JOURNAL OF EXPERIMENTAL ZOOLOGY (MOL DEV EVOL) 298B:73–90 (2003)

A Hierarchical Model of Plumage: Morphology, Development, and Evolution RICHARD O. PRUM 1nand JAN DYCK2 1Department of Ecology & Evolutionary Biology, and Natural History Museum, University of Kansas, Lawrence, Kansas 66045–2454 2Institute of Population Biology, Universitetsparken 15, DK–2100 Copenhagen, Denmark

ABSTRACT Plumage is a complex component of the avian phenotype. The plumage of an individual is composed of numerous hierarchically arranged developmental and morphological modules. We present a hierarchical model of plumage that provides an intellectual framework for understanding the development and evolution of feathers. Independence, covariation, and interaction among plumage modules create numerous opportunities for developmental and evolutionary diversification of feather complexity and function. The hierarchical relationships among plumage modules are characterized by both top-down and bottom-up effects in which properties of modules at one level of the hierarchy determine or influence the properties of modules at lower or higher levels of the hierarchy. Plumage metamodules are created by covariation or interaction among modules at different levels of the hierarchy. J. Exp. Zool. (Mol. Dev. Evol.) 298B: 73–90, 2003. r 2003 Wiley-Liss, Inc.

INTRODUCTION bristles in the definitive adult plumage (Lucas and Stettenheim, ’72). Feather follicle identity is Plumage is a complex and important component established early in development. Subsequently, of the avian phenotype. The plumage of a bird the follicle regenerates the appropriate feathers consists of thousands of feathers that vary in throughout the life of the bird. structure, shape, size, color, and chemical compo- Many of these variations in feather structure, sition (reviewed in Lucas and Stettenheim, ’72). shape, and color over the body and the life span Feathers are integumentary appendages charac- are directly correlated with the diverse functions terized by the unique tubular organization of the of the plumage in the life of the bird. Feathers feather follicle and germ (Prum, ’99), and by their are known to function in flight, temperature capacity for bipinnate structure (Figs. 1, 2) (Lucas regulation, water repellency, visual communi- and Stettenheim, ’72; Prum, ’99). The pennaceous cation, crypsis, sensory detection, sound produc- contour feathers, flight feathers, plumulaceous tion, water transport, and other functions downs, disintegrating powder downs, tiny filo- (Stettenheim, ’76). plumes, and bristles are some of the extremes of Most of what we know about the biology of feather shape and structure. The variety of feather feathers and plumage comes from extant birds. It structure is determined by variation in the size, is now clear, however, that feathers are not unique shape, and chemical composition of the keranti- to birds. Rather, feathers evolved in a coeluro- nocytes that make up the feather. saurian theropod lineage long before the origin of Avian plumage is renewed throughout the life of birds or the origin of avian flight. (Ji et al., ’98, the bird through periodic molt. Individual follicles 2001; Sereno, ’99; Xu et al., ’99a, b, 2000, 2001; can grow successive feathers that vary radically in Padian, 2001; Prum and Brush, 2002). Although structure, shape, and color between molts. For the paper repeatedly refers to birds, the plumage example, during the life of the bird, the follicles on phenotype originated in theropod dinosaurs before the head of the turkey (Meleagris gallopavo,

Phasianidae) initially grow plumulaceous natal n Correspondence to: Richard O. Prum, Natural History Museum, down feathers, which are subsequently replaced Dyche Hall, University of Kansas, Lawrence, Kansas 66045-2454. by fully pennaceous contour feathers in the E-mail: [email protected] Published online in Wiley InterScience (www.interScience.wiley. juvenile plumage, and ultimately by simple com). DOI: 10.1002.jez.b.00027 r 2003 WILEY-LISS, INC. 74 R.O. PRUM AND J. DYCK

differentiate the various levels of analysis within the fields of feather development and evolution, and, most importantly, to focus attention on new questions and motivate future research on plu- mage development and evolution. Lucas and Stettenheim (’72: 385, Fig. 241) presented a hierarchical, graphical summary of feather development. Lucas and Stettenheim’ s diagram presents the details of the complex development of a single feather in a hierarchical flow chart, an early progenitor of the conceptual model proposed here. (See also Widelitz et al., in this issue for the molecular and developmental events in different stages of feather development.) Here, the goal of our hierarchical model is to provide a heuristic framework that incorporates all components of the plumage phenotype. Further, we propose to integrate into our under- standing of feather biology the concepts modular- ity, hierarchy, and modular interactions (Mu¨ller and Wagner, ’91; Raff, ’96; Wagner, 2000), which Fig. 1. A. The structure of a typical pennaceous contour have become fundamental to the study evolution- feather with afterfeather. B. Cross-section of feather barb rami from the closed pennaceous portion of a feather showing ary novelties and developmental evolutionary the differentiation between the distal barbules (oriented biology since Lucas and Stettenheim (’72). This toward the tip of the feather) and the proximal barbules hierarchal framework applies to a wide variety (oriented toward the base of the feather). The hooks of the of proximate developmental, functional, and, pennulum of the distal ends of the distal barbules interlock ultimately, evolutionary questions. with the grooved dorsal flanges of the bases of the proximal barbules from the adjacent barbs forming the closed pennac- Our aims are to propose a unifying framework eous portion of the vane. The distal barbules of open to organize the breadth of research in feather pennaceous portions of feathers lack hooked pennula. Both biology, and to demonstrate that the plumage illustrations from Lucas and Stettenheim (’72). phenotype is an outstanding model system for understanding the modularity and hierarchy in the development and evolution of biological in- novations. the origin of birds. Thus, the model applies as well Below, we first introduce the concepts of to the non-avian plumage phenotype in feathered modularity, hierarchy, metamodules, indepen- theropods. dence, covariation, and interaction with reference All of the follicles and feathers of an individual to the plumage phenotype. We then present the birdFor, more accurately, coelurosaurian dino- hierarchical model of the plumage phenotype. saur– throughout its life can be conceived of as its Each section presents the evidence supporting plumage phenotype. The plumage phenotype is the existence of a module at that level of distinct from an individual plumageFa specific organization, and a discussion of the relationships coat of feathers of an individual – and provides a and interactions of those modules to other new concept for the entire component of pheno- modules in the hierarchy. Finally, we discuss the type, the subject of feather biology. In this paper, heuristic features of models for future investiga- we present a new conceptual model of the tion of the evolutionary developmental biology of development and evolution of the avian plumage feathers. phenotype. Intellectually, the model is an applica- tion to the plumage phenotype of the concepts of morphological modularity and developmental MODULARITY AND HIERARCHY hierarchy developed by Mu¨ller and Wagner (’91), Raff (’96), Wagner (2000), and others. Our goal Morphological modularity is a fundamental here is to provide a framework to understand feature of the plumage phenotype that contributes the complexity of plumage, to help define and both to its complexity and diversity. Morphological HIERARCHICAL MODEL OF PLUMAGE 75

Fig. 2. Histological cross-sections of a developing ridges form. D. Close-up of barb ridges from the contour feather germ from a chicken (Gallus gallus). lateral margin of the feather germ showing the ramus, A. A cross-section of the entire feather germ with the barbule plates, marginal and axial plate epithelia. The areas in B, C, and D highlighted with white boxes. B. differentiated distal and proximal barbules (Fig. 1B) Close-up of the rachis ridge on the dorsal margin of develop from the distal and proximal barbule plates of the feather germ, and the fusion of barb ridges to the each barb ridge. Abbreviations: d, distal barbule rachis ridge. C. Close-up of the new barb locus on the plates; p, proximal barbule plates. ventral margin of the feather germ where new barb

