Nitrogen and carbon isotopic dynamics of subarctic soils and plants in southern Yukon Territory and its implications for paleoecological and paleodietary studies

We examine here the carbon and nitrogen isotopic compositions of bulk soils (8 topsoil and 7 subsoils, including two soil profiles) and five different plant parts of 79 C3 plants from two main functional groups: herbs and shrubs/subshrubs, from 18 different locations in grasslands of southern Yukon Territory, Canada (eastern shoreline of Kluane Lake and Whitehorse area). The Kluane Lake region in particular has been identified previously as an analogue for Late Pleistocene eastern Beringia. All topsoils have higher average total nitrogen δ15N and organic carbon δ13C than plants from the same sites with a positive shift occurring with depth in two soil profiles analyzed. All plants analyzed have an average whole plant δ13C of −27.5 ± 1.2 ‰ and foliar δ13C of –28.0 ± 1.3 ‰, and average whole plant δ15N of −0.3 ± 2.2 ‰ and foliar δ15N of –0.6 ± 2.7 ‰. Plants analyzed here showed relatively smaller variability in δ13C than δ15N. Their average δ13C after suitable corrections for the Suess effect should be suitable as baseline for interpreting diets of Late Pleistocene herbivores that lived in eastern Beringia. Water availability, nitrogen availability, spacial differences and intra-plant variability are important controls on δ15N of herbaceous plants in the study area. The wider range of δ15N, the more numerous factors that affect nitrogen isotopic composition and their likely differences in the past, however, limit use of the modern N isotopic baseline for vegetation in paleodietary models for such ecosystems. That said, the positive correlation between foliar δ15N and N content shown for the modern plants could support use of plant δ15N as an index for plant N content and therefore forage quality. The modern N isotopic baseline cannot be applied directly to the past, but it is prerequisite to future efforts to detect shifts in N cycling and forage quality since the Late Pleistocene through comparison with fossil plants from the same region.


Introduction and background information
Stable isotopes of carbon and nitrogen are valuable for studying food webs and tracking transfer of energy and materials through trophic levels [1]. Primary producers in different ecosystems provide varied food sources with distinct carbon (δ 13 C) and nitrogen (δ 15 N) isotopic compositions at the base of the food web. The δ 13 C and δ 15 N of an animal reflect both the isotopic composition of its diet and fractionations during the building of organic tissues. Hence, an isotopic baseline for each ecosystem should be established prior to comparing animals from different regions and times [1]. This is particularly important for paleoecological and paleodietary studies of Late Pleistocene subarctic ecosystems. Such investigations have a special value in the study of whole ecosystem (structure, composition and function) responses to climate change.
Stable isotope data have been widely used in the study of Late Pleistocene subarctic mammals e.g. [2][3][4][5][6][7][8]. Determinations of appropriate modern and past local and regional, foodweb isotopic baselines, however, are still underrepresented in the literature. Subarctic ecosystems are not homogeneous. Local and regional ecological mosaics likely existed in the past because of variation in topography, soil moisture, loess deposition, altitude and animal disturbance [9] and are present currently owing to the bioclimatic subzones, patterned ground created by soilfrost processes, different plant communities, and the changes in elevation [10]. The possibility of such microhabitat diversity should be considered when establishing isotopic baselines for these ecosystems.
Soils and plants, which can show an extremely wide range of δ 13 C and δ 15 N, are two main components of all terrestrial ecosystems [11,12]. A number of studies have addressed plant δ 13 C [13,14] and δ 15 N [13,[15][16][17][18] variations in Arctic and subarctic ecosystems in North America, Eurasia and Iceland. Wooller et al. [14] conducted the most comprehensive study, analyzing the δ 13 C of herbarium modern plants (around 200 taxa) as well as fossil plants from Alaska and Yukon Territory. Schulze et al. [17] studied N isotopic and elemental compositions of different plants in northern Alaska with the aim of investigating their difference in nutrient acquisition. Similar studies have reported δ 15 N for various plants from other parts of this ecosystem [15,16]. The analysis of both δ 13 C and δ 15 N of plants has also been conducted for Carex from 15 Eurasian Coastal Arctic sites [18] and for lichens and plants from Iceland [13]. While these studies have added greatly to our general isotopic knowledge in these regions, additional local and regional isotopic studies are needed to evaluate the heterogeneity that can exist in these ecosystems.
In the present study, we measured the δ 13 C and δ 15 N of soils and plants from the south of Yukon Territory (eastern shoreline of Kluane Lake and areas around Whitehorse) (Fig 1). The study's purpose is to establish a local isotopic baseline for modern foodwebs and to gain a better understanding of main environmental factors affecting C and N isotopic dynamics in this region. The results should also enable future work in which modern and Late Pleistocene isotopic baselines for eastern Beringia are compared, in order to identify environmental factors that affected the baseline. Beringia was a largely ice-free land extending from northwest Canada to northeast Siberia (Fig 1) [19] and was located within the Mammoth Steppe Ecosystem, which was the most extensive biome on Earth during the Last Glacial Maximum [20]. Previous reconstructions of Beringia [9,19] have shown similarities in main ecosystem contexts (soil and grassland compositions) to the Kluane Lake area [21].

