Persistence of arctic-alpine flora during 24,000 years of environmental change in the Polar Urals

Plants adapted to extreme conditions can be at high risk from climate change; arctic-alpine plants, in particular, could “run out of space” as they are out-competed by expansion of woody vegetation. Mountain regions could potentially provide safe sites for arctic-alpine plants in a warmer climate, but empirical evidence is fragmentary. Here we present a 24,000-year record of species persistence based on sedimentary ancient DNA (sedaDNA) from Lake Bolshoye Shchuchye (Polar Urals). We provide robust evidence of long-term persistence of arctic-alpine plants through large-magnitude climate changes but document a decline in their diversity during a past expansion of woody vegetation. Nevertheless, most of the plants that were present during the last glacial interval, including all of the arctic-alpines, are still found in the region today. This underlines the conservation significance of mountain landscapes via their provision of a range of habitats that confer resilience to climate change, particularly for arctic-alpine taxa.


Results
Plant sedaDNA record. We obtained around 75 million paired-end raw DNA sequences for the 153 sediment samples from Lake Bolshoye Shchuchye (Supplementary Table S1). After filtering out sequencing artefacts and sequences with <98% match to the DNA reference library (see Methods), we retained 19 million reads, corresponding to 134 vascular plant and 28 bryophyte taxa. Of these, 40% are identified to species level, 45% to genus and 15% to a higher taxonomical level (Supplementary Table S2). Where possible, genera that were not identified to species level were assigned to a likely inferred species based on their biogeographic distribution (Supplementary Table S2).
A sustained, long-term increase in floristic diversity is observed in the sedaDNA record over time (Fig. 2, Supplementary Figs. S1-S3). Of the total 162 plant taxa detected using sedaDNA, 70% (114 taxa) were detected within the full-glacial period (ca. 24,000-15,000 cal. years BP), 75% (85 taxa) of which persist into the Holocene period (ca. 11,700-1,300 cal. years BP) along with the addition of 47 new plant taxa not present in the full-glacial period. The majority (87%) of the vascular plant taxa detected within the full-glacial period and 89% of all the vascular plant taxa detected by sedaDNA are still found in the vegetation today (Fig. 2, Supplementary Table S2). Thus, much of the present-day flora at Lake Bolshoye Shchuchye was already in place in the full-glacial period; subsequently, there has been underlying continuity in taxonomic composition but also a gradual addition of new plant taxa over time to form the present-day vegetation mosaic.
Saliceae, a tribe of the willow (Salicaceae) family, is common throughout the record, occurring in nearly all PCR repeats of all samples. It most likely represents dwarf shrubs (e.g. Salix polaris, S. reptans, S. nummularia) in the early full-glacial, with the addition of shrub forms (e.g. Salix lanata, S. glauca, S. phylicifolia) in the late-glacial period and potentially tree forms in the Holocene (e.g. Salix caprea, S. cinerea). In full-glacial and early late-glacial time (24,000-15,000 cal. years BP), the sedaDNA assemblage is characterized by herb-tundra vegetation with a rich diversity of forbs, such as Papaver, Draba, Bistorta vivipara and Saxifraga oppositifolia, and graminoids, such as Puccinellia, Festuca and Juncus biglumis (Fig. 3). The mat-forming dwarf shrub Dryas becomes more common from 15,000 cal. years BP, followed by a sequential arrival of additional dwarf shrubs and tall shrubs/deciduous trees (e.g. Betula, Empetrum, Vaccinium sp.).
The coniferous trees Picea sp. and Larix sibirica became common elements of the vegetation ca. 9,000-4,000 cal. years BP, alongside many boreal herbs (Fig. 3, Supplementary Figs. S1-S3). Coniferous forest withdrew around 4,000 cal. years BP; the vegetation subsequently reverted to dwarf-shrub tundra with a diverse herb flora, similar to the vegetation mosaic seen in the late-glacial and early Holocene interval, as well as the present-day vegetation mosaic of the Polar Urals.
