endochondral, ex-tending from several seemingly haphazardly po-sitioned centers within the massive cartilaginousprecursor. Furthermore, BSE SEM and CT indi-cate that the resultant cancellous (spongy) bone,unlike others in the appendicular skeleton, does

not seem oriented to match any predominant load-ing direction and lacks compact cortices, whichcould confer greater longitudinal bending stiffness.This indicates an unusually flexible ossified struc-ture that nevertheless is stiffer than the surroundingfat pad or cartilage, although even cartilaginousenlarged predigits should provide some support.

We used an indirect approach to solve thedifficult question of how elephant predigits func-tion. Elephant predigits are deeply embedded inthe digital cushions or fat pads of the feet, thustheir positions and motions are obscured. Thethick keratinized skin of elephant feet preventsultrasound or x-ray imaging at safe intensities,thus preventing in vivo investigation. We pre-viously speculated that elephant predigits mightfunction as strut-like weight supports, becausethey grow with strong positive allometry simi-lar to that of the metapodials (16). This functionwould be expected to involve a static orientationof the predigits during loading. Alternatively, pre-digits might function as dynamic levers (more

like mobile digits) if they reoriented when loaded,rotating about their joint(s). Animals variably em-ploy similar functions with their true digits (17).

We tested these hypotheses by staticallyloading cadaveric elephant feet ex vivo and CT-scanning them to examine the effects of appliedloads on their orientation (11). Predigits behav-ing in a weight-supporting role should maintaina constant orientation, whereas predigits actingas dynamic levers should display joint mobility(movie S2) that reorients them with increasingload. Our reconstructions (Fig. 3) reveal that theprepollex and prehallux act differently whenloaded: The prepollex does not move apprecia-bly even though its proximal joint allows somemobility, whereas the prehallux rotates caudo-dorsally. Internal motion contributing to this ro-tation is apparent for the prehallux which, onceossified at least, is consistently split into prox-imal (fixed to the first metatarsal and tarsal) anddistal (free to move) segments (evident in 8/8individuals with well-ossified prehalluces; movie

Fig. 1. Foot anatomy inhumans and elephants,with sesamoids shown inwhite. (Top) Diagram ofhuman manus and pes(for comparison). Dottedlines for the prepollexand prehallux show roughapproximations of wherethese structures would liein humans, but they arenormally absent. Thesepredigits are not to beconfused with the paireddigital sesamoids, whichelephants and humanshave more distally in theirdigits—the so-called “tib-ial sesamoid” in humansis one of these. (Middleand bottom) Elephantfoot anatomy in medialview of right feet. Themanus is on the left [pre-pollex (dark) and meta-carpal I shown below];the pes is on the right[prehallux (dark) and me-tatarsal I shown below].Bottom-row images arefrom CT scan reconstruc-tions of specimen no. 4(table S1). See movie S2for representative mobil-ity of a predigit. Osteo-logical terms are from(25, 26). Labels are as fol-lows: ac, accessorium (pi-siform); ca, calcaneus;D3, third digit; ds, digi-tal sesamoid(s); mc1,metacarpal I; mt, meta-tarsal I; ph, prehallux;pp, prepollex.

50mm50mm

20mm20mm

50 mmca

ca

ac

ppph

ph

pp

ph

ppds

ds

mc1 mt1

mc1mt1

ac

mc1

mt1

ds dsD3

D3D3

D3

23 DECEMBER 2011 VOL 334 SCIENCE www.sciencemag.org1700

REPORTS

on

Dec

embe

r 23

, 201

1w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

S1). The prehallux thus has a proximal portionthat statically transfers load to the tarsus (anal-ogous to the whole prepollex), and a distal, mo-bile, lever-like portion. Such segmentation wasnot apparent in any of our prepollex specimens,which behaved as simple struts. This differencein prepollex and prehallux mobility and function

may relate to the more upright manus and morehorizontal pes bone orientations (Fig. 1). Bothtypes of predigits, however, are particularly wellsuited to stiffen the highly compliant fat padagainst excessive deformation. Furthermore, thepredigits’ tight syndesmotic articulations withthe carpus and tarsus indicate that they also are

able to transfer loading proximally from the soleof the fat pad to those bones, partly bypassingthe digits. Therefore, the enlarged predigits renderelephant feet functionally plantigrade while thetrue digits remain in subunguligrade orientations.Indeed, the predigits may allow elephants to ef-fectively reduce the degrees of freedom in their

Fig. 2. Histology of elephant predigit, from speci-men no. 2 (table S1) prepollex. (A) Toluidine bluehistology of bone:cartilage interface [proximal slab4 (fig. S6); cartilage, dark blue, bone, pale blue, bonemarrow space, white; width = 1200 mm]; see also fig.S13. (B) BSE SEMmacerated slab 1 (width = 34mm).The large space in the right central area (see also fig.S9) was occupied by cartilage and shows theendochondral mineralization front [higher magnifi-cation in (C), width = 1204 mm]. (D) BSE SEM ofpolymethylmethacrylate-embedded slab 0 (width =28 mm; see also fig. S7) with a pseudocolor look-up table. The lowest backscattering coefficient(top) is level from the monobrominated standardand highest at 255 from the monoiodinated di-methacrylate standard (27); the densest phase iscalcified cartilage. (E) Higher-magnification grayimage of the calcified cartilage:bone interface(width = 900 mm). Enlarged versions of images (B)to (E) are in figs. S8 and S10 to S12.

A

B C

D E

Fig. 3. Passive motion of elephant predigits underloading. Right cadaveric manus (top row) and pes(bottom row) specimens under minimal (left) andmaximal (right) loads are shown. In the manus, theprepollex does not move noticeably relative to thevertical, whereas the metacarpal dorsiflexes up to13° at maximal load. In the pes, the distal segmentof the prehallux rotates around the static proximalsegment, dorsiflexing up to 17° as the metatarsaldorsiflexes up to 10°. Bones (Fig. 1) are colored tomatch movies S1, S3, and S4. Predigits are aqua-marine color. Specimen numbers from table S1are no. 3 (manus) and no. 5 (pes). Labels are as fol-lows: MC3, metacarpal 3; MT3, metatarsal 3; ph,prehallux; pp, prepollex.

50mm 50mm

50mm50mm

ph ph

pp

pp

MT3MT3

MC3MC3

www.sciencemag.org SCIENCE VOL 334 23 DECEMBER 2011 1701

REPORTS

on

Dec

embe

r 23

, 201

1w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

feet, by providing a more passive stabilizing sup-port that reduces need for more active and mas-sive muscular tissues, analogous to the reductionof toes in other ungulate groups (18). Yet thepersistence of musculotendinous structures an-chored to these sesamoids [such as the abductorpollicis (9)] indicates some retained ability tocontrol their position or caudolateral motion, sothe predigits are not entirely passive structures.

There is a smooth ridge on the caudomedialsurface of metacarpal I with which the prepollexarticulates, as well as a mobile ball-and-socket–like joint on the distal end of tarsal I and a ridgeon the caudomedial side of metatarsal I that botharticulate with the prehallux (15). These features,found even in juvenile elephants that lack ossifiedpredigits, are thus osteological correlates of thepresence of predigits (Fig. 1) that might be iden-tifiable in fossils. Their presence in any skeletalspecimen would corroborate the existence of en-larged predigits (cartilaginous or ossified).

Our survey of the fossil record of the cladeProboscidea revealed some evidence of predigitsin extinct forms (11), which also clarifies howelephant foot posture and function evolved. Un-fortunately the most basal proboscideans (suchas Barytherium and Numidotherium) lack suffi-ciently well-preserved metapodials (and thus po-tential evidence of predigit articulations) to more

directly test whether they had large predigits.However, their preserved proximal carpal andtarsal elements show that the feet were quiteplantigrade, leaving little space for an expandeddigital cushion or predigits (Fig. 4, movies S3and S4, and SOM text). Furthermore, the artic-ulations of more distal foot bones indicate thepresence of relatively dorsiflexed and more splayed(abducted) toes; not as adducted as in later Pro-boscidea and consistent with a more amphibiouslifestyle. Hence we infer that basal proboscideans,like many of their amphibious or wholly aquatictethytherian outgroups [Sirenia and Embrithopoda(19)] were more plantigrade than extant elephants,as is ancestral for tetrapods.