Study areas: Kluane Lake and Whitehorse, Yukon Territory
The two main regions investigated in this study are located in the southern Yukon Territory: (i) the eastern shoreline of Kluane Lake, and (ii) the Whitehorse area (Fig 1). The first site consists of grasslands located next to the southeast shore of Kluane Lake, which receives windblown loess from the Slims River delta. The second site consists of grasslands in the Whitehorse valley. Table 1 lists mean air temperature and total precipitation over the last 29 years at these sites as well as these data for 2012 and 2013, when the plant samples were collected. A few plant samples were also obtained from three sites (S13-13, -14, -15) in the Faro area (Fig 1).
The Kluane Ranges (2000-2800 masl), which are located in the southwest of Yukon Territory, effectively block penetration of Pacific air masses to the Kluane Plateau, resulting in a semiarid continental climate for the Kluane area with cold winters and warm summers [21]. The combination of ice fields at the core of the Kluane Ranges, from which glaciogenic silt and sand are delivered to the Slims River delta, strong winds off of these glaciers, and arid conditions, which amplify evapotranspiration, facilitate continuous transportation and accumulation of loess on the eastern side of Kluane Lake. Such conditions are similar to those reconstructed by Guthrie [22] for eastern Beringia loess formation during the Late Pleistocene.
Loess accumulation in this area occurred during two time periods: Late Pleistocene/early Holocene and Neoglacial [23]. The region still experiences frequent dust storms, particularly during summer months. In most soils from this area, the two loess phases are separated by a reddish-brown paleosol (Fig 2), which is named the "Slims soil" [24].
The vegetation of the Kluane region comprises a mixture of grassland and boreal forest ranging from valley-bottom elevations (781 m) to~1160 m [25]. The grasslands are composed mainly of Artemisia-Festuca communities, which are predominant in drier parts of the area on southwest-facing aspects. The forests consist mainly of white spruce [21].
The Whitehorse valley, like most of Yukon Territory, has a dry, subarctic climate characterized by long and cold winters and short and cool summers. The long-term records (Table 1) show a west to east difference in mean total precipitation, with Whitehorse receiving significantly more precipitation than Kluane Lake. The location of the City of Whitehorse in this valley makes its climate milder than other areas of the Yukon ( Table 1). The vegetation is more or less similar to the Kluane Lake area, consisting mainly of boreal forest and grasslands. Shrub communities are also present near the tree line, under a canopy of trees. Grasslands are limited to dry and south-facing slopes, while forests cover many plateaus and valleys [26].
1.2 Carbon isotopic composition of soil and terrestrial plants 1.2.1 δ 13 C of soil organic carbon. Vegetation following different photosynthesis pathways (C 3 , C 4 and CAM) imparts different δ 13 C signatures to soil organic carbon (SOC). This difference is the basis for many studies of past vegetation and climate change [27][28][29][30]. Other processes also affect the δ 13 C of SOC. A negative relationship has been observed between soil δ 13 C OC and mean annual precipitation (MAP) in C 3 -dominated ecosystems [28]. This relationship is explained by plant C isotopic response to MAP, which is transmitted to soils [31]. Discrimination against 13 C during microbial decomposition also affects soil δ 13 C OC . This effect depends mostly on the extent of decomposition, which is controlled by climate [32]. Increases in δ 13 C OC with depth are typically attributed to microbial degradation when <4 ‰ [33,34], and to C 3 /C 4 vegetation change when >4 ‰ [35]. 1.2.2 Controls on plant δ 13 C. Among vascular plants, C 3 plants are characterized by the lowest δ 13 C (-38 to -22 ‰), C 4 plants have higher δ 13 C (-21 to -9 ‰) and CAM (Crassulacean Acid Metabolism) plants lie between (-30 to -13 ‰) [36][37][38]. C 3 plants are overwhelmingly dominant in high latitudes [39,14], including our study sites (above 60ᵒN).
Other factors affecting plant δ 13 C include the δ 13 C of source CO 2 , relative humidity/water availability, temperature, light intensity and partial pressure of CO 2 (pCO 2 ) [31,36,37,40,41]. Water availability is most important. A negative correlation between MAP and C 3 plant δ 13 C [41][42][43][44] is attributed to stomatal closure in response to aridity, producing reduced C i /C a , higher water use efficiency and less negative δ 13 C [36][37][38]. Understory growth under closed tree canopies-the "canopy effect"-can lead to more negative plant δ 13 C due to fixing of 13 Cdepleted CO 2 from soil and canopy respiration [45], lower light intensity and higher pCO 2 [45]. No samples in the present study, however, grew under such conditions.
In C 3 plants, photosynthesizing tissues (e.g. leaf) have more negative δ 13 C than heterotrophic tissues (e.g. stem, root, inflorescence) [46]. Causes include different macromolecular tissue compositions, growing stage variations in photosynthetic discrimination against 13 C, and different contributions of day versus night sucrose with different δ 13 C to different tissues [46]. (a) profile S13-6, and (b) profile S13-8. In both, data points provide average values for bulk analysis of each soil interval (indicated by the vertical lines).The reddish-brown layer is called the "Slims soil", and separates Neoglacial loess deposits from underlying Late Pleistocene/early Holocene deposits (photographic credit: Tessa Plint). ) and dissolved organic compounds (e.g. simple proteins, amino acids and amino sugars) [47][48][49] are the most common forms of soil N taken up by plants. The δ 15 N varies among these sources because of different biochemical reactions in soil (N mineralization, nitrification, denitrification and volatilization) [50,51]. Bulk soil δ 15 N, however, is not always a good representation of bioavailable N [50]. Adding to the complexity is the δ 15 N gradient with soil depth [49,52]. Increasing bulk soil δ 15 N with soil depth [53][54][55] has been attributed to fresh litter input to topsoil, coupled with accumulation of 15 N-enriched, decomposing organic matter with depth [54]. Plants have diverse abilities to acquire N depending on their rooting depth and phenology [55], life form (trees, shrubs and herbs) [17], and preference for different forms of N [16,56] at different times of year [57].
Mycorrhizal fungi associations also affect the δ 15 N of source N for plants. Plants are likely more reliant on mycorrhizal fungi for N acquisition under conditions of low N availability [11], which commonly is the case in Arctic and subarctic ecosystems [58]. Arbuscular mycorrhizal-, ectomycorrhizal-and ericoid mycorrhizal-associated plants are depleted of 15 N by 2 ‰, 3.2 ‰ and 5.9 ‰, respectively, relative to non-mycorrhizal plants [11,[59][60][61].

