A test for hydrotropic behavior by roots of two coastal dune shrubs

Authors


Author for correspondence: E. Shelly Cole Tel: +1 (510) 642-6197 Fax: +1 (510) 643-5438 Email: cole@lifesci.ucsb.edu

Summary

  • • Root hydrotropism could be a means by which plants forage for limited and patchy distributions of soil water. While root hydrotropism has been induced in distinctly artificial conditions, it is unclear if it operates in natural settings. Here, we tested for this possibility in seedlings of two species of dune shrubs.
  • • Growth of individual roots in sand-filled observation chambers was monitored in response to moisture-rich patches and resultant soil water gradients. Chambers were designed so that roots could intercept the moisture gradients but not the moisture-rich patches simply through gravitropism.
  • • While up to 12% of the Eriogonum parvifolium roots grew into the moisture-rich patches, comparable root growth was observed in the control. None of the Artemisia californica roots grew into the patches.
  • • Thus, in a reasonable simulation of field conditions, we found no compelling evidence for hydrotropic root behavior in seedlings of these two dune shrubs. Our results leave the ecological significance of root hydrotropism in question.

Introduction

Heterogeneity in soil resources, varying at multiple temporal and spatial scales, is the norm rather than the exception (Stark, 1994). Fine-scale studies of soil nutrient concentrations in different ecosystems have found significant spatial variation in plant nutrients over a few meters to tens of centimeters (Jackson & Caldwell, 1993; Gross et al., 1995; Schlesinger et al., 1996; Farley & Fitter, 1999). Soil water can also be patchy at the centimeter scale, varying not only with depth (e.g. de Jong, 1979; Wan et al., 1995), but horizontally as well (Nobel et al., 1992; Dekker et al., 2001). Microtopography and the presence of plant canopies, rocks, animal burrows, root channels, cryptobiotic crusts, organic matter and roots of transpiring plants interact with a soil matrix of variably textured and structured constituents to influence water percolation, repellence, retention, and condensation (Salisbury, 1952; Brady, 1990; Nobel et al., 1992; Nobel, 1994; Yair et al., 1997; Dekker et al., 2001). These factors create fine-grained spatial heterogeneity of water availability at a scale relevant to both plants and individual roots. Within water-limited environments, acquisition of heterogeneously distributed soil water may be critical to survival and growth, and foraging may be a means by which plants can locate and take advantage of localized patches of water.

Plant foraging has been identified as ‘the processes whereby an organism searches, or ramifies within its habitat, which enhance its acquisition of essential resources’ (Hutchings & de Kroon, 1994). There are two components to foraging for patches of soil resources: (i) increasing the interception of (encounters with) resource-rich patches; and (ii) responding to intercepted patches by increasing occupation and/or utilization. Empirical studies addressing root responses to soil resource heterogeneity have focused on the latter, highlighting changes in root morphology, growth and/or physiology after a resource-rich patch has been intercepted (see reviews/syntheses by Robinson, 1994; Hodge, 2004; Kembel & Cahill, 2005). Much less is known about the mode of patch interception. Indeed, it is still unclear whether the interception process is active or passive. The idea that roots intercept patches by active searching, that is, detecting and growing towards greater resource availability, is prevalent in city planning, arborist and gardening literature (Smith, 1993; Mattheck & Bethge, 2000; Sydney Water, 2001; Paso Robles City, 2002), but the scientific literature is inconclusive regarding this possibility. Resource tropism, that is, enhanced growth towards places with greater resources, is a means by which roots could actively search for soil resource patches.

Various tropisms in roots have been examined in physiological studies. While we are unaware of any studies of nutrient tropisms, in addition to factors such as gravity (Sack, 1991; Boonsirichai et al., 2002), light (Iino, 1990) and touch (Fasano et al., 2002), directional root growth has been demonstrated in response to water (Takahashi, 1997; Eapen et al., 2005). However, such demonstrations of positive root hydrotropism (enhanced growth of roots towards places with greater available water) have been observed in distinctly artificial conditions that typically lack or minimize the effects of (i) gravity, (ii) soil and/or (iii) distance to a moisture-rich patch.

