REPORTS

A Diverse Assemblage of Late Cretaceous Dinosaur and Bird Feathers from Canadian Amber
Ryan C. McKellar,1* Brian D. E. Chatterton,1 Alexander P. Wolfe,1 Philip J. Currie2 The fossil record of early feathers has relied on carbonized compressions that lack fine structural detail. Specimens in amber are preserved in greater detail, but they are rare. Late Cretaceous coal-rich strata from western Canada provide the richest and most diverse Mesozoic feather assemblage yet reported from amber. The fossils include primitive structures closely matching the protofeathers of nonavian dinosaurs, offering new insights into their structure and function. Additional derived morphologies confirm that plumage specialized for flight and underwater diving had evolved in Late Cretaceous birds. Because amber preserves feather structure and pigmentation in unmatched detail, these fossils provide novel insights regarding feather evolution. lthough amber offers unparalleled preservation of feathers (14), only isolated specimens of uncertain affinity have been reported from the Late Cretaceous (5). This contrasts with the rich Early Cretaceous compression assemblage from northeastern China (68), leav-

insect inclusions (9). Eleven feather or protofeather specimens (10) were recovered by screening over 4000 Grassy Lake amber inclusions predominantly within the Royal Tyrrell Museum of Palaeontology (TMP) and University of Alberta (UALVP) collections. These fossils have disparate morphologies that span four evolutionary stages for feathers (11, 12). Specimens include filamentous structures similar to the protofeathers of nonavian dinosaurs that are unknown in modern birds (1315), as well as derived bird feathers displaying pigmentation and adaptations for flight and diving. The currently accepted (11, 12) evolutionarydevelopmental model for feathers (Fig. 1A) consists of a stage I morphology characterized by a single filament: This unfurls into a tuft of filaments (barbs) in stage II. In stage III, either some tufted barbs coalesce to form a rachis (central shaft) (IIIa), or barbules (segmented secondary
1 Department of Earth and Atmospheric Sciences, University of Alberta, Edmonton, Alberta T6G 2E3, Canada. 2Department of Biological Sciences, University of Alberta, Edmonton, Alberta T6G 2E9, Canada.

ing a substantial temporal gap in our understanding of feather evolution. Late Cretaceous amber from Grassy Lake, Alberta (late Campanian), is derived from lowland cupressaceous conifer forests that occupied the margin of the Western Interior Seaway and is best known for its diverse

*To whom correspondence should be addressed. E-mail: rcm1@ualberta.ca

Fig. 1. Feather evolutionary-developmental model (11), terminology (17), and stage I and II specimens from Canadian amber. (A) Feather stages outlined within text. Green, calamus or equivalent; blue, barbs; purple, rachis; red, barbule internodes; d.b., distal barbules; r., ramus; p.b., proximal barbules. (B) Field of filawww.sciencemag.org SCIENCE

ments cut obliquely (stage I), UALVP 52821. (C) Filament clusters variably oriented (stage II), UALVP 52822. (D) Close-up of (C), showing filaments that comprise clusters. Pigmentation coupled with comparatively thick outer walls produces darker color than in isolated filaments. Scale bars, (B) and (C) 1 mm, (D) 0.1 mm (10). VOL 333 16 SEPTEMBER 2011

1619

REPORTS
branches) stem from the barbs (IIIb); then, these features combine to produce tertiary branching (IIIa+b). Barbules later differentiate along the length of each barb, producing distal barbules with hooklets at each node to interlock adjacent barbs and form a closed pennaceous (vaned) feather (stage IV). Stage V encompasses a wide range of additional vane and subcomponent specializations. Most modern birds possess stage IV or V feathers or secondary reductions from these stages (11, 16). Modern feathers exhibit a range of morphologies that are associated with their various functions and remain discernible in some of their finest subunits, the barbules (17). This is particularly important in the study of amber-entombed feathers because preservation is biased toward feather subcomponents, which provide the basis for our morphological comparisons. Stage I is represented by UALVP 52821, which contains a dense forest of regularly spaced, flexible filaments with a mean diameter of 16.4 T 4.2 mm (Fig. 1B and figs. S1 to S4). Filaments are hollow with the internal cavity comprising Fig. 2. Specialized barbules in Canadian amber. (A) Coiled barbules surrounding thickened rachis (arrow), cut obliquely, TMP 96.9.334. (B) Closeup of coils in isolated barbule. (C) Semi-flattened internodes and weakly expanded node of (A). Diffuse, variable barbule pigmentation produces pale overall color. (D) Isolated barb with differentiated barbules and thickened ramus, in spiders web, UALVP 52820. (E) Barbules near distal tip of (D), with clearly defined distal and proximal barbule series (left and right sides of ramus, respectively). (F) Closeup of distal barbule in (E), showing nodal prongs and ventral tooth on basal plate (arrow) adjacent to abrupt transition into pennulum. Banded pattern of dark pigmentation within basal plate, and diffuse dark pigmentation within pennulum, suggest a gray or black feather (24). Scale bars, (A) 0.4 mm; (B), (D), and (E) 0.2 mm; (C) and (F) 0.05 mm (10). ~ 60% of total diameter, have no obvious internal pith, and taper apically. Where surface texture is observable, filaments bear a faint crosshatching pattern but lack surface topography. The filaments are not plant or fungal remains because they lack cell walls and are relatively large. Comparatively small diameters and a lack of cuticular scales imply that they are not mammalian hairs, as does direct comparison to a hair from this amber deposit. Their closest morphological match is the filamentous covering found of nonavian dinosaurs such as the compsognathid Sinosauropteryx prima (18). The amber-entombed specimens are slightly finer than those of Sinosauropteryx, which may have been distorted by compression and permineralization. The amber filaments display a wide range of pigmentation, ranging from nearly transparent to dark (fig. S2). No larger-scale color patterns are apparent. [Additional specimen details are provided in supporting online material (SOM) text.] The stage II morphotype (Figs. 1, C and D, and fig. S5) consists of tightly adpressed clusters approximately 0.2 mm in width and composed of filaments that are otherwise similar to those already discussed (10). Five clusters are preserved together in UALVP 52822. As in stage II primitive feathers (11), filaments in each cluster appear to diverge from a common basal region without branching, but no rachis is visible where the clusters exit the amber. These filaments bear some resemblance to fibrils that compose pycnofibers (tufted filaments) in pterosaur compression fossils (19), except the amber specimens lack the secondary organization observed in pycnofiber bundles. The most morphologically comparable compression fossils are protofeathers associated with the dromaeosaurid Sinornithosaurus millenii (10, 20). These clusters exhibit generally comparable sizes and shapes to the amber specimens and even have the more loosely bundled appearance distally where individual filaments have more variable lengths. In contrast to stages I and II, additional specimens from Canadian amber have barbules specialized for discrete functions. In TMP 96.9.334 (Figs 2, A to C, and figs. S6 and S7) (10), a thick-

1620

16 SEPTEMBER 2011

VOL 333

SCIENCE

www.sciencemag.org

REPORTS
ened rachis is surrounded by numerous barbules with tightly coiled bases. The barbules undergo three or more complete whorls and are composed of semi-flattened internodes (~120 mm long, 9 mm wide) separated by weakly expanded nodes (~12 mm wide). This coiling cannot be attributed to interaction between barbules and resin during amber polymerization because it only occurs at the base of each barbule. Modern seedsnipes and sandgrouse (21, 22) possess belly feathers with similar basal barbule coiling, which allows water to be retained for transport to the nest for distribution to nestlings or for cooling incubating eggs. Grebes also have coiled barbules that absorb water into plumage, facilitating diving by modifying buoyancy, reducing hydrodynamic turbulence, and improving insulation (23). In all of these instances, the keratin of coiled barbules interacts with water to uncoil and absorb water through capillary action (22). The high number of coils in TMP 96.9.334 is most similar to that reported from grebes (23, 24), implying that the Cretaceous barbules are related to diving behavior. Barbules displaying all characteristics necessary for forming vaned feathers are also present in Canadian amber (Fig. 2, D to F, and fig. S8)

