Nutrient foraging strategies are associated with productivity and population growth in forest shrubs

Abstract

Background and Aims Temperate deciduous forest understoreys are experiencing widespread changes in community composition, concurrent with increases in rates of nitrogen supply. These shifts in plant abundance may be driven by interspecific differences in nutrient foraging (i.e. conservative vs. acquisitive strategies) and, thus, adaptation to contemporary nutrient loading conditions. This study sought to determine if interspecific differences in nutrient foraging could help explain patterns of shrub success and decline in eastern North American forests.

Methods Using plants grown in a common garden, fine root traits associated with nutrient foraging were measured for six shrub species. Traits included the mean and skewness of the root diameter distribution, specific root length (SRL), C:N ratio, root tissue density, arbuscular mycorrhizal colonization and foraging precision. Above- and below-ground productivity were also determined for the same plants, and population growth rates were estimated using data from a long-term study of community dynamics. Root traits were compared among species and associations among root traits, measures of productivity and rates of population growth were evaluated.

Key Results Species fell into groups having thick or thin root forms, which correspond to conservative vs. acquisitive nutrient foraging strategies. Interspecific variation in root morphology and tissue construction correlated with measures of productivity and rates of cover expansion. Of the four species with acquisitive traits, three were introduced species that have become invasive in recent decades, and the fourth was a weedy native. In contrast, the two species with conservative traits were historically dominant shrubs that have declined in abundance in eastern North American forests.

Conclusions In forest understoreys of eastern North America, elevated nutrient availability may impose a filter on species success in addition to above-ground processes such as herbivory and overstorey canopy conditions. Shrubs that have root traits associated with rapid uptake of soil nutrients may be more likely to increase in abundance, while species without such traits may be less likely to keep pace with more productive species.

INTRODUCTION

Deciduous forests in temperate regions of eastern North America (ENA) and elsewhere are undergoing dramatic changes in understorey plant composition and abundance. For example, numerous shrub species are invading successional and mature ENA forests, despite the wide range in light availability of these systems (Meiners et al., 2002; Martin et al., 2009). Several factors are known to drive plant community change in ENA understoreys, including legacies from historical land use as row-crop agriculture (Meiners et al., 2002; Mosher et al., 2009), soil alteration by exotic earthworms (Bohlen et al., 2004), deposition of nutrients from industrial and agricultural sources (Aber et al., 2003), extensive herbivory of unarmed species by white-tailed deer (Beguin et al., 2011; Begley-Miller et al., 2014) and physical disturbance to overstorey canopies (Nuttle et al., 2013). Although there are unique features to each of these processes, many increase the rate of nitrogen (N) supply to forest soils (van Breemen et al., 2002; Aber et al., 2003). An array of previous studies has established that widespread N enrichment is altering understorey communities in the ENA and elsewhere (Gilliam, 2006; Köchy and Bråkenhielm, 2008; Bobbink et al., 2010; Verheyen et al., 2012; Clark et al., 2013), but these studies have not addressed the possibility that increased N loading has altered the adaptive value of strategies for nutrient foraging.

Interspecific differences in nutrient foraging could influence adaptation to soil N enrichment, just as differences in leaf phenology and gas exchange influence adaptation to light regimes (Fridley, 2012; Heberling and Fridley, 2013). Although nutrient and light foraging strategies appear to be broadly coupled across species (Reich, 2014), the fact that N enrichment can occur in settings with a wide range of light regimes suggests that an ability to capitalize on surplus N could be a critical adaptation for both shade-tolerant and shade-intolerant species in contemporary ENA forests. Foraging strategies could likewise influence adaptation to N enrichment for species of different origins, i.e. native vs. introduced to ENA. It is particularly likely that aggressive introduced species (we use the term ‘invasive’ strictly for this group) tend to acquire N rapidly (Keser et al., 2014; Smith et al., 2014; Jo et al., 2015). However, native species and non-aggressive introduced species probably exhibit a wide range of nutrient foraging strategies such that there is substantial overlap with the invasive group (Leishman et al., 2010; Dawson et al., 2012; Caplan and Yeakley, 2013). Any advantages or disadvantages in performance (e.g. productivity, reproductive output or spread) conferred by nutrient foraging would therefore be fundamentally independent of origin (Davis et al., 2011).

Several recent studies have demonstrated that below-ground foraging strategies vary according to a trade-off between the longevity of fine roots and the return time on the investment in root tissue (Kong et al., 2014; Roumet et al., 2016). This manifests as an economics spectrum spanning acquisitive to conservative resource acquisition (also referred to as ‘fast’ to ‘slow’; Reich, 2014), much like that of leaves (Wright et al., 2004). Under the lower N loading rates that historically characterized ENA forests, plant productivity may have been lower at the ecosystem scale (Thomas et al., 2010), but maximized for species with relatively conservative root economics. Species adapted to growing under such conditions build roots that are more resistant to herbivory, decay and environmental stress; they therefore have greater longevity and can forage for nutrients and water over their longer life spans (Roumet et al., 2016). While their rates of nutrient uptake are slower (Eissenstat, 1992; Reich et al., 1998), this may be partly compensated for by a greater reliance on arbuscular mycorrhizal fungi to acquire nutrients (Kong et al., 2014; Eissenstat et al., 2015). However, under the N-enriched conditions characteristic of many contemporary ENA forests, the strategy maximizing performance may have shifted in favour of N acquisitive species. Such species absorb nutrients more rapidly, generate more absorptive area per unit soil volume, and grow roots into patches highly preferentially (i.e. have high foraging precision; Eissenstat, 1992; Yanai et al., 1995; Hodge, 2004). Of course, the trade-off is that their roots tend to have lower longevity (McCormack et al., 2012). Although phenotypic plasticity enables species to adjust their tissue construction to a degree, natural selection constrains species to a limited range of the root economics spectrum (Reich, 2014). We suggest that changes in ENA forest understorey communities could be due, at least in part, to species with a capacity for more rapid nutrient acquisition gaining advantages in performance (especially in organismal-level productivity and spread) under elevated N loading conditions, to the detriment of species with slower nutrient acquisition.