Many recent studies demonstrating hydrotropism have limited gravity or its effects. The root cap appears to be the site of both moisture and gravity perception, and it seems that responses to these two factors can interfere with each other (Jaffe et al., 1985; Takahashi & Suge, 1991; Coutts & Nicoll, 1993; Takahashi & Scott, 1993). While evidence does exist that hydrotropism can modify (Coutts & Nicoll, 1993) and overcome gravitropic root behavior in lateral and even primary roots (Darwin, 1881; Hooker, 1915; Oyanagi et al., 1995; Takahashi et al., 2003), most of the current work analyzing hydrotropism has used mutant individuals with agravitropic roots (Jaffe et al., 1985; Takahashi & Suge, 1991; Takano et al., 1995; Tsuda et al., 2003) or reduced the effects of gravity (i.e. microgravity, clinorotation or horizontal orientation) (Takahashi & Scott, 1991; Takahashi et al., 1999; Tsutsumi et al., 2003). Because all plants operate in the presence of gravity, it is unclear how these results apply to normal roots in a natural setting.

The majority of hydrotropism studies have measured the direction of root growth in air relative to a moist site, rather than in a more realistic medium (Hooker, 1915; Takano et al., 1995; Eapen et al., 2003; Takahashi et al., 2003). Roots growing in air do not encounter significant mechanical resistance or the complexities in diffusional vectors of water, oxygen and dissolved nutrients resulting from the irregular arrangement of a variety of materials in the three physical phases characteristic of soil. While there are a few hydrotropism studies that have examined root growth in vermiculite or soil, the results have varied, and in these studies, gravitropism has been artificially minimized (Tsutsumi et al., 2003; Tsuda et al., 2003) or roots have been placed directly in the moisture-rich soil (Loomis & Ewan, 1936).

It is not clear from existing hydrotropism studies that roots actively search for patches of soil water. In several studies, the ‘hydrostimulus’ (i.e. moist substrate, such as a block of agar with a high water potential) was positioned directly on the root tip (Takano et al., 1995; Takahashi et al., 1996) or roots were initially placed in moist soil and root behavior examined once they intercepted dry air or soil (Darwin, 1881; Loomis & Ewan, 1936; Jaffe et al., 1985; Coutts & Nicoll, 1993). While these methods allow for the observation of root responses to proximate moisture sources, they do not test the ability of roots to find moisture-rich locations. In other studies, the hydrostimulus or moist patch of soil is typically located only a few millimeters away from the root (Oyanagi et al., 1995; Eapen et al., 2003; Takahashi et al., 2003). It is hard to say if responses over such a small distance could constitute ‘searching’. Additionally, in air-culture conditions, this close root-patch proximity may create an artificial moisture gradient much steeper than that observed in field soil (Tsuda et al., 2003).

While a few hydrotropism studies have examined the behavior of roots of wild species in the presence of gravity (Hooker et al., 1915; Takahashi et al., 2003) and/or tested for hydrotropism in soil (Loomis & Ewan, 1936; Tsutsumi et al., 2003) and/or in response to centimeter- rather than millimeter-scale gradients in water (Hooker, 1915; Tsuda et al., 2003; Tsutsumi et al., 2003), none have included all of these factors to examine hydrotropism as a possible searching mechanism in conditions typical of the field. In this study, we attempt to address this issue by measuring root growth in soil, in the presence of gravity, in response to a moisture-rich patch located at centimeter-scale distances from roots. In addition, we chose to examine wild species from a habitat where water is a critical resource and hydrotropism may be most beneficial.

The coastal dunes of California experience a Mediterranean climate and represent a seasonally arid, resource-heterogeneous environment in which water often limits plant growth (Hesp, 1991; Alpert & Mooney, 1996). Although dune soils are relatively simple in content, the wetting and drying patterns in sandy soils are not uniform (Ritsema & Dekker, 1994; Alpert & Mooney, 1996; Yair et al., 1997; Dekker et al., 2001) and centimeter-scale heterogeneity in soil water exists (de Jong, 1979; Ritsema & Dekker, 1994; Dekker et al., 2001). For species unable to tap into the water table, acquisition of heterogeneously distributed soil water is potentially critical to growth. Thus, non-phreatophytic dune species are likely candidates for possessing hydrotropic root behavior. We selected two non-phreatophytic shrubs common to the stabilized dune region of southern and central California, Artemisia californica and Eriogonum parvifolium, and asked the following question: at the centimeter scale, does root hydrotropism operate in the presence of gravity in response to a moisture-rich patch in soil? We measured root growth of seedlings of these two species in sand-filled observation chambers under conditions of water limitation and patchy soil water availability to address this question in an ecologically relevant context.

