Chronological and genetic analysis of an Upper Palaeolithic female infant burial from Borsuka Cave, Poland

Summary Six infant human teeth and 112 animal tooth pendants from Borsuka Cave were identified as the oldest burial in Poland. However, uncertainties around the dating and the association of the teeth to the pendants have precluded their association with an Upper Palaeolithic archaeological industry. Using <67 mg per tooth, we combined dating and genetic analyses of two human teeth and six herbivore tooth pendants to address these questions. Our interdisciplinary approach yielded informative results despite limited sampling material, and high levels of degradation and contamination. Our results confirm the Palaeolithic origin of the human remains and herbivore pendants, and permit us to identify the infant as female and discuss the association of the assemblage with different Palaeolithic industries. This study exemplifies the progress that has been made toward minimally destructive methods and the benefits of integrating methods to maximize data retrieval from precious but highly degraded and contaminated prehistoric material.


INTRODUCTION
Human remains from the Upper Palaeolithic (UP) in Eurasia are sparse, but the period is marked by the increasing visibility of human burials in the archaeological record, typically associated with the Gravettian culture ($34 -24 ka cal BP) and occasionally even earlier ($39 -34 ka cal BP).Eastern Central Europe has a rich record of Palaeolithic archaeology with several sites containing human remains (Figure 1), which are associated with different technological assemblages.Directly dated human remains at Mlade c Cave (Czech Republic), 1 Oblazowa Cave (Poland), [2][3][4][5][6] Peștera Muierii 7 and Peștera Cioclovina 8 in Romania, and Troisie `me Caverne of Goyet (individuals Q116-1 and Q376-3) in Belgium 9,10 all fall within the $37 -33 ka cal BP range and are tentatively associated with Aurignacian artifacts.The Kostenki-Borshchevo complex in Russia spans different periods of the UP and has yielded numerous human remains, including the burial of Kostenki 14 ($39 -37 ka cal BP), which is also associated with the Aurignacian. 11,12Isolated human remains from Buran Kaya III in Crimea 13 are associated with Gravettian assemblages, but this is in contrast with their early direct age of $36.8-35.7 ka cal BP.In Central Europe, the Gravettian is split into the Early Gravettian ($34 -30 ka cal BP), Pavlovian ($31 -29 ka cal BP) and Late Gravettian/Willendorf Kostenkian ($29 -24 ka cal BP). 14,15The Pavlovian period is characterized by burials containing red ochre and grave goods such as seen at Dolnı ´V estonice-Pavlov [16][17][18][19] and P redmostı ´14,20 in Czech Republic.Whilst UP child burials have been identified in the region, including the burial of infants at Krems-Wachtberg (Austria), [21][22][23] they are extremely rare.Much more elaborate UP burials have been found at Sunghir (Russia), containing multiple individuals, abundant ochre, and an exceptionally rich assemblage of grave goods. 24,25Several attempts at directly dating these individuals produced vastly differing results, likely due to issues of contamination, thus contributing to uncertainty around the cultural association of the site.The most recent and rigorous compound specific dates place the burials at Sunghir between 35.8 and 32 ka cal BP. 11,[24][25][26][27][28][29][30] Borsuka Cave (50 9 0 53.94 00 N; 19 42 0 12.23 00 E), located in the Szklarka River valley in southern Poland (Figure 2A), adds important data to this discussion.Excavations next to the entrance of the cave in 2008-2010 (Figure 2B) uncovered six deciduous human teeth in association with 112 pierced teeth of Steppe wisent/aurochs (Bison priscus/Bos bison; n = 78) and European elk (Alces alces; n = 34), which have traces of ochre on ll OPEN ACCESS them. 31,32Seven layers were excavated and separated into a lower Pleistocene unit (Layers VII -V) and an upper Holocene unit (Layers IV -I).At the base of the sequence, overlying bedrock, Layer VII was archaeologically sterile.In Layer VI (from 150 to 250 cm depth), the human remains and pierced teeth were found spread over several m 2 in an NW linear distribution down the slope from the cave entrance.The pendants were concentrated by the southern wall of the trench at depths of 160-220 cm, with a few in the central and northern parts of the trench (Figure 2).The human teeth were located to the west of the concentration of pendants, in the northwest corner of the trench within squares C7, D7, and E7 (Figures 2C and 2D).These teeth were all identified as belonging to a child aged 12-18 months.Given the young age, it was not possible to determine the sex based on morphology, but the consistent phase of dentition indicated they likely belonged to one individual.Layer VI contained the human remains, perforated ungulate teeth, and one distal fragment of a flint blade but was otherwise archaeologically sterile.The overlying Layer V was also archaeologically and palaeontologically sterile but associated with Last Glacial Maximum (LGM) cooling.Layers IV-I contained Holocene assemblages spanning from the Mesolithic to modern (see 31,33 ).
