, Volume 96, Issue 1, pp 81–86

Feather-like development of Triassic diapsid skin appendages


  • Sebastian Voigt
    • Geologisches InstitutTechnische Universität Bergakademie Freiberg
    • Geologisches InstitutTechnische Universität Bergakademie Freiberg
  • Jan Fischer
    • Geologisches InstitutTechnische Universität Bergakademie Freiberg
  • Daniel Krause
    • Institut für GeowissenschaftenMartin-Luther-Universität Halle-Wittenberg
  • Robert Georgi
    • Geologisches InstitutTechnische Universität Bergakademie Freiberg
Original Paper

DOI: 10.1007/s00114-008-0453-1

Cite this article as:
Voigt, S., Buchwitz, M., Fischer, J. et al. Naturwissenschaften (2009) 96: 81. doi:10.1007/s00114-008-0453-1


Of the recent sauropsid skin appendage types, only feathers develop from a cylindrical epidermal invagination, the follicle, and show hierarchical branching. Fossilized integuments of Mesozoic diapsids have been interpreted as follicular and potential feather homologues, an idea particularly controversially discussed for the elongate dorsal skin projections of the small diapsid Longisquama insignis from the Triassic of Kyrgyzstan. Based on new finds and their comparison with the type material, we show that Longisquama’s appendages consist of a single-branched internal frame enclosed by a flexible outer membrane. Not supporting a categorization either as feathers or as scales, our analysis demonstrates that the Longisquama appendages formed in a two-stage, feather-like developmental process, representing an unusual early example for the evolutionary plasticity of sauropsid integument.


Fossilized integumentSkin appendageFeather evo-devoLongisquama


The only skeletal specimen of the diapsid Longisquama insignis was discovered in lacustrine deposits of the Triassic Madygen Formation and described by A. G. Sharov in 1970. It features a series of seven elongated skin projections attached in fan-like arrangement along the animal’s back. These have been ambivalently assessed either as modified scales with corrugated surfaces (Sharov 1970; Reisz and Sues 2000; Prum 2001; Unwin and Benton 2001) or pennaceous feathers with outwardly fused barbs (Jones et al. 2000). Given their complexity and transitional state, the appendages attracted some attention when feathers and reptilian skin came into the focus of evolutionary developmental biology (e.g., Chuong et al. 2003; Wu et al. 2004). The unsolved status of the skin structures and the lack of a comprehensive description and reconstruction after Sharov’s initial approach have motivated our study and the search for further fossil evidence.

Here, we provide a detailed analysis of Longisquama’s appendages involving the type series and the recent finds. We are addressing crucial aspects of their structure, including their alleged high-level homology with avian feathers. Following the description, we are demonstrating how the complex morphology of the skin projections constrains feasible developmental scenarios and derive a model of development that we compare to feather morphogenesis. As functional and phylogenetic interpretations have been linked by some authors (Jones et al. 2000; Martin 2004), we discuss the appendage function briefly at the end.

Materials and methods

In addition to the holotype PIN 2584/4 (PIN for Palaeontological Institute of the Russian Academy of Sciences) and four specimens with fragmentarily preserved isolated appendages, representing the paratype series (PIN 2584/5–7, -9), we introduce three more specimens of the elongated skin projections: FG 596/V/1, -2, and -3 (FG for Geological Institute of the TU Bergakademie Freiberg) have been recovered at the Madygen type locality in southwest Kyrgyzstan during fieldwork in 2007.

Considering the incompleteness of all 19 hitherto known appendages, our comprehensive reconstruction particularly involves FG 596/V/1 (Fig. 1a,b) of the new material, a single exemplar that is fully preserved over a length of 28.9 cm, exceeding the size of the other fragments by at least 100%. We are measuring the step-by-step change in morphological parameters, e.g., the distance of characteristic lines, along the middle axis of FG 596/V/1 (electronic supplementary material S1 and S2) and use these data for multivariate analyses in S-PLUS 6.0: Principal component analysis delivers the eigenvalue and the proportion of variance for each principal component (electronic supplementary material S3a). From these values, the degree of correlation of the measured morphological parameters can be inferred. Cluster analysis demonstrates the differentiability of subunits along the proximal–distal axis of the appendage (electronic supplementary material S3b).
Fig. 1

Appendage morphology. a Left slab of the specimen FG 596/V/1. Frame marks correspond to section figured in b. b Detail of the right slab of FG 596/V/1. ce Right slab of PIN 2584/5. Frame marks in c highlight the position of the detail figured in d. Abbreviations: ab anterior bar, af anterior flange, al anterior lobe, bl breakage trail, drl distal rear lobe, ml middle lobe, mx middle axis, pb posterior bar, pl posterior lobe, ru ruga, sc sediment core, sp spine. Scale bars: 2 cm


