http://cro.sagepub.com/Critical Reviews in Oral Biology & Medicine

http://cro.sagepub.com/content/9/4/369The online version of this article can be found at:

 DOI: 10.1177/10454411980090040101

1998 9: 369CROBMK.M. Weiss, D.W. Stock and Z. Zhao

Dynamic Interactions and the Evolutionary Genetics of Dental Patterning  

Published by:

http://www.sagepublications.com

On behalf of: 

International and American Associations for Dental Research

can be found at:Critical Reviews in Oral Biology & MedicineAdditional services and information for    

  http://cro.sagepub.com/cgi/alertsEmail Alerts:

 

http://cro.sagepub.com/subscriptionsSubscriptions:  

http://www.sagepub.com/journalsReprints.navReprints:  

http://www.sagepub.com/journalsPermissions.navPermissions:  

What is This? 

- Jan 1, 1998Version of Record >>

at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

DYNAMIC INTERACTIONS AND THEEVOLUTIONARY GENETICS OF DENTAL PATTERNINGK.M. Weiss*D.W. StockZ. Zhao

Deportment of Anthropology, and Graduate Program in (ell and Developmental Biology, Penn State University, 409 (arpenter, University Park, Pennsylvania 16802;*To whom correspondence should be addressed

ABSTRACT: The mammalian dentition is a segmental, or periodically arranged, organ system whose components are arrayedin specific number and in regionally differentiated locations along the linear axes of the jaws. This arrangement evolved fromsimpler dentitions comprised of many single-cusp teeth of relatively indeterminate number. The different types of mammalianteeth have subsequently evolved as largely independent units. The experimentally documented developmental autonomy ofdental primordia shows that the basic dental pattern is established early in embryogenesis. An understanding of how geneticpatterning processes may work must be consistent with the different modes of development, and partially independent evo-lution, of the upper and lower dentition in mammals. The periodic nature of the location, number, and morphological struc-ture of teeth suggests that processes involving the quantitative interaction of diffusible signaling factors may be involved.Several extracellular signaling molecules and their interactions have been identified that may be responsible for locating teethalong the jaws and for the formation of the incisor field. Similarly, the wavelike expression of signaling factors within devel-oping teeth suggests that dynamic interactions among those factors may be responsible for crown patterns. These factors seemto be similar among different tooth types, but the extent to which crown differences can be explained strictly in terms of vari-ation in the parameters of interactions among the same genes, as opposed to tooth-type-specific combinatorial codes of geneexpression, is not yet known. There is evidence that combinatorial expression of intracellular transcription factors, includinghomeobox gene families, may establish domains within the jaws in which different tooth types are able to develop. An evolu-tionary perspective can be important for our understanding of dental patterning and the designing of appropriate experimen-tal approaches, but dental patterns also raise basic unresolved questions about the nature of the evolutionary assumptionsmade in developmental genetics.

Key words. Genes, dental patterning, segmentation, homeosis, evolution, homeobox gene,; morphogens, Dlx genes, growthfactors, positional specification, axial patterning, complex systems, genome organization, multigene families.

Abbreviations. NC, neural crest; EMI, epithelial-mesenchymal interaction; PD, proximo-distal; AP, anterior-posterior; DV,dorso-ventral; OC, organizing center; mes-met, mesencephalon-metencephalon boundary; AER, apical ectodermal ridge; ZPA,zone of polarizing activity; PZ, progress zone; IEE, inner enamel epithelium; OEE, outer enamel epithelium; EK, enamel knot;BTP, Bateson-Turing processes; Fgf, fibroblast growth factor; Bmp, bone morphogenic protein. For simplification in this paper,specific gene names are denoted in italics (e.g., "Dlxl"), making no typographic distinction by species or between genes andtheir products. Gene family names are in standard font (e.g., "Dlx genes").

Introduction: When Teeth were King

"The older works and ideas of Cuvier, Owen, Huxley and othersare of comparatively little service now, for they treat the teeth of eachorder of mammals as of so many distinct types, whereas they

must now be treated as modifications of one type."(H.F Osborn, 1897)

T he dentition was long among the most importanttraits in vertebrate biology. Leading figures in the for-

mation of modern biology in Europe devoted consider-able attention to the subject-Georges Cuvier (and his

brother Francis) and Richard Owen, for example. Of par-ticular interest was the question of dental patterning, bywhich we refer to the location, number, and morphologyof teeth in the dentition. Fig. 1 illustrates several aspectsof dental patterning that have long been of interest inresearch on the dentition and its evolution. We ask thereader to keep this Fig. in mind as a conceptual aid forthe points we wish to make, especially about the seg-mental, or periodic, nature of dental patterning.

Dental patterning was important in the constructionof vertebrate taxonomy. Teeth reflected function (afterDarwin, this would also be equated with adaptation) and

9(4) 369-398 (1998)Crit Rev Oral Biol Med 369369Crit Rev Oral Biol Med9(4):369-398 (1998) at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

:4>J. ,,1

*1*

Procyoi& Ailunus Urma Aiiuropo&

A

B

Figure 1. Collage representing various aspects of dental patterning. (1, top left) The Mesozoic pantothere Amphitherium, an early mammalshowing primitive heterodonty; (2, center left) size gradients within tooth regions are another aspect of heterodonty, as shown by a series ofmolar-premolar gradients in the upper cheek teeth of several carnivores [left to right, raccoon (Procyon), red panda (Ailurus), cave bear(Ursus), giant panda (Ailuropoda)]; (3, bottom left) there can be secondary loss of heterodonty in mammals, as shown by the sealLeptonychotes weddelli (The three Figs. on the right column illustrate the evolution of the periodicity in the patterning of complex molariformteeth in proboscidians.); (4, top right) Moeritherium from Egyptian Tertiary (upper cheek teeth shown above lower); (5, center right) a morerecent species, Paleomastodon, from Egyptian Oligocene (A, upper; B, lower) showing evolution toward (6, bottom right) the crown surfaceof modern elephant molars (A, Indian; B, African). Assembled from illustrations in Peyer (1 968) from various earlier 20th century sources.

370 Crit Rev Oral Biol Med 9(4):369-398 (1998)

.1 3 IL.0. Ii:_-

I' i

.4 f toI

at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

were among the major structures found in fossils. Of par-ticular interest were aspects of the skeleton that reflectedserial homology among their constituent parts. This termreferred to the repetition of similar parts, which in somecases showed regional variation in their details-like thevertebrae and teeth. After Darwin convinced biologiststhat similarities among species could be explained in his-torical terms-that is, homology is equated to commonancestry-teeth were among the more important struc-tures used for the construction and attempted under-standing of vertebrate and, in particular, mammalian phy-logeny. Darwin hypothesized that repetition of the traitcame first, forming the working material for subsequentregionalization. He and others of his time worked on atype of idealized morphology: Existing diversity is a varia-tion on a common theme that Darwin assumed once exist-ed as a common ancestral form. As indicated by the lead-in quote from the pioneer in the establishment of dentalhomologies, the diversity of mammals was held to reflectvariation on common components, shared since thatancestry, and genetics adopted the same assumptionswith regard to underlying genetic mechanisms. This isimportant in a practical sense; for example, the value ofthe mouse as a model system for human oral biologylargely rests on these assumptions.

Darwin believed in the gradual evolution of form, sothat traits shared by two species would be shared bytheir ancestors back to a common root. Other leadingbiologists of the late 19th century-like William Bateson,one of the founders of genetics-believed that traits likethe dentition had to evolve qualitatively (Bateson, 1894).Many aspects of the dentition are qualitative: a certainnumber of teeth, in discrete, separated sites, with inter-nally discrete traits like cusps and roots. Bateson arguedthat there need not be, perhaps cannot be, the sharing ofsuch traits back to a common ancestry among qualita-tively varying dental patterns today, or strict homologybetween specific teeth or cusps in such circumstances.Discrete variation cannot go back continuously in time.Nonetheless, Darwin's view came to dominate biologicalthinking (as it still does). That, as we will argue, may havebeen misleading.

Darwin himself (in Origin of Species) thought theproblem of the evolution of serial homology to be"almost beyond investigation". But recently, molecularand developmental methods have revealed basic mecha-nisms of tissue interaction in odontogenesis (the pro-duction of the basic tissues and structures within teeth),mechanisms that seem very similar to those used byother tissues. Teeth develop through a process known asepithelial-mesenchymal interaction (EMI), in which cellsof neural crest (NC) origin respond to signals from theoverlying epithelium at an appropriate time to triggerdental morphogenesis. Teeth are a powerful experimen-tal system by which to study the general EMI process,

regardless of the structure that results. Similar genes andperhaps developmental processes seem to be involved inearly developmental stages of the limb and facial process-es [see Section (V.l.i)l. In the case of teeth, an enamellayer derived from the epithelium develops over a mes-enchymally derived layer of dentin supplied with innerva-tion and vascularization through the pulp cavity, anchoredto the surrounding bone. [For reviews of the genetics ofodontogenesis, see Thesleff and Nieminen, 1996; Slavkinand Diekwisch, 1996; Maas and Bei, 1997; Thesleff andSharpe, 1997; and the online database (Toothbase, 1997).1

Once a tooth develops, all teeth seem to be made ofthe same material, produced by a generic process ofodontogenesis. But this similar dental material is dis-tributed in a nested, periodic way along the jaws: amongteeth separated by spaces, and in the wavelike distribu-tion of cusp and roots within teeth. What is the relation-ship between the genetic determining mechanisms andthis pattern? How much of the pattern is the result of aprocess rather than a genetic specification of individualstructures within it? What determines where and howmany teeth will be initiated along an apparently uniformjaw epithelium? This paper concerns what those pattern-ing mechanisms might be.

(I) Teeth, First and ForemostThe basic dental histology was present in the earliestknown craniate fossils over a half billion years ago(Smith and Hall, 1993; Butler, 1995; Smith et al., 1996;Huyssene and Sire, 1998; Smith and Coates, 1998).Diagnostic features-such as tubules for odontoblastprocesses in dentin, and a pulp cavity-are clearly foundin fossils of that age, in small basic structural unitsknown as odontodes or denticles. Sub-ectodermal NCmigration and EMI seem very likely to have been used (ifnot invented) to generate these structures. An exoskele-ton of denticle scales was characteristic of most earlychordates, and a classic story of evolutionary oppor-tunism had it that anterior scales were recruited to formteeth as part of the evolution of jawed vertebrates.However, the conodonts, among the earliest chordatelineages, possessed anterior denticles apparently usedfor rasping feeding (Janvier, 1995; Smith et al., 1996).Thus, teeth may be foremost in time as well as in the ani-mal. In any case, because of the similarity in the histol-ogy of scales and teeth in present-day chordates, thesestructures may also share genetic mechanism(s)-atleast, that is the Darwinian assumption.

(11) Heterodonty: "By the Nature ofthe Cusps of a Single Molar Tooth"

For a long time after the evolution of jaws, the dentitionwas homodont: composed of numerous rather similar, and

9(4~369981998 Crt Re Ora Bil Me 37371Crit Rev Oral Biol Med9(4):369-398 (1998)

at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

continually replaced, teeth. The number was highly vari-able, and the dental lamina was the locus of stem cellscapable of periodically producing an entire tooth ininteraction with underlying NC-derived, odontogenicmesenchyme (e.g., Peyer, 1968; Luckett, 1993). Teeth werenot necessarily limited to a single row along the jaw mar-gin as in present-day mammals; numerous bones of theoral cavity and pharynx were capable of bearing teeth.

A characteristic of the evolution of mammals hasbeen the addition of heterodonty, or regional structural dif-ferentiation among teeth. Homodont species may exhib-it teeth with some differences in size, shape, or orienta-tion of curvature, but these are generally graded andquantitative; shape differences among mammalian toothclasses are more qualitative. The definition of the toothclasses is not necessarily on the basis of shape, as dis-cussed in the next section, but in many mammalianspecies, heterodonty involves spatulate incisors distal toa conical canine, followed by the box-like premolars andmolars.

Heterodonty was already present in some mammal-like reptiles (Peyer, 1968; Osborn, 1973). Osborn (1993)suggested that in some modern homodont reptiles(crocodiles and alligators), teeth develop from three sep-arate populations of NC cells. Such populations wereproposed to be reflected in wavelike regions of size dif-ference. If the existence of separate populations of odon-togenic cells in homodont species is correct, jaw region-alization may have antedated and provided the basis forthe evolution of mammalian heterodonty. An early evo-lution of jaw regionalization may be the basis for theindependent appearance of heterodonty in several othervertebrate lineages (Peyer, 1968; Butler, 1995). Other axi-ally regionalized, segmented systems (such as the verte-bral column, limbs, and gut) have undergone a regional-izing history very similar to the transition from homo-donty to heterodonty (Weiss, 1990, 1993).

A major feature of heterodonty is the pattern ofcusps and ridges that characterize the crowns of teeth,especially the molars and premolars (e.g., Osborn, 1897;Hershkovitz, 1971; Hillson, 1986; Butler, 1995). Thesecrown patterns are so distinctive that, as one paleontol-ogist has recently expressed it, "most mammalianspecies can be distinguished by the nature of the cuspsof a single molar tooth" (Carroll, 1988). But is a cusp aseparate structural, genetic, or phylogenetic entity? Cancusp homology be identified between species? What isthe relationship among incisors, canines, and molars?We will see that these are interesting questions, not easyto answer.

(11. 1) DIFFERENCES AMONG TEETHWITHIN MAMMALIAN DENTITIONS

Each tooth in the dentition of a particular mammal dif-fers in shape from every other tooth (Osborn, 1993). The

universally acknowledged designation of four toothclasses (incisors, canines, premolars, and molars, withthe latter two classes sometimes jointly termed 'cheekteeth') implies that some of the differences are more fun-damental than others. In many species, the differencesbetween classes correspond to the typical shape cate-gories mentioned above. However, some mammals havehighly modified dentitions, and in these it can be diffi-cult to decide to which class individual teeth belong.Three basic criteria have been proposed for assigningteeth to classes (Luckett, 1993): (1) location relative tobony sutures coupled with replacement pattern, (2)shape (relative either to idealized morphology or toneighboring teeth), and (3) order of appearance in devel-opment [see Section (11.2)1.

Classification of teeth based on ontogenetic criteriagenerally results in assignments similar to those pro-duced by the positional method, but that may differ fromthose obtained by shape-based criteria (Luckett, 1993).These differences in interpretation imply differentprocesses of evolution of the mammalian dentition andpossibly different developmental genetic control mecha-nisms. For example, changes identified as homeotic (atooth resembling teeth from a different position in thejaw) based on developmental criteria would be interpret-ed as changes in the number of teeth within a class byshape-based models. Homeotic change has been usedas evidence that tooth shape is influenced by position inthe jaw and is at least partly independent of tooth initia-tion (Butler, 1995); re-interpretation of these evolution-ary transformations as changes in the numbers of mem-bers of tooth classes does not necessarily support suchan interpretation.

While shape differences between members of differ-ent tooth classes may be qualitative, those betweenmembers of the same class are typically more quantita-tive, as can be seen in Fig. 1. Different features (specificcups, lobes of the crown outline, and roots) reach theirmaximum development on different teeth within a class(Butler, 1939, 1995). The degrees of development of thesefeatures form gradients such that teeth tend to resembletheir immediate neighbors and are progressively moredifferent from more distant tooth class members. Theseshape gradients can be somewhat independent of gradi-ents in tooth size. The contrast between shape differ-ences among tooth classes and those within classesforms the basis of theories for the control of heterodon-ty [see Section (111. 1) 1.

(11.2) VARIATION IN DENTAL PATTERNSAMONG MAMMALS

Features of evolutionary changes in mammalian denti-tions suggest a number of basic rules that may implysomething about the underlying developmental controlmechanisms. Common modifications include changes in

372 Grit Rev Oral Biol Med (1998)372 Crit Rev Oral Biol Med 9(4).-369-398 (1998) at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

the number and location of cusps (Bateson, 1894; Peyer,1968; Carroll, 1988; Miles and Grigson, 1990). Based onthe Darwinian assumption of historical continuity,homologous cusps have been identified among diversetaxa, and under this assumption, the order in whichcusps appear during development often generally paral-lels the order in which the 'same' cusps arose in evolu-tion and is the reverse of the order in which they are lost(Hershkovitz, 197 1; Butler, 1995). Many features areprone to parallelism, however; for example, a fourthupper molar cusp known as 'the' hypocone has appearedindependently a great many times during evolution(Hunter and lernvall, 1995; Jernvall, 1995). A develop-mental rather than a historical-Darwinian view maychange the interpretation and the genetic inferencemade, as will be discussed in Sections (III) and (VII).

The most likely ancestral dental formula (number ofteeth of each type in each jaw quadrant) of placentalmammals is three incisors, one canine, four premolars,and three molars (Carroll, 1988; Luckett, 1993). With afew exceptions, change in tooth number during the evo-lution of placental mammals is the result of tooth lossrelative to this presumed ancestral formula. This losstends to occur in the reverse of the order in which teethare initiated during development.

In each jaw quadrant of most species, three toothbuds are formed almost simultaneously along a continu-ous dental lamina (Fig. 2; Luckett, 1993). These give riseto the most mesial incisor, the canine, and the most dis-tal (or second most distal) premolar. Additional budsthen appear from mesial to distal in the incisor regionand from distal to mesial in the premolar region. Distalgrowth of the lamina is followed by the progressivemesial to distal appearance of molars (and the most dis-tal premolar if not one of the initial buds). Secondaryteeth develop when a successional lamina forms as anextension of the epithelial component of a primarytooth. The secondary teeth first appear at the tips ofthese laminae prior to the completion of morphogenesisof the primary teeth. Loss of the last members of a toothclass to be formed during development also character-izes many natural anomalies within humans and otherspecies (Bateson, 1894; Miles and Grigson, 1990) andsuggests that a quantitative rather than a trait-specificgenetic mechanism may be responsible for specifyingtooth number. In other words, a generative process mayfall below a threshold during development. Loss of struc-tures in the reverse of the order in which they appear alsocharacterized the evolution of tetrapod limbs (Shubin etal., 1997) and the tail region of the vertebral column (e.g.,as in humans). This has been suggested to be the resultof a reduction in the number of precursor cells for par-ticular skeletal elements.