modules are serially homologous, or homonomous, ’99; Brush, 2000). Replicate modules are nested replicate morphological entities within the within other, more inclusive modules. For exam- phenotype (Raff, ’96). Examples of well recognized ple, each barb often includes numerous modular morphological modules include limbs, digits, ver- barbules. Most feathers are composed of numer- tebrae, leaves, flowers, and stamens. The plumage ous modular barbs. The plumage phenotype is phenotype is composed of numerous replicate composed of the hierarchical aggregation of morphological and developmental modules. For numerous nested, replicate morphological mod- example, the avian integument is organized into ules within all the feathers of all of the pterylae of the feathered areas, the pterylae or feather tracts, integument. with interposed largely featherless areas, or A fundamental question in the evolution of the apteria. Within the pterylae, there are numerous plumage phenotype is what developmental and replicate feather follicles. Within a typical pennac- evolutionary mechanisms determine the proper- eous feather, there are numerous modular feather ties and relationships within and among morpho- barbs which themselves are composed a central logical modules? It is clear that developmental ramus and numerous replicate barbules (Fig. 1). mechanisms operating within morphological mod- The plumage phenotype is the sum of these ules can contribute to or determine the properties numerous replicate morphological modules over of the modules nested within it. We call these the life of the individual. top-down effects because properties at the higher, A second fundamental feature of the plumage or more inclusive, levels of the hierarchy deter- phenotype is that its morphological modules have mine properties of modules nested at lower levels a hierarchical organization (Brush, ’93, ’96; Prum, of the hierarchy. For example, developmental 76 R.O. PRUM AND J. DYCK mechanisms operating at the level of the entire and appropriate genetic variation must precede follicle may determine which feather keratino- the evolutionary diversification among modules. cytes incorporate melanosomes, and thus deter- Independence of modules is not complete, or the mine the within-feather pigmentation pattern. integration of the phenotype and its development Thus, developmental mechanisms operating with- would breakdown. in a module can contribute fundamentally to Covariation among modules is the consequence determining properties of the modules nested of constraints on the independence of modules, or within it, or within a single lineage of the a derived loss of independence among modules. hierarchy of the plumage phenotype. However, Covariation can create metamodules within in- the relationships among modules within a single dividual feathers or at other levels of the plumage lineage of the hierarchy are not only driven from phenotype. Metamodules are novel, emergent the top of the hierarchy down. Rather, modules units that are composed of modules from across may create bottom-up effects in which the proper- different lineages or levels of the hierarchy. For ties of lower level modules contribute to or example, a distinctly colored plumage patch is a determine the properties of higher level modules plumage metamodule created by covariation in that they are nested within. For example, mor- pigmentation or structural coloration of feathers phology or shape of barbule cells greatly deter- or parts of feathers, grown from follicles that are mines whether a feather has a closed pennaceous distributed across specific portions of the integu- vane or an open plumulaceous structure. Or, ment, perhaps incorporating different pterylae. variation in the expression of different feathers Thus, covariation in color patterning of feathers keratins within individual keratinocytes will cre- from the follicles within a single or multiple ate emergent physical properties at the level of the pterylae can create an integrated plumage patch. entire feather. Thus, individual modules can also In some cases, the genetics of covariation within determine some properties of the larger modules the plumage phenotype are well understood, such that they are nested within through bottom-up as in the Columbian Restriction (Co) locus in effects or emergent properties. chickens, which inhibits eumelanin deposition in feathers of the hackle, wing, tail, and feet INDEPENDENCE, COVARIATION, (Smyth, ’90). AND INTERACTION Interactions among modules can also create metamodules across lineages or levels of the The hierarchical modularity of the plumage hierarchy. These interactions can be developmen- phenotype creates intrinsic potential for the tal or physical processes that occur after develop- development and evolution of additional morpho- ment is complete. As a developmental example, logical complexity. These opportunities arise the rachis ridge is created or specified by the through at least three mechanisms: independence fusion of barb ridges on the anterior side of the of modules, covariation among modules, and feather germ and ultimately becomes the rachis of interactions among modules (Mu¨ller and Wagner, the feather (Fig. 2B)(Lillie and Wang, ’41; Lucas ’91; Raff, ’96; Wagner, 2000). These mechanisms and Stettenheim, ’72; Prum, ’99; Harris et al., operate among modules that can be either 2002). Thus, the rachis is a unique morphological within individual level or among multiple levels novelty that develops as a metamodule by an of the hierarchy, and are not entirely mutually interaction (i.e., fusion) of component modules exclusive. within the feather germ (Harris et al., 2002). Independence of morphological modules pro- Proximate interactions among covarying modules vides the opportunity for development and evolu- after the completion of development can also tion of structural diversity among homologous create novel metamodular functional structures. modules within the plumage (Mu¨ller and Wagner, The premier example is the zippering interconnec- ’91; Raff, ’96; Wagner, 2000). For example, tions between the hooked pennula of the distal independence among follicles permits variation barbules and the grooved bases of the adjacent in structure, shape, size, and color of feathers on proximal barbules that create the closed pennac- different parts of the body. Independence among eous portion of the feather vane (Fig. 1B). barbs within a single feather and of barbules Furthermore, the proximate physical interactions within a barb permits the development and among the remiges and rectrices create the evolution of structural variations that create the physical properties for the wing and tail to form pennaceous vane. Developmental independence air foils. HIERARCHICAL MODEL OF PLUMAGE 77

A HIERARCHICAL MODEL OF THE PLUMAGE PHENOTYPE

Here we present a hierarchical model of plu- mage with a description of the modular and metamodular components of the plumage, their major developmental and physical interactions. The model is illustrated in a series of schematic figures that depict the proposed hierarchical relationships (Figs. 3–6). The modules and meta- modules are hypothesized based upon the classic descriptions of feather morphology and develop- ment (e.g., Lucas and Stettenheim, ’72), and upon recent advances in feather development and evolution (Prum, ’99; Chuong et al., 2000; Harris Fig. 4. A schematic diagram of the hierarchical relation- ships among the basic morphological modules of a feather. The et al., 2002). feather germ initially includes the dermal pulp, the sheath The presentations of each module below include (outer epidermal layer), and the basal epidermal layer. The an anatomical description of the module and its basal layer (including the marginal plate epithelium) controls variations; the developmental and anatomical the proliferation and organization of an intermediate epider- evidence for its recognition as a phenotypic mal layer into a series of longitudinal barb ridges. Each barb ridge develops into a single barb ramus and a series of barbule module; a description of its independence from, plates, which are separated by the cells of the axial plate. The and its covariation and interactions with other barbule plates can be differentiated into distal and proximal modules; and sometimes discussions of outstand- barbule plates. ing research questions pertaining to that module. This paper is a linear description of a hierarchical model; without hypertext, it cannot capture the distributed over the integument, but are concen- essential organization of the model. As a compro- trated into specific tracts, called pterylae. Pterylae mise, we present the descriptions of the modules are interspersed with apteria, which may be in the order from the top down– from the most entirely featherless or include down feathers. inclusive to least inclusive. Then, we present Some apteria occur as unfeathered areas within metamodules at the level of organization of a single pteryla. The pterylae and apteria are the individual feathers, and lastly metamodules at the level of the plumage. Pterylae and apteria With few exceptions (e.g., in penguins, Sphenis- cidae), contour feathers are not continuously

Fig. 5. A schematic diagram of the modular interactions involved in the creation of the feather rachis, a metamodular Fig. 3. A schematic diagram of the hierarchical relation- component of a feather. The rachis is created by the fusion of ships among the morphological modules within the plumage barb ridges. Subsequent barb ridges fuse to the rachis to phenotype. The pterylae include numerous feather placodes create a pennaceous structure. Finally, the rachis ridge and which each develop into a feather follicle. Each follicle grows a sheath dedifferentiate to become the calamus. A series of pulp series of feathers, which are replaced over time by periodic caps are produced by the inner-most germinative layer of the molts. Apteria are spaces between pterylae which lack contour basal epithelium through interactions (dashed lines) with the feathers or which lack feathers entirely. dermal pulp. 78 R.O. PRUM AND J. DYCK