Species persistence and floristic richness. We assigned a classification, where possible, to vascular plant taxa detected by sedaDNA based on their present-day native distribution (see Methods). In total, 31 taxa were   www.nature.com/scientificreports www.nature.com/scientificreports/ classified as "arctic-alpine", 49 as "boreal" and 22 as occupying a dual "arctic-boreal" distribution (full details of the classifications are presented in Supplementary Table S2). A steady increase in the diversity of arctic-alpine plants is observed throughout the full-glacial and late-glacial period, peaking around 12,700 cal. years BP before a distinct decline to low values (Fig. 4a). The arctic-boreal taxa, which include widespread genera such as Empetrum and Vaccinium, show low diversity in the full-glacial period but then increase, with fluctuations similar to the arctic-alpine pattern from around 17,000 cal. years BP. The initial decline in arctic-alpine plants occurs slightly before the main increase in the diversity of boreal taxa, at the time when floristic richness of arctic-boreal taxa is high and Salix is abundant. A further decline is seen ca. 10,000 cal. years BP, when the boreal taxa expand and maintain a high diversity until around 4,500 cal. years BP, with four prominent peaks visible in the Holocene period (Fig. 4a). Overall, the proportion of DNA reads of the three distribution categories (Fig. 4b) show similar trends of floristic diversity, with the exception that arctic-alpines reached their maximum proportion (>98% of total DNA reads) around 20,000 cal. years BP, considerably earlier than their maximum diversity was reached.
We investigated the response of individual arctic-alpine plant taxa to the Holocene establishment of dwarf shrub and tall shrub/ trees (e.g. Dryas octopetala, Vaccinium sp., Empetrum nigrum, Betula, Alnus) and later forest trees (e.g. Larix sibirica, Picea sp.) in the lake's vicinity between ca. 15,000-4,000 cal. years BP (Fig. 2). In total, 32 arctic-alpine plant taxa were detected by sedaDNA, of which 31 are detected within the full-glacial interval; Arabis alpina was recorded during the Holocene interval only (see Supplementary Table S2). Of the 31 arctic-alpines that sedaDNA reads in a single PCR replicate are also indicated. Plant taxa inferred to species level based on their biogeographic distribution are marked with an asterisk (*). The x-axis of the incidence matrix refers to the sedaDNA sample number, with a decreasing age (towards the present-day) from left to right. The y-axis presents each plant taxon detected using sedaDNA sorted according to their median distribution within the sedaDNA samples. Plant taxa that are still present within the vegetation of the Polar Urals today are marked with a grey square on the far right-hand side of the incidence matrix; bryophytes are excluded as the present-day biogeographic distribution of many of the identified taxa are poorly resolved. Green shaded boxes indicate the timing of past shrub and forest tree establishment in the vicinity of the lake. Botanical drawings of key plant taxa were created using Adobe Illustrator CC 2018 (https://www.adobe.com/uk/products/illustrator.html#).
are detected within the full-glacial interval, 27 (28 including occurrences with <10 DNA reads) are recorded after dwarf shrubs and tall shrubs/trees began to establish from 15,000 cal. years BP and 15 (21) are recorded in the later period when forest tree taxa established between ca. 9,000-4,000 cal. years BP. Of the 17 arctic-alpines that . Turnover in plant functional groups over the past 24,000 years. Selected functional groups are presented as a percentage of total sedaDNA reads per sample (histogram) and maximum number of PCR replicates (out of eight) per sample (diamond symbols) for the Lake Bolshoye Shchuchye record. Note that the height of the y-axis varies amongst panels. The grey dashed line indicates the Pleistocene-Holocene boundary. Botanical drawings of key plant taxa were created using Adobe Illustrator CC 2018 (https://www.adobe.com/uk/ products/illustrator.html#).
are not detected in the sedaDNA during the period of forest tree establishment, all disappear prior to the establishment of forest tree taxa during the preceding shrub period (Fig. 2). Nevertheless, three (four) of these taxa reappear when coniferous forest withdrew in the more recent period (ca. 4,000-1,300 cal. years BP), and all of the 32 arctic-alpine plant taxa detected in the sedaDNA record are found in the local vegetation today.

