Foraging for space and avoidance of physical obstructions by plant roots: a comparative study of grasses from contrasting habitats

Authors


Author for correspondence:
Marina Semchenko
Tel: +372 737 6188
Fax: +372 737 6222
Email: marina.semchenko@ut.ee

Summary

  • • Physical obstructions that reduce space for root growth can profoundly affect plant performance. The aim of this study was to investigate the ability of roots to avoid obstructions and forage for usable space, and to reveal the mechanism involved.
  • • Eight grass species from four genera were examined. Each genus included species characteristic of habitats with high and low nutrient availability. The ability to limit root mass and to adjust morphology within substrate containing obstructions in the form of gravel was investigated. A treatment with activated carbon, which adsorbs organic compounds, was used to examine the possible involvement of root exudates in responses to obstructions.
  • • Only species characteristic of nutrient-poor habitats restricted placement of root mass in substrate containing obstructions, and this response disappeared in the presence of activated carbon. Root morphological responses to obstructions differed from those shown in response to nutrient-poor conditions or compacted soil.
  • • These results suggest that the ability to avoid obstructions is dependent on the sensitivity of roots to their own exudates accumulating in the vicinity of obstructions. This is similar to other behavioural responses in which cues or signals are used to adjust growth before stressful conditions are encountered.

Introduction

It is known that the availability of below-ground space can strongly influence plant performance. Several studies have shown that the restriction of root systems to small substrate volumes can have profoundly detrimental effects on plant growth and cause significant physiological changes, even when nutrients, water and oxygen are not limiting (Carmi & Heuer, 1981; Tschaplinski & Blake, 1985; Hameed et al., 1987; McConnaughay & Bazzaz, 1991; Kharkina et al., 1999). Moreover, the presence of impenetrable obstructions in the substrate has been shown to affect root growth and plant performance (McConnaughay & Bazzaz, 1992; Semchenko et al., 2007). Despite increasing awareness of the importance of space for root growth, foraging and interactions between plants (Schenk, 2006; Hess & de Kroon, 2007; Semchenko et al., 2007), the ability of roots to avoid growth towards physical obstructions, and to adjust morphology when growing through substrate containing obstructions, has rarely been examined. However, studies on chemical signalling suggest that plants may possess mechanisms conferring such capabilities (Goss & Russell, 1980; Falik et al., 2005).

Under natural conditions, root systems are likely to experience spatial heterogeneity not only in nutrient concentration and water availability, but also in the physical properties of the substrate. For example, roots can encounter obstructions in the form of compacted soil, neighbouring roots, organic debris, stones and other objects with a variety of sizes and densities. Such obstructions can interfere with root growth in a number of ways. Firstly, they reduce space for root growth and deployment, and can cause roots to become crowded, leading to a reduction in resource acquisition efficiency (McConnaughay & Bazzaz, 1992; Nielsen et al., 1994; Rubio et al., 2001). Secondly, obstructions fragment the usable parts of the substrate and hamper root access to resources – for example, roots may be hindered from locating and exploiting resources positioned behind obstructions, or may need to elongate much more before they can do so (McConnaughay & Bazzaz, 1992; Semchenko et al., 2007). In addition, roots may need to exert more pressure on the surrounding substrate to pass between obstructions, and this may require changes in root physiology and morphology (Goss, 1977; Goss & Russell, 1980; Bengough, 2003). Plants could therefore benefit from being able to forage for usable space, avoiding substrate with a high density of obstructions in favour of substrate that their roots can explore and exploit more fully. However, there have been no studies to assess the ability of roots to avoid obstructions, to determine how widespread any such ability might be, or to explore the attributes that might confer this ability.

Identifying mechanisms conferring foraging ability in heterogeneous conditions has become one of the most challenging areas in plant science. There is a growing body of evidence showing that plants can regulate their behaviour in response to abiotic and biotic factors not only by directly assessing resource availabilities, but also by employing complex chemical signalling. For example, allelopathic exudates and nontoxic signals have been identified as important mechanisms by which roots can detect and react to the proximity of neighbouring roots (Mahall & Callaway, 1992, 1996; Ridenour & Callaway, 2001). Our understanding of the mechanisms that plants could use to assess the density of obstructions in the substrate, and to adjust root behaviour appropriately, is poor. Several studies have shown that plant hormone production plays a key role in mediating root responses to physical impedance and restricted substrate volumes (Goss & Russell, 1980; Carmi & Heuer, 1981; Moss et al., 1988; Sarquis et al., 1991; Clark et al., 2003). It has also been demonstrated that the ability of roots to limit growth towards obstructions, and to reduce proliferation in small substrate volumes, can be controlled by sensitivity of roots to accumulation of their own allelopathic exudates in the vicinity of obstructions (Falik et al., 2005; Semchenko et al., 2007). These mechanisms, which have hitherto only been demonstrated under highly artificial conditions, might also be used by plants to forage for usable space in soil where obstructions are heterogeneously distributed.

