Raw isotopic data and collagen quality criteria.

Guiry and Szpak [43] recommend using a threshold of 3.28 in atomic C:N ratio to assess the quality of modern collagen samples of mammals. According to their study, samples with higher values result in a significant correlation between atomic C:N and δ13C, and δ13C values for these samples may be contaminated by non-collagen material. Among our samples, 48 had C:N values > 3.28. Not to exclude excessively many samples we decided to use a threshold of 3.33, that corresponds to the maximum ratio determined based on amino-acid composition by Guiry and Szpack [43]. Including samples up to this threshold resulted in a significant correlation between C:N and δ13C and we may thus have included samples with slight contamination. However, given the coefficient of the regression of δ13C against C:N was -8.13, a difference in C:N of 0.05 (3.28 to 3.33) would only lead to a change in δ13C of 0.4, corresponding to less than 0.2 standard deviations of our δ13C values. The bias this might have introduced would thus be negligible.
(XLSX)