Discussion
Several studies have emphasised the advantage of sedaDNA for higher taxonomic resolution and richness compared to traditional methods of pollen and plant macrofossils 30,35,47 . The sedaDNA record from Lake Bolshoye Shchuchye is taxonomically diverse, even compared with most other sedaDNA studies 39,48 ; this may be explained by its long temporal record, optimised methods and use of a more complete local reference library. Moreover, the sediment lithology is predominantly fine-grained clays and silts; these tend to be highly suitable for the preservation of DNA since extracellular DNA can bind to the relatively large and charged surface areas of clay colloids [49][50][51] . The lake has a large (215 km 2 ) and topographically complex hydrological catchment with steep slopes, high rates of erosion and runoff and considerable riverine input to the lake 42 , all of which probably contribute to the floristically rich sedaDNA record obtained. The sedaDNA signal in Lake Bolshoye Shchuchye likely reflects sediment inputs from large source areas that capture a range of plant communities occupying the lake's hydrologic catchment. Since sedaDNA was able to identify many taxa that are often poorly represented in traditional palaeorecords, the Lake Bolshoye Shchuchye record provides a unique insight into the response of the artic-alpine flora to environmental change, including the millennial-scale, phased expansion of woody growth forms in the late-glacial and Holocene.
There was both continuity and change in plant community composition at Lake Bolshoye Shchuchye over the past 24,000 years; the lake's catchment supported a growing set of plant communities, and species diversity increased over millennia with few taxon losses to the present day. It is possible that the observed pattern of increasing species diversity could be explained by better DNA preservation and thus detection of sedaDNA within the uppermost sediments compared to older sediments. Data patterns suggest this is not the case, however. A sustained increase in species diversity is observed until the early to middle Holocene, after which species diversity appears to stabilise (with the exception of four anomalous peaks) towards the most recent samples, with little change that could be attributed to better DNA preservation (Supplementary Fig. S5).
More than two thirds of all recorded plant taxa identified were already in place in the full-glacial period (ca. 24,000-15,000 cal. years), when the study region remained free of extensive ice cover 42,52,53 , and the data provide a rich floristic record of full-glacial vegetation close to local glaciers but outside of the Eurasian Ice Sheet margin, which was situated further to the north 53 . Subsequently, new taxa appeared over time, and there was a shift in the dominance of plant functional groups during a long period of changing and (predominantly) warming climate. While the timing and nature of the general vegetation changes observed since the last glacial period at Lake Bolshoye Shchuchye are comparable to those documented in nearby pollen records from the region 54-56 , our record is longer, well-dated and more complete. Furthermore, the sedaDNA permits the identification of a sequential arrival of different dwarf-shrub taxa from 15,000 cal. years BP and the clear identification of larch (Larix sibirica) forest in the vicinity of the lake between ca. 9,000 and 4,000 cal. years BP (Larix is notorious for its low pollen productivity and underrepresentation in pollen records).
It is well accepted that arctic-alpine and boreal-steppe taxa are primarily sensitive to competition for light (rather than sensitive to heat per se), and that they can be shaded out by larger, shrubby growth forms [57][58][59] . From the sedaDNA data, it appears that the stepwise addition of shrub and sub-shrub taxa during the late-glacial may well have led to competition and the reduction of populations of many arctic-alpine taxa, resulting in a reduced www.nature.com/scientificreports www.nature.com/scientificreports/ diversity of arctic-alpines. Notably, the reduction of arctic-alpine plants occurred prior to the arrival of trees, and changes in plant dominance were well underway by the time of forest establishment. Thus, the current expansion of shrub-tundra 58,60 may pose a threat to the abundance and diversity of arctic-alpines, especially in homogenous tundra habitats where much of this expansion has been observed.