responsible for the origin of the variation in the distribution of pterylae. However, it is also clear that many pterylogical patterns are functionally redundant, and that all can produce a complete plumage. With the exception of the derived loss of apteria in the cold-adapted penguins, there are few adaptive hypotheses about the evolution of pterylae. For example, nearly all avian species have a prominent ventral apterium which func- tions proximately in brooding eggs. However, the ventral apterium is not maintained by proximate Fig. 6. A schematic diagram of example plumage metamo- selection for brooding since many birds indepen- dules created by morphological and developmental correla- tions among feathers within and among different pterylae. dently develop a featherless, edematous brood Right– A distinct plumage patch created by correlated patch to facilitate incubation. Furthermore, the coloration among feathers from different pterylae. Left– ventral apterium has not been lost in any lineage Remiges (wing flight feathers) comprise a plumage metamo- of brood parasites (R. B. Payne, pers. comm.). dule created from correlation among feathers of pterylae on Apparently, there are constraints on the loss of both wings. Center– Rectrices (flight tail feathers) are a plumage metamodule created by correlation among feathers this apterium that have been overcome only in the from a single pteryla. Center Left– All the flight feathers penguin clade which often brood their eggs on comprise a more inclusive plumage metamodule created by their feet or shanks of the legs. Some species have correlations among the remiges and rectrices. evolved conspicuous novel apteria through sexual or social selection for display behavior– e.g. the bare bright blue crown skin of Cicinnurus and most general, inclusive units in the hierarchy of Diphyloides birds-of-paradise (Paradiseaidae) and the plumage phenotype (Fig. 3). The presence of Perrisocephalus cotingas (Cotingidae). Interest- the apteria and regions with other differentiated ingly, even though these features have clearly integumentary appendages (e.g., scutate or reti- evolved by sexual selection for function in male culate scales on the legs) indicate that the pteryla courtship displays, there is no sexual dimorphism and apteria are genuine developmental entities in these apteria. Due to the development of with a distinct autonomous identity, and not pterylae early in embryogenesis (beginning day merely epiphenomena of the process of establish- 6) before any sexual differentiation has taken ment of feather placode spacing (see below). place, these derived display apteria are con- Recent molecular developmental studies have strained from evolving sexual dimorphism. There established that feather placodes and follicles may be no available genetic variation for sexual develop in a temporal pattern within a pteryla by differentiation in components of the phenotype a temporal wave of epithelial competence to that develop prior to the sexual differentiation. respond to induction (Noralmy and Morgan, ’98; Conversely, many apparently featherless areas are Jiang et al., ’99). not apteria, but are pteryale filled with small Pterylae vary extensively in shape and distribu- bristle feathers (e.g. the head of the turkey). tion among different avian families and orders (Lucas and Stettenheim, ’72). These variations Feather placode and follicle have been used traditionally in taxonomic studies of birds. Pterylae and the distribution of feathers Within the pterylae are the many thousands of on the integument have a complex evolutionary feathers (Fig. 3). These feathers initially develop history within birds, and certainly preceding birds from placodes during the first 6–15 days of within other coelurosaurs as well (Prum and incubation. A placode is the epidermal primor- Brush, 2002). Apteria are often viewed as merely dium of a feather follicle and feather germ. The the areas remaining when the feather develop- first signal to form placode comes from the dermis ment is complete, but as Lucas and Stettenheim which determines the site of the feather. The (’72) implied, apteria are likely developmental induced epidermis then forms the feather placode entities with their own modular identity and by the elongation of epidermal cells. More dermal developmental determination. It is possible that cells aggregate below the epidermal placode soon selection at the level of the entire plumage for after its formation. The epidermal placode and its thermoregulation or communication could be dermal condensation together develop into the HIERARCHICAL MODEL OF PLUMAGE 79 feather germ and follicle. Feather placodes have Although ‘‘feather follicle’’ sometimes refers been traditionally recognized as morphologically only to the socket that holds the developing distinct because of the associated dermal conden- feather (Lucas and Stettenheim, ’72), we refer to sation (Maderson and Alibardi, 2000), which is the follicle as the organ of feather growth and also present in avian scutate scales but not regeneration. Developmental mechanisms operat- in avian reticulate scales or alligator scales ing at the level of the feather follicle have (Maderson and Alibardi, 2000). tremendously influential effects on the individual Although this placode-associated dermal con- feather and the entire plumage phenotype. Feath- densation has an evolutionary history that is of er structures develop through differentiation of substantial interest, the placode exists as a the tubular epidermis of the feather germ which molecular phenomenon among all epidermal ap- itself is generated by the internal epidermal layer pendages as well. It is the site of molecular signals of the follicleFor follicle collarFthrough contin- that interact with the dermis and lead to integu- ued induction and nourishment by the papillary mentary appendage morphogenesis. The avian dermis in the center of the follicle and germ reticulate scales, which lack this morphologically (Fig. 2A) (Lucas and Stettenheim, ’72). Classical defined placode, share distinct placode-specific experiments bisecting and transplanting part of patterns of gene expression with feather and the epidermal collar and the dermal papilla among scutate scale placodes (Chuong et al., ’96; Widelitz feather follicles demonstrate the important role et al., ’99). Furthermore, Harris et al. (2002) of this organ in determining feather structure documented that the placode of feathers, avian (Lillie and Wang, ’41; Cohen and ’Espinasse, ’61). scutate scales, and alligator scales all share an Examples of feather morphologies that are likely early anterior-posterior polarized pattern of ex- substantially determined by follicle-level processes pression of Sonic Hedgehog (Shh) and Bone include the growth of feather shape (Prum and morphogenetic protein 2 (BMP2). The broad dis- Williamson, 2001), and the growth of within- tribution of this mechanism of placode specifica- feather pigmentation pattern (Prum and tion and development indicates that this Williamson, 2002). plesiomorphic, or primitive, feature was shared The diversity in structure within modular by the integumentary appendages of the common components of individual feathers and among the archosaurian ancestor of birds and crocodiles. The series of feathers grown from the same follicle shared details of the spatially polarized expression demonstrates the flexibility of the developmental of Shh and Bmp2 confirm that the placodes of mechanisms within a follicle, and indicates that archosaur scales and avian feathers are homo- many features of feather structure are determined logous, and that the dermal condensation of the within lower levels modules of the hierarchy. avian feather and scale rudiments is subsequently There do appear to be at least two distinct classes derived in feather and avian scutate scale pla- of follicles whose repertoires of developmental codes. Feather placodes have a distinct, uniform mechanisms are quite distinct. Lucas and Stet- hexagonal distribution within the pterylae. Pla- tenheim (’72) documented thoroughly that filo- code spacing is determined by a cascade of plumes are a distinct class of feathers inhibitory and activating molecular signaling characterized by a small rachis with a terminal events that are coordinated by the overlying tuft of barbs. Filoplumes develop from a distinct temporal wave of competence to respond to class of follicles that are each associated with and induction within the pteryla (Chuong et al., ’96; adjacent to a contour feather follicle (Lucas and Noralmy and Morgan, ’98; Jiang et al., ’99; Stettenheim, ’72). In Figure 3, filoplumes are Widelitz et al., ’99). represented by a specific subset of the follicles and After its appearance, the placode elongates into feathers within each pteryla. The origin of a short budFa tubular epidermal structure with a filoplume placodes is delayed by several days after central cylinder of dermis. The short bud is the the associated feather germs have already devel- first feather germ. The epidermis of the short bud oped follicles (Lucas and Stettenheim, ’72). Even soon differentiates into the barb ridges that will though most feather follicles are capable of become the barbs of the first natal down. Very growing feathers with a diversity of morphologies, shortly thereafter, the epidermis around the base filoplume follicles are distinct in growing exclu- of the tubular feather germ proliferates and sively filoplumes. Thus, nested within the most invaginates, forming the cylindrical feather general concept of feather placodes and follicles follicle. are the two distinct classes of follicles which grow 80 R.O. PRUM AND J. DYCK

filoplumes and all other feathers. The evolutionary birdsF and subsequent feather generations. The and developmental origins of filoplume follicles exhibition of different plumage coloration with remains to be investigated. distinct patches in different parts of the year is another example of developmental independence among feather generations. Feathers The feather and feather germ consist of nested The essential components of the plumage sets of modular developmental and evolutionary phenotype that develop from follicles are the units (Fig. 4). feathers themselves. The development of the first natal feathers is crucial to the development of the Dermal pulp feather follicle itself, but subsequently the feath- The center of the feather germ is occupied by ers are sequentially replaced throughout the life of the dermal pulp (Fig. 2). Although not usually the bird through periodic molt (Fig. 3). Subse- considered a component of the plumage, the quent feathers grown from a single follicle are dermal feather pulp is a critical component of hypothesized to be temporal iterations of a the feather germ. The pulp supplies nutrients to homologous tubular feather germ. The feather is the developing feather epidermis. Also, melano- regenerated by the follicle through proliferation cytes and pigment cells that transfer pigments to and differentiation of the follicle collar. The the feather epidermis migrate into the feather continuity of the tubular epidermis between germ from the central dermal pulp. The dermal feather generations is demonstrated by the physi- pulp is produced continuously but is reabsorbed cal interconnection between the calamus and the periodically (Lillie, ’40; Lucas and Stettenheim, distal tips of the initial barb ridges of subsequent ’72). This process occurs with the periodic produc- feather generations. This feature is easiest to tion of pulp caps by the germinative, or innermost, observe in the natal down feathers connected to layer of the basal layer epithelium (see Basal the tips of the first juvenile contour feathers Layer below). Although the pulp caps consist of (Lucas and Stettenheim, ’72; e.g., Figs. 229–231). keratinized epidermal cells, the interactions be- Based on experiments with regenerating plucked tween the dermal pulp and the germinative basal feathers, Cohen and ’Espinasse (’61) argued that a layer are critical to the development of this feather develops from the elongation of the follicle modular component (Fig. 5). collar, and is not generated by the collar. Their conclusions were inherently based upon the Sheath damage that comes from experimental plucking of a mature feather, and further leaves the The sheath is a distinct modular component of question unanswered as to how the follicle the feather germ that is produced by the outer repeatedly regenerates the feathers of the appro- epidermal layer of the feather germ (Figs. 2, 4) priate phenotype. We agree with Lucas and (Lucas and Stettenheim, ’72). The sheath forms a Stettenheim (’72) that the collar refers to that deciduous outer layer to the feather germ that enduring epidermal tissue at the base of the permits the feather germ to emerge smoothly from follicle that supports regeneration of subsequent the follicle, protects it during development but feathers during molt. then flakes apart, and falls off after the feather The identity of a follicle occurs early in its components have completely matured. The sheath development, and is retained permanently. Con- is ultimately subsumed by the production of the sequently, a follicle consistently generates appro- calamus at the base of the feather (see below). priate feathers with each molt. Independence The sheath is distinct in that it is composed of among different follicles within and among pter- a keratins, in addition to the b keratins that ylae provides the opportunity for the evolution of exclusively comprise the rest of the components of the functional diversity within the plumage. the mature feather (Maderson and Alibardi, 2000). Independence and decoupling of the develop- Maderson and Alibardi (2000) imply that the ment of feather generations grown from a single pattern of vertical alternation of a and b keratins follicle permits the extraordinary morphological in the feather germ could be homologous with the and functional diversity through the life of the vertical stratification of keratin classes in the bird. Such diversity includes the differences scales of shedding lepidosaurs. However, these between natal downFthe first plumulaceous protein expression features are clearly convergent plumage to cover the integument of hatchling between birds and lepidosaurs (Prum and Brush, HIERARCHICAL MODEL OF PLUMAGE 81

