Structure, attachment properties, and ecological importance of the attachment system of English ivy (Hedera helix)

Abstract

Root climbers such as English ivy (Hedera helix) rely on specialized adventitious roots for attachment, enabling the plants to climb on a wide range of natural and artificial substrates. Despite their importance for the climbing habit, the biomechanical properties of these specialized adventitious roots compared with standard roots and their performance in the attachment to different host species or inert substrates have not been studied. Here organs and tissues involved in the attachment are characterized and their significance in regard to a broader functional and ecological aspect is discussed. Depending on the substrate, the root clusters show different types of failure modes at various frequencies, demonstrating the close interaction between the climber and its substrates. With a Young’s Modulus of 109.2 MPa, the attachment roots are relatively stiff for non-woody roots. The central cylinders of the attachment roots show a high tensile strength of 38 MPa and a very high extensibility of 34%. In host trees naturally co-distributed with English ivy, a ‘balanced’ occurrence of failure of the attachment system of the climber and the bark of the host is found, suggesting a co-evolution of climber and host. Maximum loads of root clusters normalized by the number of roots match those of individually tested attachment roots. In comparison with most subterranean roots the properties and structure of the attachment roots of English ivy show distinct differences. There exist similarities to the properties found for roots of Galium aparine, suggesting a trend in not fully self-supporting plants towards a higher extensibility.

Attachment system, biomechanics, English ivy, Hedera helix, maximum strain, root climber, substrate, tensile strength, Young’s Modulus

Introduction

English ivy (Hedera helix L., Araliaceae) is an evergreen root-climbing liana which has been used for centuries as greenery. In its Old World natural habitat, H. helix typically grows in gallery forests (Schnitzler, 1995) but behaves as an invasive species in the New World (Larocque, 1999). Recent interest in the protective and insulating properties of house greenery (Stec et al., 2005; Wong et al., 2010) has brought the permanent attachment systems of climbing vines and lianas also into the focus of applied sciences. English ivy has an adaptable attachment system which allows the plant to climb on various substrates such as tree barks, rocks, and mortar. English ivy develops specialized unbranched adventitious roots at the substrate-facing side of its shoots. If the shoots are in contact with soil, typical nourishing subterranean branched roots are developed. The unbranched attachment roots enable English ivy to attach to the substrate and to climb to heights of up to 30 m, provided that appropriate conditions of light, moisture, and touch stimuli are provided (Bruhn, 1909; Negbi et al., 1982; Metcalfe, 2005). The mechanisms involved in this permanent attachment system were recently described (Melzer et al., 2010) and consist of four different phases: contact formation; form closure of the root with the substrate; chemical adhesion by excreting glue from the root hairs which densely cover the attachment roots; and a passive shape change of the root hairs.

Biomechanical studies on subterranean roots and root systems have been of interest for a long time regarding their implications on slope stabilization and soil reinforcement (Watson et al., 1999; Norris, 2005; Reubens et al., 2007), their importance to reduce the risk of lodging in cereals (Baker et al., 1998; Berry et al., 2004; Oladokun and Ennos, 2006), their ability to contract and to pull plants partly into the soil (Pütz, 2006; Schreiber et al., 2010), and in the context of ecological influences on root structure (Niklas et al., 2002; Wang et al., 2009; De Micco and Aronne, 2010). However, very little is known about the biomechanical properties of aerial attachment roots. Recently Steinbrecher et al. (2010) presented a novel approach to quantify the attachment strengths of climbing plants, in which first data on the attachment roots of H. helix were shown. The main focus of their study was on the interpretation of the resulting force–displacement curves. The present study focuses on the biomechanics of the attachment system of English ivy and its modes of failure in relation to the respective substrates. As shown earlier (Melzer et al., 2009), English ivy is able to climb effectively only on structured surfaces such as wood, cork, or mortar, whereas its attachment system does not work on most smooth surfaces such as glass and aluminium. An exception is the smooth surface of Mylar foil where attachment can also be achieved by English ivy (Melzer et al., 2010). In this study, data are presented on the attachment properties of root clusters tested via tensile tests on a range of structured artificial, semi-artificial, and natural substrates. These results were compared with tensile tests and detachment tests of isolated attachment roots in the laboratory. Finally the results are compared with those known for subterranean anchoring systems (Coutts, 1983; Ennos, 1989, 1990; Commandeur and Pyles, 1991; Ennos et al., 1993a, b; Gartner, 1994; Speck et al., 1998; Mickovski and Ennos, 2003; Speck and Spatz, 2003; Goodman, 2005).

Materials and methods

Plant material and substrates

All tested H. helix plants were cultivated in the Botanic Garden Freiburg, Germany. In the sampled shoots the attachment onto the different substrates was fully established Table 1. The substrates were grouped into three categories: (i) artificial (mortar); (ii) semi-artificial (cork); and (iii) natural (tree bark) substrates.