The dorsal appendages of Longisquama have a hockey-stick-like shape. Their continuous outer surface, interpreted as the outside of an enveloping membrane, surrounds a spacious interior, now filled with sediment (e.g., in FG 596/V/1, -3, PIN 2584/5, -7; sensu Reisz and Sues 2000). For the first time described by Sharov (1970), the conspicuous subdivision into a long and narrow proximal portion and a shorter distal portion, which is expanded anteroposteriorly and flattened laterally, is apparent from the longer exemplars (FG 596/V/1, PIN 2584/4, -7). While the proximal section of an appendage is longitudinally tripartite, consisting of a smooth anterior lobe, a corrugated middle lobe, and a smooth posterior lobe, the distal section has a bipartite morphology with two corrugated lobes separated by a distinct middle axis (Figs. 1 and 2a). The two distal lobes are transitional to the anterior and middle lobes of the proximal section. The proximal posterior lobe tapers off distally where only a narrow rim is delimiting the rear lobe (Sharov 1970). A broadening edge (anterior flange, Fig. 1c–e) originating in the distal third of the proximal section is running anteriorly up to the apex of the appendage (Sharov 1970; Haubold and Buffetaut 1987).
Fig. 2

Descriptive model and axial morphological transition. a Model appendage. b Depiction of characteristic lines and regularly-spaced cross sections (dashed lines) for FG 596/V/1 and, on that basis, diagram of proximal–distal variation (see electronic supplementary material S1 and S2). Parameters: 1 curvature (degree per cm × 5), 2 total width (mm), 3 and 4 percentage of the anterior and posterior lobe (%), 5 obliquity of rugae (degree), 6 density of rugae (number per cm × 10). I distal section, II proximal section. c Schematic longitudinal cross section of the proximal portion of FG 596/V/1 (Fig. 1b). The dashed contour line marks the position of the counter slab. Labels as in Fig. 1

In both lobes of the distal section, individual rugae of the slab and counter slab correspond in a convex–convex/concave–concave fashion, creating a series of more voluminous chambers and less voluminous interjacent zones. As noted by Reisz and Sues (2000), up to the distal end of the appendage, the rugae become narrower and more filigree mimicking barbs of a feather vane. In PIN 2584/5 (Fig. 1c–e), a fragmentary distal appendage of high fidelity, the middle axis forms a prominent roof-like ridge on the right slab and a deep furrow on the counter slab. The most striking feature of the distal section, preserved only in this specimen, is a series of about 20, up to 1.5-mm-long, spine-like impressions on the anterior lobe of the right slab. They root close to the middle axis and branch off obliquely (Fig. 1d, arrows). Corresponding structures on the counter slab are short, rounded impressions that run directly into the groove of the middle axis. We interpret them as the negative relief of rigid spines. If these were connected to the rod-like middle axis, they formed a single-branched internal frame. The insertion of the spines appears to be synchronous to the corrugation pattern, suggesting that the voluminous internal chambers and possibly intervening tissue enclosed them entirely in life.

The transition from the proximal to the distal section represents a major morphological changeover. For the specimen FG 596/V/1, covering both portions of an appendage, we measure the parameters curvature, total width, percentage of the anterior and posterior lobes, and tilting angle and density of the rugae in order to quantify the morphological change along the proximal–distal axis (Fig. 2b). These parameters are highly correlated and data points for the proximal and distal sections form distinct clusters (see electronic supplementary material S3). This is in agreement with a single switch-over in the development of a unidirectionally growing appendage. A characteristic feature related to the proximal–distal transition of the appendage is a narrow, slightly curved seam running from the posterior margin of the proximal section to the middle axis of the distal section, thereby crossing the rugae of the rear lobe (in PIN 2584/4–5, -7; FG 596/V/1, see Figs. 1a and 2a). This structure probably represents a breakage trail caused by the diagenetic compression of the appendage where it is transitional from a rather tubular, straight proximal to a flattened, increasingly curved distal section.

