On the purported presence of fossilised collagen fibres in an ichthyosaur and a theropod dinosaur

Since the discovery of exceptionally preserved theropod dinosaurs with soft tissues in China in the 1990s, there has been much debate about the nature of filamentous structures observed in some specimens. Sinosauropteryx was the first non-avian theropod to be described with these structures, and remains one of the most studied examples. Despite a general consensus that the structures represent feathers or feather homologues, a few identify them as degraded collagen fibres derived from the skin. This latter view has been based on observations of lowquality images of Sinosauropteryx, as well as the suggestion that because superficially similar structures are seen in Jurassic ichthyosaurs they cannot represent feathers. Here, we highlight issues with the evidence put forward in support of this view, showing that integumentary structures have been misinterpreted based on sedimentary features and preparation marks, and that these errors have led to incorrect conclusions being drawn about the existence of collagen in Sinosauropteryx and the ichthyosaur Stenopterygius. We find that there is no evidence to support the idea that the integumentary structures seen in the two taxa are

2 THE debate on the origins of birds is one of the oldest in palaeontology (Huxley 1870;Heilmann 1926), and there has been a substantial increase in research on this question in the past two decades thanks to the discovery of numerous exceptionally preserved fossils in China (reviewed, Zhou et al. 2003;Xu 2006). These discoveries included extraordinarily well preserved stem birds and theropod dinosaurs complete with integumentary preservation (Ji and Ji 1996;Zhou et al. 2003;Xu 2006;Zhang et al. 2006). The first of these theropods to be described with features attributed to filamentous integumentary structures was the compsognathid Sinosauropteryx (Ji and Ji 1996;Fig. 1). These authors noted the unique integumentary structures found mostly across the dorsum and tail of the animal, which have subsequently been observed in all known specimens (Ji and Ji 1996;Currie and Chen 2001;Ji et al. 2007). Through the original descriptions and a series of subsequent studies, a general consensus has arisen that the structures represent feather homologues or 'protofeathers' (see Prum 1999 andXu 2006 for discussions on feather evolution and nomenclature). This consensus conforms to previous phylogenetic hypotheses about the origins of birds and their nesting within Theropoda, established long before any theropod with proposed feathers had been discovered (Ostrom 1976;Prum 2002;Prum 2003;Norell and Xu 2005;Smith et al. 2015).

Evidence supporting a feather affinity of integumentary structures in Sinosauropteryx
includes morphological analysis of the structures, their distribution across the body which is conserved in multiple specimens (Fig. 1), SEM data showing the presence of melanosomes within the structures, and both morphological and molecular phylogenetics (Currie and Chen 2001;Prum 2002;Prum 2003;Norell and Xu 2005;Li et al. 2010Zhang et al. 2010Smith et al. 2015). Alternative explanations have however been put forward for the 3 structures, with suggestions that they represent partially degraded collagen fibres from the skin rather than feathers (Ruben and Jones 2000;Lingham-Soliar 2003a;Feduccia et al. 2005;Lingham-Soliar et al. 2007).

The idea that the integumentary structures first identified as feathers in
Sinosauropteryx could actually represent degraded collagen fibres originated soon after the original description of the theropod, based on similarities to modern snake and lizard collagen (Gibbons 1997;Ruben and Jones 2000). The claim was then further fuelled by the observations of similar structures in ichthyosaur fossils (Feduccia 1999;Lingham-Soliar 2003a). A description of the integumentary structures on a Sinosauropteryx specimen (IVPP V12415: Fig. 1C) using light microscopy was used to suggest that they were collagen fibres (Lingham-Soliar et al. 2007). This work has formed the basis of most criticisms and rebuttals of the feathered theropod hypothesis (Lingham-Soliar et al. 2007;Lingham-Soliar 2010aFeduccia 2013). One of the key observations made about the structures of IVPP V12415 as evidence of a collagen affinity is the beaded appearance of the fibres ( Fig. 2; Lingham-Soliar et al. 2007, fig. 2). Beading is said to occur in modern collagen due to contraction caused by dehydration (Lingham-Soliar et al. 2007;Lingham-Soliar and Wesley-Smith 2008). Further evidence for a collagen affinity came from the pattern and orientation of their arrangement. In certain regions, an overlap of "geometrically precise bands of parallel fibres" is said to be consistent with structural reinforcement of the skin when compared to modern animals (Lingham-Soliar et al. 2007). Observations of decaying dolphin tissues were also used to demonstrate how collagen can show a similar morphological appearance to feathers (Lingham-Soliar 2003b).