Nutrient foraging strategies are associated with fine root morphology, tissue chemistry and mycorrhizal associations in multiple plant forms (Roumet et al., 2016). Because the relationships are particularly well established for woody species (Kong et al., 2014; Eissenstat et al., 2015), it is possible to determine nutrient foraging strategies for ENA shrubs by characterizing syndromes of root traits (Iversen, 2014). The morphological traits most strongly indicative of conservative vs. acquisitive foraging are specific root length (SRL; the ratio of root length to dry mass) and mean root diameter. These traits are closely associated with fine root life span among ENA trees via negative and positive relationships, respectively (McCormack et al., 2012). SRL is also known to correlate (positively) with N uptake rates among tree species (Reich et al., 1998). Root tissue density (RTD) has been correlated with root life span (Ryser, 1996), but predicts life span less reliably than SRL or diameter (Huang et al., 2010; McCormack et al., 2012). Among chemical traits, root tissue N (expressed as a concentration or C:N ratio) is arguably the most readily measured and useful indicator of conservative vs. acquisitive foraging; it is associated with metabolic rates (e.g. respiration) and longevity, and is correlated with the above morphological traits (Reich, 2014; Roumet et al., 2016). In woody plants that form arbuscular mycorrhizae, dependence on fungi for supplying nutrients can be determined from the extent of mycorrhizal colonization (Eissenstat et al., 2015).

There is some indirect evidence suggesting that differences in foraging strategies have led to differences in productivity and spread among ENA forest shrubs. For example, several studies have identified interspecific differences in growth responses to N or organic matter addition (Cassidy et al., 2004; Gurevitch et al., 2008; Elgersma et al., 2012). Other studies have demonstrated that there are differences in functional traits among ENA forest shrubs, with root traits included in a few studies (Smith et al., 2014; Jo et al., 2015), but the vast majority focusing on above-ground traits. While above-ground traits such as leaf N are easily measured and provide some information on below-ground economics (Katabuchi et al., 2012; Laliberté et al., 2012), evaluating nutrient foraging via root traits directly is likely to yield greater insight (Laughlin, 2014). Further, root traits are known to correlate with organismal-level growth rates among ENA trees (Comas and Eissenstat, 2004). However, the few previous studies of ENA forest shrubs have established neither if root foraging strategies are associated with organismal-level productivity nor if they influence patterns of population growth.

Our objectives were 3-fold. Firstly, we sought to determine how widely ENA forest shrubs range in their nutrient foraging strategies. We used fine root traits to infer the foraging strategies of six species grown in a common garden, expecting species to range from highly conservative to highly acquisitive. Secondly, we sought to determine if interspecific variation in root traits is associated with organismal-level productivity in understorey shrubs. We used additional data from the common garden for this evaluation, expecting positive correlations between biomass production, both above- and below-ground, with SRL, foraging precision and the skewness of the diameter distribution, but negative correlations with RTD, C:N ratio, mean diameter and mycorrhizal coverage. Thirdly, we sought to determine if variation in root traits and productivity among species is associated with rates of population growth. We expected population growth, measured using data on above-ground cover expansion from a long-term study of community dynamics, to be slower for species with conservative root traits and with lower organismal-level productivity but faster for species with acquisitive root traits and higher productivity.

MATERIALS AND METHODS

Plant material

We selected six woody shrub species for inclusion in the study. Because we were interested in trait variation associated with adaptation to diverse soil N conditions, we focused on species that are highly abundant in ENA forests, or were historically. Although phylogenetically controlled comparisons are useful in determining whether trait expression differs for contrasting species groups (e.g. introduced vs. native species; van Kleunen et al., 2010; Godoy et al., 2011), this approach would not have been tenable for assessing patterns across continua of species traits and performance. We instead chose six species that span a range of rates in expected growth and spread, and that vary in their requirements for light and N to achieve those rates (Table 1). We used a combination of plants grown in the field and purchased from nurseries. To remove legacy effects of plant source, root systems were washed free of soil and roots were clipped such that each plant’s root and shoot system had approximately equal mass. Plants were 2–3 years old when they were acquired in late 2010, and about 1 m tall. Plants were submerged in a dilute bleach solution (0·4 % NaOCl) for 15 min before planting to minimize differences in microbial load on their superficial tissues.

Table 1.

Species included in the study and select characteristics

Species	Common name	Origin	Growth and spread	Light requirement	Nitrogen requirement
Berberis thunbergii	Japanese barberry	Introduced	High	Moderate	High
Lonicera maackii	Amur honeysuckle	Introduced	High	Moderate	High
Rubus phoenicolasius	Wineberry	Introduced	Moderate	High	Moderate
Rubus allegheniensis	Common blackberry	Native	Moderate	High	Moderate
Viburnum dentatum	Arrow-wood viburnum	Native	Low	Low	Low
Lindera benzoin	Spicebush	Native	Low	Low	Low

Rates of growth and spread reflect those achievable under optimal conditions, including the light and nitrogen requirements listed. Ratings are based on the authors’ previous experience with the species.

