Beavers today are renowned for their woodcutting behaviour and construction abilities. The extant genusCastor harvests trees and shrubs for sustenance (particularly during the winter¹), but also for the purpose of lodge and dam building. These behaviours have a profound effect on regional hydrology, nutrient flow across the landscape, and local biodiversity, thus making them exemplary “ecosystem engineers”^2,3,4. Modern beavers also are known to feed on trees. This food source is particularly important for northern populations that survive freezing winters, subsisting on their underwater caches of leafy branches^5,6.

Castoridae is a diverse family of herbivorous Holarctic rodents, originating during the late Eocene^7,8,9. The only definitive evidence of woodcutting in extinct beavers is from the Pliocene-aged High Arctic Beaver Pond fossil site, which preserves evidence of beaver-cut wood and a possible dam-core associated with the extinct beaver genus,Dipoides (species not yet described)^9,10,11. Examination ofDipoides sp. incisors and cut marks on wood from the Beaver Pond site suggest thatDipoides sp. woodcutting performance was poorer than that of modernCastor¹¹. The appearance of woodcutting inDipoides sp., a distant relative ofCastor, implies that this behaviour originated 20 to 24 Ma ago, in a group of semiaquatic beavers that includes both extant species (Castor canadensis andCastor fiber), the small Holarctic genusDipoides, and the North American Ice Age giant beaverCastoroides^7,9,12. And yet recent research shows that the diet of Ice AgeCastoroides, a close relative ofDipoides, was dominated by submerged plants, not trees and shrubs¹³. This finding, alongside a lack of definitive evidence of lodges or dams at the Beaver Pond site highlights the question: Does the cut wood at the Beaver Pond site represent a means of gathering food?

Here, we present stable carbon and nitrogen isotope data for Pliocene-age (i) plant macrofossils and (ii) bone collagen from High ArcticDipoides sp. subfossil remains. The specimens originate from a peat deposit at the Beaver Pond fossil site, located on Ellesmere Island (locally known as Umingmak Nuna, meaning “land of muskoxen”), situated within the Canadian Arctic Archipelago (78° 33′ N, 82° 25′ W) (Fig. 1). We reconstruct High ArcticDipoides sp. palaeodiet within the context of coeval terrestrial and freshwater plant macrofossil remains excavated from the same ~ 4 Ma old peat layer. The Beaver Pond site provides a very rare opportunity for such a palaeodiet reconstruction using coeval herbivore and plant remains.

Map of the Canadian Arctic Archipelago. The Beaver Pond site is located near the head of Strathcona Fiord, Ellesmere Island (78° 33′ N, 82° 25′ W). Base map created in Adobe Illustrator (version 24.2.1),https://www.adobe.com/uk/products/illustrator.html.

Full size image

The Beaver Pond fossil site

The Beaver Pond site is a succession of fossiliferous peat deposits interspersed within 40 m of sandy, cross-bedded, fluvial deposits and capped with glacial till¹⁴. The thickest peat deposit within this sequence, which yielded material for this study, sits 380 m above present day sea-level overlooking Strathcona Fiord. Most recent terrestrial cosmogenic nuclide burial dating of sands above this peat layer have yielded a date of 3.9 + 1.5/− 0.5 Ma¹⁵, placing the minimum age of peat formation during the late Early Pliocene (late Zanclean). This peat unit is characterized by an abundance of beaver-cut sticks^10,11 and has yielded a wide array of beautifully preserved plant macrofossils, invertebrates, and vertebrate fossil remains (Fig. 2). Rybczynski and Harington¹⁶, Matthews and Fyles¹⁷, Hutchison and Harington¹⁸, Tedford and Harington¹⁰, Dawson and Harington¹⁹, Murray et al.²⁰, Mitchell et al.²¹, Gosse et al.²², and Wang et al.²³ provide detailed descriptions of Pliocene terrestrial flora and faunal assemblages associated with the Beaver Pond site.

(A) An in-situ macrofossil cone within the Beaver Pond fossiliferous peat deposit. Scale bar is 1 cm. (B) Excavation of the peat deposit at the Beaver Pond site (2008), Strathcona Fiord, Ellesmere Island. White arrow indicates person for scale. (C) A beaver-cut stick excavated from the Beaver Pond site. Cut marks produced byDipoides sp. Photographs by M. Lipman.

Full size image

Ellesmere Island today is a polar desert, with sparse flora and very little precipitation²⁴. The landscape was very different during the Early Pliocene warm period, when the climate supported wetland habitat surrounded by open larch forest^10,21,22,25. During the Pliocene, Ellesmere Island was on the eastern edge of a large coastal plain, where intense Neogene thawing and weathering liberated sediment to create a thick, continuous clastic wedge across the Canadian Arctic Archipelago (referred to as the Beaufort Formation in the western Canadian Arctic islands). It is hypothesized that northwest passages did not exist during the Pliocene, as they had yet to be incised by fluvial and glacial erosion^22,26. This unbroken coastal plain altered ocean circulation patterns in the High Arctic, and along with the Bering Isthmus that connected North American and Eurasia until 5.5 Ma ago²⁷ would have enabled the migration of terrestrial species across the Neogene High Arctic¹⁰.

Late Early Pliocene mean global temperature was 2–3 °C above modern²⁸ and high latitude regions experienced amplified warming. Pliocene Arctic mean annual temperature was near freezing, which is ~ 15–22 °C warmer than present day, and tree line was ~ 2000 km further north^{15,24,29,30,31,32,33}. Summer temperatures at the Beaver Pond site reached 20 °C, while winter temperatures were more moderate than present day, with a low of ~ – 12 °C²⁴. Despite the relatively mild conditions, the Beaver Pond site still experienced total darkness and subzero temperatures during the winter months.

There is no modern analog for the ecological community found at the Beaver Pond site, although the flora of present-day Labrador (Canada) is considered to be similar^22,34,35,36. The Beaver Pond macrofossil assemblage indicates a larch forest, although birch, alder, spruce, pine, cedar, and cold-adapted woody shrubs were also present (see Matthews and Fyles¹⁷ for a complete list of identified flora). The Beaver Pond site supported higher faunal biodiversity than any modern near-tree line communities.