The integumentary structures in Sinosauropteryx were interpreted by Lingham-Soliar et al. (2007) as remnants of a frill along the dorsum, reinforced with structural collagen 4 fibres, thus accounting for the way in which the integument extends dorsally from the vertebrae. This was suggested to add stiffening support to the long tail, to protect against injury, as well as to provide a decorative display organ (Lingham-Soliar et al. 2007). The banded pattern of the integument was interpreted as "scalloping", analogous to features in modern basilisk lizards such as Basiliscus basiliscus and B. plumifrons (Lingham-Soliar et al. 2007;Lingham-Soliar 2012, 2013. It has been further suggested that the purported frill could be used to aid swimming in Sinosauropteryx as the dinosaur was found in a lacustrine environment (Lingham-Soliar 2012, 2013. In addition to the hypothesis that the structures on Sinosauropteryx are collagen fibres rather than feathers, it has been claimed that scales are present overlying the fibres making them unlikely to represent feathers. Martin and Czerkas (2000) were the first to suggest that scales were present in Sinosauropteryx four years after its original description, when it was stated that a slab with scale impressions was originally removed from a specimen. No evidence to support this claim was ever provided, however, nor any images (Martin and Czerkas 2000 p 688;Lingham-Soliar 2013 p. 460). More recently, scales have been purportedly identified on another specimen and used as evidence of a frill as well as to refute the presence of feathers (Lingham-Soliar 2013).
Many of the flaws in arguments given by proponents of the collagen hypothesis have been noted (e.g. Mayr 2010;Smith et al. 2015;van der Reest et al. 2016), but claims of a likely collagen affinity have propagated through the literature. Often, when avian origins and theropod integuments are discussed, both feather and collagen hypotheses are treated as equally parsimonious and credible (Perrichot et al. 2008;Dhouailly 2009;Geist 2009;Ruben 2010;Dove and Straker 2012). The debate is clearly an important one, with major implications for our understanding of the early evolution of birds, feathers and flight, and 5 thus scrutiny over proposed evidence is essential. As Sinosauropteryx has been the focus of the majority of the debate on theropod integumentary structures, a clearer understanding of the known specimens is essential.
Here, we review the evidence given for the collagen model of theropod integumentary structures in Sinosauropteryx as well as similar structures observed in ichthyosaurs. We highlight a shortcoming in this model, in that nearly all the evidence comes from lowresolution photographic images. We present our results, based on close study of the three best preserved Sinosauropteryx specimens (Fig. 1). From our observations and photographs, previous work using lower resolution images is reconsidered, and errors arising in earlier interpretations discussed. The aim of this paper is to clarify a number of misconceptions surrounding soft tissue preservation in ichthyosaurs and the theropod Sinosauropteryx as well as to highlight errors used in many of the arguments surrounding theropod integumentary structures.

SINOSAUROPTERYX REAPPRAISED
While the evidence for a feather affinity of the integumentary structures in Sinosauropteryx has been criticised heavily since the hypothesis was first put forward (Gibbons 1997;6 Feduccia 1999;Lingham-Soliar et al. 2007;Feduccia 2013), the evidence for the collagen model has never been rigorously tested. The statements provided by proponents of the collagen hypothesis will be evaluated in turn, highlighting errors that include misidentification of sedimentary structures, surface topography and marks made during the fossil preparation process. New, high resolution images (60-160mb) of three Sinosauropteryx specimens (IVPP V12415, NIGP 127586 and NIGP 127587; Fig. 1) were taken using a Nikon D800 camera with a Micro Nikkor 60 mm lens mounted on a tripod with a ten second delayed timer to maximise image sharpness. Specimens were illuminated with a mounted tungsten light source with a polarising filter attached. Images were taken under both normal lighting conditions and using a polarising filter on the camera to allow cross-polarisation to reduce glare. These methods provided very accurate and detailed images. Images of the ichthyosaur Stenopterygius quadriscissus (SMF R 457) were taken by Sven Tränkner (SMF).

