Increased ecological amplitude through heterosis following wide outcrossing in Banksia ilicifolia R.Br. (Proteaceae)



    1. School of Plant Biology, Faculty of Natural and Agricultural Sciences, The University of Western Australia, Crawley, WA, Australia
    Search for more papers by this author
  • S. L. KRAUSS,

    1. School of Plant Biology, Faculty of Natural and Agricultural Sciences, The University of Western Australia, Crawley, WA, Australia
    2. Botanic Gardens and Parks Authority, Kings Park and Botanic Garden, West Perth, WA, Australia
    Search for more papers by this author

    1. School of Plant Biology, Faculty of Natural and Agricultural Sciences, The University of Western Australia, Crawley, WA, Australia
    Search for more papers by this author

    1. School of Plant Biology, Faculty of Natural and Agricultural Sciences, The University of Western Australia, Crawley, WA, Australia
    Search for more papers by this author

    1. School of Plant Biology, Faculty of Natural and Agricultural Sciences, The University of Western Australia, Crawley, WA, Australia
    Search for more papers by this author

Erik Veneklaas, School of Plant Biology, Faculty of Natural and Agricultural Sciences, The University of Western Australia, Crawley, WA 6009, Australia.
Tel.: +61 86488 3584; fax: +61 86488 1108;


To assess whether wide outcrossing (over 30 km) in the naturally fragmented Banksia ilicifolia R.Br. increases the ecological amplitude of offspring, we performed a comparative greenhouse growth study involving seedlings of three hand-pollinated progeny classes (self, local outcross, wide outcross) and a range of substrates and stress conditions. Outcrossed seedlings outperformed selfed seedlings, with the magnitude of inbreeding depression as high as 62% for seed germination and 37% for leaf area. Wide outcrossed seedlings outperformed local outcrossed seedlings, especially in non-native soils, facilitated in part by an improved capacity to overcome soil constraints through greater root carboxylate exudation. Soil type significantly affected seedling growth, and waterlogging and water deficit decreased growth, production of cluster roots, root exudation and total plant P uptake. Our results suggest that the interaction of narrow ecological amplitude and the genetic consequences of small fragmented populations may in part explain the narrow range of local endemics, but that wide outcrossing may provide opportunities for increased genetic variation, increased ecological amplitude and range expansion.


The species-rich Southwest Australian Floristic Region (Hopper & Gioia, 2004), is a global biodiversity hotspot, the status given for regions which have many endemic species under threat (Myers et al., 2000). Characterized by a Mediterranean climate and an ancient, weathered and nutrient-deficient landscape (Hopper, 1979), this region is endowed with 7380 native vascular plant species/subspecies, of which 49% are endemic and 2500 are of conservation concern (Hopper & Gioia, 2004). Despite the global significance of this region, there is still only a poor understanding of the factors influencing high diversity and endemism, and especially the population genetic consequences of narrow endemism and naturally fragmented species distributions.

For example, narrow endemism due to narrow ecological amplitude can impose limitations on population size and fragmentation if the specific habitat is restricted and/or patchy, with consequences for population genetic variation and patterns of mating. Two major genetic consequences of small isolated populations are increased genetic drift and inbreeding (Ellstrand & Elam, 1993). Genetic drift and inbreeding can lead to a loss of genetic variation and fitness through inbreeding depression, or the loss of fitness with increasing homozygosity (Charlesworth & Charlesworth, 1987). This may serve to reinforce the narrow ecological amplitude of the species by reducing genetic variation. Thus, narrow ecological amplitude, small population size, fragmented populations, reduced genetic variation and increased inbreeding may interact to affect the rarity or endemism of species.

Conversely, in some circumstances a distribution of naturally small and fragmented populations can show a high inter-population gene flow (Young et al., 1996). Although the products of wide outcrossing from mating between individuals from different populations often display heterosis, or hybrid vigour, and improved fitness, they can also have negative effects through outbreeding depression (Keller & Waller, 2002; Hufford & Mazer, 2003). This range of outcomes is most evident in interspecific hybrids (Keller et al., 2000). There is also a growing realization of the importance of metapopulation dynamics for naturally fragmented species (Hanski & Gaggiotti, 2004) and the important role that gene flow into an inbred population can have for population fitness (Tallmon et al., 2004).