This study was designed to examine the ability of plant roots to avoid obstructions and to forage for usable space in the substrate. Plants were grown in pots in such a way as to allow their roots the choice of growing into substrate containing only sand (particle diameter < 1 mm) or into substrate containing a mixture of sand and larger obstructions in the form of gravel (nominal particle diameter 4 mm). The larger size of gravel particles presents greater potential for obstruction of roots than that of sand particles. Larger particles also reduce the space (per unit volume of substrate) that roots can occupy and from which they can obtain resources. Treatments were established with different nutrient availabilities, and with or without the addition of activated carbon, to determine the role of nutrient availability and root exudates in modulating root placement and morphology. Activated carbon adsorbs organic root exudates from the substrate, reducing their possible effects on root growth (Mahall & Callaway, 1992; Ridenour & Callaway, 2001). The following hypotheses were tested:

Hypothesis 1.  Plants are able to limit root mass within substrate containing obstructions compared with substrate free of obstructions.

Hypothesis 2.  Species from nutrient-poor habitats are more effective than those from nutrient-rich habitats in limiting root mass in patches containing obstructions. This prediction is based on the results of previous studies that have demonstrated higher root construction and maintenance costs in species from nutrient-poor habitats (Poorter et al., 1990; Wahl & Ryser, 2000), and stronger reductions in the growth of plants constrained to densely obstructed substrate at low than at high nutrient concentrations (McConnaughay & Bazzaz, 1992). To minimize the confounding effects of phylogenetic differences between species from different habitats, eight grass species from four genera were examined (Felsenstein, 1985; Baskauf & Eickmeier, 1994). One species from each genus was characteristic of nutrient-poor habitats and the other was characteristic of nutrient-rich habitats.

Hypothesis 3.  Self-inhibition of root growth, owing to the local accumulation of root exudates, contributes to avoidance of obstructions. If self-inhibition via such a mechanism is involved, the addition of activated carbon to the substrate should impair the ability of plants to limit root mass placement in substrate containing obstructions.

Hypothesis 4.  The morphological responses of roots to obstructions will differ from those documented for roots subjected to low nutrient availability or compacted soil. Reduced lateral root production and elongation are the most common responses of roots to low nutrient concentration (reviewed in Hutchings & de Kroon, 1994; Robinson, 1994; Hutchings & John, 2003), whereas common responses to mechanical impedance within the substrate include reduction in elongation rate of the main root axis, and increases in root diameter and branching density (Goss, 1977; reviewed in Bengough, 2003; Clark et al., 2003). However, obstructions in the substrate directly affect the availability of space for growth. They present a different type of environmental challenge than change in either nutrient supply or soil compaction. The experimental design used in this study minimized differences in nutrient concentration and compaction between the substrates with and without gravel, enabling us to focus directly on the effects of obstructions and usable space on root growth.

Hypothesis 5.  Any obstruction avoidance response will be dependent on overall nutrient concentration in the substrate. We predicted that further root growth into substrate fragmented with obstructions should be precluded only if root crowding occurs, for example, when root length per unit of usable substrate volume is relatively large. As we expected plants to be much larger in the nutrient-rich treatment than in the nutrient-poor treatment by the end of the experiment, we expected more root crowding, and consequently more evidence of obstacle avoidance, in the nutrient-rich treatment.

Materials and Methods

Species

Eight species from four genera were used in the experiment: Agrostis stolonifera L., Agrostis vinealis Schreber, Festuca pratensis Hudson, Festuca ovina L., Phleum pratense L., Phleum phleoides (L.) Karsten, Poa trivialis L. and Poa bulbosa L. Within each genus, the species listed first is characteristic of nutrient-rich habitats and that listed second is characteristic of nutrient-poor habitats. Ellenberg nitrogen values are 6 for all the species of nutrient-rich habitats, and 2 for all the species of nutrient-poor habitats (Hill et al., 1999).