Table 1.

Substrates from which the attached English ivy shoots were displaced during the tests

Substrate	No. of test samples	Degree of bark structuring	Natural distribution	Natural co-distribution with Hedera helix
Mortar 1 (facing north)	43
Mortar 2 (facing west)	31
Cork	187
Acer rubrum	12	Medium	North America	No
Aesculus hippocastanum	12	Smooth	Balkans	Yes
Amelanchier lamarckii	14	Smooth	North America	No
Corylus avellana	8	Smooth	Europe, Asia Minor	Yes
Ginkgo biloba	13	Medium	East Asia	No
Picea engelmannii	25	Rough	North America	No
Prunus serrulata	11	Smooth	East Asia	No

The plants growing on the artificial substrate (mortar) were sampled from two different sites. At both sites the plants were attached to walls plastered with regular mortar. One of the walls was facing north, the other was facing west. At both locations tensile tests were carried out in situ. Pieces of mortar from the climbing sites were collected for tensile tests of the substrate in the laboratory.

The specimen climbing on cork (sensuJunikka, 1994) were purchased H. helix. plants cultivated outdoors in pots equipped with climber racks. Cork boards were glued onto PVC plates and mounted into the frames of the climber rack where they were grown over by the plants during a 12 month period. For the tests, the attached climbing shoots were cut. Next the PVC–cork plates were demounted. For the mechanical tests, these plates were mounted indoors onto vertical wooden boards. The shoots were then defoliated, and all cut surfaces were sealed with Vaseline to prevent desiccation. The detached ivy shoots were wrapped in moist paper towels until subsequent testing.

Shoot segments of English ivy attached to the outer barks (sensuJunikka, 1994) of seven tree species were tested. These trees were growing outdoors in the Botanic Garden Freiburg, and the tensile tests were carried out at the growing sites in situ. The barks were classified into three categories following their macroscopic qualitative structuring, namely smooth, medium, and rough (see also Bauer et al., 2010). Two tree species were selected that are naturally co-distributed with English ivy (Coryllus avellana L. and Aesculum hippocastanum L.) and five species that are not, but come from different regions where English ivy is an invasive plant (Acer rubrum L., Amelanchier lamarckii F.G.Schroed., Picea engelmannii Parry ex Engelm., Prunus serrulata Lindl., and Ginkgo biloba L.). The five hardwood species belong to different angiosperm groups; two species (G. biloba and P. engelmannii) represent the two gymnosperm lineages occurring in a temperate climate. The natural distribution areas of the trees are listed in Table 1.

The samples for the individual root tests were taken from an ivy plant growing in the Botanic Garden attached to the outer bark of a P. engelmannii tree.

Mechanical testing I: root clusters

A custom-made mobile testing machine which allows the simultaneous recording of displacement and force was used to test root clusters (Steinbrecher et al., 2010). The testing machine was equipped with a tension–compression load cell with a maximum load of 20 N (model 8523-20, Burster Praezisionsmesstechnik GmbH and Co KG, Gernsbach, Germany) and a potentiometric displacement transducer with a range of 0–50 mm (model 8712-50, Burster Praezisionsmesstechnik GmbH and Co KG, Gernsbach, Germany). For a detailed description of the testing device, see Steinbrecher et al. (2010). The testing machine was adjusted to apply forces normal to the selected shoot to ensure comparable results. A pair of lockable tweezers was placed in the middle part of the ivy shoot segment over the cluster of attachment roots. After fastening the tweezers, the shoot was cut above and below the cluster of attachment roots. Then the tensile test was conducted and the force–displacement data were recorded to derive the force necessary to detach the root cluster.

For samples which showed failure of the roots, the average maximum detachment force per single attachment root was calculated by dividing the maximum force by the number of attachment roots in the root cluster. It was assumed that all roots were attached to the substrate, which holds true in good approximation as visual inspection proved. According to the force–displacement curves, the root clusters were classified into three categories: (i) clusters showing a sudden failure (cf. Fig. 5A.1); these data were used for comparison with the tests on isolated roots; (ii) clusters showing few preliminary failure events of the root cluster before complete failure (cf. Fig. 5A.2); and (iii) clusters with several distinct preliminary failure events and a stepwise force decline until complete failure (cf. Fig. 5A.3).

The tested samples were fixated with FAA (32% formaldehyde, 63% ethanol, 5% acetic acid) for the analysis of failure modes as described below.

Mechanical testing I: analysis of failure modes

The tested samples were classified according to the mode of failure. Three different basal types of failure modes were discerned: (i) failure of the substrate; (ii) failure of the attachment roots; and (iii) failure of the shoot (Fig. 1). Mixed modes of failure were classified accordingly. In addition, the number of roots per tested root cluster was recorded.