Fig. 3. Pigmentation in Canadian amber feathers. (A to D) Semi-pennaceous feathers, TMP 96.9.997: (A) six barbs; (B) close-up of box in (A), arrow indicates unpigmented ramus; (C) detail of ramus and barbule bases; (D) dark-field microphotograph of (C), showing brown coloration with ramus and basal internodes minimally pigmented. Density and distribution of pigments (24, 25) are consistent with medium- to dark-brown modern feathers. (E) Unpigmented downy barbules, TMP 79.16.12. (F to K) Poorly differentiated, flattened barbules: (F) partial overview of 16 pennaceous barbs, TMP 96.9.553; (G) close-up of (F), www.sciencemag.org SCIENCE

showing variable, diffuse pigmentation within barbule bases (ventral plates translucent, dorsal flanges pigmented); (H) unpigmented, isolated barb with juvenile mite, TMP 96.9.546; (I) central portion of isolated barb, TMP 94.666.15; (J) dark-field microphotograph of (I), showing overall color; (K) banded pigmentation within basal plate of proximal barbule in (I), indicating 5 to 6 component internodes. (L) Reduced pennaceous barbs from non-interlocking region of dark brown and white mottled chicken contour feather for comparison. Scale bars, (A) 0.5 mm; (B), (E), (F), (H) to (J), and (L) 0.2 mm; (C), (D), (G), and (K) 0.04 mm. VOL 333 16 SEPTEMBER 2011

1621

REPORTS
(10). These were probably borne by an animal capable of flight. Within UALVP 52820, barbules of unequal lengths arise from either side of the barb, producing a differentiated series of longer proximal (~ 0.42 mm) and shorter distal (~ 0.24 mm) elements, all having spinose nodal projections. Barbules are widely spaced along a thick ramus (barb shaft) adapted for rigidity and are strongly differentiated to interlock with adjacent barbs to form a vane (10). On the basis of the presence of a rachis in TMP 96.9.334 and differentiated barbules in UALVP 52820, these specimens can be assigned conservatively to stages IV and V and are attributed to Late Cretaceous birds. The remaining six feathers are fragmentary downy and contour feathers (Fig. 3). Although they offer limited insight concerning the identity or behavior of their bearer, their structure and pigmentation bear directly on feather evolutionary stages. Four of the six feather fragments in TMP 96.9.997 (Fig. 3, A to D, and fig. S9) are aligned. Superficially, these exhibit an intermediate morphology (stage IIb) proposed for an Early Cretaceous (late Albian) French amber specimen (4). In the Canadian specimens, as in the French material, the main axis preserved is short (3.7 mm) and weakly defined, dorsoventrally flattened, and composed of fused secondary branches in an opposite arrangement. However, in the Canadian specimens intense pigmentation in each internode produces a beaded appearance, highlighting segmentation that is otherwise difficult to discern based on barbule diameter variation (Fig. 3C). Segmentation identifies the finest branches as barbules attached to narrow rami, and not barb equivalents attached to a rachis. This interpretation identifies these small specimens as subcomponents of a larger feather, such as basal barbs on a contour feather (17), and not a distinct stage in feather evolution lacking barbules (4). This interpretation probably extends to the French material as well. Pigmentation is preserved with fidelity in all additional specimens. Although downy feathers are consistently transparent, and would have been white in life, pennaceous feathers are more variable, with diffuse, transparent, and mottled patterns of pigmentation (Fig. 3, E to L) that match those observed in modern birds (10, 24, 25). Although neither avian nor dinosaurian skeletal material has been found in direct association with amber at the Grassy Lake locality, fossils of both groups are present in adjacent stratigraphic units. Hadrosaur footprints are found in close association with the amber, and younger (late Campanian and Maastrichtian) strata of western Canada contain diverse nonavian dinosaur (26) and avian (27, 28) remains. There is currently no way to refer the feathers in amber with certainty to either birds or the rare small theropods from the area (26). However, the discovery of endmembers of the evolutionary-developmental spectrum in this time interval, and the overlap with structures found only in nonavian dinosaur compression fossils, strongly suggests that the protofeathers described here are from dinosaurs and not birds. Given that stage I filaments were present in densities relevant for thermoregulation and protection, and that comparable structures are preserved as coronae surrounding compression fossils, it becomes apparent that protofeathers had important nonornamental functions. Specialized barbule morphologies, including basal coiling, suggest that Campanian feather-bearers had already evolved highly specialized structures similar to those of modern grebes to enhance diving efficiency. Canadian amber provides examples of stages I through Vof Prums (11) evolutionary-developmental model for feathers. None of the additional morphotypes observed in compression fossils of nonavian dinosaurs (8, 15) or amber (4) were found here, suggesting that some morphotypes may not represent distinct evolutionary stages, or may not have persisted into the Late Cretaceous. The snapshot of Campanian feather diversity from Canadian amber is biased toward smaller feathers, subcomponents of feathers, feathers that are molted frequently, and feathers in body positions that increase their likelihood of contacting resin on tree trunks. Despite these limitations, the assemblage demonstrates that numerous evolutionary stages were present in the Late Cretaceous, and that plumage already served a range of functions in both dinosaurs and birds.
References and Notes
1. P. G. Davis, D. E. G. Briggs, Geology 23, 783 (1995). 2. D. Schlee, W. Glckner, Stuttg. Beitr. Naturkd. C 8, 1 (1978). 3. J. Alonso et al., J. Paleontol. 74, 158 (2000). 4. V. Perrichot, L. Marion, D. Nraudeau, R. Vullo, P. Tafforeau, Proc. Biol. Sci. 275, 1197 (2008). 5. D. Grimaldi, G. R. Case, Am. Mus. Novit. 3126, 1 (1995). 6. Q. Ji, P. J. Currie, M. A. Norell, S.-A. Ji, Nature 393, 39 (1998). 7. M. Norell, X. Xu, Annu. Rev. Earth Planet. Sci. 33, 277 (2005). 8. X. Xu, Y. Guo, Vertebrata PalAsiatica 47, 311 (2009). 9. R. C. McKellar, A. P. Wolfe, in Biodiversity of Fossils in Amber from the Major World Deposits, D. Penney, Ed. (Siri Scientific Press, Manchester, 2010), pp. 149166. 10. Materials and methods are available as supporting material on Science Online. 11. R. O. Prum, J. Exp. Zool. 285, 291 (1999). 12. R. O. Prum, A. H. Brush, Q. Rev. Biol. 77, 261 (2002). 13. A. H. Brush, Am. Zool. 40, 631 (2000). 14. P. J. Currie, P.-J. Chen, Can. J. Earth Sci. 38, 1705 (2001). 15. X. Xu, X. Zheng, H. You, Nature 464, 1338 (2010). 16. P. Stettenheim, Living Bird 12, 201 (1973). 17. A. M. Lucas, P. R. Stettenheim, Avian Anatomy: Integument (U.S. Department of Agriculture, Washington, DC, 1972). 18. P.-J. Chen, Z.-M. Dong, S.-N. Zhen, Nature 391, 147 (1998). 19. A. W. A. Kellner et al., Proc. Biol. Sci. 277, 321 (2010). 20. X. Xu, Z.-H. Zhou, R. O. Prum, Nature 410, 200 (2001). 21. T. J. Cade, G. L. Maclean, Condor 69, 323 (1967). 22. G. L. MacLean, Bioscience 33, 365 (1983). 23. J. Fjelds, The Grebes: Podicipedidae (Oxford Univ. Press, New York, 2004). 24. A. C. Chandler, Univ. Calif. Publ. Zool. 13, 243 (1916). 25. C. J. Dove, Ornith. Mono. 51, 1 (2000). 26. P. J. Currie, in Dinosaur Provincial Park: A Spectacular Ancient Ecosystem Revealed, P. J. Currie, E. B. Koppelhus, Eds. (Indiana Univ. Press, Bloomington, 2005), pp. 367397. 27. N. Longrich, Cretac. Res. 30, 161 (2009). 28. E. Buffetaut, Geol. Mag. 147, 469 (2010). Acknowledgments: We thank the Leuck family and M. Schmidt (donated specimens); M. Caldwell, S. Ogg and M. Srayko (microscopy); E. Koppelhus and H. Proctor (discussions); and J. Gardner, B. Strilisky, A. Howell, and J. Hudon (TMP, Redpath Museum, and Royal Alberta Museum collections). Research was funded by Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Grants to B.D.E.C., A.P.W., and P.J.C. and NSERC and Alberta Ingenuity Fund support to R.C.M.