Materials and Methods

Experimental setup

We grew seedlings of Artemisia californica Less. (Asteraceae) and Eriogonum parvifolium Smith (Polygonaceae) (California sagebrush and coastal buckwheat, respectively) in root observation chambers (Fig. 1) and monitored root growth. The chambers were c. 42 cm × 42 cm × 2.5 cm and constructed from grey polyvinyl chloride (PVC), with one clear Plexiglas side to allow for root observation. These chambers were held at a 45° angle from vertical to promote root growth along the clear side (Mahall & Callaway, 1991). To eliminate light infiltration to the roots and insulate the chamber, the Plexiglas was covered with an aluminum plate, and the entire chamber was then covered in a reflective bubble-wrap sleeve. Thirteen chambers for each species were set up on an outdoor wooden deck equipped with a rain shelter on the UCSB campus during the winter of 2003.

Figure 1.

Schematic of the root observation chamber. ‘Experimental regions’, grey areas; ‘chevrons’, stippled areas; locations of the soil water samples, circles; partitions for root counts, thick dashed lines. The treated chevron that received the injection of water is indicated by the darker stipple.

Seedlings were grown from seed collected from plants growing on stabilized dunes in Montaña de Oro State Park, Los Osos, CA, USA, in September 2003. After growing in pots for c. 11 wk, seedlings of each species were transplanted into the root chambers in matched pairs, two per chamber. At this time, most root systems extended no more than 10 cm in depth. The chambers were filled with 2/12 grit (c. 1–2 mm particle size) washed and kiln-dried Monterey beach sand (‘Lapis Lustre’, Cemex, Marina, CA, USA). The bottom 3.5 cm of the chamber was filled with coarser 12 grit sand (c. 1–5 mm particle size) (‘Silver Sand’, P.W. Gillibrand, Simi Valley, CA, USA) to encourage air flow and drainage. The two individuals per chamber were grown in opposite halves of the chamber, separated by a vertical waterproof barrier. In each half, roots grew downward past line ‘1’, and into the ‘experimental region’ of the chamber (9.5 cm × 9.5 cm) (Fig. 1). To the side of each experimental region, was a ‘chevron’ whose outer edge included an injection port, with an asymmetric PVC ‘arm’ to direct injections and roots inside the chamber (Fig. 1). For each chamber, one of the halves was randomly selected to serve as the ‘control’ and the other as the ‘treatment’. The control side contained sand in the chevron with a low moisture content similar to that in the experimental regions, while the treatment side had supplemental water injected to create a moisture-rich patch within the arms of the chevron. These arms maintained a moisture gradient perpendicular to gravity and prevented roots from intercepting the patch by vertical, gravitropic growth alone. Therefore, some degree of horizontal or diagonal root growth in the experimental region was required for roots to enter the chevron and the moisture-rich patch. Thus, if a greater percentage of roots entered the chevron in the treatment, as compared with the control, this could be interpreted as hydrotropism exceeding gravitropism.

Watering and fertilizing regime

During transplantation and the following week, the seedlings were watered lightly from the upper soil surface, and given a one-time nutrient addition of 5 ml of 1/6 strength modified Hoagland's solution (Epstein, 1972). After this, limited amounts of additional water were given to maintain a homogeneous background level of low soil moisture (c. 1% (g g−1)) in the chamber. Once roots had grown into the experimental region (Fig. 1, past line 1), the water treatment was administered by sliding a needle through the injection port and slowly syringing 3 ml of water along the length of the upper chevron arm up to the distal 1 cm. Pilot studies indicated that this method created a moisture-rich patch evenly distributed within the chevron. An additional 1 ml was injected in this manner, within 1–2 wk following the first injection to maintain the soil water treatment.

Determination of the distribution of soil water

To measure the distribution of water in the chambers, samples of sand were extracted from 10 positions in both the control and treated sides of the chambers (Fig. 1). Samples were collected from three chambers immediately following patch creation (within 4–11 d) to describe conditions roots initially encountered. Percent soil moisture (g g−1) of these sand samples was determined by weighing (Brady, 1990) before and after oven-drying at 105°C for 48 h. The same soil moisture determinations were made at the end of the experiment on the remaining 14 chambers. Soil samples were collected from these chambers once root growth had significantly diminished or when root growth extended beyond the experimental regions (Fig. 1, past line 2). Therefore, soil samples were collected on average 36 d after the patch creation; however, this ranged from 20 to 50 d after the initial injection, because of the variable growth rates of roots. One anomalous chamber was excluded from this analysis because of higher than average root densities, and thus much higher water depletion by the end of the experiment. Soil water potentials were estimated from the percent soil moisture measurements using a soil moisture release curve for the 2/12 grit experimental sand generated with a dew point hygrometer (WP-4 Dewpoint PotentiaMeter, Decagon Devices Inc., Pullman, WA, USA) by Decagon technical staff.