2002). As elsewhere in the reptilian integument, differentiated in situ. Most recently, Yu et al. the expression of a -keratin in the feather sheath (2002) hypothesize that the intermediate layer is is associated with its flexible properties and its ‘‘cleaved’’ into barb ridges by the marginal plate. deciduous function. All these models posit that the intermediate layer As the cells of the sheath become a component of itself precedes the formation of the barb ridges, the developing calamus, the sheath ceases to be a and that the process is essentially the subdivision distinct entity and cannot be identified histologi- of this intermediate layer. cally as a distinct layer of the b keratin calamus Recently, Harris et al. (2002) have described the (Lucas and Stettenheim, ’72: 381). This observa- developmental roles of Sonic Hedgehog (Shh) and tion reinforces the conclusion that the sheath is Bone morphogenetic protein 2 (Bmp2) expression not a lineage of a keratin expressing epidermal in the basal layer epithelium and in the creation of cells continuous through all feather generations, barb ridges. Early in the formation of a barb ridge, but a modular component of the feather germ that Shh is generally expressed in all marginal plate is regenerated with each new feather in response cells in the folded portions of the basal layer to molecular signaling to differentiate the outer between barb ridges, whereas BMP2 is expressed epithelial layer. in a restricted zone at the periphery of the fold of the marginal plate epithelium. During the forma- Basal layer tion of a new barb ridge, proliferation of barb ridge The basal layer of the feather germ does not cells begins at the site of the peripheral fold in the become a conspicuous component of the mature marginal plates of neighboring ridges (Harris feather, but it plays a crucial role in the morpho- et al., 2002). The lineages of intermediate cells genesis of the barb ridges, barbules, barb rami, the that form the barb ridges initially proliferate rachis, and pulp caps of the feather (Fig. 2). The between the marginal plate epithelium and the basal layer epithelium of the feather germ is outer epithelium (or sheath). As they proliferate, a homologous with the general basal layer of the notch forms in the peripheral fold in the marginal epidermis (Lucas and Stettenheim, ’72). It forms plate. The new barb ridge then expands balloon- the boundary between the barb and rachis ridges, like through cell proliferation into the basal layer, which develop into the mature feather, and and is flanked by a new set of Shh expressing the central dermal pulp (Fig. 2B-D). Due to the marginal plate cells. Similarly, the fusion of barb tubular configuration of the feather germ, the ridges (or the loss of distinct differentiation basal epidermal layer is the internal or central- between neighboring barb ridges) is initiated most epidermal layer of the feather germ, and the peripherally through the gradual peripheral to intermediate and outer layers are peripheral to it. central retraction or diminishment of the marg- The barb ridges are the longitudinal masses of inal plate epithelium (Harris et al., 2002). Further, intermediate epithelium cells that are bordered Harris et al. (2002) indicate that polarized internally by a layer of cells that comprise the Shh-Bmp2 signaling within the marginal plate basal layer (Fig. 2). The basal layer forms a series epithelium orchestrates the proliferation of the of folds toward the periphery of the feather germ barb ridge cells between the marginal plate and that separates adjacent barb ridges with a double the outer epithelial layer. cell layer. It also forms an intervening single cell These data from Harris et al. (2002) are layer that cover the internal surface of the barb incompatible with the classic notion that barb ridge. The folded portions of the basal layer are ridges are formed by the subdivision of a pre- also referred to as the marignal plate, or marginal existent intermediate epidermal layer (e.g. Lucas plate epithelium. Lucas and Stettenheim (’72) and Stettenheim, ’72: 376. Fig. 238). Rather, the summarize the confusion among classic studies intermediate layer that becomes the feather is over the role of the basal layer in barb ridge created through controlled morphogenesis orche- formation. Lucas and Stettenheim (’72), Strong strated by signaling from the basal layer. Harris (’02), and Hosker (’36) hypothesized that the basal et al. (2002) hypothesize that Shh plays a role in layer cells rearrange themselves around the bulge fostering proliferation of barb ridge cells, and that of intermediate epidermal cells of the barb ridges, Bmp2 plays a role in controlling Shh signaling, and that they may be forced into position by the limiting proliferation, and ultimately fostering cell dermal pulp. Griete (’34) alternatively proposed maturation. The basal layer and marginal plate that the axial organization of the barb ridges signaling have an additional crucial role in the originated first, and that the basal layer cells morphogenesis of the rachis and pennaceous 82 R.O. PRUM AND J. DYCK structure (Harris et al., 2002). Thus, the same Within the developing barb ridge, the more developmental mechanisms that create the differ- central cells develop into a series of longitudinally entiated intermediate epidermal layer of the interconnected cells that form the ramus, or shaft feather germ are responsible for morphogenesis of of the barb (Fig. 2, 4). The morphology of the barb many of the diverse structures that are created ramus is complex, multicellular, and differentiated from this layer. into cortical, medullary, and central layers (or The germinative layer is the inner-most compo- pith). The outer, more peripheral cells within the nent of the basal epidermal layer (see Lucas and barb ridge differentiate laterally into a pair of Stettenheim, ’72: 384–385), and forms the kerati- peripheral barbule plates and the central axial nized pulp caps through interactions with the plate. The axial plate tissue separates the two dermal pulp and the process of dermal pulp barbule plates during their development and then reabsorption. The other basal membrane cells, disintegrates (Lucas and Stettenheim, ’72). The including the marginal plate cells, ultimately barbules develop from a single row of cells in a undergo apoptosis (Yu et al., 2002), and do not single layer of feather germ cells. The more basal, become a component of the mature feather. Yu innermost cells become the base of the barbule et al. (2002) hypothesize that apoptosis of the and ultimately connect the barb ramus to form the marginal plate cells permits the barbs to open up. branched structure of the barb (Fig. 1B). The Strictly speaking, however, only those keratino- more peripheral cells become the elongate distal cytes that establish tight cell connections (through cells of the barbule pennulum (Fig. 1B). As the desmosomes or other structures) during keratino- barbule keratinocytes elongate, they are con- genesis have the potential to remain intercon- strained from growing outward by the feather nected as the feather emerges (Mantulionis, ’70; sheath, so they grow upward within the feather R. O. Prum, pers. obs.). For example, the cells of germ. This physical pattern is well illustrated in adjacent barbules on a barb are immediately Fig. 239 of Lucas and Stettenheim, ’72. Although adjacent to one another without any intervening barbules develop from a single row of cells, the layer of apoptotic cells (Fig. 2D), yet they show no distal elongation of the barbules means that a problem opening up upon expanding. Thus, horizontal cross-section of a maturing, keratiniz- apoptosis appears to be the way of ultimately ing barb ridge will sample cross-sections of eliminating the basal layer cells, but it does not numerous barbules, with the oldest toward the provide a crucial mechanism for the separation of center and the youngest toward the periphery. barbs themselves (i.e. experimentally suppressing The axial plate is composed of the barb ridge apoptosis of marginal plate cells should not cells along the central axis of the barb ridge produce a feather that could not open). between the two barbule plates (Fig. 2D). The axial plate cells do not form a component of the Barb ridgeFbarbs, ramus, barbules, mature feather. According to illustrations of Yu and axial Plate et al. (2002), the axial plate cells also die by The fundamental, differentiated, modular unit apoptosis. of the intermediate epidermal layer of the feather Barbule morphology is a very conspicuous germ are the barb ridges (Figs. 2, 4)(Lucas and component of feather morphology because bar- Stettenheim, ’72). Barb ridges first develop on the bules can have important bottom-up effects on the anterior (or dorsal) side of the initial feather germ structure of the feather vane. Plumulaceous (Lucas and Stettenheim, ’72), but in at least some feathers are characterized by barbules with a pterylae of the chick (Gallus gallus) barb ridge series of simple cells within the pennulum (often origin is simultaneous around the entire feather with nodal prongs at cell junctions that create germ (R. O. Prum, pers. obs.). The barb ridges space-filling tangles among barbule filaments), have complex modular components nested within and little lateral differentiation among barbules. them: the presumptive barbs ramus (the main In contrast, a feather with a planar pennaceous shaft of the barb); the paired, peripheral barbule vane is characterized by strongly differentiated plates, which develop into the paired barbules; and barbules on either side of the barb ridge (Fig. 2D). the axial plate (Figs. 2, 4). Interactions among The barbules that extend from the ramus toward barb ridges (mediated by basal layer signaling) are the tip of the featherFcalled distal barbu- also responsible for the creation of additional lesFhave conspicuous hooks on the pennulum meta-modules including the rachis and the cala- cells, whereas the barbules that extend toward mus (Fig. 5). the base of the entire featherFcalled proximal HIERARCHICAL MODEL OF PLUMAGE 83 barbulesFhave a prominent groove in the cells at ridge (Lucas and Stettenheim, ’72). This special their bases (Fig. 1B). The differentiated proximal name has contributed to the erroneous notion that barbules reach over the obverse (outer) surface of the rachis is a distinct entity from the barb ridges the planar vane and the appropriately oriented (Maderson and Alibardi, 2000), sometimes erro- hooklets interact with the grooved bases of the neously hypothesized to be homologous with the proximal barbules to create the coherent structure central ridge of a reptilian scale. A simple of the pennaceous vane. This differentiation is a inspection of the diversity of feather morphologies, consequence of decoupling the development of the however, demonstrates that the identity of the paired barbule plates within a cell layer of the rachis ridge is established during development. barb ridge from one another. The distal tip of nearly every pennaceous or Interestingly, the first cells in the barb ridge to plumulaceous feather is comprised of a set of mature are the peripheral barbule plate cells and equivalently sized barbs that lack a presumptive ultimately the more central ramus cells. It has rachis (e.g. Lucas and Stettenheim, ’72: Figs. 229– been hypothesized that this sequence is derived 231). Describing rachis formation in chicken from a developmental mechanism that facilitates feathers, Lucas and Stettenheim (’72:371) wrote nutrient supply to the peripheral tissues first ‘‘two to four barb ridges in the middorsal region (Prum and Brush, 2002). If the central tissues of of the blastema [or feather germ] fuse at their the ramus were to keratinize first, it would proximal ends, thereby creating the apex of the prevent nutrient supply to the peripheral cells. rachis.’’ Harris et al. (2002) have further estab- The importance of nutrient transport and pigment lished that the rachis ridge of the pennaceous cell transport is demonstrated by the existence of embryonic duck rectrix is established through the numerous gaps between barb ridge cells until they initial fusion of barb ridges on the anterior side of finally begin to keratinize. the feather follicle. The initial fusion of barb Little is known about the molecular mechan- ridges to form the rachis (and the later fusion of isms of barbule development and differentiation. barb ridges to the rachis) proceed from the Harris et al. (2002) hypothesize that this gradient periphery of the marginal plate to the center in differentiation is established and controlled by (Harris et al., 2002). Lastly, the distinct or polarized Shh-Bmp2 signaling in the adjacent singular identity of the rachis ridge has been marginal plate. Yu et al. (2002) establish that falsified by classical experimental developmental Bmp2 signaling switches from the peripheral fold studies in which multiple rachi are formed within of the marginal plate epithelium to the peripheral a single feather germ by experimental bisection of portions of the paired barbule plates later in a regenerating feather follicle (Lillie and Wang, development. As hypothesized by Harris et al. ’41; Cohen and ’Espinasse, ’61). (2002), localized Bmp2 signaling in barbule plates Once established, the rachis ridge develops a later in development may play a similar role in distinct identity with a distinct internal structure, controlling the response to the persistent Shh including a medullary layer and pith. After the signaling, which favors cell proliferation, and rachis ridge is formed initially, barb ridges con- fostering the maturation and keratinization of tinue to fuse to it as they reach the dorsal side of barbule cells. Nothing is known about the the feather follicle by helical growth (Lucas and mechanism responsible for the differentiation of Stettenheim, ’72; Prum, ’99). Barb ridges that fuse distal and proximal barbules, which must depend to the rachis constitute the lateral cortex of the upon establishing a rachis vs. new barb locus side developing rachis, as shown elegantly by Lillie and identity to the two barbule plates within a barb Wang (’41). In this fashion, helical growth of barb ridge. Further, no primary histological descrip- ridges around the follicle, rachis ridge establish- tions or experimental studies have been done on ment, serial fusion of barb ridges to the rachis the development of barbs in feathers that lack ridge, and continued origin of new barb ridges on barbules (such as the display plumes of certain the posterior (ventral) new barb locus (Fig. 2C) egrets Casmerodius). create the complex structure of the pinnate feather with a planar vane (Lucas and Stetten- Rachis heim, ’72; Prum, ’99; Prum and Williamson, In a pennaceous feather, the rachis is the main 2001). shaft of the feather vane to which the barbs are Occasionally, the rachis bears barbules in the fused at their bases. The rachis has long been internodes between barb fusions (e.g. Pavo rec- described as developing from a rachis, or rachidial, trices). Apparently, the barbule developing 84 R.O. PRUM AND J. DYCK capacity of the peripheral tissue of the barb ridge beginning of the calamus is marked by the super- is retained after fusion to the rachis, until the next ior umbilicus. As elegantly stated by Lucas and fusion event. This phenomenon reaches its ex- Stettenheim (’72), the superior umbilicus is the treme in the racket-tailed parrots (Prioniturus, opening through which the dermal pulp, that is Psittacidae) in which the rachis bears barbules for wrapped in the feather vane and will be released the length of the intermediate barb-free zone of its as the feather unfolds from the sheath, passes into racket-shaped central rectrices (Bleiweiss, ’87). the inside of the keratinized tube that becomes the Interestingly, the development of these rachis- calamus. Calamus formation is characterized by born barbules has never been described histologi- the loss of differentiation between the outer and cally. intermediate epidermal layers. The sheath unites Toward the base of the vane as the last barb with the calamus and ceases to be a distinct, a ridges fuse to the rachis, the rachis enlarges in size keratin expressing cell lineage. During growth of providing additional structural support to the the feather vane, the pulp caps are only weakly feather. keratinized and are typically destroyed as the feather emerges from the sheath. Within the calamus the pulp caps become permanent struc- Afterfeather and hyporachis tures that are integrated into the feather. The afterfeather is a ventrally oriented feather vane that grows simultaneously from the same Feather keratinocytes feather germ and follicle (Fig. 1A). The vane of the afterfeather is usually plumulaceous and much The mature feather is composed of the feather smaller in size than the main feather vane, but in type b keratins produced by the keratinocytes of cassowaries (Casuarius) and emus (Dromaius) the the developing feather germ and the pigments afterfeather is identical in size and morphology to they contain (Brush, ’78, ’93). Feather keratino- the main vane. The afterfeather has a patchy cytes are the ultimate modular, cellular compo- distribution among modern bird orders, but it is nents of all the larger units of the feather. There invariably absent in remiges and rectrices are a number of critical developmental features (Ziswiler, ’62; Lucas and Stettenheim, ’72). The that are determined at the level of the keratino- hyporachis (the shaft of the afterfeather) and the cytes. These include, but are not limited to, cell afterfeather are formed by a duplication of the number, cell size, cell shape, the types and helical growth and serial barb ridge fusion that proportions of keratins to be expressed, whether creates the main vane that is oriented toward the or not to incorporate eumelanin or phaeomelanin posterior (ventral) margin of the follicle and granules, whether to accept lipid soluble carote- feather germ (Lillie and Juhn, ’38; Ziswiler, ’62; noid pigments, and the internal structure of Lucas and Stettenheim, ’72). Accordingly, the keratin deposition within the cell. Little is known hyporachis ridge is established by the fusion of about the mechanisms of feather b keratin synth- barb ridges on the posterior (ventral) margin of esis and self assembly within feather keratinocytes the follicle. As a consequence of posteriorly- beyond a few transmission electron microscope oriented helical growth by barb ridges in a studies (Mantulionis, ’70; Bowers and Brum- posterior quadrant of the follicle, the posterior baugh, ’78; Alibardi, 2002). locus of new barb formation duplicates into two The number, size, and shape of any keratino- laterally displaced centers of new barb ridge cytes will critically effect how the larger units of formation (Lillie and Juhn, ’38; Ziswiler, ’62; the hierarchy function in the completed feather. Lucas and Stettenheim, ’72). Prum (’99) hypothe- Although the diversity of feather keratins and the sized that the afterfeather and hyporachis are variation of expression between pennaceous and secondarily derived, because the mechanism of plumulaceous feathers (or even pennaceous or hyporachis and afterfeather formation is a dupli- plumulaceous portions of a single feather) support cation of the mechanism that creates the main the hypothesis that variation in feather keratin vane. expression is functional or adaptive (Brush, ’78, ’93), we do not yet know enough about the physical properties of these keratin variations to Calamus test this hypothesis critically. The calamus is the tubular base of the feather. The emergence of the feather from the tubular The end of the rachis (if present) and the sheath to form the planar vane requires that the HIERARCHICAL MODEL OF PLUMAGE 85 mature but ensheathed barbs and barbules must that feather keratinocytes have the capacity to be ‘‘spring-loaded’’ and ready to assume their selectively absorb carotenoids, and in some cases appropriate angles (see Pennaceous Vane below) to differentiate among some different classes of (’Espinasse, ’39; Prum and Williamson, 2001). carotenoids. How the angles of the barbs to rachis and the A complex of features determines whether a barbules to the ramus are specified by developing mature keratinocyte produces a structural keratinocytes is a fundamental question that has colorFi.e., creates visible hue as a result of never been addressed. We do not know of any interactions of light with the physical structure previous hypotheses addressing how these physi- of the cell. Iridescent structural colors are gen- cal properties are determined during develop- erally produced by constructive interference, or ment. Apparently, the deposition of keratin coherent scattering, of light by layers of melanin within cells, the cell shapes, and the junctions granules in feather barbules (usually only the among cells are appropriately engineered within distal barbules which cover the obverse surface of the cylindrical feather germ so that upon emer- the feather) (Dyck, ’76; Durrer, ’86). These hues gence these structures can assume a coherent, are determined by a number of factors: the size of functional shape. This keratin engineering is the melanosomes produced by melanocytes, final an additional, potentially important function of spatial arrangement of the melanosomes within diversity of b-keratin composition of feather the barbule keratinocytes, and the amount of components. keratin deposited between the melanosomes. In contrast, typically noniridescent colors are pro- duced by coherent scattering of light by the spongy Color of feather keratinocytes keratin-air matrix within the mature medullary The color of feather keratinocytes can come cells of feather barb rami (Prum et al., ’98, ’99). In from pigments or structural color (Lucas and these feathers, the hue is produced by selective Stettenheim, ’72). Pigments in birds include constructive interference of specific wavelengths melanin, carotenoids, and others. Feather mela- of light waves, which are determined by the size nins come from melanocytes of neural crest origin and spatial distribution of the air-bubbles (origin- (Lecoin et al., ’98) that migrate into the developing ally cytoplasm of the keratinocyte) within the feather germ from the dermal pulp of the feather medullary ramus keratinocytes (Prum et al., ’98, germ (Greite, ’34; Watterson, ’42; Strong, ’02). ’99). Little is known about the development of These melanocytes extend pseudopodia among the color producing, spongy medullary barb keratin epidermal cells of the barb ridges, and transfer (Auber, ’71, 72). fully developed melanosomes to the developing feather keratinocytes (Greite, ’34; Watterson, ’42; The pennaceous vane Lucas and Stettenheim, ’72; Strong, ’02). Melano- somes are actively taken into the keratinocytes by The closed, pennaceous vane of a contour phagocytosis (Greite, ’34; Watterson, ’42; Lucas feather is a complex and functionally important and Stettenheim, ’72; Jimbow and Sigiyama, ’98; feather metamodule (Figs. 1, 5). The vane itself Sharlow et al., 2000), and are incorporated into does not exist during development but is created the keratin of that cell as it completes its only after the feather emerges from its tubular development and dies. sheath and expands to assume a planar shape. Feather carotenoids are introduced into feather Only at this time, after keratinocyte development keratinocytes in a solution of fat globules that is completed, do the barbs expand to assume an come from the blood supply in the dermal appropriate angle with the rachis, and the bar- feather pulp (Lucas and Stettenheim, ’72; Brush, bules expand to interconnect with one another, ’78). The precise mechanism by which fat soluble and create the closed vane. The metamodular, carotenoids travel from the blood vessels of the closed pennaceous vane is a consequence of dermal pulp into specific keratinocytes of the numerous diverse and covarying features at all feather germ has not been described (A. H. Brush, levels of the follicle and feather hierarchy, from pers. comm.). Although the capacity for patterning keratinocyte to follicle. For example, these details in carotenoid pigments within feathers is not as must include keratinocyte size and shape, precise as for melanin pigments, differential proximal and distal barbule differentiation, basi- distribution of carotenoid pigments within and lar-outer differentiation within barbule plates, among feathers of the body does demonstrate differentiation of the barbule plates from the 86 R.O. PRUM AND J. DYCK ramus, the helical growth of barb ridges, creation Plumage level metamodules of the rachis, subsequent barb ridge fusion to the In addition to the modular units of the plumage rachis, and the predetermined spatial relation- hierarchy, the plumage phenotype is composed of ships assumed by these components upon emer- additional metamodules that include multiple gence from the tubular sheath. feathers from different or multiple pterylae Although feathers with a closed pennaceous (Fig. 6). Examples include any group of feathers vane constitute only one of the many structural or parts of feathers that combine to form a classes of feathers, they are functionally, critically distinctly colored plumage patch, such as the wing important, given their role in flight, creating the bars of a wood warbler or the epaulets of a Red- contour of the body and the outward appearance of winged Blackbird (Fig. 5). The remiges and the plumage. rectrices covary in structure, shape, size, colora- Within-feather pigment pattern tion, and timing of molt to create the air foils that function in flight. Additional components of Another important feature of the pennaceous covariation among metamodulesFe.g., between feather vane is pigmentation patterning. There is the wings and the tailFcan create additional an enormous variety in within-feather pigment hierarchical relationship among metamodules. patterns that contribute importantly to the plu- Thus, the left and right sets of remiges, all mage phenotype. Prum and Williamson (2002) remiges, all rectrices, and all remiges and rectrices have provided the first unified models of the together (i.e., all flight feathers) comprise a