Fibres in ichthyosaurs
Further study of purported ichthyosaur collagen fibres used in comparisons to the integumentary structures of Sinosauropteryx has revealed a number of misinterpretations in the original descriptions. Identified fibres on Stenopterygius quadriscissus (SMF R 457) said to show three-dimensional preservation and resemblances to theropod fossil fibres are, in many cases, actually preparation marks made when the matrix of the ichthyosaur was being smoothed around the genuine preserved soft tissue (Fig. 3). The new images of the same specimen clearly show that these marks are deep scratches within the matrix, which is not so obvious from the original black and white images (Fig. 3A). These scratch marks, frequently described as "class 2" or "radial" fibres were said to show superficial resemblance to a 7 feather rachis (referred to as pseudo-rachis), an argument against genuine feather preservation in theropods (Lingham-Soliar 2001, 2003a figure 1d). These fibres are noted as only being present on SMF R 457, which is presumably because the other fossils studied do not have these preparation marks still on the matrix or fossil. Further purported "radial" fibres can now also be shown to be misidentified cracks running through the matrix which had been repaired when the multiple blocks of the fossil were put back together (Figs. 3A and B).
It was claimed that mineralisation patterns of a "dorsal longitudinal fibre" (DLF) in SMF R 457 also resembled that of filaments in Sinosauropteryx (Fig. 3A and C;Lingham-Soliar 2001, 2003a. This feature can be seen to run parallel to the body, separating the soft tissue from the preparation marks misidentified as "radial fibres" in the ichthyosaur ( Fig. 3A; Lingham-Soliar 2001, figs. 5 and 9). The "DLF" runs continually, marking the boundary between soft tissue and matrix (Lingham-Soliar 2001). Re-examination of the specimen indicates that rather than a single primary feature, this is in fact a deep trench within the matrix, with the three-dimensional structure creating a light centre with darker ridges due to shadowing ( Fig. 3B and D). This is therefore not a pattern created by mineralisation, but is likely also a preparation mark, and is therefore not an appropriate analogue for any structure seen on Sinosauropteryx.
While many of the soft tissue features preserved within the halo of the body of SMF R 457 do appear to be remnants of organic structures, none that resemble the filaments found on of the new images however shows that these are three dimensionally preserved parts of the bones themselves, likely part of the internal structure as they appear broken when compared to a number of complete bones adjacent to them, which readily cast shadows that appear to have been misidentified as integument ( Fig. 5B and D). One of these skeletal structures was digitally cut from the bone and again presented as an isolated collagen fibre ( Fig. 5C; Lingham-Soliar et al. 2007, fig. 3b). It cannot be assumed that any of these cut fibres are integumentary features showing beading, as the only two that have their original location shown are not fibres, as noted. In conclusion, it seems that the illustrations of supposedly beaded collagen fibres from the Sinosauropteryx specimens are based on undulations of the matrix which create differential lighting and shadowing. Furthermore, we could find no examples of these beaded structures on close inspection of the actual specimens.
Orientation of the filaments in Sinosauropteryx: parallel and cross-fibre patterning 10 Cross-fibre patterning of the dermis reported from NIGP 127587 and IVPP V12415, used to support the collagen hypothesis, also fails to stand up to scrutiny (Lingham-Soliar et al. 2007;Lingham-Soliar 2012, 2013. In Lingham-Soliar (2013), figure 4 shows a section of the tail of IVPP V12415 which has broken and separated, with arrows indicating the proposed opposing direction of fibre orientation suggested to represent the original collagen structural arrangement ( Fig. 6A; Lingham-Soliar 2013 p. 457). Our photographs, however, show that the so-called 'fibres' are in fact deep scratches made during preparation of the specimen, likely from a pneumatic air scribe, and so with no biological significance (Fig. 6C). Similar preparatory scratch marks can be seen across the matrix around the entire animal (Fig. 7).