Agrostis stolonifera is a fast-growing species that is common in a wide range of fertile, moist, moderately disturbed conditions. It is particularly abundant on road verges, spoil heaps, pastures and arable land. A. vinealis is most common in dry, undisturbed and relatively unproductive grasslands. Festuca pratensis is a short-lived species found in moist grassland of intermediate productivity, including meadows, pastures, road verges and wasteland. F. ovina is a slow-growing species found in a wide range of unproductive grasslands and rocky habitats, and is particularly abundant in upland pastures, heaths and moors. Phleum pratense occurs in vegetation of intermediate to high productivity in moderately disturbed habitats, and is common in meadows, pastures and arable land (Grime et al., 2007). Pphleoides is characteristic of dry sandy and chalky pastures and rough ground (Stace, 1997). Poa trivialis is a fast-growing species found most frequently in damp and fertile conditions, including wetlands, woods, grassland and arable fields (Grime et al., 2007). P. bulbosa is found in short grasslands, open ground on sandy soils and shingle habitats, and on limestone in coastal regions (Stace, 1997).

Experimental design and measurements

Seeds of all species were germinated in trays of moist vermiculite in February 2007. For each species, seedlings of equal size were transplanted individually into the centre of 15-cm-diameter (1.6 l) pots. Each pot was filled with two different substrates separated by a partition placed vertically across the centre of the pot. The partition was removed after the pots had been filled. One half of each pot was filled with pre-washed sand (particle size < 1 mm, Fargro Ltd, Littlehampton, UK) and the other half was filled with a mixture of sand and gravel, with 20% of the substrate volume consisting of washed and graded lime-free quartzite gravel, nominal size 4 mm (William Sinclair Horticulture Ltd, Lincoln, UK). The relatively low density of obstructions enabled us to maintain the same amount of substrate compaction and nutrient concentration per unit of sand volume in both halves of a pot, ensuring that the main difference between the halves was the availability of space for growth. Root placement might also be affected by water availability, which may depend on particle size distribution in the substrate. However, a recent study in which plants were grown in a substrate similar to that used in our experiment, but subject to water limitation, found little evidence that roots could forage for substrate patches where the availability of water was greater (Cole & Mahall, 2006). Water retention in the mixture of sand and gravel used in this experiment was 16% less than that in the pure sand, but the mixture also contained 20% less sand. Thus, water retention per unit volume of sand was only marginally different between the two substrate types. As pots were also watered regularly to avoid any water deficit, it is unlikely that difference in water availability between the substrate types affected patterns of root placement.

Four treatments were established that combined two factors: the addition or absence of activated carbon, and low or high nutrient concentration. In treatments with carbon addition, activated carbon powder (Sigma-Aldrich Inc., St Louis, MO, USA) was added at the rate of 20 ml l−1 of sand. Nutrients were supplied in the form of controlled-release fertilizer pellets (Osmocote Mini Plus; Scotts Professional Ltd, Nottingham, UK). The fertilizer contained 16% N, 8% P, 11% K, 2% Mg and micronutrients. These pellets release nutrients at a constant rate over a period of 3–4 months. Fertilizer pellets were added at the start of the experiment by mixing them thoroughly with the sand in quantities designed to produce the same concentration of nutrients per unit volume of sand in both halves of the pot. In the nutrient-poor treatment, 200 mg of fertilizer was added per l of sand. In the nutrient-rich treatment, 1000 mg of fertilizer was added per l of sand. The amount of fertilizer to be added was decided on this basis because roots react to the nutrient concentration in their immediate vicinity (Zhang & Forde, 1998, 2000). This design therefore allowed roots to perceive the two halves of a pot as being equal in nutrient availability, even though total nutrient content differed between the two halves of the pot. Semchenko et al. (2007) also confirmed that root growth was affected by the presence of gravel, even when nutrient concentration and the total amount of nutrients available to a plant were equal in substrate with and without gravel.

The experiment was carried out in a glasshouse maintained at a temperature of 20 ± 5°C, with additional light supplied by Osram 400 W lamps on a 16 : 8 h light : dark cycle. Pots were arranged in a single block and their positions were randomized at the start of the experiment and three times during the experiment. There were 12 replicate pots per treatment × species combination, except for P. phleoides and P. bulbosa. These species could only be replicated seven and six times per treatment, respectively, because of poor germination. One seedling each of A. vinealis and F. ovina died during the experiment.