Fig. 1.

Basal modes of failure. Schematic drawing of shoot segments of English ivy with one attachment root each. The arrows indicate the direction of the force during the tests. (A) Failure of the substrate; (B) failure of the root; (C) failure of the shoot. ar, attachment root; cc, central cylinder; st, substrate; rb, root cortex; s, shoot. (This figure is available in colour at JXB online.)

Statistical tests were carried out with R v2.11.1. The testing routine used is given in the Results for each set of data. For scanning electron microscope (SEM) analyses, a LEO 435 VP SEM (Leica, Wiesbaden, Germany) microscope was used.

Mechanical testing II: individual roots

All tests on individual attached roots, isolated roots, central cylinders, and pieces of mortar were carried out on a modified microtensile testing machine (for details, see Burgert et al., 2003) equipped with a compression–tension load cell with a maximum load of 10 N and a resolution of 1 mN (model 31E, Honeywell, Columbus, OH, USA). Two aluminium platelets on opposite sides were used, and the samples were glued onto these supports. The platelets were fitted with holes for easy mounting onto the tensile apparatus via a pinhole assembly (Fig. 2).

Fig. 2.

Tensile testing of individual attachment roots. (A) Schematic drawing of the test set-up. The arrow indicates the direction of displacement. (B) Mounted sample, scale bar: 2 mm. ap, aluminium platelet; ca, cyanoacrylate adhesive; ct, compression–tension load cell; ir, isolated root; lt, linear table.

For sample preparation, the position of the aluminium platelets was fixed using sticky tape. The ends of the samples were glued on one of the platelets by a rapid cyanoacrylate adhesive (UHU Sekundenalleskleber Gel, UHU GmbH and Co. KG Bühl, Germany). For glue hardening and storage, the samples were transferred into a moist chamber with a humidity of >95%. After glue hardening the platelets were mounted onto the tensile apparatus, the sticky tape was removed, and the displacement tests were conducted.

Individual attached roots were tested in different arrangements (cf. Fig. 5B, C). Samples of ivy shoots, with individual attachment roots still fastened to pieces of tree bark were carefully detached from the tree, and the bark pieces and the shoot segment were each glued onto one of the aluminium platelets. By this procedure it was ensured that there was no contact between glue and attachment root that could influence the results. Parts of the samples were tested in the direction normal to the bark surface (cf. Fig. 5B) and parts parallel to it (cf. Fig. 5C).

For testing isolated intact roots (cf. Fig. 5D), the roots were removed from shoot and substrate, and both ends were glued onto the aluminium platelets, ensuring a straight and undamaged piece of root between the platelets without any contamination due to the gluing. The roots were carefully arranged parallel to the tension forces to ensure an even strain field over the diameter of the root. Due to the distinct morphological separation into cortex and central cylinder, the central cylinder was often found to rupture within the root cortex during the tensile tests and it could not be excluded that the central cylinder was pulled out of the cortex tissue. Therefore, only Young’s Modulus and tensile strength were calculated from the initial slope of the force–displacement curves of the isolated intact roots, but no maximum strain could be derived.

Central cylinders of attachment roots were prepared by removing the root cortex. Afterwards they were glued onto aluminium platelets and tested in the same way as described for the isolated roots (cf. Fig. 5E).

Images taken before the start of the tensile tests allowed measurement of the diameter of the intact isolated roots and of the central cylinders. These data were used to calculate the cross-sectional area. The original sample length was also recorded. Due to the plastic deformation of the plant tissues as a result of the tensile tests, it was not possible to measure directly the initial area fractions of cortex to central cylinder in the tested root samples. Therefore, the average area fractions of the two tissues in cross-sections of 12 fresh attachment roots were calculated and used for estimating the mechanical properties of the root cortex quantitatively.

The images were taken with a Olympus SZX7 dissecting microscope via an Altra20 (Olympus Europa GmbH, Hamburg, Germany) digital camera and the software cell A 2.8. The measurements were conducted with the software ImageJ 1.41o.

The recorded force–displacement data of the tested isolated intact roots and of the central cylinders were used to assess root strength, breaking stress, and Young’s Modulus. The intact isolated roots, like the central cylinders, can be considered in good approximation cylindrical over the testing distance, with an original cross-sectional area (A₀) calculated from the original unstrained diameter. The strain (ϵ) was calculated (1) as the fraction of the displacement (ΔL) divided by the original sample length (L₀).

(1)

The stress (σ) was calculated (2) as the force (F) per original cross-sectional area (A₀) of the intact root or central cylinder, respectively.