The same marks are found in NIGP 12586, with some even cutting through the genuine integumentary structures (Fig. 7C). These marks cast shadows, which appears to have led to their misidentification. Interestingly, these preparation marks are clear and obvious in the region where the tail has been displaced, in IVPP V12415 (Fig. 6D), a feature blocked out on the original image by a black square with no explanation in Lingham-Soliar (2013, fig. 1).
Further evidence for geometrically precise bands of parallel collagen fibres comes from observations made in an image of NIGP 127587 (Lingham-Soliar et al. 2007, fig. 4) first published in Chen et al. (1998, fig. 6). The preservation of the structures in this area appears markedly different from those in other areas, with a more three-dimensional and organised appearance, which is explained by Lingham-Soliar et al. (2007, fig. 4) as showing different stages of collagen preservation (Fig. 8A). The layer containing the purported fibres clearly sits underneath the main integumentary layer, as can be seen from the shadow cast by the overlying layer. Our specimen photographs show that subsequent preparation of the specimen has removed most of the overlying layer of integumentary structures, and the matrix now sits entirely at the level of the aforementioned purported parallel fibres (Fig. 8).

11
This new preparation work was presumably done in order to reveal more depth around the bones themselves, as can be seen in the more exposed ventral side of the vertebrae in the new image ( Fig. 8) in comparison to the pre-preparation image. Due to this preparation, information from the integument in the region has been lost. However, it reveals that the identified purported fibres are actually sedimentary layers of the matrix (Fig. 8) 1). From a reconstruction in the same paper (Lingham-Soliar 2012, fig. 6) it appears that this halo has been misinterpreted as the remnants of the outline of the skin of the animal in the form of a frill. This has been used as evidence in support of a collagen affinity (Lingham-Soliar et al. 2007;Feduccia 2013). The halo present on NIGP 127587 is not seen in the other two fossils (Fig. 1), which both sit on a more uniform flat surface layer, highlighting the fact that only where the sediment layers are cut through in NIGP 127587 can the lines suggested as 'fibres' be seen. Shadowing effects on fossil specimens from images in previous work have also been misinterpreted as genuine biological structures in a similar fashion, as noted by Mayr (2010).
The orientation of the genuine integumentary filaments (outwards and posteriorly along the body axis), conforms to pliable integumentary appendages such as feathers homologues rather than collagen fibres, especially as they often curve back in towards the body distally, showing apparent flexibility (Currie and Chen 2001). This is most obvious in the tail region in IVPP V12415 (Fig. 4A-B), and is also clear in the tails of NIGP 127586 ( Fig. 4D-E) and NIGP 127587 (Fig. 4C). A schematic figure in Lingham-Soliar et al. (2007, fig. 2f) ignores the actual pattern of orientation of the filaments and depicts them as being aligned straight and parallel, running posteriorly and outwards from the skeletal axis, in support of a structural collagen fibre affinity. This is in spite of the fact that the genuine orientation of the fibres is presented in a panel in the same figure (Fig. 2E). Our high-quality images of the filaments (Fig. 4) also clearly refute the suggestion that they become progressively more degraded towards their tips, a feature suggested to support identification as collagen (Lingham-Soliar et al. 2007).