We therefore hypothesize that the evolution ofmore subunguligrade toes in elephants is linkedwith the expansion of the manual and pedal digitalcushions and their supporting predigits. In thisscenario, the predigits increasingly adopted thesupportive roles that were played by the carpals(e.g., pisiform) and tarsals (e.g., calcaneus) in moreplantigrade basal Proboscidea. Representativeelephantiform and deinothere taxa along the phy-logeny (Fig. 4) before Elephantidae support thishypothesis (SOM text, figs. S14 to S16, andmoviesS3 and S4): All well-preserved taxa exhibit smallerproximal carpal/tarsal bones and foot bone articu-lations that are more consistent with increased

dorsiflexion of the toes, and thus amore subunguli-grade toe posture relative to the ancestral conditionfor Proboscidea. All of these taxa display osteolog-ical correlates for the articulation of predigits in themanus and pes. Thus, we conclude that the pre-digits have served to stiffen the expanded fat padand maintain a plantigrade-like foot function, trans-ferring loads from the substrate to the carpus/tarsus,since early in elephantiform evolution.

Extant elephants have remarkable feet thatcombine advantages of plantigrady [such as thepotential for damping impacts at heelstrike (20),larger foot surface area and thus moderatedpressures (21), large translations of the center ofpressure during the stance phase involving pro-nounced heelstrike, dynamic gearing, and toe-off dynamics (17)] with those of digitigrady orsubunguligrady [such as reasonable mechanicaladvantage of the toes to keep supportive tissuestresses at safe levels (22), or even potential ben-efits to metabolic economy from elastic energystorage (23)]. These changes occurred while earlyelephantiforms attained gigantism (>2000 kgof body mass or shoulder height >2 m) in theEocene epoch (~40 million years ago, Fig. 4) andoccupied a wider range of terrestrial habitats, be-coming less amphibious around the node joiningDeinotheriinae and Elephantiformes (Fig. 4).Hence, there is probably a link between the in-

Fig. 4. Evolution of pro-boscidean foot posture.A stratigraphically time-calibrated axis is shownat top, using the phyloge-netic tree from (28–30),with clades Proboscidea,Elephantiformes, and El-ephantoidea labeled atnodes; the Sirenia (sea-cows; manatees and du-gongs) extant outgroupis shown. Manus (on left)and pes (on right) speci-mens are shown in ap-proximate osteologicallyneutral poses in lateralview (more explanationand images are in theSOM text and figs. S14 toS16). Movies S3 and S4show three-dimensionalfoot reconstructions andpredigit articular surfaces(where present). A shiftfrom a relatively moreplantigrade manus andpes in Numidotheriumand Barytherium to moresubunguligrade feet inlater taxa is evident, es-pecially when articularsurfaces are compared.Shoulder heights (top ofscapula) for each genus are roughly estimated in parentheses, as a proxy for body size changes. Representative skeletons of Barytherium (top) and Deinotherium(bottom) are shown with approximate relative size differences.

PLIOPaleP

Sirenia

Erytherium

Numidotherium(1m)

Barytherium(2m)

Phiomia(<2m)

Deinotheriinae(3m)

Mammutamericanum(3m)

Gomphotherium(3m)

Elephantidae(3m)

60 my 50 my 40 my 30 my 20 my 10 my

Paleocene Oligocene Miocene Plio IVEocene

Proboscidea

Elephantiformes

Elephantoidea

Increasingterrestriality& gigantism(>2m)

23 DECEMBER 2011 VOL 334 SCIENCE www.sciencemag.org1702

REPORTS

on

Dec

embe

r 23

, 201

1w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

creasing demands of supporting and movinggreater weight on land and the benefits of havingmore upright toe bones but directing some loadsaway from the toes with the predigits and fatpad, which resulted in the peculiar compromisethat persists in the feet of extant elephants.

The recognition of elephant predigits as en-larged sesamoids that perform digit-like functionsfuels inspiration for examining the evolution offoot function, terrestriality, and gigantism in otherlineages. Sauropod dinosaurs had expansive footpads, particularly in their pedes (24); however,no evidence of predigits has been found. Con-sidering that the predigits form on the medialborder of the feet, they would tend to be lost ifdigit I is lost or reduced, as it was in early peris-sodactyls and artiodactyls. This loss might limitfoot pad expansion and thereby explain whyrhinos and hippos seem to lack predigits [but see(18) for a possible rudimentary pollex in hippos]and have less expanded foot pads than elephantsdo (8). Regardless, the previously misunderstoodand neglected predigits of elephants now deserverecognition as a remarkable case of evolutionaryexaptation (4), revealing how elephants evolvedtheir specialized foot form and function.

References and Notes1. D. D. Davis, Fieldiana 3, 1 (1964).2. H. Endo et al., J. Anat. 195, 295 (1999).3. S. J. Gould, Nat. Hist. 87, 20 (1978).4. S. J. Gould, E. S. Vrba, Paleobiology 8, 4 (1982).

5. M. Fabrezi, Zool. J. Linn. Soc. 131, 227 (2001).6. M. R. Sánchez-Villagra, P. R. Menke, Zoology 108, 3

(2005).7. P. Blair, Philos. Trans. 27, 53 (1710).8. H. Neuville, Arch. Mus. Natl. Hist. Nat. Paris 13, 6e Serie,

111 (1935).9. K. von Bardeleben, Proc. Zool. Soc. 1894, 354

(1894).10. C. Mitgutsch et al. Biol. Lett., 10.1098/rsbl.2011.0494

(2011).11. Materials and methods are available as supporting

material on Science Online.12. G. E. Weissengruber et al., J. Anat. 209, 781 (2006).13. F. Galis, J. J. M. van Alphen, J. A. J. Metz, Trends Ecol.

Evol. 16, 637 (2001).14. J. Prochel, P. Vogel, M. R. Sánchez-Villagra, J. Anat. 205,

99 (2004).15. J. R. Hutchinson, C. E. Miller, G. Fritsch, T. Hildebrandt,

in Anatomical Imaging: Towards a New Morphology,R. Frey, H. Endo, Eds. (Springer, Berlin, 2009),pp. 23–38.

16. C. E. Miller, C. Basu, G. Fritsch, T. Hildebrandt,J. R. Hutchinson, J. R. Soc. Interface 5, 465 (2008).

17. D. R. Carrier, N. C. Heglund, K. D. Earls, Science 265,651 (1994).

18. A. B. Clifford, J. Vertebr. Paleontol. 30, 1827 (2010).19. N. Court, Palaeontogr. Abt. A 226, 125 (1993).20. D. E. Lieberman et al., Nature 463, 531 (2010).21. F. Michilsens, P. Aerts, R. Van Damme, K. D’Août,

J. Zool. (London) 279, 236 (2009).22. A. A. Biewener, Science 250, 1097 (1990).23. M. N. Scholz, M. F. Bobbert, A. J. van Soest, J. R. Clark,

J. van Heerden, J. Exp. Biol. 211, 3266 (2008).24. M. F. Bonnan, in Thunder-Lizards: the Sauropodomorph

Dinosaurs, K. Carpenter, V. Tidwell, Eds. (Indiana Univ.Press, Bloomington, IN, 2005), pp. 346–380.

25. M. M. Smuts, A. J. Bezuidenhout, Onderstepoort J. Vet. Res.60, 1 (1993).

26. M. M. Smuts, A. J. Bezuidenhout, Onderstepoort J. Vet. Res.61, 51 (1994).

27. A. Boyde, R. Travers, F. H. Glorieux, S. J. Jones,Calcif. Tissue Int. 64, 185 (1999).

28. C. Delmer, Acta Palaeontol. Pol. 54, 561 (2009).29. E. Gheerbrant, Proc. Natl. Acad. Sci. U.S.A. 106, 10717

(2009).30. E. Gheerbrant, P. Tassy, C. R. Palevol. 8, 281 (2009).

Acknowledgments: We thank the staff of the Structureand Motion Laboratory of the Royal Veterinary College forassistance and three anonymous reviewers for constructivecriticism. Many individuals assisted with the collection ofthe cadaveric data; we particularly thank the European-basedzoos that provided the specimens and G. Fritsch for CTscans done in Germany. O. Cosar, R. Weller, A. Wilson, andK. Jespers assisted with the ex vivo loading experiments.J. Molnar assisted with Figs. 1 to 4 and the movies. This projectwas funded by the Biotechnology and Biological Sciences ResearchCouncil (BBSRC) (grants BB/C516844/1 and BB/H002782/1to J.R.H.). Additionally, A.A.P. appreciates funding fromArthritis Research UK and the BBSRC, and A.B. was supportedby the Veterinary Advisory Committee of the UK HorseraceBetting Levy Board. The data reported in this paper aretabulated in the SOM. The authors declare no conflictsof interest.