The timing of forest tree establishment in the vicinity of Lake Bolshoye Shchuchye (~9,000 cal. years BP and continuing until ~4,000 cal. years BP) is concurrent with the known northward treeline expansion into the lowlands of the Polar Urals and surrounding areas, which many authors consider representing warmer summer temperatures [61][62][63][64] . During this period, when arctic-alpine taxa would have been most disadvantaged, nearly half of the 31 arctic-alpine taxa were detected, at least in small quantities, which strongly suggests their continuous presence in the catchment from full-glacial times; all 31 are found in the region today.
Pollen and plant macrofossil records seldom show such clear evidence of continuity, because of the swamping effect of Holocene anemophilous pollen at virtually all northern sites and differential preservation of identifiable plant remains 65,66 . A survey of European late-Quaternary pollen data showed no increase in diversity into the Holocene in northern sites, in part probably due to this feature 67 . However, exceptional records with high pollen counts in regions never colonized by trees in the Holocene do demonstrate the local persistence of arctic-alpines from the late-glacial period through the Holocene, for example, that from Hanging Lake in mountains of the northern Yukon 26,68 . Although sedaDNA cannot fully resolve all aspects of the arctic-alpine plant community, it can greatly enhance the documentation of species persistence in the face of changing environmental conditions. Intermittent absences inevitably present an interpretational problem in palaeorecords 22,69 . For the subset of taxa which do show absences during the periods of shrub and/or tree expansion, the possibilities are either that they became locally extinct, or that they were present but their DNA input to the sediment dropped below the threshold for detection, probably as a result of their decreased biomass on the landscape at a time when the biomass of woody taxa expanded. Furthermore, taxa that may have been growing near the lake in full-glacial time most likely shifted their range up slope with increasing warmth and associated competition, removing them from the lake edge and valley floor, which is a prime locality for DNA recruitment to sediment 45 . Thus, as with pollen of entomophilous taxa, but probably not to the same extent, there is likely a limitation in detectability for rare taxa in the sedaDNA signal.
A complementary line of evidence used to evaluate the likelihood of species persistence is that of modern genetic diversity patterns, where, generally, regions that have supported a flora over a relatively long period show higher levels of genetic diversity than regions recently recolonised. Genetic diversity shows an increasing gradient from Fennoscandia eastward to unglaciated areas of northern Russia and Siberia. Within this gradient, the Polar Urals are classified as intermediate 70,71 . This suggests a degree of long-term persistence (compared with Fennoscandia) and could reflect the presence of plant communities since the last major deglaciation of the northern segment of the Ural Mountains at ~60,000 cal. years BP 42,52,53 .
The arctic-alpine flora has been identified as one that is exceptionally threatened by projected future climate change 3 ; it is also one that is often difficult to track via traditional palaeorecords. The Lake Bolshoye Shchuchye record provides unusually robust empirical evidence of the persistence of an arctic-alpine flora through a long period of environmental change, demonstrating the buffering capacity of a spatially heterogeneous landscape. However, a distinct decline in abundance, which is likely related to biomass, began as soon as woody taxa started to expand, suggesting that in a future warming scenario, local species loss may occur long before tree establishment occurs. These empirical results strongly support the need for conservation awareness of potential biodiversity loss in arctic environments, particularly those with less topographic buffering capacity. We conclude that in the Arctic, as elsewhere, conservation priorities should go beyond protecting individual species or areas that are deemed important today and instead move towards monitoring and protecting areas that have shown to be resilient to past changes and thus tend to harbour high genetic diversity and biodiversity.