Species of nutrient-rich habitats had higher growth rates and were harvested earlier than the congeneric species from nutrient-poor habitats to reduce differences in size between species at the time of harvest. Consequently, A. stolonifera, F. pratensis, P. pratense and P. trivialis were harvested 55, 57, 60 and 60 d after planting, respectively, and A. vinealis, F. ovina, P. phleoides and P. bulbosa were harvested 70, 70, 80 and 65 d after planting, respectively.

At harvest, the above-ground parts of each plant were separated from the roots. The substrate in each half of the pot was then separated using a sharp metal partition inserted along the boundary between the two substrates. Roots were separated from the substrate in each half of the pot by careful washing over a sieve. A representative primary root axis with all its associated laterals was selected from each pot half for morphological analysis, and stored at 4°C. The remaining parts of each plant were dried separately at 70°C for 48 h and weighed. The stored primary roots were stained with neutral red, spread out on glass trays and scanned at 400 dpi. The images were analysed using WinRhizo 4.1c (Regent Instruments Inc., Quebec, Canada). Total root length, length of primary axis, total number of branches and number of first-order laterals were determined for each sample. After scanning, root samples were dried and weighed as described earlier, and these weights were then added to the root fractions from which they had been taken. Based on these measurements, the specific root length (total root length/mass of root axis plus all laterals), the mean lateral length ((total root length – length of primary axis)/total number of branches) and the density of laterals on the primary axis (number of first order laterals/length of primary axis) were calculated.

Statistical analysis

The effects of type of habitat which is characteristic for each species (fixed factor with two levels: nutrient-rich or nutrient-poor), activated carbon (fixed factor with two levels: carbon present or absent), nutrients (fixed factor with two levels: low or high concentration) and substrate type within a pot (repeated-measures factor with two levels: substrate without gravel and substrate with gravel) on root mass, and on traits measured on the primary root axis samples, were estimated using general linear mixed models. All interactions between these factors were included in the models. Genus was included in the models as a random effect since the aim of this study was to investigate general trends in obstruction avoidance between species characteristic of contrasting types of habitat, rather than to concentrate on differences between genera or species that were not attributable to habitat type (Littell et al., 2006).

We also analysed the data using a fixed-effects model with all the factors listed in the previous paragraph, but with genus as a fixed effect. If significant differences in root responses to gravel were observed between habitats in the mixed-effects model, this fixed-effects analysis would indicate whether there was also significant variation in root behaviour between genera or species (reflected in gravel × genus × (other factors) or gravel × habitat × genus × (other factors) interactions, respectively). Examination of these interaction terms in the case of significant habitat effect showed that neither genera nor species varied significantly in their root placement responses to gravel. Therefore, only results from mixed-effects models are presented.

When necessary, variables were ln-transformed before analysis to improve normality of residuals and homogeneity of variances. The data were analysed using SAS v. 9.1 (SAS Institute Inc., Cary, NC, USA).

We measured the effect of activated carbon on nutrient concentrations in water solution using an ion chromatography system (ICS-1000, Dionex Corp., CA, USA). Whereas some of the ions examined (K+, inline image, inline image, inline image) were not affected by the presence of activated carbon, the concentrations of inline image, inline image and inline image were significantly reduced in its presence, especially in the nutrient-poor treatment. Changes in root morphology and reductions in total mass were observed when activated carbon was added to the substrate; however, these results are not presented here, as it would be difficult to distinguish between the effects of carbon on the concentrations of nutrients and root exudates in the substrate. Instead, we confine our examination of the effects of carbon to comparisons of root mass placement and root morphology in substrate with and without gravel. If a significant interaction between the effects of activated carbon and gravel is observed, it can be attributed to the effect of activated carbon on root exudates. It cannot be ascribed to the effects on nutrients because both nutrients and activated carbon were homogenously distributed throughout the pots. Thus, the same changes in root growth should be observed in both substrates if activated carbon only affected nutrient concentration (i.e. there would not be a significant interaction between carbon and gravel).