More dramatic changes in tooth number haveoccurred during evolution (Peyer, 1968); some lineages

dll dC dP4

dll dl2 d13 dC dPi dP2 dP3 dP4 Ml

Figure 2. Order of tooth appearance in mammals. Dental lami-nae with epithelial thickenings and epithelial components oftooth germs. Upper lamina shows near-simultaneous appear-ance of first incisor, canine, and last premolar buds. In lowerportion of Fig., order of initiation of subsequent tooth germs isshown by decrease in size with later initiation. Dashed line indi-cates posteriad (lateral) growth to produce bud of first molar.Additional molar buds form with further posteriad growth of thelamina (not shown). Fig. indicates only anterior (medial) to pos-terior appearance of incisors, posterior to anterior appearanceof premolars, and anterior to posterior appearance of molars,and the time of appearance. Abbreviations: dl, deciduousincisor; dC, deciduous canine; dP, deciduous premolar; M,molar. All teeth are part of the "primary" dentition. Deciduousteeth, but not molars, are replaced by "secondary" teeth. AfterLuckett, 1 993.

have lost complete tooth classes, such as rodents, whichlack premolars and canines. Based implicitly on a classicDarwinian interpretation, it has been suggested thattransient epithelial thickenings present during mousedevelopment represent vestiges of teeth present inancestral forms (Peterkova et al., 1996; Ruch et al., 1997).As an exception to the general rule of evolutionary toothloss, some toothed whales have greatly increased toothnumbers of essentially homodont morphology, clearlynot consistent with classic Darwinian continuity.Evolutionary modifications in tooth number have alsoaffected tooth succession. Some forms have becometoothless (e.g., baleen whales and pangolins) by the fail-ure of eruption of vestigial primary teeth coupled withthe complete suppression of the secondary dentition.The adult dentition of rodents is believed to consistentirely of primary teeth (Luckett, 1993), while in bats,primary and secondary teeth can erupt at about the sametime, leading to the presence of both in the adult ante-molar dentition.

Evolutionary modifications have occurred in toothstructure (Peyer, 1968). The incisors of rodents growthroughout life and have lost enamel from their inneredges. Rabbits have ever-growing molars and incisors,both of which possess enamel surrounding all surfaces(Hillson, 1986). Enamel is completely lacking in the teethof armadillos and aardvaarks.

The teeth within a class typically evolve as a unit,such that new features arising in a group often affect

9(4) 369 398 (1998) Grit Rev Oral Biol Med373373Crit Rev Oral Biol Med9(4) 369-398 (1998)

at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

more than a single tooth (Peyer, 1968; Carroll, 1988; Butler,1995; lernvall, 1995). Correlated evolution of cheek teethmay also be reflected by changes in the shape gradient,which can take two general forms (Butler, 1939, 1995). Theshape gradient may change its slope so that the differ-ences between adjacent teeth are less in one species thanin another. Homeotic changes may occur, in which thetooth at one position comes to resemble a tooth at a dif-ferent (but nearby) position in another species. Suchchanges may be thought of as moving the shape gradientrelative to the tooth positions.

Although teeth in different classes are more inde-pendent in evolution than are members of the sameclass, homeotic changes may also occur across toothclasses. Examples include molarization of the premolars(e.g., horses), which involves shifting of the already-exist-ing molar cusp pattern anteriorly, premolarization of themolars in seals and bats, incisiform canines (cows andother artiodactyls), and the incisiform canine and canini-form premolar of the mole Talpa europaea (Peyer, 1968;Butler, 1978; Luckett, 1993). In the case of the mole, thehomeotic changes have occurred in the lower but not theupper jaw, suggesting some degree of independence ofthe two portions of the dentition. An extreme example ofhomeotic change is the secondary homodonty oftoothed whales; determining what type of tooth remainsin this group is problematic.

Homeotic variation characterizes other segmentedorgan systems, like the vertebral column (Bateson, 1894;Weiss, 1993; Weiss et al., 1994). As in the evolutionaryhomeotic transformations of teeth described above andthose occurring as anomalies within species (Miles andGrigson, 1990), homeotic changes in the vertebral col-umn occur between neighboring identities (e.g., cervicaland thoracic vertebrae). This pattern may explain theabsence to date of homeotic changes in mice with inac-tivation of genes proposed to play a role in tooth typespecification (see below); because mice lack premolarsand canines, the only obvious transformations would bebetween incisor and molar morphologies. But sincethese teeth occupy the termini of the dentition of mostmammals, it is possible that they are too distant toexchange form. There is at least one possible example ofsuch a change. Kantaputra and Gorlin (1992) reportedmolar-shaped teeth in the incisor region of a human fam-ily; however, the teeth were imperfect molars and mustbe interpreted with caution.

Most naturally occurring human dental formulaanomalies are typified by a variable number of teeth inone or both jaws, but other members of the same classare usually present somewhere in the dentition. Similareffects characterize anomalies in several experimentalgenetic interventions in mice. This will be the subject ofSection (V) below and has been reviewed recently (Maasand Bei, 1997).

(111) Dynamic Interactions: Dental Patterningand Self-organizing Processes

Variations in dental patterns have the potential to guideideas about possible controlling mechanisms.Segmented, periodic, or serially repeated biological sys-tems-like the teeth shown in Fig. 1-must be estab-lished by processes capable of producing quantitative,wavelike patterns. The siting (location) of teeth and thedetermination of tooth form, both of which occur beforethe processes of histogenesis and mineralization, seemto be generally similar among teeth. The most likelymechanisms for primary patterning involve dynamic inter-action among factors that lead to differential cell growthand/or its inhibition. The challenge is to identify thenature of such interactions.

(111.1) CONCEPTS OF DYNAMIC MECHANISMS

In 1830, the French physiologist Felix Savart adaptedideas of Ernst Chladni to evaluate the musical quality ofthe top and back plates of violins. If a source of vibrationis applied to plates on which a fine powder has beenplaced, wavelike mechanical interference patternsappear in the powder that reflect the acoustical qualitiesof the plates. In such physical fields, the sound generatesthe interference pattern, but there is not an independentcause for each wave in the pattern. For example, achange in the vibration frequency does not change theprocess or the plate, but can bring about discretechanges in the interference pattern, reflecting differentlatent response characteristics of the plates.

A series of highly publicized and vitriolic debates,attended by the Parisian glitterati, occurred in the winterevenings of that same year, 1830, at the Academie desSciences in Paris. Etienne Geoffroy St Hillaire defendedhis idea that the morphological diversity of animals-such as the repetition of skeletal elements-representedvariations on a unitary underlying form (the archetype)or developmental process. Georges Cuvier vigorouslyopposed this idea with his view that each structure var-ied independently, its adult form following the demandsof its function. Savart may well have attended thesedebates.

The later 19th century was a time of discovery andunderstanding of electromagnetic fields, which can gen-erate similar interference patterns. Several investigatorstransferred these mechanistic concepts to biological pat-terns. Of course, after Darwin, explanations in biologyhad also shifted, to an evolutionary framework. Bateson(1894)-working when evolution had become the formalmatrix of biology-noted a resemblance between pat-terns found in serially homologous traits, like the skele-ton and dentition, and the wavelike mechanical Chladnifigures (Waller, 1961)-ideas he extended to includeaspects of regional differentiation (Bateson, 1913).Bateson likened discrete evolutionary changes in seg-

374 Crit Rev Oral Biol Med (1998)374 Crit Rev Oral Biol Med 9(4):369-398 (1998) at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

mented traits to changes in Chladni figures. He contrast-ed this type of change with the gradual change of indi-vidual traits invoked by contemporary Darwinian think-ing, arguing that such gradualism could not explain dis-cretely varying traits and did not explain species evolu-tion. Patterned variation, he argued, may arise "by a sin-gle event" (Bateson, 1913). If this is what occurred inteeth, should we be looking for mutations affecting theproduction or protein kinetics of regulatory molecules,rather than for genes or combinations of genes specificto individual traits?

By the 1930s, almost exactly a century after theGeoffroy-Cuvier debates, experimental embryologistsborrowed from physics to suggest that fields within tis-sues may explain the development of patterned traits(Gilbert et al., 1996). Gradients within a field would speci-fy position within a tissue, information that otherwise-identical cells could translate into position-specificstructure by activating appropriate genes. A half-centuryafter Bateson's mechanical speculations, and unaware ofthat work, Alan Turing (1952) suggested that quantitativewavelike patterns could be produced by chemical inter-actions over space if substances, called morphogens, inter-acted in reaction diffusion processes to produce spatial vari-ation (see also Murray, 1993; Meinhardt, 1996; Sager,1996). If diffusible activator(s) move from a source,through a tissue, and induce or interact with diffusinginhibitor substance(s), for example, the relative rates ofproduction, autocatalysis, degradation, diffusion, induc-tion, and inhibition of these substances lead to dynamicwavelike patterns. If there were a cellular response to theconcentration of one of these substances, patternedgrowth, mineralization, pigment production, and the like,as observed in natural species, could result. LewisWolpert applied Turing's mechanism to the patterning ofthe limb by diffusible morphogens (Wolpert, 1969).

Morphogenetic field models can be applied to thepatterning of the dentition. For example, Butler (1939;van Valen, 1970) used the field concept to argue that gra-dients of tooth shape within a class could result from ini-tially identical primordia responding to morphogen gra-dients along the jaw, while the different tooth classescould be the result of separate morphogens.Alternatively, the different types of teeth could be pro-duced by the progeny of three or four different sets, orclones, of cells. Osborn (1978) suggested such a model,and a means by which the gradual exhaustion of thepotential of these clones to divide might lead to the sizeand shape gradients within each tooth class.

Reaction diffusion processes could provide a meansof establishing either the different sets of cells or differ-ent morphogens responsible for each tooth type.Differences between field and clone models illustratepossible relationships between the signals for tooth ini-tiation and those for tooth type selection (ignoring

Pre-initiation (A

Initiation

Pre-initiation C

B I

Initiation (3)

Pre-initiation (1

C Initiation ()

iPost-initiation(i

1~

010I0I

.1.Time

AO Time

10I

01031

Time

1

Figure 3. Possible relationships between signals for tooth initia-tion and tooth type selection. The three circles in each row indi-cate cells that give rise to incisors (i), canines (c), and molars (m)from left (mesial) to right (distal). No attempt is made to indicatemultiple members of tooth classes, the number of cells, or inwhich tissue the signaling molecules are located. S and I refer tosignals for tooth type selection and tooth initiation, respectively.Subscripts indicate signals specific to each tooth type. (A) Cellsin different regions along the proximal-distal axis of the jaw aredetermined to form different types of teeth prior to initiation,with the initiation signal being the same among tooth types. (B)Cells along the jaw axis are equivalent prior to receiving threedifferent types of signals, each of which initiates tooth develop-ment and selects for a particular tooth type. (C) Tooth primordiaare equivalent even after tooth initiation, with tooth type selec-tion occurring in later stages of development as a result of dif-ferent signals in different regions in the jaw.

shape differences within each tooth type for themoment). One possibility (Fig. 3A) is that moleculesresponsible for tooth type selection might already be dif-ferentially expressed among cells that have the potentialto form teeth, prior to any signals for tooth initiation.Such selector molecules would represent a pre-pattern,and the initiation signals could be identical in all tooth-forming regions. Spatially differing combinatorial codes ofgenes such as homeobox transcription factors might bethe selectors, for example.

As illustrated in Fig. 3B, another possible relation-ship between signals for tooth type selection and initia-tion is that three different signals are responsible for ini-

3759(4) 369-398 (1998)Oral Biol Med

Crit Rev Oral Biol Med9(4):369-398 (1998) at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

tiation in the different tooth classes, so that moleculardifferences among tooth types appear at, but not before,the time of initiation. Finally, tooth type selection mightoccur subsequent to the initiation of identical tooth pri-mordia (Fig. 3C). This possibility is not consistent withexplant and tissue recombination experiments suggest-ing that mouse molar and incisor regions are determinedat least as early as the time of initiation (Lumsden,1988).

There are various ways that patterning may be laiddown during the development of a structure. A patterncan be imposed on a pre-existing region by an organizingcenter (OC) which serves as a source of diffusible sub-stances (morphogens). Such substances may form a gra-dient that decreases with distance from the OC. Anexample of an OC that imposes pattern on an already-formed structure is the midbrain-hindbrain boundaryregion, which plays a role in patterning the midbrain(Lumsden and Krumlauf, 1996). Alternatively, patterninginformation can be imparted to cells during outgrowth ofa structure, as is believed to occur in the limb buds(Tabin, 1991; Tickle, 1995) and facial processes (Weddenet al., 1988; Richman and Tickle, 1992). In the progresszone (PZ) model for patterning the proximo-distal axis ofthe limb, cells form different elements of the limb pat-tern based on the amount of time they spend in the PZ,a region of rapidly dividing mesenchymal cells near thedistal edge of the limb (Summerbell et al., 1973). Suchcells receive signals from the overlying ectoderm; duringthe process of cell division, some cells are left behind,out of the range of these signals. Once the cells leave theprogress zone, they differentiate according to the pat-terning information they received while in the PZ.Dynamic processes in the cells of the PZ may changeover time; cells left behind at later stages of develop-ment represent a different "frozen record" of suchprocesses than those left behind earlier. The possibilitythat the different tooth types correspond to differentlengths of time in a PZ is discussed further in Section(V.1 .i). As in Fig. 3A, the pre-pattern or positional infor-mation laid down during outgrowth of the jaw primordiamight be interpreted only after a subsequent tooth initi-ation signal is received; i.e., the cells already "know" whattype of tooth to make. The latter idea receives supportfrom Lumsden (1979), who obtained complete molardentitions after culturing explanted fragments of mousemandibles from gestational stages E II to El 5 (dayspost-conception), free from any morphogenetic fieldsthat might exist in the jaw.

(111.2) SELF-ORGANIZING SYSTEMSSystems of the kind just described become self-organizedonce a set of differentiation factors is established, in thesense that further exogenous signaling is no longerneeded. The elements of the final pattern are 'contained'

within the parameters of the process, rather than beingindividually specified genetically. To recognize their orig-inal ideas, that intrinsic processes may generate regularform in self-organizing ways, in this paper we refer tosuch interactions as Bateson-Turing Processes (BTPs).Reaction diffusion systems are one major example, butthere are many alternatives (Murray, 1993), some ofwhich involve only physical cellular processes, fluid pres-sures, and the like, rather than chemical interactions.Computer simulations of reaction diffusion processeshave persuasively mimicked a diversity of biologicaltraits with pattern attributes similar to those of the den-tition as represented earlier in Fig. 1. These includeseashell (Meinhardt, 1995, 1996), butterfly (Murray, 1993;Nijhout, 1991), feather patterning (Jung et al, 1998), andmammal coloration patterns (Murray, 1993). As withChladni figures, changes in the parameter values ofrather simple processes can generate pattern variationlike that observed naturally.

Exploratory simulation of BTPs has been applied todental patterning. Osborn (1993) modeled the produc-tion of cusps and valleys by cell-mechanical forces alonga linear axis within the constrained physical matrix of adeveloping molar crown, arguing that cusp morphologyneed involve no exogenous signaling by reaction diffu-sion processes involving morphogens [we will see inSection (VI) that such signaling in fact occurs within thetooth, but the point here is that quite different dynamicprocesses can generate similar resultsl. Kulesa et al.(1996) used a reaction diffusion model involving mor-phogens to simulate the location and timing of tooth ini-tiation along the linear axis of the homodont alligatorjaw. Although both of these studies produced quite plau-sible patterns, both-like the examples from otherorganisms cited above-involved many ad hoc (or post hoc)parametric specifications. Another problem with these(and other similar) mathematical models is their instabil-ity: Slight parametric changes can have major patterningeffects. Dental patterns seem, intuitively at least, to berather stable phylogenetically and within species, so it isdifficult to know how to interpret the apparent fit of suchsimulations: We need actual genes to work with, andrecent advances have identified some candidates for us.

Reaction diffusion mechanisms can generate a vari-ety of distributions of substances (e.g., gene products),including gradients and stripes (Nijhout, 1991; Murray,1993; Meinhardt, 1996). As was stressed by O'Farrell(1994), there are essentially two schools of thought onhow distributions of gene products might be interpretedin the elaboration of pattern. One is that the concentra-tion of a morphogen distributed in a gradient causescells to adopt a particular fate. Another is that cell fate isspecified by a code of regulatory molecules (e.g., tran-scription factors). O'Farrell (1994) pointed out that thetwo views are not mutually exclusive and that there

376 Grit Rev Oral Biol

9(4):369-398 (1998)376 Crit Rev Oral Biol Med

at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

might be a sequential process of interpretation of gradi-ents to produce stripes of regulatory molecules, withineach of which further gradients are established. One ofthe questions that arise if such a view is applied to thedentition is, Which elements of the pattern are specifiedby unique codes of molecules and which are encoded byconcentration gradients? In other words, is there a spe-cific code (expression of a unique gene or set of genes)for each molar cusp, for each tooth, or only for eachtooth type? The variations in numbers of cusps and teethwithin mammals make the former two possibilities high-ly unlikely (jernvall, 1995). However, is it possible thateven the different tooth types are specified by quantita-tive differences in expression of identical gene products?Aspects of the pattern of the dentition (e.g., cusps vs. val-leys, toothed vs. diastemal regions of the jaw, differenttooth types) may simply represent nested waves andtroughs of a BTP If so, there may not be any genetic differ-ence between these mutually exclusive pattern elements,which may instead be produced by spatially or temporal-ly different concentrations of the same factors at earlierstages of development (e.g., Murray, 1993; Webster andGoodwin, 1996). Genes expressed at various times andplaces in dental patterning could be the consequence ofsuch a process, but not its cause, as we will see when weconsider the nature of the enamel knot in crown mor-phogenesis [Sections (VI. 1) and (VI.2)l. We stress thesepoints because there is a difference between seeking toexplain pattern in terms of variation in a quantitativeprocess and expecting there to be genes 'for' each ele-ment of the pattern.