In this study, we assess whether seedlings generated by wide outcrossing demonstrate greater ecological amplitude than relatively more inbred seedlings across a range of local and nonlocal substrates and stress conditions in the widespread but naturally fragmented Banksia ilicifolia R.Br. (holly-leaved banksia). The performance of individual seedlings was assessed through plant vigour and root-exudation properties. Cluster (proteoid) roots of Proteaceae exude carboxylates, which enhance nutrient uptake, especially phosphorus (Lambers et al., 2003; Shane et al., 2005). In the extremely phosphorus-depleted soils of Western Australia, greater expression of this mechanism potentially leads to greater plant vigour across a wider range of soils. This study was conducted in the context of attempting to understand interactions between genetic variation and environmental specificity that may influence the restricted distribution of B. ilicifolia, and perhaps in general, rarity and/or habitat specialization.

Banksia ilicifolia is a widespread species in south-western Australia, found mainly on the sandy coastal plain from Mt Lesueur to Augusta, and east to Cordingup River between Albany and Bremer Bay (Taylor & Hopper, 1988). Although widespread, B. ilicifolia is locally restricted to swales and wetland fringes with depth to groundwater <10 m (Groom et al., 2000; Groom, 2004). As a consequence, populations are naturally fragmented and typically small (<100 plants), although some large stands do exist (Taylor & Hopper, 1988). B. ilicifolia is self-compatible, but preferentially outcrossing, with marked inbreeding depression at seed production, seed germination and seedling growth (Heliyanto et al., 2005). Flowering is from winter to spring, with extremely low fruit set (c. 0.04–1% of flowers setting fruit) following natural pollination (Whelan & Burbidge, 1980; Lamont & Collins, 1988; Heliyanto et al., 2005).

The key questions that we address for selfed and narrowly and widely outcrossed seedlings of B. ilicifolia are: (1) what is the magnitude of inbreeding depression? (2) does wide outcrossing lead to heterosis across a range of local and nonlocal substrates and stress conditions? (3) which growth parameters and root functional traits are associated with these trends?

Materials and methods

We investigated the growth responses of B. ilicifolia seedlings, originating from three pollen sources, to a range of challenging soil and environmental conditions. The three pollen sources were selfed (S), narrow outcrossed, or local (L) and wide outcrossed (over 30 km), or nonlocal (NL) (Heliyanto et al., 2005). These three classes represent increasing levels of outbreeding, and presumably, increasing average levels of individual heterozygosity.

Seeds of each group were generated by performing hand-pollination manipulations on plants within a natural population of B. ilicifolia in banksia woodland at the Harry Waring Marsupial Reserve, 32°10′S, 115°50′E, The University of Western Australia, from 26 September to 15 December 2002. Nonlocal pollen was sourced from five arbitrarily selected plants at an ecologically matched site approximately 30 km away at Kensington Bush Reserve (31°57′S, 115°49′E). Local pollen was sourced from 15 plants within the population that were at least at 100 m from the maternal plants. Selfed seeds were produced using self-pollen from the maternal plant. Hand-pollinated flowers were bagged with glacine bags to exclude pollinators and hence pollen contamination (Heliyanto et al., 2005). Mature fruits (follicles) were collected on 11 April 2003, and closed follicles were heat treated to extract seeds. Further details of hand pollinations are in Heliyanto et al. (2005). As B. ilicifolia is preferentially out-crossing (Heliyanto et al., 2005), the number of seeds produced for each group varied significantly, i.e. 71 for the nonlocal outcrossed, 65 for local outcrossed and 24 for the selfed, respectively, despite hand pollinating 2000 flowers with each pollen source. However, average seed mass (32.7 mg for selfed seeds, 38.5 mg for local outcrossed seeds, 39.0 mg for nonlocal outcrossed seeds) did not differ significantly among treatments (Kruskall–Wallis H4,162 = 8.04, P = 0.09).

This paper presents the results of three glasshouse experiments, all of which assessed growth, plant morphometric traits and rhizosphere chemistry.