(2)

The breaking stress (σ_max) is given by (3) the root strength (F_max) per original cross-sectional area (A₀).

(3)

The Young’s Modulus for the intact attachment roots (E_root) and the Young’s Modulus for the central cylinders (E_cc) is a measure of material stiffness and is calculated (4) from the slope of the initial elastic part of the respective stress–strain curves (σ/ϵ).

(4)

The Young’s Modulus of the root’s cortex (E_cortex) is derived from the ‘isostrain rule of mixture’ for composite materials, that describes the combined Young’s Modulus for the intact attachment root (E_root) as the sum of E_cc times V_cc plus E_cortex times V_cortex. E_cortex and V_cortex are the volume fraction and Young’s Modulus of the cortex, and E_cc and V_cc the volume fraction and Young’s Modulus of the central cylinder (Rammerstorfer, 1994; Speck et al., 1996).

(5)

Means and standard deviations were calculated with EXCEL 2007; further statistical analyses were carried out with the software R v2.11.1.

Results

Analysis of failure modes

The attachment interface did not fail in any of the tests. Tensile tests from the two tested mortars showed no significant differences either in the maximal displacement forces (U-test, W=385.5, P-value=0.2475) or in the distribution of failure modes (Pearson’s χ² test with simulated P-value—based on 2000 replicates—χ²=0.9129, df=NA, P-value=1). Therefore, the data of the two sets were pooled. The tests on the artificial substrate (mortar) showed 94.6% (n=70 of 74 tests) substrate failure, 1.4% (n=1) shoot failure, and 4.1% (n=3) of the mixed modes of failure in shoot and root (Fig. 3A). In the by far dominant case of substrate failure, macroscopic to microscopic pieces of substrate remained on the surface of the attachment roots.

Fig. 3.

Comparison of failure modes of attachment between different substrate groups. The different modes of failure are shown on the x-axis and the relative frequencies of the failure modes per substrate are shown on the y-axis as a percentage. (A) Results from tests on mortar (pooled data for two types of mortars). (B) Results from tests on cork. (C) Results from tests on tree barks (pooled data for all bark species).

While the shoots of English ivy usually developed two rows of attachment roots on mortar and tree bark, shoots of English ivy climbing on cork were often anchored by up to five rows of attachment roots. The total number of roots per root cluster between the other substrates and cork were not significantly different (Wilcoxon rank sum test with continuity correction, W=3491, P-value=0.3222) and varied around a median of 38 [interquartile range (IQR) 24, n=191], meaning that the root clusters on cork were shorter than those on the other substrates.

The samples pulled from the semi-artificial substrate (cork) showed a more diverse distribution of failure modes. All seven possible types of basal failure modes and mixed failure modes were observed. Of the 187 displacement tests from this substrate, 32.6% (n=61) showed a mixed failure of root and substrate, in 30.5% (n=57) the shoot failed, in 20.9% (n=39) the substrate failed, in 13.4% (n=25) the root failed, and in 2.8% (n=5) of the tests mixed modes of shoot failure in combination with root failure and/or substrate failure occurred (Fig. 3B).

Pooled data of the 95 displacement tests carried out on the natural substrate (tree bark) showed failure of the substrate in 51.6% (n=49), mixed failure of root and substrate in 25.3% (n=24), failure of the roots in 21.1% (n=20), and failure of the shoot in 2.1% (n=2) (Fig. 3C). A comparison of the failure modes within the seven different tested tree barks (Fig. 4) showed predominantly failure of the substrate and of the substrate in combination with failure of the root in the five species P. engelmannii, P. serrulata, A. hippocastanum, C. avellana, and G. biloba. In A. lamarckii and A. rubrum, failure of the bark substrate was less pronounced than failure of the root. Failure of the shoot was recorded once in the latter two species. There was no correlation between the observed modes of failure and the native co-distribution of the trees with English ivy.

Fig. 4.

Comparison of failure modes of attachment between the seven tested tree barks. The different species of tested tree barks are shown on the x-axis and the relative frequency of the failure modes is shown on the y-axis. Descriptions above the columns indicate the level of bark structuring.

Fig. 5.

Schematic drawings of force application in various testing set-ups and typical force–displacement curves of mechanical tests in which root failure occurred. (A) Test of root clusters. The load is applied to a shoot segment of English ivy, normal to the surface. (A.1) Typical force–displacement curve with a slightly convex force increase followed by a sudden and complete failure of the sample, (A.2) Typical force–displacement curve for samples that showed preliminary failure events in the decreasing part of the curve before complete failure. (A.3) Typical force–displacement curve with a stepwise force decline showing several preliminary failure events after root strength is reached. (B) Tensile test of complete individual attached roots; the load is applied normal to the substrate. (C) Test of complete individual attached roots; the load is applied parallel to the substrate. (D) Test of isolated intact roots with cortex; the load is applied parallel to the longitudinal axis of the root. (E) Test of isolated central cylinders; the load is applied parallel to the longitudinal axis of the root. ar, attachment root; cc, central cylinder; cs, substrate; is, ivy shoot segment; rc, root cortex. (This figure is available in colour at JXB online.)