(IV) Histological and DevelopmentalConsideration of the Dental Arches

Ideas about the control of dental patterning were derivedfirst from experiments that complemented the compara-tive and paleontological work discussed earlier (Butler,1995; Ten Cate, 1995; Lewin, 1997; Kollar, 1998) (we thankone of the pioneering participants, E.l. Kollar, for first-hand help in this summary). In the 1930s, Dahlberg(Huggins et al., 1934) and Glasstone (1936) opened thedoor to experimental embryology on teeth by demon-strating that tooth germs could be cultured in vitro, andshowing that EMI was a requisite for their development.Understanding of aspects of the control of tooth devel-opment progressed in the 1960s, as Slavkin and col-leagues studied aspects of the control of cultured wholetooth germs and developed serum-free culturing tech-niques (e.g., Slavkin and Bevetta, 1968; Slavkin et al., 1968;Slavkin, 1974, 1982; Hata and Slavkin, 1978; Cummings etal., 1981), in which various conditions could be manipu-lated and the presence or distribution of specific struc-tural components could be measured or manipulated;this work was advanced by several laboratories, notably

those of J.-V. Ruch and I. Thesleff (see their papers in thebibliography), who have concentrated on signaling andstructural factors, the extracellular matrix that separatesthe two major tissues, and the genes responsible forthem.

A major advance in studies of the control of toothmorphogenesis and patterning occurred when Rawles(1963), working on feather buds, found that trypsin treat-ment could dissociate the two primary tissues, allowingfor in vitro culturing of very early dental tissue. This led toa number of experiments with cultured recombinant tis-sues of epithelium and mesenchyme from the same ordifferent tooth germs (incisor, molar) or from the same ordifferent origins (first branchial arch, which will becomethe jaws, or from elsewhere in the embryo). The power ofsuch experiments was augmented when Kollar andLumsden cultured first-arch explants and/or recombi-nant tooth germs in the eye chamber of a host mouse.The essential findings (Kollar and Baird, 1968, 1970;Miller, 1969; Lumsden, 1979, 1988; Mina and Kollar,1987) were that dental programming instructions resideinitially in the epithelium but by about E12.5 (in themouse) are transferred to the responding NC-derivedmesenchyme. Subsequently, Lumsden, Hall, and othershave shown that NC from beyond the head region canrespond adequately to epithelial signal (Lumsden, 1988;Graveson et al., 1997). We will be reviewing the relevantaspects of the morphological evidence in Section (IV)and the genetic evidence in Sections (V) and (VI).

The origins of odontogenic tissues suggest ways thedentition may be patterned, and may help us to interpretor pursue the above kinds of models. For example, achallenge for gradient concepts is that the upper andlower dentitions develop differently with respect to phys-ical continuity.

(IV. 1) FACIAL PROCESSES AND THE DENTITIONThe adult mammalian dentition consists of two parallelmesio-distal (incisor-molar) axes, corresponding to theupper and lower jaws. Across a diversity of species, theteeth in the two jaws are similarly arranged and usuallysimilar in morphology. An obvious possibility is that theaxes in both jaws may be patterned by a similar or iden-tical process, and Bateson (1894) toyed with the ideathat an original single axis develops and is then slit intoupper and lower halves. How these axes are establishedmight help us understand the underlying geneticprocesses.

The jaws of amniotes develop from a number offacial processes which undergo outgrowth and subse-quent union. The relevant structures are shown in Fig. 4.The first arch develops into (1) the left and rightmandibular processes, which fuse in the midline to formthe lower jaw, and (2) the maxillary processes, whichform only a part of the upper jaw, the remainder of which

9(4 36-9 19)Ci RvOa i e 7

377Crit Rev Oral Biol Med9(4):.369-398 (1998) at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

oral

I m mesia distal

aboral

Figure 4. Elements of the early embryonic head with relevancefor the dentition, as described in the text. Md, mandibularprocess of the first branchial arch; mx, maxillary process; In,laternal nasal process; mn, mesial nasal process; hy, hyoid(second) arch.

mes-met OCR C

A BFigure 5. Schematic diagram to show the idea of early pre-patterning of tparallel axes, labeled on an El 0.5 embryo. Enlarged first arch imagesunidirectional axis, analogous to the way that the Hox genes program tior body axis; (B) two parallel unidirectional axes, analogous to the axEhind-limbs; (C) a mirror-image axis programmed by an OC (dark dot), ththe final parallel orientation, analogous to programming around thedetails, see text. i, m = presumptive incisor and molar regions; FB, forebr(HB, hindbrain; mes-met OC, mesencephalon-metencephalon organizingbranchial arch; R, rostral; C, caudal; D, distal; P, proximal relative to thfuture hinge region of the jaw.

is derived from the frontonasal mass. The latter, with theformation of the nasal pits, becomes subdivided into leftand right lateral nasal processes and left and right medi-al nasal processes. The tooth-bearing margin of theupper jaw is formed from the union of the maxillaryprocesses with the medial nasal processes. The junctionbetween these processes is probably at the suturebetween the premaxillary and maxillary bones (Miles andGrigson, 1990). Upper incisors are the only teeth gener-ally accepted as forming in the premaxilla (Miles andGrigson, 1990), which is derived from the medial nasalprocess, while the other upper teeth are derived from themaxillary process. Lumsden and Buchanan (1986)showed experimentally that this is so in the mouse,although others have argued that the maxillary processis involved in the formation of at least the distal incisorsof mammals that have them (e.g., Ruch et al., 1997).

Let us now consider how pattern-ing mechanisms such as those illus-trated in Fig. 3 might be applied inthe first arch to generate the non-incisor dentition, that is, the caninethrough molar part (mice do not pos-sess canines, but the following canbe related conceptually to mice andother mammals). We will then con-sider the incisors. Fig. 5 shows sever-al plausible scenarios. In discussingthis Fig., to be consistent with otherauthors, we use 'proximal' and 'distal'in relation to the first arch itself, cor-responding to distal and mesialpoints, respectively, relative to the

D final dentition. One scenario is basedon analogy with the way that position

P along a single unidirectional axis isprogrammed by the Hox genes IseeSection (V.2)] (Fig. 5A).

Alternatively, the adult dentitionD can be viewed as two unidirectional

parallel axes developing from agrowth edge early in the body, withpresumptive molar regions develop-

C ing first and the canine regions (orincisors, depending on the model)

the dentition as two last and most distal (again, meaningshow: (A) a single 'mesial' in the dentition itself), ashe anterior-posteri- indicated in Fig. 5B. Depending ones of the fore- and whether incisors are part of this pro-en folding to attain gramming, the two ends resultingmes-met OC. For from this single axis may not be iden-ain; MB, midbrain; tically programmed, and in fact thecenter; 2A, second maxillary and mandibular branchese embryo; and H, of the first arch grow out asymmetri-

cally in time and size. This model is

378 Grit Rev Oral Biol Med (1998)378 Crit Rev Oral Biol Med 9(4):369-398 (1998) at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

analogous to the parallel axes of the fore- and hind-limb,which grow out from separate sites along the lateral mid-line. Finally, we (Stock et al., 1997; Weiss et al., 1997) sug-gested that the dental arches might be viewed as a sin-gle, bidirectional mirror-image axis programmed by acentral OC in the primordial hinge region (betweenfuture upper and lower jaws), with symmetric patterninggradients spreading out from the OC and folding forward(Fig. 5C).

An analogy may be the midbrain mentioned earlier,which is patterned by an OC near the midbrain-hindbrainboundary from which symmetrically decreasing gradientsof the Enl and En2 homeoproteins spread rostrally andcaudally (Joyner, 1996; Lumsden and Krumlauf, 1996).This region of the neural tube contributes NC cells to thefirst branchial arches (Osumi-Yamashita et al., 1997). Fgf8in the OC is one factor responsible for establishing thesegradients, a gene also expressed in ventral epithelialstructures that will form the oral cavity and dentition,including the future jaw-hinge region; this gene is neces-sary for locating and initiating teeth [see Section(V.l.iv)). The original orientation of the presumptive jawtissue may vary from that shown in Fig. 5, and it is notclear if the known migration paths of NC cells coincidewith the programming of the dental axes.

If patterning information is laid down early enough,an initial rostro-caudal axis could give rise to two proxi-mal-distal axes (in the terminology used for Fig. 5) byfolding about the hinge region between the future jaws.In fact, Kbntges and Lumsden (1996) showed that in thechick, initially rostro-caudal organization of NC cellsalong the neural tube is converted into proximal-distalorganization along the branchial arches. A simple pat-tern of folding a symmetrical axis might imply that hind-brain neural crest populated the mandibular process andmidbrain neural crest populated the maxillary process.However, Kbntges and Lumsden (1996) showed that thiswas not the case, because hindbrain crest contributed tothe proximal regions of both maxillary and mandibularprocesses and midbrain crest to the distal portions ofboth.

One can reconcile this result with the folding of asymmetrical axis by postulating that distal portions ofboth processes arise from more rostral NC cells and thatthe maxillary process grows out from a more proximalregion of the first arch than does the mandibularprocess. In fact, the mandibular processes do appearbefore the maxillary processes, the latter of which appearto arise from a rostral, proximal region of the mandibu-lar process. The tooth-bearing ectodermal regions andtheir underlying NC could be derived from the midbrainregion, and all could be distal to the hindbrain-derivedNC that contributes to bone, cartilage, and muscle.However, this geometry seems to rule out a direct role forthe mes-met OC itself in patterning the first arch.

What about the incisors? The derivation of the upperincisors from the frontonasal mass, rather than the firstbranchial arch, raises the question of whether the entiredentition (or even just the upper) is patterned as a unit.Even from early stages of development, both the epithe-lium and the NC of the upper incisor region are likely tobe non-contiguous with similar cells in the first arch.Based on work in the chick, early (0-3 somite stage) ecto-derm of the first arch processes comes from a region lat-eral to the mesencephalic neural fold, while premaxillaryepithelium is from a more anterior region (Couly and LeDouarin, 1985, 1990). The source of NC in these regionsis similar to the source of their epithelium, since the twogrow out together; prosencephalic neural crest con-tributes to the medial nasal process, anterior midbraincrest to the lateral nasal process, and posterior midbraincrest to the first branchial arch, and the frontonasalepithelium of the mouse came from the anterior neuralridge of the E8.0 embryo, consistent with the chick data(Osumi-Yamashita et al., 1994).

Most authors treat the lower dentition as being pat-terned along a single axis. If similar induction and tooth-type specification processes apply to the upper dentalarch, they would be expected to act when upper jaw pre-cursors also represent a single axis. If this occurs prior tothe outgrowth of the facial processes, then there wouldhave to be a gap in the pattern, with no correspondingelement in the lower jaw, because of the tissue originsdiscussed above. Such patterning could not occur afterthe fusion of maxillary and medial nasal processesbecause of Lumsden and Buchanan's (1986) demonstra-tion that incisors can be obtained from medial nasalprocesses explanted prior to this fusion. An alternativeview is that the incisors in both jaws are induced andpatterned by mechanisms other than those producingother teeth. There is evidence that incisors in both jawsform from an originally single field that is continuousacross the midline Idiscussed in Section (V.1.ii)l.Differences between the two sets of incisors may relateto the different timing of the formation of this midline.

From the point of view of variation, the case could beargued that the upper and lower incisors are part of thesame, or an independent, mechanism. Upper and lowerincisors do typically resemble each other within species,as in the single, ever-growing incisor in each mouse jawquadrant. On the other hand, there is some evolutionaryindependence between them, as in the mortise (upper)and tenon (lower) arrangement in some seals that dove-tail to grasp fish prey, loss of upper but not lower incisorsin some ungulates, elephant tusks, and lemur lowertooth combs (Peyer, 1968). As will be discussed inSection (V.3), several mouse mutations affect the lowerand upper incisors differently. In very anomalous denti-tions such as the secondary homodonty of some whales,a patterning change has affected all of the teeth.

9(4)369981998 Crt Re Ora Bil Me 37

379Crit Rev Oral Biol Med9(4) .369-398 ( 1998) at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

In sum, the degree of parallelism of the evolution ofthe upper and lower dentitions suggests that some sim-ilar process may be at work in both-for example, paral-lel or mirror-image BTPs acting on a contiguous tissueaxis. This is an attractive, economical view, but thenwhen is such an axis programmed, given the develop-mental differences between the two jaws? The alterna-tive is that different inductive mechanisms operate inboth jaws, but activate similar downstream develop-mental cascades. At present, we do not know which ofthese is true.

(IV.2) NEURAL CREST AND EPITHELIUMIN DENTAL PATTERNING

The origin of teeth from ectodermal epithelium and NC-derived mesenchyme raises the question of the locationof the dental initiation and patterning information.Determining the initial source of this information is com-plicated by the fact that, during the development of thefacial processes, the NC cells migrate in association withexpanding and proliferating epithelium, as well as meso-derm (Noden, 1991). All three tissues have been shownto exhibit an initial organization along the rostro-caudalaxis that predicts their final localization in developingcranial structures, including the jaws (Couly and LeDouarin, 1990; Serbedzija et al., 1992; Osumi-Yamashita etal., 1994, 1997; Trainor and Tam, 1995). Kontges andLumsden (1996) showed that in the chick, rostro-caudalorganization of pre-migratory NC cells along the neuraltube was converted into mesio-distal organization in thebranchial arches. In the mouse, Imai et al. (1996) showedthat midbrain NC cells are included in mandibularmolars, but the exact source of the NC cells of the otherteeth has not been directly demonstrated.

Recall that, despite the highly patterned migrationof NC cells, tissue recombination experiments in themouse suggest that information for tooth initiation andtooth type determination initially resides in the epitheli-um, prior to the morphological appearance of teeth(Lumsden, 1988; Kollar and Mina, 1991). Furthermore,NC migration is not necessary for tooth development,and mandibular epithelium is capable of inducing odon-togenesis in trunk NC cells (Lumsden, 1988; Graveson etal., 1997). The control of tooth type subsequently switch-es from the epithelium to the mesenchyme (Lumsden,1988).

(IV.3) THE AUTONOMY OF THE TOOTH GERMAs noted earlier, cultured explants in the mouse showthat teeth are autonomous units, even before their pla-codes are morphologically visible. Individual teeth, andindeed whole sets of molars, develop in isolation fromthe jaw (Lumsden, 1979; Luckett, 1993; Butler, 1995; TenCate, 1995). If sectioned early enough, partial tooth germrudiments can generate fully formed teeth, an informa-

tive fact about scale invariance in the patterning process[see Section (VII)l. Within teeth, individual cusps alsoappear and develop independently (Snead et al., 1988;Butler, 1995; Jernvall, 1995). The incisor and molarregions by E1O.5 bear within them whatever instructionsare needed to form teeth of the proper shape.

(V) Gene Expression and PatternEstablishment in the Dental Arches

Recent work has begun to identify genetic mechanismsthat may be involved in different stages of dental pat-terning, which can be roughly characterized as (a) dentalpre-patterning, (b) tooth initiation, and (c) tooth mor-phogenesis and terminal histogenesis. We look first atextracellular signaling factors which, if they act at suffi-ciently long cellular distances, are candidates for the ele-ments of BTPs to establish periodicities along the dentalarches. Then we will look at intracellular (nuclear) tran-scription factors that may respond to fields establishedby the former.

(V. 1) THE ROLE OF EXTRACELLULARSIGNALS IN DENTAL PATTERNING

There are several times and places when extracellularsignaling could be involved in patterning the dentition.

(V.1 .1) Similarities between facialprimordia and limb buds

The limb bud is one of the most extensively studiedmodels for pattern formation, and patterning of thisstructure is believed to be associated intimately withoutgrowth (Tabin, 1991; Tickle, 1995). Numerous extra-cellular signaling molecules that play a role in out-growth and patterning have been identified (Shubin etal., 1997). Several authors have pointed out similaritiesin the outgrowth of facial processes and limb budsextending to the expression of extracellular signalingmolecules (Shubin et al., 1997; see references in Stock etal., 1997). Stock et al. ( 1997) suggested that the dentitionmight be patterned during the outgrowth of the facialprocesses.

In the limb, a thickened apical ectodermal ridge(AER) is required for outgrowth and the maintenance ofproliferating, undifferentiated mesenchymal cells that lieunderneath the AER. Proximal-distal (PD) patterningdepends on the time cells spend in this mesenchymalprogress zone before being left behind and differentiat-ing. A region of posterior, distal mesenchyme, the zone ofpolarizing activity (ZPA), controls anterior-posterior (AP)patterning. Information for patterning the dorso-ventral(DV) axis resides in the ectoderm. AP and PD limb pat-terns are controlled by Hox genes (Nelson et al., 1996),whose expression there is induced by extracellular sig-naling molecules. Epithelial Fgf8 provides a signal for the

380 Crit Rev Oral Biol Med 9(4)369-398380 Crit Rev Oral Biol Med 9(4):369-398 (1998) at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

initial limb budding (Crossley et al., 1996), while Shh inthe ZPA controls AP pattern by inducing Bmp2 and thenHox genes (Tickle, 1995; Duprez et al., 1996). Continuedoutgrowth requires AER expression of Fgf4 and Fgf8(Tickle, 1995; Crossley et al., 1996), and pattern along theDV axis is controlled by Wnt7a in the dorsal ectoderm(Tickle, 1995). Additional molecules similarly deployed inlimbs and facial processes are described in Shubin et al.(1997).