Common garden

Plants were grown for 1 year in an outdoor common garden. The garden was located at a research farm operated by Rutgers University, in central New Jersey, USA (40·4608°N, 74·4304°W). The region has a temperate climate, with the warmest and coolest months (July and January) averaging 23·7 and –0·8  °C, respectively, and a mean precipitation rate of 116 cm year⁻¹ (Office of the New Jersey State Climatologist). Hurricane Irene delivered record rainfall to the region in August 2011, though no adverse effects on plants were observed. Soil collected from the site in March 2011 had a loamy texture and contained 16 mg N kg⁻¹ as NO₃ + NO₂ and 5 mg N kg⁻¹ as NH₄; total Kjeldahl N was 0·12 % (Rutgers Soil Testing Laboratory). These N concentrations are in the range found currently in forest soils throughout ENA (Small and McCarthy, 2005; Johnson et al., 2014). Additional soil characteristics are provided in Supplementary Data Table S1. Deer were excluded from the site by a 3 m fence.

The common garden was established in a mowed field that had not been used experimentally for several decades. After removing herbaceous plants with glyphosate and tilling surficial soil, four rows (30 × 1·2 m, 3·4 m apart) were established with a weed barrier at ground level and shadecloth hung in a pitched configuration above. Photosynthetic photon flux density (PPFD) under the shadecloth ranged from 34 to 274 μmol m⁻² s⁻¹ (23–31 % of incident radiation) when measured during mid-day in early autumn. This PPFD range is comparable with moderately shaded understorey conditions in deciduous forests (Ricard et al., 2003). Plants were randomly assigned to positions within the garden and planted in the soil in early October 2010, 1·2 m apart (n = 12–13 plants per species). Positions at the ends of rows were left empty to minimize variation in light availability among plants.

Root measurements

Root traits and below-ground production were measured simultaneously from roots generated during the 2011 growing season. Roots were sampled using ingrowth bags installed at the base of each plant in late March (n = 6 per plant). Ingrowth bags were prepared by placing cylindrical columns of soil (10 cm tall, 5 cm diameter) in polyethylene mesh tubes (approx. 6 × 6 mm openings) that conformed to the shape of columns and were sealed at either end. Half of the soil columns consisted strictly of soil collected from the site that had been sieved and picked through for roots and large gravel (n = 3 per plant). The mass of soil used to form columns (195 g) was that needed to match the bulk density of the soil before sieving (1·35 g cm⁻³). To evaluate foraging precision and to determine how root traits and productivity were altered by growth in organic patches, an additional set of columns was prepared with the same soil but mixed with chopped and dried Lolium multiflorum (0·75 % m/m dry, equivalent to a 10 % increase in organic matter; n = 3 per plant). Ingrowth bags were buried in the soil vertically (0–10 cm depth) at the margin of the elliptical region known to contain roots (mean ± s.d. 14·9 ± 4·1 cm from plant bases). Ingrowth bags containing soil with and without additional organic material (hereafter ‘organic patch’ and ‘field soil,’ respectively) were alternatingly positioned around each plant, such that the probability of roots encountering the two types of bags was equal.

All root ingrowth bags were unearthed from eight plants per species (n = 288 bags; six bags from each of 48 plants) in November, approx. 7 months after being buried. Allowing growth to proceed through a full growing season maximized the probability that roots would proliferate into all bags and avoided the potential influence of phenological differences among species (Steinaker et al., 2010). Because some fine roots may have died during the growing season, captured root material reflected net rather than gross production. However, median fine root life spans in ENA woody plants typically range from 200 to 300 d (McCormack et al., 2012; Smith et al., 2014), so it is unlikely that there were large differences in mortality among species. Further, we observed few dead roots upon harvest, although the root systems of two Lindera benzoin plants were highly decayed. This appeared to be caused by Phytophthora infection; all ingrowth bags from both plants (n = 12 bags) were excluded from analysis.

For ingrowth bags containing healthy roots (n = 276 bags), soil was washed from roots by hand. Sieves (≤0·5 mm) were used to maximize the fraction of roots recovered. Small grasses and herbs were occasionally found in bag tops, whereas roots from adjacent shrubs were rare. Extraneous roots could be identified easily based on architecture and colour, and were discarded when found. Roots from each bag were imaged on an Epson 10000XL flatbed scanner (600 dpi) with an overhead light source used to prevent shadows. Roots were subsequently dried at 70  °C for at least 48 h and weighed.

Root length production was measured from images using WinRhizo Pro 2007d (Regent Instruments, Québec, Canada). Total length was determined as the sum of segment lengths in 41 diameter classes; 40 of these spanned 0·05 mm increments up to 2 mm, and the uppermost class contained segments >2 mm. Note that root length was equivalent to root length density because the volume of ingrowth bags was constant. Root surface area and volume in each class were also determined by WinRhizo under the assumption of cylindrical geometry. For each bag, the distribution of length among diameter classes was derived by fitting log-normal distribution functions to cumulative lengths, up to 2 mm (Scanlan and Hinz, 2010); we report mean and skewness parameters subsequently. Skewness was included as an indication of species’ level of investment in the thinnest fine roots. Sub-samples of fine roots (≤2 mm in diameter) were ground and analysed for tissue C and N concentrations (Ecosystem Analysis Lab, Nebraska, USA).

The SRL of fine roots was determined after excluding coarse roots (i.e. diameter >2 mm) from length and mass data. Length was adjusted in each sample by summing values from all but the uppermost diameter class. Mass was adjusted by calculating the fraction of root volume with diameter ≤2 mm, and applying this fraction to the sample’s mass. These adjustments ensured that most morphological trait values reflected only fine roots (≤2 mm in diameter). The exception was RTD, which was calculated from the mass and volume of all roots in each ingrowth bag. However, species ranks for mean RTD were nearly identical between this and a separate data set collected strictly from roots ≤2 mm (J. S. Caplan, unpubl. data). Rankings were also nearly identical for SRL, suggesting that adjustments were successful. Foraging precision was calculated for each plant as the sum of the length (or mass) of roots in the three organic patches divided by the total length (or mass) in all six ingrowth bags. Separate precision values for the two bag types were therefore not available.

