Rapid range shifts and megafaunal extinctions associated with late Pleistocene climate change

Large-scale changes in global climate at the end of the Pleistocene significantly impacted ecosystems across North America. However, the pace and scale of biotic turnover in response to both the Younger Dryas cold period and subsequent Holocene rapid warming have been challenging to assess because of the scarcity of well dated fossil and pollen records that covers this period. Here we present an ancient DNA record from Hall’s Cave, Texas, that documents 100 vertebrate and 45 plant taxa from bulk fossils and sediment. We show that local plant and animal diversity dropped markedly during Younger Dryas cooling, but while plant diversity recovered in the early Holocene, animal diversity did not. Instead, five extant and nine extinct large bodied animals disappeared from the region at the end of the Pleistocene. Our findings suggest that climate change affected the local ecosystem in Texas over the Pleistocene-Holocene boundary, but climate change on its own may not explain the disappearance of the megafauna at the end of the Pleistocene.

-independent ordination analysis based on ASV diversity. The two upper panels represent bulk bone data analysed with mitochondrial assays targeting vertebrates (Mam16S and 12Sv5; subsampled to 7247 reads per sample), whereas the two lower panels represent sediment samples analysed with chloroplast assays targeting plants (rbcL and trnL; subsampled to 5374 reads per sample). As opposed to figures 3b and 4b in the main text which are based on the taxonomic record inferred from the DNA data, these ordination analyses are based on ASVs (amplicon sequence variants). Hence, this approach compares DNA sequences across samples without assigning these to taxa first. Accordingly, unknown sequences without matches in the database can be taken into account. LGM (n=11), Bølling-Allerød (n=5), YD (n=3) and Early Holocene (n=11). Center line: median. Box limits: upper and lower quartiles. No data points were excluded. Source data are provided as a Source Data file.  Number of contaminant taxa in samples excavated following ancient DNA guidelines (bulk bone data from this study, n=72, and sediment data from this study, n=15) compared with laboratory controls (n=10) and material that was excavated for morphological analyses by Toomey (n=10). Contaminant taxa include human (Homo sp.), dog/wolf (Canis sp.), sheep (Ovis sp.), cattle (Bos sp.), goat (Capra sp.), cat (Felis sp.), chicken (Gallus gallus) and pig (Sus scrofa). Center line: median. Box limits: upper and lower quartiles. Whiskers extends to 1.5xIQR (inter quartile range), no data points were excluded. Source data are provided as a Source Data file.

Supplementary Note 1. DNA preservation and other taphonomic biases
Unlike other caves in North America, Hall's cave exhibits exceptional DNA preservation across the chronosequence. There are several lines of evidence for a uniform level of DNA preservation throughout the sequence that we investigated: (1) Both the bulk bone and the sediment samples all contained amplifiable endogenous DNA. (2) As opposed to what would be expected in a setting of increasing DNA damage through time, diversity is positively correlated with sample age for bulk bone assays ( Figure 3c) and while diversity drops in the Younger Dryas for sediment samples, it increases in the Pleistocene (Figure 4c). (3) The presence of certain taxa throughout the sequence, such as Sylvilagus sp. and Peromyscus sp., serves as taphonomic controls, illustrating that there is no detectable decrease in DNA preservation over time. (4) There is no evidence of systematic changes in the relative DNA content when comparing qPCR results across assays and sample type. As depicted in Supplementary Figure 1, the 16S assay does appear to display higher relative DNA concentrations at depths 100 to 150 cm BDT, however, if this pattern represented variable DNA preservation in the cave, we would expect the same pattern to be present in the 12S assay, which is not the case. Similarly, for the sediment samples (Supplementary Figure 2), sequence A displays a small increase in DNA concentration around 140 cm BDT, but this pattern is not reflected in sequence B. Hence, we do not find evidence of any significant changes in DNA preservation over time that would affect the interpretation of our results.
The depositional processes of a faunal assemblage, such as Hall's Cave, must be understood to correctly infer paleoenvironmental changes from each accumulation. In Hall's Cave, the taphonomic processes were thoroughly investigated by Toomey 24 , who found that the assemblage was mainly accumulated by predation or bone gathering. Raptors, in particular large owls, were found to be an important contributor to the assemblage, preying mainly on small mammals the size of rabbits and smaller. Furthermore, both small and large mammalian carnivores contributed to the assemblage by dragging prey into the cave. In our data, the presence of lagomorphs and rodents in all time periods indicate that raptor accumulation most likely occurred throughout the sequence, while the disappearance of most carnivores and large mammals at the beginning of the Younger Dryas period indicates that accumulation by large carnivores diminished over time. This change is unlikely to be a result of a change in the shape of the cave entrance as large animals such as Odocoileus and Canis latrans are found in Holocene layers (though infrequently). Furthermore, as carnivores represents the top of the food chain, their disappearance is unlikely to be a result of a change in deposition but must reflect a change in the surrounding ecosystem. The disappearance of large herbivores at the onset of the Younger Dryas, on the other hand, could be linked to the disappearance of carnivores. Still, the continued presence of raptors in the cave suggests that the loss of diversity in frogs and reptiles in the Holocene reflects a loss of these species in the area surrounding the cave. Compared to the faunal assemblage, the plant data are less affected by taphonomic processes. As noted by Toomey, hackberry seeds might have been washed into the cave from the surrounding land surface, but other dispersal routes, such as wind or transportation on birds, insects and other animals would also have contributed to the plant assemblage 25 . However, as the depositional processes for the plant and animal assemblages are very different, the two assemblages serve as important validations of each other.
In combination, data from pollen, plytholiths, sedimentary DNA, bulk bone DNA and morphological bone identifications details how the landscape changed around Hall's Cave from the Pleistocene to the Holocene. While certain species groups could be affected by a change in depositional processes, it is very unlikely that all species are.

