look for evidence of gastric digestion. This bone fragment was analyzed, along with 15 bones from modern hyena scats as controls, by means of a FEI Inspect Low Vacuum scanning electron microscope (SEM), which is housed at the Museo Nacional de Ciencias Naturales. Observations were done in backscattered electron mode, combined with secondary electron emission mode, at 30 kV, 0.6–0.33 Torr. This type of SEM enabled us to analyze specimens directly with no coating or any other pre-treatment.

Chemical Analysis of the Coprolites

Both coprolites from Azokh 1 Unit II were chemically analyzed at the MNCN laboratories. Sample labeled 5153A corresponds to the outer layer of the coprolite with sediment attached to the surface (residue from pollen cleaning) and 5153B contains exclusively the inner part of the coprolite (also paleogenetically analyzed here). Coprolite 5246 was taken as a whole and the analyzed sample is a mixture of the outer and inner layers of the coprolite. A modern hyena scat (2160) from Burungi (Tanzania) was analyzed as control. All samples were ground to a fine powder using an agate pestle and mortar to be chemically analyzed. These samples were analyzed for X-Ray Diffraction (XRD, Philips PW-1830) and X-Ray Fluorescence measurements (XRF, Philips PW-1404) to obtain their mineral and element compositions respectively.

Paleogenetics and Paleogenomics

The intact half of the coprolite 5153 that was not used for the plant microfossil studies, has been subject to a paleogenetic and paleogenomic analysis. The pre-PCR experiments were carried out in the high containment laboratory of the Institut Jacques Monod (http://www.ijm.fr/ijm/plates-formes/pole- paleogenomique/), the post-PCR experiments in a series of separated laboratories of the Institut Jacques Monod designed to minimize carry-over contamination as described previously (Bennett et al. 2014). The surface of the coprolite was removed to reduce contamination with exogenous, environmental DNA and the quantity of inhibitors that can be enriched on the surface. Removal of the outer layer was performed with a sterile scalpel and 583 mg recovered from the inner part of the coprolite with a slowly moving drill. The powder was extracted in 10 ml extraction buffer (0.5 M EDTA pH 8.0, 0. 25 M potassium dihydrogenphosphate, 0.14 M betamercaptoethanol) and purified over silica columns (Qiagen) as described (Charruau et al. 2010). The total DNA quantity (measured on a Qubit® 2,0 Fluorometer and comprising environmental and endogenous DNA) was 0.93 ng/µl. The

purified DNA was amplified via quantitative real-time PCR (Pruvost and Geigl 2004).

Several procedures to prevent contamination were implemented in the protocol, such as elimination of contamination due to carry-over (Pruvost et al. 2005) and to reagents (Champlot et al. 2010). The extract strongly inhibited the polymerase in the polymerase chain reaction (PCR) with a 5 cycle delay at 10% reaction volume, 2.8 cycle delay at 5% reaction volume, and 0.1 cycle delay at 2.5%. Two PCR primer pairs targeting a 111 bp and a 84 bp (Bon et al. 2012) fragment of the mitochondrial cytochrome B region of the hyaenidae were used. When the primer sequences are removed from the PCR product sequences, the 84 bp and the 111 bp leave 43 bp and 64 bp, respectively, of informative sequence. Moreover, two primer pairs were designed that targeted 103 and 106 bp regions of the hypervariable region of Ursus spelaeus. In addition, a third primer pair was designed to amplify an 88 bp region of the NADH dehydrogenase 2 gene (ND2) with equal efficiency between both Ursus and Hyaenidae, the 36 bp internal sequence of which would differentiate between the two.

When using the hyena-specific primers, PCR products of 84 bp and 111 bp were obtained from the extracts at 2.5%, 5%, and 10% of reaction volume (20 µl total volume reactions), despite the inhibition with the larger extract volumes. A single 88 bp product was amplified using the universal bear-hyena primers. No product was obtained when using the bear-specific primer pairs. The PCR products were directly sequenced after purification.

In order to compare the obtained sequences with those of the three extant hyena species, DNA from the hair of two male brown hyenas from the Zoo “Fauverie du Mont Faron”, France, was extracted as described previously (Charruau et al. 2010) and analyzed using the same primers and PCR conditions as described above, but in a laboratory of the “Institut Jacques Monod” where modern DNA is analyzed. For next-generation high throughput sequencing, a library was prepared in the high-containment laboratory using the double-stranded DNA procedure as described in Bennett et al. (2014). The size-selected library was then amplified and sequenced on an Illumina MiSeq with paired-end 100 bp-long reads, using the manufacturer’s workflow.