Supporting Online Material

www.sciencemag.org/cgi/content/full/333/6049/1619/DC1 Materials and Methods SOM Text Figs. S1 to S12 References (2949) 25 January 2011; accepted 22 July 2011 10.1126/science.1203344

Trace Metals as Biomarkers for Eumelanin Pigment in the Fossil Record

R. A. Wogelius,1,2* P. L. Manning,1,2,3 H. E. Barden,1,2 N. P. Edwards,1,2 S. M. Webb,4 W. I. Sellers,5 K. G. Taylor,6 P. L. Larson,1,7 P. Dodson,3,8 H. You,9 L. Da-qing,10 U. Bergmann11 Well-preserved fossils of pivotal early bird and nonavian theropod species have provided unequivocal evidence for feathers and/or downlike integuments. Recent studies have reconstructed color on the basis of melanosome structure; however, the chemistry of these proposed melanosomes has remained unknown. We applied synchrotron x-ray techniques to several fossil and extant organisms, including Confuciusornis sanctus, in order to map and characterize possible chemical residues of melanin pigments. Results show that trace metals, such as copper, are present in fossils as organometallic compounds most likely derived from original eumelanin. The distribution of these compounds provides a long-lived biomarker of melanin presence and density within a range of fossilized organisms. Metal zoning patterns may be preserved long after melanosome structures have been destroyed.

eather color in birds stems mostly from chemical pigments, of which the most widely used are melanins (1). Resolving VOL 333 SCIENCE

color patterns in extinct species may hold the key to understanding selection processes that acted during crucial evolutionary periods and

1622

16 SEPTEMBER 2011

www.sciencemag.org

www.sciencemag.org/cgi/content/full/333/6049/1619/DC1

Supporting Online Material for

A Diverse Assemblage of Late Cretaceous Dinosaur and Bird Feathers from Canadian Amber
Ryan C. McKellar,* Brian D. E. Chatterton, Alexander P. Wolfe, Philip J. Currie

*To whom correspondence should be addressed. E-mail: rcm1@ualberta.ca

Published 16 September 2011, Science 333, 1619 (2010) DOI: 10.1126/science.1203344 This PDF file includes: Materials and Methods SOM Text Figs. S1 to S12 References (2949)

Materials and Methods Established methods were employed for the collection and preparation (29) of amber inclusions. Epoxy-embedded amber nodules were slide-mounted and polished, and cover slips were applied to optimize views and ensure long-term preservation of the inclusions. Total slide thickness ranged from 1.8 mm to 8.5 mm, with the thickest mounts at times challenging the resolving power of compound microscopy. A suite of modern bird feathers and hair samples were directly compared to the amber-entombed specimens, as were morphological atlases on the microscopic structure of mammalian hairs (30, 31) and feathers (24, 25). Modern comparative specimens were either epoxy-embedded or examined unaltered, depending on the degree of magnification required. All specimens were photographed using a Canon PowerShot A640 camera attached to a Zeiss Stereo Discovery.V8 microscope, or Zeiss Axio Imager.A1 compound microscope (b.f. denotes bright field photographs, d.f. denotes dark field photographs). Images usually encompass multiple focal planes and were compiled using Axiomat or Helicon Focus software. All measurements were taken either digitally using Axiomat, or on a Wild M5 dissecting microscope equipped with an ocular micrometer. The inherent limitations of working with amber governed our approach to the Canadian amber specimens, and consequently we focused our work on morphological comparisons and morphometric analyses. The nature and rarity of these specimens precludes destructive sampling until additional specimens are recovered. Potentially contentious specimens, such as the Stage I and II morphotypes, were subjected to additional non-destructive sampling. Spinning disk confocal microscopy (SDCM) and laser scanning confocal microscopy (LSCM) were utilized. SDCM data were obtained using a Hamamatsu Orca R2 camera on an inverted Olympus IX81 microscope with a Yokogawa CSU-10 spinning disc confocal head (examining excitation at 491 nm and 561 nm). LSCM data were obtained with a Leica SP5 microscope using a 20x 0.5 Na objective and acousto-optical tunable filters (examining excitation at 405 nm). Results of these analyses, as well as additional morphological details on all specimens are presented here. Institutional abbreviations: TMP, Royal Tyrrell Museum of Palaeontology, Drumheller, Alberta, Canada; RAM, Royal Alberta Museum, Edmonton, Alberta; RM, Redpath Museum, McGill University, Montreal, Quebec, Canada, modern bird collection; UALVP, University of Alberta Laboratory of Vertebrate Palaeontology, Edmonton, Alberta.

SOM Text Additional details of Stage I and II morphotype identifications A major concern regarding the specimens identified as Stage I or II morphotypes is whether other possible interpretations are tenable. Within the main text we briefly summarize these possibilities and the bases for their rejection. Here we provide full details of the work underpinning our conclusions. Comparison to modern mammalian hairs In general, the Stage I and II morphotypes reported are of a smaller diameter than most mammalian hairs and do not appear to possess cuticular scales. More specifically, the filaments measured from UALVP 52821 have a mean diameter of 16.44.2 m (n=80), with minimum and maximum diameters of 6.2 m and 27.1 m, respectively. In UALVP 52822, filaments have a mean diameter of 17.95.0 m (n=28), and range between 10.7 m and 31.0 m. The UALVP 52822 filaments are loosely bundled into five distinct clusters. The three clusters that have definite edges and appear to represent a complete cross-section of the bundle measure 213 m, 233 m and 325 m in diameter at their narrowest. The diameters observed for Stage I and II filaments therefore fall just within the lowest range of values known for modern mammal hair. Mammal hair has been studied extensively, and the two main types that have been documented across a wide range of taxa, with attention to both overall diameter and cuticular scale patterns, are underhairs (understory fur) and guard hairs. Given that the Stage I and II filaments overlap with only the finest known mammal hairs, and furthermore given differences between modern and Cretaceous mammalian faunas, we conducted detailed comparisons to pelages that represent both the smallest known underhair diameters (30) and contain the widest taxonomic range of organisms, including numerous marsupials (31). Because the latter work was based mainly upon guard hairs (typically of slightly larger diameter than underhairs (31), but more likely to enter in contact with tree resin), measurements were taken from the narrowest part of each exemplar. Underhair diameters listed for 162 species of mammals (30) yielded a mean value of 59.582.3 m, ranging from 6.8 m to 680 m. Measurements of guard hair diameters for 75 species of Australian mammals (31) yielded a mean value of 48.237.8 m, a minimum of 9.4 m, and a maximum of 168 m. Although these samples clearly display a wide range of diameters, all modern specimens within the low end of the spectrum were united by two morphological features. In almost all cases of diameters below 25 m, the medulla (hollow core) of the hair was discontinuous, being subdivided along its length into either a uniserial ladder or aeriform lattice arrangement (30, 31). This was typically observed in conjunction with coarse, diamond-shaped cuticular scales arranged with a maximum of two to three scales fitting within one hair-width and resulting in a jagged margin on the hair when viewed in longitudinal section (30, 31). Stage I and II filaments differ markedly from this arrangement. The filaments are hollow, with an outer wall that comprises approximately 40% of the total diameter, and is further reduced within apical portions of the filaments. The hollow nature of the filaments is best illustrated in UALVP 52821, where patchy translucency and broken edges demonstrate that the filaments have a circular cross3