Supporting Online Materialwww.sciencemag.org/cgi/content/full/334/6063/1699/DC1Materials and MethodsSOM TextFigs. S1 to S16Tables S1 to S3References (31–41)Movies S1 to S4

20 July 2011; accepted 8 November 201110.1126/science.1211437

Global Seabird Response to ForageFish Depletion—One-Third for the BirdsPhilippe M. Cury,1* Ian L. Boyd,2* Sylvain Bonhommeau,3 Tycho Anker-Nilssen,4

Robert J. M. Crawford,5 Robert W. Furness,6 James A. Mills,7 Eugene J. Murphy,8

Henrik Österblom,9 Michelle Paleczny,10 John F. Piatt,11 Jean-Paul Roux,12,13

Lynne Shannon,14 William J. Sydeman15

Determining the form of key predator-prey relationships is critical for understanding marineecosystem dynamics. Using a comprehensive global database, we quantified the effect offluctuations in food abundance on seabird breeding success. We identified a threshold in prey(fish and krill, termed “forage fish”) abundance below which seabirds experience consistentlyreduced and more variable productivity. This response was common to all seven ecosystems and14 bird species examined within the Atlantic, Pacific, and Southern Oceans. The thresholdapproximated one-third of the maximum prey biomass observed in long-term studies. Thisprovides an indicator of the minimal forage fish biomass needed to sustain seabirdproductivity over the long term.

Public and scientific appreciation for therole of top predators in marine ecosystemshas grown considerably, yet many upper

trophic level (UTL) species, including seabirds,marine mammals, and large predatory fish, re-main depleted owing to human activities (1–4).Fisheries impacts include direct mortality of ex-ploited species and the more subtle effects ofaltering trophic pathways and the functioning ofmarine ecosystems (5). Specifically, fisheries forlower trophic level (LTL) species, primarily small

coastal pelagic fish (e.g., anchovies and sar-dines), euphausiid crustaceans (krill), and squid(hereafter referred to as “forage fish”), threatenthe future sustainability of UTL predators inmarine ecosystems (6, 7). An increasing globaldemand for protein and marine oils contributespressure to catch more LTL species (8). Thus,fisheries for LTL species are likely to increaseeven though the consequences of such activityremain largely unknown at the ecosystem level. Itremains challenging, however, to assess fishing

impacts on food webs because numerical re-lationships between predators and prey are oftenunknown, even for commercially valuable fish(9, 10). Ecosystem models and ecosystem-basedfisheries management, for which maintaining

1Institut de Recherche pour le Développement, UMR EME-212,Centre de Recherche Halieutique Méditerranéenne et Tropi-cale, Avenue Jean Monnet, BP 171, 34203 Sète Cedex, France.2Scottish Oceans Institute, University of St Andrews, St AndrewsKY16 8LB, UK. 3Ifremer, UMR EME 212, Centre de RechercheHalieutiqueMéditerranéenne et Tropicale, Avenue JeanMonnet,BP 171, 34203 Sète Cedex, France. 4Norwegian Institute forNature Research, Post Office Box 5685 Sluppen, NO-7485Trondheim, Norway. 5BranchOceans and Coasts, Department ofEnvironmental Affairs, Private Bag X2, Rogge Bay 8012, SouthAfrica. 6College of Medical, Veterinary and Life Sciences, Uni-versity of Glasgow, Glasgow G12 8QQ, UK. 710527 A SkylineDrive, Corning, NY 14830, USA. 8British Antarctic Survey, HighCross, Madingley Road, Cambridge CB3 0ET, UK. 9Baltic NestInstitute, Stockholm Resilience Centre, Stockholm University,SE-106 91 Stockholm, Sweden. 10Fisheries Centre, AquaticEcosystems Research Laboratory (AERL), 2202 Main Mall, TheUniversity of British Columbia, Vancouver, BC, Canada V6T 1Z4.11U.S. Geological Survey, Alaska Science Center, 4210 Uni-versity Drive, Anchorage, AK 99508, USA. 12Ecosystem AnalysisSection, Ministry of Fisheries and Marine Resources, LüderitzMarine Research, Post Office Box 394, Lüderitz, Namibia.13Animal Demography Unit, Zoology Department, University ofCape Town, Private Bag X3, Rondebosch, Cape Town 7701,South Africa. 14Marine Research Institute and Zoology Depart-ment, University of Cape Town, Private Bag X3, Rondebosch,Cape Town 7701, South Africa. 15Farallon Institute for AdvancedEcosystem Research, Post Office Box 750756 Petaluma, CA94952, USA.

*To whom correspondence should be addressed. E-mail:philippe.cury@ird.fr (P.M.C.); ilb@st-andrews.ac.uk (I.L.B.)

www.sciencemag.org SCIENCE VOL 334 23 DECEMBER 2011 1703

REPORTS

on

Dec

embe

r 23

, 201

1w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

www.sciencemag.org/cgi/content/full/334/6063/1699/DC1

Supporting Online Material for

From Flat Foot to Fat Foot: Structure, Ontogeny, Function, and Evolution Of Elephant “Sixth Toes”

John R. Hutchinson, Cyrille Delmer, Charlotte E. Miller, Thomas Hildebrandt, Andrew A. Pitsillides, Alan Boyde

Published 23 December 2011, Science 334, 1699 (2011)

DOI: 10.1126/science.1211437

This PDF file includes:

Materials and Methods SOM Text Figs. S1 to S16 Tables S1 to S3 References (31–41) Captions for Movies S1 to S4

Other Supporting Online Material for this manuscript includes the following: (available at www.sciencemag.org/cgi/content/full/334/6063/1699/DC1)

Movies S1 to S4

1

Supporting Online Material For the paper by J. R. Hutchinson, C. Delmer, C. E. Miller, T. Hildebrandt, A. A. Pitsillides and A. Boyde, “From flat foot to fat foot: the structure, ontogeny, function, and evolution of elephant “sixth toes.”

Materials and Methods Gross Anatomy and Dissections: We obtained 63 fore and hind feet of 29 individual Asian (Elephas maximus Linnaeus 1758) and African (Loxodonta africana Blumenbach 1797; i.e., bush/savannah) elephant cadavers, from neonatal to adult (tables S1,S2). These were sourced from EU (mainly UK) zoos and safari parks after natural mortality or euthanasia for reasons unrelated to this study. Pathologies were observed in some specimens, but these did not relate to the factors described here and are the basis of another study. Selected feet were dissected to determine the relationships of the predigits with surrounding tissues, to manually investigate predigit mobility and joints (Movie S2), and to remove the predigits for imaging and histology.

Three-dimensional (3D) Reconstructions: All specimens were scanned using computed tomography (CT), using a GE LightSpeed Ultra 8-detector scanner; typical settings: 2.5-5mm axial slices, x-ray beam intensities of 120 kVp and 100 mA, 512x512 pixel and ~1.067 pixels/mm images, bone algorithm). A laser surface scanner was used for some fossil specimens (see table S3). Bone; including predigit; positions and mineralization patterns then were visually inspected by segmenting the CT DICOM slices in Mimics 15.0 (Materialize, Inc.; Leuven, Belgium) software following previous methods (15), with semi-automatic thresholding (bone default cutoff; 226 Hounsfield units), to create 3D images of each bone element (Movie S1).

Histology and Imaging: Four adult and two juvenile elephant predigits (tables S1,S2) were inspected with the following methods. The results from these analyses consistently matched those from our CT imaging (above). Hence we consider the general qualitative conclusions drawn about predigit structure and ontogeny robust enough to broadly apply to elephants. Slabs selected for SEM were either macerated in 4% alkaline bacterial pronase for 5 days and then 8h in 1% KOH – both at 50C - to remove all non-mineralized tissues (including hyaline cartilage), washed, dried from ethanol and carbon coated; or embedded in PMMA, and the blocks trimmed and polished to 1um diamond and carbon coated. Both types of preparation were imaged using backscattered electrons at 20kV in an SEM and also by 30kV point projection digital microradiography. To standardize the BSE signal from the polished samples, halogenated dimethacrylate standards were employed (27).