In the facial primordia, there is no analog to theAER, but outgrowth is dependent on EMI (Wedden et al.,1988; Richman and Tickle, 1992), and a similar set ofgrowth factors is deployed (Francis-West et al., 1994; Walland Hogan, 1995). Delineation of the arches has beenproposed to be the result of growth-inhibiting effects ofBmp7 (also found in the limb) in the branchial cleftregions (Wall and Hogan, 1995), and outgrowth may bemaintained by epithelial Fgf8 and Fgf4 (Niswander andMartin, 1992; Crossley and Martin, 1995). Bmp2 and Bmp4are initially expressed in the distal epithelium, then inthe underlying mesenchyme (Francis-West et al., 1994;Wall and Hogan, 1995; Helms et al., 1997). There is a PDgradient of Wnt5a expression in both the limb bud andthe frontonasal process (Gavin et al., 1990). Oral-aboralpolarity (tooth-forming vs. non-tooth-forming surfaces)of the mandibular process may be related to expressionof Wnt3 and Wnt3a in the aboral epithelium (Roelink andNusse, 1991).

An interesting difference between signaling factordeployment in the limb and facial processes relates topolarizing activity and Shh expression. Polarizing activityin the frontonasal process of the chick (Helms et al., 1997;assayed by grafting tissues to the limb bud and scoringfor gene expression and digit duplication), or the mousetooth germ (Koyama et al., 1996), was limited to theepithelium. In the limb, such activity is located in themesenchyme (the ZPA). Shh expression was also limitedto the frontonasal epithelium rather than to the mes-enchyme, as in the limb (Helms et al., 1997).

(V.1 .ii) Sonic hedgehog and the subdivisionof a single incisor field

In the limb, Shh may be acting as a morphogen in theclassic sense of that term. Whether it does so in thefacial processes or dentition is not known. But a role forShh in incisor patterning in relation to the midline is sug-gested by human holoprosencephaly (Roessler et al.,1996). This syndrome is characterized by the incompletesubdivision of structures across the midline of the head,such as the forebrain, eyes, nose, and, significantly, theupper dentition. In mild cases, a single central maxillaryincisor may be located on the jaw midline (i.e., there arethree rather than four upper incisors). This phenotypehas been found in human heterozygotes for Shh muta-tions, indicating a quantitative dose effect.

Classic studies suggested that, in vertebrates, an ini-tially single eye field is subdivided by influences from theprechordal plate [axial mesendoderm underlying theforebrain (Chiang et al., 1996)1 The cyclopic phenotype ofShh knockout mice led Chiang et al. (1996) to propose thatShh in the prechordal plate (Marti et al., 1995; Shimamuraet al., 1995) contributes to this signal. The role of the pre-chordal plate and/or Shh in subdividing the eye field isfurther supported by experiments in zebrafish (Ekker etal., 1995; MacDonald et al., 1995), Xenopus, and the chick(Lietal., 1997).

We suggest that the single maxillary central incisorin holoprosencephaly similarly represents an incompletesubdivision of an initially single upper incisor field andthat Shh is responsible for this subdivision across themidline in normal development. Lumsden and Buchanan(1986; Lumsden, 1982) listed a number of observationsin the mouse that are consistent with an initially singleincisor field in the mandible.

In addition to being expressed in the initial dentalepithelial thickening (Bitgood and McMahon, 1995), Shhexpression has been reported in a number of locationsthat could influence early patterning of the dentition(Wall and Hogan, 1995; Helms et al., 1997). In the chickembryo, Shh is expressed at early stages in an epithelialregion that later gives rise to the oral cavity. Subsequentexpression is found in the stomatodeal roof ectoderm,followed by expression in the midline ectoderm of thefrontonasal process. Kronmiller et al. (1995b) describedexpression in the midline region of the EIO mousemandibular arch, and Helms et al. (1997) reported Shhexpression in a region of the chick mandible correspond-ing to the mammalian dental lamina. Helms et al. (1997)cited unpublished data that inhibition of Shh expressionin the frontonasal process caused phenotypes reminis-cent of holoprosencephaly.

This latter result suggests that Shh patterning of thedentition may be due to expression at relatively late devel-opmental stages. However, because the structures affect-ed in holoprosencephaly (forebrain, nasal epithelium,retina, premaxillary epithelium) arise from regions thatare located close to one another at early somite stages ofmouse and chick development (Couly and Le Douarin,1985, 1987; Osumi-Yamashita et al., 1994), it is possiblethat the effects of Shh on the upper incisor dentition aredue to expression at this stage or earlier. Specifically, thedefects in holoprosencephaly may all be due to reducedlevels of Shh protein in the prechordal plate, since thisregion is close to the affected structures. This hypothesisis developed below in connection with a hypothesizedinteraction between Shh and Pax6 in patterning the denti-tion. It is also possible that the dental defects in holo-prosencephaly are the result of insufficient levels of Shh atseveral developmental stages, with frontonasal expres-sion downstream of expression in the prechordal plate.

9)4) 369 398 (1998) Grit Rev Oral Biol Med3813819(4):369-398 (1998) Crit Rev Oral Biol Med

at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

(V.1 .iii) Shh and Pax genes may functionin a midline signaling system for the dentition

Stock et at. (1997; Weiss et al., 1998) suggested that Shhmay pattern the dentition through its regulation of Paxgenes. This proposal was based on the dental phenotypeof mice with a mutation in the Pax6 transcription factor(Kauffman et at., 1995), the effects of Shh mutations inhumans (Roessler et al., 1996), and a previously hypothe-sized signaling system for patterning the optic vesiclealong its proximal-distal axis (Ekker et at., 1995;MacDonald et al., 1995; Chiang et al., 1996). Analysis ofPax2 and Pax6 expression in zebrafish overexpressing ordeficient in Shh (Ekker et at., 1995; MacDonald et al., 1995)and mice homozygous for an inactivation of Shh (Chianget al., 1996) supported a model in which Shh activity in theventral midline of the forebrain and/or the underlyingprechordal plate promotes proximal fates in the opticvesicle (optic stalks) by stimulating Pax2 expression andinhibits distal fates (optic cups) through repression ofPax6 expression.

Mice homozygous for a mutation in Pax6 exhibitdefects of the optic vesicle and fail to develop lens ornasal placodes (Grindley et al., 1995; Kauffman et al.,1995). These mice also frequently show additional upperincisors distal (lateral) to the normal pair (Kauffman etat., 1995), a pattern clearly affected by other genes,because its strength of effect depends on the geneticbackground of the affected mice (Quinn et at., 1997).Stock et al. (I1997) interpreted these additional teeth as anexpansion of mesial (toward the midline) fates at theexpense of fates distal to the midline (no teeth).Interpretation of the dental phenotype of Shh mutationsin humans described above as a reduction of mesial fatesled to the proposal that Shh promotes mesial fates in theupper incisor dentition and inhibits distal fates in partthrough repression of Pax6 expression.

Previously published expression patterns of Pax6 inthe mouse (Grindley et al., 1995) and the chick (Li et al.,1994) suggest that any patterning effects of this gene onthe dentition must be the result of expression wellbefore the appearance of tooth primordia, and that pat-terning of the lower jaw would have to differ at least inpart. At E8.0-E8.5 in the mouse (early somite stages),Pax6 is expressed in a broad region of the head surfaceectoderm covering the forebrain, but not the presump-tive first branchial arch (Grindley et al., 1995). This expres-sion domain may overlap or be contiguous with pre-sumptive premaxillary (and hence upper incisor) epithe-lium, as defined in the mouse (Osumi-Yamashita et al.,1994) and chick (Couly and Le Douarin 1985, 1987). In thechick, Pax6 expression has been reported to be lower orabsent in ventral midline epithelium at similar stages (Liet al., 1994), consistent with a role for this gene in con-trolling distal fates in the upper incisor dentition. Inaddition, the lack of Pax6 expression in the presumptive

first branchial arch epithelium may explain why mutationof the mouse gene affects only the upper incisors.

Tissue transplantation experiments in the chickshowed that the prechordal plate can inhibit Pax6expression, and, as described above, this structure islikely to be responsible for the subdivision of the eyefield (Li et at., 1997). This effect on the eye field would bedue to the juxtaposition of the prechordal plate and theventral forebrain. In the anterior head, the ventral surfaceepithelium is also likely to be in proximity to the pre-chordal plate (Seifert et at., 1993). We speculate that Shhexpression in the prechordal plate is responsible for theexclusion of Pax6 from ventral midline epithelium andhence contributes to the patterning of the upper incisordentition. This patterning might be a crude regionaliza-tion of the head ectoderm. Pax6 expression would definea region competent to form lens and nasal placodes aswell as the distal portion of the premaxillary epithelium(toothless in the mouse). More precise patterning infor-mation might be provided by levels of Pax6 protein,which has been shown to be important for function(Schedl et at., 1996). Variation in premaxillary tooth num-ber caused by Shh and Pax6 mutations indicates thatthese genes are involved at least indirectly in tooth initi-ation. We suggest that teeth are initiated after cells haveinterpreted their position relative to the midline in thehead epithelium. Whether the postulated midline signal-ing system (Stock et at., 1997) controls tooth shape (i e,the incisor shape gradient; Butler, 1939) as well remainsto be determined; if so, humans with Pax6 mutationsmight exhibit homeotic transformations of distal incisorsinto central incisors. Such changes have not been report-ed, but most patients with Pax6 mutations are heterozy-gous (Glaser et al., 1994), and the phenotype could bemissed easily.

It is unknown whether other Pax genes play a role inthe dentition similar to that proposed for Pax2 in the opticvesicle (specification of mesial fate). Pax2 expression hasnot been reported in the ventral midline of the head ecto-derm, and no effects on the dentition have been reportedin association with Pax2 mutations in mice and humans(Torres et al., 1995, 1996; Sanyanusin et al., 1995; Favor et al.,1996). Paxl-deficient mice may lack upper incisors(Dietrich and Gruss, 1995), a phenotype consistent with arole for the gene in specifying mesial fates (Stock et al.,1997). However, Paxi, like a number of other Pax genes(including Pax9), is expressed in the mesenchyme of thefacial processes rather than the early head epithelium(Dietrich and Gruss, 1995; Neubuiser et al., 1995; Mansouriet al., 1996). The activation of Pax6 by Shh and possibleinhibitory effects of the latter on an unknown gene resem-ble likely components of a BTP If such a process isinvolved in patterning the incisor dentition, reaction diffu-sion models predict specific types of interactions with as-yet-undiscovered components (Meinhardt, 1996).

382 Crit Rev Oral Biol Med (1998)382 Crit Rev Oral Biol Med 9(4):369-398 (1998) at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

(V.1 .iv) Control of tooth initiation

Several extracellular signaling molecules have beenshown to be expressed in tooth germs from their earliestappearance as an epithelial thickening (e.g., Maas andBei, 1997). Recent experiments by Neubuser et al. (1997)suggest that Fgf8, Bmp4, and Bmp2 are involved in toothinitiation prior to any visible morphological signs ofteeth. Fg48 is expressed by E8.5 or earlier in what willbecome the oral ectoderm; by E9.5-10.5, expression is ina much broader domain than those in which specificteeth will form. Bmp4 is also broadly expressed in theoral ectoderm, but in more restricted areas than Fgf8, andtooth initiation appears to occur in those areas whereFgf8 but not B3mp4 (or Bmp2) is expressed. The relation-ship between the expression patterns and sites of futuretooth initiation was assessed by a comparison with Pax9expression. This gene is subsequently expressed in den-tal mesenchyme and is necessary for tooth developmentbeyond the placode stage (Peters et al., 1998). Neubuseret al (1997) showed that Fgf8 induced Pax9 expression,and that this induction was inhibited by Bmp4. Fgf8 hasalso been shown to induce the transcription factors Lhx6and Lhx7 in dental mesenchyme, where they are natural-ly expressed in tooth germs (Grigorion et al., 1998).

The interactions of Fgf8, Bmp4, and Bmp2 resemblethose expected for components of a reaction diffusionsystem. However, if their distributions are "read out" intoa simple code of Pax9 expression that determines toothlocation, then the interesting patterning question is howthe distributions of the former three genes are estab-lished. Furthermore, there a few difficulties in interpret-ing the findings described above. The link between Pax9expression and tooth initiation remains indirect, sincethe knockout mice still exhibited the earliest morpholog-ical stage of tooth initiation. There has been evidencethat Bmp4 induces rather than inhibits tooth formation(Vainio et L)., 1993; Maas and Bei, 1997). Bmp2 and Bmp4are expressed in the mouse diastema region, in thicken-ings in the dental lamina that subsequently disappear(Tureckova et al., 1995; Ruch et al., 1997). It has not yetbeen shown that Fgf8 and Bmps are interacting with eachother relative to the establishment of their expressiondomains (even if they interact in some way relative toPax9). The former is the patterning process, and isunknown.

(V.1 .v) Retinoic acid

Retinoids are among the known candidates for a role asmorphogens in embryonic development, by activatinggene expression through their receptors (Petkovich,1992; Mangelsdorf et al., 1994). Retinoic acid is a dif-fusible factor that produces pattern alterations in a num-ber of regions of the developing embryo, including theaxial skeleton, limbs, and craniofacial region (Conlon,1995). Krcnmiller and Beeman ( 1994) reported a concen-

tration gradient of all-trans retinoic acid between incisorand molar regions of the mouse mandible and an oppos-ing gradient of retinol; retinoid receptors and bindingproteins are present in the developing jaws (Dolle et al.,1990; Ruberte et al., 1990, 1991, 1992). This concentrationgradient and ability to modify tooth morphology (seebelow) suggest a role for retinoic acid in tooth develop-ment, although retinoic acid may not act strictly throughconcentration gradients (Conlon, 1995).

Retinoids affect the extent of the incisor region.Maternal administration of retinoids to mice inhibitedmolar development, produced supernumerary incisors,and fused tooth germs in their offspring (Deuschle et a!1959; Kalter and Warkany, 1961; Knudsen, 1965, 1967).The periodicity of the teeth in the dental lamina wasaffected, but there were no homeotic changes or distur-bances of tooth morphology. In vitro experiments withexogenous retinoids have led to missing molar toothgerms (Kronmiller et al., 1992, 1995a; Zhao et al., unpub-lished), and Kronmiller et al. (1995a) reported that appli-cation of all-trans retinoic acid to cultured mousemandibles produced supernumerary tooth initiationsites (epithelial invaginations) extending into the mousediastema and molar regions. Knockouts of retinoic acidreceptors in mice have led to the absence of upperincisors (Mark et al., 1995). This defect was associatedwith failure of outgrowth of the frontonasal processes.Helms et al. (1997) found that retinoic acid caused arrestof chick facial process outgrowth as well as loss of Shhexpression. Hardcastle et al. (I1998) found that ectopic Shhinduced ectopic invagination sites in the dental lamina.

Retinoids also affect structures within molar teeth.Mark et al. (1992) have generated anomalous numbersand locations of cusps in cultured molar tooth germsexposed to exogenous retinoic acid. The effects of thisdiffusible signal on the periodicity of structure locationwithin and between teeth suggest that retinoids affectpatterning by altering the parametric value (concentra-tion) of an element in a BTP.

(V.2) DOES PATTERNING OF THE DENTITIONINVOLVE NUCLEAR TRANSCRIPTION FACTOR CODES?

One mechanism for specifying position established bydiffusible signaling factors is to "switch on" regionallydifferent sets of transcription factors that act as selectorsfor alternative downstream differentiation cascades.Segmental identity along the rostro-caudal axis of theaxial skeleton and hindbrain is controlled by the summedsequential expression domains of the 39 members of theHox gene family. These homeobox-containing transcrip-tion factors are arranged as four clusters of cis-paralogs(genes descended by duplication events from an ancientancestral Hox gene, linked together on the same chro-mosome) (Hunt and Krumlauf, 1992; Ruddle et al., 1994;Maconochie et al., 1996). The set of Hox genes expressed

9(4)369 398(1998) Crit Rev Oral Biol Med 383Crit Rev Oral Biol Med9(4)-369-398 (1998) at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

MedialNasalProcess

I ;%I..,e Maxillary

Process

K\MandibularProcess

Figure 6. Domains of expression of Dlx genes in the El 0.5 mouse,drawn schematically, showing the similarities in domains of thelinked pairs of Dlx genes, identified by number in the Fig.

in a given region of the anterior-posterior axis is a com-binatorial code that specifies position along that axis(Krumlauf, 1994; Ruddle et al., 1994; Burke et al., 1995).Corresponding genes in similar positions within each ofthe Hox clusters, called trans-paralogs, have largely simi-lar expression boundaries and function. Experimentalalteration of this code can cause homeotic changes inthe morphology of structures forming at the affectedboundary (Krumlauf, 1994)-a term coined by Bateson(1894) for just these kinds of changes observed in nature.

The dentition seemed an obvious candidate for asegmental, regionally differentiated, homeotic organ sys-tem controlled by a Hox code (Weiss, 1990, 1993; Weisset at., 1995). The evolution of heterodonty is roughly par-allel in paleontological time and phylogeny to Hox-relat-ed differentiation among elements of the limb and verte-bral column, suggesting-on morphological grounds-that similar Hox mechanisms could be responsible fordental patterning. Unfortunately, paraphrasing ThomasHuxley, this is a beautiful theory slain by the ugly factthat the most anterior expression of the Hox genes doesnot include the first branchial arch (Hunt and Krumlauf,1992; Prince and Lumsden, 1994).