Mycorrhizal structures were inventoried in fine root samples of most plants with healthy roots (n = 41 plants). We focused on arbuscular mycorrhizae because all plant genera used in this study are known to form arbuscular mycorrhizae (Wang and Qiu, 2006). A root sample from each type of ingrowth bag was placed in 70 % ethanol prior to drying and was later stained with Trypan blue. The presence or absence of arbuscules, vesicles and hyphae that penetrated plant cells was noted in 150–240 fields of view per species at × 40 magnification (Vierheilig et al., 2005). For most plants, 20 fields of view were evaluated (half from field soil and half from organic patches), but a single plant per species was inventoried using 50–100 fields of view to ensure that data generated with smaller sample numbers were not spatially biased.

Organismal-level productivity

Above-ground production was determined by measuring biomass at the beginning and end of the study. Initial masses (M_i) were calculated indirectly while roots were free of soil in September 2010. This entailed measuring the total fresh mass of each plant from a bottom-loading balance while it was suspended in air and, subsequently, with its root system submersed in water. Given that water was displaced by volume, the difference in masses divided by the density of water (1·0 g cm⁻³) yielded root system volume. For 2–4 plants per species not used in the experiment, we determined volume (as above) as well as dry mass and the water content of both roots and shoots. Mean root densities for each species were then used to convert measured root system volumes to fresh masses. Shoot fresh mass was determined as total fresh mass minus root fresh mass, and finally converted to dry mass (M_i) using species-specific water content values. Final shoot dry mass (M_f) was measured following an above-ground harvest on 14 December 2011. Shoot systems were weighed after drying at 70  °C for at least 72 h. The relative growth rate of each shoot system (RGR_sht) was calculated as RGR_sht = [ln(M_f) – ln(M_i)]/Δt, where Δt was the time between planting and harvest (437 d). Because pants had largely defoliated by the time of harvest and because little growth occurred during winter months, values of RGR_sht are conservative. However, plants also had few leaves when they were weighed initially, presumably due to transplant shock or cold weather.

Expansion in cover

Measurements of population growth for the six focal species were derived from time series data on plant cover from an old field forest in central New Jersey, USA (12 km from the common garden). Plant cover has been monitored annually at this site since 1958 as part of the Buell–Small Succession study (BSS; Meiners et al., 2002). The study includes 480 plots (0·5 × 2·0 m) within a young ENA forest that is typical of a community developing on abandoned agricultural land. Plots were abandoned over an 8 year period, so we aligned data by relative age (i.e. years since abandonment) as done in previous analyses of the data set (Meiners et al., 2002; Yurkonis et al., 2005). Because the sampling design imposes a saw-tooth pattern over the temporal trend in cover (approximately half of the plots are monitored each year), we smoothed raw data using a 5 year moving average.

The rate of population expansion for each species in BSS plots was calculated over the interval spanning its first occurrence until peak cover. Temporal trends were strongly linear during this phase so we used linear regression to characterize increases in cover through time (Supplementary Data Fig. S1). An important caveat with this data set is that the target species spread through the successional forest during different periods of its transition from abandoned field to young forest. However, the plant community present when each species underwent its primary expansion (and, thus, the overstorey canopy cover) probably reflected the community in which the target species commonly establish and spread. Also, propagule availability probably differed among species, such that some species may not have achieved their maximum potential for spread at the BSS site.

Data analysis

Root traits (excluding mycorrhizal coverage) and metrics of productivity were analysed with analysis of variance (ANOVA)-type linear models. Plant-level data (i.e. means of bag-level data) were used for this analysis (n = 46). Predictor variables included species in all cases, though bag type (field soil vs. organic patch) and the species × bag type interaction were also evaluated for most traits. The exception was foraging precision, as the calculation of foraging precision requires both bag types. Likelihood ratio tests were used to determine the significance of each term, with evaluations made for sequential additions of species, bag type and their interaction to the intercept-only model. Transformations of response variables were used to normalize residuals and improve homoscedasticity; log, square root, square and negative square root were evaluated via the Box–Cox procedure (Box and Cox, 1964). Pairwise comparisons among species, bag types and species–bag type combinations, as appropriate, were performed with Tukey’s honestly significant difference (HSD) tests. Wilcoxon tests were used to determine if median foraging precision values for each species were >50 %, i.e. if roots grew preferentially into organic patches.

A logistic mixed effects model, using the logit link function, was used to determine if species, bag type or their interaction (all treated as fixed effects) influenced the probability of mycorrhizal occurrence. Plant identity was treated as a random effect to account for multiple samples (i.e. fields of view) being drawn from each of the 41 plants in the data set. The presence or absence of fungal structures (hyphae, vesicles or arbuscules) in a given field of view was used as the response variable. As above, fixed effects were evaluated using likelihood ratio tests, and pairwise comparisons among species were performed with Tukey’s HSD tests.

Principal components analysis (PCA) was used to identify species-level patterns among root traits and among productivity metrics. The PCA of root traits included SRL, the mean and skewness of the diameter distribution, the C:N ratio, RTD and length-based precision. Mycorrhizal coverage was excluded because data were not available for all plants, and mass-based precision was excluded because it was highly redundant with length-based precision. All four productivity metrics were included in the second PCA. Correlation matrices were used in both cases, as measurement scales differed among variables. Broken-stick criteria were used to determine the number of principal components to retain (Jackson, 1993). Associations between PC axis scores and original variables were quantified using Pearson’s correlation coefficients.