Selected segments of the samples (fig. 2 and fig. S13) were decalcified for up to 14 days in 10% EDTA/TRIS-HCl (pH 7.4) at 4° C, depending on degree of mineralization and size of specimen, with the solution refreshed at least twice each week. After decalcification, samples were processed and embedded by standard methods in paraffin. Transverse 8µm sections of individual

2

predigits were cut across the entire element and stained with Toluidine blue (0.1% in 0.1M solution of acetate buffer, pH 5.6) or with 0.1% Safranin O with 0.02% Fast Green to study proteoglycan content.

Functional analysis ex vivo: We loaded two manus and two pes specimens from different individual adult cadavers (sources above) with a custom-made wooden-framed loading jig to apply vertical loads to the centre of the proximal carpals/tarsals at zero (i.e., only foot weight) and near-maximal locomotor loads of approximately one body weight per limb (31). The custom loading jig used a 10-tonne capacity car jack (Ferm HB-10; Ferm B.V., Genemuiden, Netherlands) outfitted with a customized pressure gauge that was calibrated to quantify axial forces along its distal steel loading bar, which was rigidly inserted into drilled holes in the proximal bones.

We then CT-scanned the statically loaded foot once per loading condition, and segmented the images as above to examine the positional changes of foot bones with loading (fig. 3). The predigit, metapodial III and the loading bar were segmented, as above, in Mimics software. 3D vectors were then placed along their central shafts to represent segment axes. Two-dimensional angles of the predigit and metapodial III were measured in the mediolateral and craniocaudal planes, with respect to the orientation of the loading bar, and hence the applied load. Angular changes were calculated between minimum and maximum loading conditions. While quantitative, this simple analysis of limited specimens was merely intended to give generally descriptive indications of the passive loading behavior of elephant predigits, as the first functional analysis of these structures.

Repeated measurements (five repetitions of the entire 3D data analysis procedure) for one manus specimen showed maximal differences of predigit angle changes (loaded v. unloaded) of 2.1-4.2° (from 3D, mediolateral and craniocaudal perspectives). Thus in our results (fig. 3; main text), we conservatively considered any differences between loading conditions of <5° predigit angle to be inconsequential; potentially just measurement error; and simply report those angle changes in our specimens that exceeded this threshold.

Fossils and phylogeny: We surveyed museum specimens (table S3) and literature for all basal proboscidean taxa (especially Numidotherium and Barytherium) continuing into Elephantiformes (including the basal taxa Phiomia, Gomphotherium and Mammut), and integrated these data into a phylogenetic context (fig. 4). Detailed descriptions are in the supporting online text and figures.

3

Supporting Online Text and Figures Dissections Tables S1 and S2 outline the specimens we used. Additional images with descriptions follow below as figs. S1-S4.

A B Figure S1. Right feet of elephants in medial view with skin removed to show the superficial surface of predigit regions. A, Manus (some superficial tendons removed) of specimen #5 (table S1); B, Pes of specimen #10 (table S1). Not to scale. The arrows are just above, and run parallel to, the dorsal border of the predigits, showing the proximodistal axis. See also Movie S2.

A B Figure S2. Right feet of elephants in caudal view showing exposed predigit (still embedded in part of foot pad, especially in B); otherwise as in fig. S1.

4

A B Figure S3. Dissected elephant predigits and first metapodials. A, Left prepollex and metacarpal I (showing associated musculature) of specimen #15 (table S1) in oblique lateral view; B, Right prehallux and metatarsal I of specimen #10 (table S1) in oblique dorsolateral view (proximal articular surface of metatarsal I visible at bottom right). Not to scale. Arrow is just above, and runs parallel to, the dorsal border of the predigit, showing the proximodistal axis.

A B Figure S4. Bandsawn frozen specimen of the left pes of a 38 year old female Asian elephant (not tables S1,S2; these images only), in oblique lateral (A) and cross-sectional views (B). Red arrows denote the prehallux, mainly exhibiting white cartilage but with an ossifying caudal border visible in A (location of arrow; pink tissue). Also note intrinsic musculature around proximal/cranial margin of prehallux in both images.

5

Histology and Imaging

Additional images with descriptions follow below as figs. S5-S13.

Figure S5. X-ray image from the right manus of specimen #15 (see fig. S3A for photograph). The first metacarpal is the bright white object on the bottom right; the fainter object above it is the prepollex, which is largely cartilaginous but shows some small centers of ossification (higher contrast areas).

6

Figure S6. Series of photographs showing sectioning of right prepollex of specimen #2 (table S1) in preparation for histology and imaging (fig. 2). Pink/red ossified regions contrast with white cartilage islands. White numbers show section locations. Scale bars show mm lines.

7

A B Figure S7. Right prepollex, section 0 from fig. S6. A, enlarged photograph; B, 3x magnification Faxitron point projection digital x-ray image of PMMA embedded block of same specimen as prepared for SEM, width of field = 27mm.

8

Figure S8. Right prepollex, section 1 (same specimen as in fig. S7), BSE SEM montage of macerated specimen, width of field = 34mm (cf. fig. 2B; smaller image). Large space in right central area of specimen - which was occupied by cartilage - shows endochondral mineralization front (shown at higher magnification in figs. S9-11).

9

Figure S9. Right prepollex, section 1 (same specimen as in figs. S6-S8), showing magnified view of endochondral mineralization front, width of field = 4.45mm (cf. fig. S8). Center of field shown at higher magnification in fig. S10.

10

Figure S10. Right prepollex, section 1 (same specimen as in figs. S6-S9), showing detailed view of cartilage mineralization front; width of field = 1.78mm (cf. fig. 2C; smaller image).

11

Figure S11. Right prepollex, section 0 (same as in figs. S6,S7; cf. fig. 2D; smaller image), showing BSE image of density variations in block surface (scale: 1=lowest; 8=highest: see fig. S12 legend for more details). Field is 27.95 x 23.39mm. The calcified cartilage (pink and grey) has been partially resorbed on the opposite surface, and served as a scaffold for bone deposition. Bone is seen as red, orange and yellow colors towards the bottom of the image, where the black regions are bone marrow.

12

A B Figure S12. A, Right prepollex, section 0 (same as in figs. S6,S7,S11), BSE image showing density variations. Hyaline cartilage is in the black area at the top of the image and is calcifying along (and in places, ahead of) a mineralization front (cf. fig. 2E; smaller greyscale image). The calcified cartilage (pink and grey) has been partially resorbed on the opposite surface, and served as a scaffold for bone deposition. Bone is seen as red, orange and yellow colors towards the bottom of the image and the black regions here are bone marrow. B, further explanation of the pseudo-color look-up table. The grey signal level represents electron backscattering, which increases with increasing mineral concentration in the calcified tissue. For quantitative BSE imaging, signal levels are calibrated using halogenated dimethacrylate standards (27). The grey level obtained from the monobrominated compound is adjusted to zero and that from the monoiodinated is adjusted to 255. The color look-up table shows zero as black, 1-32 as dark blue, 33-64 lighter blue, 65-96 green, 97-128 yellow, 129-160 orange, 161-192 red, 193-224 pink, 225-254 grey, 255 white.

13

Figure S13. Safranin-O and Fast green stained transverse section of decalcified prepollex from specimen #10 (table S1) showing an almost entirely cartilaginous element (red), and detail showing disorganized chondrocyte orientation but some evidence of hypertrophy and chondrocyte cluster formation near non-cartilaginous tissues (blue-green).

14

Fossil Evidence Known autopodium (foot; manus and pes) fossils from basal proboscideans: Autopodial elements of early proboscideans are extremely rare in the fossil record. The oldest proboscideans, Erytherium, Phosphatherium and Daouitherium are known only from cranio-dental remains (except for one phalanx referred to Phosphatherium) found isolated in marine sediments. Discoveries of foot elements for these taxa, basal tethytherian outgroups, and more material of later taxa mentioned below will be critical for testing the inferences made in this study. Anatomical terminology follows (25,26), with alternative comparative terms in parentheses when first used.

The oldest well documented early proboscidean is Numidotherium koholense, from the late early-middle Eocene, for which numerous postcranial elements together with several skulls were collected in Algeria during the 1980s. This material includes two partial manus, possibly from the same individual, together with scattered pes remains, mainly tali (astragali) and calcanei (calcanea). Similarly, two partial manus of Barytherium were found in Dor El Talha, Libya, during the 1960s, together with a few tali.

