Temporal genomic data hold great potential for studying evolutionary processes, including speciation. However, sampling across speciation events would in many cases require genomic time series that stretch well into the Early Pleistocene (>1 million years). Although theoretical models suggest that DNA should survive on this timescale 1 , the oldest genomic data recovered so far is from a 560-780 ka old horse specimen 2 . Here we report the recovery of genome-wide data from three Early and Middle Pleistocene mammoth specimens, two of which are more than one million years old. We find that two distinct mammoth lineages were present in eastern Siberia during the Early Pleistocene. One of these gave rise to the woolly mammoth, whereas the other represents a previously unrecognised lineage that was ancestral to the first mammoths to colonise North America. Our analyses reveal that the North American Columbian mammoth traces its ancestry to a Middle Pleistocene hybridisation between these two lineages, with roughly equal admixture proportions. Finally, we show that the majority of protein-coding changes associated with cold adaptation in woolly mammoths were present already a million years ago. These findings highlight the potential of deep time palaeogenomics to expand our understanding of speciation and long-term adaptive evolution.

Evolutionary processes, including selection, can be indirectly inferred based on patterns of genomic variation among contemporary populations or species. However, this often requires unrealistic assumptions of ancestral demography and selective regimes.Sequencing ancient DNA from temporally spaced samples can inform about past selection processes, as time series data allow direct quantification of population parameters collected before, during, and after genetic changes driven by selection. In this 9 4Comment and Opinion, we advocate for the inclusion of temporal sampling and the generation of paleogenomic datasets in evolutionary biology, and highlight some of the recent advances that have yet to be broadly applied by evolutionary biologists.In doing so, we consider the expected signatures of balancing, purifying, and positive selection in time series data, and detail how this can advance our understanding of the chronology and tempo of genomic change driven by selection. However, we also recognize the limitations of such data, which can suffer from postmortem damage, fragmentation, low coverage, and typically low sample size. We therefore highlight the many assumptions and considerations associated with analyzing paleogenomic data and the assumptions associated with analytical methods. K E Y W O R D S : Adaptation, ancient DNA, natural selection, paleogenomics, time series. Impact SummaryThe search for signatures of natural selection on the genome is still most commonly based on screening modern genomes for regions of reduced diversity or increased differentiation between populations. This framework is essentially a snapshot in time of a process that may have played out over many millennia, during which changes in population size, ecology and gene flow between populations may have played a role in determining genetic variation. Here, we outline how utilising ancient DNA (aDNA) techniques to sequence time series of genomes spanning changes in natural selection can provide a more nuanced understanding of how natural selection has impacted genomic variation in present-day populations. In particular, we argue that the advent of paleo-population genomics, in which datasets of multiple individuals spanning millennia have been sequenced, offers unprecedented opportunity to estimate changes in allele frequencies through time. We outline considerations and the types of data that would be needed for the inference of positive selection on traits associated with single and many genes (polygenic), genome-wide negative (background) selection, and balancing selection. However, we recognise that there are currently few datasets existing that are suitable for these types of investigation. There is thus a bias towards study species that have undergone strong selection over relatively recent timescales that are within the scope of aDNA, such as has occurred in domesticated species. We also detail a number of caveats associated with working with aDNA data, which is by its nature comprised of short, degraded DNA fragments, typicall...