In addition to the human remains, over 2,000 faunal skeletal fragments were excavated from Layer VI.The assemblage contained an unusual collection of fauna including cold steppe-tundra species, such as Rangifer tarandus (reindeer), Vulpes alopex (Arctic fox), Equus ferus (wild horse) and Coelodonta antiquitatis (woolly rhino), alongside taxa adapted to a warmer forested environment, including Alces alces (elk), Bos primigenius (aurochs), Martes martes (pine marten), Lynx lynx (Eurasian lynx), Castor fiber (beaver) and Meles meles (badger). 33The majority of the Pleistocene skeletal remains likely represent a natural accumulation, with signs of carnivore modifications on 1.7% of the mammalian and bird bones and no signs of anthropogenic modification aside from the pierced teeth.
Two of the herbivore tooth pendants were directly radiocarbon dated to 32,890 -30,760 cal BP at 95.4% probability (Poz-32394: 27,350 G 450 14 C BP; 14 C errors reported at 1s) and 29,900-29,120 cal BP (Poz-38236: 25,150 G 160 14 C BP).A reindeer metatarsus from the same layer was dated to 31,060 -30,270 cal BP (Poz-38237: 26,430 G 180 14 C BP). 32,33 While the dates indicated a Mid-Upper Palaeolithic age for the Layer VI assemblage, the lack of agreement between the dates from the two pendants at the 95% range suggested that the pendants are not strictly contemporaneous.This, along with the distribution of the herbivore and human teeth, leaves open the question of the association between the pendants and the human remains, necessitating further chronological investigation.However, due to their small size, direct dating of the human teeth was initially determined to be impossible.
Although alternative theories have been discussed, 32 the absence of associated ''domestic'' archaeology led to the interpretation that the assemblage represents an infant burial and grave goods that were disturbed post-deposition, resulting in the absence of a burial pit and their linear spread down the slope. 32The lack of any Palaeolithic archaeology necessitates relying on the dating and stylistic analysis of the pendants for determining the cultural association of the burial.2][23] However, the typology of the pendants themselves are similar to those found at Mlade c 32 and suggest a Late Aurignacian association.5][36][37][38][39] Given the small nature of fragmented deciduous human teeth and presence of Holocene archaeology in the upper layers, the possibility of this being the case at Borsuka Cave could not be ruled out.
Significant advances have been made in recent years for the radiocarbon dating and genetic analysis of ancient and highly degraded skeletal material, including reducing sample sizes for dating, 19,[40][41][42] combining multi-method extractions from a single sample, 43 non-destructive pre-screening of organic preservation, 44 treatments for removing contaminant DNA and/or enriching for endogenous ancient DNA (aDNA), [45][46][47][48][49] and non-destructive aDNA extraction from osseous artifacts. 50Such developments enable biomolecular analyses of material which were previously impossible.Considering the expected age and very limited amount of material available from the human and pierced herbivore teeth from Borsuka, we devised a multi-disciplinary study to investigate (i) if the human teeth originate from a single individual, (ii) the chronological relationship of the human teeth and pendants, and (iii) the association of the assemblage to different Palaeolithic cultures.In order to confirm the Palaeolithic origin of the teeth and ascertain if the human remains are contemporaneous with the pendants, we undertook radiocarbon dating of two deciduous human teeth (Figure 3) and six of the pierced herbivore teeth (Figure 4).From the human material, we selected two teeth (C7/675, C7/683) with partially intact roots to preserve the crown morphology and selected herbivore pendants from different squares to test across the scattered assemblage.Ancient DNA analysis was attempted on the same two human teeth to determine if they belonged to one individual and to see how this individual related to other Upper Palaeolithic individuals.The assemblage at Borsuka Cave represents a challenge in terms of the extremely limited material available, and the inherent issues of poor preservation and contamination of ancient samples for both 14 C and aDNA.Nevertheless, we aimed to maximise data retrieval from extremely small samples to provide additional information about the Borsuka assemblage within the broader context of the UP.