section, and that their cores are hollow (Figs. S3, S4A, S11; see also SOM pp. 68, regarding specimen LSCM analyses and taphonomic considerations). Within areas where the filaments are preserved as nearly opaque masses due to darker pigmentation, they nonetheless preserve faint cross-hatching of very fine light and dark spots (Figs. S4BD). Within areas where the filaments are translucent, the outer wall clearly does not possess a jagged margin, which would be clearly observed if cuticular scales were present. Comparison to fossil mammalian hair In Canadian amber, there is currently one hair fragment known (TMP 96.9.998). This specimen (Figs. S4E, F) is in the process of being studied and described, but our preliminary analysis already indicates a number of distinctions between it and the Stage I and II morphotypes described herein. The hair fragment is significantly wider than any of the filaments preserved (approximately 56 m in diameter) and reveals faint indications of fine, closely-spaced cuticular scales when viewed with dark-field microscopy. Additional observations suggest that the specimen lacks a broad medullary cavity and that the medulla is likely discontinuous in either an aeriform lattice or multiserial ladder pattern, once again in contrast to any of the Stage I and II filaments reported. Additional Mesozoic fossil hair specimens from the Early Cretaceous of France include two fragments preserved in three dimensions within amber (32). These specimens have observed diameters that range from 3248 m and from 4978 m, respectively, and possess cuticular scales that are smoothly undulate with an intermediate spacing (32). Preservational characteristics of the hair fragments described by these authors are similar to those observed for both hair and the Stage I morphotype filaments from Canadian amber. Comparison to fungal and plant remains In general, the Stage I and II morphotypes reported can be differentiated from plant and fungal remains based upon their comparatively large size, lack of septae, and preservational characteristics. Most fungal hyphae branch and exhibit a diameter range from 115 m, but the known range extends from 0.5 m to 1 mm (33). Cell walls in hyphae are generally thin (often 0.2 m or less), with chitin as the main structural component (34). In amber, this combination of features typically results in filamentous fungi that are easily observed as mycelia (larger mats of hyphae). These appear vitreous or white when examined under reflected light (Fig. S4G). Conceivably, groups such as the Zygomycetes (bread moulds) could produce coenocytic hyphae (those lacking internal septae) of similar overall morphology to UALVP 52821. However, a subparallel, non-branching, centimeter-scale series of such hyphae lacking any adventitious septae or terminal sporangia seems highly improbable (35). Furthermore, UALVP 52821 displays pigmentation and an outer wall thickness that do not match the preservational characteristics of fungi within this amber deposit. Many of the characteristics that separate Stage I and II morphotypes from fungal remains also distinguish them from plant remains. The Stage I and II morphotypes exhibit no evidence of longitudinal subdivision within their hollow cores, and their diameters are roughly twice to thrice those of xylem cells found in the deposit. Furthermore, xylem cells are typically polygonal in cross-section and when encountered in Canadian amber, are typically present as adjoined series of cells that form blocky fragments of tissue that 4

have been carbonized or perhaps fusainized. Sclerenchyma fibers (commonly referred to as bast or plant fibers) are the most likely component of woody plants to exhibit the general shape, size, lack of pitting, thickened outer walls, and undivided elongate forms (36) observed in the amber filaments. Although some sclerenchyma fibers used in textiles have comparable mean diameters to those observed in Stage I and II filaments, these are never heavily pigmented, are nearly pentagonal or hexagonal in cross-section with uniformly thick walls, taper at both apices, and exhibit a wide array of apicular morphologies (36, 37). Furthermore, the plant remains we have recognized in Cretaceous ambers from western Canada (9), particularly those that breach the surface of their encapsulating amber nodule, are generally preserved as carbonized remains that preserve little surface detail at the cellular level. Comparison to degraded or taphonomically-altered feather remains An alternate interpretation of the Stage II morphotype we describe is that it represents a series of degraded feather rachi that have decayed to the point of exposing their internal filamentous structure. The morphology of such structures has recently been explored (38) through biodegradation, using keratin consuming fungi. This has revealed the underlying structure of the rachis, indicating that filaments that once comprised rachi bear distinct nodes directly comparable to those of barbules, quite unlike the filaments recovered from amber. The inferred Stage II clusters could also be construed as a result of poorly-preened feathers, in which the barbs have clumped together. Although this alternative is more difficult to discount, we note that, unlike typical barbs, the filaments that comprise the Stage II clusters we describe possess circular cross-sections, in absence of any indication of a rachis from which they could have originated. Comparison to pterosaur pycnofibers Pycnofibers are bushy fibers found in association with pterosaur remains: these have an average diameter between 0.2 and 0.5 mm, and are apparently composed of finer fibrils of unknown original structure or composition (19). Compared to UALVP 52821, there is an overlap in the known diameters of the clusters, and they both appear to have sub-centimeter lengths. In the case of pycnofibers, the component fibrils appear to be much more tightly bound, particularly near the apex of the pycnofiber, which makes their distinction much more difficult than the loosely-bound Stage II filaments observed in amber. Sinosauropteryx prima comparison In terms of compression fossils, the Stage I morphotype filaments observed in Canadian amber are most comparable to protofeathers from Sinosauropteryx prima. The integumentary structures of S. prima display a range of lengths, from ~4 mm to at least 4.0 cm, depending on the specimen and their body position (14, 18). These independent filaments range in thickness from easily observed 0.2 mm filaments to those that are considerably smaller than 0.1 mm (14). The filaments are hollow and round in crosssection (39) and may have been secondary branches of larger structures (14) or isolated filaments (12). Although the UALVP 52821 specimen does not display filaments with diameters as large as the maximum reported from S. prima, they are consistent with the finer filaments found in this specimen, and fall within the range of observed lengths. As 5

with the filaments from S. prima, UALVP 52821 filaments are hollow with circular cross-sections. It must be noted that compression, permineralization, and lack of definition may all have contributed to some degree of distortion of the original dimensions of filaments associated with S. prima. Compression potentially flattens otherwise cylindrical filaments, whereas permineralization may increase the apparent thickness of the outer wall. The lack of definition between individual filaments in S. prima may also yield overestimates of original filament thicknesses. Sinornithosaurus millenii comparison The UALVP 52822 clustered filaments described as a Stage II morphotype are most similar to compression fossils surrounding Sinornithosaurus millenii. In S. millenii, although there is no direct evidence of a rachis (as with the amber specimens), barbules are clearly clustered into independent tufts with compressed widths of 13 mm and lengths of up to 4.5 cm (12, 20). These clustered filaments appear to have been attached basally, or in one example, inferred to have arisen from a central rachis (12, 20). Although no direct measurements of the filaments that comprise each cluster have been presented by Xu et al (20) they appear to be of sub-millimeter diameter similar to the filaments observed in amber. As in the Sinosauropteryx prima protofeathers, the clusters found with Sinornithosaurus millenii are likely to have expanded diameters as a result of filament splaying during compression. Their displacement from the body suggests that the clusters associated with S. millenii were not immediately buried (20), so the main limitations on the degree of filament splaying would have been the length of time the clusters were allowed to decay, and the rigidity with which the filaments were fixed in the clusters. Additional morphological observation techniques Due to the current rarity of specimens, destructive sampling is not possible with the Canadian amber material (including crack-out studies utilizing scanning electron microscopy). Synchrotron x-ray microtomography has recently demonstrated great promise for studying small-scale inclusions within amber. This imaging technique has demonstrated unmatched resolution of fine structures (40), yet has been unsuccessful in the analysis of amber-entombed hair specimens (32) comparable to the Stage I and II filaments described here, likely as a result of low density contrast. This leaves, beyond light microscopy, confocal microscopy as the primary source of additional data on the Canadian amber specimens (described below). Chemical comparison to mammalian hairs As mammalian hairs constitute the most similar structures in terms of both overall morphology and preservational characteristics, we sought additional analyses to compare the chemical composition of the Stage I and II morphotypes to hair. The identification of -keratin or -keratin in the putative protofeathers would provide strong support for our structural inferences, because these proteins are specific to the integumentary structures of mammals and reptiles, respectively. The presence of -keratin has been demonstrated in filaments associated with the non-avian theropod Shuvuiia deserti through the use of immunohistochemical responses, measured utilizing -keratin specific antibodies that were tagged with fluorescent markers and subjected to LSCM (41). Such testing is, at 6