Controls on plant δ 15 N.
While nitrogen in plants mainly originates from soil, plant δ 15 N varies from total soil N [12]. This reflects the range of bioavailable versus non-bioavailable N compounds in bulk soil N, isotopic fractionation during N uptake by plants, and biological processes during N assimilation [12,50].

Nitrogen Isotopic Fractionation during N Uptake and Assimilation.
The size of nitrogen isotope fractionation (ε) during plant uptake of NO 3 and NH 4 + is controlled by two factors: (i) external N concentration [72,73], and (ii) efflux of 15 N-enriched inorganic N and/ or 15 N-depleted organic N from roots after N uptake [72,74]. Values of ε are small during plant uptake under low concentrations of NO 3 and/or NH 4 + (~0.5 mol m 3-); ε increases at higher concentrations [75,76]. Discrimination against 15 N during plant uptake is negligible under most natural conditions [72,77]. Enzymatic reactions during N assimilation typically produce large fractionations; ε of +15 ‰ and +17 ‰, respectively, have been reported for nitrate reductase and glutamine synthetase, two main enzymes involved in N assimilation [72].

Intra-plant Variation in δ 15 N.
Intra-plant variation in δ 15 N can arise from: (i) variation in plant nitrogen sources as different organs form and expand, (ii) different patterns of N assimilation with either NO 3 or NH 4 + as the primary N source, (iii) reallocation and transportation of N macromolecules between sink and source organs, and (iv) organ-specific efflux of N [72]. Leaves normally have higher δ 15 N than other organs, particularly roots [72,78], although there are some exceptions [79]. When NO 3 is the sole N source, significant intraplant variation occurs, with leaves having much higher δ 15 N than roots. This likely reflects different patterns of NO 3 vs. NH 4 + assimilation. Whereas NH 4 + is assimilated immediately after root uptake, some NO 3 is assimilated in roots while the remaining, 15 N-enriched NO 3 is transported to shoots for N assimilation in leaves [72]. Enzymatic reactions involved in reallocation of N also can produce molecules with lower δ 15 N than the original source and thus cause intra-plant δ 15 N variation [72]. Likewise, loss of NH 3 through plant leaves and efflux of organic N from roots can enrich these organs in 15 N [50, 72].