Tree barks were categorized as: A. hippocastanum, A. lamarckii, C. avellana, and P. serrulata, smooth; A. rubrum and G. biloba, medium; and P. engelmannii, rough.

Mechanical testing:

The maximum attachment force per root in tests of root clusters with root failure only does not differ between shoot segments pulled off tree barks and pulled off cork (U-test W=248, P=0.4926). Typical force–displacement curves for the three defined categories of the displacement tests of complete root clusters with root failure only are shown in Fig. 5A. The number of steps in the force–displacement curves of categories A.2 and A.3 do not match the number of roots per root cluster. They seem to represent failure of single roots as well as failure events of root subclusters. The maximum attachment force for these root clusters does not vary according to the category of failure (Kruskal–Wallis rank sum test χ²=0.2233, df=2, P-value=0.8944). Therefore, the data of the three sets of tests with root failure only were pooled. For the force–displacement curves in all categories, a root strength of 0.33 N (IQR 0.27, n=42) per single root was calculated.

Individual roots pulled normal to the bark surface (Fig. 5B) typically show a force–displacement curve with a steep increase of force after initial slippage until a peak force was reached. This peak corresponded to the initial failure of the root cortex. Afterwards, until complete failure, a stepwise decline occurs with clearly marked plateaus of force indicating preliminary failure events (Fig. 5B). During this phase the central cylinder was peeled out of the surrounding cortex tissue. Roots pulled normal to the surface have a median of the root strength of 0.23 N (IQR 0.37, n=13, Fig. 6B).

Fig. 6.

The root strength per root and isolated central cylinders experimentally measured or calculated, respectively (in the case of root clusters), for the different test set-ups. (A.1) Root cluster tests with root failure mode only (category A.1), calculation of root strength per single root. (B) Load applied normal to the substrate. (C) Load applied parallel to the substrate. (D) Tensile test on an intact isolated attachment root with cortex. (E) Tensile test on an isolated central cylinder. Lower case letters on top of A.1–D indicate statistical differences (pair-wise U-tests with Šidák correction). (This figure is available in colour at JXB online.)

Individual roots pulled parallel to the bark surface (Fig. 5C) show a slightly convex force–displacement curve that increases continuously until root strength is reached. Then typically a preliminary failure occurs before the force rapidly decreases until sudden failure of the sample (Fig. 5C). The roots pulled parallel to the surface show a median of the root strength of 0.32 N (IQR 0.27, n=11, Fig. 6C).

In the tension tests with intact isolated roots, the root cortex (Fig. 5D) typically failed after small displacements, and the central cylinder was pulled out of the cortex on one side. On the other side, the central cylinder remained firmly anchored in the cortex. The free end of the central cylinder after the test was usually longer than the piece of cortex it was drawn out of. A clear peak in the typical force–displacement curve marks the rupture of the cortex. Afterwards the curves show a relatively continuous region with a slope close to zero at around half the root strength until complete and sudden failure (Fig. 5D). The median root strength of intact attachment roots with cortex is 0.40 N (IQR 0.20, n=19, Fig. 6D). The Young’s Modulus for the intact isolated roots calculates to 109.2 MPa (IQR 141.9, n=16) and the breaking stress to 3.4 MPa (IQR 1.7, n=16). No correlation between the initial sample length of the tested isolated roots and the calculated Young’s Modulus is found (R² >0.005, see Supplementary Fig. S1 available at JXB online). The tested mortar showed a tensile strength of 0.13±0.1 MPa (median ±IQR, n=12).

After an initial slippage, a typical force–displacement curve of an isolated central cylinder of an attachment root (Fig. 5E) shows a steep (nearly) linear, elastic part and, after a short saddle region with a slope close to zero, an extended region of plastic deformation with a shallow increasing slope until failure. Isolated central cylinders show median strength of 0.61 N (IQR 0.23, n=11, Fig. 6E). The strain calculates to a median at 34% (IQR 7.1, n=10), the breaking stress to a median of 38 MPa (IQR 29.8, n=10), and the Young’s Modulus of the initial linear elastic region to a median of 220 MPa (IQR 83.5, n=10). The average median area fraction of cortex to central cylinder was 0.75:0.25 (IQR 0.17, n=12). Based on this area ratio, the Young’s Modulus of the attachment root cortex was approximated to be 74 MPa.