Further misidentification of non-organic structures occurs through the use of low resolution images of NIGP 127587. Holes in the matrix which cut through the genuine integumentary structures were identified as neural spines with geometrically parallel fibres impressed into them (Lingham-Soliar 2012, fig. 3a). The new images show that these holes 13 are not part of the theropod fossil nor do they show any evidence of fibre association (Fig.   10B). It is also claimed that a cartilaginous trachea is preserved in NIGP 127587 (Lingham-Soliar 2012, fig. 1). In fact, the high resolution images show that no tracheal structure is present (Fig. 10C). In the original figure, a vague bend has been highlighted to create the impression of a structure, which the new images reveal as a joining of unrelated features of the matrix. It is also possible that the conchostracans that appear throughout the matrix were interpreted as tracheal rings (Li et al. 2007). Genuine tracheal preservation is rare, but has been found in some exceptional fossils including a mosasaur where tracheal rings were present in the throat region of the animal (Lindgren et al. 2010, fig . 3) and the theropod Scipionyx samniticus (Dal Sasso and Signore 1998). When present in fossils however, these tracheal features bear no resemblance to any structure seen in Sinosauropteryx (Dal Sasso and Signore 1998;Lindgren et al. 2010).

Evidence of scales in Sinosauropteryx.
In IVPP V12415, a dark patch found at a break in the tail was identified as comprising four distinct scales with papulose surface patterns ( Fig. 6B; Lingham-Soliar 2013, fig. 3). The original figure is of such low resolution however, that no features can be seen. Even using the highest resolution of the new images obtained, identifiable structures cannot be made out within the organic patch (Fig. 6C).
A further area is also claimed to represent scales, this time with 'attachment fibres' as further evidence of their affinity ( Fig. 6A; Lingham-Soliar 2013, fig. 4). Superficially these appear more convincing, and scale-like shapes can be made out along with apparent fibres protruding from them (Fig. 6A). Examination of the high resolution images however reveals 14 that these scales are sedimentary features of the matrix, which continue not only between the tail break, as in the original figure, but right across the matrix far from the animal in a distinct linear arrangement, with those furthest from the fossil showing no scale-like structures (Fig.   6D). The claim that these structures also show a papulose pattern is incorrect, and no evidence of this is seen in the specimen or the new images (Fig. 6). The proposed 'attachment fibres' are also misidentified on closer inspection of the high quality images, which clearly show these to be scratch marks likely made by an air scribe during the preparation of the fossil, further matching those made throughout the matrix of the specimen (Figs. 6 and 7).

CONTEMPORANEOUS JEHOL FOSSILS REFUTE COLLAGEN PRESERVATION
Following these discussions, there is no evidence of scales on the known specimens of Sinosauropteryx. On the other hand, scaled dinosaurs have been found in the Jehol Biota, and so there is evidence available of what preserved scales look like in these lithologies. The ornithischian dinosaur Psittacosaurus shows a covering of well-preserved scales complete with pigment preservation (Mayr et al. 2002Lingham-Soliar andPlodowski 2010, 15 Smith et al. 2015;Vinther et al. 2016). Despite this, no evidence of any structures similar in appearance to the integumentary structures of Sinosauropteryx are present, other than long bristles on the dorsal side of the tail of Psittacosaurus . It seems implausible that one scaled dinosaur would be preserved with no scales but with the underlying collagen fibre structure in situ, when another in the same deposit would have a covering of scales but no evidence of collagen preservation. Furthermore, the scales of Psittacosaurus are likely preserved because of the refractory properties of melanin (in melanosomes) and calcium phosphate, residues of which are embedded in mineral salts commonly found in the vertebrate epidermis, rather than through mineralisation or organic preservation of collagen or keratin (Pautard 1964;Mayr et al. 2016;Vinther et al. 2016).
The exceptionally preserved feathers in stem birds in the same deposits show similarities to the integument of Sinosauropteryx, including pigment preservation, but no evidence of preserved collagen fibres (Clarke et al. 2006;Zhang et al. 2006Zhang et al. , 2010Wogelius et al. 2011). Furthermore, pterosaurs from the Jehol Biota show both dermal soft tissue preservation (actinofibrils) and integumentary appendages (pycnofibres) which are preserved in different manners within individual specimens (Kellner et al. 2010). The actinofibrils are lightly coloured, geometrically arranged structures (herein shown not to exist in Sinosauropteryx) while the pycnofibres are darkly coloured filaments similar to those of Sinosauropteryx, which are most likely preserved by their pigment content (Kellner 2010).