Environmental Factors and Plant δ 15 N.
A decrease in soil and plant δ 15 N generally accompanies increasing MAP and decreasing mean annual temperature (MAT) [11,80] for MAT ! −0.5˚C. This pattern may be related to changes in the rate and nature of soil and plant N cycling and/or dependence on mycorrhizal association [11]. Changes in amount of rainfall and soil water availability can affect the openness of the N cycle [81]. A more open N cycle in drier sites probably reflects greater N availability because of lower plant N demand; this can stimulate NH 4 + volatilization, leading to higher soil and plant δ 15 N [49,81]. Lower N availability [82] and greater plant reliance on mycorrhizal association for N acquisition [11] can contribute to lower soil and plant δ 15 N in wetter ecosystems. In short, N cycling in ecosystems is highly responsive to climatic factors, and the associated changes in plant δ 15 N can be traced from primary producers to consumers [6,83].

Sample collection and preparation
A total of 79 terrestrial plant samples and 15 soil samples (8 topsoil and 7 subsoil including two soil profiles) were collected during September and August 2012 and 2013 (Fig 1) with permission of the Government of Yukon and agreement of Yukon First Nations (Licenses No. 13-52S&E and No. 14-46S&E). Sample collection and field work did not involve any endangered or protected species. Plant and soil samples were placed in woven poly bags and plastic bags, respectively. At sites S13-8 and S13-10, several topsoil samples were collected in response to observed differences in soil texture and topography (shallow vs. steep slopes). All plant samples were air-dried and separated into different plant parts. These plant tissues were washed with distilled water (DW) and dried at 90˚C overnight. The dried plant materials were then ground to a very fine powder using a Crescent Wig-L-Bug and stored in small, sealed glass vials for N and C elemental and isotopic measurements.
All soil samples were air-dried, sieved (<2 mm), ground gently using a metal mortar and pestle, and then stored in plastic containers. Analysis of soil samples for physical and chemical properties (mineral fraction, OM fraction, pH, mineralogy) followed standard methods (see Supporting Information: Section A in S1 Text).
Two methods, (i) acid fumigation [84], and (ii) acid rinsing [85], were used to remove carbonates from soil samples prior to elemental and isotopic analyses of OC. Untreated soil samples were used to determine total nitrogen (TN), total carbon (TC), and δ 15 N [84].

Elemental analysis
Procedures used for elemental analysis are described in Section B of S1 Text.

Stable isotope analyses
All C and N isotopic results are presented using δ-notation [86]: where R Sa and R Std denote 13 C/ 12 C or 15 N/ 14 N of the sample and standard for δ 13 C and δ 15 N, respectively. The δ-values of all samples were calibrated to VPDB (carbon) and AIR (nitrogen) using USGS40 and USGS41 [87,88].
The δ 13 C and δ 15 N of plant samples, soil OC (after carbonate removal), and soil TN were measured by dry combustion using an EA (Costech Analytical Technologies, Valencia, CA, USA) coupled in continuous flow mode to either a Thermo Scientific Delta PLUS XL or a Thermo Scientific Delta V PLUS IRMS (Thermo Scientific Bremen, Germany). Because of the low N content of the plant samples, nitrogen isotopic analysis was performed in a separate analytical session from that of carbon; CO 2 generated in the latter sessions was scrubbed from samples using a Carbo-Sorb trap on the EA. Separate analytical sessions were also used to obtain δ 15 N for soil TN, using un-acidified samples.

Statistical analysis
All plant samples were categorized into two main functional groups: herbs (including annual and perennial grasses, forbs and sedges), and shrub/subshrubs. An independent-sample t-test was used to test for differences in the C and N isotopic and elemental compositions between: (i) different plant functional groups, and (ii) samples from two sampling years (2012 and 2013). Comparisons of C and N isotopic and elemental compositions for (i) different plant parts, and (ii) plants from different sampling sites were performed using one-way ANOVA followed by means comparison using either Tukey's HSD test, if variance was homogeneous, or Dunnett'sT3 test, if variance was not homogeneous. Assessments of correlation between (i) plant δ 13 C and δ 15 N, (ii) plant N content and δ 15 N, and (iii) soil δ 13 C OC and OC in soil profiles were performed using Pearson correlation coefficient. All statistical analyses were performed in SPSS 20.