Determination of gradients in soil water

In each chamber half, percent soil moisture and water potential gradients were estimated across the experimental region and chevron in both vertical and horizontal planes to identify the ‘overall’ gradients present in the chambers. The magnitude of a vertical gradient was examined with a linear regression of the percent soil moisture or water potential values against distance in centimeters from the upper to the lower portion of the experimental region (Fig. 1, level A to level C), and the horizontal or chevron gradient was similarly estimated across the middle of the experimental region to the edge of the chevron (Fig. 1, B2 to CV1). Data from the soil samples collected immediately following patch creation and at the end of the experiment were combined for estimating average gradients during the experiment.

The spatial patterns of percent soil moisture and water potential among sampled positions were estimated using spatial interpolation utilities in a geographic information system (ArcView 3.3, Environmental Systems Research Institute (ESRI), Redlands, CA, USA). Interpolations in general can be sensitive to the number and distribution of sampling points as well as the assumptions of the interpolation technique. We selected a general purpose method for our estimations by using the default values of the spline interpolation tension method (ESRI, 1998).

Determination of nutrient distributions

To test for nutrient gradients in the chambers, six soil samples of 8 ml each, were collected adjacent to soil moisture sample positions in both the treated and control sides of the chamber within a subset of chambers (n = 4). Thirty-five milliliters of 0.5 m KCl solution were added to each of these samples and shaken for 1 h. The supernatant was then extracted through #1 Whatman filter paper and frozen at −4°C until laboratory analysis. Each of these extracts was tested for nitrate and ammonium as a general indicator of nutrient content by the Marine Science Institute Analytic Laboratory (UCSB, Santa Barbara, CA, USA) using a flow injection analyzer (Zellweger Analytics, Sunrise, FL, USA).

Root quantification

To measure the root growth responses to the moisture-rich patch and corresponding gradient in soil water, roots were traced every 3 d on acetate sheets placed on the viewing windows of the chambers. For quantification purposes, we considered a single root to be any segment of the root system with a root tip. Therefore, all types of roots present in the experimental regions (primary and lateral roots) were included in our analysis. In both species, vertically oriented roots were the most common, but horizontally and diagonally oriented roots were also present. The percentage of roots growing into the chevron in each side of a chamber was calculated by the following formula (see Fig. 1 for details): (total number of roots crossing the ‘chevron line’/total number of roots observed in the ‘experimental region’) × 100. It is worth noting that we used the number of roots growing into the chevron rather than the total number in the chevron; therefore, if proliferation took place in response to greater soil moisture in the treated chevron, it would not influence the calculation. Data presented here are from the three chambers sampled immediately following patch creation and roots monitored during the following 18–21 d. Measurements taken at times later than 21 d did not meaningfully change the results. During the c. 3 wk time period following patch creation, roots maintained growth and had not extended notably beyond the experimental regions (i.e. > 20 mm of root length or > 6 d past line 2; Fig. 1). Because the roots of one A. californica individual grew rapidly downward, data recorded 12 rather than 18–21 d post-treatment were used for that chamber.

Statistics

Student's paired t-tests were performed to test for significant differences between the treated and control sides of the chambers in percent soil moisture, water potential and nutrients. Log10 transformations were made to the absolute water potential values to meet statistical test assumptions. Paired t-tests were also used to compare horizontal and vertical gradients in water within a chamber side. To determine if nutrients varied spatially, a one-factor anova (location as the factor) was performed within each side of the chamber. None of the standard transformations we attempted successfully normalized the heavily skewed distribution of root data; therefore, a Wilcoxon signed-rank test was used (Quinn & Keough, 2002) to compare the percentage of roots in the treated and control chevrons within each chamber for each species. Additional comparisons of chevron root growth between the treated and control sides of the chambers were made using only roots from subsections of the experimental regions. In some cases, roots did not grow into these subsections in both sides of the chamber, and therefore a Wilcoxon rank sum test was performed rather than a paired analysis to avoid omission of the ‘unpaired’ data. Statistical analyses were performed using the JMP statistics program (JMP 5.1.1 SAS Institute Inc., Cary, NC, USA). Wilcoxon test results were compared with statistical tables (McCornack, 1965; Snedecor & Cochran, 1989) rather than using the probability approximations appropriate to larger sample sizes (McCornack, 1965; Snedecor & Cochran, 1989; Quinn & Keough, 2002) generated by JMP. Results are considered significant where P  0.05.