Results

Placement of root mass

A significant four-way interaction was detected when the effects of different factors on root mass were examined (Table 1). Plants of nutrient-poor habitats placed 15% less root mass in the half of the pot with gravel than in the half of the pot without gravel in the high-nutrient and no-carbon treatment (Fig. 1). This effect disappeared when activated carbon was added to the substrate, suggesting the involvement of root exudates in an avoidance response to the presence of obstructions. No selective placement of root mass was observed in the low-nutrient treatment for species from nutrient-poor habitats. By contrast, plants of nutrient-rich habitats did not show any differences in root mass placement in any treatment (Fig. 1).

Table 1.  The results of linear mixed models for seven plant root traits
Effect/traitRoot massVariables associated with root elongationVariables associated with root branching
Total root lengthSpecific root lengthLength of primary axisMean lateral lengthTotal number of lateralsDensity of laterals on the primary axis
  1. The statistical significance of the effects of species habitat type (nutrient-rich or nutrient-poor), nutrient treatment (high or low), addition of activated carbon (present or absent) and substrate type within a pot (substrate with or without gravel – denoted as ‘gravel’ in the table) is presented. Genus was included in the model as a random effect. The effect of genus was not significant for any of the traits (P > 0.05). P-values of fixed effects are shown. Bold typeface indicates significant (P < 0.05) effects under a sequential Bonferroni criterion. Degrees of freedom for all parameters were 1 for the numerator and 657 for the denominator.

Habitat0.0819< 0.00010.63270.20260.4693< 0.00010.0405
Gravel0.09460.02020.01470.72280.67790.07040.0002
Nutrients< 0.0001< 0.00010.7066< 0.00010.3024< 0.0001< 0.0001
Carbon0.00020.00060.0810< 0.0001< 0.0001< 0.0001< 0.0001
Habitat × gravel0.85510.80190.69780.81430.24390.61130.3259
Habitat × nutrients0.17110.17400.00840.28520.82890.09090.6856
Habitat × carbon0.49700.15000.15620.58170.80200.09810.2019
Nutrients × gravel0.0023< 0.0001< 0.00010.53310.6248< 0.00010.0007
Carbon × gravel0.02710.17480.46020.18380.10210.72660.0294
Nutrients × carbon0.0214< 0.00010.30920.13160.00070.01070.1781
Habitat × nutrients × gravel0.15630.84610.48050.65230.05160.37080.4782
Habitat × carbon × gravel0.05000.88110.42650.20730.75870.68180.0331
Habitat × nutrients × carbon0.66840.99780.72060.92280.39330.64370.0956
Nutrients × carbon × gravel0.07180.03790.10280.40550.32260.16330.0064
Habitat × nutrients × carbon × gravel0.00200.55460.82650.11510.60110.38620.7355
Figure 1.

The effects of activated carbon (no C/with C), nutrient treatment (low/high), species habitat (nutrient-poor/nutrient-rich) and substrate type within a pot (pure sand, open bars; mixture of sand and gravel, closed bars – each substrate type occupying half of a pot) on root mass. Error bars denote 1 SE of the difference between means in the two halves of the pot. *, P < 0.0001 (Tukey test).

Effects of obstructions on elongation and branching of primary root axes

Plants exhibited similar morphological changes in response to the presence of obstructions regardless of habitat type (no significant habitat × gravel × (other factors) interactions in Table 1). Of all the morphological traits measured, the addition of activated carbon only affected the density of laterals on the primary root axis in substrate with and without obstructions in the nutrient-rich treatment (significant nutrient × carbon × gravel interaction in Table 1). Branching density was 11% lower in the substrate containing gravel than in the substrate without gravel, but the difference disappeared when activated carbon was added to the substrate (Fig. 2). No differences in branching were observed in the nutrient-poor treatment.

Figure 2.

The effects of nutrient treatment (low/high), addition of activated carbon (no C/with C) and substrate type within a pot (pure sand, open bars; mixture of sand and gravel, closed bars – each substrate type occupying half of a pot) on the density of laterals on the primary root axis. Error bars denote 1 SE of the difference between means in the two halves of the pot. *, P < 0.0001 (Tukey test).