The remains of a complex faunal community were discovered at the Beaver Pond site, including beaver (Dipoides sp., see below), three-toed horse, bear, badger, “deerlet”, water fowl, fish and a rabbit relative^10,22. Pliocene-age sites with similar fauna and flora community composition are very rare—Idaho in mid-continent North America, and the high altitude Yushe Basin, in northeastern China are the only known sites with similar (but not equivalent) faunal assemblages^37,38.

Dipoides ecology

The most common vertebrate remains at the Beaver Pond site belong toDipoides, an extinct genus of beaver known from the Neogene of Eurasia and North America, represented by 12 different species^7,39,40,41. Although not directly related to the extantCastor, both genera share semiaquatic and woodcutting behaviours^9,11,42.

The Beaver Pond site is the only known locality with sufficiently well-preserved plant macrofossils to record evidence ofDipoides sp. woodcutting behaviour. The peat deposits are hypothesized to be the remnants of an ancient beaver pond due to the presence of many beaver-cut sticks, and even a cluster of cut sticks, cobbles, and silt that resemble the core of a beaver dam¹⁰.

Here we use stable isotope data to understandDipoides sp. diet and elucidate the purpose of their woodcutting behaviour. Our study aims to describe the relative contributions of woody vegetation and aquatic plants to the diet ofDipoides sp. using stable carbon and nitrogen isotope analysis of contemporary sub-fossil skeletal and plant macrofossil material from the Beaver Pond site. This new information is used to better interpret (i) the ecological impact ofDipoides sp. on the Pliocene landscape, (ii)Dipoides sp. potential for winter survival strategies such as underwater food caching, and (iii) the evolutionary context of tree-exploitation within the Castoridae family.

Stable isotopes and palaeodiet

The stable carbon (δ¹³C) and nitrogen (δ¹⁵N) isotope compositions of an animal’s bodily tissues correlate closely with that of its diet, when adjusted for¹³C- and¹⁵N-enrichment that occurs during collagen formation and with each successive trophic level^43,44. Well-preserved bone collagen is therefore a useful integrator of an animal’s diet. In addition, sufficient context is required to accurately describe the nutrient flow between subsequent trophic levels of an ecosystem. In particular, the diet of an organism must be interpreted within the context of an appropriate dietary baseline. This baseline is composed of isotopically defined food or “menu-items” available to the organism.

The isotopic composition of primary producers at the base of the food chain control theδ¹³C andδ¹⁵N of the dietary baseline for herbivores. Theδ¹³C andδ¹⁵N of primary producers depend on physiology (i.e. which photosynthetic pathway the plant employs) and the isotopic composition of bioavailable sources of C and N (i.e. atmospheric CO₂). Casey and Post⁴⁵ provide a thorough review of how primary producerδ¹³C andδ¹⁵N vary with physiology and local terrestrial and aquatic environmental conditions. A particular challenge in many forested-wetland environments, however, is that the carbon and nitrogen isotope range of terrestrial and freshwater plants overlap. There are, however, sufficient differences between theδ¹³C andδ¹⁵N of terrestrial vegetation utilizing the C3-photosynthetic pathway and vascular freshwater plants (macrophytes) for them to serve as useful endmembers of herbivore diet in such environments (see Methodology).

Another challenge is that the isotopic composition of regional and global C and N baselines (and subsequently, that of primary producers that use them) can change over time^46,47. Hence, reconstructing the diet of herbivores that lived thousands or millions of years ago can be problematic when using isotopic data, as is very rare to find sufficiently preserved coeval plant material and faunal remains from the same geologic locality. Typically, isotopic data for modern plants are all that are available in palaeodiet studies. Fortunately, much of this concern is alleviated at the Beaver Pond site, given the excellent organic preservation of coeval plant and animal tissues.

Bone collagen stable carbon and nitrogen isotopes

Dipoides sp. (n = 5) bone collagen stable isotope results (δ¹³C_col andδ¹⁵N_col) are presented in Table1.Dipoides sp.δ¹³C_col ranges from − 20.8 to − 19.1‰, with a mean of − 20.3‰, andδ¹⁵N_col ranges from + 3.2 to + 5.8‰, with a mean of + 4.7‰ (Fig. 3).

Comparison of bone collagenδ¹³C andδ¹⁵N among PlioceneDipoides sp. (4 Ma, from the Beaver Pond Site, Ellesmere Island), modernCastor canadensis (collected 2013–2014, from Yukon Territory), and PleistoceneCastoroides (late Pleistocene, from Beringia, Yukon Territory and Ohio, USA).Castor canadensis andCastoroides carbon isotope compositions are corrected for Suess effects appropriate to their time period (see text).Castor canadensis andCastoroides isotope data from Plint et al.¹³.

Full size image

Atomic C:N ratio and carbon and nitrogen contents (wt%) were used to assess collagen preservation forDipoides skeletal material (Table1). All specimen parameters are within the accepted range for well-preserved archaeological or palaeontological skeletal material from temperate or polar regions reported by van Klinken⁴⁸ (acceptable parameters: C wt% = 34.8 ± 8.8%; N wt% =  ~ 11–16%; atomic C:N = 3.1–3.5).

Plant macrofossil species diversity

By volume, the Beaver Pond peat sample examined in this study consisted of 85% bryophytes, 10% wood and twigs, and 5% macrofossils. Eleven genera were identified, representing a diverse assemblage of terrestrial and freshwater plants (Table2). Seven taxa were analyzed for stable carbon and nitrogen isotope compositions (Scorpidium scorpioides,Larix,Betula, Stuckenia filiformis,Scheuchzeria sp.,Cornus sericea,Menyanthes trifoliata) (Table2). The dominant moss type wasScorpidium scorpioides (hooked scorpion moss).Larix (larch—a deciduous conifer) was the only conifer species identified from this peat sample (although many other tree species have been previously identified from the Beaver Pond site—seeIntroduction). Plant macrofossils from multiple genera (Myrica,Shepherdia,Potamogeton, andHippuris) and from three species ofCarex (Carex aquatilis,Carex diandra, andCarex maritime) were also recognized, but in insufficient quantities for stable isotope analysis.