Results

Nutrients

No significant differences in nitrate, ammonium or total nitrogen were found between the treated and control sides of the chambers at each location (Student's paired two-tailed t-tests, P ≥ 0.20, d.f. = 3) or among locations within a chamber side (one-factor anova, P ≥ 0.30, d.f. = 21), indicating that nitrogen distributions were homogeneous irrespective of water treatment. The total inorganic nitrogen present in the chambers was low and averaged 2.48 ± 0.12 p.p.m.

Percent soil moisture and water potential

A moisture-rich patch of soil was successfully created and maintained under conditions of limited water availability elsewhere in the chambers. Immediately following the patch creation and at the end of the experiment, percent soil moistures were on average 3.3- and 2.4-fold greater, respectively, in the treated chevrons than in the experimental regions of the chamber (Table 1), while soil water potentials were 4.4- and 3.9-fold greater (Table 2), respectively. Within the experimental regions, percent soil moistures and water potentials were low and similar between the treated and control sides throughout the experiment (Tables 1 and 2). However, at one location near the chevrons (B1), percent soil moistures and water potentials were consistently higher in the treated sides of the chambers (Tables 1 and 2). Although this difference was not significant at the P  0.05 significance level, it was indicative of the water gradient generated by the moisture-rich patch.

Table 1.  Average percent soil moisture (g g−1) in the treated and control sides of the chambers at each sampling location immediately following patch creation and at the end of the experiment Thumbnail image of
Table 2.  Average soil water potential (MPa) in the treated and control sides of the chambers at each sampling location immediately following patch creation and at the end of the experiment Thumbnail image of

Higher soil moistures in the treated chevrons created a strong horizontal soil moisture gradient (Fig. 2). From the initial chamber watering, a downward, vertical gradient in percent soil moisture was also present, but it did not significantly differ among the treated and control sides of the chambers, and it was on average approx. one-third that of the horizontal gradient towards the chevron in the treated sides of the chambers (Fig. 2). The strongest segment of the horizontal gradient was at the interface between the moisture-rich patch and the experimental region, creating a 0.37% cm−1 change in soil moisture between the Cv2 and the B1 sampling locations (Figs 2, 3). This gradient declined with horizontal distance (Fig. 2). Based on the spatial interpolations (Fig. 2) and the gradient estimates between the Cv2 and A1 and B1 locations (Fig. 3), the influence of the moisture-rich patch extended into the experimental region, creating a strong soil moisture gradient at least 2–3 cm beyond the chevron line.

Figure 2.

Percent soil moisture patterns in the treated and control sides of chambers. Contoured shading represents the interpolation estimates of percent soil moisture and dots indicate the center of sampling locations for soil water. Average ‘overall’ percent soil moisture gradients (Δ% cm−1) measured in the experimental regions ± 1SE are reported above the arrows indicating the gradient type (horizontally towards the chevron, vertically downward). Letters indicate significant differences between the gradients found in Student's paired two-tailed t-tests at the P  0.05 level of significance, d.f. = 15. For both the interpolation estimates and gradient analyses, n = 16.

Figure 3.

Average gradients in percent soil moisture in the treated sides of the chambers at the A1 and B1 locations. The gradients towards the Cv2 location in the chevron (horizontal or diagonal) and downward towards either the B1 or C1 location (vertical) are reported (n = 16). Bars indicate ± 1SE. Letters indicate significant differences between the gradients at each location found in Student's paired two-tailed t-tests. At the A1 location: t15 = −6.18, P < 0.0001; at the B1 location: t15 = −5.88, P < 0.0001.