There was a significant interaction between the effects of nutrient treatment and the presence of gravel in the substrate on total root length (the summed lengths of a primary axis and all of its laterals; Table 1). Total root length was, on average, 18% less for primary root axes in the half of the pot with gravel than in the half of the pot without gravel in the nutrient-rich treatment, whereas there was no significant difference in the nutrient-poor treatment (Fig. 3a). However, when the length of primary axes and mean lateral lengths were examined separately, no significant responses to obstructions were observed (Fig. 3b,c; no significant effect of gravel in Table 1). Instead, a significant interaction between the effects of nutrient treatment and the presence of gravel in the substrate was observed for specific root length and total number of laterals (Table 1). In the nutrient-rich treatment, roots in the half of the pot with gravel had 11% lower specific length and 18% fewer laterals than roots in the half of the pot without gravel (Fig. 3d,e). No significant differences in root morphology were observed between substrates in the nutrient-poor treatment.

Figure 3.

(a–e) The effects of nutrient treatment and substrate type within a pot (pure sand, open bars; mixture of sand and gravel, closed bars – each substrate type occupying half of a pot) on different root traits measured per primary root axis. Error bars denote 1 SE of the difference between means within a pot. *, P < 0.0001 (Tukey test). Note the log scale.

In summary, obstructions affected root branching and construction costs (specific root length), but not the elongation of single roots.

Discussion

This experiment showed that species characteristic of nutrient-poor habitats were able to limit root mass placement in substrate with a high density of obstructions, whereas those of nutrient-rich habitats were not (hypotheses 1 and 2 supported). Moreover, the ability to avoid obstructions was apparently mediated by chemical signalling rather than by direct reactions to differences in resources (hypothesis 3 supported). This was indicated by the fact that, in the presence of activated carbon, which adsorbs organic compounds, plants were no longer able to place roots preferentially in substrate without gravel. Self-inhibition of root growth in the vicinity of obstructions, caused by the sensitivity of roots to localized accumulation of their own exudates, has been documented in highly artificial conditions, where obstructions were represented by nylon thread (Falik et al., 2005) and by pot walls (Semchenko et al., 2007). Our results show that species of nutrient-poor habitats may employ this mechanism to forage for space in substrate with heterogeneous distribution of obstructions. Responses mediated by root exudates may facilitate efficient root placement by limiting biomass allocation in substrate with many obstructions before resource uptake becomes directly limited by a shortage of growth space. In addition to inanimate objects, obstructions to root growth also include other roots, both of the same plant and of neighbouring plants. Therefore, the ability to avoid obstructions may play an important role in root competition. Responses to the presence of below-ground competitors have usually been attributed to resource depletion and allelopathic exudation by roots of competitors and, in some cases, to nontoxic signals (Schenk et al., 1999; de Kroon et al., 2003; Schenk, 2006). However, obstruction avoidance responses to the presence of neighbouring roots have rarely been considered (McConnaughay & Bazzaz, 1992). Their effects on root competition deserve more attention.

Obstruction avoidance behaviour might also have important implications in agriculture, horticulture and silviculture, and in experiments in which plants are grown in pots of small volume. For example, it has been shown that availability of space can confound the results of experimental studies on root competition where neighbour presence and pot volume have been manipulated simultaneously (Schenk, 2006; Hess & de Kroon, 2007; Semchenko et al., 2007). The findings of this study may help to explain why there is variation in the success with which different species can be grown in pots (NeSmith & Duval, 1998), and suggest that there might be potential to breed plants that are better adapted for growth in limited pot volumes.

Several factors could render the ability to avoid obstructions more critical for species of nutrient-poor habitats. Firstly, such species have high root mass density and greater root longevity than species of fertile habitats (Wahl & Ryser, 2000, Craine et al., 2001; Van der Krift & Berendse, 2002). Moreover, slow-growing species, which are generally associated with nutrient-poor habitats, tend to fix less carbon per unit of plant mass and expend a higher proportion of assimilates in root respiration than fast-growing species (Poorter et al., 1990). These properties imply higher construction and maintenance costs of roots in nutrient-poor habitats, and consequently a stronger selective pressure for traits that minimize nutrient loss, including efficient placement of roots in heterogeneous soil (Aerts, 1999). The suggestion that the ability to express an adaptive growth pattern is more critical in resource-poor than in resource-rich conditions is also supported by studies of the adaptive value of root traits and shade-avoidance responses in Arabidopsis thaliana mutants at different levels of resource supply (Ballare & Scopel, 1997; Fitter et al., 2002).