It is possible that other homeobox genes areinvolved in patterning the dentition. We and othersshowed that multiple members of the Dlx and Msx class-es of homeobox genes are expressed in developing toothgerms (Mackenzie et al., 1991, 1992; Porteus et al., 1991;Robinson et al., 1991; Dolle et al., 1992; Weiss et al., 1994,1995). In mammals, there are at least six Dlx homeoboxgenes, which are arranged as three pairs of adjacent

genes, each linked to a Hox cluster. These linkage groupsprobably arose by joint Hox-Dix cluster duplicationevents (Nakamura et al., 1996; Stock et al., 1996a).Whereas the Hox genes are expressed posterior to themes-met OC, the Dlx genes are expressed anterior to thatpoint, leading us to the irresistible hypothesis that com-binatorial codes of Dlx expression are involved in toothand jaw evolution and development (Weiss, 1993; Weisset al., 1994, 1995). Subsequently, Sharpe (1995) includedDlxl and Dlx2 spatial expression in a specific code bywhich they might work, ideas that have been updated bymore recent work (Qiu et al., 1997; Thomas et al., 1997,1998; Thomas and Sharpe, 1998).

(V.2.i) Dlx, other homeobox genes,and dental patterning

The expression pattern of the Dlx genes reflects, at leastas markers, the different stages in dental patterning.

(V.2.a) Dental pre-patterningThe first observed expression of Dlx genes potentiallyrelevant to dental patterning is the expression of Dlx2 atE9.5, coinciding with NC cells migrating from the poste-rior midbrain toward the first arch (Robinson andMahon, 1994). This expression pattern is shared at leastas far back as the common ancestor between mammalsand teleost fish (Akimenko et al., 1994).

By about E10.5, when the maxillary and mandibularprocesses of the first arch have formed, all six Dlx genesare expressed, mainly in the mesenchyme (Dlx2 is alsoexpressed faintly, and Dlx3 in a restricted region near thelower midline, in the oral epithelium). As shown in Fig. 6(Qiu et al., 1997; Zhao et al., manuscript in preparation),the genes in each pair of cis paralogs have highly over-lapping expression domains. DlxI and Dlx2 are expressedin both mandibular and maxillary processes of the firstarch, though probably not extending mesially as far asthe incisor domains. The cis pair Dlx5 and Dlx6 isexpressed in the mandibular process only. Dlx3 and Dlx7(Nakamura et al., 1996) are found only in non-dentalregions of the first arch. This expression pattern appearsto reflect regulation of Dlx genes at this stage in a wayrelated to their genomic organization.

Sharpe and colleagues (Sharpe, 1995; Qiu et al., 1997;Thomas et al., 1998) have proposed combinatorial homeo-box codes for dental patterning at this stage, involvingDlx, Msx, and Barxl genes, with overlapping expressiondomains specifying tooth location and type along thedental lamina. This model has incisors specified by Msxgenes, and molars by Dlx and Barxl genes. Below, we dis-cuss the case for the involvement of Dlx genes in suchcodes. Knockout experiments involving either Dlxi orDlx2 alone yielded no apparent dental effects, but inmice lacking both genes, upper molars (defined by loca-tion) did not progress beyond the stage of epithelial

384 Grit Rev Oral Biol Med 9(4)369-398 (1998)384 Crit Rev Oral Biol Med 9(4):369-398 (1998) at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

thickenings (Qiu et al., 1995, 1997; Thomas et al., 1997).Lower molars and all other teeth developed normally inthe dual-knockout mice, and no effect on the expressionof early epithelial markers (Shh and Fgf8) was observed.Ectopic cartilage was found in the region normally occu-pied by molar mesenchyme.

Qiu et al. (1997) interpreted the knockout results andexpression data as evidence for a patterning system sim-ilar to the summed expression of the Hox genes to estab-lish a single proximal-distal axis along the first arch (asin Fig. 5A). In this patterning system, the code for the dis-tal portion of the arch (their terminology, correspondingto the mesial tip of the mandibular process) is theexpression of Dlxl, Dlx2, Dlx3, Dlx5, and Dlx6. More prox-imal regions of the mandibular process express all ofthese genes except Dlx3, and the most proximal portionof the first arch (maxillary process) is specified by theexpression of only Dlxl and Dlx2. This Dlx code washypothesized by Qiu et al. (1997) to be involved in pat-terning all of the structures of the first arch. The exactcorrespondence of the expression boundaries to toothtypes was not described. Nevertheless, the authors inter-preted the Dlxl/Dlx2 knockout result as evidence thatthese two genes form a code that specifies odontogenicneural crest cells as the upper molar.

There are several problems with the above interpre-tation, and alternative hypotheses remain to be exclud-ed. If combinatorial patterns of Dlx expression serve asselectors for molar tooth type, loss of function mutationsmight be expected to result in homeotic changes towarda more proximal identity (using the analogy between Dlxand Hox codes suggested by Qiu et al., 1997). The factthat Dlxl and Dlx2 are hypothesized as the code for themost proximal tooth types leaves unclear what thedefault state of odontogenic mesenchyme (in theabsence of Dlx genes) might be. Perhaps it is cartilage, asis found in the upper molar region of the dual-knockoutmice. Another possibility, however, is that the defaultstate is the incisor. Sharpe (1995) and our unpublisheddata suggest that both upper and lower incisors developin a region where mesenchymal Dlx expression is lacking.The lack of an upper-molar-to-incisor transformation inthe dual-knockout mice might be due to the inability ofmesenchyme to respond to epithelial signals in theupper molar region rather than to a defect in the mecha-nism of tooth type selection. The latter explanation isconsistent with the fact that epithelial thickenings devel-op in the upper molar region, but tooth developmentdoes not progress beyond this stage (Qiu et al., 1997;Thomas et al, 1997). A way of testing whether the Dlxcode of Qiu et al. (1997) functions in the selection oftooth type is to construct a Dlx5/Dlx6 dual-knockout. Thiswould make the Dlx code in the lower molar region iden-tical to that in the upper molar region. Assuming thatdevelopmental anomalies in other structures were not so

severe as to preclude analysis of the dentition, Dlxinvolvement in tooth type selection would predict thatthe lower molars will acquire the cusp pattern of uppermolars. Conversely, the absence of an effect on lowermolar teeth would be consistent with the involvement ofDlx genes in processes downstream of tooth type selec-tion, coupled with functional redundancy among familymembers.

A role for nested domains of mesenchymal Dlx geneexpression seems inconsistent with the results of the tis-sue recombination experiments described by Lumsden(1988) and Kollar and Mina (1991). These suggest thatinformation for tooth type selection is present in theepithelium at the time that nested Dlx expressiondomains are found in the mesenchyme. Such informa-tion is believed to be transferred to the mesenchyme atlater stages, but only after Dlx gene expression is moresimilar between different types of teeth (Zhao et al.,unpublished). It should be pointed out, however, that theconclusions of Lumsden (1988) are not universally sup-ported (Ruch, 1995). Finally, we note that the Dlx code ofQiu et al. (1997) differs somewhat from that proposed forthe Hox gene family. In the former case, cis-paraloguesshare expression domains and may be functionallyredundant, while in the latter case, the more closelyrelated trans-paralogues are more similar in expressiondomain and possibly in function (Sharkey et al., 1997).

(V.2.b) Tooth initiation

By about E12.5, general jaw mesenchymal expression ofDlx genes has waned, and tooth initiation is beginning inspecific sites, the dental placodes, in the two jaws. Inthose sites, only the trans paralogs Dlx2 and Dlx3 areexpressed, in upper and lower jaws and both types ofteeth (Thomas et al., 1995; Zhao et al., manuscript inpreparation). If Dlx genes have general redundancy, wecan hypothesize that a Dlx3 knockout will not affect toothinitiation, but that a dual D1x2/3 knockout will preventany teeth from initiating. Recently, Dlx3 has beenknocked out in the mouse, but most -/- embryos fail tosurvive beyond E10.5, so these animals have not beeninformative regarding tooth development or jaw pro-gramming (T. Sargent, personal communication).

(V.2.c) Tooth morphogenesis and terminal histogenesis

Once initiated, teeth develop as independent units. Allsix Dlx genes are expressed during these stages, but indifferent regions of each tooth type, at different times,also varying within a tooth during its odontogenesis(Zhao et al., manuscript in preparation); we do not find asimple pattern of expression during these processes thatwe can relate to genome organization. Among our obser-vations are that only the trans-paralogs Dlx2 and Dlx3 areexpressed in incisor epithelium during most of odonto-genesis, and that Dlx7 first appears in teeth at about

9)41 369 398 (1998) Grit Rev Oral Biol Med385385Crit Rev Oral Biol Med9(41-.369-398 (1998)

at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

E15.5. The expression of the Dlx genes is typically con-fined to the tooth germ or dental follicle (and not sur-rounding tissue). Unlike their differences during jaw pat-terning at E10.5, the complex expression of the Dlx genesduring tooth morphogenesis is essentially the same inthe upper and the lower teeth.

TPwo features of Dlx expression during dental pattern-ing also characterize many if not most of the other genesdiscussed in this review: (1) multiple use of the samegenes during dental patterning and development, and (2)independent expression and hence regulation of theseindividual genes at different stages (the same combina-tions are not always co-expressed). Unless tissue-specificsplicing variation is involved, it seems likely that tissue-specific function-rather than protein function-residesin the patterns of regulation of these genes. Unless targetDNA binding regions are accessible in only some tissuecontexts-for example, by tissue-specific patterns of chro-mosomal packaging-the multiple specificities of thesame gene probably depend on the other regulatory genesthat are co-expressed in each context. And the multiplecontexts in which a given Dlx gene is itself activated aredriven by the presence of multiple enhancer sequencesflanking the gene (McGuinness et al., 1996; our work,unpublished).

It is tempting to speculate that the evolution of den-tal patterning may correlate directly with the duplicationevents that produced the Dlx-Hox complexes, especiallygiven the relationship between the anterior-posteriordomains of their respective expression. However, theduplication events occurred long before the evolution ofheterodont differentiation in the mammalian dentition(Ruddle et al 1994; Burke et al., 1995; Stock et al., 1996a).

Overall, on present evidence, the most likely role ofDlx genes in jaw patterning seems to be to relate to theregional responsiveness of mesenchyme to epithelial ini-tiation signals, and subsequently to participate in initia-tion and morphogenesis in ways not yet known. Whetherthese genes have anything to do with tooth type directly,rather than indirectly, is not yet clear.

(V.2.d) Other transcription factors knownto be present during tooth development

Miscellaneous other transcription factors are expressedduring dental patterning (and more will likely be found).The paired-box gene Ptx2 (also called Otlx2) is expressedin the early stomatodeal ectoderm (Mucchielli et al., 1997),and that gene is responsible for Rieger syndrome, whichaffects tooth development (Semina et al., 1996). Barxl isexpressed in the mesenchyme in the region that will latercontain molars (Tissier-Seta et al., 1995). At the tooth initi-ation stage, other transcription factors found in dentalplacodes include Msx2 and Lefl. Barxl continues to beexpressed only in the molar region mesenchyme. A relat-ed gene, Barx2, is also expressed in dental tissue (Jones et

al., 1997). During tooth morphogenesis, as the bud stageprogresses to the bell stage and final histogenesis, Msxland Msx2, Barxl, Lhx6, Lhx7, and Ptx2, and at least 15 Soxgenes also appear to be expressed in tooth germs (Stocket al., 1996b; Maas and Bei, 1997; Toothbase, 1997).

(V.3) GENE INTERVENTION EXPERIMENTSWITH EFFECTS ON DENTAL PATTERNING

Several genes have been inactivated that affect tooth germmaturation. Knockout mice lacking Msxl or a combinationof Msxl and Msx2 initiate but do not complete incisor andmolar development (Satokata and Maas, 1994; Maas andBei, 1997). This effect is similar to that on upper molars ofDlx1/2 knockouts. A natural mis-sense mutation in humanMsxl caused agenesis of second premolars and thirdmolars of the permanent dentition (Vastardis et a)., 1996);however, the missing teeth are also the last to form duringdevelopment and might be the most susceptible todefects in developmental processes common to all teeth(Thesleff, 1996). Activin is a member of the TGF3 family;inactivation of activin3A, the activin binding protein follis-tatin, or type II activin receptors have caused cessation ofdevelopment of all teeth (except upper molars) at the birdstage (Ferguson et a., 1998). Lef 1 is a member of the HMGbox family of genes, which alter DNA conformation tofacilitate the binding of other transcription factors.Missing or developmentally stunted teeth are found inknockouts of Lefl (van Genderen et a)., 1994); this pheno-type resembles that for Msxl (Maas and Bei, 1997) andPax9 knockouts (Neubuser et al., 1997). Ectopic Leflinduces additional tooth development in dental laminaand lip furrow, as well as ectopic hair in these regions,suggesting a general affect in tooth-competent epithelium(Zhou et al., 1995). Recent tissue recombination experi-ments have shown that Lefl is transiently necessary in theepithelium when the instructive/inductive signal is passedto the mesenchyme (Kratochwil et al., 1996).

In these experiments, teeth initiate and/or develop inthe right number and in the right place, suggesting thatthe type of tooth is also not affected, especially if theepithelium is the primary instructive tissue. It is interest-ing that none of these mutations produces homeotic effects.Instead, they are either generally dysgenic, resemble nat-ural and phylogenetic variation in tooth number, or affectonly one component of a tooth class (e.g., in only one jaw,or in only some members of a series). The effects appearto affect the competence of the dental lamina or theunderlying mesenchyme, but suggest that the specifica-tion of tooth identity itself involves processes not yetinvestigated or identified.

(VI) What is a Segment? Genesand Patterning within a Tooth

What about patterning within teeth? How do the instruc-tions that enable teeth to develop as autonomous units

386 Crit Rev Oral Biol Med 9(4) 369 398386 Crit Rev Oral Biol Med 9(4):369-398 (1998) at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

generate a three-dimensional crown pattern, such asthose shown in Fig. 1? Although knowledge in this arealags behind that of basic signaling and gene expression,we are beginning to assemble indications about whatcrown patterning involves.

(VI. 1) MOLAR MORPHOGENESIS

If one looks at the changing shape of a developing tooth,the resemblance to the formation of wave patterns-likethose so effectively modeled by reaction diffusionBTPs-is striking. Form arises from a slightly thickenedbut otherwise formless dental lamina, leading to thefamiliar bud, cap, and bell stages of crown development(Fig. 7A. One might say, subjectively, that the structurethat emerges is latent within the initial conditions of thetooth germ.

At the cap stage of molar development, lingual andbuccal ridges called enamel grooves form from the develop-ing inner enamel epitheliunm (IEE), wrapped along a con-densed, rod-like, intervening epithelial ridge of cellsknown as the primary enamel knot (EK), which runs alongthe mesial-distal axis of the jaw (Schour, 1962; Lesot etal., 1996; Peterkova et al., 1996; Vaahtokari et al., 1996a).Gradually, cusps appear as secondary 'waves' insequence along each enamel groove (shown in cross-section in Fig. 7A; Jernvall et al., 1994; Butler, 1995).During the bell stage, the primary EK disappears, and thesecondary EKs appear on the tip of each cusp. Unlike thelongitudinal rod-like primary EK, the secondary EKs arecell clusters that remain on the tips of the cusps until thecrown pattern is nearly completed (Fig. 7B; lernvall et al.,1994).

(VI.2) THE ENAMEL KNOT AS AN ORGANIZINGCENTER FOR THE DEVELOPMENT OFINDIVIDUAL MOLARS

The EK is a cluster of non-dividing cells derived from thedental epithelium. A potential role for this structure incrown pattern formation has been inferred for a longtime (see Kirino et al., 1973; Mackenzie et al., 1992; Butler,1995, lernvall, 1995). The EK may serve as an OC to con-trol relative growth and growth inhibition in neighbor-hoods within the developing tooth (Thesleff andSahlberg, 1996; Thesleff and Jernvall, 1997).

Analogy between the patterning of the jaws and limbwas discussed earlier, in Section (V). Similar analogieshave been suggested for the role of EMI or of the EK inthe development of individual teeth (Lumsden, 1979;Jernvall, 1995). The EK expresses a number of signalingfactors such as Shh, Bmp2, Bmp4, Bmp7, Fgf4, and thehomeobox transcription factor Msx2 (Mackenzie et al.,1992; Niswander and Martin, 1992; Thesleff andSahlberg, 1996; Thesleff and Jernvall, 1997). This expres-sion resembles the AER and ZPA in the limb bud, noto-chord, floor plate of the spinal cord, and facial processes

A

sr

lee

oeesek

dp

cI

BFigure 7. Schematic representation of the structures in molar toothgerms at the cap (A) and bell (B) stages. Cl, cervical loop; dm,dental mesenchyme; dp, dental papilla; eg, enamel grooves; en,enamel navel; pek, primary enamel knot; sek, secondary enamelknots; iee, inner enamel epithelium (where ameloblasts develop);oee, outer enamel epithelium; sr, stellate reticulum.

(Niswander and Martin, 1992; Fietz et al., 1994; Hogan etal., 1994; Kingsley, 1994; see Thesleff and Sahlberg, 1996;Zhao et al., in press).

In the cap-stage enamel organ, the enamel cord andenamel navel (Figs. 5B, 6A) also express Msx2(Mackenzie et al., 1992) and perhaps help specify theposition of the first buccal cusp, as a reference point forthe later cusps (Ruch et al., 1982; Mackenzie et al., 1992;Butler, 1995). Msx2, the only transcription factor found inthe EK so far (Mackenzie et al., 1992), has been shown tobe regulated by Bmp4 in tooth development (Vainio et al.,1993; Chen et al., 1996; Maas and Bei, 1997). Bmp4 andMsx2 are associated with apoptosis in some contexts(e.g., Marazzi et al., 1997) and may be responsible for thedisappearance of the EKs by this process (Tureckov6 etal., 1995; Vaahtokari et al., 1996b). Dlx genes do notappear to be expressed within EKs (Zhao et al., manu-script in preparation).