A potential talus and calcaneus have been referred to the amphibious Moeritherium: a talus (Yale Peabody Museum specimen YPM 24186); noted by (32); and a calcaneus (YPM 24125); referred to Moeritherium but apparently unpublished. The two elements seem to be from different individuals; if not species; because they do not articulate well (Delmer, pers. obs.; 33). The talus bears a tuberculum mediale quite like that of other early proboscideans, but bears no diagnostic features allowing its confident assignment to Moeritherium.

Thus because no unambiguous remains of autopodia exist for many basal taxa, Numidotherium koholense and Barytherium grave are the only two taxa that give us some indication of the manus and pes morphology of basal proboscideans. The morphology of deinotheres, the only other non-elephantiform proboscideans, is very well understood based on almost complete skeletons from Europe (34-36). Yet their postcranial anatomy is strikingly derived; similar to that in elephantoids (e.g., fig. 4). Most of their postcranial elements, if found isolated, are not easily distinguishable from coeval proboscideans such as Gomphotherium.

The morphology of the manus of early Elephantiformes is fairly well known. It is exemplified by a semi-complete associated manus of Phiomia from the Fayum region of Egypt, but the pes of Phiomia remains still poorly known, essentially from its talus and calcaneus (like Palaeomastodon). However, the remains of all known deinothere and elephantiform manus and pedes are sufficiently similar to draw robust inferences about the ancestral morphology for this clade. Movies S3,S4 show the manus and pes morphology for some of the best preserved specimens of representative taxa along this lineage (see also fig. 4).

15

Figure S14. Right manus of Barytherium grave (table S3), in dorsal view, showing plantigrade orientation and other features noted in the text. Not to scale.

Morphology of the autopodia in basal Proboscidea: The autopodia of Numidotherium and Barytherium, although quite different from those of Elephantiformes, display characteristics that still link them to the seriated (taxeopode) anatomical structure that is typical of Proboscidea and its close relatives, particularly the following four traits. First, The intermedium (lunar/semilunar) does not articulate with carpal IV (unciform/hamate), and carpal III (magnum) does not articulate with the ulnar carpal (trichetrum/cuneiform). The intermedium of Numidotherium koholense has a proximal articulation that is almost solely with the radius (the ulna is almost excluded from this joint), as in other more derived proboscideans. Barytherium, however, has an intermedium with a strong contact for the ulna, quite similar to that in Arsinoitherium, but this is probably an autapomorphy (33). Second, the articulations between the metacarpals (Mc) and distal row of carpals are identical to those in later proboscideans: Mc II articulates mainly with carpals II and III, Mc III with carpals III and IV, and Mc IV and V with carpal IV. Third, the talus articulates only with the centrale (navicular). It also displays a well developed tuberculum mediale, identified as a diagnostic character for Proboscidea (32). Finally, the tuber calcanei also exhibits a strong mediolateral widening that is also found in Elephantiformes.

However, overall, the manus and (fragmentary) pes of Numidotherium and Barytherium (fig. S14; Movies S3,S4) are strikingly different from those of deinotheres and Elephantiformes (Movies S3,S4) in the following seven features. First, Carpal III has a strong caudodistal extension, which creates a strong craniodistal orientation of its articular facet for Mc II and III. In deinotheres and Elephantiformes, that facet is much more parallel to that of the proximal

16

articulation. In association with the related strong caudal extension of the proximal articular surfaces of the metacarpals observed in Barytherium and Numidotherium, such a disposition supports the inference that both taxa have almost plantigrade manus, very dissimilar from the stereotypically vertical (digitigrade/subunguligrade) skeletal orientation of the manus in deinotheres and Elephantiformes.

Second, the metacarpals are much more flattened craniocaudally, and proportionally wider than in deinotheres and Elephantiformes; especially in Barytherium. They therefore seem to have been mediolaterally splayed (abducted) more widely, confirmed by their articulations. In conjunction with the plantigrady of the manus, this could have facilitated swimming as well as locomotion over soft ground such as swamps. It is unclear whether this state is plesiomorphic for Proboscidea as one might suspect. The feet of basal taxa within outgroup clades such as Pezosiren (Sirenia), Ashoroa (Desmostylia), and Anthracobune are either already highly specialized, insufficiently described or unpreserved. This possible amphibious specialization is, however, congruent with recent isotope analysis of the enamel of Barytherium that concluded that its diet was mainly composed of aquatic plants (35). In later Proboscidea, the metacarpals become strongly adducted, forming a near-vertical colonnade (12,15,16,25).

Third, the Mc V of Numidotherium and Barytherium is massive, very asymmetrical and has an almost completely lateral orientation. It bears a strong proximolateral articular facet for the ulnare. Considering this and the massive development of the accessory carpal (pisiform/accessorium; below), we infer that there was a lateral bias for forefoot forces in basal Proboscidea. This condition is reduced in Elephantiformes, with support seemingly becoming more balanced toward the middle of the feet (but see 16).

Fourth, the accessory carpal of Barytherium is the largest carpal of the manus. It displays a very strong contact for the ulnare, but an even larger contact, almost vertically oriented, for the ulna. Considering the almost plantigrade posture of the manus, this intimates that the accessory carpal of Barytherium was the major support for the caudal border of the manus. Thus it likely contributed to limb support in a similar way as the prepollex and foot pad do in deinotheres and Elephantiformes. No accessory carpal is known for Numidotherium, but the size, shape and concavity of the facet for the accessory carpal on the ulnare suggest that its morphology was similar to that of Barytherium. In later Proboscidea, the accessory carpal is comparatively smaller and does not have potential to contact the ground, serving more as a lever/attachment point for carpal and digital flexor muscles.

Fifth, the intermedium in Barytherium and Numidotherium displays a craniomedially positioned, triangular facet interpreted as a contact for a free centrale (38). It however lacks a contact for carpal II (trapezoid), as deinotheres and Elephantiformes have, which is congruent with the hypothesis that strict taxeopody/serial arrangement of the carpals is a synapomorphy of Paenungulata (39). In this case, the contact between the intermedium and carpal II displayed by deinotheres and Elephantiformes must have appeared later in the history of the clade.

Sixth, the main difference between the calcaneus of Barytherium and Numidotherium v. Phiomia, deinotheres and elephantoids is that the ectal and sustentacular facets of the calcaneus are perpendicular to the facet for tarsal IV (cuboid) in the latter taxa, whereas they are slightly cranially inclined in the former taxa (apparently plesiomorphic for tethytheres). The fibular facet is somewhat smaller in Barytherium and Numidotherium.

17

Finally, the central (navicular) facet of the talus is almost perpendicular to the tibial facet in Barytherium. This is less true for Numidotherium, but still more than in Phiomia, deinotheres or elephantoids, in which it is more cranioventrally and mediodistally inclined.

A B Figure S15: Osteological correlates (joint surfaces) for the articulation of the prepollex in fossil proboscideans; metacarpal I in proximal view. A, Phiomia sp. (right side); B, Gomphotherium sp. (right side). Arrows note a palmar ridge on Mc I that is the articular surface. Not to scale. Specimen details in table S3.

18

A B Figure S16: Osteological correlates (joint surfaces) for the articulation of the prehallux in a representative fossil elephantiform; tarsal I (above) and metatarsal I (below) of Gomphotherium sp. in caudal (A) and medial (B) views. Arrows note articular surfaces for the prehallux. Not to scale. Specimen details in table S3.

Deinotheres and Elephantiformes: evidence for the ancestral presence of an expanded foot pad and predigits: Although it is much more gracile than the contemporaneous Barytherium, Phiomia displays a manus with all the typical characteristics of the subunguligrade and taxeopod “elephantine” structure (Movie S3); even its specific articulations (cranial contact between the intermedium and carpal II) that have prompted descriptions of the taxeopody of proboscideans as ‘non-strictly serially arranged’ (40). Although still displaying some plesiomorphic traits (e.g., very flat ulnare with an extended lateral hook), its manus displays the same basic arrangement as the manus of an extant elephant or other elephantiform outgroups (Movie S3), with markedly similar contacts. It also lacks a free centrale. With very few differences (mainly the shape of carpal III; as previously emphasized (41)), carpal III of deinotheres is proportionally thicker than that of elephantoids, especially in its caudal half), the same features are exhibited by the manus of deinotheres.