Soils
General information for each site sampled is presented in Table 2.
Seven topsoils, two soil profiles and one topsoil sample of the loess source area (Slims River delta) were analyzed for basic physical and chemical properties (see Supporting Information: Section C and Tables A and B in S1 Text), and OC and TN isotopic compositions (Table 3).
All topsoils, except for S13-9, are dominated by silt (avg. 48.9 ± 16.5 wt. %, all ± errors reported hereafter are one standard deviation (SD)). The mineralogy of most soil samples is similar to that of Slims River deltaic sediment, which is representative of the sources of windblown deposits in the area (see Supporting Information: Table B in S1 Text).
The total nitrogen isotopic compositions (δ 15 N TN ) of the topsoils range from +2.1 to +5.5 ‰. The δ 13 C OC results derived from both types of pretreatment to remove carbonate are very similar except for samples S13-9 and Slims River for which acid fumigation showed more efficacy (Table 3). Accordingly, only results produced using acid fumigation are considered further. The range of δ 13 C OC (-25.2 to -24.5 ‰) obtained for topsoils is much smaller than measured for δ 15 N TN , and is characteristic of C 3 vegetation. Slims River sediment, by comparison, has δ 15 N TN and δ 13 C OC of +1.6 ‰ and -20.6 ‰, respectively.
The δ 13 C OC of profile S13-6 increases with depth from -24.5 ‰ for topsoil to -22.5 ‰ in the subsoil (Fig 2a). The change in pH with depth in profile S13-6 strongly correlates with inorganic carbon (IC) content, which was calculated by subtracting OC from TC (r = 0.960, p <0.01). The δ 15 N TN in profile S13-6 also show a positive shift from +4.3 to +6.7 ‰ with increasing depth from topsoil to 60 cm, but then decreases to +4.1 ‰ between 60-70 cm   Fig 2a). A positive, albeit smaller, shift in both δ 13 C OC (-25.0 to -24.4 ‰) and δ 15 N TN (+4.6 to +5.3 ‰) with increasing depth is also observed for the second soil profile (S13-8) (Fig 2b).  arctica) were sampled from 18 sites distributed for the most part just east of Kluane Lake, but also including some locations near the City of Whitehorse and in the Faro area, Yukon Territory (Fig 1, Table 2). The samples represent two main functional groups: herbs (including both annual and perennial grasses, forbs and sedges) (n = 66) and shrub/subshrubs (n = 13) [89]. Subshrubs are smaller than shrubs, but still have a woody base and bushy shape; their soft stems die back during cold seasons.

Plants
The δ 13 C and δ 15 N of leaves (L) and stems (S) were measured for all samples, with additional plant parts analyzed as available (70 root crowns (RC), 68 fine roots (FR), 73 inflorescences (I)) (see Supporting Information: S1 Table). In this study, the root crown is considered as the top part of the root system in herbs and subshrubs. Subshrubs and perennial grasses regrow each spring from buds produced by the root crown. In the herbs, the major vascular changes required for formation of new stems in spring occur in the root crown [90].