  • 1The first experiment assessed the performance of seedlings from the three progeny classes (S, L and NL) in their native soil (sandy native; Table 1). Sample sizes at harvest were 3, 6 and 6 (for S, L and NL respectively).
  • 2In the second experiment, seedlings of two progeny classes (L and NL) were grown in the native soil as well as two non-native soils, sandy non-native and lateritic (Table 1). The sandy non-native soil also originated from the Harry Waring reserve, but c. 650 m from the study population and c. 5 m higher in the landscape, where B. ilicifolia does not occur (presumably because of inadequate access to groundwater). The lateritic soil originated from the habitat of another Banksia species [B. seminuda (A.S. George) Rye] that, like B. ilicifolia, is also associated with wetter parts of the landscape. In contrast to the sandy soils occupied by B. ilicifolia, B. seminuda occupies loamier, gravely lateritic soils on the Darling Scarp, approximately 250 km from the Harry Waring population. Treatment sample sizes were 12 for native soil (six for each progeny class), eight for sandy non-native (four for each progeny class) and nine for lateritic soil (four for L and five for NL).
  • 3The third experiment assessed the effect of stress conditions that B. ilicifolia is likely to be exposed to from time to time (water stress and waterlogging), on two progeny classes (L and NL) in their native soil. Sample sizes for each treatment were 12 (six for each progeny class).
Table 1.   Physical and chemical properties of different soils used in the experiments, their locations and codes.
Soil codeSandy nativeSandy non-nativeLateriticMethod
  1. Methods: 1, Rayment & Higginson (1992); 2, Walkley & Black (1934); 3, Allen & Jeffery (1990); 4, Colwell (1963); 5, Parfitt & Childs (1988).

LocationHarry Waring 1Harry Waring 2Darling Scarp 
Latitude, longitude32°10′S, 115°50′E32°21′S, 115°37′E48°05′S, 116°08′E 
LandscapeSandy damp swaleSandy dry slopeLateritic 
Gravel (%)0021.5
pH CaCl24.354.654.951
Organic C (%)1.451.264.692
Total phosphorus (mg kg−1)1933.51333
Bicarbonate-extractable phosphate (mg kg−1)256.54
Reactive Fe (mg kg−1)22529017055
Oxalate-extractable Al (mg kg−1)15024319895
Bulk density (g cm−3)

Germination, growth condition and experimental design

Germination occurred in a growth chamber. Seeds were germinated on plastic trays filled with sandy soil. To maintain moisture content, water was added every second day. The average temperature and light intensity (400–700 nm PAR) during the germination process were 15 °C and 14 μmol m−2 s−1, respectively. The progeny classes did not differ in their time to germination, but germination rates varied significantly. Outcrossed seeds had similar rates of germination (88% of seeds germinated) which was greater than that of selfed seeds (33%) (Heliyanto et al., 2005). Approximately 60 days after sowing, when all seedlings had produced opened cotyledons, they were transferred into pots of 15 × 15 × 15 cm (one per pot), each pot containing one of three different soils (Table 1) that had been passed through a 4-mm sieve. Pots were placed randomly on two neighbouring benches in a controlled greenhouse (20 °C/15 °C, day/night, light intensities 65% ambient). Pots were flushed with deionized water as required (daily in summer, less frequently in winter), and no additional nutrients were given.


For the first and second experiment, no further treatments beyond soil type were applied, and seedlings were grown on the assigned substrates until harvest. For the third experiment, treatments were imposed on 10-month-old seedlings. Stress treatments were established as follows:

  • (i) To achieve waterlogging, pots were placed inside bigger containers and filled, from the top, with de-oxygenated deionized water to a level approximately 20–30 mm above the soil. To replenish evaporative water loss, the lost amount was compensated by adding de-oxygenated deionized water. During waterlogging, redox potentials (Eh) at a depth of 60 mm below the surface of the soil were monitored regularly. Platinum electrodes (Patrick et al., 1996) were pushed into the sand to the desired depth, a few days before pots were waterlogged. Measurements were recorded by connecting the platinum electrode and a calomel reference to a millivolt meter, and recording the potential between the half cells, after calibration. Values were then converted to the hydrogen electrode standard by adding 245 mV to the reading (Patrick et al., 1996; Poot & Lambers, 2003). During the experiment, the redox potential values of the soils at 7-cm depth rapidly declined to values around 0 mV, 2 weeks after the onset of the waterlogging treatment, and gradually decreased to −50 mV towards harvesting time.
  • (ii) To establish water stress, replenishment of water to pots was reduced to 50% of the amount lost through evapotranspiration, on a daily basis, for each plant separately. Once the plants showed stress symptoms (wilting of the youngest developing leaf, at 5–7% of field capacity), pots were held at that soil moisture level.