Subdivision of the proximal section into three lobes begins above a short, proximally tapering basis—a feature emphasized by Jones et al. (2000)—which appears to be attached right to the bone of the thoracal vertebral spines (in appendages no. 2–4 of PIN 2584/4). A slight convexity of the proximalmost section led to its interpretation as having a tubular shape (Jones et al. 2000; Reisz and Sues 2000; Prum 2001). This is in agreement with the proximally thicker sediment core in FG 596/V/1. Of the three longitudinal lobes in the proximal section, only the middle lobe, which accounts for 40–60% of the total width, bears a relief of rugae and deeper encarved interstices. Our analysis shows that the proximal rugae vary considerably in appearance: They are square-edged with a parallelepiped outline and merlon-like arranged in FG 596/V/1 (Fig. 1b), but roundish convex folds are found in PIN 2584/4 and PIN 2584/6, described by Sharov (1970) as “rosary-bead-like.” In all specimens, the long axes of the rugae are dipping caudoventrally transversal to the trend of the proximal–distal axis. The rugae run in and overlap with an anterior and posterior longitudinal bar forming, a latter-like pattern. This complex relief of the middle lobe occurs mirror-inverted on the left and right sides of the proximal section—rugae and interstices are synchronously arranged in a convex–convex/concave–concave fashion as in the distal section (Fig. 2c). The relief of the proximal section varies in distinctness, probably affected by the preservation of a smooth outer lamella. That such an envelope exists has been suggested by Jones et al. (2000) for the third appendage of the holotype in which a proximal cover layer with shallow relief only flimsily reproduces the topography of the underlying tripartite structure.


Following our description, Longisquama’s appendages possess several anatomical features that do not occur in recent elongate reptilian scales but are reminiscent of avian feathers and their developmental stages: (1) a proximal–distal differentiation with a single major morphological transition, (2) distinctive internal and external structures, (3) a complex internal organization with voluminous chambers and a branched frame in the distal section, and (4) a high length/proximal width ratio—up to 50 in FG 596/V/1. Thus, we are interpreting aspects of the appendage development in analogy to that of feathers (as characterized by Lucas and Stettenheim 1972; Widelitz et al. 2003; Yu et al. 2004): Their growth was unidirectional, requiring a clearly defined zone of cell proliferation. The appendages descended from a multilayered epidermal collar whose differentiation governed the shaping of the complex appendage topography. A switch from distal to proximal section marks the succession of distinct developmental phases analogous to the transition from feather vane to calamus. Furthermore, the deep anchoring and the probable tubular nature of the proximal section, also described by Jones et al. (2000), Reisz and Sues (2000), and Prum (2001), may be indicative for derivation from a cylindrical epidermal invagination, i.e., a follicle (Jones et al. 2000; Martin 2004). However, a cylindrical organization also occurs in nonfollicular appendages, such as the dorsal frill scales of recent iguanids (Wu et al. 2004).

In our developmental model for the distal section, the spines and tissue of the internal chambers, the anterior flange, and the middle axis develop from an interior layer, which undergoes partitioning and termination of intervening tissue while the enveloping membrane originates from an exterior layer. We are regarding the differentiation in the interior layer as a process analogous to barb and rachidial ridge formation in the ramogenic zone of a feather follicle (Lucas and Stettenheim 1972; Yu et al. 2002, 2004). Left–right axial polarization driven by morphogene gradients leads to the formation of the middle axis at the left or right end of the collar—we compare this to the effects of anterior–posterior polarization in feathers (Harris et al. 2005; Yue et al. 2005, 2006). At the opposing pole, inner-layer termination and outer-layer invagination increase the appendage outline at constant follicle diameter, thereby realizing the relatively thin and wide appearance of the distal section (Fig. 3a). During the eruption of the appendage, the distal section unfolds in a feather-like fashion. We are explaining the transversal rugae following the left–right axial trend as a consequence of helical organization in the follicle, a process responsible for the orientation of first-order branches in feathers (Lucas and Stettenheim 1972; Prum 1999; Yue et al. 2006). The transition from distal to proximal section materializes in a total reorganization of follicle topology. In the proximal section, no invagination occurs; it retains a tubular cross section. Its smooth anterior and posterior sectors, plus corrugated left and right sectors, have a tripartite appearance when being laterally compressed. The four sectors differentiated from an inner layer of the epidermal collar in continuation to the one that gave rise to the distal rugae and middle axis in an earlier developmental stage. However, unlike the distal condition, there is a proximal symmetry to the sagittal plane and polarization does not occur along the left–right axis (Fig. 3b). Possibly, individual pairs of proximal rugae represent stages in a cyclic growth mode of the proximal section, analogous to the calamus growth and step-by-step retreat of the dermal papilla during the resting phase of the feather development cycle (Lucas and Stettenheim 1972; Yu et al. 2004). In contrast to earlier suggestions (Sharov 1970; Jones et al. 2000; Martin 2004), we find no close morphological similarity of the proximal rugae to feather pulp chambers and caps. In the proximal-most section, the separation of lobes vanishes, and close to the root, the appendage tapers analogous to the proximal end of a feather calamus.
Fig. 3