Kluane Lake Soils
All topsoil samples examined along the eastern shoreline of Kluane Lake are rich in silty eolian sediment that likely originated from the Slims River delta [21,24] except for S13-9, which contains a larger abundance of sand (Supporting Information: Table A in S1 Text). Its higher sand content likely reflects a larger contribution of underlying sandy glaciofluvial deposits resulting from bioturbation or post-fire redistribution [24]. The soil mineralogy is similar to the Slims River sediment, consistent with an eolian source (Supporting Information: Table B in S1 Text). The average δ 13 C OC (-24.8 ± 0.3 ‰) of all topsoils is typical of soil organic matter generated by C 3 vegetation, except for the higher δ 13 C OC (-20.6 ‰) of the Slims River deltaic sediment. This site (Fig 6), which is one of the two main outlets of the Kaskawulsh glacier [92], is subject to flooding and seasonal and diurnal fluctuations in the water flow, and supports algae and macrophytes growth at low flow stages. Aquatic plants in Yukon Territory have a wide range of foliar δ 13 C OC (−41 to −15 ‰), particularly for submerged macrophytes [93]. Less negative δ 13 C OC (−16 to −13 ‰) for aquatic plants and algae have also been reported for Arctic continental shelf sediments [94]. Contributions of CAM or C 4 plants at modern, northern high-latitude site like this one are unlikely [39]. Incomplete removal of carbonate can also be ruled out based on the acid fumigation results (Table 3).
In general, the topsoil samples have higher average δ 15 N TN and δ 13 C OC than plants from the same sites (Fig 7), which is typical of many other terrestrial ecosystems [16,71]. This can be explained by the general enrichment in 13 C and 15 N of plant tissues in soil during organic degradation [12,32]. The larger isotopic variation among the foliar isotopic compositions relative to topsoil samples likely reflects differences in plant sample size, number of plant species sampled, and the age of plants at different sites. In particular, enzymatic variability among different plant species and different patterns of N acquisition among co-occurring species [72] serves to increase the range of foliar δ 15 N among sites and between plants within sites and their topsoil. In contrast, the continuous activities of soil decomposers (mainly fungi and bacteria), the strong role of soil minerals in stabilizing soil chemistry [95] and the open N dynamics of soil input and output reactions reduce N isotopic variations in topsoil from site to site within the same general region.
The increase in δ 13 C OC with depth in two soil profiles (Fig 2a and 2b) is in agreement with earlier results for subarctic ecosystems [27,96]. This pattern likely reflects the input of fresh litter with lower δ 13 C at the soil surface and accumulation of decomposed and hence more 13 Crich OM at greater depths [27,33]. This interpretation is consistent with the strong negative correlation between δ 13 C OC and OC content (r = −0.950, p <0.05) and the decrease in OC content with depth (Fig 2a and 2b).
The change in δ 15 N TN with depth observed in both soil profiles (Fig 2a and 2b) has been reported previously for such soils from Siberia [27]. The negative shift in δ 15 N TN below 60 cm in profile S13-6 may arise from the very small amount of OM (2.9 wt. %) and TN (0.1 wt. %) at this depth, in which case the contribution of inorganic nitrogen to the δ 15 N TN signal may be more important at this depth. In soils, organic N typically has higher δ 15 N (+5 to +7 ‰) than inorganic forms (NO 3 and NH 4 + ) (−2 to +5 ‰) [12,56]. Values of δ 15 N TN are positive in all topsoil samples, which is typical of alpine and tundra ecosystems [27,53]. Soil δ 15 N TN >0 ‰ point to inputs with compositions higher than those produced by fixation of nitrogen from air and/or N-loss processes that leave the soil N pool enriched in 15 N [97]. Two samples (S13-9, S13-10-2) sit at the lowest end of the N isotopic range (+2.1 to +5.5 ‰). S13-9 has the highest sand and lowest clay and silt contents of the soils examined in this study. The content, structure and function of OM associated with mineral particles generally vary with grain size. Sand-sized particles are typically associated with less humified and lower abundances of OM than silt and clay [95]. Several studies of soils underlying grasslands or forest have reported higher δ 15 N TN for clay-sized fractions (~+9 to +12 ‰), which generally have a higher content of stable, humified OM than silt (+5 to +9 ‰) and sand (+2 to +7 ‰) particles [98][99][100][101].
Topography may explain the low δ 15 N TN (+2.5 ‰) of sample S13-10-2 at site S13-10. Sample S13-10-2 was collected from top of a steep slope, which made it more susceptible to erosional disturbance, while sample S13-10-1 (δ 15 N TN = +5.5 ‰) (from the same site) was collected at the bottom of the slope from a flat and more stable location. Steeply sloping soils can have δ 15 N close to atmospheric inputs because of continuous soil removal and soil organic matter rejuvenation, which maintains the soil's N status far from steady state [80].