Plant macrofossil stable carbon and nitrogen isotopes

Plant macrofossil stable isotope results are presented in Table2 and Fig. 4. Macrofossilδ¹³C,δ¹⁵N, C (wt%), N (wt%), C/N (wt%), and atomic C:N ratios are all within the range expected for terrestrial and freshwater plants (Table2, Figs. 4 and5). Beaver Pond plant macrofossilδ¹³C andδ¹⁵N range from − 36.6 to − 22.7‰, and + 0.1 to + 4.8‰, respectively.

Stable carbon and nitrogen isotope results for the Beaver Pond site plant macrofossils.

Full size image

Beaver Pond plant macrofossil carbon and nitrogen content represented by C/N (wt%).

Full size image

The chemical and elemental compositions of plants vary widely by species, life history stage (i.e. senescence), and environment conditions. For these reasons, atomic C:N ratio and carbon and nitrogen content are not considered to be infallible indicators of organic preservation in subfossil plants⁴⁹. That said, the plant macrofossil C (wt%) values obtained in the present study are within or close to the mean carbon content for modern plants (between ~ 40 and 47%) (Metcalfe and Mead⁴⁹, and references therein). Plant macrofossil N (wt%) values are lower than the mean nitrogen contents of modern plants (between ~ 1 and 3%) but are not outside the reported range for modern plants.

Stable isotope analysis in R (SIAR) mixing model

Next, we evaluateDipoides sp.δ¹³C_col andδ¹⁵N_col within the context of the isotopic dietary baseline composed of coeval terrestrial and freshwater vegetation from the High Arctic Beaver Pond fossil site. The faunal and plant macrofossil isotope data were incorporated into a Bayesian mixing model to determine the relative input of terrestrial versus freshwater plants toDipoides sp. diet (Fig. 6). This also allowed us to better assess the connection betweenDipoides sp. woodcutting behaviour and its consumption of woody plants.

Stable carbon and nitrogen isotope compositions of plant functional groups andDipoides sp. bone collagen generated using SIAR mixing model. Plant functional groupings include terrestrial woody plants (larch, birch, and red-osier dogwood), vascular freshwater macrophytes (pondweed, podgrass, and bogbean), and bryophytes (Hooked scorpion moss).Dipoides sp. bone collagenδ¹³C andδ¹⁵N are corrected for trophic enrichment factors to render them comparable to the three plant functional groups (represented by their mean and a range of 2SD). TheDipoides sp. data, once so corrected, overlap with the plant functional groups that contributed significantly to their diet.

Full size image

The SIAR model is a statistical tool that uses biotracers (stable isotopes) to estimate the relative input of different sources to a product or mixture. In (palaeo)ecology, mixing models use the stable isotope compositions of different food sources to infer their relative contributions to the composition of overall diet, and assess the probability that the inferred proportions are correct. There are systematic differences in the isotopic composition between a consumer’s collagen and its diet, both for carbon and for nitrogen. Hence, a correction factor must be applied to render data for consumers and possible diet items directly comparable.

Plant macrofossils were divided into three sources, or functional types: terrestrial woody plants, vascular freshwater macrophytes, and bryophytes (mosses). These groupings were created to assess how terrestrial and freshwater resources contributed separately toDipoides sp. diet. The functional types were statistically defined and their relative contribution to diet was assessed using scripts from SIAR V4 in R Studio 3.1.2 (Stable Isotope Analysis in R: An Ecologist’s Guide). Based on existing literature, respective bone collagen-to-diet offsets of + 4.2‰ and + 3.0‰ were subtracted fromDipoides sp.δ¹³C_col andδ¹⁵N_col when incorporated into the mixing model^50,51,52,53.

Dipoides sp. palaeoecology

The Bayesian mixing model indicates thatDipoides sp. consumed both woody plants and freshwater macrophytes in approximately equal proportions (Figs. 6 and7), although it relied slightly more on freshwater macrophytes. This suggests thatDipoides sp. spent a greater proportion of time feeding in the water than on land.

(a) Proportion versus Source Boxplot generated using SIAR, indicating the relative proportion that moss, woody vegetation, and aquatic macrophytes contributed to the diet ofDipoides sp. at the Beaver Pond site. Darker shaded areas indicate highest probability of source proportion. The Proportion versus Source Boxplots for (b) extantCastor canadensis and (c) late PleistoceneCastoroides have been included for comparison. Note the differences in dietary Source data used to distinguishC. canadensis andCastoroides diet (primarily the sub-division of aquatic plants into categories based on habitat within the water column). b and c from Plint et al.¹³.

Full size image

The distribution ofDipoides sp.δ¹³C_col andδ¹⁵N_col is not entirely enclosed within the three primary producer functional groups analyzed (Fig. 6). This is likely the result of the relatively small plant macrofossil sample size. Submerged aquatic macrophytes, for example, are under-represented in the plant macrofossils available for stable isotope analysis. Macrophytes have highly variableδ¹³C and may have contributed more toDipoides sp. diet than the mixing model suggests. Submerged macrophytes can be highly enriched in¹³C because of physiological differences (primarily the use of¹³C-enriched dissolved bicarbonate) or environmental conditions in the water column (i.e. boundary-layer effect)^54,55,56. In addition, tree bark is more enriched in¹³C than tree foliage⁵⁷ and may have been a key resource forDipoides sp.

The results of the dietary mixing model support the interpretation that woody plants were an important contributor toDipoides sp. diet. It is likely thatDipoides sp. also used shrubs and trees as a source of construction material^10,11, but more evidence is needed to confirm this. Similar to extantCastor,Dipoides sp. may have also demonstrated regional differences in diet, where northern and southern populations utilized different resources according to their availability.