Plants were harvested for growth analysis 6 weeks after the commencement of stress treatments.


At the time of harvest, shoots were cut at the soil surface. For each plant, five rhizosphere soil-bearing cluster roots per pot were taken for rhizosphere extracts (Veneklaas et al., 2003). Cluster roots were selected on the basis of their apparent stage of development, mature but not senescent (i.e. white to yellowish, densely branched and covered with root hairs; Shane et al., 2004). Briefly, cluster roots were transferred to a 30 mL plastic container, in which the rhizosphere soil was washed off the roots by gently shaking the container after adding a measured amount of 0.2 mm CaCl2 solution, i.e. 15–20 mL depending on cluster size. A subsample of the extract was filtered through 0.22 μm Pal Gelman Acrodisc® syringe filter (Pal, East Hills, NY, USA) into 1 mL HPLC vials, preserved by adding two drops of acid and then frozen. Morphometric traits observed on the plants were number of leaves, oven-dried (70 °C for 48 h) masses of roots, stems and leaves, as well as total root length and total projected leaf area (both measured using scanner and image analysis system: Win Rhizo V3.9, Regent Instrument, Quebec, Canada). For carboxylate analysis, a reversed-phase column liquid chromatography method was used for the separation and quantification of 10 low-molecular-mass organic acids (malic, malonic, lactic, acetic, maleic, citric, cis-aconitic, succinic, fumaric, and trans-aconitic) in plant root exudates (Cawthray, 2003).

Inbreeding depression

The magnitude of inbreeding depression (δ) for selfed vs. outcrossed and local outcrossed vs. wide outcrossed seedlings were calculated as:


where ωS and ωO are the growth parameters of relatively inbred and outbred offspring, respectively (Johnston, 1992). Inbreeding depression following selfing was assessed by comparing growth parameters of selfs to the average value across both classes of outcrossed progeny. Significantly positive values (assessed by t-test) of δ indicate inbreeding depression.

Soil analysis

General soil analyses (with two replicates per soil sample) were carried out by CSBP FutureFarm analytical laboratories (Bibra Lake, Australia) after air drying and sieving at 2 mm. The analytical methods used are listed in Table 1.

Leaf phosphorus content

The whole shoot of each replicate was ground in a stainless steel ball mill for P analyses. A subsample of 0.1 mg was digested in a nitric acid/perchloric acid mixture. The P content of the digest was then determined using the molybdo-vanado-phosphoric acid method (Kitson & Mellon, 1944).

Relative water content (%)

Relative water content (RWC) of discs from fully expanded leaves was measured, in six replicates, by determining field fresh weight, saturated fresh weight after overnight saturation between moist paper at room temperature, and dry weight after oven drying for 48 h at 70 °C. RWC is defined as the ratio between water content at the sampling and water content after saturation.

Statistical analysis

Statistical analysis was performed using Statistica 6 (Statsoft Inc., Tulsa, OK, USA). Parametric anova was performed after confirming homogeneity of variance using Cochran's C-test. Effect of pollen source alone (first experiment) and pollen source, substrate and their interaction (second experiment) were evaluated using one-way and two-way unbalanced anova, respectively, followed by a posthoc unequal N HSD Test. For this purpose, a type III Sum of Squares was employed (Zar, 1999; Quinn & Keough, 2002). To compare the performance of genotype (progeny group) under different environmental conditions (waterlogging, water-stressed and control) a balanced two-way anova was applied, followed by a post hoc Duncan Multiple Range Test.


Growth and rhizosphere carboxylates of seedlings across substrates

Source of pollen (self, local cross, nonlocal cross) significantly affected the performance of seedlings grown in their native soil (Fig. 1; Table 2). Nonlocal outcrossed seedlings outperformed local outcross seedlings, and selfed seedlings, for total root length and shoot dry mass (Fig. 1). The magnitude of inbreeding depression, however, varied across the parameters measured, and was significantly >0 for total projected leaf area (δ = 0.37) and shoot dry mass (δ = 0.34), whereas all others traits showed positive values that were not significantly different from 0 (Table 3).

Figure 1.