Associations among root traits, metrics of organismal-level productivity and rates of cover expansion were evaluated with simple linear regression. Although multiple regression can reveal more complex relationships than we evaluated, correlations among root traits were sufficiently strong that variance inflation would have rendered results unreliable (Graham, 2003). Species means were used for the analysis (n = 6), with productivity metrics and expansion rate used as response variables. PC axis scores (denoted PC1_root, PC2_prod, etc.) were included in the analysis since they represented integrated measures of nutrient acquisition and productivity. If needed, response variables were log-transformed to improve residual normality and homoscedascity. All data analyses were carried out using R 3.2.2. Statistical significance was assessed using α = 0·05.

RESULTS

Root functional traits

Species fell into two groups with respect to SRL (Fig. 1A; Table 2); Berberis thunbergii, Lonicera maackii, Rubus allegheniensis and Rubus phoenicolasius had relatively high SRL means (24–31 m g⁻¹), whereas Vibernum dentatum and Lindera benzoin had relatively low means (8–10 m g⁻¹). Differences in SRL corresponded closely to difference in mean root diameter, as these were 0·2–0·3 mm in species with high SRL, but 0·5–0·7 mm in L. benzoin and V. dentatum (Figs 1B and 2). Diameter distributions were positively skewed for all species, with skewness coefficients 1·5–3·5 times greater for Rubus than for other species (Figs 1C and 2). Root C:N was least in B. thunbergii and greatest in V. dentatum; mean C:N ratios were 13·5 and 29·0 at these extremes, respectively (Fig. 1D). RTD ranged over a factor of five among the species investigated; means were greatest in Rubus species, at 1·3–1·7 g cm⁻³, and least in L. benzoin, at 0·3 g cm⁻³ (Fig. 1E). None of the above traits was altered by growth in organic patches (Table 2).

Fig. 1.

Root functional traits and metrics of productivity (mean ± s.e.) for the six species included in the study. All root traits are measured on fine roots (<2 mm diameter) except RTD. Within each panel, bars that do not share a letter have statistically separable means based on Tukey’s HSD tests. Metrics with two sets of bars (I–L) differed by enrichment status or the species × enrichment interaction (Table 2); bar heights depict sums for the three ingrowth bags of each type. C:N, carbon to nitrogen ratio; Diam_mean, mean parameter of diameter distribution; Diam_skew, skewness parameter of diameter distribution; Length_rt, root length production; Mass_rt, root mass production; Mycor, mycorrhizal coverage; Precision_lngth, length-based foraging precision; Precision_mass, mass-based foraging precision; RGR_sht, relative growth rate of shoot biomass; RTD, root tissue density; SRL, specific root length; Surf Area_rt, root surface area production.

Table 2.

Results of likelihood ratio tests for root traits and metrics of organismal-level productivity

Category	Variable	Transform	Term	d.f.	χ2	P-value
Root traits	SRL	Neg sqrt	S	5	176·4	<0 ·001
			BT	1	<0·01	0 ·99
			S × BT	5	3·9	0 ·56
	Diammean	Sqrt	S	5	236·1	<0 ·001
			BT	1	0·3	0 ·58
			S × BT	5	3·1	0 ·69
	Diamskew	Log	S	5	165·5	<0 ·001
			BT	1	0·3	0 ·59
			S × BT	5	4·1	0 ·54
	C:N	Sqrt	S	5	175·2	<0 ·001
			BT	1	0·0	1 ·00
			S × BT	5	0·0	1 ·00
	RTD	Neg sqrt	S	5	165·4	<0 ·001
			BT	1	0·1	0 ·81
			S × BT	5	3·8	0 ·58
	Precisionmass	None	S	5	10·4	0 ·06
	Precisionlngth	Square	S	5	10·1	0 ·07
	Mycor	Logit	S	5	54·4	<0 ·001
			BT	1	5·2	0 ·02
			S × BT	5	6·9	0 ·23
Productivity metrics	RGRsht	Sqrt	S	5	121·5	<0 ·001
	Lengthrt	Sqrt	S	5	30·9	<0 ·001
			BT	1	45·9	<0 ·001
			S × BT	5	6·9	0 ·23
	Massrt	Sqrt	S	5	43·7	<0 ·001
			BT	1	18·3	<0 ·001
			S × BT	5	10·1	0 ·07
	Surf Areart	Sqrt	S	5	78·9	<0 ·001
			BT	1	44·4	<0 ·001
			S × BT	5	13·3	0 ·02

S, species; BT, bag type; S BT, species by bag-type interaction.

Other abbreviations are defined in the legend of Fig. 1.

Fig. 2.

Distributions of fine root diameter, computed on the basis of length, for the species included in the study. Curves depict log-normal distributions whose parameters are the mean (± s.e.) of the fitted parameter estimates across individuals of the species.

Median foraging precision, when calculated on the basis of length, was significantly greater than 50 % in all cases (P = 0·004–0·039). For most plants, 55–75 % of net root length production occurred in organic patches (Fig. 1F). Results for mass-based foraging precision were similar (Fig. 1G;P = 0·004–0·020), though the median for R. phoenicolasius was not differentiable from 50 % (P = 0·098). We found no interspecific differences in foraging precision (Table 2).

Across all species, evidence of fungal colonization was found in 82 % of the fields of view sampled; hyphae were present in 80 %, arbuscules were present in 46 % and vesicles were present in 24 %. Although the probability of mycorrhizal occurrence was influenced by species identity (Table 2), the only pairwise difference that our analysis identified was between L. benzoin (92 ± 2 %; mean ± s.e.) and R. allegheniensis (68 ± 2 %; Fig. 1I). Mycorrhizal colonization was also greater in organic patches than in unamended field soil (85 ± 4 % vs. 79 ± 4 %). The species × bag type interaction was not statistically significant.