Nitrogen content and C/N as indicators of forage quality

Plant macrofossil nitrogen content (N wt%) and C/N are indicators of forage quality and may be used to interpret the relative nutrition of dietary inputs. Plants with high N (wt%) contain more protein and energy—likewise, low N (wt%) correlates with low plant digestibility, high fiber and high lignin compound content⁵⁸. Beaver Pond plant macrofossil N (wt%) and C/N are highly variable (Table2, Fig. 5). Although there is considerable variability in C/N ratios depending upon which plant part was analyzed (i.e. seeds versus woody tissue), woody vegetation tends to have higher C/N ratios than macrophytes, and thus tends to be of lower food quality. However, the increased structural tissues in woody plants may have rendered them more effective winter cache foods.

In extremely seasonal environments such as the High Arctic, herbivores must use plant resources in a highly efficient manner. Herbivores must consume the highest quality forage possible during the brief growing season to maximize nutrient and energy gain. High quality forage typically includes young leaves with high nitrogen content, minimal structural (fibrous) tissues, and low defense compound content^59,60.

Within the Beaver Pond macrofossil assemblage, pod grass (an emergent macrophyte) and birch have the highest nitrogen content and lowest C/N (Fig. 5). A larger sample set is necessary to confirm this observation; however, current data supports the conclusion that emergent macrophytes and deciduous broadleaf trees were among the more nutritious types of forage available toDipoides sp. at the Beaver Pond site. It should be noted that forage quality is not the only factor that governs herbivore feeding behaviour. Animals may preferentially target plants with higher biomass to minimize energy expenditure traveling between forage sites or select plants that grow in locations that minimize the risk of predation.

The C/N of high Arctic shrubs decreases over the course of the growing season⁵⁸. As there is no time-constraint on macrofossil deposition at the Beaver Pond site, variation in C/N may also be due to differences in plant phenological stage at time of incorporation into the peat layer. The incorporation of senescent plants into the peat deposit at the end of each growing season may in part account for the lower than expected macrofossil N (wt%) values reported from this site.

Beaver Pond site flora δ¹³C and δ¹⁵N

The Beaver Pond macrofossil assemblage contains a diverse range of terrestrial and freshwater plant species. The identified plant species in this study concur with previous interpretations that this was an open-forest landscape interspersed with shallow wetlands. Larch trees and cool-climate woody shrubs dominated the forest community. The wetlands supported both vascular macrophytes and dense assemblages of bryophytes.

The macrofossilδ¹³C are all within the range expected for primary producers utilizing the C₃ photosynthetic pathway and accessing either ambient or dissolved atmospheric CO₂ as their dominant carbon source. Theδ¹⁵N of the macrofossils are also within the expected range for a riparian ecosystem in a cool climate biome.

WhileDipoides sp. most likely consumed leafy tree branches and woody tissues (cambium), it is worth noting that plant seeds and cone bracts were analyzed in this study due to ease of macrofossil taxonomic identification. Leafδ¹³C is typically lower than that of other plant parts⁶¹, although there is no clear pattern in intra-plant variation ofδ¹⁵N.

Moss

Samples of the dominant Beaver Pond site bryophyte,Scorpidium (hooked scorpion moss), have very lowδ¹³C for a primary producer (− 36.6 to − 34.6‰). This pattern is consistent with modern mosses collected from freshwater habitats in Subarctic and Arctic regions^13,62,63,64.

Environmental conditions dictate mossδ¹³C rather than species-specific physiological differences. Peat mosses can grow partially or fully submerged in water. Given that mid-Pliocene atmospheric CO₂ concentration levels were similar to modern (~ 400 ppm)^15,65,66, moss exposed to the atmosphere would preferentially have used the abundant¹²CO₂, resulting in lowδ¹³C. Alternatively, lowδ¹³C in peat mosses can also indicate an underwater growing environment rich in¹³C-depleted respired CO₂ from surrounding plants⁶².

Moss macrofossilδ¹⁵N is relatively high for a photosynthetic organism (mean =  + 4.8‰). This is indicative of either the presence of¹⁵N-enriched sources of bioavailable N (i.e. dissolved nitrates, organic proteins such as urea or amino acids), or increased nutrient availability^45,61. Unlike vascular plants, mosses do not uptake compounds through their roots. Rather, they obtain nutrients from wet or dry deposition through their leaves^67,68. Today, beaver ponds are considered to be N sinks, with elevated rates of bacterially mediated denitrification⁶⁹. These bacterial processes result in¹⁵N-enriched products that are readily dissolved and used by plants (including moss) living in an aqueous environment. Decomposition processes also increase plantδ¹⁵N over time⁴⁷ and remineralized organic debris decomposing in wetlands may be particularly¹⁵N-enriched.

Beaver Pond bulk peat samples and moss macrofossils show a similar isotopic pattern (lowδ¹³C and highδ¹⁵N), which suggests that hooked scorpion moss contributed substantially to peat biomass accumulation at the Beaver Pond site.

Macrophytes

Beaver Pond macrophyteδ¹³C fall well within the albeit very wide known range for modern freshwater plants (− 50 to − 11‰, see Osmond et al.⁷⁰, Keeley and Sandquist⁵⁴, Mendonça et al.⁵⁵, and Chappuis et al.⁵⁶). It is reasonable to assume, however, that the very small sample size in this study hides the potential extent of the carbon isotope variability of macrophytes at the site.

Pod grass and bogbean are classified as emergent macrophytes (they grow rooted in water-logged substrates, but their leaves are exposed to the atmosphere), while pondweed grows entirely submerged. Submerged macrophytes become more enriched in¹³C as the dissolved CO₂ pool (the dominant carbon source) becomes increasingly limited⁵⁴.

The Beaver Pond site pondweedδ¹³C is relatively low (− 26.5‰) for a submerged macrophyte. This indicates that it grew in an aquatic environment with adequate dissolved CO₂. This is in keeping with the interpretation that the Beaver Pond was a fen (near neutral pH, cool water temperature) during the Pliocene. A lowδ¹³C may also indicate high influx of terrestrial organic biomass or mosses (with lowδ¹³C) into the water that subsequently remineralized and contributed to the dissolved inorganic carbon pool.