Figure 6 shows the root strength per root measured with the different testing set-ups. The data of root cluster tests, measurements on individual attachment roots, and tests on intact isolated attachment roots show significant differences only between intact roots pulled normal to the surface and intact isolated attachment roots (pairwise U-test with Šidák correction, W=55, P-value=0.0077).

Discussion

Modes of failure with regard to the substrate

Attachment strength depends on the interaction between the attachment system and the substrate. This is reflected by the different frequencies of failure modes obtained from the detachment tests from artificial, semi-artificial, and natural substrates. The typical mortar used as an artificial substrate showed almost exclusively failure of the substrate in the tensile tests (Fig. 3A). The attachment system (Melzer et al., 2010) provides the anchorage needed, but the cohesion of the substrate itself represents the ‘weak spot’—around 27times weaker than the attachment roots—if subjected to tensile loading.

The semi-artificial substrate (cork) showed the most shoot failures in tensile tests, implying that the attachment system and all its parts are stronger than the structural integrity of the shoot. Since the root strength per root in tests with root failure only do not differ from those of plants attached to other substrates, the reason for this very strong attachment can most probably be attributed to the distribution of attachment roots. On cork the same number of attachment roots as on other substrates is developed by a shorter section of ivy shoot. This condensed attachment results in the shoot being the weakest part in displacement tests (Fig. 3B).

The natural substrates (tree barks) showed failure of substrate, failure of root, and combination of substrate and root failure, whereas failure of the shoot was only observed in two cases (Fig. 3C). The bark of the various tree species tested showed differing proportions of failure modes reflecting the quality of attachment of ivy onto the respective tree barks. In two cases—A. lamarckii and A. rubrum—failure of the root was predominant and accompanied by sporadic shoot failures. This suggests a strong attachment of ivy onto these tree barks, and a ‘sufficient’ mechanical stability of the bark. In C. avellana and A. hippocastaneum, mixed failures of substrate and root prevail, indicating from the mechanical point of view a ‘balanced’ attachment–surface system. Contrasting results are found for the tests on P. serrulata, G. biloba, and P. engelmannii where predominantly substrate failure is found, indicating a strong attachment of ivy to the bark but an ‘insufficient’ mechanical stability of the bark.

No evident differences in terms of getting overgrown could be found between the two tested species that are naturally co-distributed with English ivy (C. avellana and A. hippocastanum) and the five species that are not. However, it is striking that in C. avellana and A. hippocastanum that are naturally co-distributed with English ivy a mechanically ‘balanced’ attachment–surface system exists, in which the attachment organs and the bark mostly fail simultaneously. This finding may be interpreted as a result of co-evolution. When comparing the failure modes with the macroscopic structuring of the tree barks no correlation was found.

The present study shows that English ivy is able to grow on a wide variety of bark structures, semi-artificial substrates such as cork, and artificial ones such as mortar. Failure of the attachment interface did not occur in the tests. Thus the attachment system of English ivy can be interpreted as successfully adapted for functioning on most vertical substrates the plant is likely to encounter. This non-specialization regarding the substrates may also explain partly why H. helix became an invasive species when introduced to non-native territories (Larocque, 1999). One way to reduce the risk of getting overgrown by H. helix is an easy-to-shed outer layer, which peels off together with the attached ivy, for example by producing sacrificial ever-renewing barks.

Mechanical testing of isolated roots and central cylinders

The attachment roots of English ivy show the common basic anatomical structure of roots: a stiff central cylinder embedded in a relatively soft cortex. This cable-like structure is—from an evolutionary point of view—inherited from the subterranean tension-loaded water-uptaking roots, which also show a stripping of the central cylinder when under tension (Ennos, 1991). Regarding the mechanical analysis, the method used, namely gluing isolated roots onto the aluminium platelets, in combination with relatively short sample lengths, could lead to an uneven strain field over the diameter of the tested root sample and may cause an overestimation of the Young’s Modulus. However, tests with root samples of different lengths showed no correlation between Young’s modulus and root sample length. This indicates that there exists no or only a negligible influence of the ‘end effect’ at the fixation points on the holders on the measured mechanical properties (Supplementary Fig. S1 at JXB online).

Comparing the results of tensile tests on isolated complete roots composed of a central cylinder surrounded by a cortex layer (Fig. 5D) with those of isolated central cylinders (Fig. 5E), some differences were found. Force–displacement curves for isolated central cylinders typically show a region of plastic deformation with a slow but steady increase of force after the steep force increase in the initial elastic region of the curve. In the case of the isolated complete roots, the force–displacement curve starts with an initial linear increase in the elastic region followed by a sudden drop of force when the cortex ruptures. Thereafter the curve shows a long shoulder region with a slope close to zero and not an increase as could be expected due to the continuous tensile loading of the central cylinder. A possible explanation for the constant force in the long shoulder region might be the occurrence of additional damage of the interface between the cortex and central cylinder. Further load acts mainly on the central cylinder and causes it to be strained and simultaneously sheared out of the cortex until complete failure of the system. This could explain the differences in the shape of curves as, in isolated central cylinders, after the initial elastic region, only plastic straining will occur.

