The recent discovery of titanosaurid dinosaur eggs containing embryonic remains in Argentina shed light on the phylogeny (Chiappe et al. 1998, 2000, 2001), development (Chiappe et al. 2001) and behaviour (Chiappe et al. 2004) of this important group of extinct vertebrates. Dinosaur eggs from this locality preserve not only embryonic bone material, including articulated skulls (Chiappe et al. 2001), but also the first reported fragments of embryonic dinosaur skin, preserved not simply as impressions, but in three dimensions (Chiappe et al. 1998; figure 1). The preservation of embryonic soft tissues is highly significant, because it indicates unique depositional and geochemical conditions that resulted in rapid and complete mineralization of these very labile soft tissues before degradation could occur. These taphonomic conditions extend to the eggshell, and as a result, both eggs and their contents represent a unique opportunity to elucidate taphonomic conditions resulting in exceptional preservation. We define exceptional preservation as a taphonomic mode preserving soft tissues, original biomolecules or their altered fragments, original mineralogy and/or other features normally lost during diagenesis. The exceptional morphological preservation of these specimens encouraged analytical experiments designed to determine if this preservation extended to the molecular level.

Figure 1 Transmitted light microscopy image of skin fragment from sauropod embryo. Intricate scale pattern is clearly visible.

The microscopic, molecular and elemental analyses we describe herein are limited to eggshell, rather than embryonic remains, because of the destructive nature of these analyses. Eggshell was more prevalent (though less morphologically informative) than embryonic remains, and serves as a proxy for the preservational state of the rarer embryonic bone and skin. Immunological data support the presence of at least some immunogenic molecules in these fossil shells that, although altered, share characteristics in common with organic material in extant eggshells.

Sauropod eggs and embryos were recovered from the Anacleto Formation (uppermost unit of Neuquén Group) near Auca Mahuevo (Chiappe et al. 1998, 2000, 2001, 2004). This unit has been assigned by Legarreta & Gulisano (1989) to middle Campanian, a date recently confirmed by Dingus et al. (2000) for Auca Mahuevo outcrops using palaeomagnetic data.

Sedimentological evidence suggests a palaeoenvironment characterized by mixed-load meandering rivers developed over a low-gradient, extensive floodplain. Thick overbank deposits show signs of recurrent, periodical flooding events that buried the nests and contributed to preservation of the eggs over time. Pedogenic caliche levels and heavy phytoturbation (root-traces) influence the overbank and older channels deposits, indicating well-developed palaeosols and abundant vegetation. However, the only fossil evidence of this vegetation is in the form of small- and medium-sized plants.

Caliche formation and vertisol development, together with the presence of evaporitic sediments in the sequence, suggest warm and relatively arid palaeoclimatic conditions with a pronounced alternation of wet and dry seasons (Esteban & Klappa 1983; Goudie 1983; Bridge 1984; Jerzykiewicz & Sweet 1987).

The eggs are arranged in clutches distributed across at least four stratigraphic layers in an 85m thick sequence of sandstone, siltstone and mudstone (Chiappe et al. 2000, 2004). In each horizon, clutches typically contain 15–30 eggs arranged in layers, and are found in depressions underlain by either mudstone or sandstone. After the eggs were deposited, flooding events caused the depressions to be filled with silty mud that that settled out of suspension. Channel and crevasse-splay deposits, represented by thin sandstone lenses that interfinger laterally with mudstone units, are found within some regions of the egg-bearing intervals. The clutches are located within silty, reddish-brown mudstone units with slickensided surfaces that indicate vertisol development (Chiappe et al. 2000). The depressions containing several clutches truncate bedding and sedimentary structures, suggesting excavation by the adult sauropod (Chiappe et al. 2004).

See Electronic Appendix.

The quality of morphological preservation of these fossil shells is demonstrated using transmitted light and scanning electron microscopy (figure 2a,b). External shell morphology consists of closely spaced knobby projections, or tubercles (T; figure 2a), with straight, unbranched pores (P) for gas exchange arranged in the depressions between tubercles (Chiappe et al. 1998). Radial ground sections of some specimens show intact shell units (figure 2a) as vertical columns with roughly parallel, distinct margins, showing minimal alteration of original structure. Fine laminations seen in transmitted light microscopy (figure 2a, arrowheads) represent periodic accretion of mineral upon the organic matrix of the shell during biomineralization, tracking outward growth of the shell from nucleation sites within the external shell membrane (SM) (Vianey-Liaud et al. 1994; Carrino et al. 1996; Chiappe et al. 1998; Carpenter 1999; Grellet-Tinner et al. 2004). Erosion seen at the base of the shell units (figure 2a,b) may be the result of demineralization by the embryo or of subsequent early diagenetic dissolution. The borders of these eroded areas are lined with a dark brown, non-translucent material consistent with diagenetically altered organics. In addition, apparent organic cores (OC, figure 2) can be seen embedded in a laminated material (SM) that lies immediately internal to the mineralized shell, features confirmed in electron microscopy (figure 2b). Based upon location, structure and comparison with extant taxa (e.g. figure 2c), we hypothesize that this material may represent remnants of preserved SM, permineralized by calcite precipitation during fossilization (Jackson et al. 2002).