Environmental conditions strongly influence aquatic plantδ¹⁵N. Beaver Pond macrophyteδ¹⁵N (range =  + 0.2 to + 2.7‰) indicate interspecific access and use of a variety of different sources of bioavailable N within the water column and substrate. The most likely N sources are microbial-fixed atmospheric N₂ (which ranges from –2 to + 2‰), the products of nitrification/denitrification processes (¹⁵N-enriched NH₄⁺ or NO_x), and remineralized¹⁵N-enriched organic material (either terrestrial or aquatic)^71,72,73.

Larch

Larch (the extinct speciesLarix groenlandii) is the most common vascular plant species in this macrofossil assemblage.

There is an offset of ~ 2‰ between theδ¹³C of (i) larch shoots/buds (which bear the needles) and cone bracts, and (ii) larch seeds. Larch shoots and cones (δ¹³C range = ‒25.4 to − 25.1‰; mean = − 25.3‰) are more depleted of¹³C than larch seeds (δ¹³C range = − 23.3 to − 22.7‰, mean = − 23.1‰). This could be indicative of seasonal physiological or environmental conditions experienced by larch trees at the Beaver Pond site. The cones and shoots of extant larch trees begin growing in the early spring and have lowerδ¹³C, whereas their seeds (higherδ¹³C) do not develop and mature until mid to late summer⁷⁴.

A number of physiological and environmental conditions could be responsible for this offset between needle/-bearing structures and seeds. Atmospheric vapor pressure deficit (aridity) induces stomatal closure in vascular plants⁷⁵. This restricts not only the rate of water leaving the needle/tree, but also that of atmospheric CO₂ entering it. Stomatal closure reduces CO₂ entry and results in less discrimination against¹³CO₂. High levels of solar irradiance in the summer increase the rate of CO₂ assimilation. Plants growing at very high latitudes experience 24-h of daylight during the summer. This creates a greater demand for CO₂ to maintain photosynthesis and less discrimination against¹³CO₂. Both aridity and increased light levels could contribute to why Beaver Pond larch tissues grown late in the summer are more¹³C-enriched than those grown in the early spring/the previous fall.

Alternatively, trees can use water and carbon (in the form of sugars) stored during the previous year to promote new growth during the early spring when leaves are absent and light levels are low. Tissues that develop early in the growing season (i.e. needle-bearing buds and shoots) can therefore reflect theδ¹³C of photosynthetic conditions from the previous growing season^76,77. In addition, differences in the macromolecular (lipid, protein, sugar) composition of larch buds/needles versus seeds could account for their offset inδ¹³C (i.e. lipids are typically more¹³C-depleted than proteins).

Larchδ¹⁵N (mean =  + 2.7‰) indicate that these conifer trees had access to N sources other than “light” fixed atmospheric N₂. Given the proximity of wetlands, the root systems of larch trees may have had access to¹⁵N-enriched dissolved nitrates in the surrounding water-logged soils. Increasing foliar N concentration due to atmospheric N deposition also drives up plantδ¹⁵N^78,79.

Aridity may also have influenced terrestrial plants growing at the Beaver Pond site. Higher rainfall is inversely correlated withδ¹⁵N, where rainier ecosystems tend to produce more¹⁵N-depleted plants⁸⁰.

Similar toδ¹³C, there is an offset inδ¹⁵N (and N wt%) between larch needle-bearing structures (mean =  + 3.8‰; 0.9%) and larch seeds (mean =  + 2.1‰; 0.3%). This could indicate differences in the macromolecular composition of these different tissue types (where high N content typically indicates higher tissue protein content).

Comparison ofDipoides within Castoridae

The composition ofDipoides sp. diet differs from that of other members of Castoridae that lived in North America during the late Cenozoic. Pliocene High ArcticDipoides sp. (n = 5), modern subarcticCastor canadensis (n = 4) (Table1), and late PleistoceneCastoroides ohioensis (n = 11) (Table1)δ¹³C_col andδ¹⁵N_col are compared in Fig. 3. A correction for the Suess effect was first necessary render theδ¹³C of all three genera comparable. The carbon isotope composition of atmospheric CO₂ has changed over time with global climatic conditions. More recently, anthropogenic burning of fossil fuels that has rapidly released CO₂ enriched in¹²C into the atmosphere^46,81. Hence, a correction is needed when comparingδ¹³C of organic samples from different time periods to account for this isotopic variation in the primary carbon source of photosynthetic organisms at the base of the food web.

Suess effect corrections of + 2.02‰ and − 0.1‰ were applied to theδ¹³C_col of modernC. canadensis (collected in 2013 and 2014) andCastoroides (late Pleistocene in age), respectively. These corrections were based on the averageδ¹³C of atmospheric CO₂ (δ¹³C_CO2) calculated from Pliocene dual-benthic and planktonic foraminifera proxy records, spanning from ~ 4.1 to 3.8 Ma (averageδ¹³C_CO2 = − 6.55‰)⁷⁶. These foraminifera proxy records are approximately contemporary with the Beaver Pond site. Averageδ¹³C_CO2 for 2014 (− 8.57‰) was compiled from the Scripps CO₂ monitoring program. Averageδ¹³C_CO2 for the late Pleistocene (− 6.45‰) was compiled using ice core data from Schmitt et al.⁸².

Plants growing during these three different time periods (Pliocene, late Pleistocene, and modern/2014) would reflect theδ¹³C of contemporary atmospheric CO₂. Therefore, changes inδ¹³C_CO2 help explain differences inδ¹³C betweenDipoides sp. and modernC. canadensis. Additional factors, however, are important in explaining the wide range ofδ¹³C and large enrichment in¹³C measured forCastoroides.

Dipoides sp. diet composition differs from that ofCastoroides (the Pleistocene giant beaver) (Figs. 3 and7).Castoroides’ highδ¹³C_col andδ¹⁵N_col (meanδ¹³C_col = − 17.6‰ and meanδ¹⁵N_col =  + 5.8‰) indicate a diet composed predominantly of aquatic (particularly submerged) macrophytes and minimal woody plant material (Table1)¹³.