Radiocarbon dating
The collagen yield of the two human teeth (4.3-8.5%) was within the range of the herbivore teeth (3.0-9.1%),consistent with degraded Palaeolithic material but all falling well above the $1% minimum required for dating.The elemental values (C%, N%, C/N) of all extracts fall within the commonly accepted ranges of well-preserved collagen (C%: $30-45%; N%: $11-16%; C/N: 2.9-3.6),indicating that the collagen extracts were suitable for radiocarbon dating 51 (Table 1).However, several fall at the limit (3.5-3.6) of the accepted C/N range, indicating low levels of contamination may be present in the samples. 52,53For C7/675, C7/658, C6/494, sufficient collagen was available for additional quality analysis using Fourier Transform Infrared Spectroscopy (FTIR).All extracts had FTIR spectra characteristic of ancient bone collagen, [54][55][56][57] but the higher intensity of the peak at $2900 cm À1 in the spectra of the human tooth C7/675 may be indicative of external contamination (Figure S1).
The bulk collagen d 13 C and d 15 N values of the two human teeth fall within the range of stable isotopic values seen for other Mid-Upper Palaeolithic humans in Eurasia, 19,23,[58][59][60][61][62] with very high d 15 N values that are well above the herbivorous signature of the Bos/Bison and elk teeth (Figure S2).Although the d 13 C values of the two human teeth agree within instrumental error, the d 15 N value of C7/675 is higher than C7/683 by 1.3&.Given that the teeth are both deciduous molars with similar formation times, the difference is more likely to be indicative of contamination, or of separate individuals, rather than palaeodietary.The elevated d 15 N values are at the highest end of the range seen in other Mid-UP humans and in the range of two trophic levels (trophic level increase: $3&-5&) above the herbivores.This is consistent with a breastfeeding signal given the age of the infant, but without comparative adult data and further chemical analyses it is not possible to elucidate this further.
Given the small size of the dentine samples (23.4-67.4mg) and extracted collagen (1-4.5 mg), we followed a previously established approach for small samples. 41,42Multiple 14 C dates were obtained from each collagen extract using both graphitization and CO 2 gas ion source (GIS) methods, and subsequently combined.X 2 tests were carried out to determine the statistical agreement between replicate dates from the same extract.All of the 14 C dates obtained fall within the Mid-UP range, from 31,070 G 770 14 C BP to 24,830 G 290 14 C BP, or $35,290 to 29,090 cal BP (Figure 5; Table 1; Tables S1 and S2).For the herbivore pendants, all six pendants were dated to 35,290 -33,260 cal BP (95.4% probability), much older than dates obtained previously on two of the pendants, overlapping with the Late Aurignacian/Early Gravettian periods.For the two human teeth, replicate measurements agree statistically for each collagen extract but the dates from C7/683 (Aix-12058 combined age: 26,610 G 240 14 C BP; 31,160 -30,320 cal BP at 95.4% probability) are older than C7/675 (Aix-12047 combined age: 25,100 G 140 14 C BP; 29,860 -29,090 cal BP at 95.4% probability) by $1500 14 C years (Figure 5; Table S1).When the dates from Layer VI obtained thus far are modeled together in one phase in OxCal, the assemblage spans 6,290-4,770 years, between 36,650 and 27,620 cal BP (Figure 5; Table S3).No dates are available from the overlying and underlying layers to constrain the modeled range, due to a lack of material suitable for dating so none of the dates were identified as outliers.The radiocarbon data is included in Table S1, the background collagen measurements in Table S2 and the modeled ranges of the Borsuka data in Table S3.