Hypothetical follicular design. a Distal section. b Proximal section. Abbreviations: A anterior, L left, P posterior, R right, as anterior sector; caf chamber of the anterior flange, cdc constricted dermal cylinder, ch internal chamber, dc dermal cylinder, ls left sector, ps posterior sector, rs right sector. Other labels as in Fig. 1

Remaining discrepancies, in particular the flexible enveloping membrane and the lateral position of the middle axis, demonstrate the phylogenetic distance of Longisquama’s appendages to avian feathers. Considering the occurrence of derived and possibly follicular skin appendages in some archosaur groups, which are only distantly related to birds (Ji and Yuan 2002; Wang et al. 2002; Mayr et al. 2002), and the uncertain archosaurian relationship of Longisquama (as discussed by Unwin et al. 2000; Senter 2004), we agree with others that the dorsal appendages are unlikely to be homologous with avian feathers (Reisz and Sues 2000; Unwin and Benton 2001; Prum 2001; Prum and Brush 2002). The question of whether the convergent acquisition of a certain morphological feature, such as a tubular basis or branching, is sufficient to call the structures “feathers” has been raised and answered in the negative (Chuong et al. 2003). In agreement with Prum and Brush (2002), we advocate the restriction of the term “feather” to skin derivates regarded as homologous with avian feathers at least on the level of the feather follicle. Distinct from both, feathers and scales, Longisquama’s highly derived skin projections exemplify the plurality of sauropsid integument evolution.

The theory of an airborne Longisquama using its elongate appendages as gliding devices was introduced for the first time by Sharov (1970) and later adopted and adapted by others (Haubold and Buffetaut 1987; Jones et al. 2000; Martin 2004). It is rejected here for two reasons: (1) According to our and previous observations (Sharov 1970; Unwin et al. 2000; Unwin and Benton 2001), the holotype comprises only a single fanned-out row of appendages with no indication for the postmortem bending and loss of a second row, as assumed by Jones et al. (2000). (2) Unlike reptilian gliders, such as Coelurosauravus, Sharovipteryx, kuehneosaurids, and draco lizards, whose wing membranes are spanned close to the trunk and supported by limbs, ribs, or bony spines (Fig. 4a, see Schaller 1985), a gliding Longisquama with two rows of dorsal appendages would possess a continuous airfoil only distally where the expanded portions may partially overlap (Fig. 4b). With the bulk lift created far from the center of mass, the strain on the anchoring would be high, maximizing the risk of structural failure, especially if a flexible joint existed as in the reconstruction of Haubold and Buffetaut (1987). The orderly arrangement of appendages in PIN 2584/9 (Fig. 4c) has been interpreted as a “thoracic wing” by Martin (2004), but preservation effects or an alternative function can explain it as well: If the animal could fan out a single row of appendages by rotation in the sagittal plane (Fig. 4d), it might have used this mechanism only occasionally—e.g., for protection mimicry or for sexual display—most of the time, the appendages might have rested in a horizontal position. As in PIN 2584/9, they would overlap and form a series of successively smaller elements. Erection would have been realized by a system of longitudinal muscles attached to the deeply countersunk follicles.
Fig. 4

Functional models. a, b Simplified transversal sections of gliding Coelurosauravus (after the reconstruction of Schaumberg et al. 2007) and Longisquama. In black color: thoracal vertebra, ribs, and bony spines; dashed circle: heavily strained anchoring; FL lift, FG weight. c Line drawing of PIN 2584/9, representing a series of five appendages in a “wing-like” arrangement. Scale bar: 1 cm. d Rotation of appendages in the sagittal plane by traction of dorsal longitudinal muscles (gray)


This work was supported by the German Research Foundation (DFG II - VO 1466/1–1) and by the Society of Vertebrate Paleontology (Patterson Memorial Grant to JF). We are thankful to Evgenii N. Kurochkin and Vladimir R. Alifanov for access to the type material in Moscow in February 2007, to Susan Turner and Jörg W. Schneider for suggestions and comments, and to Ilja Kogan for his help with the translation of Russian publications and reports.

Supplementary material

114_2008_453_MOESM1_ESM.xls (116 kb)
Electronic supplementary material S1Complete data table. (XLS 115 KB)
114_2008_453_MOESM2_ESM.xls (80 kb)
Electronic supplementary material S2Table of data displayed in Fig. 2b. (XLS 79.5 KB)
114_2008_453_MOESM3_ESM.xls (84 kb)
Electronic supplementary material S3Multivariate analysis. (XLS 84.0 KB)

Copyright information

© Springer-Verlag 2008