In comparison withCastoroides, bothDipoides sp. andC. canadensis have a relatively small range ofδ¹³C_col andδ¹⁵N_col (Table1) (Fig. 3).Dipoides sp. meanδ¹³C_col andδ¹⁵N_col are higher than those of modernC. canadensis (Fig. 3). This is attributable to either variation in diet between the two species, or changes in global C and N baselines over geologic time.

Previous mixing model studies predict that extantC. canadensis diet is composed of approximately equal proportions of woody terrestrial plants and aquatic macrophytes¹³. However, this can vary by latitude and season. For example, extantC. canadensis in the Canadian subarctic vary their winter diet significantly depending on habitat⁸³. It is worth noting that extantC. canadensis does not occur north of 70° latitude and High ArcticDipoides sp. living at 78° latitude may have employed different dietary strategies.

Dipoides sp. may have relied more heavily thanC. canadensis on underwater stores of tree branches to survive the long, dark polar winter. Tree bark is more¹³C-enriched than leafy vegetation⁵⁷ and increased consumption could account for the higherδ¹³C_col seen inDipoides sp. Variation in the quantity and type of macrophytes consumed by each beaver species could also account for this difference (i.e. emergent and floating macrophytes are, on average more¹⁵N-enriched than submerged macrophytes).

Changes in the isotopic composition of the C and N baseline between the Pliocene and the present could also account for the isotopic offset between beaver species. Further investigation of possible changes in theδ¹⁵N baseline of flora in terrestrial high latitude environments during the Pliocene would be a valuable avenue of future research.

Dipoides sp. behaviour and evolutionary implications

Evidence from the Beaver Pond site has implications for our understanding ofDipoides sp. ecology. These data also contribute to our understanding of the evolution of behavioural transitions within Castoridae. In particular, howCastor’s distinctive complex of behavioural traits (tree harvesting, underwater food caching, and construction behaviour) may have evolved. A new hypothesis of behavioural evolution in castorids based on evidence from the fossil record (i.e. fossil burrows, cut wood, and stable isotope measurements) and skeletal-dental morphology is mapped onto a simplified phylogenetic tree in Fig. 8^{42,84,85,86,87}.

Simplified Castoridae phylogeny showing behavioural reconstructions, including new evidence of woody plant consumption inDipoides sp. Diagram based on phylogenetic analysis by Rybczynski⁹, which used a matrix of 88 morphological characters and 38 taxa. The origination of dam building is a minimum age (~ 7–8 Ma), corresponding to the time of divergence ofCastor canadensis andC. fiber, inferred from molecular evidence⁹⁶ and supported by fossil evidence⁹⁷. Legend: CIRCLE—taxa that burrowed (Dipoides andCastoroides may have burrowed, but direct fossil evidence is currently lacking); WP—taxa with significant woody plant contribution to their diet; NWP—taxa that did not generally consume woody plants (the terrestrial burrowing clade is associated with open plains and unforested habitat, and therefore assumed to have not consumed significant amounts of woody plants); Plio—Pliocene; Q—Quaternary. Age range sources:Castor^96,97,103, nowdatabase.org;Steneofiber eseri¹⁰⁴; Fossorial clade^84,86,94,105;Eutypomys⁹⁴, Fossilworks.org, nowdatabase.org;Dipoides, includingD. tanneri: Fossilworks.org, nowdatabase.org;Castoroides¹⁰⁶, Fossilworks.org, nowdatabase.org. Fossil taxa behavioural evidence sources:Steneofiber eseri¹⁰⁴;Castoroides¹³;Dipoides (this study); Fossorial clade⁹⁰.

Full size image

Castoridae is a group of herbivorous rodents comprising roughly two dozen genera. Most fossil castorids fall within two major groups: a clade of fossorial specialists (Palaeocastorinae) and a semiaquatic clade^{42,84,86,88,89,90}. The latter includesCastor andDipoides. Members of the fossorial clade (~ 7 genera) possess striking specializations such large digging claws, extremely reduced tails, and broad, procumbent incisors for digging. In some cases, specimens have been found within fossil burrows (i.e.Palaeocastor, or “The devil’s corkscrew” burrows discovered in the plains of North America⁸⁸). The semiaquatic group comprises two subfamilies, Castorinae (~ 6 genera, includingSteneofiber and the extantCastor), and Castoroidinae (~ 7 genera, includingDipoides and the giant beaver,Castoroides). The oldest definitive Castorinae in the fossil record isSteneofiber eseri from the early Miocene (France, MN2, ~ 23 Ma).S. eseri shows evidence of living in family groups and swimming specializations⁹¹. This, in combination with aDNA evidence¹², suggests Castorinae and Castoroidinae are derived from a semiaquatic ancestor in the early Miocene.

Digging behaviour was not just characteristic of the fossorial group and appears within the semiaquatic clade as well.Castor, though not morphologically highly specialized for the task, digs bank burrows and creates extensive canal systems⁹². In addition, the extinct semiaquatic beaverSteneofiber eseri was found within a burrow⁹¹. Considering the phylogenetic distribution of burrowing behaviour within the Castorid tree (Fig. 8), it is likely that the common ancestor of the fossorial and semiaquatic clades also burrowed. Thus, the appearance of burrowing behaviour withinCastor andSteneofiber are seen as a retention of a primitive trait⁹.

If burrowing behaviour in semiaquatic castorids is the primitive condition, it is likelyDipoides burrowed as well, as seen in other semiaquatic rodents today such asCastor, but alsoCrossomys (earless water rat),Myocastor (nutria), andOndatra (muskrat)⁹². It is also possible thatDipoides constructed lodges. ExtantCastor andOndatra are known to construct burrows and lodges, depending on the characteristics of the habitat. Bank burrows are associated with stream environments, whereas lodges are better suited to calmer waters⁹². UnlikeCastor, extantOndatra construct their push-up lodges using cattails and other fibrous vegetation rather than wood. The abundance of cut wood at the Beaver Pond site¹¹ suggests thatDipoides sp. had the option to incorporate wood into their nesting structures, and possibly built lodges.