Methods
Study site. Lake Bolshoye Shchuchye (67°53′24″N, 66°18′36″E) is located at 187 m a.s.l. in the central part of Polar Urals mountain chain, ca. 105 km north-east of the mining town of Vorkuta (Fig. 1). The lake itself is ~13 km long and ~1 km wide and has a maximum water depth of 140 m in the centre of the basin. Seismic reflection profiling revealed that the basin fill contains more than 160 m of acoustically laminated sediments with minimal disturbances from mass movements 40,42 . The lake has a catchment area of 215 km 2 , with a deltaic inlet at its northern shore caused by inflow of the River Pyriatanyu and an outlet along its southern shore draining into the Bolshoye Shchuchya River, a tributary of the Ob River. The lake is surrounded by steep-sided valley slopes and exposed rock faces with high mountain peaks reaching 500-1,100 m a.s.l. at its north-western shore, intersected by valleys.
Present-day climate conditions are characterised as cold and continental, with a mean summer temperature (June-July-August) of 7 °C at the Bolshaya Khadata station 72 , located 25 km to the south of Lake Bolshoye Shchuchye at 260 m a.s.l. The lake lies within arctic bioclimatic sub-zone E (Shrub tundra), which is characterised by mean July temperatures of 10-12 °C 73 . The catchment vegetation mosaic comprises dwarf shrub-tundra with patchy thickets of Alnus viridis growing on south-facing slopes up to an elevation of 300 m a.s.l. At the higher elevations, the vegetation is discontinuous with exposed rocky surfaces supporting alpine grass and forb communities. The lake is situated close to the northern distributional limit of Larix sibirica, with isolated trees observed a few kilometres to the southeast of the lake.
Lake Bolshoye Shchuchye is considered to have been formed by glacial erosion during repeated past glaciations, following weaknesses along ancient NW-SE striking faults 42 . The lake and its catchment are located well outside the maximum extent of the Barents-Kara Ice Sheet during the LGM ca. 25,000-17,000 cal. years BP and the region is thought to have remained ice-free for at least the last 50,000-60,000 cal. years BP 42 www.nature.com/scientificreports www.nature.com/scientificreports/ periods around 90,000 cal. years BP (MIS 5b) and 70,000-60,000 cal. years BP (MIS 4), large ice sheets formed over the Barents-Kara Sea region, creating major ice-dammed lakes that flooded the adjacent lowlands on both sides of the Ural Mountains chain 56,74 . It remains controversial whether large ice caps formed over the Polar Urals, the northernmost part of the mountain chain, during this time 75 but it is clear that sizeable outlet glaciers from mountain valleys reached the foothills of the Polar Urals ca. 60,000 cal. years BP. In contrast to previous glaciations, the Barents-Kara Ice Sheet did not reach the northern rim of the mainland during the LGM and almost the entire Arctic seaboard of Russia to the east of the Arkhangelsk region remained ice-free during this glaciation 74 . Cosmogenic ( 10 Be) exposure dating revealed that the small cirque glaciers that exist today within the Polar Urals were only slightly larger during the LGM than today 42,76 . Sediment core retrieval. Six separate cores were retrieved from Lake Bolshoye Shchuchye, full details of which are provided by an earlier accompanying paper 42 . A 24-m long core (number 506-48) retrieved in July 2009 from the southern end of the lake (67 o 51′22.248″N, 66 o 21′30.096″E) is the focus for this study. The core was retrieved with a UWITEC piston corer using a combination of 2-m long by 10-cm diameter PVC sample tubes for most sections and 2-m long by 9-cm diameter steel tubes for the deepest sections. Since collection in July 2009, the core sections remained unopened and sealed within light-proof cylinders in a cold storage facility until they were opened and split longitudinally in the winter of 2014, when subsamples for radiocarbon dating were taken. The core sections were sealed and placed back into cold storage until subsampling for sedaDNA analyses was undertaken in the winter of 2015. A detailed sedimentological description and the core chronology ( Supplementary Fig. S4), which is based on a series of AMS radiocarbon dates supported by a sequence of annual laminations (varves), is presented in two earlier publications 41,42 . Radiocarbon dating and sampling of the sediment core. In total, 26 samples of terrestrial macrofossils were radiocarbon ( 14 C) dated with Accelerator Mass Spectrometry (AMS) at the Poznań Radiocarbon Laboratory of the Adam Mickiewicz University, Poland. All radiocarbon ages were calibrated according to the terrestrial IntCal13 curve 77 using the online Calib program 78 . Subsampling of the core was undertaken at ca. 15-cm resolution in a laminar flow cabinet within a clean laboratory within the Centre for Geobiology and Microbiology at the Department of Earth Science, University of Bergen, Norway, using sterile tools, full bodysuit, facemask, and gloves. Subsampling was undertaken in the presence of subsampling controls (open water samples) in order to detect potential lab aerial contamination. Following the protocol described by Parducci et al. 36 the outer 10 mm of sediment was avoided and an ~20 g subsample was retrieved from inside the freshly exposed centre only.