Figure 2 (a) Tangential ground section of sauropod eggshell, showing external topography in the form of small knobs or tubercles (T), shell units, accretion lines indicating periodic mineral deposition (black arrowheads), and organic cores (OC, arrows) embedded (more ...)

This membrane-like feature is also visible in figure 3, which shows a cross-section of a double-shelled egg. These pathological eggs, similar to other described pathological dinosaur eggs (Zelenitsky & Hills 1997; Hirsch 2001), were quite common in this nesting horizon. The fibrous material interpreted as membrane (M) divides the two sequentially biomineralized eggshell layers (Jackson et al. 2002, 2004) consistent with occurrences in extant birds (Jackson & Varrichio 2002 and references therein) and reptiles (e.g. Ewert et al. 1984).

Figure 3 Scanning electron micrographs of double-layered (pathological) sauropod eggshell in tangential section. (a) Tubercles (T) delineate the external surface. Preserved shell membrane (M) in between and separating the two layers of shell. Arrow indicates a (more ...)

Energy dispersive X-ray (EDX) elemental analyses shows that element distributions in both sauropod shell and membrane are within the range of variation seen in comparable extant samples (figure 4a–d; table 1), suggesting minimal diagenetic alteration. Freshly fractured surfaces of sauropod eggshell (figure 4a) also show that the elemental content of the shell differs from that of the membrane (figure 4b) in both weight per cent and atomic per cent (table 1). These compositional differences were most probably present during formation of the egg, as they are consistent with those observed in extant specimens (figure 4c,d). The membrane is enriched in traces of Mn, Mg, Al, Si and K, elements that are also present in one or more examples of extant SMs, but not present in the shell itself. Similar distributions of minor elements are seen in extant shell and membrane from chicken, ostrich and crocodile (table 1), suggesting that these distributions reflect original composition rather than diagenetic artefact. Ratios of the most common elements in both shells and membranes (Ca, C, O) show that extant ostrich shell values are more similar to sauropod shells than to other extant shells in both atomic and weight percentages. These values for membranes across taxa, however, demonstrate wider variation, and reflect the greater degree of mineralization of sauropod membrane relative to other taxa. Weight per cent ratios are depicted graphically in figure 5.

Figure 4 Energy dispersive X-ray (EDX) elemental profile taken through clean fracture of sauropod shell (a) and membrane (b), compared with extant ostrich shell (c) and membrane (d). Element profiles are similar, with a similar content and distribution of trace (more ...)

Table 1 Table showing weight and atomic per cent values for element distribution.

Figure 5 Graphic display of elemental (EDX) data, showing relative values of Ca/C, O/C and O/Ca. These ratios are virtually identical for ostrich and sauropod shell, while ratios obtained from other extant shells vary significantly in these values. Ca/C and O/C (more ...)

The dialysed supernatants of fossil shell extracts did not lyophilize to completion but left a thick, oily residue with a brownish hue. Chicken eggshell likewise left an oily but colourless residue. Conversely, while the sediments had a greater mass to begin with than any eggshell, both sediment extracts and buffer controls had minimal residue after lyophilization, and what remained was colourless, and crystalline rather than oily.

Figure 6 Representative ELISA showing reactivity of pre-immune (white, dark grey bars) and post-immunization (light grey, black) test sera against multiple antigens. Both test sera show significant reactivity with purified ovalbumin, a predominant constituent (more ...)

To demonstrate localization of antigens present in fossilized specimens, and to demonstrate that the antibodies are specific for endogenous antigens and not materials that infiltrated during or after the mineralization process, antisera and pre-immune sera were incubated with ground sections of sauropod, chicken and crocodile eggshell. As in ELISA, pre-immune sera did not react with any of the eggshells tested, but immune sera raised against both chicken and dinosaur shell extracts demonstrated specific and localized reactivity (figure 7). Both sera have similar binding patterns, whether on sauropod (figure 7a,b) or chicken eggshell (figure 7c,d) ground sections. Relatively homogeneous distribution of fluorescent signal across the shell (palisade layer; Jackson & Varrichio 2002) reflect localization of matrix proteins in extant material (figure 7c,d). Binding of both dinosaur antibodies and chicken antibodies is greater in regions high in protein content (e.g. external shell, corresponding to thin layer of organic cuticle, and interior of shell corresponding to mammillary layer (ml) and OC. Similar distribution is seen in ground sections of chicken shell. Very bright and patchy distribution may reflect non-specific adherence of antiserum to roughened regions of the dinosaur shell or infilling mineral, but in most regions corresponding to calcite precipitate (CP, arrows), antibody binding is decreased over the rest of the shell, and signal is localized to the shell matrix, membrane, and OC.