Most importantly, the first metacarpal of all deinothere/elephantiform taxa inspected has a tuberosity located caudal (palmar) to its proximal articulation, with a shape and orientation

19

similar to that observed in later elephantoids and extant elephants (fig. S15). In Phiomia and deinotheres, this tuberosity, which we have identified as the location where the prepollex articulates with the rest of the manus in extant elephants (15) (see main text), indicates a contact with another element that is caudolaterally oriented (i.e., into the fat pad). That element could only be the radial sesamoid bone; i.e., prepollex. Similar features for the articulation of the prehallux (tibial sesamoid bone) with metatarsal I and tarsal I are evident in the same taxa (fig. S16; Movie S4).

The above observations, together with the subunguligrade disposition of the manus and pes, support the hypothesis that all deinotheres and Elephantiformes possessed predigits and expanded foot pads in their more upright manus and pedes. Our inspections of numerous specimens of Mammuthus and Palaeoloxodon (not shown) as well as our extant samples of Elephas and Loxodonta (tables S1,S2) confirm that this pattern is plesiomorphic for Elephantidae.

Conclusions: We infer that the early proboscideans Barytherium and Numidotherium had an almost completely plantigrade manus and pes, with an enlarged accessory carpal and a plantarflexed calcaneus contributing to the support of the body. This foot orientation most parsimoniously represents the ancestral condition for Proboscidea (fig. 4; Movies S3,S4). It does not leave space for the presence of an enlarged foot pad similar to that observed in extant elephants. However, as in essentially all other tetrapods with autopodia, basal proboscideans should still have had some compliant (but small and superficial) foot pads that were the plesiomorphic predecessors to the giant fatty-fibrous pads of later Proboscidea. These probably extended over the entire plantigrade foot’s palmar/plantar surface including the “heel.”

The enlarged predigits, and therefore the expanded foot pads, should have appeared early in the history of Proboscidea, in the mid-late Eocene (fig. 4). This is because the basal elephantiform Phiomia bears evidence of the enlarged prepollex and a derived manus morphology suggesting the presence of an expanded foot pad. Furthermore, deinotheres have osteological correlates on their Mc I and Mt I (and tarsal I) showing the attachments of the prepollex and prehallux. The more subunguligrade manus and pes postures of all these later Proboscidea (fig. 4; Movies S3,S4) should have evolved in conjunction with this change. Thus foot posture, enlarged sesamoids (predigits) and expanded foot pads seem to have coevolved in Proboscidea.

It is interesting to note that Phiomia was not the largest proboscidean representative in the Eocene, and that the latter, Barytherium, lacks this derived foot posture. Perhaps foot pad evolution was more greatly influenced by the habitat or ecology of proboscideans rather than just increasing body size. Regardless, a gradient toward decreasing amphibious habits and increasing terrestriality extends from basal Proboscidea (and other tethytheres) to Phiomia and deinotheres as well as Elephantoidea. Across this trend, body size changed frequently.

20

Acknowledgements for specimen access We thank Bernard Marandat, Universite Montpellier II, France, for access to manus and pes elements of Numidotherium koholense; William Sanders, University of Michigan, for providing the cast of the calcaneum of Barytherium grave; the Museum d’Histoire Naturelle of Toulouse France, for access to the Prodeinotherium bavaricum pes, and Docteur Eric Jouan, from the Villeneuve St Georges Hospital, France, for providing the CT-scan of the same pes; the Palaeontology department of the Natural History Museum, London for access to specimens and the use of their Konica Minolta laser scanner; Ursula Gohlich, Natural History Museum of Vienna, Austria, for access to the pes of Gomphotherium cf. steinheimensis; Dr Reinhart Ziegler, Staatliches Museum fur Naturkunde, Stuttgart, for access to the manus of Gomphotherium sp.; and Dr Gertrud Rossner, Bayerische Staatssammlung für Paläontologie und Geologie und Geobio-Center der Ludwig-Maximilians-Universität München, Germany, for access to a manus of Phiomia sp.

Specimen Species Element left/right

Body

mass

(kg) Age (yr) Gender MRI dissected

ex vivo

loaded

cleaned

bones

predigits

ossified

CT

predigits

predigit

histology

1 Asian manus L,R

20+

y y n y y y

1 Asian pes L,R

20+

y y n y y y

2 Asian manus R

20+

n y y y y n y

2 Asian pes L

20+

n y n y y n y

3 Asian manus R

~20

y y y y n y

4 Asian manus L,R 3400 26 F y y n y y y

4 Asian pes L,R 3400 26 F y y y y y y n

5 Asian manus L

50 M n y n y y y n

5 Asian pes L

50 M y y y y n y

6 Asian pes R

18

n y n y y y

7 African manus L 2500? 24 F n y n y n y

7 African pes R 2500? 24 F n y n y p y

8 African manus R

28 M y n n n n n n

9 Asian manus L,R 3550 40 F y y n y p y y

9 Asian pes L,R 3550 40 F

y n y p y y

10 Asian manus L,R 3920 35 F y y n y n y y

10 Asian pes L,R 3920 35 F

y n y p y y

11 Asian manus L,R 3690 55 F

y

11 Asian pes L,R 3690 55 F

y

12 Asian manus L,R

35 F n n n n n n n

12 Asian pes R

35 F n n n n p n n

13 African manus R

20+

y

n

13 African pes R

20+

p

14 Asian manus L

40+ F

n

n

15 Asian pes R

50+ F

n

15 Asian manus L

50+ F y y n y p y

Table S1. Adult elephant specimens examined in this study. Body mass with ? indicates an uncertain estimate from incomplete measurements. The columns from “MRI?” to the far

right column indicate, respectively, whether (y/n= yes/no) the specimen’s anatomy was inspected with an MRI scan, had a complete dissection (i.e. disarticulated), was used for ex

vivo loading for functional analysis (11), had its bones defleshed and bleached to check skeletal morphology, determined to have an ossified (p= partially; y= extensively) prepollex

(manus) or prehallux (pes), had its predigit(s) removed and CT scanned at higher resolution to check ossification patterns, or had more detailed histology and imaging techniques

applied to its predigit(s) (11). Blank entries indicate that a procedure or check has not yet been applied to the specimen or was inconclusive, whereas “n” entries for procedures

indicate that the specimen was destructively analyzed before such procedures could be applied.

Specimen Species Element left/right

Body

mass

(kg) Age (yr) Gender MRI dissected

ex vivo

loaded

cleaned

bones

predigits

ossified

CT

predigits

predigit

histology

16 Asian manus L,R 116? 0.1

y y

n

16 Asian pes L,R 116? 0.1

y y

n

17 African pes L 500? 7.5 F n y n y n n n

18 Asian manus L,R 838 2.5 F y y n y n n y

18 Asian pes L,R 838 2.5 F y y n y n n y

19 Asian manus L 110 0.17 F n n n n n n n

19 Asian pes R 110 0.17 F n n n n n n n

20 Asian pes R

2 to 4

y y n y n

21 Asian manus R 116? 0.1

n n n n n n n

22 Asian manus L 452 1.5 M n y n y n n n

22 Asian pes L 452 1.5 M n y n y n n n

23 Asian manus L 833 2 F

n

23 Asian pes L 833 2 F

n

24 Asian manus L 950 2.5 M

n

24 Asian pes L 950 2.5 M

n

25 Asian manus R

3.5

n

26 Asian pes L

12

p

27 Asian manus R,L

0.1

n

27 Asian pes R

0.1

n

28 Asian manus R

0.1

n

29 Asian manus R

0.1

n

Table S2. Juvenile (12 years or younger) elephant specimens examined in this study. Body mass with ? indicates an uncertain estimate from incomplete measurements. The columns

from “MRI?” to the far right column indicate, respectively, whether (y/n= yes/no) the specimen was checked with an MRI scan, had a complete dissection, was used for ex vivo

loading for functional analysis (11), had its bones defleshed and bleached to check skeletal morphology, determined to have an ossified (p= partial; y= extensive) prepollex (manus)

or prehallux (pes), had its predigit(s) removed and CT scanned at higher resolution to check ossification patterns, or had more detailed histology and imaging techniques applied to

its predigit(s) (11). Blank entries indicate that a procedure or check has not yet been applied to the specimen or was inconclusive, whereas “n” entries for procedures indicate that the

specimen was destructively analyzed before such procedures could be applied.