present, impossible for specimens such as UALVP 52821 and UALVP 52822 because they do not provide enough volume for analysis, and furthermore cannot be dissociated from the entombing amber matrix. Moreover, the gymnospermous resin has permeated filaments during amber polymerization, which is problematic for such analyses because it is autofluorescent, impermeable, highly insoluble, and contains trace quantities of various amino acids of botanical origin (42, 43). Finally, Canadian amber is not readily sectioned as it fractures conchoidally. Taken together, these characteristics temper our expectations for successful immunohistochemical analyses of amber-borne filaments at present, should additional specimens be located to allow destructive sampling. Fortunately, we can interrogate this issue non-destructively with confocal microscopic approaches. Analysis by LSCM and SDCM Given these caveats, we turned to LSCM and SDCM to assess the composition of the filaments. Keratin is known to autofluoresce with a predictable emission profile (44). This makes possible a comparison of fluorescence patterns amongst UALVP 52821, UALVP 52822, and unambiguous feather fragments within the deposit. Ideally, differences between the excitation and emission profiles of the specimens would permit comparison between these specimens, as well as a wider range of inclusions within the deposit, in order to rule out conclusively the alternative origins for the filaments discussed above. UALVP 52821 was compared to TMP 96.9.997 with both SDCM and LSCM. TMP 96.9.997 is both strongly-pigmented and has completely transparent barbule sections in close proximity to the slides cover slip. It has a total slide thickness of approximately 2.5 mm, and as one of the thinnest specimens in the feather series is the most likely to produce a clear excitation response from keratin alone. These specimens were exposed to a wide range of excitation wavelengths (405 nm, 488 nm, and 561 nm UV was not possible due to the pronounced autofluorescence of amber at these wavelengths). The responses of the pigmented keratin, clear keratin, and surrounding amber were contrasted in TMP 96.9.997 and compared to areas of similar visible response in UALVP 52821. Analysis of TMP 96.9.997 illustrated the limitations of this approach, as autofluorescence from the amber was strong at all observed excitation wavelengths. Focusing on keratin within TMP 96.9.997 did not provide an emission profile that was distinguishable from that of the amber in terms of peak values (Figs. S10AC), but the intensity produced by keratin provided additional visibility of anatomical details. When UALVP 52821 was analyzed, an identical pattern emerged (Figs. S10DF), but the background interference from the surrounding amber was much greater (because the total slide thickness in the area sampled was approximately 5 mm). Although these data do not demonstrate conclusively the presence of keratin within either specimen, LSCM imaging confirmed the hollow structure of the filaments in UALVP 52821 (Fig. S11). Furthermore, three-dimensional viewing indicated that the pigmented portion of each filament is surrounded by a thin layer that emits with slightly greater intensity than the surrounding amber. This layer may represent either a reflective surface where the specimen has pulled away from the amber, or a region of different composition. The latter appears more likely given the apparent thickness of this feature (35 m where measurable). Similar elevated emission intensities were observed from both keratin in TMP 96.9.997 and some narrow fractures within the amber of UALVP 52821; thus the 7

observations are inconclusive. The pigmented layer within UALVP 52821 is readily visible and appears to be thin and nearly circular in cross-section. The internal area bounded by the pigmented layer lacks any visible structures, and appears to have been hollow in life. These observations are strongly supported by instances where the filaments are cross-cut by either the polished surface of the amber (Fig. S11B), or the edge of the amber piece itself. Where the filaments are cut cleanly, the hollow core appears as an oblong shape free of structure. This would be expected if the filaments possess a nearly circular cross-section, as their orientation within the amber specimen typically produces oblique sections. Where the filaments breach the surface of the amber nodule, their outlines are rounded and appear circular. If and when additional representatives of the Stage I and II morphotypes are recovered from Canadian amber, we plan to pursue chemical analyses to a much greater extent, particularly once a sufficient archive exists to allow destructive sampling. In the interim, we are open to suggestions for additional techniques from the community. Detailed descriptions of individual specimens and consideration of taphonomy and preservation UALVP 52821 (Stage I morphotype): UALVP 52821 exhibits complex taphonomy: resin remobilization prior to hardening has sheared off the basal portions of the filaments, and has introduced a series of offsets or micro-faults running through many of them. It also appears as though minor decay and the escape of trapped gasses have resulted in fragmentation of the outer wall in many filaments. This has produced a number of perforations in some of the filaments: these are visible as semicircular incisions of filament margins that correspond to fragments of the outer wall found floating in the amber (Figs. S2, S4A). The complete margin of some of these holes is also visible in some places (Fig. S4A inset), providing a clear indication of the thickness of the outer wall, and confirming the hollow interior of the filaments. Additionally, the filaments appear to have been arranged in rows at the time of inclusion within the amber mass, which may reflect either their original arrangement or clumping within the resin (Fig. S2). Their form of preservation, particularly their patchy translucency, is similar to that of both feather remains and a hair fragment recovered from the deposit (Figs. S4E, F). The alternation of fine light and dark spots that appears to form a cross-hatch pattern on the surface of some filaments may have a taponomic origin. This pattern is similar to that observed in insect cuticles that have pulled away from the encapsulating amber within the deposit, and does not necessarily indicate genuine primary topography. UALVP 52822 (Stage II morphotype): The clusters of filaments that run parallel to the longest axis of UALVP 52822 (Fig. 1C) interact with a dark drying line, partly obscuring the separation between individual filaments at the apex of the cluster. Also, all clusters breach the exterior surface of the amber nodule, limiting their observed lengths and any potential to observe basal attachments. The filaments within each cluster converge basally, regardless of orientation or their preserved lengths. Within the same amber nodule are a single aphidoid hemipteran, potential insect frass pieces, and a few partial strands of a spiders web. TMP 96.9.997 and TMP 96.9.1036 (superficially Stage IIb morphotype): TMP 96.9.997 is close to, but does not extensively contact a drying line within the amber. This has produced a few small areas where dark staining appears to spread outward from 8

individual barbules. In TMP 96.9.1036, the contact with a drying line is greater, as is the areal extent of the darkened surroundings. Exposure and weathering of the latter specimen explains at least some of the lack of pigmentation in barbules near the apex of the barb (Fig. S9). TMP 96.9.553, TMP 94.666.15 and TMP 96.9.546 (pennaceous barbs): In TMP 96.9.553, resin flow and interaction with a drying line has caused barbules on at least three of the barbs to draw inward toward the ramus. In TMP 94.666.15, barbules near the apex and base of the barb are similarly swept inward due to resin flow (Figs. 3I, J). In this specimen, interaction with the drying line is fairly extensive, and may have caused the darker color. In TMP 96.9.546, interaction with a drying line has created dark margins surrounding basal barbules (Fig. 3H), but has had little other effect. A mite that appears to be a juvenile oribatid (H. Proctor det.) is found in association with TMP 96.9.546, but appears to be within a different flow region in the amber, and not directly associated with the feather fragment. TMP 79.16.12 (down feather): The tuft of downy barbules within TMP 79.16.12 converges basally (Fig. 3E), but each is truncated at the edge of the amber nodule. Within the amber nodule are six specimens of the dipteran Adelohelea glabra Borkent, at least 4 partial aphidoid hemipteran specimens, and a few isolated strands of spider web. TMP 96.9.334 (coiled barbules): Although the feather portion preserved within the amber nodule does not appear to encompass any barbs, the presence of specialized barbules and a broad rachis suggest an advanced Stage IV morphotype for TMP 96.9.334 (Figs. S6, S7). A prominent drying line within the nodule suggests resin flowed toward the apex of the rachis, sweeping many of the barbules inward toward the rachis, and causing some of the barbules to tear free and rotate (their nodes show that they face the opposite direction). Most of these barbules were probably attached to an unpreserved barb ramus that was basal to the preserved section of feather (as the barb ramus is not preserved). There is a fragment of what may be barb ramus preserved on the surface of the amber nodule (Fig. S7B), but preservation is too poor for identification. Those barbules that do not terminate on this questionable fragment exit the edge of the amber piece basally with no indication of attaching to the rachis segment (Fig. S7). The amber slice that entombs the feather is slightly less than 2.25 mm thick, so it is possible that the window of preservation occurred between barb rami, but this would require a relatively wide spacing of barb rami. Posterior to the microphysid hemipteran in the amber nodule, and well removed from rachis, a second set of barbules splays outward in the opposite direction (Fig. S6B). Unless the rachis has completely folded back on itself outside the window of preservation (and against the direction of resin flow), this second set of barbules is difficult to explain as anything other than the remains of a second feather. UALVP 52820 (Stage V, vaned feather fragment): UALVP 52820 is caught within a large mass of tangled spiders web (Fig. S8). To preserve the web, the amber nodule was not polished to a thin wafer. As a result, there are numerous drying lines to contend with. Aside from the feather, only a few potential insect frass pellets are found within the amber nodule.