Ancient DNA analysis
Two microsamples between 4.7 and 12.9 mg of dentine powder were drilled from each of the two human teeth (C7/675 and C7/683).The resulting four tooth powder sub-samples underwent DNA extraction, single-stranded library preparation and shotgun sequencing (see method details).Between 1,628,190 and 8,575,605 sequences were generated per library, of which between 2,495 and 37,436 mapped to the human reference genome (hg19 63 ) for each library.While no ancient DNA was detected in C7/683, one of the sub-samples from C7/ 675 was identified as containing ancient DNA based on the observation of between 12.4% and 20.6% C-to-T substitutions on the terminal ends of recovered DNA (Table S4), which is characteristic of ancient DNA. 64,65The duplication rates, or number of times each DNA fragment was sequenced, for these libraries were all around 1 (Table S4), indicating that not all recovered DNA fragments were sequenced.Modern human DNA contamination for libraries produced from this subsample was estimated at 54-56% (Table S4) using the method AuthentiCT 66 which calculates levels of contamination based on deamination patterns.
Despite not detecting ancient DNA in C7/683 to the limits of our resolution, libraries from both teeth were subsequently enriched for human mitochondrial (mt) DNA via hybridization capture, 67 in an attempt to determine if both likely stem from the same individual (Table S5).Libraries from C7/683 were still included in hybridization captures as the shotgun sequencing was not exhaustive, and the content of nuclear and mitochondrial DNA has been shown to defer in ancient specimens, 68 leaving the possibility that small amounts of endogenous DNA remained undetected.The enriched libraries contained between 243,078 and 938,984 sequences longer than 34 base pairs that mapped to the human mtDNA revised Cambridge Reference Sequence 69 with elevated C-to-T substitutions (21.3-35.1% on the 5 0 end and 14.5 to 24.6% on the 3 0 end).As modern human contamination was detected in each library, ranging from 2.93% to 39.94%, all subsequent analyses were restricted to putatively deaminated fragments that contained a C-to-T substitution within the first three and/or last three terminal bases.Between 15,269 and 60,728 deaminated fragments (corresponding to mtDNA coverages of 46-and 223-fold, respectively) were used to reconstruct near-complete consensus mtDNA genomes.The resulting mtDNA haplotypes from each tooth were the same, consistent with the teeth either being from the same individuals or the same maternal lineage.However, as neither mtDNA genome was complete, we cannot exclude that individually discriminating variants were not detected.The more complete mtDNA genome from C7/675 was then used for tree building and molecular branch shortening (Tables S6 and S7).The resulting molecular date was estimated at 33,533 years BP (95% highest posterior   S1. density interval: 28,200 -38,935 years BP) and the mtDNA genome falls within haplogroup U6 (Figure 6A).Haplogroup U6 is most commonly observed in Northern Africa in present-day humans, 70,71 and has previously only been observed in Palaeolithic Europe in the specimens from the site of Pes xtera Muierii in Romania ($34,000 years old). 9,72,73Similar to the Muierii 1 and 2 mtDNA genomes, which were subsequently determined to belong to the same individual, 73 the Borsuka mtDNA genome is basal to the U6 haplogroup, sharing the 3348G, 10517A, and 16172C positions.Importantly, among the overlapping positions (6 missing positions) no differences are observed between the Muierii 2 and Borsuka mtDNA genomes.Even if both Borsuka teeth and Muierii 2 share an mtDNA genome, this only indicates that they likely lived within $2,500 years of each other (Figure S3, method details).In order to determine the genetic affinity of the child buried in Borsuka Cave to other individuals from the Upper Palaeolithic, we enriched the libraries from C7/675 for $1.2 million single nucleotide polymorphisms (SNPs) in the human genome known to be informative for studying human population history. 74,75While the resulting libraries covered 513,237 SNPs, they were also identified as containing approximately 56% modern human contamination (Table S8).This required all downstream analysis to be restricted to the 46,286 putatively deaminated fragments, which were used to determine the sex of the child.As the ratio of autosome, X, Y chromosome SNPs in the 1240k does not follow conventional expectations due to the respective target size, we used an adjustment of the X and Y-rate as has been performed previously with this type of data. 9This resulted in an ''X-rate'' of 0.68 and ''Y-rate'' of 0.04, consistent with the individual being female (Table 2).When projected onto a principal component analysis (PCA) of modern-day individuals from West Eurasia, Central Asia and East Asia, the Borsuka individual clustered with previously published Upper Palaeolithic Western Eurasians and near modern day Western Eurasians (Figure 6B).The genetic sharing between the Borsuka individual and previously published modern and ancient individuals was calculated with f3-statistics, using present-day Mbuti individuals as an outgroup population (Figure 6A).Consistent with the PCA analysis, the greatest affinity among modern day populations was to Western Eurasians.Basal Eurasian ancestry was not detected in the Borsuka individual (Tables S9 and S10, method details).Among ancient individuals, the Borsuka individual shared the most alleles with the $35 ka cal BP Bacho Kiro Cave individual (BK1653), $34 ka cal BP Muierii 1, $31 ka cal BP V estonice 16 and $34 ka cal BP Sunghir 3 individuals when tested with f3-statistics (Table S11).When directly comparing the genetic affinity of the Borsuka individual with other ancient individuals with D-statistics, the Borsuka female has a greater affinity to the Gravettian and Aurignacian individuals (Figures S4 and S5; Table S12).However, no significant difference was observed in the affinity of the Borsuka child to the V estonice vs. Sunghir vs. BK1653 vs. Muierii 1 individuals, precluding the direct association of the Borsuka girl with one of these groups.