Taxon Collection # Original/cast Element Side Scan Method Equipment Resolution/Lens Scanning Location

Numidotherium koholense UO-K 065 Cast Ulnare R Laser scan Konica Minolta VI-910 digitizer Minolta lens TELE 25mm NHM, London, UK Numidotherium koholense UO-K 066 Cast Intermedium R Laser scan Konica Minolta VI-910 digitizer Minolta lens TELE 25mm NHM, London, UK

Numidotherium koholense UO-K 068 Cast Carpal IV R Laser scan Konica Minolta VI-910 digitizer Minolta lens TELE 25mm NHM, London, UK

Numidotherium koholense UO-K 070 Cast Carpal III L Laser scan Konica Minolta VI-910 digitizer Minolta lens TELE 25mm NHM, London, UK Numidotherium koholense UO-K 076 Cast Carpal II L Laser scan Konica Minolta VI-910 digitizer Minolta lens TELE 25mm NHM, London, UK

Numidotherium koholense UO-K 072 Cast Metacarpal III L Laser scan Konica Minolta VI-910 digitizer Minolta lens TELE 25mm NHM, London, UK

Numidotherium koholense UO-K 075 Cast Metacarpal V R Laser scan Konica Minolta VI-910 digitizer Minolta lens TELE 25mm NHM, London, UK Numidotherium koholense UO-KX 18 Cast Talus L Laser scan Konica Minolta VI-910 digitizer Minolta lens TELE 25mm NHM, London, UK

Numidotherium koholense UO-KA 1-31 Cast Calcaneus R Laser scan Konica Minolta VI-910 digitizer Minolta lens TELE 25mm NHM, London, UK

Barytherium grave NHMUK. M84010 Original Carpal III R Laser scan Polhemus Fastscan NHM, London, UK Barytherium grave NHMUK. M84011 Original Accessorium R Laser scan Polhemus Fastscan NHM, London, UK

Barytherium grave NHMUK. M84012 Original Metacarpal III R Laser scan Polhemus Fastscan NHM, London, UK

Barytherium grave NHMUK. M84013 Original Metacarpal V R Laser scan Polhemus Fastscan NHM, London, UK Barytherium grave NHMUK. M84014 Original Metacarpal IV R Laser scan Polhemus Fastscan NHM, London, UK

Barytherium grave NHMUK. M84015 Original Intermedium R Laser scan Polhemus Fastscan NHM, London, UK

Barytherium grave NHMUK. M84016 Original Talus R Laser scan Polhemus Fastscan NHM, London, UK Barytherium grave UM. 13973 Cast Calcaneus L Laser scan Polhemus Fastscan NHM, London, UK

Phiomia sp. BSP.1905.XIII 502 Original Intermedium R CT scan Marconi Medical Systems 0.742 mm per pixel RVC, Hatfield, UK

Phiomia sp. BSP.1905.XIII 501 Original Ulnare R CT scan Marconi Medical Systems 0.742 mm per pixel RVC, Hatfield, UK Phiomia sp. BSP.1905.XIII 503 Original Radiale R CT scan Marconi Medical Systems 0.742 mm per pixel RVC, Hatfield, UK

Phiomia sp. BSP.1905.XIII 504 Original Carpal IV R CT scan Marconi Medical Systems 0.742 mm per pixel RVC, Hatfield, UK

Phiomia sp. BSP.1905.XIII 505 Original Carpal III R CT scan Marconi Medical Systems 0.742 mm per pixel RVC, Hatfield, UK Phiomia sp. BSP.1905.XIII 506 Original Carpal II R CT scan Marconi Medical Systems 0.742 mm per pixel RVC, Hatfield, UK

Phiomia sp. BSP.1905.XIII.b.3a Original Metacarpal I R CT scan Marconi Medical Systems 0.742 mm per pixel RVC, Hatfield, UK

Phiomia sp. BSP.1905.XIII 507 Original Metacarpal II R CT scan Marconi Medical Systems 0.742 mm per pixel RVC, Hatfield, UK Phiomia sp. BSP.1905.XIII 508 Original Metacarpal III R CT scan Marconi Medical Systems 0.742 mm per pixel RVC, Hatfield, UK

Phiomia sp. BSP.1905.XIII 509 Original Metacarpal IV R CT scan Marconi Medical Systems 0.742 mm per pixel RVC, Hatfield, UK Phiomia sp. NHMUK. M8513 Original Talus R Laser scan Polhemus Fastscan NHM, London, UK

Palaeomastodon beadnelli NHMUK. M8484 Original Calcaneus L Laser scan Polhemus Fastscan NHM, London, UK

Deinotherium sp. NHMUK. M9261 Cast Carpal III R Laser scan Konica Minolta VI Minolta lens TELE 25mm NHM, London, UK Deinotherium sp. NHMUK. M9261 Cast Carpal IV R Laser scan Konica Minolta VI Minolta lens TELE 25mm NHM, London, UK

Deinotherium sp. NHMUK. M9261 Cast Carpal II L Laser scan Konica Minolta VI Minolta lens TELE 25mm NHM, London, UK

Deinotherium sp. NHMUK. M9261 Cast Intermedium L Laser scan Konica Minolta VI Minolta lens TELE 25mm NHM, London, UK Deinotherium sp. NHMUK. M9261 Cast Ulnare L Laser scan Konica Minolta VI Minolta lens TELE 25mm NHM, London, UK

Deinotherium sp. NHMUK. M9261 Cast Carpal I L Laser scan Konica Minolta VI Minolta lens TELE 25mm NHM, London, UK

Deinotherium sp. NHMUK. M9261 Cast Metacarpal I L Laser scan Konica Minolta VI Minolta lens TELE 25mm NHM, London, UK Deinotherium sp. NHMUK. M9261 Cast Metacarpal III R Laser scan Konica Minolta VI Minolta lens TELE 25mm NHM, London, UK

Deinotherium sp. NHMUK. M9261 Cast Metacarpal IV L Laser scan Konica Minolta VI Minolta lens TELE 25mm NHM, London, UK

Prodeinotherium bavaricum MNHT.PAL.2010.0.3 Original Pes L CT scan Philips\Mx8000 0.627 mm per pixel Villeneuve St Georges Hospital, France Prodeinotherium bavaricum uncatalogued Original Metacarpal I CT scan GE LightSpeed 16 0.813 mm per pixel RVC, Hatfield, UK

Gomphotherium sp. SMNS 47443 Original Manus R CT scan Marconi Medical Systems 0.742 mm per pixel RVC, Hatfield, UK

Gomphotherium cf. steinheimensis BSP 1971 I 275 Original Pes L CT scan GE LightSpeed 16 0.813 mm per pixel RVC, Hatfield, UK Mammut americanum PRI 49820 Original Manus L CT scan TOSHIBA\Acquilion\LB 0.175-0.456 mm per pixel Cayuga Medical Center, USA

Mammut americanum PRI 49820 Original Pes L CT scan TOSHIBA\Acquilion\LB 0.112-0.781 mm per pixel Cayuga Medical Center, USA

Table S3. Fossil specimens examined. For further details see supplementary online text: Fossil Evidence. All specimens except Palaeomastodon were used in

Movies S3,S4.

1

1

References and Notes

1. D. D. Davis, The giant panda: A morphological study of evolutionary mechanisms. Fieldiana 3, 1 (1964).

2. H. Endo et al., CT examination of the manipulation system in the giant panda (Ailuropoda melanoleuca). J. Anat. 195, 295 (1999). doi:10.1046/j.1469-7580.1999.19520295.x Medline

3. S. J. Gould, The panda’s peculiar thumb. Nat. Hist. 87, 20 (1978).

4. S. J. Gould, E. S. Vrba, Exaptation: A missing term in the science of form. Paleobiology 8, 4 (1982).

5. M. Fabrezi, A survey of prepollex and prehallux variation in anuran limbs. Zool. J. Linn. Soc. 131, 227 (2001). doi:10.1111/j.1096-3642.2001.tb01316.x

6. M. R. Sánchez-Villagra, P. R. Menke, The mole’s thumb—evolution of the hand skeleton in talpids (Mammalia). Zoology 108, 3 (2005). doi:10.1016/j.zool.2004.07.006 Medline

7. P. Blair, Osteographia elephantina. Philos. Trans. 27, 53 (1710). doi:10.1098/rstl.1710.0008

8. H. Neuville, Sur quelques caracteres anatomiques du pied des Elephants. Arch. Mus. Natl. Hist. Nat. Paris 13, 6e Serie, 111 (1935).

9. K. von Bardeleben, On the bones and muscles of the mammalian hand and foot. Proc. Zool. Soc. 1894, 354 (1894).

10. C. Mitgutsch et al. Circumventing the pentadactyly‘constraint’: The mole’s ‘thumb’. Biol. Lett., 10.1098/rsbl.2011.0494 (2011).