Supplementary Note 2. Sampling and laboratory contamination
To monitor contamination from sampling through to sequencing, we included non-template controls at each stage of laboratory processing. For both bulk bone and sediment samples, at least one extraction blank and one PCR-blank was included in each batch of sample preparation. For bulk bone samples, a total of 10 blanks were sequenced. From these we identified Homo sapiens, Sus scrofa and Canis sp., which were marked as laboratory or field contaminants and excluded from downstream data analysis. Moreover, although not identified in the controls, we identified a number of other common contaminants in the data. These include: Gallus gallus, Bos sp., Ovis sp., Capra sp. and Felis sp. Although some of these identifications could be from endogenous DNA (e.g. Capra), they are widely reported as common laboratory contaminants in ancient DNA studies 14,26,27 and were therefore marked as possible contamination and excluded from downstream analyses. Furthermore, the identification of turkey (Meleagris gallopavo) and caballine horse (Equus caballus/lambei/scotti) could represent contamination because these species are often associated with human everyday life (although rarely reported as contaminants in the literature). Hence, to confirm that these identifications represented endogenous DNA, we re-extracted and amplified the samples in which they were detected. Reassuringly, all re-processed samples confirmed their presence. Lastly, as we have marked Canis sp. as contamination in our data, the identification of Canis latrans could pose a problem. Hence, we confirmed that sequences assigned to C. latrans, were in fact distinguishable from the contaminant sequences from Canis (most likely Canis lupus familiaris) that we detected. In cases where reads could not be distinguished between different Canis species, the reads were marked as possible contamination and excluded from downstream analysis.
From the bulk bone samples, we did not find any evidence of cross contamination between samples, however, in one sample (HCB23_B) we identified significant contamination from fauna of New Zealand, which could stem from samples processed in the same laboratory for a different project. To identify the source of this contamination, we re-extracted and amplified the bone powder from HCB23_A and HCB23_B, which confirmed the presence of significant DNA contamination in the bone powder of both of these samples. Next, we returned to the original bulk bone samples, sampled and processed another 2x50 bones each. These samples contained only species from Texas and were comparable to other samples from the surrounding layers. Hence, we conclude that the bulk bone powder from the two subsamples of HCB23, was contaminated during the bone grinding stage. Most likely, this contamination stems from the reuse of a grinding pot that had not been cleaned properly. To confirm that this contamination event was a single incidence, we processed two 'grinding blanks' in which 15 mL of ultrapure water was run in the ball mill under the same configuration as the bulk bone samples. After grinding, the water sample was concentrated to 500 µL in an Amicon®Ultra-4 Centrifugal Filter (Millipore) and processed as a bulk bone sample. From these samples, we only identified background contamination from Homo sapiens.
This study includes samples that were excavated following strict ancient DNA guidelines (excavated for this study), and samples that were excavated solely for morphological analyses where no measures were taken to limit DNA contamination (Toomey et al. 1993). Hence, our dataset offers a unique opportunity to compare the level of contamination between the two. Not surprisingly, we found that the fossils excavated by Toomey exhibited a higher number of contaminant taxa (mean: 3.0 contaminant taxa per sample) than those excavated for this study (mean: 1.3 contaminant taxa per sample; Supplementary Figure 3). The level of contaminant taxa for the data excavated for this study is comparable to that of the controls (mean: 1.3 contaminant taxa per sample) and the sediment samples analysed with vertebrate assays (mean: 0.8 contaminant taxa per sample), indicating that the only source of contamination for these samples is background laboratory contamination. The higher level of contamination for the samples excavated by Toomey, on the other hand, suggests that excavations that do not follow ancient DNA guidelines have significantly increased risk of contamination. For example, the two contaminants Felis sp. and Gallus gallus are only detected in the Toomey samples. Fortunately, this level of contamination does not affect the interpretation of the data, as the common contaminants are easily distinguishable from the local fauna in Texas. This does, however, highlight the need for secondary authentication in metabarcoding studies where domesticates are detected -in particular, in ancient DNA studies on samples excavated for other purposes than ancient DNA.
For the plant data, a total of three extraction blanks were sequenced. From these we identified three contaminant reads (taxa: Cicer sp., Brassicaceae, and 'no blast hit'), which were abundant in the extraction blanks, and present in low concentrations in some of the test samples-these reads were excluded from downstream analysis. Furthermore, in sediment extraction blank 3 we identify 9 copies of the Celtis read that is present in high numbers in all of the test samples. This most likely represents cross contamination during the PCR-reaction. However, this low level of cross contamination is unlikely to affect the sediment samples, as they all exhibit high concentrations of endogenous DNA. Extraction and PCR blanks are likely to overestimate cross contamination levels as low-level background contamination is more likely to amplify in samples with no endogenous DNA.