DISCUSSION
The chronological and genetic data presented here confirms the Upper Palaeolithic origin of the infant remains from Borsuka Cave, making these remains currently the oldest female infant burial identified to date.While it is not possible to confirm that the teeth come from the same individual without nuclear DNA, the mitochondrial analysis is consistent with the two teeth belonging to the same individual or the same maternal lineage.
Radiocarbon dating is based on the exponential decay of 14 C over time.As the concentration of 14 C in modern carbon is therefore much larger than in ancient samples, any contamination with modern carbon will make 14 C dates younger than the true age of the sample, with the effects getting progressively worse for more ancient material.Given the ubiquity of modern carbon introduced via the burial environment, handling, storage and analysis of artifacts, when inconsistent dating results are obtained from Palaeolithic material, older 14 C ages are generally considered more accurate. 29,76The direct date from human tooth C7/683 is older than the date from C7/675.If the two teeth originate from the same individual, this implies that either 1) the older age of C7/683 is correct and the date from C7/675 is an under-estimation of the true age due to the presence of external carbon contamination, or 2) both dates are under-estimations.S3.Alternatively, the teeth could originate from two infants and both dates could be correct.In addition to the discrepancy in age between the direct human dates and the range of ages from the pendants, this may also be indicated by the difference in d 15 N values between the teeth and the scatter of the assemblage over several square meters.Given the consistent form of pendant manufacture and that the mtDNA analysis indicates the teeth are at least from maternal relatives (if not from the same individual), a group from the same maternal lineage would need to have used the site for 4,000-6,000 years to account for the dating results in this scenario.The elevated C/N value and additional FTIR peak at $2900 cm À1 in the C7/675 extract do suggest contamination may have affected the date, and potentially the d 15 N value.Thus, the most parsimonious interpretation of the results is that the two teeth originate from the same individual and are affected by contamination.If we consider the teeth as originating from one individual, the older date from C7/683 is more likely to be accurate than the younger date from C7/675 as contamination typically decreases the age.Given the small amount of material available for sampling and the C/N values from both teeth of 3.5-3.6,we consider the $31.2-30.3ka cal BP date from C7/683 to be a minimum age for the human remains.
The older date from C7/683 is contemporaneous with the dates previously obtained from two Borsuka pendants and the $31 ka cal BP Pavlovian burials at Dolnı ´V estonice-Pavlov and Krems-Wachtberg, which would support the association of Borsuka with the Pavlovian culture in Central Europe (Figure 7).However, the new 14 C dates on the herbivore tooth pendants range in age from 35,290 to 33,260 cal BP (95.4% probability).These new dates for the pendants are older than the ages obtained from the human infant teeth, but also much older than the dates previously obtained on two pendants and a reindeer bone from the same layer (Figure 5).
It is possible that the burial assemblage dates to $31 -30 ka cal BP, and the necklace was constructed from a collection of herbivore teeth of varying age, including some that were pierced and used for the construction of an ornament long after the death of the animal.The solitary lifestyle of European elk (Alces alces) implies that the elk teeth were unlikely to have been collected during one event.A recent study of the Sr isotopic compositions and trace element analysis of the enamel from four of the elk tooth pendants (at least 3 individuals) demonstrated a non-local origin of the teeth, indicating that the pendants were transported $250 km to Borsuka Cave from an area near the Austria/Slovakia border or northern Hungary, on the southern side of the Western Carpathians. 779][80][81] The presence of Trans-Carpathian raw stone materials in Central Europe also shows high human mobility around the Western Carpathian mountain region. 80,82This evidence suggests that the collection of the herbivore teeth over a period of time is feasible, however, it seems unlikely that the pendants represent animals varying in age by over 5,000 years.Alternatively, low levels of contamination may be affecting some of the 14 C results with the $32 -29 ka cal BP ages under-estimating the true age of the pendants and human teeth.
The older $35.3-33.3ka cal BP range of dates from the pendants overlap in time with the $36.8-34.7 ka cal BP range of directly dated human remains from Mlade c Cave (Czech Republic) 1 (Figure 7).It has been suggested that the relatively shallow human remains at Mlade c could be the result of intentional deposition, 35,83,84 but the lack of contextual information means this hypothesis cannot be tested.Despite their uncertain context, based on the dates, the Mlade c humans have been associated with Aurignacian artifacts found elsewhere in the cave, including Mlade c-type bone points and pendants made from the teeth of large ungulates (e.g., elk, horse, bison/aurochs). 856][87] The entire roots of the perforated teeth were heavily scraped on both surfaces, remarkably reducing the root thickness and the perforations were made very close to the root apex by drilling from both sides.The only difference between pendants discovered at Borsuka and Mlade c can be observed in the larger range of species from which teeth were used at Mlade c. Thereby the assemblage from Borsuka cave represents the largest collection of pendants made of large ungulate teeth potentially associated with the Aurignacian culture.Currently, no aDNA is available from the Mlade c individuals to investigate this link further.
While the origin of the Gravettian has been widely debated, 14 the relatively large collection of burials across Eurasia has provided a significant body of morphological and behavioral data on Gravettian humans across a wide geographical and chronological range.Recent aDNA studies 10,88 have demonstrated a genetic distinction between the Gravettian cultures in Central Eastern Europe and Southwestern Europe.The oldest humans discovered to date with genetic continuity to Gravettians and modern West Eurasians are represented by three different genetic ancestry components from the $35,000-year-old GoyetQ116-1 individual, 9 the $35,000-year-old individual from Bacho Kiro Cave (BK1653) 89 and the $38,000-year-old Kostenki 14 9, . 90The BK1653 individual is related, but not identical, to the genetic ancestry of GoyetQ116-1 which contributes to the genetic ancestry of the Gravettian individuals in Southwestern Europe, while the former contributes ancestry to the Central European Gravettian individuals.Genetic ancestry from Kostenki 14 has been found in all individuals associated with the Gravettian culture.We aimed to clarify the genetic relationship of the Borsuka individual to individuals of different Palaeolithic cultures, however these associations were limited due to the low amounts of data recovered.The Borsuka individual shares a basal U6 mtDNA haplotype with Pes xtera Muierii 2, but there is not sufficient nuclear data to determine the full genetic relationship between these individuals in context with other ancient humans.The available data show that the Borsuka individual has the most nuclear genetic affinity to the Bacho Kiro 1653 individual (presumed to be associated with the Aurignacian), as well as individuals from Dolnı ´V estonice (Gravettian) and Sunghir (early UP).Interestingly, the Borsuka individual is significantly closer to the Bacho Kiro individual than they are to Kostenki 14.This is consistent with Borsuka being an UP Central European, but does not determine a closer affinity to Gravettian or Aurignacian individuals within Central Europe.The 14 C dates provide a minimum age of 31.3-30.3ka cal BP for the human remains, which agrees with the molecular date estimation of 33.5 ka that falls roughly at the border between the Late Aurignacian and Early Gravettian periods.
The Borsuka Cave assemblage represents an important data point in the catalog of Palaeolithic human remains for investigating the geographical and chronological range of UP burial practices.Of the very limited number of Upper Palaeolithic infant burials discovered and successfully analyzed with aDNA, this is the first female identified.The previously oldest confirmed female infant burial was found in Mesolithic Italy, 91 at least 20,000 years after the burial of the Borsuka girl.This study therefore provides the first evidence that child burials incorporating grave goods were not limited to males during the UP.This study further highlights the challenges of working with limited and poorly preserved ancient materials, and emphasizes that employing an interdisciplinary, multi-method approach can still provide valuable insights to allow in-depth investigations of Palaeolithic contexts.