Results

Bone and Coprolite Morphometry

The two complete and undamaged coprolites (5153 and 5246) measure 50 × 49 × 33 mm and 48 × 47 × 30 mm (major axis by minor axis by length) respectively. Coprolite

sizes were compared to modern and fossil scats of spotted hyena (Crocuta crocuta) and brown hyena (Hyaena brunnea) by Fernández-Jalvo et al. (2010a). We added the raw values of modern spotted hyena scats from Colchester Zoo and hyena coprolites from European sites (Larkin et al. 2000;

Pesquero et al. 2011), as well as more measurements from hyena and other carnivore coprolites from Laetoli (Tanzania) measured by Harrison (2011) see Fig. 12.1a.

The fossil bone fragment found in the residues from sawing and cleaning the coprolite 5153 is 5 mm long and,

therefore, taxonomically unidentifiable. The small piece of fossil bone shows signs of moderate digestion (Andrews 1990), with slightly rounded edges (Fig. 12.2a) and a smooth surface (Fig. 12.2b), and in addition there is heavy diagenetic cracking on the bone surface (Fig. 12.2c). In contrast, bones contained in modern hyena scats (both spotted and brown hyenas) have higher degrees of rounding (Fig. 12.2d), enlarged bone porosity (Fig. 12.2e) and a characteristic “torn-like” damaged surface (Fig. 12.2f).

Chemical Analyses of Coprolites

and Modern Scat

The XRD diagrams provide information of the mineral content in the sample and crystalline traits. Broadness of the peaks indicates low crystalline structure of the mineral content (Kolska Horwitz and Goldberg 1989; LeGeros 1994; Mulla et al. 2012). Diagrams of XRD shown in Fig. 12.3 have broad and irregular curves for the coprolites and modern scat and indicate that none of these samples are highly crystalline.

Amorphous phases obtained by XRD (Table 12.1) mainly refer to poor-crystallized minerals, but it may also be influ- enced by organic matter or volatile content. Azokh samples have higher amorphous content, especially sample 5153B (17.1% inner coprolite), than the modern scat (Table 12.1). The Loss on Ignition (LOI, Table 12.2) is the weight loss before and after heating. LOI values are also closely correlated to the organic matter and volatiles content in the sample (Heiri et al. 2001). Chemical analysis by XRD of the recent scat has yielded only hydroxyapatite from bones ingested by

Fig. 12.3 X-Ray diffraction diagram of modern scat (below), Azokh coprolite 5153 fraction from outer layers and sediment attached (5153A). Azokh coprolite 5246 including both outer and inner layers. Azokh coprolite inner fraction of 5153 is the top XRD profile. Note hydroxyapatite peaks at 2q 44–54 region show some double simultaneous peaks (close to hydroxyapatite and fluorapatite), although in most cases the peak corresponds to the hydroxyapatite mineral

modern hyenas; amorphous phases are absent and the LOI is relatively high. Azokh coprolites have a low LOI compared to modern scats and moderately high amounts of amorphous phase (especially sample 5153B with 17.1% inner coprolite). Samples 5153A and 5246 have also calcium phosphate content as well as other minerals (feldspars and micas) from the surrounding sediment attached to the coprolites, but sample 5153B (cleaned inner layer of the coprolite) mainly contains hydroxyapatite (72.9%) and quartz (10.1%). However, influence of neo-formed minerals from the diagenetic processes cannot be fully discarded (see discussion).

Fluorescence results are displayed in Table 12.2. The XRF analysis has yielded similar results in all samples (fossil and modern scats). Some differences are observed with regard to silica and some elements, such as aluminum or potassium (components of feldspars) that are more abundant in Azokh coprolites. Trace elements, such as strontium, are more abundant in modern scat. The low value of Loss on Ignition (LOI) from Azokh coprolites (below 16.2%) indicates low proportions of organic materials and/or volatiles.

Paleogenetic Analysis of the Coprolite

In five out of eighteen attempts, PCR amplification of the coprolite extract yielded products of the 84 bp long mitochondrial DNA fragments of the cytochrome B. The longer, 111 bp fragment, however, yielded only non-specific amplifications of modern human contaminants. Depending on primer specificity, amplification of human sequences is an expected result, due to the low copy number and size degradation characteristic of targeted ancient DNA, as well as the ubiquity of modern human DNA in reagents and samples, particularly those that have not been aseptically excavated. One of nine attempts to amplify the 88 bp universal bear/hyena sequence of the ND2 gene was successful.