development of within feather pigmentation pat- complex nested set of plumage metamodules terns. Their analyses indicate that barb ridge (Fig. 6). A complex example of a plumage identity plays no role in the determination of metamodule comes from the remiges and rectrices feather pigment patterning. Rather, it appears of the Golden-winged Manakin (Masius chrysop- that spatial position within the feather germ and terus), which are all black with brilliant yellow time during development determine whether inner vanes. These distinctly patterned flight particular keratinocyte cells will incorporate mel- feathers appear black when the bird is perched, anin pigments and how they will contribute to but they combine to create a brilliant visual overall feather pigment patterns. Applying stan- display when the wings and tail are spread dard reaction diffusion equations to a computer during the courtship display behavior (Prum and model of the growing feather germ, Prum and Johnson, ’87). Williamson (2002) documented that many feather Origin and evolution of plumage meta-modules pigment patterns can be simulated by hypothesiz- requires genetic covariation among the appropri- ing gradients of activating and inhibiting morpho- ate components of the plumage hierarchy, and gens within the feather germ and follicle. They likely subsequent selection on the emergent were also able to document detailed congruence functions of these plumage metamodules. Recent between the dynamics of the models and the research on the evolution of plumage patterning in documented variation in pigment patterning. orioles by Omland and Lanyon (2000) documents Experimental research on the molecular basis of the dynamic pattern in the evolution of the feather pigmentation patterning is required to test expression of pigment pattern metamodules with- these models. Traditional experiments on avian in this clade. Many distinctive plumage patches, embryos may be limited because of lack of strong such as wing bars, epaulets, hoods, tail patches, pigmentation patterning in most natal downs. etc.– have evolved convergently and been lost Although barb ridge identity itself is not critical to repeatedly within oriole phylogeny. This evolu- the establishment of feather pigment pattern, tionary pattern indicates that these plumage color other keratinocyte identities do play a crucial role. patch metamodules are plesiomorphic entities Thus, in many birds the distal barbules that cover that have been retained, redeployed, and replaced the outer (obverse) surface of the pennaceous frequently during the history of sexual and social feather vanes are more heavily pigmented than selection on oriole species. the proximal barbules (e.g., Strong, ’02). This detailed patterning in pigment deposition requires Molt and feather wear patterning mechanisms that distinguish between distal and proximal barbule plate cells that are Molt is the periodic replacement of feathers separated only by a single layer of axial plate cells within each follicle. Within this model, molt is the (Fig. 2D). renewal of an entire plumage through the growth HIERARCHICAL MODEL OF PLUMAGE 87 of serially homologous replacement feathers. Most plumage is composed of entirely cryptically co- feathers are molted once a year, but timing may lored, sexually monomorphic feathers of simple vary from once every two years to two to three shape. times a year. Normal molt maintains the func- Another effect that can change plumage appear- tional continuity and integrity of the plumage ance is feather wear. If the tips of the contour during the process of complete or partial plumage feather barbs are distinctly colored from the base replacement. Molt requires an elaborate coordina- of these barbs, then the gradual wear of the tion, or covariation, in developmental timing contour feathers can create a dramatic change in among feather follicles throughout the body. The plumage appearance without any molt. For exam- phenomenon of fright molt (or Schreckmauser), in ple, the male Bobolink (Dolichonyx oryzivorus; which all the feathers are dropped from their Icteridae) molts in the winter into a light brownish follicles within a few minutes of a severe dis- plumage, but as the tan tips of these feathers wear turbance (Dathe, ’55; Lucas and Stettenheim, ’72), off, a striking pattern of white rump and wing demonstrates that an important primary function coverts and black chest is revealed. These contour of the feather follicle is to retain the feather feathers may be structured for convenient break- between molts. (A Kansas ornithology student age at the point of pigmentation change. once observed entirely featherless chickens run- ning around minutes after a tornado hit a DISCUSSION AND CONCLUSION neighbor’s farm. Everyone concluded that the tornado had blown the feathers off.) The timing A hierarchical perspective on plumage develop- of feather replacement is under strict hormonal ment and evolution provides an important control. Maintenance of flight function during heuristic framework for understanding the diver- molt requires the replacement of remiges and sity and complexity of the plumage phenotype rectrices in a series without completely compro- (Figs. 3–6). This conceptual framework can be mising the function of the entire airfoil. Different used to generate new testable hypotheses and can groups of birds have evolved different sequences of itself be subjected to testing and revision as new feather replacement (Stresseman and Stresseman, data become available. ’66). The pattern of primary remige molt in Recent application of hierarchical modularity to cuckoos (Cuculidae) Freplacement of the pri- the evolutionary origin of feathers implies that maries in a series two alternating waves: # 9, 7, 5, this framework is realistic. Brush (’93, ’96) was 3, 1, 10, 8, 6, 4, and 2Fis the most strikingly the first to emphasize the hierarchical nature of complex wing molt pattern (Stresseman and feather development, which led him to recognize Stresseman, ’66). The simplest alternativeF and emphasize the numerous features that distin- simultaneous molt of all flight feathers with the guish feathers from plesiomorphic scales. Prum temporary loss of flight abilityFis a derived (’99) later proposed a developmental theory of the condition found only in waterfowl (Anseriformes). evolutionary origin of feathers which hypothesized The many complex details of molt sequence have that the causal hierarchy of feather ontogeny evolved by covariation of feather replacement could provide specific predictions about the se- among follicles from various parts of the plumage. quence of developmental novelties required for the More than annual molt coupled with indepen- evolution of complexly branched feathers. The dence of the plumages grown in different molts model predicted a transition series of feather can create distinctive plumages in different times morphologies from simple, undifferentiated tubu- of the year that may function in mate choice, lar appendages, to a basally branched tuft of intrasexual reproductive communication, or cryp- barbs, to a pennaceous vane, and other subse- sis depending upon the season. Thus, molts can quent structural complexity (Prum, ’99). create an entirely different plumage appearance, Recent discoveries of the diversity of feathers in or distinct alternating plumage modules that are non-avian theropod dinosaurs have confirmed expressed in only one of the annual plumages. The some of the Prum’s (’99) predictions about the male Ruff (Philomachus pugnax, Scolopacidae) morphology (Chen et al., ’98; Ji et al., ’98, 2001; Xu has a dramatic change in appearance in its two et al., ’99a, b, 2000, 2001), and the evolutionary annual plumages. The alternate (breeding) plu- polarity of primitive feathers (Padian, 2001; Sues, mage includes boldly patterned, sexually di- 2001; Ji et al., 2001; Xu et al., 2001; Prum and morphic feathers that form a conspicuous ruff Brush, 2002). Furthermore, a preliminary analysis around the head, whereas the basic (winter) of distribution of variation in feather morphology 88 R.O. PRUM AND J. DYCK among theropod lineages also provides support to more discussion on the relationship between the transition series, or evolutionary order, in feathers and scales.) primitive feather morphologies (Prum, ’99; Prum There is also genetic evidence for the existence and Brush, 2002). Support for the predictions of covariation among plumage modules in birds as of the hierarchical developmental model of hypothesized by the hierarchical model. As men- the feather origins supports the realism of the tioned above, the chicken plumage coloration locus hierarchical modularity of feathers that is the known as Columbian Restriction (Co) and other basis of the model proposed here. (Also refer to similar eumelanin inhibitors limit melanin deposi- Sawyer and Knapp, Chuong et al., and Homberger tion to feathers of the hackle, wing, tail, and feet in this issue for more discussion on feather (Smyth, ’90). Similarly, the White Crested breed of evolution.) duck has a novel modular tuft of longer feathers More recently, Harris et al. (2002) provided on the top of the head. Interestingly, this mutation strong molecular developmental support for has no other known phenotypic effects, but it does Prum’s model of feather evolution and for a raise egg mortality to nearly 25% (Metzer Farms, hierarchical conception of plumage complexity. pers. comm.). In both instances, it is known that Specifically Harris et al. (2002) document that novel metamodules within the plumage phenotype feather evolution preceded, through the repeated are determined by single genetic loci. Other co-option of a plesiomorphic molecular Shh-Bmp2 domestic breeds, such as Frizzle and Silky chick- module. It appears that the co-option of this ens (Somes, ’90), exhibit plumage phenotypes that plesiomorphic signaling module, with the inherent are entirely altered by single or few mutations. feature of decoupling of signal function at differ- These mutations with widespread effects indicate ent times of development, contributed directly to the potential of bottom-up effects in keratinogen- the hierarchical complexity of the plumage phe- esis or feather morphogenesis. notype. These advances in our understanding of The hierarchical model provides an intellectual feather development and evolution were facili- framework for organizing our current under- tated by a hierarchical conception of the plumage standing of the plumage phenotype, and for phenotype, and we anticipate that this formalized focusing attention on those subjects which deserve hierarchical model will provide further opportu- attention. Furthermore, the concise hierarchical nities for conceptual and experimental discoveries modularity of avian plumage indicates that avian in this area. plumage is an excellent model system for experi- Among the most fundamental insights estab- mental, comparative, and phylogenetic analysis of lished by the hierarchical model of the plumage modularity, hierarchy, and novelty in evolutionary phenotype is the evolutionary relationship be- developmental biology. It also raises the compel- tween feathers and scales– essentially the bound- ling question of why feathers are so hierarchically ary between the plumage phenotype and the total modular? The answer to this question would be an of all integumentary appendages. Brush (’93, ’96) interesting subject of study, but initial solutions emphasized the differences between the feather can be seen from several directions. As with other development of feather and scales, and questioned integumentary appendages, the iterative repeti- their homology. Prum (’99) emphasized that the tion of feather follicles over the body provides most fundamental feather noveltyFthe tubular numerous opportunities for independence and feather follicle and feather germFprovided a differentiation, but it is really the tubular novel definition of a feather, distinguishing feath- organization of the epidermal follicle and feather ers from scales. Harris et al. (2002) document that germFwith its inherent capacity for complex feathers, bird scales, and crocodilian scales all nested modular replication, the superimposed share detailed molecular homologies at the level of spatial polarities within and among modules, and the placode, but that all subsequent stages of the the continuous inductive and nutritional role of feather development occur through derived, novel the central dermal pulpFthat creates the genu- mechanisms of morphogenesis. The hierarchical inely unique hierarchical complexity of feathers. perspective on feather development and evolution has clarified the limited homology of scale and LITERATURE CITED feather placodes, leading to the obvious rejection Alibardi L. 2002. Keratinization and lipogenesis in epidermal of numerous traditional, elongate scale-based derivatives of the zebra finch, Taeniopygia guttata theories of the origin of feathers (Prum, ’99; Prum castanotis (Aves, Ploecidae) during embryonic development. and Brush, 2002). (Also see Sawyer and Knapp for J Morphol 251:294–308. HIERARCHICAL MODEL OF PLUMAGE 89