3879(41 369 398 119981Oral Biol Med

Crit Rev Oral Biol Med9(4) 369-398 1998)

enecpek

dm

at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

A~peksek

cI

B

CFigure 8. Schematic of tooth morphogenesis to show sources ofpressure (A) and cell proliferation (B,C). B is cap stage; A,C arethe bell stage. Cl, cervical loop; pek, sek, primary and sec-ondary enamel knots. Open arrowheads in A indicate pressureto the enamel epithelia. Filled arrowheads show directions oftissue growth.

(Vl.3) THE MECHANISMS FOR EPITHELIAL FOLDINGBoth secreted and mechanical factors are involved inepithelial folding. Fgf4 is produced by the EK and maypromote mitosis in adjacent IEE and mesenchymal cells,eventually causing mechanical folding of the epithelium(Jernvall et al., 1994; Thesleff et al., 1996; Thesleff andSahlberg, 1996). Proliferation of the IEE bilateral to theEK gives rise to the cap-shaped enamel organ with the

enamel grooves and the cervical loops (Figs. 8B, 8C). Atthe same time, the dental mesenchyme is in a highmitotic state, suggesting that it plays a physical role inepithelial folding (Vaahtokari et al., 1996a; Thesleff andJernvall, 1997). Later, at the bell stage, the primary EKdisappears through apoptosis (Vaahtokari et al., 1996b).With the appearance of the secondary EKs, differentialproliferation in the IEE deepens the valley between theenamel grooves, forming cusps with the downgrowth ofthe cervical loops (Fig. 8C). An increased mitotic index ofthe IEE during the period of epithelial folding was alsodemonstrated by Ruch (1987), who suggested that theactivity was related to cusp formation. Cells within theenamel organ differentiate to secrete hydrophilic gly-cosaminoglycans which accumulate fluid in the intercel-lular spaces, increasing pressure in the interior of theenamel organ, to maintain its shape and volume (Butler,1956).

(VII) Some Final Thoughts about theInvolvement of Bateson-Turing Processes

All of embryogenesis, from the single fertilized eggonward, can be viewed as basically 'self-organizing', andthe concept is specifically useful only to the extent thatit helps identify aspects of development that are freedfrom some point onward from the external specificationof internal structures. The patterns shown earlier in Fig.1 are strikingly like the patterns that can be generated bycomputer simulations of reaction diffusion BTPs, and themorphological dynamics described above support thisnotion impressionistically. The domains of Fgf8, Bmp4, orDlx in the jaws may be induced by such processes.

Early rabbit tooth germs transversely sectioned(bucco-lingually) can develop complete crown patterns(Glasstone, 1963), whereas tooth germs halved longitu-dinally (mesio-distally) did not do so. Proximal first archtissue dissected at E1O.5 from just caudal to the locationof the future MI develop into molars (Lumsden, 1979).These experiments suggest the presence of field symme-tries and asymmetries within tooth germs. For example,the enamel groves and intervening primary EK run longi-tudinally, and transverse sectioning, which does not pre-vent tooth development in culture, probably does notinterrupt any vital initial periodicity in the tooth germ,whereas longitudinal sectioning destroys the EK oressential components of its context. Later, mesio-distalperiodicity arises in the form of three pairs of adjacentcusps (in the mouse lower molars); cultured tooth germssectioned too late do not develop normally.

Two obvious problems for quantitative dynamicmodels relate to the size and time scale of mammaliandentitions within and between species. Related mam-mals with similar dentition (e.g., small and large rodents)have dissimilar adult tooth and body size. But many of

388 Crit Rev Oral Biol Med

9(4).369-398 (1998)388 Crit Rev Oral Biol Med

at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

the mathe-m a t i c a Imodels thatseem plau- Paracone Metaconesible for thistype of pat-terning arevery sensi- Hypocon4tive to para- (from meimetric con-ditions-forexample, in Athe time at } Metaconulewhich the Protoconefirst pattern,or 'prepat-tern', isimposed U Hypocon( M u r ra y, (from pos1993). So 4.Anterior-J (fo pohow could Figure 9. Primordial left upper mammalian mcthe pattern sources of a hypocone added to this form (centEdynam ics top, deer and kangaroo; right middle, human, rbe con sis- sion from Hunter and Jernvall, 1995.tent acrossthe very dis-similar dis-tances involved in teeth in very different-sized animals?

A likely answer is that tooth growth and maturationseem to take place only after crown shape has beenestablished (e.g., lernvall, 1995). The future jaws and theearly tooth germs of mammals are programmed early indevelopment, when the embryos, and their tooth pri-mordia, are all of roughly the same size across species(Luckett, 1993, and personal communication). Similarly,the successional (adult) teeth of most mammals developyears after the tiny primordia are formed; it seems clearthat persistent morphogens consistent across the vastembryo-to-adult size range could not operate. Whatevercauses the crown pattern must be self-contained and,resident' within the primordium.

If field symmetries discussed above suggest a degreeof scale invariance in tooth germ development, a degreeof earlier scale sensitivity of BTPs could be turned to evo-lutionary advantage. Slight differences in the size of thetissue at the pre-patterning time, brought about by sim-ple differences in growth factor regulation but by notooth-specific genetic changes, could be responsible fordental patterning differences among species. This raisesthe issue of developmental stability. Teeth vary, but thatvariation is constrained, especially within genus orspecies If BTPs really are responsible, they must either(1) lay down their patterning information and thenbecome inactive or (2) produce pattern states that arerobust to perturbations of parameters such as the timing

stprotocingulum)

olar form with three cusps (left). Proposed ways in which two commoner) can diversify to form modern upper molar crown morphologies. Rightrhinoceros; right bottom, mammoth. Not to scale. Reprinted with permis-

or concentration of the interacting factors, or geneticmutations affecting their kinetics. This concept ofrobusticity is called by terms like 'attractor' these days;a century ago, by Galton and Bateson, 'positions oforganic stability'; 'archetypes' fifty years earlier by Owenand St Hillaire in the great debate of the 1830s; and 2,500years ago Plato and Aristotle's differing notions of 'ideal'forms. Natural selection, acting over evolutionary timescales, would be responsible for establishing this stablestate (e.g., if mutations leading to stable states did notarise, we would not have teeth). If Felix Savart did attendthe Geoffroy-Cuvier debates, he might have recognizedthe archetype as analogous to the classic basic shape ofstringed instruments-produced and kept in tune by nat-ural selection.

Fig. 9 shows Hunter and Jernvall's (1995) scheme fortransforming simple ancestral molar crown morpholo-gies to the diversity of current forms, by adding a fourthcusp (a hypocone) to the generative process, which hasoccurred many times during mammalian evolution, andmodifying the result. lernvall (1995) described a modelfor the initiation of secondary EKs, which would controlthe development and evolution of cusp patterns.According to him, the primary factor controlling the ini-tiation of EKs is the number of epithelial cells availableto form them. This number might change in a wavelikefashion over time as cell proliferation directed by molec-ular signals from existing EKs increased the number ofepithelial cells above a threshold. This threshold number

9(4 369 398 (1998) Grit Rev Oral Bio1 Med389389Crit Rev Oral Biol Med9(4):369-398 (1998)

at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

of cells might represent that necessary to accumulate aparticular level of a signaling molecule. Formation of anew EK temporarily decreases the number of availablecells, creating a zone of inhibition (of formation of newEKs) surrounding it. The processes involved in thismodel-control of cell proliferation through signalingmolecules, zones of inhibition based on cell availabili-ty-are parameters in what we are terming a BTP.Alteration of these parameters to produce different pat-terns is likely to affect the epithelial cell population as awhole, and therefore more than one cusp. lernvall (pers.comm.) has studied the size relationship among the dis-tinct, linearly arrayed cusps found in seal molars (Fig. 1,lower left). The relative sizes, numbers, and heights ofthe cusps are symmetrically correlated, anterior and pos-terior, to a major central cusp, as if the development ofthese additional cusps depends on rate of growth rela-tive to a threshold in time, beyond which further devel-opment does not occur.

This work supports the idea that molar cusps aregenerated by BTPs. Gene expression differences in timeand space within a molar tooth germ would be found inassociation with these changes, but their presencereflects the process itself, not gene expression specific toa particular cusp or to the tooth identity. If this really isa reaction diffusion process, one would need only toalter the initial conditions of the tooth germ to producedifferent crown morphologies. All could then use the samegenes in a similar way.

(VIII) Considerations for the Designof Studies of Dental Patterning

Mathematical models show us that BTPs are likely, butthey do not tell us what the parameters are in terms ofspecific gene products. To the extent that dental pattern-ing lies in the interactions among factors, the most inter-pretable experiments will involve simultaneous consid-eration of those factors. The Dlx genes show the issuesclearly. Most studies to date have involved one, or atmost two, genes, even when there are six candidates forthe naming or generative system. Even with the com-plexity of dual simultaneous Dlxl/Dlx2 knockouts, it isdebatable what the role of these genes in teeth may be.Only the transition from early epithelial thickening tolater development of the upper molars was affected,despite the normal expression of both genes in multipleother stages of tooth development in all teeth, and thefact that the remaining teeth were unaffected. is thisbecause these genes are not functioning in teeth, orbecause the right combinations have not yet beenknocked out to eliminate redundancy?

The challenge of simultaneous multi-gene studies iseasy to illustrate. Assuming that constitutive knockoutgenotypes would tell the tale we seek, there are 36= 729

possible Dlx- genotypes (+/+, +/-, -/- at each of the sixloci). Three cis and two trans paralog multiple knockoutsshould be done to examine redundancy of function.Similar challenges face the study of several other genefamilies in teeth.

It is perhaps from a combination of the practicalrealities of research, and the impatient confidence ofmodern science, that most work in this area todayinvolves known genes, or models based on only a subsetof a family of equally plausible genes. Effort is still need-ed to find genes with tooth-specific expression (e.g.,cDNA screens and comparisons). Unfortunately, with dif-ferent combinations of the same genes going on and offduring the process, it is not clear at what developmentalstage such work would be most informative or for whatone would look.

Knockouts of candidate genes may be too crude abludgeon to be easily interpretable. Knockout animalstypically generate partial, variable, or syndromic effects.Knockouts are only a partial experiment, in that they maynot adequately elucidate the function of the gene inquestion; for example, background affects the phenotypeof Pax6 mutations (Quinn et al., 1997).

If the same gene is expressed in multiple ways dur-ing dental development, how is one to interpret theresults of such studies? Perhaps it is better to interveneonly at selected times. Antisense experiments in organculture to inhibit a specific gene, or the application ofexogenous or ectopic gene product, are not always easyto do but can be informative. Similarly useful would bethe development of context-specific quantitative over-and under-expression experiments in transgenic ani-mals. To do the latter, we need to identify appropriatecontext-specific enhancers.

It might seem curious to question whether thereneed ever be a difference in the genes expressed in dif-ferent types of teeth, and direct cDNA comparisonshould be able to detect such a difference. That differ-ence might persist, since successional tooth germsremember what to be for many years after birth. If self-organizing BTPs are responsible for crown patterning,tooth type could be wholly determined by the initial con-ditions; these could have to do only with levels of mor-phogens, size or shape of the clone, etc., and with no, oronly very fleeting, genetic differences, or with the ratiosof gene expression. For example, Fgf4 is expressed inEKs; finding Fgf4 expressed in the first cusp of an E15molar germ but not elsewhere in the tooth does notmean that the gene is a cusp-specific code. Whether asingle quantitative process could be responsible for theentire dentition is unclear, given the evolutionary inde-pendence of the different tooth classes. For example,even if a single process were not responsible for toothtype, aspects like shape differences and size gradientswithin class may be due to such mechanisms.

390 Crit Rev Oral Biol Med(1998)

390 Crit Rev Oral Biol Med 9(4):369-398 (1998) at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

Essentially, at this stage, we simply do not know, and thedifficulties of this challenge are clear (Ruch, 1995).

ConclusionThe basic patterning questions are: What determines thenumber, location, and morphological differences amongteeth? The work reviewed here shows that there has beenconsiderable progress in our understanding of the genet-ic basis of these phenomena, although we know muchmore about specific gene induction pathways than we doabout ultimate patterning mechanisms. Specifically, westill do not know whether different tooth types requiredifferent genes or just different parameters of self-con-tained generating processes. We have noted similaritiesin the genes expressed, and perhaps even the geneticregulatory pathways, in teeth at different times andplaces as well as in other developing organs like thelimb. This raises a question as to how similar sets ofgenes can be expressed in different tissues and yet con-trol tissue specificity. The likely explanation is that thesets of factors expressed in each context are not com-pletely overlapping. Additional factors that have not yetbeen discovered are probably also involved.

These considerations and the similarity of mostknown natural or experimental effects on teeth (that onlysome of a given tooth type are affected, or only the mat-uration) raise the question whether we have failed toidentify any of the genes involved in primary dental pat-terning. Is this possible? One way it could be so is if weare looking for specific genes rather than processes-amajor point of this paper.

We have discussed dental patterning in comparativeand evolutionary terms. Even to identify such a thing asthe 'general' problem implies some assumption(s) aboutthe unity of life. Clearly, we rely on evolution for its para-mount concept, common ancestry. Comparative and evolu-tionary approaches to the genetics of a trait like the den-tition are useful mainly by extending this notion to homol-ogy, which ever since Darwin has usually implied sharedancestry, despite some long-standing objections(Bateson, 1894; Hall, 1992; Webster and Goodwin, 1996).

Without the general idea of homology, we would beless interested in dental patterning-heterodonty, forexample-in other species as models for the same typeof patterning in humans. Why else would there be inter-est in the mouse diastema region or its 'missing' caninesor premolars? Conceptually not far from the essentialismof past centuries (or millennia), we assume that themouse is a representation of a plan that in some sensescontains, or at least contained, these phantom struc-tures (or perhaps not-so-phantom: Tureckova et al., 1995;Ruch et al., 1997). We look at the mouse to identify geneswe assume have similar function to (assumed) earlierstates such as fish teeth and scales and to (assumed)

more directly homologous traits like human teeth. It istacitly assumed that there is some genetic differencebetween molars and incisors, and that this difference isshared and thus homologous among heterodontspecies. If we are wrong, much research effort will bewasted.

These issues have direct biomedical relevance. In prin-ciple, the greatest advance in the history of dentistry wouldoccur if natural regenerative therapy could replace pros-thetic restorative implants for damaged teeth. Progress isclearly being made toward our improved understanding ofthe inductive pathways that produce competent odonto-genic cells. But if the shape of a tooth is determined by BTPsas described above, it may not be sufficient to paint thecleaned surface of a decayed tooth with (say) competentpre-ameloblasts. The temporal and three-dimensionaldynamics of the processes needed to generate useful occlu-sion may require very specific, perhaps non-uniform, initialconditions on such a surface. A mass of new but unshapedenamel will not be of much use.

A considerable amount is now known about genesexpressed in teeth and gene-gene induction andresponse pathways. Much less is known about how thesegenerate initial pattern and morphology. We do not knowwhat structures are specified by individual genes orgenetic codes, and what are strictly process-dependent,and little experimentation has been done on the latter.Questions of this type are among the most challenging inall of biology. Teeth are as important as ever as traits onwhich to work on those questions, and the working mate-rial for such studies is rapidly developing.

AcknowledgmentsOur work on dental patterning has been supported by the NationalScience Foundation (Physical Anthropology Progracm, SBR9408402), and the National Institutes of Health (NIDR, DE 10871 ),and includes collaboration with Frank H Ruddle and CooduvaliShashikant at Yale University, and Anne Buchanan at Penn State.

REFERENCESAkimenko MA, Ekker M, Wegner J, Lin W, Westerfield M

(1994). Combinatorial expression of three zebrafishgenes related to distal-less: part of a homeobox genecode for the head. I Neurosci 14:3475-3486.

Bateson W (1894). Materials for the study of variation,treated with special regard to discontinuity in the ori-gin of species. London: Macmillan.

Bateson W (1913). Problems of genetics. New Haven, CT.Yale University Press.

Bitgood MJ, McMahon AP (1995). Hedgehog and bmp genesare coexpressed at many diverse sites of cell-cellinteraction in the mouse embryo. Dev Biol 172:126-138.

3919(41369981998Cri Re Ora Bio Me

Crit Rev Oral Biol Med9(4)-369-398 (1998) at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

Burke A, Nelson C, Morgan B, Tabin C (1995). Hox genesand the evolution of vertebrate axial morphology.Development 121:333-346.

Butler P (1939). Studies of the mammalian dentition.Differentiation of the post-canine dentition. Proc ZoolSoc Lond 109:1-36.

Butler PM (1956). The ontogeny of molar pattern. Biol Rev31:30-71.

Butler PM (1978). The ontogeny of mammalian het-erodonty. J Biol Buccale 6:21 7-227.

Butler PM (1995). Ontogenetic aspects of dental evolu-tion. lnt I Dev Biol 39:25-34.

Carroll RL (1988). Vertebrate paleontology and evolution.New York: W.H. Freeman.

Chen Y, Bei M, Woo 1, Satokata I, Maas R (1996). Msxlcontrols inductive signaling in mammalian toothmorphogenesis. Development 122:3035-3044.

Chiang C, Litingtung Y, Lee E, Young KE, Corden JL,Westphal H, et al. (1996). Cyclopia and defective axialpatterning in mice lacking Sonic hedgehog gene func-tion. Nature 383:407-413.

Conlon RA (1995). Retinoic acid and pattern formation invertebrates. Trends Genet 11:314-319.