Plant C and N isotopic and elemental compositions
The range of δ 13 C (−32.5 to −23.5 ‰) measured for all plant parts in this study (Fig 8) is typical of C 3 vegetation, which dominates high latitude ecosystems [39]. Variation in environmental factors (slope aspect, light, water availability and topography) even within small microhabitats may cause this large spread in δ 13 C. Wooller et al. [14] reported a very similar range of foliar δ 13 C for sedges and grasses from Alaska and Yukon Territory.
The plants analyzed here show great variation in δ 15 N, spanning~40 ‰ (Fig 8). Variation in δ 15 N of vegetation tends to be more pronounced in N-limited ecosystems, which is typical N and C isotopic dynamics of subarctic soils and plants in southern Yukon Territory of Arctic and subarctic regions, and points to utilization of different soil N resources by plants depending on their life forms, type of mycorrhizal association and rooting depth and morphology [16]. Coexisting plant species are known to partition N resources with different δ 15 N in these ecosystems [16,17,56,102]. For example, the grass Calamagrostis canadensis (δ 15 N = +0.9 ‰) in Alaska acquires N from deeper soil horizons, while the evergreens Picea glauca and Picea mariana (−7.7 ‰) likely utilize ammonium or organic N from fresh litter [17].
There is no statistically significant difference in foliar δ 13 C and δ 15 N between herbs and shrubs/subshrubs in this study. Considering that all plants studied here utilize the C 3 photosynthetic pathway, the lack of any clear distinction in foliar δ 13 C is not surprising. The absence of systematic differences in δ 15 N between these two groupings is less expected, given potential differences in N conservation and resorption efficiency [103] and patterns of root distribution and rooting depth, which affect resource acquisition [104]. The absence of a statistically meaningful difference in δ 15 N might arise from the unequal sample size of herbs vs. shrubs/subshrubs or confounding environmental factors such as plant growth stage, the range of available N sources, and the varied environmental conditions in the study area.
The shrubs and subshrubs have statistically significantly higher C and N contents than the herbs at both foliar and whole plant levels, as has been observed previously, especially for N https://doi.org/10.1371/journal.pone.0183016.g008 [104]. The difference in N content measured here might be related to different N conservation strategies during the late growing season. Arid and semi-arid perennial grasses have lower N contents in senesced leaves and overall higher N conservation efficiency than shrubs, which is an important adaptive trait for plants from nutrient-limited ecosystems [103]. Plants in this study were sampled in very late growing season (mid-late September in 2012 and late August in 2013), with perennial grasses comprising most herb samples. Hence, the difference in N contents may indicate higher N resorption efficiency from senescing tissues in grasses than in shrubs.
There are likely four main environmental factors that control the isotopic compositions of the herbaceous plants, shrubs and subshrubs sampled in this study: (i) water availability, (ii) N availability, (iii) spacial differences, and (iv) intra-plant variation. Each factor is discussed next.
4.2.1 Water availability. The very weak, but statistically significant positive correlation between foliar δ 15 N and δ 13 C of all samples analyzed (Fig 4) may point to water availability as at least a minor factor affecting both δ 15 N and δ 13 C of plants in this ecosystem. Plants capture atmospheric CO 2 through leaf stomata and fix it using enzymatic reactions, while N is mainly obtained through roots from soil or symbiotic associations and then assimilated. Given that the sources and pathways determining the δ 13 C and δ 15 N of plants are different, the observed, albeit weak, correlation may indicate a common environmental factor affecting both isotopic signals. As noted earlier, a change in MAP can affect both δ 13 C and δ 15 N of plants in the same direction (N: [11,54,80]; C: [42,105,106]). Ma et al. [43], for example, have reported an aridity-associated positive correlation between δ 13 C and δ 15 N in plants from northern China and noted that plant isotopic sensitivity to water availability can vary among ecosystems and plant species. Such variations warrant further consideration in paleoecological and paleodietary reconstructions of Arctic and subarctic regions.

N availability.
The positive correlation between foliar δ 15 N and N content suggests a key role for N availability and N cycling in determining the N isotopic signal acquired by plants in this ecosystem. Such a correlation has been reported previously on local [107], regional [54,108] and global [11] scales. Higher plant δ 15 N reflects higher N availability and a more open N cycle in terrestrial ecosystems [107,[109][110][111]. The globally observed positive correlation between soil and foliar δ 15 N suggest foliar δ 15 N as an index for N availability in soils and therefore ecosystems [11]. Nonetheless, the best way to describe N availability for plants in different ecosystems remains unclear [11]. It can be defined in several ways including: (i) an increase in N inputs into the soil from different sources (animal dung, plant materials, microbial N fixation), (ii) increased OM decomposition and N mineralization, (iii) increased NO 3 − production through more nitrification [111], and (iv) less N demand by plants, particularly in drier localities [81]. In any of these scenarios, higher inorganic N availability in soils means that extra N is available to fuel N loss processes (e.g. denitrification and volatilization), which leave the bioavailable N in soils (NO 3 − , NH 4 + ) enriched in 15 N [49,50] (Fig 9a). In such ecosystems, plants acquire both higher δ 15 N and N content in their leaves, which is characteristic of a more open N cycle. In N-limited ecosystems, by comparison, there is less N-bearing material available for N loss, and hence less opportunity for 15 N enrichment of the system through such processes. Plants also rely more heavily on mycorrhizal fungi for N acquisition, which is a more 15 N-depleted source [11,58] (Fig 9b).

