Detailed methods descriptions for each section are provided in the Supplementary Information.
Archaeological sampling
The majority (22 out of 28) of ancient human remains analysed in this study were housed at the National Museum of Bloemfontein at Florisbad Quaternary Research Station, Free State, South Africa, whereas the remaining human remains were housed at the School of Anatomical Sciences at Wits Medical School University of Witwatersrand, Gauteng, South Africa. Sampling of teeth and bone elements was done on site in a mobile clean-laboratory and the sampled elements were immediately returned. The samples were transported to the Ancient DNA Laboratory at Uppsala University, Sweden, for further analyses.
Radiocarbon dating
In total, 20 human remains were sampled for accelerator mass spectrometry (AMS) radiocarbon dating and sent to Beta Analytic where bone collagen extraction were performed for AMS radiocarbon dating and for stable dietary isotope analyses using isotope-ratio mass spectrometry. Conventional radiocarbon dates obtained from Beta Analytic were modelled using BetaCal 3.21 and SHCal13 (ref. 51), whereas radiocarbon ages for three previously dated individuals were modelled using OxCal v.4.4 and SHCal20 calibration curves52,53.
Ancient DNA retrieval
DNA was extracted either as in ref. 54 with modifications as described in ref. 55 or ref. 56. DNA extracts were prepared from the 28 human remains (between 1–7 DNA extracts from each individual). One blunt-end library was prepared for each DNA extract and sequenced for screening of endogenous human DNA content and verification of ancient DNA using deamination profiles57,58. Authentic ancient DNA (and >1% human DNA) was found in 32% of the sequencing libraries that were used for screening. For DNA extracts for which the proportion of ancient human DNA was >2%, sequencing libraries were prepared using UDG treatment to minimize post-mortem deaminations and increase the sequencing depth59. For DNA libraries for which the proportion of human ancient DNA was <2%, additional blunt-end libraries were prepared to increase the sequencing depth. Owing to low amounts of endogenous human DNA, some of the libraries were enriched using MY-bait African Human Whole Genome Capture Kit (MYcroarray) according to the manufacturer’s instructions (MYbaits manual v.2.3.1) and amplified as described previously4. All libraries were sequenced on either the HiSeq X10 or a NovaSeq 6000 (SP flow cell) Illumina sequencer with either 100 bp or 150 bp paired-end chemistry.
Data processing and authentication
Adapters and low-quality bases were trimmed from the sequencing data and paired-end reads were merged if an overlap of at least 11 bp was detected between the forward and reverse read, using either the script MergeReadsFastQcc.py60 or AdapterRemoval (v.2.1.7)61. The reads were then mapped against the human reference genome build 37 (hs37d5) using bwa aln62,63. BAM files from resequenced libraries were merged using Samtools merge (v.0.1.19)64 before PCR duplicates were identified and collapsed using a slightly modified version of FilterUniqeSAMCons.py60. Non-UDG and UDG-treated libraries were then separately merged per individual and reads shorter than 35 bp and with <90% consensus with the reference sequence were filtered out using percidentity_threshold.py65. Mitochondrial contamination was estimated using two different methods—Green and contamMix66,67. Additional contamination estimates were performed on all individuals with genome-wide coverage >2× using VerifyBamID68. Contamination estimates were generally low for both the nuclear and mitochondrial genome (Supplementary Data 1 and 5).
Determining sex, uniparental haplogroups
... continue reading