The overall vertical gradient in soil water potential was on average approximately double that of the horizontal gradient in the treated side of the chambers (Fig. 4). This pattern appears to be largely driven by the low water potentials in the upper experimental region (Fig. 4). At a finer scale, a strong gradient in soil water potential towards the moisture-rich patch was observed proximal to the treated chevron. In the upper portion of the experimental region, at the A1 location, the downward vertical gradient towards the B1 location (3.67 MPa cm−1) was similar to the diagonal soil water potential gradient towards the patch (3.54 MPa cm−1) (Fig. 5). However, in the mid-experimental region, B1 location, the soil water potential gradient toward the treated chevron was several-fold greater horizontally towards the moisture-rich patch (1.83 MPa cm−1) than vertically towards the C1 location (0.71 MPa cm−1) (Fig. 5).

Figure 4.

Soil water potential patterns in the treated and control sides of chambers. Contoured shading represents the interpolation estimates of soil water potential and dots indicate the center of sampling locations for soil water. Average ‘overall’ soil water potential gradients (Δ MPa cm−1) measured in the experimental regions ± 1SE are reported above the arrows, indicating the gradient type (horizontally towards the chevron, vertically downward). Letters indicate significant differences between the gradients found in Student's paired two-tailed t-tests at the P  0.05 level of significance, d.f. = 15. For both the interpolation estimates and gradient analyses, n = 16.

Figure 5.

Average gradients in soil water potential in the treated sides of the chambers at the A1 and B1 locations. The gradients towards the Cv2 location in the chevron (horizontal or diagonal) and downward towards either the B1 or C1 location (vertical) are reported (n = 16). Bars indicate ± 1SE. Letters indicate significant differences found between the gradients at each location in Student's paired two-tailed t-tests. At the A1 location: t15 = 0.28, P = 0.783; at the B1 location: t15 = −4.13, P < 0.001. Note that the large SE bars at the B1 location reflect the absolute differences in gradient values between chambers. Within a chamber, the difference between gradients was highly significant.

Root growth

The roots in the experimental regions encountered relatively low soil water contents. The percent soil moisture and water potential values here ranged from an average of 0.4 to 1.5% and −33.7 to −5.1 MPa, while in the treated chevron these ranged from 2.0 to 3.9% and −4.0 to −2.0 MPa (Tables 1 and 2). Under these conditions, roots grew into the experimental region in both sides of the chamber in eight of the 13 A californica and nine of the 13 E. parvifolium chambers. Initial root growth resulted in roots being located throughout a given experimental region, ranging from a few mm to 9.5 cm away from the chevrons. In the majority of chambers, roots grew within 2 cm of the treated ‘chevron line’ (Fig. 1) and therefore encountered the strong horizontal gradient in soil water created by the moisture-rich patch. However, none of the A. californica individuals grew roots into the treated chevrons, although two individuals grew roots into the control chevrons (Fig. 6a). Within 3 d of patch creation, two E. parvifolium individuals grew roots into the treated chevrons. Root growth of these individuals was not followed for > 3 d after the response since soil samples were collected from their chambers to measure the distribution of soil water at the time root responses were first observed. Only one other E. parvifolium individual grew roots into the treated chevron, eventually locating 12% of its roots there. However, similar root growth was observed in the control chevrons in two other individuals (Fig. 6a). Therefore, considering all roots within the experimental regions, no significant difference in the percentage of roots growing into the chevrons was found between the treated and control chevrons in either species (Fig. 6a) (Wilcoxon signed-rank test, A. californica: T7 = 15.0, P  > 0.70; E. parvifolium: T8 = 14.5, P > 0.40).

Figure 6.

A comparison of the percentage of study species’ roots from the entire experimental region (a), within 2 cm of the chevron line (b) or within 2 cm and intercepting the B1 location (c) that entered the treated vs. control chevrons ∼3 wk after the patch was created. No significant treatment effect was found in either species based on a Wilcoxon signed-rank test (a) and Wilcoxon rank sum tests (b, c) at the P  0.05 level of significance. Bars indicate ± 1SE. Values in parentheses: total number of individuals, total number of roots. Note that data include responses before c. 3 wk for one Artemisia californica and three Eriogonum parvifolium chambers since they were completed earlier either to capture the distribution of soil water immediately following the patch creation or because roots had grown past line 2.

Because the soil water gradient created by the patch was strongest in close proximity to the treated chevron (Figs 2, 4), we compared the percentage of roots crossing into the treated and control chevrons using only roots in the experimental regions that grew within 2 cm of the chevron lines. A total of 14 roots from five A. californica individuals and 59 roots from nine E. parvifolium individuals grew into this region in the treated side of the chambers (Fig. 6b). Differences between treatments and controls revealed no evidence for hydrotropic root behavior (Fig. 6b; Wilcoxon rank sum test, A. californica: T8 = 25, P > 0.05; E. parvifolium: T13 = 47.5, P > 0.05; note that from published statistical tables, rank sums (T) could be compared with critical values to determine significance at the P 0.05 level, but probability levels could not be derived from T).