Limitations of the study
Palaeolithic human remains are rare, and the remains of UP infants even more so.The small size and partial fragmentation of the infant teeth presented an exceptionally limited amount of suitable material for analysis.We aimed to take as little material as possible and therefore limited destructive sampling to two of the six teeth to preserve the morphology of the crowns and ensure their preservation for future studies.Modern carbon contamination in a 14 C sample produces an under-estimation of the true age of the material.As sample size decreases, the risk of contamination increases, which is clearly a consideration for the data presented.This is highlighted by the lack of agreement in dates from the two human teeth from the (potentially) same individual and their slightly elevated C/N values, indicating that low levels of C contamination may be present either due to incomplete removal of environmental contaminants during pretreatment or introduced in the laboratory.Given the ubiquity of carbon contaminants in the environment and laboratory, older ages produced from the same sample material are generally considered to be more accurate and reliable.Rigorous compound specific dating approaches can be employed to isolate endogenous hydroxyproline in bone collagen to overcome contamination issues. 27,92,93However, the small proportion of hydroxyproline in collagen means this approach requires large starting sample sizes, thus making the approach not feasible for this study.As contamination of the human dates is indicated we therefore consider them to be minimum ages.The molecular dating technique (calibrated using other directly radiocarbon dated human remains) indicates an age of $33.5 ka BP (albeit with a very large confidence interval), which supports the $31 ka cal BP being a minimum age.The new data presented nevertheless confirm the UP origin of the human remains.Given the challenges and the limited data available, we cannot conclude a definitive age for the burial assemblage or a more definitive association with a specific Palaeolithic culture.However, even using very small sample sizes for dating (in the order of aDNA sample sizes), we confirm the placement of the assemblage around the border of the Late Aurignacian/Early Gravettian, supported by the aDNA analysis and the typology of the pendants.
The high levels of modern human contamination and limited amount of endogenous DNA present in both infant teeth limited the power of determining different genetic affinities beyond broad population genetic relationships.This resulted in reduced coverage of both the mtDNA and nuclear genomes.As shown in Figures S6-S9, decreased amounts of data result in missing some genetic affinities.It is likely that with more data it would have been possible to deconvolute the association of the Borsuka girl to Bacho Kiro 1653, or the V estonice and Sunghir clusters, but unfortunately the limited data prevented this.
The limitations arising due to contamination emphasize the importance of taking efforts to minimize the contamination of archaeological materials.This includes wearing gloves and face masks during excavation and handling of human remains and artifacts, appropriate storage and consideration of sampling prior to consolidants or glues being applied.