Couly GF, Le Douarin NM (1985). Mapping of the earlyneural primordium in quail-chick chimeras. I.Developmental relations between placodes, facialectoderm, and prosencephalon. Dev Biol 110:422-439.

Couly GF, Le Douarin NM (1987). Mapping of the earlyneural primordium in quail-chick chimeras. II. Theprosencephalic neural plate and neural folds: impli-cations for the genesis of cephalic human congenitalabnormalities. Dev Biol 120:198-214.

Couly G, Le Douarin NM (1990). Head morphogenesis inembryonic avian chimeras: evidence for a segmentalpattern in the ectoderm corresponding to the neu-romeres. Development 108:543-558.

Crossley PH, Martin GR (1995). The mouse Fgf8 geneencodes a family of polypeptides and is expressed inregions that direct outgrowth and patterning in thedeveloping embryo. Development 121:439-451.

Crossley PH, Martinez S, Martin GR (1996). Midbraindevelopment induced by FGF8 in the chick embryo.Nature 380:66-68.

Cummings E, Bringas P, Grodin M, Slavkin H (1981).Epithelial-directed mesenchyme differentiation invitro. Differentiation 20:1-9.

Deuschle FM, Geiger JF, Warkany J (1959). Analysis of ananomalous oculodentofacial pattern in newborn rats bymaternal hypervitaminosis. Am J Dent Res 18:149-155.

Dietrich S, Gruss P (1995). undulated phenotypes suggesta role of Pax-l for the development of vertebral andextravertebral structures. Dev Biol 167:529-548.

Dolle P, Ruberte E, Leroy P, Morriss-Kay G, Chambon P(1990). Retinoic acid receptors and cellular retinoidbinding proteins. l. A systematic study of their differ-

ential pattern of transcription during mouse organo-genesis. Development 1 10: 1133-1151.

Dolle P, Price M, Duboule D (1992). Expression of themurine Dlx-1 homeobox gene during facial, ocularand limb development. Differentiation 49:93-99.

Duprez DM, Kostakopoulou K, Francis-West PH, Tickle C,Brickell PM (1996). Activation of Fgf-4 and HoxD geneexpression by BMP-2 expressing cells in the develop-ing chick limb. Development 122:1821-1828.

Ekker SC, Ungar AR, Greenstein P, von Kessler DP, PorterIA, Moon RT, et al. (1995). Patterning activities of ver-tebrate hedgehog proteins in the developing eye andbrain. Curr Biol 5:944-955.

Favor J, Sandulache R, Neuhauser-Klaus A, Pretsch W,Chatterjee B, Senft E, et al. (I1996). The mouse Pax2INeUmutation is identical to a human PAX2 mutation in a

family with renal-coloboma syndrome and results indevelopmental defects of the brain, ear, eye, and kid-ney. Proc Natl Acad Sci USA 93:13870-13875.

Ferguson CA, Tucker AS, Christensen L, Lau AL, MatzukMM, Sharpe PT (1998). Activin is an essential earlymesenchymal signal in tooth development that isrequired for patterning of the murine dentition. GenesDev 12:2636-2649.

Fietz MJ, Concordet J-P, Barbosa R, Johnson R, Krauss S,McMahon AP, et al. (1994). The hedgehog gene family inDrosophila and vertebrate development. Development(Suppl):43-5 1.

Francis-West PH, Tatla T, Brickell PM (1994). Expressionpatterns of the bone morphogenetic protein genesBmp-4 and Bmp-2 in the developing chick face suggesta role in the outgrowth of the primordia. Dev Dynam201:168-178.

Gavin BJ, McMahon JA, McMahon AP (1990). Expression ofmultiple novel Wnt-1/int-l-related genes during fetaland adult mouse development. Genes Dev 4:2319-2332.

Gilbert S, Opitz J, Raff R (1996). Resynthesizing evolution-ary and developmental biology. Dev Biol 173:357-372.

Glasstone S (1936). The development of tooth germs invitro. Anat Lond 70:260-266.

Glasstone S (1963). Regulative changes in tooth germsgrown in tissue culture. I Dent Res 42:1364-1368.

Glaser T, Jepeal L, Edwards JG, Young SR, Favor J, MaasRL (1994). PAX6 gene dosage effect in a family withcongenital cataracts, aniridia, anophthalmia and cen-tral nervous system defects. Nature Genet 7:463-471.

Graveson AC, Smith MM, Hall BK (1997). Neural crestpotential for tooth development in a urodele amphib-ian: developmental and evolutionary significance. DevBiol 188:34-42.

Grigoriou M, Tucker AS, Sharpe PT, Pachnis V (1998).Expression and regulation of Lhx6 and Lhx7, a novelsubfamily of LIM homeodomain encoding genes, sug-gests a role in mammalian head development.Development 125:2063-2074.

392 Crit Rev Oral Biol

9(4):369-398 (1998)392 Crit Rev Oral Biol Med

at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

Grindley IC, Davidson DR, Hill RE (1995). The role of Pax-6 ineye and nasal development. Development 12 1:1433-1442.

Hall BK (1992). Evolutionary developmental biology.London, UK: Chapman & Hall.

Hardcastle Z, Mo R, Hui C-C, Sharpe PT (1998). The Shhsignalling pathway in tooth development: defects inGli2 and Gli3 mutants. Development 125:2803-2811.

Hata R, Slavkin H (1978). De novo induction of a geneproduct during heterologous epithelial-mesenchymalinteractions in vitro. Proc Natl Acad Sci USA 75:2790-2794.

Helms J, Kim C, Hu D, Minkoff R, Thaller C, Eichele G(1997). Sonic hedgehog participates in craniofacial mor-phogenesis and is down-regulated by teratogenicdoses of retinoic acid. Dev Biol 187:25-35.

Hershkovitz P (1971). Basic crown patterns and cusphomologies of mammalian teeth. In: Dental morphol-ogy and evolution. Dahlberg AA, editor. Chicago, IL:University of Chicago Press.

Hillson S (1986). Teeth. Cambridge, UK: CambridgeUniversity Press.

Hogan BLM, Blessing M, Winnier GE, Suzuki N, Jones CM(1 994). Growth factors in development: the role ofTGF-f related polypeptide signalling molecules inembryogenesis. Development (Suppl):53-60.

Huggins C, McCarroll H, Dahlberg A (1934).Transplantation of tooth germ elements and theexperimental heterotypic formation of dentin andenamel. J Exp Med 60199-210.

Hunt P, Krumlauf R (1992). HOX codes and positionalspecification in vertebrate embryonic axes. Ann RevCell Biol 8 227-256.

Hunter J, Jernvall 1 (1995). The hypocone as a key innova-tion in mammalian evolution. Proc Natl Acad Sci USA92:10718- 10722.

Huyssene A, Sire l-Y (1998). Evolution of patterns andprocesses in teeth and tooth-related tissues in non-mammalian vertebrates. Eur J Oral Sci 106(Suppl1)437-481.

Imai H, Osumi-Yamashita N, Ninomiya Y, Eto K (1996).Contribution of early-emigrating midbrain crest cellsto the dental mesenchyme of mandibular molar teethin rat embryos. Dev Biol 1 76:151-165.

Janvier P (1995). Conodonts join the club. Nature 374:761-762.

Iernvall 1(1995). Mammalian molar cusp patterns: devel-opmental mechanisms of diversity. Acta Zool Fennica198:1 -61.

lones F, Kioussi C, Copertino D, Kallunki P, Holst B,Edelman G (1997). Barx2, a new homeobox gene ofthe Bar class, is expressed in neural and craniofacialstructures during development. Proc Natl Acad Sci USA94 2632-2637.

loyner A (1996). Engrailed, Wnt, and Pax genes regulatemidbrain-hindbrain development. Trends Genet 12:15-20.

lung H, Francis-West P, Widelitz R, Jiang T, Ting-Berreth S,Tickle C, et al. (1998). Local inhibitory action of BMPsand their relationships with activators in feather for-mation: implications for periodic patterning. Dev Biol196:11-23.

Kalter H, Warkany J (1961). Experimental production ofcongenital malformations in strains of inbred mice bymaternal treatment with hypervitaminosis. Am J Pathol38:1-21.

Kantaputra PN, Gorlin RI (1992). Double dens invagina-tus of molarized maxillary central incisors, premolar-ization of maxillary lateral incisors, multituberculismof the mandibular incisors, canines and first premo-lars, and sensorineural hearing loss. Clin Dysmorphol1:128-136.

Kauffman MH, Chang H-H, Shaw IP (1995). Craniofacialabnormalities in homozygous Small eye (Sey/Sey).embryos and newborn mice. I Anat 186:607-617.

Kingsley DM (1994). The TGF-4 superfamily: new mem-bers, new receptors, and new genetic tests of functionin different organisms. Genes Dev 8:133-146.

Kirino T, Nozue T, Inoue M (1973). Deficiency of enamelknot in experimental morphology. Okaj Fol Anat50:117-132.

Knudsen PA (1965). Fusion of upper incisors at bud stageor cap stage in mouse embryos with exencephalyinduced by hypervitaminosis. Acta Odontol Scand23:549-565.

Knudsen PA (1967). Congenital malformations of lowerincisors and molars in exencephalic mouse embryos,induced by hypervitaminosis. Acta Odontol Scand25:669-691.

Kollar E (1997). Odontogenesis: a retrospective. Eur IOral Sci 106(Suppl 1):2-6.

Kollar E, Baird G (1968). The influence of the dentalpapilla on the development of tooth shape in embry-onic mouse tooth germs. J Embryol Exp Morphol 21:131 -

148.Kollar E, Baird G (1970). Tissue interactions in embryon-

ic mouse tooth germs. 11. The inductive role of thedental papilla. I Embryol Exp Morphol 24:173-186.

Kollar E, Mina M (1987). The induction of odontogenesisin non-dental mesenchyme combined with earlymurine mandibular arch epithelium. Arch Oral Biol32:123-127.

Kollar E, Mina M (1991). Role of the early epithelium inthe patterning of the teeth and Meckel's cartilage. JCraniofac Genet Dev Biol 11:223-228.

Kontges G, Lumsden A (1996). Rhombencephalic neuralcrest segmentation is preserved throughout craniofa-cial ontogeny. Development 122:3229-3242.

Koyama E, Yamaai T, Iseki S, Ohuchi H, Nohno T,Yoshioka H, et al. (1996). Polarizing activity, Sonic hedge-hog, and tooth development in embryonic and post-natal mouse. Dev Dynam 206:59-72.

9(4)369981998 Grt Re Ora Bil Me 39

393Crit Rev Oral Biol Med9(4) 369-398 (1998) at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

Kratochwil K, Dull M, Farinas 1, Galceran J, Grosschedl R(1996). Lef 1 expression is activated by BMP-4 and reg-ulates inductive tissue interactions in tooth and hairdevelopment. Genes Dev 10:1382-1394.

Kronmiller JE, Beeman CS (1994). Spatial distribution ofendogenous retinoids in the murine embryonicmandible. Arch Oral Biol 39:1071-1078.

Kronmiller JE, Upholt WB, Kollar E (1992). Alteration ofmurine odontogenic patterning and prolongation ofexpression of epidermal growth factor mRNA byretinol in vitro. Arch Oral Biol 37:129-138.

Kronmiller JE, Nguyen T, Berndt W (1995a). Instruction byretinoic acid of incisor morphology in the mouseembryonic mandible. Arch Oral Biol 40:589-595.

Kronmiller JE, Nguyen T, Berndt W, Wickson A (1995b).Spatial and temporal distribution of Sonic hedgehogmRNA in the embryonic mouse mandible by reversetranscription/polymerase chain reaction and in situhybridization analysis. Arch Oral Biol 40:831-838.

Krumlauf R (1994). Hox genes in vertebrate development.Cell 78:191-201.

Kulesa P, Cruywagen G, Lubkin S, Maini P, Sneyd J,Ferguson M, et al. (1996). On a model mechanism forthe spatial patterning of teeth primordia in the alliga-tor. J Theoret Biol 180:287-296.

Lesot H, Vonesch J-L, Peterka M, Tureckova J, PeterkovaR, Ruch J-V (1996). Mouse molar morphogenesisrevisited by three-dimensional reconstruction. 11.Spatial distribution of mitoses and apoptosis in capto bell staged first and second upper molar teeth. IntJ Dev Biol 40:1017-1032.

Lewin D (1997). Evolutions: mammalian tooth develop-ment. J NIH Res 9:75-80.

Li H-S, Yang J-M, Jacobson RD, Pasko D, Sundin 0 (1994).Pax-6 is first expressed in a region of ectoderm ante-rior to the early neural plate: implications for step-wise determination of the lens. Dev Biol 162:181-194.

Li H-S, Tierney C, Wen L, Wu JY, Rao Y (1997). A singlemorphogenetic field gives rise to two retina primordiaunder the influence of the prechordal plate.Development 1 24:603-615.

Luckett WP (1993). An ontogenetic assessment of dentalhomologies in therian mammals. In: Mammal phy-logeny: mesozoic differentiation, multituberculates,monotremes, early therians, and marsupials. SzalayFS, Novacek MJ, McKenna MC, editors. New York:Springer-Verlag, pp. 182-204.

Lumsden AGS (1979). Pattern formation in the molardentition of the mouse. J Biol Buccale 7:77-103.

Lumsden AGS (1982). The developing innervation of thelower jaw and its relation to the formation of toothgerms in mouse embryos. In Teeth. Kurten B, editor.New York: Columbia University Press, pp. 32-43.

Lumsden AGS (1988). Spatial organization of the epitheli-um and the role of neural crest cells in the initiation of

the mammalian tooth germ. Development103(Suppl): 155-169.

Lumsden AGS, Buchanan JAG (1986). An experimentalstudy of timing and topography of early tooth devel-opment in the mouse embryo with an analysis of therole of innervation. Arch Oral Biol 31:301-311.

Lumsden A, Krumlauf R (1996). Patterning the vertebrateneuraxis. Science 274:1109-1115.

Maas R, Bei M (1997). The genetic control of early toothdevelopment. Crit Rev Oral Biol Med 8:4-39.

MacDonald R, Barth KA, Xu Q, Holder N, Mikkola 1,Wilson SW (1995). Midline signalling is required forPax gene regulation and patterning of the eyes.Development 121:3267-3278.

Mackenzie A, Leeming G, Jowett A, Ferguson M, Sharpe P(1991). The homeobox gene Hox 7.1 has specificregional and temporal expression patterns duringealy murine craniofacial embryogenesis, especiallytooth development in vivo and in vitro. Development111:269-285.

Mackenzie A, Ferguson M, Sharpe P (1992). Expressionpatterns of the homeobox gene, Hox-8, in the mouseembryo suggest a role in specifying tooth initiationand shape. Development 11 5:403-420.

Maconochie M, Nonchev S, Morrison A, Krumlauf R(1996). Paralogous Hox genes: function and regula-tion. Ann Rev Genet 30:529-556.

Mangelsdorf DJ, Umesono K, Evans RM (1994). Theretinoid receptors. In: The retinoids: biology, chem-istry, and medicine. Sporn MB, Roberts AB, GoodmanDS, editors. New York: Raven Press, pp. 319-349.

Mansouri A, Hallonet M, Gruss P (1996). Pax genes andtheir roles in cell differentiation and development.Curr Opin Cell Biol 8:851-857.

Marazzi G, Wang Y, Sassoon D (1997). Msx2 is a tran-scriptional regulator in the BMPr-mediated pro-grammed cell death pathway. Dev Biol 186:127-138.

Mark M, Bloch-Zupan A, Ruch I-V (1992). Effects ofretinoids on tooth development and cytodifferentia-tions in vitro. Int J Dev Biol 36:517-526.

Mark M, Lohnes D, Mendelsohn C, Dupe V, Vonesch J-L,Kastner P, et al. (1995). Roles of retinoic acid receptorsand of Hox genes in the patterning of the teeth and ofthe jaw skeleton. Int I Dev Biol 39:111 -121.

Marti E, Takada R, Bumcrot DA, Sasaki H, McMahon AP(1995). Distribution of Sonic hedgehog peptides in thedeveloping chick and mouse embryo. Development121:2537-2547.

McGuinness T, Porteus M, Smiga S, Bulfone A, KingsleyC, Oiu M, et al. (1996). Sequence, organization, andtranscription of the Dlx-I and Dlx-2 locus. Genomics35:473-485.

Meinhardt H (1995). The algorithmic beauty of seashells.Berlin, Germany: Springer-Verlag.

Meinhardt H (1996). Models of biological pattern forma-

394 Crit Rev Oral Biol Med (1998)394 Crit Rev Oral Biol Med 9(4):369-398 (1998) at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

tion: common mechanisms in plant and animaldevelopment. Int J Dev Biol 40:123-134.

Miles AEW, Grigson C (1990). Colyer's variations and dis-eases of the teeth of animals. Cambridge, UK:Cambridge University Press.

Miller W (1969). Inductive changes in early tooth develop-ment: 1. A study of mouse tooth development on thechick chorioallantois. J Dent Res 48(Suppl):719-725.

Mucchielli M-L, Mitsiadis T, Raffo S, Brunet I-F, Proust J-P, Goridis C (1997). Mouse Otlx2/RIEG expression inthe odontogenic epithelium precedes tooth initiationand requires mesenchyme-derived signals for itsmaintenance. Dev Biol 189:279-284.

Murray (1993). Mathematical biology. 2nd ed. New York:Springer-Verlag.

Nakamura S, Stock DW, Wydner KL, Bollekens JA,Takeshita K, Nagai BM, et al. (1996). Genomic analysisof a new mammalian distal-less gene: Dlx7. Genomics38:3 14-324.

Nelson CE Morgan BA, Burke AC, Laufer E, DiMambro E,Murtaugh LC, et al. (1996). Analysis of Hox geneexpression in the chick limb bud. Development122: 1449-1466.