Additional notes on pigmentation and structure of Canadian amber feather specimens 9

One of the most interesting aspects of the Canadian amber assemblage is the preservation of pigments within the specimens. Pigmentation has recently been described from a number of non-avian theropods (45, 46) and fossil birds (4649). This work has hinged upon the identification of melanosomes (pigment bodies within organelles) with distinctive shapes and arrangements, through the use of scanning electron microscopy (49). Although preservation is exceptional within amber, examination of the insect assemblage has demonstrated that diagenetic alteration has had a profound effect on the coloration of the insect remains, and techniques such as melanosome observation are likely the only way to precisely identify the original colors of the feather specimens. Unfortunately, it is not possible to subject the Canadian amber to the destructive sampling required to access the melanosomes for SEM examination. This limits the discussion to pigment intensity and distribution, and comparison with works that have mapped these patterns in modern feathers (24, 25). UALVP 52821 (Stage I morphotype): The filaments of UALVP 52821 exhibit a wide range of diffuse (non-localized) pigmentations, ranging from near-transparency to heavily-pigmented, nearly opaque (Fig. S2). Pigmentation along the length of each filament appears to be relatively consistent, but taphonomic influences complicate this observation, and limit any inferences of the original colors. These specimens appear to have ranged in color from near-white (unpigmented) to near-black (heavily pigmented). No large-scale pigmentation patterns, such as banding created by a series of neighboring filaments with similar pigmentation can be inferred, although this may be an effect of the small sample size. UALVP 52822 (Stage II morphotype): Much of the dark coloration in stage II morphotype specimens (UALVP 52822) is attributable to preserved pigments; however, it is not possible to observe the distribution of pigments within these structures as they are nearly opaque (Fig. S5). This, combined with a lack of modern analogues, limits our interpretation to suggesting tentatively a dark brown or black overall color for the filament clusters. TMP 96.9.997 and TMP 96.9.1036 (superficially Stage IIb morphotype): Dark-field microphotography (Fig. S9) and comparison between the Canadian amber specimens (TMP 96.9.997 and TMP 96.9.1036) and epoxy-embedded modern feathers shows that the density and distribution of pigments (24, 25) preserved in the fossil material is consistent with a medium- to dark-brown plumage (Fig. S12). The ramus and proximal three to four barbule nodes lack or have reduced pigmentation, as do the basal sections of distal barbule nodes (Fig. S9B). Within distal barbule nodes, pigment is concentrated in oblong masses, leaving clear nodes in addition to clear bases within the internodes (Fig. S9E). Barbules are approximately 6 m in diameter and gradually taper away from their nodes, which appear to bear three elongate (3 m) prongs (25). This barbule type conforms to the reduced plumulaceous barbules described within the basal regions of some contour feathers (17), but the barbs and their barbules are present in a sparse and strictly aligned pennaceous pattern that does not match well with observed modern exemplars. TMP 96.9.553, TMP 94.666.15 and TMP 96.9.546 (pennaceous barbs): In these specimens, the barbules on both sides of each barb are of pennaceous morphology (17), with ventral plates (blade-like bases) that gradually narrow and become more cylindrical toward their apices (Figs. 3FK). Barbules range in length from 0.15 mm to 0.35 mm and 10

appear to be composed of 10 to 12 distinct nodes. Barbules on these partial feathers appear to lack differentiation into the smooth proximal and hooked distal series on either side of the barb. The barbules do not correspond well to modern reduced pennaceous morphologies, because barbules on either side of the barb are of relatively even lengths, comparable shapes, and lack hooklets at their nodes. Within one amber piece containing 16 examples of this morphotype (TMP 96.9.553, Fig. 3F), the individual barbs appear to converge upon a shared base. These specimens might be identified as a variation on the open pennaceous (non-interlocking) terminal regions of barbs within contour feathers (17), but this interpretation would require that the bases of individual barbs were drawn together taphonomically, due to torsion within viscous resin. Pigmentation is present within two of the three specimens with flattened barbs. In both of these specimens, the dorsal flange (cylindrical portion) of the basal internode is darkly pigmented, while the ventral plate bears reduced pigmentation within its ventral margin (Figs. 3IK). Interrupted pigmentation is apparent within many of the ventral plates, reflecting segmentation within the base of each barbule (Fig. 3K), as pigmentation is only present within the apical portions of the subsequent internodes. In each of the two specimens that possess pigmentation, its intensity and distribution are comparable to dark brown modern feathers (24); however, amber thickness and interactions with drying lines within the amber preclude more detailed analysis. TMP 79.16.12 (down feather): TMP 79.16.12 possesses tufted barbules that lack pigmentation, with thin, flattened internodes (approximately 8 m in width, 18 m in breadth, and 170 m in length) ending in moderately inflated nodes (25 m in diameter) with three weak nodal points (Fig. 3E). Individual barbules appear to converge on a short rachis, although none is apparent within the amber itself. Taken together, these features suggest an understory position within the plumage, and the overall appearance of the specimens is similar to that of natal or juvenile down (17). These barbules appear transparent, and would have been white in life. TMP 96.9.334 (coiled barbules): Pigmentation is diffuse and variable within the barbules of TMP 96.9.334 (Fig. S6): the overall color would likely have been pale or white. Interestingly, the basal internodes within each barbule appear to be consistently of a slightly darker color than their apical equivalents. Structurally, these barbules exhibit a form of basal coiling that is analogous to that found in some modern birds, such as sandgrouse, seedsnipes, and grebes. In these modern examples, the coils are used to sequester water within the plumage. This coiling differs significantly from the curled barbule bases observed in many taxa (e.g., Fig. S12D), in that the barbules undergo full rotations, and when exposed to water, they straighten, drawing water in by capillary action (21). In the modern taxa that exhibit basal coiling, this structure is either used for transport of water to the nest (in sandgrouse, possibly in seedsnipes) or as a means of altering the hydrodynamic properties of the bird, in order to facilitate diving (in grebes) (21, 22, 23). Although neither of these groups exhibit as many basal coils as those observed in TMP 96.6.334, grebes appear to exhibit slightly more coils than the other taxa. Grebes possess 23 basal coils (23, 24, Figs. S12FG), while sandgrouse possess 1.52 full coils (22), and seedsnipes have less developed basal coiling (21). The basal coils in TMP 96.9.334 may have served either of the functions known in the modern avifauna, but the high number of coils suggests that they are more likely to have been

11

employed in diving behavior, as they would have sequestered a comparatively large volume of water. UALVP 52820 (Stage V, vaned feather fragment): The preserved feather section of UALVP 52820 is entombed within a thick piece of amber and crosses multiple drying lines, making color observations difficult. Transmitted light microphotographs (Fig. S8) reveal a banded pattern of dark pigmentation within the basal plate and diffuse dark pigmentation within the pennulum, suggesting perhaps a grey or black feather (24). Although this is only a partial feather, the ramus (barb shaft) is expanded dorsoventrally, with a distinct dorsal ridge bordered by ledges, a characteristic of rami adapted to form strong vanes for flight (17). Furthermore, distal series barbules each display a distinct, narrow pennulum, and a moderately elongate, narrow ventral tooth on the apex of a broad basal plate. These are adaptations for interlocking with adjacent barbs to form a vane (17).

12

Supplementary figures Fig. S1 Graph of specimen diameters for filamentous structures (Stage I and II) and barbules in Canadian amber, compared to other possible sources. Circles indicate mean value, vertical lines 1 SD, boxes show observed ranges or reported ranges for majority of specimens (33), and arrows indicate ranges beyond graph area accompanied by maximum value.

13

Fig. S2 Photomicrographs of Stage I filaments in UALVP 52821. (A) Field of individual filaments cut obliquely, illustrating distribution of filaments; (B) close-up of boxed area within A, showing apparent grouping of filaments (arrow) and color variation between filaments when illuminated from above.

14

Fig. S3 Compound microscope images (b.f.) of Stage I filaments in UALVP 52821. (A) Area where filaments are truncated by outer surface of amber nodule (pebbled amber surface in upper-right of figure), arrow indicates one of the faults running through the filaments; (B) hollow central region of a filament (arrow); see figures S4 and S11 also.

15

Fig. S4 Dissecting and compound microscope images of Stage I filaments, fungi, and mammalian hair. (A) Degraded portion of Stage I filament apex in UALVP 52821; vertical arrows indicate regions where there are holes in the outer wall, angled arrows indicate pieces of the outer wall floating within the amber; inset shows holes with complete outlines at double the magnification of A (d.f.); (BD) apparent surface texture of Stage I filament in UALVP 52821, (B) filament adjacent to arrow displays faint cross-hatching pattern of light and dark areas, (C) filament adjacent to arrow displays clearer cross-hatching, perhaps as a result of a nearby bend in the filament (d.f.), (D) multiple filaments display faint texture where they have pulled away from the surrounding amber (d.f.); (E) TMP 96.9.998, mammalian hair from Canadian amber, with thick cortex and discontinuous medulla, most likely displaying a multiserial ladder or aeriform lattice pattern adjacent to arrow (b.f.); (F) TMP 96.9.998, showing faint traces of cuticular scales adjacent to arrows (d.f.); (G) mat of fungal hyphae (white filaments near bottom of image) contrasted against pair of Stage I filaments (larger, dark filaments near top of image) in UALVP 52821.