The sequences obtained for the 84 bp hyena-specific fragment were unambiguous and identical. Their comparison with the mitochondrial cytochrome B sequences in GenBank (NCBI Blast search) showed that the closest match was the cytochrome B sequence of the hyena (Rohland et al. 2005), rather than that of the cave bear (Krause et al. 2008). As a precaution, the sequences obtained were also compared with bovine (Bos taurus) and human sequences, the DNA of which can often contaminate ancient DNA analysis through reagents and handling. These sequences showed no similarity to either of these potential contaminants. The sequence from the ND2 gene fragment amplified with bear/hyena universal primers was also determined to be hyena sequence. Since the informative sequences of the coprolite obtained with the 84 bp CytB fragment most closely matched the brown hyena (Hyaena brunnea) and no ND2 sequences for

brown hyena were available in Genbank, we subsequently determined the sequence of the analyzed ND2 gene fragment from modern brown hyena using hair samples from two brown hyenas from a zoo (Fauverie du Mont Faron, France). We also amplified and sequenced the two CytB fragments from these individuals, and their sequences were identical to the brown hyena sequences deposited in Genbank. Alignments of both the CytB and ND2 fragments were concatenated, compared to those of extant bears, the extinct cave

Table 12.2 Fluorescence (XRF) results from Azokh coprolites and modern scat

bear Ursus spelaeus, the cheetah (Acinonyx jubatus), the tiger (Panthera tigris), and to the different extant hyena species and the extinct cave hyena (Fig. 12.4a). A maximal likelihood phylogenetic tree was constructed using PHYML (Guindon & Gascuel 2003) (Fig. 12.4b).

It can be seen from both the alignment and the tree that the DNA sequences recovered from the coprolite using the highly sensitive targeted PCR approach clearly belong to Hyaena brunnea and can be unambiguously distinguished from the other hyena species as well as from other Feliformia. The Ursidae sequences are even more distantly related.

To explore the possibility that the coprolite could contain DNA sequences from other organisms that were not targeted with the directed PCR approach used, we constructed a library from the total DNA extracted from the coprolite and sequenced a subset of this library using the Illumina Miseq platform. High throughput sequencing is an ideal approach to analyze the DNA composition of environmental samples (Shokralla et al. 2012), including feces (Murray et al. 2011). Shotgun next-generation sequencing was performed allowing the DNA molecule present in the extract to be randomly sequenced. This approach has two advantages: first, it provides an unbiased view of the DNA sequence composition of a fossil bone extract; second, it provides sequence information of very short DNA sequences that are too short to be analyzed with the targeted PCR approach. The disadvantage of this approach is that it does not discriminate between environmental and endogenous DNA and it can generate only minute amounts of sequencing data from endogenous DNA because of the pervasive nature of environmental DNA contamination. We analyzed the sequences of 619,848 fragments from a subset of the library of the Azokh coprolite (Fig. 12.5a). Of these, only 81,063 (13.7%) sequencing reads could be uniquely mapped to sequences present in databases (Fig. 12.5b). The vast majority of these uniquely mapped sequences, i.e. 95.6%, are of bacterial, archaeal or viral origin. The sequencing reads were also mapped to the human, cat, dog, and cow genomes, as well as the cave bear and striped hyena mitochondrial genomes. None of these attempted mappings revealed any appreciable presence of mammalian DNA, apart from the low level human sequences (0.4%), which are expected background contaminants with standard excavation techniques and non-decontaminated reagents. This

result argues in favor of poor DNA preservation since the extract contains no detectable endogenous DNA from any of the likely scat producers, and indeed, no vertebrate DNA at all was detected (at least not in the 619,848 reads analyzed) beyond trace sequences that cannot be excluded from common biological reagent and handling contaminants. It is concluded that if any endogenous sequences are still preserved in this sample, they are too rare to be detected using a global approach without prior enrichment.

In contrast to the targeted PCR approach, which is highly sensitive to longer, targeted sequencing reads, no sequencing reads indicating the presence of Hyaena brunnea were obtained via next-generation sequencing. It is noteworthy that the human mitochondrial DNA sequence obtained with the PCR approach does not match any individual working in the paleogenomics laboratory in Paris indicating that human contamination occurred most likely upstream to this analysis, or was introduced by reagents.