 Number of leaves per plant (a), projected leaf area per plant (b), total root length per plant (c), shoot dry mass per plant (d), and root dry mass per plant (e) of three groups of progenies grown on native soils. S, selfed seedlings; L, local outcrossed seedlings; NL, non-local outcrossed seedlings. Error bars represent standard errors. Treatment notations were given based on homogeneous groups following unequal N HSD at P = 0.05.

Table 2. anova for growth parameters of three different Banksia ilicifolia progeny classes (self, local outcross, wide outcross) on native soil.
Source of variationd.f.SSMSFP
Number of leaves
 Source of pollen24.502.250.180.83
Projected leaf area
 Source of pollen288612433067.0<0.01
Total root length
 Source of pollen264876324384.80<0.05
Shoot dry mass
 Source of pollen235.0917.544.87<0.05
Root dry mass
 Source of pollen23.611.803.700.06
Table 3.   Estimates of inbreeding depression, and statistical significance, in Banksia ilicifolia.
ParameterInbreeding depression (δ)
Self vs. outcrossedLocal vs. non-local
  1. NA, t-test not applicable.

Seed germination (%)0.62 (NA)0.09 (NA)
Number of leaves0.06 (t2 = 2.18, P = 0.16)0.02 (t5 = 1.00,P = 0.36)
Projected leaf area0.37 (t2 = 42.33,P < 0.001)0.01 (t5 = 1.90,P = 0.11)
Total root length0.16 (t2 = 3.2,P = 0.09)0.03 (t5 = 1.27,P = 0.26)
Root dry mass0.31 (t2 = 3.14, P = 0.09)0.21 (t5 = 1.57,P = 0.18)
Shoot dry mass0.34 (t2 = 29.40, P < 0.001)0.06 (t5 = 1.91,P = 0.11)

The relative performance of nonlocal outcrossed seedlings on non-native sandy soils was not significantly different from that of local outcrossed seedlings (Fig. 2; Table 4). Although the performance of nonlocal outcrossed seeds on lateritic soils was similar to the performance of these seedlings on sandy soils, shoot and root dry weight for local outcross seedlings on lateritic soils was about half that of all other seedlings (Fig. 2). There was no significant interaction between crosses and substrates (Table 4).

Figure 2.

 Shoot dry mass per plant (a), and root dry mass per plant (b) of two groups of progenies across the three soils. L, local outcrossed seedlings; NL, non-local outcrossed seedlings. Error bars represent standard errors. Treatment notations were given based on homogeneous groups following unequal N HSD at P = 0.05.

Table 4. anova for growth parameters of two different Banksia ilicifolia progeny classes (local outcross, wide outcross) across three substrates.
Source of variationd.f.SSMSFP
Shoot dry mass
 Substrates (S)275.4937.7410.64<0.001
 Source of pollen (G)184.6784.6723.86<0.001
 S × G210.585.301.490.25
Root dry mass
 Substrates (S)29.924.969.91<0.001
 Source of pollen (G)15.585.5811.15<0.001
 S × G20.

Roots of all seedlings grown in native and non-native soils exuded a range of carboxylates, with aconitate, citrate, isocitrate and malate being the major components, depending on substrate and genetic background (Fig. 3). The concentration of some carboxylates (e.g. fumarate) was negligible, and not different across treatments. The amount of carboxylates in the rhizosphere of nonlocal outcross seedlings on lateritic soils, mainly citrate and isocitrate, was approximately 2.5 times greater than that of local outcrossed seedlings on laterite and approximately 5 times greater than all seedlings on sandy soils (Fig. 3).

Figure 3.

 Concentrations of citrate, iso-citrate, malate and aconitate (extracted using a 0.2 mm CaCl2 solution) in the rhizosphere of selfed, local outcrossed and nonlocal outcrossed Banksia ilicifolia progenies, grown in three different substrates. Error bars represent standard errors. S, selfed seedlings, L, local outcrossed seedlings, NL, nonlocal outcrossed seedlings.