Auber L. 1971/72. Formation of ‘polyhedral’ cell cavities in Ortonne JP, editors The Pigmentary System: Physiology cloudy media of bird feathers. Proceedings of the Royal and Pathophysiology. New York: Oxford University Press. Society, Edinburgh. 74:27–41. p 107–114. Bleiweiss R. 1987. Development and evolution of avian racket Lecoin L, Lahav R, Dupin E, Le Douarin N. 1998. Develop- plumes: fine structure and serial homology of the wire. ment of melanocytes from neural crest progenitors. J Morphol 194:23–39. In: Chuong C-M, editors. Molecular Basis of Epithelial Bowers RR, Brumbaugh JA. 1978. An ultrastructural study of Appendage Morphogenesis. Georgetown, TX:R. G. Landes the regenerating breast feather of the fowl. J Morphol Co. 158:275–289. Lillie FR. 1940. Physiology of development of the feather. III. Brush AH. 1978. Feather keratins. In: Brush AH, editor. Growth of the mesodermal consituents and blood circulation Chemical Zoology. New York: Academic Press. p 117–139. in the pulp. Physiol Zool 13:143–175. Brush AH. 1993. The origin of feathers. In: Farner DS, King Lillie FR, Juhn M. 1938. Physiology of development of the JS, Parkes KC, editors. Avian Biology. London: Academic feather. II. General principles of development with Press. p 121–162. special reference to the after-feather. Physiol Zool 11: Brush AH. 1996. On the origin of feathers. J Evol Biol 9: 434–448. 131–142. Lillie FR, Wang H. 1941. Physiology of development of the Brush AH. 2000. Evolving a Protofeather and Feather feather. V. Experimental morphogenesis. Physiol Zool Diversity. Am Zool 40:631–639. 14:103–133. Chen P-J, Dong ZM, Zhen SN. 1998. An exceptionally well- Lucas AM, Stettenheim PR. 1972. Avian Anatomy- Integu- preserved theropod dinosaur from the Yixian formation of ment. Washington, DC: U.S. Dept. Agricult. Handbook. China. Nature 391:147–152. Maderson PFA, Alibardi L. 2000. The development of Chuong C-M, Chodankar R, Widelitz RB, Jiang RB. 2000. Evo- the sauropsid integument: a contribution to the problem devo of feathers and scales: building complex epithelial of the origin and evolution of feathers. Am Zool 40: appendages. Curr Opin Genet Dev 10:449–456. 513–529. Chuong C-M, Widelitz RB, Ting-Berreth S, Jiang TX. 1996. Mantulionis DH. 1970. Morphology of the developing down Early events during avian skin appendage regeneration: feathers of chick embryos. Z Anat Enwickl-Gesch 132: Dependence on epithelial-mesenchyme interaction and 107–157. order of molecular reappearance. J Invest Dermat Mu¨ller GB, Wagner GP. 1991. Novelty in evolution: restruc- 107:639–646. turing the concept. Ann Rev Ecol Syst 22:229–256. Cohen J, ’Espinasse PG. 1961. On the normal and abnormal Noralmy S, Morgan BA. 1998. BMPS mediate lateral inhibi- development of the feather. J Embryol Exp Morphol 9: tion at successive stages in feather tract development. 223–251. Development 125:3775–3787. Dathe H. 1955. U¨ ber die Schreckmauser. Journal fu¨r Omland KE, Lanyon SM. 2000. Reconstructing plumage Ornithologie 96:5–14. evolution in orioles (Icterus): Repeated convergence and Durrer H. 1986. The Skin of Birds: Colouration. In: Bereiter- reversal in patterns. Evolution 54:2119–2133. Hahn J, Matoltsky AG, Richards KS, editors. Biology of the Padian K. 2001. Cross-testing adaptive hypotheses: Phyloge- Integument 2: Vertebrates. Berlin: Springer-Verlag. netic analysis and the origin of bird flight. Am Zool 41: 239–247. 598–607. Dyck J. 1976. Structural colours. Proceedings of the Interna- Prum RO. 1999. Development and evolutionary origin of tional Ornithological Congress 16:426–437. feathers. J Exp Zoolog Part B Mol Dev Evol 285: ’Espinasse PG. 1939. The developmental anatomy of the 291–306. Brown Leghorn breast feather, and its reactions to oestrone. Prum RO, Brush AH. 2002. The evolutionary origin and Proceedings of the Zoological Society, London 109:247–288. diversification of feathers. Q Rev Biol 77:261–295. Greite W. 1934. Die Strukturbildung der Vogelfeder und ihre Prum RO, Johnson AE. 1987. Display behavior, foraging Pigmentierung durch Melanine. Zeitsschrift f wissenschaf ecology, and systematics of the Golden-winged Manakin Zoologie 145:284–334. (Masius chrysopterus). Wilson Bull 87:521–539. Harris MK, Fallon JF, Prum RO. 2002. A Shh-Bmp2 Prum RO, Williamson S. 2001. A theory of the growth and developmental module and the evolutionary origin and evolution of feather shape. J Exp Zoolog Part B Mol Dev diversification of feathers. J Exp Zool (Mol Dev Evol) Evol 291:30–57. 294:160–176. Prum RO, Williamson S. 2002. Reaction-diffusion models of Hosker A. 1936. Studies on the epidermal structures of birds. within feather pigmentation patterning. Proc R Soc Lond B Philosophical Transactions of the Royal Society, London, B Biol Sci 269:781–792. 226:143–188. Prum RO, Torres RH, Williamson S, Dyck J. 1998. Coherent Ji Q, Currie PJ, Norell MA, Ji S-A. 1998. Two feathered light scattering by blue feather barbs. Nature 396: dinosaurs from northeastern China. Nature 393:753–761. 28–29. Ji Q, Norell MA, Gao K-Q, Ji S-A, Ren D. 2001. The Prum RO, Torres RH, Williamson S, Dyck J. 1999. Two- distribution of integumentary structures in a feathered dimensional Fourier analysis of the spongy medullary dinosaur. Nature 410:1084–1088. keratin of structurally coloured feather barbs. Proc R Soc Jiang TX, Jung H-S, Widelitz RB, Chuong C-M. 1999. Self Lond B Biol Sci 266:13–22. organization of periodic patterns by dissociated feather Raff RA. 1996. The Shape of Life: Genes, Development and the mesenchymal cells and the regulation of size number and Evolution of Animal Form. Chicago: Chicago University spacing of primordia. Development 126:4997–5009. Press. Jimbow K, Sigiyama S. 1998. Melanosomal translocation and Sereno P. 1999. The evolution of dinosaurs. Science 284: transfer. In: Nordlund JJ, Boissy RE, Hearing VJ, King RA, 2137–2147. 90 R.O. PRUM AND J. DYCK