Given the occurrence of woodcutting and woody plant consumption within both subfamilies of semiaquatic castorids (represented byCastor andDipoides in Fig. 8), it seems likely these behaviours appeared in the common ancestor of the semiaquatic group. Woody plant consumption may have preadapted castorids to exploit colder environments that arose during and after the late Miocene.Castor canadensis does not hibernate, but builds and sink rafts of branches and foliage to use as a source of fresh food during the winter months^1,93.Dipoides sp. may have also engaged in this behaviour and used underwater caches of branches as a primary food source to survive the consecutive months of darkness during the high latitude winter when plants become dormant. The use of woody plants in this way may have been key to allowing beavers to disperse between North American and Eurasia, which required crossing the Bering Isthmus⁹⁴, a high latitude landmass. Curiously, given that a diet rich in woody plants appears to be the primitive condition of semiaquatic castorids, the absence of woody plant consumption seen in the Pleistocene giant beaverCastoroides¹³ must be interpreted here as an evolutionary loss and potentially a leading factor in their extinction (Fig. 8).

Among living mammals,Castor’s dam construction is a unique and highly derived behaviour – an evolutionary puzzle, associated with a set of innate behavioural specializations⁹⁵. For example, dam construction is well known to be triggered by the sound of running water alone⁹⁵. The presence of such “hard-wired” behaviours may be associated with the ancient origins of this behaviour. Molecular and fossil occurrence records indicate that the split between Eurasian and North AmericanCastor arose around 7.5 Ma ago^96,97, implying that dam building behaviour itself is at least as old.

Definitive fossil evidence for dam building by an extinct beaver is currently lacking. Consequently, dam building behaviour is shown as possibly arising only on the lineage leading toCastor. Hypothetically, dam building may have arisen from beavers collecting branches near their burrow/lodge for feeding purposes and the accumulations of sticks could have dammed streams by happenstance. The effects may have been multifold. A deeper pond is an effective defense mechanism and provides a safe refuge from predators. Raised water levels also create more favourable conditions for underwater food caching of branches in sub-freezing winter conditions because the deeper water would prevent an underwater food cache from being locked in ice. As such, natural selection would have favoured animals that maintained the dam, presumably as an extension of their pre-existing nesting behaviour such as lodge building. In this scenario, the climate cooling that started around 15 Ma ago and continued into the Pleistocene would have provided an interval where behaviours promoting over-wintering survival, such as underwater food caching branches and dam building, would have been increasingly reinforced by natural selection.

It seems unlikely that the common ancestor of all semiaquatic beavers was a dam-builder. ExtantCastor is a large powerful rodent weighing 12–25 kg, with some individuals as large as 40 kg⁹². Its body size is one factor that allows the animal to harvest branches and whole trees to build lodges and maintain dams over multiple years. The Beaver Pond siteDipoides sp. was also a large rodent and was roughly two-thirds the size of an average extantCastor. In contrast, the less-derived semiaquatic beavers, such as the MioceneEucastor tortus (Castoroidinae) andSteneofiber eseri (Castorinae) were small (~ 1 kg, or less), suggesting that the common ancestor of the semiaquatic lineage was also small bodied. Although the common ancestor of the semiaquatic beaver lineage is inferred to have consumed woody plants (this study), and may have used branches in creating food piles and wood for lodge construction, it would have been too small to have had the capacity to build and maintain dams. As such, ifDipoides sp. did exhibit dam building behaviour, it would be the result of parallel evolution within the Castoroidinae and Castorinae lineages.

Here, we reconstruct Pliocene High ArcticDipoides sp. palaeodiet from bone collagenδ¹³C andδ¹⁵N within the context of an isotopic dietary baseline composed of coeval ~ 4 Ma old terrestrial and freshwater plant macrofossil remains. The Beaver Pond site provides a very rare opportunity for such a palaeodiet reconstruction using coeval herbivore and plant remains. A Bayesian mixing model indicates thatDipoides sp. diet was composed of approximately equal proportions of woody plant material and freshwater macrophytes, with slightly more emphasis on macrophyte consumption.Dipoides sp. dietary preferences lie somewhere in between those of other North American late Cenozoic semiaquatic beavers (extantCastor and extinct Pleistocene giant beaver,Castoroides).

The consumption of woody plant material suggests that a proportion of the assemblage of the wood cut byDipoides sp. at the Beaver Pond fossil site was the result of harvesting for consumption, possibly as part of an underwater winter food cache. The results also suggest that the early Miocene ancestor of the semiaquatic beaver lineage engaged in woodcutting and consumed woody plants as part of its diet. Swimming, woodcutting, and a diet of woody plants could have set the stage for the evolution of dam building behaviours—advantageous traits that may have been selected for by the cooling climate of the late Neogene, and which have resulted inCastor’s modern role as a keystone species and ecosystem engineer.

TheDipoides sp. skeletal material and the plant macrofossils used in this study originated from the Beaver Pond fossil site, Unit III, as defined by Mitchell et al.²¹. Unit III is a peat layer that yielded the majority of the beaver-cut sticks and vertebrate faunal remains discovered at the site. It is interpreted to have been a rich fen connected to open water, within a larch-dominated forest ecosystem²¹.

Plant macrofossil preparation

Plant macrofossils were isolated from bulk samples of Unit III peat and identified to taxon. Macrofossils were extracted from the peat using a combination of water-flotation and wet-sieving. Organic material greater than 0.425 mm was retained for further cleaning. Adhered sediment and moss were removed from the macrofossils using surgical forceps and repeated ultrasonic water baths. Cleaned macrofossils were dried at 26 °C for 24 h and identified to taxon using a binocular microscope.