Growth and rhizosphere carboxylates of seedlings under stress conditions


Waterlogging significantly decreased the soil redox potential (data not shown). As a result, the growth of individual progeny was decreased (Fig. 4a–c; Table 5). There was no significant difference in terms of soil redox potential readings between the two pollen source classes (data not shown). All waterlogged plants showed leaf senescence, starting from the oldest and gradually moving upwards. The two progeny classes did not differ significantly in the percentage of leaves retained (as percentage of control) (76% for nonlocal outcrossed seedlings and 69% for local outcrossed seedlings). Similarly, the two progeny classes showed comparable performance for all other growth parameters. The average projected leaf area, shoot dry mass and root dry mass (as percentage of control) were 77%, 86% and 69% for local outcrossed seedlings as against 80%, 88% and 79% for nonlocal outcrossed seedlings. There was no interaction between source of pollen and stress treatment (Table 5). Waterlogged cluster roots appeared to be decaying with a strong sulphurous odour and a greyish appearance. Carboxylate analyses of the rhizosphere soil found a considerable amount of acetate and lactate, which presumably did not result from root exudation, but from anoxic decomposition of organic matter (data not presented).

Figure 4.

 Projected leaf area per plant (a), shoot dry mass per plant (gram per plant) (b), and root dry mass per plant (c) of two different progenies following a 6-week waterlogging period. L, local outcrossed seedlings, NL, non-local outcrossed seedlings, S, stress, C, control. Shaded bars represent control treatments. Error bars are standard errors. Treatment notations were given based on homogeneous groups following Duncan Multiple Range Tests at P = 0.05.

Table 5. anova for growth parameters of two different Banksia ilicifolia progeny classes (local outcross, wide outcross) under waterlogged condition.
Source of variationd.f.SSMSFP
Projected leaf area
 Waterlogged (WL)1 775957759512.15<0.01
 Source of pollen (P)1  1708 17080.270.61
 WL × P1   499  4990.080.78
 Total23 20573   
Shoot dry mass
 Waterlogged (WL)112.3312.335.50<0.05
 Source of pollen (P)1 0.10
 WL × P1 0.05
 Error2044.87 2.24  
Root dry mass
 Waterlogged (WL)1 3.66 3.666.13<0.05
 Source of pollen (P)1 0.87 0.871.610.21
 WL × P10.0020.0020.0040.94
 Error2010.70 0.54  

Water stress

Water stress decreased projected leaf area, shoot mass and root mass compared with the control (Fig. 5; Table 6). In general, seedlings responded to water deficit by slower leaf growth (Fig. 5). Comparing pollen sources, nonlocal outcrossed seedlings did not perform differently to local outcrossed seedlings. The average projected leaf area, shoot dry mass and root dry mass (as percentage of control) for local outcrossed progeny and nonlocal outcrossed progeny were 68%, 75% and 73% and 73%, 78% and 80%, respectively. There was no interaction between source of pollen and stress condition.

Figure 5.

 Average projected leaf area per plant (a), shoot dry mass per plant (b), and root dry mass per plant (c) of two different progenies following a 6-week water stress period. L, local outcrossed seedlings, NL, nonlocal outcrossed seedlings, S, stress, C, control. Shaded bars represent control treatments. Error bars are standard errors. Treatment notations were given based on homogeneous groups following Duncan Multiple Range Tests at P = 0.05.

Table 6. anova for growth parameters of two different Banksia ilicifolia progeny classes (local outcross, wide outcross) under water stress condition.
Source of variationd.f.SSMSFP
Projected leaf area
 Water stress (WS)118083518083536.51<0.001
 Source of pollen (P)1  7977  79771.610.21
 WS × P1  4943  49430.990.33
 Error20 99074  4954  
Shoot dry mass
 Water stress (WS)1 43.50 43.5014.96<0.001
 Source of pollen (P)1 0.41 0.410.130.71
 WS × P1 0.30 0.300.110.75
 Error20 58.20 2.91  
Root dry mass
 Water stress (WS)1 3.03 3.035.66<0.05
 Source of pollen (P)1 0.92 0.921.700.25
 WS × P1 0.002 0.0020.0040.94
 Error20 10.70 0.54  
 Total23 14.65   

During the period of stress, seedlings maintained a high RWC in their leaves (88% as compared with 90% in control). There was no significant difference in RWC between the two pollen source classes (F1,20 = 0.54, P =0.47).

As found for waterlogged seedlings, the numbers of fresh active cluster roots were much less for stressed seedlings than for controls (Table 7). Roots of water-stressed seedlings exuded small amounts of carboxylates [data not presented].