Spacial differences.
The statistically significant differences observed in foliar δ 15 N among sites S13-2, S13-3, S13-7 and S13-6 points to heterogeneity in this ecosystem and the presence of different microhabitats even at small scales. These differences may be related to local variations in soil properties, slope aspects, topography, water availability, animal disturbance and grazing, which in turn can affect local N cycling. While we might predict higher foliar δ 15 N at S13-7 than S13-6 given the former's higher topsoil OM content (9.2 vs. 7.5 wt. %) and TN (0.4 vs. 0.2 wt. %), the opposite result is obtained (S13-7-avg. δ 15 N = −2.0 ‰, n = 6) vs. (S13-6-avg. δ 15 N = +1.4 ‰, n = 8). Other factors such as slope aspect and topography may have overshadowed the influence of soil properties. In S13-6, plant samples were collected from north side of the road on a SW-facing shallow slope, while in S13-7, plants were sampled on south side of the road on a NE-facing, shallow slope. This difference influenced the microclimate and vegetation pattern [24], as is indicated by the dominance of subshrubs at S13-7 and herbs at S13-6.

Intra-plant variations.
The average Δ 13 C other plant part-leaf measured in this study are positive, consistent with previous studies [46,113]. The differences measured here for most individual specimens, however, are small (<1 ‰) ( Supporting Information, S1 Fig), and should not affect the use of δ 13 C to infer diet for animals that may prefer one plant part over another as forage.
No clear pattern in Δ 15 N other plant part-leaf was observed, likely because of confounding factors such as differences in root morphology, depth and distribution, microhabitat, mycorrhizal associations and growing stage. The root crowns examined here have higher N contents ( Fig  5) and δ 15 N (Fig 3) than other plant parts; this difference is statistically significant for N content, but not for δ 15 N. As discussed earlier, nitrogen contents of different plant parts commonly show variations between active growing and senescent stages in herbaceous plants [114,115]. The higher N content of the root crowns in this study is best explained by N re-allocation from leaves to below ground parts late in the growing season.

Summary and implications
All plants analyzed from this ecosystem follow the C 3 photosynthetic pathway, and have an average whole plant δ 13 C of −27.5 ± 1.2 ‰ and foliar δ 13 C of -28.0 ± 1.3 ‰. The plants analyzed here showed very little variability in δ 13 C among different plant parts and sampling sites. Their isotopic composition, with suitable correction for the Suess effect arising from fossil fuel combustion since the Late Pleistocene, provides a suitable baseline for interpreting the diet of ancient herbivores that lived in these grasslands.
The nitrogen isotopic data for these modern plants (average whole plant δ 15 N = −0.3 ± 2.2 ‰ and foliar = -0.6 ± 2.7 ‰) provide a good baseline for the region's vegetation under present conditions of nitrogen cycling. The wide range of intra-plant and inter-plant δ 15 N variations likely arises from multiple factors, including water availability, N availability and topography. Such variation is likely typical of vegetation from different microhabitats within Arctic and subarctic ecosystems. Given this complexity, modern vegetation δ 15 N is unlikely to adequately approximate Late Pleistocene vegetation in the same region.
A growing body of studies utilizes the isotopic composition of bone collagen and other tissues from Late Pleistocene megaherbivores in high latitudes to reconstruct their diet. These works are hampered by insufficient isotopic data for plants at the base of food web in these now vanished ecosystems. More data for both modern and well-dated fossil plants from these ecosystems are needed to fully document and understand the variations in plant isotopic baselines both through time and space. The present study provides a starting point for modern vegetation in south and central Yukon Territory, Canada. While the plant N isotopic compositions presented here cannot be applied directly to the past, it provides a baseline for comparison with its ancient equivalent, which can be obtained by analyzing fossil plants from the same region. We note that the significant positive correlation between foliar δ 15 N and N content observed for this region has potential as an index for forage quality. Comparison of the modern N isotopic vegetation baseline for this region with its Late Pleistocene equivalent could serve as a proxy for tracing changes in forage quality, which is tightly connected to ecosystem productivity and potentially, the sustainability of megaherbivore populations.
Supporting information S1 Text. Section A. Methods for analyzing physical and chemical properties of soil. Section B. Analysis of carbon and nitrogen content. Section C. Results of soil analysis.