Stable isotope analysis

Dipoides sp. bone collagen was extracted and itsδ¹³C andδ¹⁵N measured at the Alaska Stable Isotope Facility (UAF). Collagen extraction was performed using a modified Longin⁹⁸ method of gelatinization. Organic contaminants were removed using XAD-2 resin^99,100 and collagen purification was performed according to methods developed by Matheus¹⁰¹. Non-soluble collagenous portions were rinsed to neutral pH, but not subjected to base treatment. Collagen was gelatinized in weak HCl (pH 3) under N₂ gas at 105° C until dissolved (2 to 6 h). The solution was centrifuged and filtered with a 0.45 µm syringe-type PTFE filter and the supernatant containing dissolved collagen was lyophilized and weighed to determine the percent collagen yield. Lyophilized collagen was then hydrolyzed in 6 N HCL under N₂ gas for 4 h at 120 °C. The hydrolyzates were passed by gravity flow through 2 cc of compacted Serva XAD-2 HPLC resin in syringe columns to extract humates and other long-chain organic contaminants that can adhere to fossil collagen. The hydrolyzates were passed through a 0.45 µm PTFE filter placed at the distal end of each syringe column and were dried by rotary evaporation. Stable carbon and nitrogen isotope analysis of the hydrolyzed collagen was performed using a GC-Isolink gas chromatography combustion system coupled to a Thermo Scientific Delta V Plus isotope ratio mass spectrometer operated in continuous flow mode, using helium as the carrier gas.

Plant macrofossils were powdered using a ball-bearing mill and weighed into tin capsules (0.38 ± 0.02 mg). Stable carbon and nitrogen isotope analysis of macrofossil remains was conducted at the LSIS-AFAR facility at the University of Western Ontario (London, Canada). Samples were analyzed in continuous flow mode using a Costech elemental analyzer (ECS 4010), coupled to a Thermo Scientific ConFlo IV and Delta V Plus isotope ratio mass spectrometer in continuous flow mode, using helium as the carrier gas. One method duplicate (complete duplication of sample preparation and isotopic analysis) and one analytical duplicate (separate isotopic analysis of sample powder) were included for every ten samples. The carbon and nitrogen isotope measurements of the plant macrofossils were completed in separate analytical sessions. The first session was used to determineδ¹³C and nitrogen content (weight percent, N wt%); values ofδ¹⁵N were determined in the second session, using individually tailored weights based on each sample’s N wt%.

All isotopic results are reported inδ-notation in per mil (‰) relative to international standards. Collagenδ¹³C andδ¹⁵N were calibrated to VPDB and AIR, respectively. Analytical accuracy and precision were 0.0‰ forδ¹³C measurements, and 0.2‰ forδ¹⁵N measurements.

Plant macrofossilδ¹³C andδ¹⁵N were calibrated to VPDB and AIR, respectively using USGS40 (acceptedδ¹³C = − 26.39‰, SD =  ± 0.0‰; acceptedδ¹⁵N = − 4.52‰, SD =  ± 0.2‰) and USGS41a (acceptedδ¹³C =  + 36.55‰, SD =  ± 0.1‰; acceptedδ¹⁵N =  + 47.55‰, SD =  ± 0.2‰). Additional reference materials IAEA-CH-6 (acceptedδ¹³C = − 10.45‰, SD =  ± 0.0‰) and NIST-1547 (Peach Leaves) (internally calibratedδ¹⁵N =  + 1.98‰, SD =  ± 0.1‰) were used to evaluate instrument precision and accuracy forδ¹³C andδ¹⁵N, respectively. A keratin powder (Spectrum Chemicals Mfg. Corp., derived from pig skin and hair) was also included to monitor instrument drift. Combined instrument and analytical errors were ± 0.1‰ forδ¹³C, and ± 0.2‰ forδ¹⁵N.

TheDipoides sp. and plant macrofossil isotopic results were incorporated into a statistically-based Bayesian mixing model (SIAR V4). This approach provides a statistically robust means of evaluating the relative dietary contributions of woody plants and aquatic primary producers toDipoides sp. Diet¹⁰².

The authors declare no limitations on data or standard operating protocol availability.

The Pliocene peat sample was collected in 2006, supported by the Canadian Museum of Nature (N.R.), National Geographic Exploration Grant, Scientific Research Grant # 7902-05 (N.R.) and with logistical support by the Polar Continental Shelf Program (N.R.). Collection permit from Nunavut’s Department of Culture, Language, Elders and Youth (Permit No. 2006-002P), and with permission from the Qikiqtani Inuit Association and the Hamlet of Grise Fiord (Aujuittuq). Financial support for this research was provided by a Natural Sciences and Engineering Research Council Discovery grant (F.J.L.), Canada Research Chairs program (F.J.L.), Canada Foundation for Innovation (F.J.L.) and the Ontario Research Fund (F.J.L.). This is Western’s Laboratory for Stable Isotope Science Contribution #378.Dipoides sp. bone collagen stable isotope measurements provided by Paul Matheus (University of Alaska Fairbanks). We thank Martin Lipman for providing photographs of the Beaver Pond site, Donna Naughton for provision of map templates, and Michael Burzynski for additional advice concerning plant physiology.

1. Tessa Plint

   Present address: The Lyell Centre, Heriot-Watt University, Edinburgh, EH14 4AS, UK

2. Alice Telka is deceased.

Authors and Affiliations

1. Department of Earth Sciences, The University of Western Ontario, London, ON, N6A 5B7, Canada

   Tessa Plint & Fred J. Longstaffe

2. Department of Ecosystem and Conservation Sciences, University of Montana, Missoula, MT, 59812, USA

   Ashley Ballantyne

3. Paleotec Services, 574 Somerset St W Suite 1, Ottawa, ON, K1R 5K2, Canada

   Alice Telka

4. Department of Palaeobiology, Canadian Museum of Nature, Ottawa, ON, K1P 6P4, Canada

   Natalia Rybczynski

5. Department of Biology, Department of Earth Sciences, Carleton University, Ottawa, ON, K1S 5B6, Canada

   Natalia Rybczynski

Contributions

T.P., F.J.L., and N.R. conceived the research design. N.R. conducted the sampling. A.T. provided palaeobotanical identification. T.P. conducted plant macrofossil stable isotope analyses and wrote the initial draft. F.J.L., N.R., and A.B. provided valuable comments and advice, and revised the various drafts of the manuscript. F.J.L. and N.R. provided funding to support the research.

Corresponding author

Correspondence toTessa Plint.

Competing interests

The authors declare no competing interests.

Publisher's note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visithttp://creativecommons.org/licenses/by/4.0/.

Reprints and permissions