Under natural conditions this mechanical behaviour can be interpreted as an effective fail-safe mechanism. As long as complete failure of the system has not developed, the attachment system remains functionally intact, anchoring the plant to its substrate.

The data for root strength per root show very little differences between the various testing methods used. The results calculated for single roots of the tested root clusters (Fig. 6A) and the data for single roots with a load applied parallel to the substrate (Fig. 6C) differ significantly neither from each other nor from the results obtained with the two other testing methods. The only significant difference is found between tests of single roots with a load applied normal to the substrate (Fig. 6B) and measurements of intact isolated roots tested under tension (Fig. 6D). The slight (often not significant) differences in the medians of root strength may be explained by differences in load application. Whereas in tension tests of intact isolated roots (Fig. 6D)—rendering the highest median value for the root strength—only the root material was tested, in all other set-ups a combination of root material (under tension and/or bending) and attachment strength (under shearing and/or peeling) was tested. In the case of single roots with a load applied parallel to the substrate (Fig. 6C) due to the testing geometry, the force application mainly causes tension loads in the root and shearing in the attachment zone, resulting in the second highest median value for root strength. In the cases of the tested root clusters (Fig. 6A) and of single roots with a load applied normal to the substrate (Fig. 6D), due to the testing geometry the force application causes tension and bending loads in the root and peeling and shearing in the attachment zone, resulting in the lowest median value for root strength. Due to the spatial arrangement of the single roots in the tested root clusters (Fig. 6A), the effect of bending and peeling is in this case less detrimental than in the case of single roots with a load applied normal to the substrate (Fig. 6B), resulting for single roots of the root clusters in a median value for the root strength very similar to the value found for tests of single roots with a load applied parallel to the substrate. Even if most of the differences found are not significant, it may be hypothesized that the spatial arrangement of single roots in the root clusters is mechanically beneficial. Due to these considerations, the only significant difference found—between single roots with a load applied normal to the substrate (Fig. 6B) and intact isolated roots tested under tension (Fig. 6D)—may indicate that the attachment interface is weaker under peeling and shearing than the root tissues under tension and bending. This unfavourable load type is avoided by the usually pair-wise arrangement of the roots. Nevertheless, if shearing and peeling still occur, overall failure is delayed by a stepwise failure of the cortex, whereby more energy can be dissipated (Fig. 5B). The overall high similarity of root strength between the different tests of single roots and of the tests with root clusters indicates that the attachment system of English ivy as a whole is functioning at a level very close to the maximum loads the root tissues are able to bear, which represents the highest possible fail-safe for the system.

Mechanical properties of attachment roots in comparison with subterranean roots

When comparing the mechanical properties of the attachment roots of English ivy with those of the roots of woody plants (Coutts, 1983; Commandeur and Pyles, 1991; Mickovski and Ennos, 2003; Table 2), it can be seen that the stiffness and the breaking stress of the roots of woody plants are clearly higher and their maximum strain lower than the values found for attachment roots of English ivy. The structural reason for these differences may be the development of highly lignified secondary xylem (wood) that significantly contributes to the stiffness and strength but will reduce maximum strain.

Table 2.

Biomechanical properties of roots and root tissues of different species, with row 1 showing the results of the present study

Species, details (reference)	Young’s Modulus	Tensile strength	Maximum strain
Hedera helix, intact attachment root (present study)	109 MPa	3.4 MPa	34 %
Hedera helix, central cylinder (present study)	220 MPa	38 MPa
Hedera helix, cortex (present study)	∼74 MPa
Woody plants/roots (Coutts, 1983; Commandeur and Pyles, 1991; Mickovski and Ennos, 2003)	204–800 MPa	13.5–63 MPa	11–20%
Allium porrum/radicles (Ennos, 1990)	16 MPa	1.6 MPa	10%
Arundo donax, rhizome longitudinal (Speck and Spatz 2003)	112 MPa	4.3 MPa	11.4%
Arundo donax, adventitious roots (Speck et al., 1998)	∼700 MPa	∼25 MPa (20–30 MPa)	8–9%
Galium aparine, first-order lateral roots (Goodman, 2005)	235 MPa	16.5 MPa	28%
Helianthus annuus, roots (Ennos et al., 1993a)	68 MPa
Helianthus annuus, radicle (Ennos, 1989)	10.8 MPa	0.7 MPa	7%
Impatiens glandulifera adventitious roots (Ennos et al., 1993a)	18.6 MPa	4.24 MPa
Solanum lycopersicum, upper lateral root (Gartner, 1994)	49.6 MPa	7.6 MPa	10.4%
Zea mays, adventitous roots (Ennos et al., 1993b)	72 MPa