Sharlow ER, Paine CS, Babiariz L, Eisinger M, Shapiro S, Watterson RL. 1942. The morphogenesis of down feathers Seiberg M. 2000. The protease-activated receptor-2 with special reference to the developmental history of upregulates keratinocyte phagocytosis. J Cell Sci 113: melanophores. Physiol Zool 15:234–259. 3093–3101. Widelitz RB, Jiang T-X, Chen CJ, Stott NS, Jung H-S, Chuong Smyth JR, Jr. 1990. Genetics of plumage, skin, and eye C-M. 1999. Wnt–7a in feather morphogenesis: involvement pigmentation in chickens. In: Crawford RD, editor. Poultry of anterior-posterior asymmetry and proximal-distal elonga- Breeding and Genetics. Amsterdam: Elsevier. p 109–167. tion demonstrated with an in vitro reconstitution model. Somes RG, Jr. 1990. Mutations and major variants of Development 126:2577–2587. plumage and skin in chickens. In: Crawford RD, editor. Xu X, Tang Z-l, Wang X-l. 1999a. A therizinosauroid dinosaur Poultry Breeding and Genetics. Amsterdam: Elsevier. with integumentary structures from China. Nature 169–208. 399:350–354. Stettenheim P. 1976. Structural adaptation in feathers. Xu X, Wang X-l, Wu X-C. 1999b. A dromaeosaurid dinosaur Proceedings of the 16th International Ornithological with a filamentous integument from the Yixian Formation Congress 16:385–401. of China. Nature 410:262–266. Stresseman E, Stresseman V. 1966. Die Mauser der Vogel. Xu X, Zhou Z, Prum RO. 2001. Branched integumental Journal fu¨r Ornithologie 107:1–445. structures in Sinornithosaurus and the origin of feathers. Strong RM. 1902. The development of color in the definitive Nature 410:200–204. feather. Bulletin of the Museum of Comparative Zoology Xu X, Zhou Z, Wang X. 2000. The smallest known non-avian 40:147–185. theropod dinosaur. Nature 408:705–708. Sues H-D. 2001. Ruffling feathers. Nature 410:1036–1037. Yu M, Wu P, Widelitz RB, Chuong C-M. 2002. The Wagner GP. 2000. What is the premise of developmental morphogenesis of feathers. Nature 420:308–312. evolution? Part I: Why is developmental biology necessary to Ziswiler V. 1962. Die Afterfeder der Vo¨gel. Untersuchungen explain evolutionary innovations? J Exp Zool (Mol Dev Evol) zue Morphogenese un Phylogeneses de sogenannten After- 288:95–98. schaftes. Zool Jb Abt f Anat 80:245–308.