16

Fig. S5 Compound microscope images (b.f.) of Stage II clusters in UALVP 52822. (A) Distal tip of cluster in Fig. 1C, showing tapered apices of filaments and loose bundling within a cluster, also with apparent dark, diffuse pigmentation; (B) proximal truncation of cluster in Fig. 1C, showing tightly adpressed filaments at point where cluster is cross-cut by the edge of the amber nodule (arrow); (C) loose bundling apparent within other clusters in the same piece of amber, these clusters are more obliquely oriented within the nodule, and may show variable pigmentation within filaments (toward upper-left of figure).

17

Fig. S6 Compound microscope images of coiled barbules (TMP 96.9.334). (A) Specimen overview showing coiled barbule bases (predominantly within the lower left of figure) surrounding thick, flattened rachis (arrow); reddish-brown areas are the result of a prominent drying line within the amber (b.f.); (B) oblique section through cluster of coiled barbules surrounding a microphysid hemipteran, with portions of second feather posterior to microphysid (TMP 96.9.334, microphotograph); (C) straight apical barbule sections exhibiting variable diffuse pigmentation (b.f.); (D, E) close-ups of straight barbule nodes and internodes, showing flattened internodes that twist slightly along their length and exhibit a linear pattern as a result of either ultrastructure or pigment granule distribution (b.f.).

18

Fig. S7 Dissecting microscope images of coiled barbules (TMP 96.9.334). (A) Specimen overview (opposite to Fig. S6A), showing coiled barbule bases (predominantly within lower, central part of figure) surrounding thick, flattened rachis (vertical arrows); base of rachis (lower arrow) recessed with respect to surface of amber piece as a result of weathering; reddish-brown areas are the result of a prominent drying line within the amber; (B) close-up of rachis base, vertical arrow indicates fragment of possible barb ramus that is too poorly preserved to permit confident identification, inclined arrows indicate a few of the many individual barbules that exit the edge of the amber piece without making contact with the rachis (this lack of attachment appears to be characteristic of most of the barbules, although they are crowded basally).

19

Fig. S8 Compound microscope images of differentiated barbules with distinct pennulae in UALVP 52820, indicating preservation that is visually identical to Stage I and II morphotypes. (A) Isolated barb with differentiated barbules and thickened barb shaft ensnared in spider web (microphotograph) (B) overview of barbules near base of barb, and surrounding spider web, (b.f.); (C) overview of barbules near distal tip of barb, with clearly defined distal and proximal barbule series (left and right sides of ramus, respectively), distinguished by the sharp transition between the base and pennulum within the distal series barbules (arrow), (b.f.); (D) close-up of proximal barbule, showing distribution of pigmentation, and nodal prongs, (b.f.); (E) close-up of distal barbule, showing distribution of pigmentation, nodal prongs, and ventral tooth upon basal plate (arrow) adjacent to abrupt transition into pennulum (b.f.).

20

Fig. S9 Compound microscope images of pennaceous barbs with reduced plumulaceous barbules. (A) Overview of pigmented barb, TMP 96.9.997 (b.f.); (B) close-up of boxed area in A, showing weak ramus and unpigmented basal barbules, as well as distribution of pigment within subsequent barbules (b.f.); (C) dark-field image of same feather region, showing apparent feather color created by pigmentation, as well as distribution of pigmentation within barb components; (D) overview of variably pigmented barb with elongate ramus tip, TMP 96.9.1036, (b.f., micro-panorama compiled using Helicon Focus); (E) close-up of pigment distribution within basal barbules of D.

21

Fig. S10 SDCM and LSCM data for Stage I morphotype and TMP 96.9.997 emission response microphotographs and emission spectra. (A) Autofluorescence of TMP 96.9.997 when exposed to laser based excitation at 491 nm (green, filter for emission wavelength et525/50) and 561 nm (red, filter for emission wavelength et620/60), showing marginally brighter spots where only keratin is preserved; (B) normalized emission spectrum for sampling points on TMP 96.9.997 when excited at 405 nm, emission from amber (ROI 1) peaks between 480490 nm, similar, but progressively more muted peaks for unpigmented keratin (ROI 3) and pigmented regions of barbule (ROI 2); (C) map of sampling points and emission intensity between 540 and 550 nm TMP 96.9.997; (D) autofluorescence of UALVP 52821 when exposed to laser based excitation at 491 nm (green) and 561 nm (red); (E) emission spectrum for sampling points on UALVP 52821 when excited at 405 nm, emission from amber (ROI 1) peaks broadly near 540 nm; similar, but progressively more muted peaks for unpigmented outer wall of filament when cut obliquely (ROI 2); unpigmented outer wall of filament when cut longitudinally (ROI 4); and pigmented layer (ROI 3); (F) map of sampling points and emission intensity between 540 and 550 nm in UALVP 52821, arrow indicates rounded outline produced where filament breaches surface of amber piece (the pebbled surface at the lower right of the image, also visible in Fig. S10D).

22

Fig. S11 LSCM and additional photomicographs of UALVP 52821. (A) Three-dimensional scan of UALVP 52821 at 405 nm excitation (mapping emission intensity between 411 nm and 766 nm); fine white lines correspond to vertical section planes presented in panels to the left of and below the main figure. Arrow indicates circular cross-section of one filament apex, directly comparable to B. Brackets delimit filaments cut obliquely, demonstrating outer wall (bright) surrounding thin pigmented layer (dark) and hollow core (comparable to the surrounding amber), this pattern is also found within longitudinal sections of the filaments within this piece of amber. (B) Dissecting microscope photomicrograph of filaments in the same region of the amber specimen as A, illustrating appearance of filaments when sectioned obliquely along polished surface, arrows indicate oblong internal voids exposed at the section plane.

23

Fig. S12 Photomicrographs of modern bird feathers for comparison of barbule structure and pigmentation patterns. (A) Plumulaceous barbules from the afterfeather of a pheasant (b.f.); (B) dark-field image of A, showing pigment distribution and resulting dull-brown coloration; (C) close-up of barbules in A, showing pigment concentration near nodes although somewhat more diffuse, this is comparable to pigmentation in Fig. S8 (b.f.); (D) partially curled barbule bases in the plumulaceous basal barbs within a body contour feather of a kiwi (Apteryx owenii, RM 5440), for comparison to coiled barbule bases in Figs. S6 and S7, and also an example of diffuse pigmentation (b.f.); (E) single barb from white belly feather of a grebe (Aechmophorus occidentalis, RAM Z279.78.2), illustrating coiled barbule bases, predominantly with two basal coils (dissecting microscope); (F) combined focal-plane image of different barbule from same feather as E, providing overview of coiling (d.f.); (G) single image of the barbule bases that are partly obscured in G, due to the orientation of the barbules (d.f.).