The comparison with data from subterranean anchoring systems of non-woody plants (Table 2) is more complex. The rhizome of giant reed (Arundo donax) which also develops no secondary xylem (Speck and Spatz, 2003; Table 2) shows a similar stiffness, a slightly lower breaking stress, and three times the maximum strain for the ivy attachment roots. Compared with preliminary data of the adventitious roots of the giant reed which originate from the rhizome (Speck et al., 1998), the attachment roots of English ivy show only a seventh of the stiffness and of the breaking stress but possess an ∼4 times higher maximum strain. The combination of the less stiff rhizome and the stiff adventitious roots contributes to the effective dissipation of energy in subterranean anchoring structures of A. donax when its culms are swaying in the wind (Speck and Spatz, 2004). This may explain (together with the missing secondary xylem) the similarities in mechanical properties of the rhizome of A. donax to those of the attachment roots of English ivy. In both plants loads caused by wind are transferred from one stiff structure (the stem) to another. In the case of giant reed the loads of the swaying stem are transferred to the stiff roots and soil by the rhizome, and in the case of English ivy the loads of the stem deflected by wind acting on the leaves are transferred to the stiff substrate by the attachment roots.

Compared with the radicles of Helianthus annuus and Allium porrum (Ennos, 1989, 1990; Table 2), the attachment roots of H. helix are stiffer, and have a higher breaking stress and maximum strain. Since radicles have to stabilize only small seedlings and are not lignified, this is to be expected. A comparison with the roots of sunflower (H. annuus; Ennos et al., 1993a) and tomato (Solanum lycopersicum; Gartner, 1994), and with the adventitious roots of balsam (Impatiens gladulifera; Ennos et al., 1993a) and corn (Zea mays; Ennos et al., 1993b) proves that the attachment roots of English ivy are stiffer by a factor of 1.5–5 (Table 2), have a lower breaking stress, and a 3–4 times higher maximum strain. The lower breaking stress is probably compensated by the higher maximum strain of the central cylinder, which most probably plays an important role in the fail-safe mechanism of the attachment system of English ivy.

The combination of a high stiffness and a high maximum strain can also be found in the roots of semi-self-supporting cleavers (Galium aparine; Goodman, 2005). It seems to fit well with the requirements for climbing or leaning plants, which are regularly exposed to changing loads from wind and relative motion of the host plants. In addition to a two times higher stiffness and a similar maximum strain, the roots of G. aparine have a breaking stress almost five times higher than that of the attachment roots of English ivy. This may be due to differences in the attachment mode of cleavers and English ivy, which lead to higher strains and loads acting on the subterranean roots of G. aparine than on the attachment roots of H. helix when their host plants begin to sway.

The attachment roots of English ivy differ considerably in their morphology and anatomy from its nourishing subterranean roots (Bruhn, 1909). These differences can most probably be attributed to the differing functions of these two root types, with the attachment roots being highly specialized in anchoring. A closer look at the comparative morphology, anatomy, ultrastructure, and biochemistry of attachment roots, roots hairs, and the attachment interfaces will help in a better understanding of nature’s ‘evolutionary strategies’ for developing highly efficient attachment systems (cf. Melzer et al., 2010).

Conclusions

For the permanent attachment system of English ivy it could be shown that attachment performance depends on the interaction between attachment system and substrate. This holds true for all tested artificial, semi-artificial, and natural substrates, as well as for subcategories of the latter. The attachment system is functioning on a wide variability of substrates and the best way to avoid a permanent attachment of ivy is to develop an easy to shed outer layer which can be peeled of together with the climber.

Tensile tests showed that the overall performance of the attachment system of English ivy for a given substrate is close to the structural integrity of the involved plant tissues. The present studies prove that the attachment system of English ivy is very effective especially regarding its intrinsic fail-safe mechanism. The attachment roots have evolved towards specialized anchorage organs with mechanical and structural properties showing distinct differences from those known from subterranean roots.

The authors would like to thank Martina Goldmann for her help with the outdoor tests. We are particularly grateful to Jürgen Schmitt and Dieter Schächtele from the technical workshop of the Institute of Biology II/II, University of Freiburg for technical support during the development of the tensile tester and the modification of the micro-tensile testing device. Financial support from the ‘Baden-Württemberg Stiftung’ (formerly ‘Landesstiftung Baden-Württemberg’) within the scope of the research programme ‘Neue Materialien aus der Bionik’ is gratefully acknowledged.