We further refined this comparison to include only those roots that grew into the B1 location, where they encountered a strong gradient in both soil water potential and soil moisture. This refinement further reduced our sample size. From the B1 location, E. parvifolium roots grew into the treated chevron in two of six chambers, and none grew into the control chevrons in three chambers, suggesting a possible weak, positive response (Fig. 6c). However, the percentage of roots in the treated chevrons did not significantly differ from the control (Wilcoxon rank sum test, T7 = 12, P > 0.05). Because none of the A. californica roots had grown into the treated chevrons, there was no evidence for hydrotropism in this species; however, it should be noted that root density in the treated side of the A. californica chambers was more limited (six roots of four individuals) than observed in E. parvifolium (18 roots of six individuals) (Fig. 6c).

Discussion

Hydrotropism has been observed in distinctly artificial conditions that have minimized the effects of gravity, soil and distance to the hydrostimulus or moist patch of soil. In this study, we successfully created a moisture-rich patch, and roots encountered a corresponding centimeter-scale gradient in soil water in a controlled environment that was realistic in substrate, gravity and moisture conditions to the field. Because hydrotropism may be particularly important when water availability is limited as well as heterogeneously distributed, we tested for hydrotropism in our chambers under relatively dry conditions. None of the roots of A. californica grew into the treated chevron towards greater moisture and water potential in the patch. While we did observe a small percentage of E. parvifolium roots growing towards the moisture-rich patch, the response did not significantly differ from the control or a zero response. Therefore, in these species, we did not find persuasive evidence for hydrotropic root behavior under field-like conditions when water was limiting.

Seventy years ago, the work of Loomis and Ewan (1936) also questioned the applicability of air-culture studies and tested for hydrotropic root behavior in the primary roots of 29 genera of primarily cultivated species under normal gravity in response to a relatively steep moisture gradient in soil (estimated as c. 2.5% cm−1 at the moist/dry soil interface). However, in their study, distance to the moist soil was minimized and roots did not need to detect and grow towards soil water. With the advantage of placement in close proximity to, or actually in, the moist soil, they observed responses interpreted to be hydrotropism in the radicles of several cultivated species, but concluded that ‘hydrotropism is not universal and probably under field conditions, not a common plant response’. Our findings are not at variance with this conclusion.

Conditions encountered by roots

It appears that the roots in our study encountered a reasonable simulation of field conditions when water availability is limited. Total inorganic nitrogen was low in our experimental chambers and similar to averages of field measurements for dune soils (Olff et al., 1993; Alpert & Mooney, 1996; Cain et al., 1999). While, to our knowledge, soil water potentials have not been extensively measured in California sand dunes (de Jong, 1979), there are some published measurements of soil moisture in dunes of California (Purer, 1936; de Jong, 1979; Alpert & Mooney, 1996) and other Mediterranean climates (Zencich et al., 2002). With the exception of de Jong (1979), these measurements are similar to our chambers, with background amounts similar to dry season measurements and treated chevrons similar to field averages during the rainy season.

The moisture-rich patches created soil moisture and water potential gradients that extended horizontally several centimeters into the experimental regions of our chambers. Roots growing within a few centimeters of the treated chevron encountered these gradients. Here, the estimated horizontal gradient in soil water was 0.37% cm−1 and 1.83 MPa cm−1 towards the treated chevron. In coastal dunes, we estimated linear gradients in soil moisture from published fine-scale measurements. Below the surface 10 cm of soil, gradients ranged from 0.007% cm−1 to 0.8% cm−1 (Purer, 1936; de Jong, 1979; Ritsema & Dekker, 1994; Alpert & Mooney, 1996; Dekker et al., 2001). It appears our study species therefore encountered gradients in the chambers plausible in their natural habitat.

Possible controls on hydrotropism

The experimental gradients in soil water did not induce hydrotropism in either A. californica or E. parvifolium, although our conclusions regarding these results are based on only a small number of roots encountering a horizontal gradient in soil water potential (Fig. 6c). The exact cue that triggers hydrotropism, however, remains unknown (Takahashi, 1997; Eapen et al., 2005). If soil moisture rather than soil water potential is the cue, an ample number of roots of both A. californica and E. parvifolium encountered the stimulus and did not display significant hydrotropic root behavior (Fig. 6b).