24

References and Notes 1. P. G. Davis, D. E. G. Briggs, Fossilization of feathers. Geology 23, 783 (1995). doi:10.1130/0091-7613(1995)023<0783:FOF>2.3.CO;2 2. D. Schlee, W. Glckner, Bernstein: Bernsteine und Bernstein-Fossilien. Stuttg. Beitr. Naturkd. C 8, 1 (1978). 3. J. Alonso et al., A new fossil resin with biological inclusions in Lower Cretaceous deposits from lava (northern Spain, Basque-Cantabrian basin). J. Paleontol. 74, 158 (2000). doi:10.1666/0022-3360(2000)074<0158:ANFRWB>2.0.CO;2 4. V. Perrichot, L. Marion, D. Nraudeau, R. Vullo, P. Tafforeau, The early evolution of feathers: Fossil evidence from Cretaceous amber of France. Proc. Biol. Sci. 275, 1197 (2008). doi:10.1098/rspb.2008.0003 Medline 5. D. Grimaldi, G. R. Case, A feather in amber from the Upper Cretaceous of New Jersey. Am. Mus. Novit. 3126, 1 (1995). 6. Q. Ji, P. J. Currie, M. A. Norell, S.-A. Ji, Two feathered dinosaurs from northeastern China. Nature 393, 39 (1998). 7. M. Norell, X. Xu, Feathered dinosaurs. Annu. Rev. Earth Planet. Sci. 33, 277 (2005). doi:10.1146/annurev.earth.33.092203.122511 8. X. Xu, Y. Guo, The origin and early evolution of feathers: Insights from recent paleontological and neontological data. Vertebrata PalAsiatica 47, 311 (2009). 9. R. C. McKellar, A. P. Wolfe, in Biodiversity of Fossils in Amber from the Major World Deposits, D. Penney, Ed. (Siri Scientific Press, Manchester, 2010), pp. 149166. 10. Materials and methods are available as supporting material on Science online. 11. R. O. Prum, Development and evolutionary origin of feathers. J. Exp. Zool. 285, 291 (1999). doi:10.1002/(SICI)1097-010X(19991215)285:4<291::AID-JEZ1>3.0.CO;2-9 Medline 12. R. O. Prum, A. H. Brush, The evolutionary origin and diversification of feathers. Q. Rev. Biol. 77, 261 (2002). doi:10.1086/341993 Medline 13. A. H. Brush, Evolving a protofeather and feather diversity. Am. Zool. 40, 631 (2000). doi:10.1668/0003-1569(2000)040[0631:EAPAFD]2.0.CO;2 14. P. J. Currie, P.-J. Chen, Anatomy of Sinosauropteryx prima from Liaoning, northeastern China. Can. J. Earth Sci. 38, 1705 (2001). doi:10.1139/e01-050 15. X. Xu, X. Zheng, H. You, Exceptional dinosaur fossils show ontogenetic development of early feathers. Nature 464, 1338 (2010). doi:10.1038/nature08965 Medline 16. P. Stettenheim, The bristles of birds. Living Bird 12, 201 (1973). 17. A. M. Lucas, P. R. Stettenheim, Avian Anatomy: Integument (U.S. Department of Agriculture, Washington, DC, 1972). 18. P.-J. Chen, Z.-M. Dong, S.-N. Zhen, An exceptionally well-preserved theropod dinosaur from the Yixian Formation of China. Nature 391, 147 (1998). doi:10.1038/34356 1

19. A. W. A. Kellner et al., The soft tissue of Jeholopterus (Pterosauria, Anurognathidae, Batrachognathinae) and the structure of the pterosaur wing membrane. Proc. Biol. Sci. 277, 321 (2010). doi:10.1098/rspb.2009.0846 Medline 20. X. Xu, Z.-H. Zhou, R. O. Prum, Branched integumental structures in Sinornithosaurus and the origin of feathers. Nature 410, 200 (2001). doi:10.1038/35065589 Medline 21. T. J. Cade, G. L. Maclean, Transport of water by adult sandgrouse to their young. Condor 69, 323 (1967). doi:10.2307/1366197 22. G. L. MacLean, Water transport by sandgrouse. Bioscience 33, 365 (1983). doi:10.2307/1309104 23. J. Fjelds, The Grebes: Podicipedidae (Oxford Univ. Press, New York, 2004). 24. A. C. Chandler, A study of the structure of feathers, with reference to their taxonomic significance. Univ. Calif. Publ. Zool. 13, 243 (1916). 25. C. J. Dove, A descriptive and phylogenetic analysis of plumulaceous feather characteristics in Charadriiformes. Ornith. Mono. 51, 1 (2000). 26. P. J. Currie, in Dinosaur Provincial Park: A Spectacular Ancient Ecosystem Revealed, P. J. Currie, E. B. Koppelhus, Eds. (Indiana Univ. Press, Bloomington, 2005), pp. 367397. 27. N. Longrich, An ornithurine-dominated avifauna from the Belly River Group (Campanian, Upper Cretaceous) of Alberta, Canada. Cretac. Res. 30, 161 (2009). doi:10.1016/j.cretres.2008.06.007 28. E. Buffetaut, A basal bird from the Campanian (Late Cretaceous) of Dinosaur Provincial Park (Alberta, Canada). Geol. Mag. 147, 469 (2010). doi:10.1017/S0016756810000129 29. P. Nascimbene, H. Silverstein, in Studies on Fossils in Amber, With Particular Reference to the Cretaceous of New Jersey, D. Grimaldi Ed. (Backhuys Publishers, Leiden, 2000), pp. 93102. 30. L. A. Hausman, Structural characteristics of the hair of mammals. Am. Nat. 54, 496 (1920). doi:10.1086/279782 31. H. Brunner, B. J. Coman, The Identification of Mammalian Hair (Inkata Press, Melbourne, 1974). 32. R. Vullo, V. Girard, D. Azar, D. Nraudeau, Mammalian hairs in Early Cretaceous amber. Naturwissenschaften 97, 683 (2010). doi:10.1007/s00114-010-0677-8 Medline 33. H. J. Hudson, Fungal Biology (Edward Arnold Publishers, London, 1986). 34. J. H. Burnett, Fundamentals of Mycology (Edward Arnold Publishers, London, ed. 3, 1986). 35. P. H. B. Talbot, Principles of Fungal Taxonomy (Macmillan, London, 1971). 36. D. Catling, J. Grayson, Identification of Vegetable Fibres (Chapman and Hall, London, 1982). 37. W. Von Bergen, W. Krauss, Textile Fiber Atlas: A Collection of Photomicrographs of Common Textile Fibers (American Wool Handbook Company, New York, 1942). 2

38. T. Lingham-Soliar, R. H. C. Bonser, J. Wesley-Smith, Selective biodegradation of keratin matrix in feather rachis reveals classic bioengineering. Proc. Biol. Sci. 277, 1161 (2010). doi:10.1098/rspb.2009.1980 Medline 39. Q. Ji, S.-A. Ji, Advances in the study of the avian Sinosauropteryx prima. Chin. Geol. 242, 30 (1997). 40. C. Soriano et al., Synchrotron x-ray imaging of inclusions in amber. C. R. Palevol 9, 361 (2010). doi:10.1016/j.crpv.2010.07.014 41. M. H. Schweitzer et al., Beta-keratin specific immunological reactivity in feather-like structures of the cretaceous alvarezsaurid, Shuvuuia deserti. J. Exp. Biol. 285, 146 (1999). doi:10.1002/(SICI)1097-010X(19990815)285:2<146::AID-JEZ7>3.0.CO;2-A Medline 42. G. O. Poinar, Jr., Life in Amber (Stanford Univ. Press, Stanford, 1992). 43. J. H. Langenheim, Plant Resins: Chemistry, Evolution, Ecology, and Ethnobotany (Timber Press, Portland, 2003). 44. Y. Wu, J. Y. Qu, Autofluorescence spectroscopy of epithelial tissues. J. Biomed. Opt. 11, 054023 (2006). doi:10.1117/1.2362741 Medline 45. Q. Li et al., Plumage color patterns of an extinct dinosaur. Science 327, 1369 (2010). doi:10.1126/science.1186290 Medline 46. F. Zhang et al., Fossilized melanosomes and the colour of Cretaceous dinosaurs and birds. Nature 463, 1075 (2010). doi:10.1038/nature08740 Medline 47. J. A. Clarke et al., Fossil evidence for evolution of the shape and color of penguin feathers. Science 330, 954 (2010). doi:10.1126/science.1193604 Medline 48. J. Vinther, D. E. G. Briggs, R. O. Prum, V. Saranathan, The colour of fossil feathers. Biol. Lett. 4, 522 (2008). doi:10.1098/rsbl.2008.0302 Medline 49. J. Vinther, D. E. G. Briggs, J. Clarke, G. Mayr, R. O. Prum, Structural coloration in a fossil feather. Biol. Lett. 6, 128 (2010). doi:10.1098/rsbl.2009.0524 Medline Acknowledgments: We thank the Leuck family and M. Schmidt (donated specimens); M. Caldwell, S. Ogg and M. Srayko (microscopy); E. Koppelhus and H. Proctor (discussions); and J. Gardner, B. Strilisky, A. Howell, and J. Hudon (TMP, Redpath Museum, and Royal Alberta Museum collections). Research was funded by Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Grants to B.D.E.C., A.P.W., and P.J.C. and NSERC and Alberta Ingenuity Fund support to R.C.M.