STAR+METHODS
5][56] Roughly 0.3 mg collagen was homogenised and mixed with $40 mg of IR grade KBr powder in an agate mortar and pestle.The powder was pressed into a pellet using a manual hydraulic press (Wasserman) and analysed with an Agilent Technologies Cary FTIR with a DTGS detector.Spectra were recorded in transmission mode at 4 cm À1 resolution with averaging of 34 scans between 4000 and 400 cm À1 using Resolution Pro software (Agilent Technologies) and were compared to library spectra of well preserved collagen.

AMS dating
The extracts were dated on the AixMICADAS 108 installed at CEREGE (Centre de Recherche et d'Enseignement de Geosciences de l'Environnement) in Aix-en-Provence, France.Each collagen extract was split and measured in several aliquots as per the protocol for small collagen extracts described in Fewlass et al. 41,42 (Table S1).Where sufficient material was available, collagen (combustion weight $700 ug C) was weighed into aluminium cups and graphitized using the AGE 3 109 (Automated Graphitisation Equipment) prior to dating on the AixMICADAS.Oxalic acid II standards and background collagen samples (Table S2) were measured in the same session and used in the age calculation of the archaeological samples in BATS. 105A relative error of 30% was applied to the blank value and an additional uncertainty of 1.6& was propagated in the F 14 C error calculation as per standard practice.
Gas measurements were performed using the protocol described in Tuna et al. 110 and Fewlass et al. 41 Small aliquots of collagen (combustion weight $60 ug C) were introduced into cleaned silver cups (800 C, 2 h) and combusted in an Elementar Vario MICRO cube EA (Elementar Analysensysteme GmbH, Germany) which was directly coupled to the gas ion source (GIS) of the AixMICADAS. 111,112Oxalic acid II NIST standards (from a gas canister) were measured to normalize and correct samples for fractionation.The age of each sample was corrected with background measurements from the same batch (Table S2).A relative error of 30% was applied to the blank value and an additional uncertainty of 4& was propagated in the F 14 C error.

Ancient DNA analysis Data generation and shotgun sequencing
Sampling of the teeth for radiocarbon dating and genetic analysis was performed in a dedicated clean room at the Max Planck Institute for Evolutionary Anthropology.A microsampling approach was used for sample collection from the two of the infant teeth (C7/675 and C7/683) for downstream genetic analyses.Between 7.2 and 12.9 mg of dentine was collected from the two roots of C7/675 and two microsamples of 4.7 and 5.2 mg of dentine were drilled from the crown of C7/683 for a total of four samples.To extract DNA from the tooth powder, 500 uL of lysis buffer (for 5 mL lysis buffer: 4.5 mL 0.5M EDTA pH 8.0, 2.5 uL of Tween 20,125 uL of 10mg/mL proteinase K, 373 uL water) was added to each tooth powder aliquot for overnight lysis at 37 C.The resulting lysate was purified following the automated extraction protocol described, 113 using binder buffer 'D'.Each DNA extract was then converted into dual-indexed single-stranded libraries 106 and sequenced on Illumina MiSeq or HiSeq2500 platforms in pools of 63-96 libraries that included libraries from other projects not discussed in this paper.Negative controls were included starting at both the DNA extraction and library preparation steps.Base calling was performed using Bustard (Illumina) and leeHom 104 was used to trim adapters and merge overlapping forward and reverse sequences.The resulting sequences were mapped to the human reference genome (hg19 63 ) using BWA version 0.5 (https://github.com/mpieva/network-aware-bwa)with ancient DNA parameters (-n 0.001 -o 2 -l 16500).Reads that were not a perfect match to expected index combinations, shorter than 35 base pairs, or had a mapping quality less than 25 were removed using SAMtools version 1.3.1 104 .Bam-rmdup version 0.6.1 (https://github.com/mpieva/biohazard-tools) was used to remove PCR duplicates.Each library (including controls) was evaluated for the presence of ancient human DNA based on the presence of elevated C-to-T substitutions on the terminal ends of DNA fragments (Table S4).Potential cross-contamination due to index swapping between libraries was evaluated using a previously published method 67 that evaluates all sequenced index pairs in a sequencing run to estimate the number of reads assigned to a library that may originate from another library.All libraries were estimated to have less than 6 cross-contaminated reads.The negative controls were also examined for evidence of ancient DNA based on elevated C-to-T substitutions and no evidence of aDNA was identified in these libraries (Table S4).