The PCA of root functional traits yielded three principle components that were interpretable based on broken-stick criteria; these represented 49, 24 and 16 % of variance in the data, respectively. Species were tightly clustered in the biplot of PC1_root and PC2_root, and were separated into two structural forms (Fig. 3A). PC1_root most strongly distinguished these forms; it was correlated with SRL (r = –0·78; all reported values have P < 0·001), RTD (r = –0·79), and the mean and the skewness of the diameter distribution (r = 0·95 and –0·82, respectively). The two species with high PC1_root scores (V. dentatum and L. benzoin) thus invested in roots that had greater mass and were more consistently wide relative to their length; the remaining four species had roots that were substantially more elongated. Although species with thin roots had a moderately wide range of PC1_root scores (Fig. 3A), both structural forms varied with respect to PC2_root. PC2_root was most strongly correlated with the C:N ratio (r = –0·86), SRL (r = 0·51) and RTD (r = –0·51). Low PC2_root scores (exhibited by V. dentatum and the Rubus species) were therefore associated with more durable tissue constructions, presumably corresponding to longer life spans but less metabolic activity. In contrast, B. thunbergii, L. maackii and L. benzoin had PC2_root scores corresponding to shorter life spans but greater metabolic activity. PC3_root was most strongly correlated with foraging precision (r = –0·98). Although PC3_root scores indicated that V. dentatum had marginally greater foraging precision than other species, the univariate analysis of precision did not provide conclusive evidence of this.

Fig. 3.

Principal component (PC) biplot of (A) root functional traits and (B) productivity metrics for the six species included in the study. Points represent individual plants (n = 46). Abbreviations are defined in the legend of Fig. 1.

Organismal-level productivity

Root productivity varied by species and ingrowth bag type (Table 2). Lonicera maackii produced the greatest root length (Fig. 1J), V. dentatum produced the most root mass (Fig. 1K) and L. benzoin produced the most surface area (Fig. 1L); means were significantly greater than most, but not all, other species. However, root production values for V. dentatum and L. benzoin were probably influenced by a methodological artefact. Because root length density decreases radially outwards from the centre of root systems, ingrowth bag placement (at the margins of clipped root systems) probably influenced the spatial density of roots sampled. Given that V. dentatum and L. benzoin had particularly small diameter root systems when they were planted (data not shown), their ingrowth bags were placed in a relatively high-density region of final root systems, and root productivity metrics are probably inflated. Root length, mass and surface area were greater in organic patches than in field soil across all species (Fig. 1J–L). The magnitude of this effect was similar among species for length and mass production. However, there was a significant species × bag type interaction for surface area, such that L. maackii, V. dentatum and L. benzoin had greater responses to organic patches than the remaining species (Table 2). Rubus species had the lowest, or among the lowest, net root production by all metrics considered (Fig. 1J–L). However, Rubus species and L. maackii had the fastest shoot growth rates, whereas V. dentatum and L. benzoin had the slowest (Fig. 1H). Berberis thunbergii had moderate productivity, both above- and below-ground, compared with the other species investigated.

In the multivariate analysis, PC1_prod summarized 62 % of the variance in the data set, while PC2_prod summarized 22 %. However, PC2_prod was not important according to broken-stick criteria. PC1_prod was strongly correlated with shoot growth rate (r = 0·74), as well as the production of root mass (r = –0·78) and surface area (r = –0·96; Fig. 3B). Given that absolute biomass production for all species was substantially greater above-ground than below-ground (as indicated by extrapolations of root ingrowth biomass throughout the soil volumes under plant canopies; data not shown), higher values of PC1_prod simultaneously reflected greater organismal-level growth rates and greater allocation to shoots vs. roots. Rubus species had the highest PC1_prod scores, followed by B. thunbergii and L. maackii; V. dentatum and L. benzoin had the lowest PC1_prod scores.

All productivity metrics had statistically significant linear or log-linear relationships with at least one root trait (Table 3). Moreover, the primary multivariate axes associated with productivity (PC1_prod) and root traits (PC1_root) were strongly, and inversely, related (Fig. 4). This relationship was largely driven by shoot growth rate being positively associated with SRL and diameter skewness, and inversely associated with mean diameter (Table 3). PC1_prod scores were also positively associated with RTD but negatively associated with mycorrhizal coverage. Individual root productivity metrics were correlated with sub-sets of the root traits with which PC1_prod was correlated, although relationships were opposite in direction (Table 3). The only exception was that root mass production was positively related to root C:N ratio whereas PC1_prod was not.

Table 3.

Coefficients of determination (R²) for linear regressions of productivity metrics and population growth on root traits

Root trait	Productivity metric					Population growth
	PC1prod	RGRsht	Lengthrt	Massrt	Surf Areart
PC1root	0 ·92↓	0 ·82↓	0 ·36	0 ·55	0 ·91↑	0 ·65↓
PC2root	0 ·00	0 ·12	0 ·33	0 ·57	0 ·13	0 ·11
SRL	0 ·89↑	0 ·71↑	0 ·01	0 ·74↓	0 ·60	0 ·28
Diammean	0 ·94↓	0 ·68↓	0 ·24	0 ·51	0 ·91↑	0 ·50
Diamskew	0 ·80↑	0 ·85↑	0 ·49	0 ·24	0 ·73↓	0 ·75↑
C:N	0 ·23	0 ·48	0 ·05	0 ·68↑	0 ·05	0 ·00
RTD	0 ·66↑	0 ·56	0 ·78↓	0 ·11	0 ·84↓	0 ·77↑
Precisionlngth	0 ·17	0 ·28	0 ·13	0 ·13	0 ·03	0 ·00
Mycor	0 ·69↓	0 ·28	0 ·69↑	0 ·32	0 ·79↑	0 ·50

Statistically significant relationships are in bold and the direction of their relationships is indicated (↑ = increasing, ↓ = decreasing).