Figure 7 Immunohistochemical localization of antigens in ground sections of extant and fossil eggshell. First two columns are imaged using confocal fluorescence microscopy, third column is transmitted light image. (a),(d) Sauropod shell, and (g),(j) chicken shell, (more ...)

Finally, both antisera were tested against ground sections of crocodile egg, to test antigen similarity and distribution from that seen in extant birds. Figure 8 shows that antibodies to chicken shell react strongly with both crocodile shell and the SM (figure 8a), and fluorescent signal concentrates in the ml, which is rich in organics. Although reactivity of dinosaur antibodies to crocodile shell is significantly less, binding can be seen in the lower portions of the ml and innermost portion of the SM, regions similar to the pattern seen with chicken shell serum. To further demonstrate specificity, dinosaur antibodies were incubated with an excess of dinosaur shell extract to block the binding sites on the antibodies that are specific to sauropod antigen. The antibodies specific to sauropod antigens would then be unavailable to bind antigen in the crocodile shell that was similar to sauropod, but would leaving other, non-specific antibodies free to bind. Additionally, if antibodies raised against sauropod antigens were not specific for those antigens, the more concentrated antigens present in the extant shell would out-compete the dinosaur inhibitor, and binding would be unaffected.

Figure 8 Immunohistochemical localization of antigens in ground sections of crocodile eggshell probed with antichicken and antidinosaur shell antibodies. Left panels are imaged using confocal fluorescence microscopy, right panels are transmitted light images. (more ...)

The inhibited antibodies were then incubated with the crocodile shell as described. The binding pattern seen in figure 8b disappears when specific antibodies are blocked (figure 8c). This supports specificity as well as molecular similarity between antigenic components in the shells of both taxa, and demonstrates the shared nature of some of the components of crocodile shell matrix with those of sauropod eggshells.

The cross-reactivity of chicken antiserum with sauropod shell components, and sauropod antiserum with chicken eggshell was demonstrated by two different assays (figures 6 and 7). These results indicate molecular similarity of shell-derived immunogens. This point is further emphasized by the localization of antibody binding (figures 7 and 8), suggesting that similar biochemical structures have comparative sites of localization in both chicken and sauropod eggshells. Both cross-reactivity and pattern of localization provide evidence that the antisera are detecting endogenous sauropod antigens.

To propose that this pattern arises from contamination would require that the contaminant, molecularly similar to eggshell proteins, infiltrated the shell and distributed in a pattern similar to the distribution of extant shell organics. Additionally, the reactivity of both antisera to ovalbumin (figure 6) would require that the contaminant also be present in this biochemically purified protein. A far more parsimonious explanation is that ovalbumin is one of the cross-reactive materials recognized by the antisera, and that sauropod eggshells retain fragments of original, immunogenic and antigenic organic material which are homologous with some components in extant shell matrices.

We have shown that in these well-preserved fossils, the molecular structure of endogenous antigens may be preserved for more than 70Myr. The results show that antigens preserved within fossil material retain immunogenicity, and that antisera made against both extant and fossil materials can detect these antigens in fossils. Although the exact chemical nature of the antigenic material has not been demonstrated, it has retained similarities to extant native material in regard to those structures recognized by the immune system.

Evidence supporting the preservation of endogenous biomolecules in the pre-Cenozoic fossil record has generally been met with scepticism, because it is assumed that primary organic molecules cannot withstand the alterations and breakdown that occur during diagenesis (e.g. Bada 1985; Runnegar 1986; Logan et al. 1991; Lindahl 1993). Laboratory experiments designed to approximate molecular diagenesis apply physical and chemical parameters not normally encountered in nature (e.g. pH=1, T≥300°C) and do not account for the protective effects of mineral association (e.g. Weiner et al. 1989; Glimcher et al. 1990; Sykes et al. 1995). Therefore, their utility as a proxy for diagenetic processes at the molecular level in naturally preserved samples is somewhat limited. We propose that, while such experiments proved useful information regarding possible degradation pathways, and some recent findings indicate that protein persistence has been underestimated (Collins et al. 2000), a more direct test of molecular longevity is the application of multiple and varied analyses for endogenous molecular components to extraordinarily preserved fossils. We propose that exhaustive investigation of a variety of these exceptional assemblages, from a spectrum of depositional and diagenetic environments, is an important way of assessing biomolecular preservation, because normal diagenetic processes are both slower and less extensive than those simulated in laboratory experiments.