DNA extraction, amplification, library preparation and sequencing. DNA extraction, PCR amplification, PCR product pooling and purification, and sequencing followed the protocols of Alsos et al. 35 and Clarke et al. 31 unless otherwise stated. DNA was extracted from 153 sediment subsamples within the dedicated ancient DNA facility at Tromsø University Museum (TMU), Norway. In addition, DNA was extracted from 17 negative extraction, nine negative PCR and nine negative water (subsampling) controls, which contained no sediment and were used to monitor for contamination. Aliquots of DNA extracts were then shipped to the Laboratoire d'ÉCologie Alpine (LECA, Université Grenoble Alpes, France) for metabarcoding. Each DNA extract and negative extraction control was independently amplified using uniquely-tagged generic primers that amplify the trnL P6 loop of the plant chloroplast genome 79 in eight PCR replicates to increase confidence in the results and improve the chance of detecting taxa with small quantities of template in the DNA extracts 80 . Pooled and cleaned PCR products were then converted to four Illumina-compatible amplicon libraries using the single-indexed, PCR-free MetaFast method (FASTERIS SA, Switzerland). These libraries were then sequenced on the Illumina HiSeq-2500 platform for 2 × 125 cycles at FASTERIS. DNA sequence analysis and taxonomic assignments. Next-generation sequence data were filtered using the OBITools software package (Boyer et al., 2016; http://metabarcoding.org/obitools/doc/index.html) following the protocol and criteria defined by Alsos et al. 35 . Taxonomic assignments were performed by first matching sequences against a reference library of 2445 sequences comprising 815 artic 32 and 835 boreal vascular plants 29 , in addition to 455 arctic/boreal bryophytes 33 . Sequences were then matched to a second reference library generated after running ecopcr on the global EMBL database (release r117).
In order to minimise any erroneous taxonomic assignments, only taxa with a 98% match, or greater, to a reference sequence were retained. A minimum threshold of at least 10 reads per PCR replicate was used, except in the case of the arctic-alpine plant taxa (see Fig. 2) where <10 reads per PCR replicate occurrences are highlighted. We further removed sequences that displayed higher average frequency in the negative subsampling (water), extraction or PCR controls than in the lake sediment samples in which they were present. Identified taxa were compared to the circumpolar flora 81 , Pan Arctic Flora Checklist 82 and flora of the Polar Urals [83][84][85] . Sequences with <100% match to a reference sequence and/or those assigned to taxa that are not present in northern Russia today were carefully crosschecked against the NCBI BLAST database for multiple or alternative taxonomic assignments (http://www.ncbi.nlm.nih.gov/blast/). Taxonomy follows the Pan Arctic Flora Checklist 82 , if present.
Distributional classification of taxa. Where possible, we assigned a classification of either "arctic-alpine", "boreal" or "arctic-boreal" to each vascular plant taxon identified by sedaDNA based on their present-day native distribution 81,86 . Plant taxa classified as "arctic-alpine" had two-thirds or more of their main distributional area located north of the forest limit for boreal tree taxa Pinus, Picea, Larix and Betula. Those classified as "boreal" had two-thirds or more of their distributional area located within the boreal forest limits. Finally, those assigned an "intermediate" classification occupied an area where approximately half of their main distributional area was located within each biome. For those taxa which were identified to genus or family level only by sedaDNA, a