Relationships derived from log-transformed response variables are italicized.

Abbreviations are defined in the legend of Fig. 1.

Fig. 4.

Associations among the primary variables measured in the study: the principal component (PC) axes representing root structure (PC1_root) and productivity (PC1_prod), as well as the rate of cover increase (Spread) in the Buell–Small Succession Study. Filled symbols depict mean species-level data (mean ± s.e. for PC scores) while open symbols depict plant-level data. Abbreviations are defined in the legend of Fig. 1.

Population growth

Rates of cover expansion ranged by nearly a factor of ten in the BSS study for the shrub species considered. Rubus allegheniensis exhibited the most rapid expansion in cover, at a rate of 0·23 % year⁻¹ (percentages are based on the total area of all plots), while L. maackii and R. phoenicolasius had rates of 0·091 and 0·076 % year⁻¹, respectively (Fig. 4). Rates for B. thunbergii, V. dentatum and L. benzoin were between 0·024 and 0·027 % year⁻¹. In addition to having the fastest rate of cover expansion, R. allegheniensis was the first of the focal species to appear on the site (it initially occurred in plots 5 years of age), reached the highest peak cover (3·59 % after smoothing) and did so in younger plots (i.e. with less overstorey canopy cover) than any species except B. thunbergii (Fig. S1). Most species first occurred when plots were young (<25 years since abandonment) and reached peak cover 12–19 years after the first occurrence. However, R. phoenicolasius first occurred when plots were more mature (36 years since abandonment), and peaked after 6 years. Lonicera maackii had 4–5 times the peak cover of any species (at 2·23 % vs. 0·28–0·53 %) except R. allegheniensis, and may still be increasing in frequency at the BSS site (data not shown). Berberis thunbergii began expanding rapidly in recent years, but was subject to control efforts beginning around 2010 (Fig. S1).

Rates of cover expansion were correlated with a sub-set of root traits (Table 3). More rapid expansion was associated with more strongly skewed root diameter distributions, greater RTD and lower PC1_root scores (Fig. 4). Associations of cover expansion rate with PC1_root and the skewness coefficient for diameter were described substantially better by log-linear than by linear relationships. There were no statistically significant correlations between cover expansion rate and metrics of productivity.

DISCUSSION

We documented substantial variation in root functional traits, and thus nutrient foraging strategies, among ENA forest shrubs. This variation corresponded well to the proposition that root tissue construction is largely determined by a trade-off between resource assimilation rate and longevity, or a fast vs. slow return on investment (Kong et al., 2014; Reich, 2014; Eissenstat et al., 2015; Roumet et al., 2016). Moreover, species either had relatively high or relatively low SRL. A similar dichotomy has been called ‘thick’ vs. ‘thin’ root forms in studies of trees (Eissenstat, 1992; McCormack et al., 2012) and may also involve a fundamental trade-off between constructing fine roots with greater diameter or greater branching density (Pagès, 2014). Given that the variation in root traits that we observed corresponds to established strategies of tissue construction and nutrient foraging, our results indicate that the forest shrubs we studied differ considerably in the soil nutrient regimes to which they are best adapted.

Both of the historically dominant understorey shrubs we included (L. benzoin and V. dentatum) had the thick, low SRL fine roots characteristic of the conservative resource acquisition strategy. This root form requires a greater energetic investment but yields greater longevity and carbohydrate storage than the thin root form (Ostonen et al., 2007; McCormack et al., 2012; Eissenstat et al., 2015). Our results were also consistent with the possibility that species with thick roots rely more strongly on mycorrhizae for nutrient acquisition (Kong et al., 2014; Eissenstat et al., 2015), though mean colonization values were only statistically separable for one pair of species (L. benzoin and R. allegheniensis). While the resource-conservative foraging strategy may have been highly adaptive under historical, predominantly low-nutrient conditions, it may be mismatched to contemporary soil conditions, which are characterized by increasing N availability over increasingly widespread areas (Aber et al., 2003). For instance, species with thick root morphologies are unable to increase root life span when they are grown in enriched N conditions (Adams et al., 2013), potentially making them less competitive in contemporary forests.

Fine roots of L. maackii, B. thunbergii and the two Rubus species were distinctly thinner and had higher SRL than the other two species, indicating that they use more acquisitive nutrient foraging strategies. Roots built for rapid nutrient uptake and processing may enable these and species with similar traits to capitalize on the greater N loading rates now characteristic of ENA forests. Coupled with other factors that increase productivity (e.g. photosynthetic efficiency; Heberling and Fridley, 2013) or alter interspecific interactions (e.g. allelopathy; Hale et al., 2016), root traits may thus facilitate acquisitive species displacing conservative species in contemporary forests. However, thin-rooted species exhibited a broad range of root traits; this variation manifested most strongly in RTD, but was also evident in diameter distribution parameters, C:N ratio and mycorrhizal coverage (Fig. 1). Some of the variation in RTD may have been due to both transporting and absorbing roots being included in samples (McCormack et al., 2015). Given that Rubus species had the thinnest roots, their RTD values would have been disproportionately affected. However, data on lignin content from a sub-set of plants indicated that L. maackii roots had the highest lignin content in the group (38 % vs. 25–31 %) despite being approximately half as dense as Rubus roots. The contrasting patterns for RTD and lignification may reflect different strategies of extending root life span within the constraints of a thin-root morphology. Similarly, variation in diameter distributions and mycorrhizal coverage may reflect differences in the degree to which these species take up nutrients directly vs. depend on mycorrhizae (Eissenstat et al., 2015).