11. Materials and methods are available as supporting material on Science Online.

12. G. E. Weissengruber et al., The structure of the cushions in the feet of African elephants (Loxodonta africana). J. Anat. 209, 781 (2006). doi:10.1111/j.1469-7580.2006.00648.x Medline

13. F. Galis, J. J. M. van Alphen, J. A. J. Metz, Why five fingers? Evolutionary constraints on digit numbers. Trends Ecol. Evol. 16, 637 (2001). doi:10.1016/S0169-5347(01)02289-3

14. J. Prochel, P. Vogel, M. R. Sánchez-Villagra, Hand development and sequence of ossification in the forelimb of the European shrew Crocidura russula (Soricidae) and comparisons across therian mammals. J. Anat. 205, 99 (2004). doi:10.1111/j.0021-8782.2004.00321.x Medline

15. J. R. Hutchinson, C. E. Miller, G. Fritsch, T. Hildebrandt, in Anatomical Imaging: Towards a New Morphology, R. Frey, H. Endo, Eds. (Springer, Berlin, 2009), pp. 23–38.

2

2

16. C. E. Miller, C. Basu, G. Fritsch, T. Hildebrandt, J. R. Hutchinson, Ontogenetic scaling of foot musculoskeletal anatomy in elephants. J. R. Soc. Interface 5, 465 (2008). doi:10.1098/rsif.2007.1220 Medline

17. D. R. Carrier, N. C. Heglund, K. D. Earls, Variable gearing during locomotion in the human musculoskeletal system. Science 265, 651 (1994). doi:10.1126/science.8036513 Medline

18. A. B. Clifford, The evolution of the unguligrade manus in artiodactyls. J. Vertebr. Paleontol. 30, 1827 (2010). doi:10.1080/02724634.2010.521216

19. N. Court, Morphology and functional anatomy of the postcranial skeleton in Arsinoitherium (Mammalia, Embrithopoda). Palaeontogr. Abt. A 226, 125 (1993).

20. D. E. Lieberman et al., Foot strike patterns and collision forces in habitually barefoot versus shod runners. Nature 463, 531 (2010). doi:10.1038/nature08723 Medline

21. F. Michilsens, P. Aerts, R. Van Damme, K. D’Août, Scaling of plantar pressures in mammals. J. Zool. (London) 279, 236 (2009). doi:10.1111/j.1469-7998.2009.00611.x

22. A. A. Biewener, Biomechanics of mammalian terrestrial locomotion. Science 250, 1097 (1990). doi:10.1126/science.2251499 Medline

23. M. N. Scholz, M. F. Bobbert, A. J. van Soest, J. R. Clark, J. van Heerden, Running biomechanics: Shorter heels, better economy. J. Exp. Biol. 211, 3266 (2008). doi:10.1242/jeb.018812 Medline

24. M. F. Bonnan, in Thunder-Lizards: the Sauropodomorph Dinosaurs, K. Carpenter, V. Tidwell, Eds. (Indiana Univ. Press, Bloomington, IN, 2005), pp. 346–380.

25. M. M. Smuts, A. J. Bezuidenhout, Osteology of the thoracic limb of the African elephant (Loxodonta africana). Onderstepoort J. Vet. Res. 60, 1 (1993). Medline

26. M. M. Smuts, A. J. Bezuidenhout, Osteology of the pelvic limb of the African elephant (Loxodonta africana). Onderstepoort J. Vet. Res. 61, 51 (1994). Medline

27. A. Boyde, R. Travers, F. H. Glorieux, S. J. Jones, The mineralization density of iliac crest bone from children with osteogenesis imperfecta. Calcif. Tissue Int. 64, 185 (1999). doi:10.1007/s002239900600 Medline

28. C. Delmer, Reassessment of the generic attribution of Numidotherium savagei and the homologies of lower incisors in proboscideans. Acta Palaeontol. Pol. 54, 561 (2009). doi:10.4202/app.2007.0036

29. E. Gheerbrant, Paleocene emergence of elephant relatives and the rapid radiation of African ungulates. Proc. Natl. Acad. Sci. U.S.A. 106, 10717 (2009). doi:10.1073/pnas.0900251106 Medline

30. E. Gheerbrant, P. Tassy, L'origine et l'evolution des elephants. C. R. Palevol. 8, 281 (2009). doi:10.1016/j.crpv.2008.08.003

3

3

31. L. Ren, C. E. Miller, R. Lair, J. R. Hutchinson, Integration of biomechanical compliance, leverage, and power in elephant limbs. Proc. Natl. Acad. Sci. U.S.A. 107, 7078 (2010). doi:10.1073/pnas.0911396107 Medline

32. P. Tassy, Les principales dichotomies dans l'histoire des Proboscidea (Mammalia): Une approche phylogénétique. Geobios Mem. Spec. 15, 225 (1982). doi:10.1016/S0016-6995(82)80116-2

33. C. Delmer, thesis, Muséum National d’Histoire Naturelle, Paris (2005).

34. K. Huttunen, U. B. Gohlich, A partial skeleton of Prodeinotherium bavaricum (Proboscidea, Mammalia) from the Middle Miocene of Unterzolling (Upper Freshwater Molasse, Germany). Geobios 35, 489 (2002). doi:10.1016/S0016-6995(02)00042-6

35. K. Huttunen, On a Prodeinotherium bavaricum (Proboscidea, Mammalia) skeleton from Franzensbad, Czech Republic. Annal. Naturhist. Mus. Wien 105A, 333 (2004).

36. D. Kovachev, I. Nikolov, Deinotherium thraceiensis sp. nov. from the Miocene near Ezerovo, Plovdiv District. Geologica Balc. 35, 3; 5 (2006).

37. A. G. S. C. Liu, E. R. Seiffert, E. L. Simons, Stable isotope evidence for an amphibious phase in early proboscidean evolution. Proc. Natl. Acad. Sci. U.S.A. 105, 5786 (2008). doi:10.1073/pnas.0800884105 Medline

38. N. Court, Limb posture and gait in Numidotherium koholense, a primitive proboscidean from the Eocene of Algeria. Zool. J. Linn. Soc. 111, 297 (1994). doi:10.1111/j.1096-3642.1994.tb01487.x

39. M. S. Fischer, Die stellung der schliefer (Hyracoidea) im phylogenetishen system der Eutheria. Zugleich ein beiträg zur anpassungsgeschichte der Procaviidae. Courier Forsch. Senckenberg 84, 1 (1986).

40. M. S. Fischer, P. Tassy, in Mammal Phylogeny: Placentals, F. S. Szalay, M. J. Novacek, M. C. McKenna, Eds. (Springer, New York, 1993), pp. 217–234.

41. P. Tassy, Nouveaux Elephantoidea (Mammalia) dans le Miocène du Kenya–Essai de Réévaluation Systématique (Éditions du Centre National de la Recherche Scientifique, Paris, 1994).

Movie S1. Foot anatomy of an extant elephant (Elephas maximus; specimen #4 in table S1). Manus, then pes anatomy are shown, with bones labeled using terms from (25, 26)

(alternative comparative terms in parentheses). Movie S2. Prehallux mobility in a cadaveric hindfoot. Right pes of specimen #5 in table S1 shown in medial view. Outlines at beginning show rough positions of some anatomical structures (see movie S1); labels: c = centrale; ca = calcaneus, D2 = distal end (ungual/nail) of digit II, mt1,2,3 = metatarsals I-III, ph = prehallux, t1 = tarsal I. Some tissue around the proximal prehallux has been cut and part of the fat pad has been removed around the distal prehallux so overall mobility is exaggerated, but the video is representative of potential caudodorsal mobility of the predigits under loading (fig. 2). Movie S3. Evolution of manus anatomy in Proboscidea. Rotating movies of 3D forefoot anatomy in representative taxa along the line to extant elephants (fig. 4); introductory image explains the format. Still images in caudal and cranial views show some major structures as per movie S1 (MC = metacarpals). Specimens and scanning modalities are listed in table S3. Movie S4. Evolution of pes anatomy in Proboscidea. Rotating movies of 3D hindfoot anatomy in representative taxa along the line to extant elephants (fig. 4); introductory image explains the format. Still images in caudal and cranial views show some major structures as per movie S1 (MT = metatarsals). Specimens and scanning modalities are listed in table S3.