Nutrient-poor soils often contain higher proportions of large particles (i.e. obstructions) than fertile soils (Borchers & Perry, 1992; Hassink, 1992). This may also contribute to stronger selective pressure on species of nutrient-poor habitats to evolve the ability to forage for space. However, the adaptive value of obstruction avoidance may depend on other habitat characteristics, including water regime and the types of obstructions present. Recent studies on root growth patterns of species characteristic of different soil types suggest that self-inhibition of root growth when physical barriers are encountered may be disadvantageous in shallow soils overlying bedrock prone to summer drought (Poot & Lambers, 2003, 2008). In such environments, the absence of self-inhibition may be an adaptation allowing extensive growth of roots along the surface of the bedrock in search of crevices that can provide access to additional water during summer drought.

In our experiment, plants exhibited a set of morphological responses to obstructions in the substrate that were independent of the nutritional status of their characteristic habitat. Primary root axes that encountered obstructions produced fewer laterals and had lower specific root length than root axes growing in gravel-free substrate. Such changes are common responses to unfavourable nutrient conditions and soil compaction, respectively (Robinson, 1994; Hutchings & John, 2003; Bengough, 2003). Lower specific root length was associated with larger root diameter (data not shown), which increases the ability of roots to exert pressure on the substrate and penetrate the gaps between obstructions (Wilson et al., 1977; Materechera et al., 1991, 1992; reviewed in Clark et al., 2003). Reduced branching indicates that root axes growing in substrate with a high density of obstructions invested less in exploitation of the local substrate than roots growing in substrate without obstructions. However, obstructions did not affect primary axis length or mean lateral length in our experiment, whereas changes in nutrient availability and soil compaction generally cause changes in root elongation (Goss, 1977; Hutchings & de Kroon, 1994; Zhang & Forde, 1998). Moreover, changes in the density of laterals on the primary axis in response to obstructions were mediated by root exudates: the addition of activated carbon rendered root branching density unresponsive to the presence of gravel. This supports the prediction that physical obstructions, and the resultant changes in availability of space, induce a set of responses in plant roots that cannot be entirely attributed to changes in nutrient availability or soil compaction (hypothesis 4).

While there were significant changes in the placement of root mass, and in root morphology, in response to obstructions in the nutrient-rich treatment, no differences were observed in the nutrient-poor treatment, supporting hypothesis 5. The main effect of obstructions in the substrate for plant roots is probably the reduction of usable space, which causes root crowding and hampers the ability of roots to explore the substrate. In addition to possible direct effects of nutrients on root responses to obstructions, plants grown in nutrient-poor conditions were, on average, three times smaller at the end of the experiment than plants grown in nutrient-rich conditions, and their root systems may have been too small to experience crowding in the presence of obstructions. Consequently, they may have perceived substrates with and without gravel as similar in quality. Had plants from low- and high-nutrient treatments been harvested when their root systems were the same size, rather than at the same time, results might have been different.

The results of our study suggest that the ability of species from unproductive habitats to avoid obstructions was determined by the sensitivity of roots to their own exudates accumulating in the vicinity of obstructions. This mechanism may ensure efficient root placement in obstructed substrate before resource uptake becomes directly affected by the aggregation of roots in limited space. This phenomenon is similar to the ability of plants to detect incipient above-ground competition from neighbours by assessing red/far-red light ratio and to adjust morphology to avoid or reduce shading (Ballare et al., 1987, 1990; Ballare, 1999). Similarly, plants can use volatiles released from herbivore- or pathogen-attacked neighbours as a cue to induce their own defences before they are directly attacked (Shulaev et al., 1997; Baldwin et al., 2006; Heil & Silva Bueno, 2007). Mechanisms that enable plants to change their behaviour before adverse conditions have an impact upon them are likely to play a central role in plant adaptation. Previous findings suggest that the expression of stress-avoidance behaviour is more critical for plant performance in resource-poor than in resource-rich conditions (Ballare & Scopel, 1997). This may explain why species of nutrient-poor habitats appear, from the results of this study, to have evolved the ability to forage more effectively for rooting space than species of nutrient-rich habitats. More research on costs and benefits of various behaviours in different abiotic and biotic settings is needed to reveal the evolutionary forces that shape plant-foraging behaviour in different communities.

Acknowledgements

We gratefully acknowledge technical assistance from Kalle Kirsimäe and statistical advice from Alex J. Dumbrell and Jaan Liira. We thank Sonia Sultan and four anonymous referees for helpful comments on the manuscript. This study was supported by the University of Sussex, University of Tartu (0119) and grant 7576 from Estonian Science Foundation.

Ancillary