Neubuser A, Koseki H, Balling R (1995). Characterizationand developmental express'on of Pax9, a paired-box-containing gene related to Paxl. Dev Biol 170:701-716.

Neubuser A, Peters H, Balling R, Martin G (1997).Antagonistic interactions between FGF and BMP sig-naling pathways: a mechanism for positioning thesites of tooth formation. Cell 90:247-256.

Nijhout HF (1991). The development and evolution ofbutterfly wing patterns. Washington, DC:Smithsonian Institution Press.

Niswander L, Martin GR (1992). Fgf-4 expression duringgastrulation, myogenesis, limb and tooth develop-ment in the mouse. Development 114:755-768.

Noden D (1991). Vertebrate craniofacial development:the relation between ontogenetic process and mor-phological outcome. Brain Behav Evolut 38:190-225.

O'Farrell P 11994). Unanimity waits in the wings. Nature368 188- 189.

Osborn HF (1897). Trituberculy: a review dedicated to thelate Professor Cope. Am Nat XXXI:993-1016.

Osborn 1W (1973). The evolution of dentitions. Am Sci61:548-559.

Osborn 1W (1978). Morphogenetic gradients: fields ver-sus clones. In: Development, function and evolutionof teeth. Butler PM, loysey KA, editors. New York:Academic Press, pp. 171-201.

Osborn JW (1993). A model simulating tooth morphogen-esis without morphogens. I Theoret Biol 165:429-455.

Osumi-Yamashita N, Ninomiya Y, Doi H, Eto K (1994).The con-tribution of both forebrain and midbrain crestcells to the mesenchyme in the frontonasal mass ofmouse embryos. Dev Biol 164:409-419.

Osumi-Yamashita N, Ninomiya Y, Eto K (1997).Mammalian craniofacial embryology in vitro lnt I DevBiol 41:187-194.

Peterkova R, Lesot H, Vonesch J-L, Peterka M, Ruch J-V(1996). Mouse molar morphogenesis revisited bythree-dimensional reconstruction. 1. Analysis of ini-tial stages of the first upper molar developmentrevealed two transient buds. Int Dev Biol 40:1009-1016.

Peters H, Neubuser A, Balling R (1998). Pax genes andorganogenesis: Pax9 meets tooth development. Eur IOral Sci 106(Suppl 1):38-43.

Petkovich M (1992). Regulation of gene expression byvitamin A: the role of nuclear retinoic acid receptors.Ann Rev Nutr 12:443-471.

Peyer B (1968). Comparative odontology. Chicago, IL:The University of Chicago Press.

Porteus M, Bulfone A, Ciaranello R, Rubenstein (1991).Isolation and characterization of a novel cDNA cloneencoding a homeodomain that is developmentally inthe ventral forebrain. Neuron 7:221 -229.

Prince V, Lumsden A (1994. Hoxa-2 expression in normaland transposed rhombomeres: independent regula-tion in the neural tube and neural crest. Development120:911-923.

Qiu M, Bulfone A, Martinez S, Meneses JI, Shimamura K,Pedersen RA, et al. (1995). Null mutation of Dlx-2results in abnormal morphogenesis of proximal firstand second branchial arch derivatives and abnormaldifferentiation in the forebrain. Genes Dev 9:2523-2538.

Qiu M, Bulfone A, Ghattas 1, Meneses J, Christensen L,Sharpe P, et al. (1997). Role of the Dlx homeoboxgenes in proximodistal patterning of the branchialarches: mutations of Dlx-l, Dlx-2 and Dlx-1 and -2alter morphogenesis of proximal skeletal and soft tis-sue structures derived from the first and second arch-es. Dev Biol 185:165-184.

Quinn J, West J, Kauffman M (1997). Genetic backgroundeffects on dental and other craniofacial abnormalitiesin homozygous small eye (Pax6seY/Pax6seY) mice. AnatEmbryol 196:311 -321.

Rawles M (1963). Tissue interactions in scale and featherdevelopment as studied in dermal-epidermal recom-binations. J Em6ryol Exp Morphol 11:765-789.

Richman IM, Tickle C (1992). Epithelial-mesenchymalinteractions in the outgrowth of limb buds and facialprimordia in chick embryos. Dev Biol 154:299-308.

Robinson GW, Mahon KA (1994). Differential and over-

lapping expression of Dlx-2 and Dlx-3 suggest distinctroles for distal-less homeobox genes in craniofacialdevelopment. Mech Dev 48:199-215.

Robinson G, Wray S, Mahon K (1991). Spatially restrictedexpression of a member of a new family of murine dis-tal-less homeobox genes in the developing forebrain.New Biol 3:1183-1194.

9(4)3693%(998 Cri Re Ora Blt Me 39395Crit Rev Oral Biol Med9(4)-369-398 (1998)

at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

Roelink H, Nusse R (1991). Expression of two membersof the Wnt family during mouse development-restrict-ed temporal and spatial patterns in the developingneural tube. Genes Dev 5:381-388.

Roessler E, Belloni E, Gaudenz K, Jay P, Berta P, SchererSW, et al. (I1996). Mutations in the human Sonic Hedgehoggene cause holoprosencephaly. Nature Genet 14:357-360.

Ruberte E, Dolle P, Krust A, Zelent A, Morriss-Kay G,Chambon P (1990). Specific spatial and temporal dis-tribution of retinoic acid receptor gamma transcriptsduring mouse embryogenesis. Development 108:213-222.

Ruberte E, Dolle P, Chambon P, Morriss-Kay G (1991).Retinoic acid receptors and cellular retinoid bindingproteins. II. Their differential pattern of transcriptionduring early morphogenesis in mouse embryos.Development I 1 1:45-60.

Ruberte E, Freiderich V, Morriss-Kay G, Chambon P(1992). Differential distribution patterns of CRABP Iand CRABP 11 transcripts during mouse embryogene-sis. Development 115:973-987.

Ruch JV (1987). Determinisms of odontogenesis. Cell BiolRev 14:1-81.

Ruch IV (1995). Tooth crown morphogenesis and cytodif-ferentiations: Candid questions and critical com-ments. Connect Tissue Res 32:1-8.

Ruch JV, Lesot H, Karcher-Djuricic V, Meyer JM, Olive M(1982). Facts and hypotheses concerning the controlof odontoblast differentiation. Differentiation 21:7-12.

Ruch JV, Lesot H, Peterkova R, Peterka M (1997). Evolvingrodent dentition (letter). BioEssays 19:1041.

Ruddle FH, Bentley KL, Murtha MT, Risch N (1994). Geneloss and gain in the evolution of the vertebrates.Development (Suppl):155-161.

Sager B (1996). Propagation of traveling waves inexcitable media. Genes Dev 10:2237-2250.

Sanyanusin P, Schimmenti LA, McNoe LA, Ward TA,Pierpont MEM, Sullivan MI, et al. (1995). Mutation ofthe PAX2 gene in a family with optic nerve colobo-mas, renal anomalies and vesicoureteral reflux. NatureGenet 9:358-363.

Satokata 1, Maas R (1994). Msxl deficient mice exhibitcleft palate and abnormalities of craniofacial andtooth development. Nature Genet 6:348-356.

Schedl A, Ross A, Lee M, Engelkamp D, Rashbass P, vanHeyningen V, et al. (1996). Influence of PAX6 genedosage on development: overexpression causessevere eye abnormalities. Cell 86:71-82.

Schour 1 (1962). Oral histology and embryology. 8th ed.Philadelphia, PA: Lea & Febiger.

Seifert R, Jacob M, Jacob HJ (1993). The avian prechordalhead region: a morphological study. J Anat 183:75-89.

Semina E, Reiter R, Leysens N, Alward W, Small K,Datson N, et al. (I1996). Cloning and characterization ofa novel bicoid-related homeobox transcription factor

gene, RIEG, involved in Rieger syndrome. Nature Genet14:392-399.

Serbedzija GN, Bronner-Fraser M, Fraser SE (1992). Vitaldye analysis of cranial neural crest cell migration inthe mouse embryo. Development 116:297-307.

Sharkey M, Graba Y, Scott M (1997). Hox genes in evolu-tion: protein surfaces and paralog groups. Trends Genet13:145-151.

Sharpe PT (1995). Homeobox genes and orofacial devel-opment. Connect Tissue Res 32:17-25.

Shimamura K, Hartigan DJ, Martinez S, Puelles L,Rubenstein JLR (1995). Longitudinal organization ofthe anterior neural plate and neural tube. Development121:3923-3933.

Shubin N, Tabin C, Carroll S (1997). Fossils, genes andthe evolution of animal limbs. Nature 388:639-648.

Slavkin H (1974). Embryonic tooth formation. A tool fordevelopmental biology. Oral Sci Rev 4:1-4.

Slavkin H (1982). Experimental dissection of avian andmurine tissue interactions using organ culture in aserumless medium free from exogenous (nondefined)factors. Prog Clin Biol Res 101:217-228.

Slavkin H, Bavetta L (1968). Organogenesis: prolongeddifferentiation and growth of tooth primordia on thechick chorio-allantois. Experientia 24:192-204.

Slavkin H, Diekwisch T (1996). Evolution in tooth devel-opmental biology: of morphology and molecules.Anat Rec 245:131-150.

Slavkin HC, Beierle J, Bavetta CA (1968). Odontogenesis:cell-cell interaction in vitro. Nature 2 17:269-270.

Smith MM, Coates MI (1998). Evolutionary origins of thevertebrate dentition: phylogenetic patterns anddevelopmental evolution. Eur J Oral Sci 106(Suppl1):482-500.

Smith MM, Hall BK (1993). A developmental model forevolution of the vertebrate exoskeleton and teeth:The role of cranial and trunk neural crest. Evol Biol27:387-448.

Smith MP, Sansom II, Smith MP (1996). 'Teeth' beforearmour: the earliest vertebrate mineralized tissues.Modern Geology 20:303-3 19.

Snead M, Luo W, Lau EC, Slavkin HC (1988). Spatial- andtemporal-restricted pattern for amelogenin geneexpression during mouse molar tooth organogenesis.Development 104:77-85.

Stock DW, Ellies D, Zhao Z, Ekker M, Ruddle FH, WeissKM (1996a). The evolution of the Dlx gene family. ProcNatl Acad Sci USA 93:10858-10863.

Stock DW, Buchanan AV, Zhao Z, Weiss KM (1996b).Numerous members of the Sox family of HMG box-containing genes are expressed in mouse teeth.Genomics 37:234-237.

Stock DW, Weiss KM, Zhao Z (1997). Patterning of themammalian dentition in development and evolution.Bioessays 19:481-490.

396 Grit Rev Oral Biol Med 9(4):369 398 (1998)396 Crit Rev Oral Biol Med 9(4):369-398 (1998) at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

Summerbell D, Lewis IH, Wolpert L (1973). Positionalinformation in chick limb morphogenesis. Nature244:492-496.

Tabin C (1991). Retinoids, homeoboxes, and growth fac-tors: toward molecular models for limb development.Cell 66:199-217.

Ten Cate AR (1995). The experimental investigation ofodontogenesis. Int I Dev Biol 395-1 1.

Thesleff (1996). Two genes for missing teeth. NatureGenet 13:379-380.

Thesleff 1, Nieminen P (1996). Tooth morphogenesis andcell differentiation. Curr Opin Cell Biol 8:844-850.

Thesleff 1, Sahlberg C (1996). Growth factors as inductivesignals regulating tooth morphogenesis. Cell Dev Biol7:185-193.

Thesleff 1, Sharpe P (1997). Signalling networks regulat-ing dental development. Mech Dev 67:111-123.

Thesleff 1, Vaahtokari A, Vainio S, lowett A (1996).Molecular mechanisms of cell and tissue interactionsduring early tooth development. Anat Rec 245:151-161.

Thomas BL, Sharpe, PT (1998). Patterning of the murinedentition by homeobox genes. Eur I Oral Sci106(Suppl 1):48-54.

Thomas BL, Porteus MH, Rubinstein ILR, Sharpe PT(1995). The spatial localization of Dlx-2 during toothdevelopment. Connect Tissue Res 32:27-34.

Thomas BL, Tucker AS, Qui M, Ferguson CA, HardcastleZ, Rubenstein IL, et al. (1997). Role of Dlx-1 and Dlx-2 genes in patterning of the murine dentition.Development 124:4811-4818.

Thomas B Tucker A, Ferguson C, Qiu M, Rubenstein J,Sharpe P (1998). Molecular control of odontogenicpatterning: positional dependent initiation and mor-phogenesis. Eur J Oral Sci 106(Suppl 1):44-47.

Tickle C (1995). Vertebrate limb development. Curr OpinGenet Dev 5:478-484.

Tissier-Seta J-P, Mucchielli M-L, Mark M, Mattei M-G,Goridis C, Brunet I-F (1995). Barxl, a new mousehomeodomain transcription factor expressed incranio-facial ectomesenchyme and the stomach. MechDev 51:3-15.

Toothbase (1997). An online website for data on gene expres-sion in teeth, at http://honeybee.helsinki.fi/toothexp.Administered by the laboratory of Dr I. Thesleff.

Torres M, G6mez-Pardo E, Dressler GR, Gruss P (1995).Pax-2 controls multiple steps of urogenital develop-ment. Development 121:4057-4065.

Torres M, G6mez-Pardo E, Gruss P (1996). Pax2 con-tributes to inner ear patterning and optic nerve tra-jectory. Development 122:3381-3391.

Trainor P, Tam P (1995). Cranial paraxial mesoderm andneural crest cells of the mouse embryo: co-distributionin the craniofacial mesenchyme but distinct segregationin branchial arches. Development 121:2569-2582.

Tucker A, Markham H, Green P, Doherty P, Sharpe P

(1998). A novel approach for inhibiting growth factorsignalling in murine tooth development. Inhibitionof FGFs. Eur J Oral Sci 106(Suppl 1): 122-125.

Tureckova 1, Sahlberg C, Aberg T, Ruch JV, Thesleff 1,Peterkova R (1995). Comparison of the expression ofthe msx-1, msx-2, bmp-2 and bmp-4 genes in themouse upper diastemal and molar tooth primordia.Int I Dev Biol 39:459-468.

Turing A (1952). The chemical basis of morphogenesis.Phil Trans R Soc B 237:37-72.

Vaahtokari A, Aberg T, Iernvall J, Keranen S, Thesleff I(1996a). The enamel knot as a signaling center in thedeveloping mouse tooth. Mech Dev 54:39-43.

Vaahtokari A, Aberg T, Thesleff I (1996b). Apoptosis in thedeveloping tooth: association with an embryonic sig-naling center and suppression by EGF and FGF-4.Development 122:121-1 29.

Vainio S, Karavanova 1, Jowett A, Thesleff 1 (1993).Identification of BMP-4 as a signal mediating sec-ondary induction between epithelial and mesenchy-mal tissues during early tooth development. Cell75:45-58.

van Genderen C, Okamura R, Farinas 1, Quo R-G, ParslowT, Bruhn L, et al. (1994). Development of severalorgans that require inductive epithelial-mesenchymalinteractions is impaired in LEFl-deficient mice. GenesDev 8:2691-2703.

van Valen L (1970). An analysis of developmental fields.Dev Biol 23:456-477.

Vastardis H, Karimbux N, Guthua SW, Seidman JG,Seidman CE (1996). A human MSXI homeobox mis-sense mutation causes selective tooth agenesis.Nature Genet 13:417-421.

Wall NA, Hogan BLM (1995). Expression of bone mor-phogenetic protein-4 (BMP-4), bone morphogeneticprotein-7 (BMP-7), fibroblast growth factor-8 (FGF-8)and sonic hedgehog (SHH) during branchial arch devel-opment in the chick. Mech Dev 53:383-392.

Waller M (1961). Chladni figures: a study in symmetry.London: G. Bell & Sons.

Webster G, Goodwin B (1996). Form and transformation.Cambridge, UK: Cambridge University Press.

Wedden SE, Ralphs JR, Tickle C (1988). Pattern formationin the facial primordia. Development 103(Suppl):31-40.

Weiss KM (1990). Duplication with variation: metamericlogic in evolution from genes to morphology. YearbookPhys Anthropol 33:1-23.

Weiss KM (1993). A tooth, a toe, and a vertebra: thegenetic dimensions of complex morphological traits.Evol Anthropol 2:121-134.

Weiss K, Bollekens J, Ruddle F, Takeshita K (1994). Distal-less and other homeobox genes in the developmentof the dentition. I Exp Zool 270:273-284.

Weiss KM, Ruddle FH, Bollekens JA (1995). Dlx and otherhomeobox genes in the morphological development

9(41369398998 Cri Re Ora Bil Me 39

397Crit Rev Oral Biol Med9(4)-369-398 (1998) at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from

and evolution of the dentition. Connect Tissue Res3 1:1-6.

Weiss K, Stock D, Zhao Z, Buchanan A, Ruddle F,Shashikant C (1998). Perspectives on genetic aspectsof dental patterning. Eur J Oral Sci 106(Suppl 1):55-63.

Wolpert L (1969). Positional information and the spatialpattern of cellular differentiation. I Theoret Biol 25:1-47.

Zhou P, Byrne C, Jacobs J, Fuchs E (1995). Lymphoid

enhancer factor 1 directs hair follicle patterning andepithelial cell fate. Genes Dev 9:570-583.

Zhao Z, Weiss K, Stock D. Development and evolution ofdentition patterns and their genetic basis. In:Development, function, and evolution of teeth.Teaford M, Ferguson M, Smith M, editors.Cambridge, UK: Cambridge University Press (inpress).

398 Crit Rev Oral Biol Med 9(4)369-398 (1998)398 Crit Rev Oral Biol Med 9(4):369-398 (1998) at PENNSYLVANIA STATE UNIV on February 28, 2014 For personal use only. No other uses without permission.cro.sagepub.comDownloaded from