Radiocarbon data analysis and reporting
The radiocarbon dates were calibrated and analysed using OxCal 4.4 114 and the IntCal20 calibration curve. 115The dates were input into a one phase General t-type Outlier Model with a 5% prior likelihood of being an outlier, except C7/675 which was given an outlier likelihood of 100% due to the C/N value and FTIR analysis 116 (Table S3).As per convention (e.g.Stuiver & Polach, 94 Millard 95 ), throughout the text uncalibrated 14 C ages are reported with the abbreviation 'BP', meaning 'radiocarbon years before AD 1950' and are reported with 1s errors, whereas exception.The V estonice complex includes multiple individuals with varying qualities of genetic data.Four of these individuals were included for analysis to serve as a proxy for Borsuka for the input of limited data on each analysis.
Basal Eurasian ancestry evaluation D-statistics was used to explore how the Borsuka individual's genetic affinities compared to other previously published ancient humans.It has previously been shown that the D statistic is $ 0 for both D(Kostenki14, X; Han/Dai, Mbuti) and D(W, Stuttgart; Han, Mbuti), when W/X is a European from $39,000 to 14,000 years ago without Basal Eurasian ancestry [9] .Calculating these D-statistics with Borsuka in the X or W position is consistent with an individual that also does not contain basal Eurasian ancestry (Tables S9 and S10).

Figure 2 .
Figure 2. Visuals of Borsuka Cave location, site plans, and stratigraphic profile (A) Map showing the location of Borsuka Cave within Europe.(B) Plan view of Borsuka Cave showing the 2008-2010 excavation trench (gray shaded rectangle).(C) Stratigraphic profile with the location of ungulate tooth pendants (black circles) and human remains (red circles).(D) Plan view of Layer VI showing the location of human remains and pendants.Note that three human teeth found in situ are plotted but three were found during wet-sieving so lack exact 3D coordinates.In C and D, each square is 1 m 2 .

Figure 5 .
Figure 5. Single phase outlier model showing the posterior probability distributions of the 14 C dates from Borsuka Cave Layer VI generated in this study (Aix-) and previously (Poz-) The bracket beneath each distribution shows the 95% probability range.The outlier probability (O) is given in the format ''[O: posterior outlier probability/prior outlier probability].''The ranges are given in TableS3.

Figure 6 .
Figure 6.The genetic analysis of the Borsuka C7/675 individual in relation to other modern and ancient humans (A) The calculated allele sharing between the Borsuka individual and a selection of modern and ancient individuals (X) where warmer colors represent higher amounts of shared alleles.The f3 statistic f3(X, Borsuka; Mbuti) was calculated using 28,213 -31,393 overlapping SNPs for modern populations and 6,697 -28,642 SNPs for ancient individuals.Ancient individual key: 1-Goyet Q116, 2-EL Miron, 3-Dolnı ´V estonice16, 4-Bichon, 5-Oase1, 6-Bichon, 7-Mal'ta1, 8-Karelia, 9-Yana2, 10-Tianyuan, 11-Ust-Ishim, 12-Kostenki14, 13-Sunghir3.(B) The projection of 20 ancient individuals onto a principal component analysis of 1,267 modern individuals.Ancient individuals are colored based on their association with the Aurignacian (dark blue), Gravettian (orange), or IUP (light blue) archaeological assemblages or in green if their technological association is uncertain.(C) A subset of the phylogenetic tree generated with BEAST2.Ancient humans are colored in blue (Borsuka in red).Nodes are labeled with their posterior probability.The x axis indicates calibrated years before present.

Figure 7 .
Figure 7. Calibrated radiocarbon dates (95.4% range) from Borsuka Cave infant teeth and pendants compared to UP human burials and human remains in Central and Eastern Eurasia mentioned in the text Human remains tentatively associated with Aurignacian assemblages are shown in blue and Gravettian assemblages in orange while uncertain cultural associations are shown in gray.The Borsuka infant dates are conservatively considered minimum ages.

Table 1 .
Pretreatment and radiocarbon dating of humans and herbivore pendants from Borsuka Cave Sample ID Material Sampled (mg) Collagen (mg) % Collagen d 13 C (&) d 15 N (&) %C %N C/N AMS lab number combined BP combined err 95.4% cal BP n Combined dates of multiple replicates (n) are reported here with individual replicate dates and X 2 values reported in Table

Table 2 .
Sex determination based on coverage of nuclear SNPs in the '1240k 0 array using putatively deaminated reads