Original colour patterns have been found in pterosaur integuments indicating the high likelihood of pigment preservation (Vinther 2015a). The presence of pigment (via melanosomes) is a key piece of evidence uniting the integumentary appendages of vertebrates including theropods, ornithischians, pterosaurs and early representatives of Aves in the Jehol Biota. Thus far no satisfactory alternative explanation for the dark colour of these structures has been provided (Vinther 2015a(Vinther , 2015b, although it has been suggested by some using antibody immunisation experiments that keratin should also preserve (Pan et al. 2016), which would need direct chemical confirmation. Melanosomes are characteristic for harbouring melanin in vertebrates and are often present in high abundance in hair and feathers (Vinther 2015a). They are now known to be the predominant reason that hair and feather outlines preserve in exceptional fossils, and are responsible for their preservation as organic (dark coloured) residues (Colleary 2015;Vinther 2015aVinther , 2015b. Alongside morphological evidence of similarity between fossil and modern melanosomes, a number of recent papers have confirmed the presence of melanin in fossil integumentary structures chemically (Colleary et al. 2015;Gren et al. 2016;Pan et al. 2016). The melanosomes found in Sinosauropteryx conform to the morphology of modern phaeomelanosomes, which are known to impart rufous or chestnut hues to feathers (Zhang et al. 2010;Vinther 2015a).
Criticisms of the presence of melanosomes in the integument of Sinosauropteryx (Lingham-Soliar 2011) are poorly supported, and include observations that a chestnut colour appears throughout the matrix, which does not conform to our understanding of the nature of pigment preservation and the taphonomy of melanin (Colleary et al. 2015;Vinther 2015a). Preserved pigments rarely retain their original colour through diagenesis, and no case is known where phaeomelanosomes provide rufous or chestnut colours in a fossil (Li et al. 2010;Colleary et al. 2015;Vinther 2015a). This can be seen in the integument of all specimens of Sinosauropteryx, which is preserved as dark brown-black pigment, differing markedly from any other features of the matrix except for the eyes and abdominal soft tissues (Fig. 1), which also contain abundant melanin in modern animals (Lindgren et al. 2012;Vinther 2015a). In Sinosauropteryx, the preserved dark integumentary pigment is confined to the filaments themselves, marking out their structure, and is not found outside the filaments as would likely be the case if claims that leaching from overlying skin were true (Lingham-Soliar and Plodowski 2010). Orange colours in the matrix are likely produced by iron oxides and are not related to biological pigments. Claims that SEM images of melanosomes in Sinosauropteryx resemble collagen 67 nm D-banding are based on low resolution expanded images of the originals, which show little more than groups of pixels highlighted to infer a desired pattern (Zhang et al. 2010;Lingham-Soliar 2011, fig. 2).
Recently, an ornithischian dinosaur (Kulindadromeus) from Siberia was described with both scales and feather-like structures preserved together, with extensive comparison made between the preservation and morphology of each structure type (Godefroit et al. 2014a). This is further evidence that scales and feathers have a similar propensity for preservation within specific deposits, and the specimens show that when scales and feathers are preserved in the same specimen they can readily be distinguished. Reinterpretations of the feathers in Kulindadromeus as collagen fibres by Lingham-Soliar (2014) have been rebutted and cannot be considered likely in the absence of evidence (Godefroit et al. 2014b).

CONCLUSIONS
The debate about avian origins has generated a great deal of controversy in the past few decades. Despite an accumulation of substantial evidence that birds are dinosaurs, some voices continue to challenge this evidence. Our focus here has been on the papers that criticise the evidence for feathers in Sinosauropteryx, and we show that these studies have largely misinterpreted sedimentary and preparation structures as primary anatomical features.
We believe that the use of low quality images instead of first-hand study of the specimens has made it difficult to determine which traces in the fossils might be bone, feathers, sedimentary 18 features or preparation marks. Our examination of the specimens and use of high quality images has shown how these features have been misidentified. Our reanalysis shows that the studies arguing for collagen preservation in the integument of Sinosauropteryx and Stenopterygius are thus refuted.