It is possible that the soil water gradients typical of the field are not strong enough to induce hydrotropic root behavior sufficient to overcome gravitropism. In air-culture studies, the strength of the water potential or water vapor gradients has been shown to influence the degree of root curvature towards the wet surface (Hooker, 1915; Takahashi & Scott, 1991; Takahashi & Suge, 1991; Takahashi & Scott, 1993; Oyanagi et al., 1995; Hirasawa et al., 1997) with more gravitropic roots requiring greater moisture gradients to overcome gravitropism (Takahashi & Scott, 1993; Takahashi, 1994). For the induction of hydrotropism, a minimum gradient of 0.3 MPa mm−1 has been reported (Tsuda et al., 2003). This is considerably steeper than those in our study and possibly the field (Tsuda et al., 2003). Two recent studies (Tsuda et al., 2003; Tsutsumi et al., 2003) reported hydrotropic root behavior in response to weaker moisture or water potential gradients in soil; however, the counteracting effects of gravitropism were removed or limited.

Recent work by Takahashi et al. (2003) suggests a physiological mechanism by which exposure to a moisture gradient or water stress may partially release roots from gravitropic behavior, allowing them to respond to gradients in water availability. Roots encountering the soil water gradient in our study were exposed to a moisture gradient and low soil water potentials (−7 to −12 MPa mid-region) but did not behave hydrotropically. While our conditions were dry relative to many hydrotropism studies (Coutts & Nicoll, 1993; Takano et al., 1995; Tsuda et al., 2003), they were similar to air-culture studies that have demonstrated hydrotropism in roots exposed to relative humidities (83–94%) which would have produced low water potentials (−25 MPa to −8 Mpa; Takahashi & Scott, 1991, 1993; Takahashi & Suge, 1991; Oyanagi et al., 1995; Takahashi et al., 2002).

Hydrotropism as a foraging mechanism

It appears that the roots of A. californica and E. parvifolium seedlings may not employ root hydrotropism to actively search for and increase interception of moisture-rich patches in soil, one of the two components of foraging. To our knowledge, no other studies have directly tested for such active searching in response to patchy resource availability in an ecological context. St. John et al. (1983) did raise the possibility of roots using a ‘locational mechanism’ to intercept nutrient-rich patches. Based on their correlative work, they suggested that, instead, through ‘random’ growth, roots of their study species passively intercepted nutrient-rich patches. We hypothesize that the roots of our selected species encounter moisture-rich patches in the dune environment serendipitously by passively following a genetically conferred architectural design (Weaver, 1919; Purer, 1936; Canadell & Zedler, 1995). Once moist patches are intercepted, root foraging mechanisms, such as selective aquaporin expression (see North & Nobel, 2000) and root proliferation, may facilitate patch utilization, although further research is necessary to determine exact root responses.

In conclusion, little is known about resource-tropic behavior in the context of soil resource heterogeneity. Conceptually, the idea that roots forage for soil water through hydrotropic root behavior is appealing. However, the artificial experimental conditions that have elicited hydrotropism in previous studies may not pertain to most natural conditions. In this study, we specifically chose dune scrub species, because the characteristics of their environment suggested a strong advantage for hydrotropism. In field-like conditions, we created water gradients roughly perpendicular from the direction of gravity, so that hydrotropism could be differentiated from gravitropism, since the possible role of hydrotropism under natural conditions would be severely limited were it always dominated by gravitropism. Under these conditions, we found no compelling evidence for hydrotropic root behavior. Whether it occurs in other species or under other natural circumstances must await future research.

Acknowledgements

We thank J. Gremer for her hard work and dedication to the chamber study and J. Kim for assistance with pilot work. This study benefited from our encouraging discussions with M. Moritz and D. Bush, and constructive comments on the manuscript by M. Moritz, J. Schimel and E. O’Brien. We would also like to thank the Andrew W. Mellon Foundation for financial support of this work, V. Cicero and the California Department of Parks and Recreation for permission to collect seeds, E. Boyer for support at UC Berkeley and D. Cobos, T.-J. Clevenger and B. Teare of Decagon Devices who generously conducted the soil water potential measurements. This work is in partial fulfillment of the requirements of a PhD at the University of California, Santa Barbara.

Ancillary