Despite scepticism regarding the premise of long-term molecular survival based on laboratory experiments and predictions of molecular kinetics, many studies have shown that amino acids, short peptides, and amino sugars can persist within fossils over a wide geological age distribution (e.g. Weiner et al. 1976; Westbroek et al. 1979; Armstrong et al. 1983 and references therein; Lowenstein 1981, 1985; Ostrom et al. 1990; Collins et al. 1991; Gurley et al. 1991; Muyzer et al. 1992; Stankiewicz et al. 1997a,b, 1998; Schweitzer et al. 1997a,b, 1999a,b, 2002; Poinar et al. 1998; Collins et al. 1999). Immunological techniques have identified antigenic compounds in fossils of varying ages and from various source taxa (e.g. Rowley et al. 1986; Muyzer & Westbroek 1989; Baird & Rowley 1990; Collins et al. 1991; Lowenstein & Scheuenstuhl 1991; Child & Pollard 1992; Nerlich et al. 1993; Franc et al. 1995; Borja et al. 1997; Schweitzer et al. 2002) including Cretaceous fossils (Collins et al. 1991; Muyzer et al. 1992; Schweitzer et al. 1997b, 1999a,b).

To preserve fossils in an exceptional manner requires early cessation of diagenetic processes, and is attributable to unusual physical and chemical conditions from death through diagenesis, and minimal alteration of fossil material at the macro and microscopic levels. This, in turn, is correlated with the preservation of endogenous biomolecules, fragments of molecules, or biomarkers (i.e. altered molecular fragments that can be traced to the original source, e.g. Hagelberg & Clegg 1991; Hedges 2002; Schweitzer et al. 2002). Because molecular preservation has been correlated with morphological and microstructural preservation (e.g. Hagelberg & Clegg 1991; Marota & Rollo 2002), optimizing the search for endogenous organic components within fossil specimens should involve careful selection of fossil specimens for study. Analyses of such extraordinary samples are a direct test of survivability of biomolecules.

However, molecular analyses of fossils present unique challenges. In part, this is because chemical modifications occur during diagenesis (Mycke & Michaelis 1985; Rafalska et al. 1991; Poinar et al. 1998). These modifications include breaking peptide bonds, removal or alteration of original amino acid side chains, and cross-linking of peptide fragments to other organic degradation products, a process that makes organic material insoluble and difficult to separate into constituent compounds (Macko & Engel 1991; Poinar et al. 1998). Condensation reactions along this pathway may result in the formation of hydrocarbons from proteinaceous precursors. Because these products are hydrophobic and because the original organics are contained within a biomineralized matrix (Weiner et al. 1989; Glimcher et al. 1990; Sykes et al. 1995), the chances of retention of organic material that contains remnants, however altered, of the original compounds are greatly increased. The strong petroliferous odour released upon decalcification of the eggshell supports this pathway of degradation in these eggshells. A lipid-containing molecular complex of degraded organics and antigenic material may explain some of the anomalies seen in our data. For example, low reactivity of antibodies with dinosaur antigen by ELISA may be owing to insufficient immobilization of antigen resulting from lipid or hydrocarbon mixing with the antigens.

Finally, while the organic material extracted from dinosaur eggshells shows characteristics consistent with extant material similarly derived, it is recognized that the antigenic material may or may not be derived from proteinaceous precursors, and is surely diagenetically altered from its original state. We do not claim here that this material represents complete proteins. Indeed, epitopes are known to be only a few amino acids in length (Child & Pollard 1992); therefore, it is possible that antigenic response may be owing to selective preservation of a few peptides, or even altered, fossilized derivatives of peptides.

The antisera that we have prepared may be used to purify the antigenic material through affinity isolation techniques. It may then be possible to perform further biochemical analyses upon these materials, thus offering the promise of subjecting dinosaur tissues to modern molecular techniques, including amino acid sequencing (Schweitzer et al. 2002).

These eggs, containing such fragile and labile elements as embryonic bone and fragments of embryonic skin, attest to unusual taphonomic and diagenetic conditions and provide an opportunity to expand the correlation between unusual morphological preservation and the presence of endogenous molecules.

This research was supported by a grant from the National Science Foundation (EAR 99-73073 to MHS). The Fundacion Antorchas, Infoquest Foundation, National Geographic Society, and Municipalidad de Plaza Huincul provided funding and logistic support during fieldwork.

We thank F. Jackson and the Museum of the Rockies for ground thin section preparation and comments/discussions that greatly enriched this manuscript. We also thank K. Padian for editorial advice, D. Loope, R. Coria, J. Schmitt and L. Dingus for field expertise, and S. Kruger, J. Wittmeyer and L. McLandress for assistance with analyses. Finally, we recognize and thank an anonymous reviewer for his/her thoughtful and insightful comments that contributed greatly to the readability and cohesiveness of this paper.