Rubus species generated surprisingly little root length relative to the shoot biomass that they produced (Fig. 1H, J). This was not likely to be an artefact of clipping roots, given that, among ENA trees, species with thin roots had stronger responses to clipping than species with thick roots (Eissenstat et al., 2015). Thus, unless R. allegheniensis and R. phoenicolasius are far more plastic in their root:shoot allocation than other Rubus species (Caplan and Yeakley, 2013), the minimal investment in roots exhibited by these Rubus species could yield reduced competitive ability when nutrient concentrations are low. A requirement for high nutrient supply by Rubus would be consistent with the association of these species with disturbed habitats such as forest edges and treefall gaps (Aikens et al., 2007; Gorchov et al., 2011), where soil nutrients and light are often abundant. Several other Rubus species are similarly known to achieve high productivity only under conditions of high resource availability (Caplan and Yeakley, 2006, 2010, 2013).

Foraging precision may not be a strong determinant of nutrient acquisition in ENA forest understoreys. Although all species grew roots preferentially into organic patches, our results provided no evidence that acquisitive species had greater foraging precision than conservative species, as we had expected. It is possible that some species were capable of greater foraging precision than they exhibited in our study, but that differences in nutrient availability between organic patches and field soil were too small to warrant additional investment within patches. If this occurs widely in the context of nutrient-enriched forests, finding organic patches in the soil (i.e. foraging scale) may be more important for shrub success than foraging precision. In that case, species would be more likely to differ in traits that we did not measure, such as root placement (e.g. vertical and radial distribution), branching topology or soil exploration efficiency (Fitter and Stickland, 1991; Berntson, 1994; Hodge et al., 2009).

Our analysis demonstrated that interspecific variation in root functional traits is associated with productivity and population growth in at least some ENA forest shrubs. As expected, species with the resource acquisitive syndrome of root traits (especially higher SRL) had more rapid above-ground growth and greater rates of cover increase. In contrast, species with lower SRL had much slower growth rates and spread less quickly. Although we have not demonstrated that the relationships are causal, traits conferring rapid nutrient acquisition as well as rapid or prolonged carbon gain occur together in a number of understorey shrub species (Smith et al., 2014; Jo et al., 2015), providing support for this possibility.

Neither root traits nor metrics of productivity differed strictly by species origin in this study. Like other cases reported in the literature, our results (particularly from the pair of Rubus species) demonstrate that native and introduced species adapted to resource-rich habitats can have similar life history strategies, functional traits and population dynamics (Thompson et al., 1995; Meiners, 2007; Leishman et al., 2010; Dawson et al., 2012). However, other studies have identified differences in functional traits between invasive (i.e. not only introduced, but aggressive) species when comparing them with native shrubs from ENA forests (Heberling and Fridley, 2013; Jo et al., 2015). We attribute this, in part, to the selection of species from a pool of known aggressive species, rather than a cross-section of all introduced species. Nevertheless, our results are consistent with the possibility that introduced shrubs are more likely to become invasive in regions with nutrient-enriched soils if they have acquisitive root traits (Peltzer et al., 2016).

Using an acquisitive foraging strategy may be a necessary but insufficient condition for some shrub species to achieve high productivity in nutrient-enriched forest soils. Berberis thunbergii may be an example of this, as it had substantially slower rates of shoot growth and cover increase than other species with high SRL. Berberis thunbergii is known to induce plant–soil feedbacks wherein its litter causes soil microbes to accelerate nitrification, and high nitrate reductase allows it to capitalize on the additional nitrate (Ehrenfeld et al., 2001; Elgersma et al., 2012). Slow growth in our common garden may therefore have been due to a lack of time for plant–soil feedbacks to establish. This may have also contributed to the slow rate of cover increase for B. thunbergii at the BSS site, though low propagule pressure from the surrounding landscape probably played a greater role (Eschtruth and Battles, 2009). The recent rise in B. thunbergii cover in some plots at the site (Fig. S1) suggests that it is capable of expanding as rapidly in that system as other species with high SRL.

The results of this study contribute to a growing body of research that is connecting changes in nutrient regimes to changes in plant abundance patterns in forest communities globally (Gilliam, 2006; Köchy and Bråkenhielm, 2008; Bobbink et al., 2010; Verheyen et al., 2012; Clark et al., 2013). Most of the high SRL species in this study have increased in abundance in ENA forests in recent decades, a pattern which our data suggest was aided by these species having a strong ability to acquire widely available soil nutrients. We propose that nutrient enrichment may act as an ecological filter in ENA forest understoreys, favouring species with root traits capable of accessing and processing nutrients rapidly (Katabuchi et al., 2012). As such, nutrient enrichment would operate in concert with other ecological filters such as deer herbivory and canopy disturbance (Beckage et al., 2008; Nuttle et al., 2013) to shape community composition. We further suggest that the increased frequency of understorey invasions in ENA is partly symptomatic of widespread nutrient enrichment, though we note that some native species may also benefit from nutrient enrichment, and that many non-aggressive introduced species are probably maladapted. In the face of intensifying eutrophication of forest soils due to urbanization and other processes, it will be important to determine if eutrophication favours species with acquisitive nutrient foraging strategies globally.

SUPPLEMENTARY DATA

Supplementary data are available online at https://academic.oup.com/aob and consist of the following. Table S1: soil characteristics for the common garden. Figure S1: species cover in the Buell–Small Succession Study.