Table 7.   Cluster-root formation in Banksia ilicifolia as dependent on treatment.
Treatments (soil, water status, genotype)Presence/absence of fresh (whitish) cluster roots Classification*
  1. *Classification is based on the presence or absence of carboxylates following collection and analysis of root exudates collected from rhizosphere soil.

Sandy native, field capacity, localPresent, abundantActive
Sandy native, field capacity, nonlocalPresent, abundantActive
Sandy non-native, field capacity, localPresent, abundantActive
Sandy non-native, field capacity, nonlocalPresent, abundantActive
Lateritic, field capacity, localPresent, abundant, smallVery active
Lateritic, field capacity, nonlocalPresent, abundant, smallVery active
Sandy native, waterlogged, localAbsentNonactive
Sandy native, waterlogged, nonlocalAbsentNonactive
Sandy native, water stress, localPresent, limitedActive
Sandy native, water stress, nonlocalPresent, limitedActive
Sandy native, field capacity, selfedPresent, abundantActive

Leaf P concentration under different soil conditions

Leaf P concentration varied among soil and stress conditions. Plants grown in the two sandy soils had higher leaf P concentration than those grown in lateritic soil or in stress conditions (Fig. 6). There were no big differences amongst the three pollen source classes, either in favourable or in nonfavourable conditions.

Figure 6.

 Leaf P concentration of three Banksia ilicifolia progenies across three substrates and stress conditions. Error bars represent standard errors. S, selfed progeny, L, local outcrossed progeny, NL, nonlocal outcrossed progeny.


Inbreeding depression

We have shown previously that B. ilicifolia demonstrates preferential outcrossing and strong early acting inbreeding depression, where outcrossed seedlings are more vigorous than selfed seedlings on native soils (Heliyanto et al., 2005). In the present study, the intensity of inbreeding depression was found to be as high as 62% for seed germination and 37% for projected leaf area. These measures underestimate the true strength of inbreeding depression in B. ilicifolia. For example, seed set following self-pollination by hand is <30% that following cross-pollination by hand, and self-seedling survival following fungal attack was less than half that of outcrossed seeds (Heliyanto et al., 2005). Ultimately, survival to 16 weeks of age for selfed progeny under glasshouse conditions was approximately 7% of that of outcrossed progeny, equating to an inbreeding depression of 93% (Heliyanto et al., 2005). Inbreeding depression is almost certainly more severe under natural field conditions than in the glasshouse conditions of the current experiment (Dudash, 1990). Therefore, few, if any, selfed offspring are expected to survive to reproductive age.

Consequently, elevated inbreeding may severely impact on the short-term viability, and long-term evolutionary potential, of small fragmented populations of B. ilicifolia. Although realized outcrossing rates are not yet known for B. ilicifolia, other species in the genus are typically completely, or highly, outcrossed (Goldingay & Carthew, 1998). In the closely related species B. cuneata, outcrossing rates varied between 0.65 and 0.95, with increased inbreeding detected in small, highly disturbed populations (Coates & Sokolowski, 1992). As a result, high outcrossing rates might be expected in B. ilicifolia populations, but small, and particularly recently fragmented, populations may show higher selfing rates and increased inbreeding (Sampson et al., 1989; Coates & Sokolowski, 1992; England et al., 2001).

We also demonstrate that inbreeding effects extend beyond the comparison of self vs. outcross progeny. Although not significantly different across all individual measures, we found a consistent tendency of superior performance of seedlings produced by wide outcrossing compared with seedlings produced by outcrossing between plants within the study population. Wide outcrossing can result in heterosis and/or outbreeding depression, and these effects can vary from F1 to F2 generations, and beyond (Waser, 1993; Hufford & Mazer, 2003). Our study was restricted to the F1 generation, and heterotic effects observed could be maintained, or hybrid breakdown may occur, in F2 or even F3 generations (Fenster & Galloway, 2000; Keller et al., 2000). Heterosis and outbreeding depression are obviously relative terms, influenced by the extent of inbreeding depression within populations (Fenster & Galloway, 2000). The relative performance of narrow outcross seedlings to wide outcross seedlings suggests that the study population of B. ilicifolia is inbred, with heterosis a consequence of wide outcrossing in F1 seedlings. Outbreeding depression has not been detected, suggesting that an optimal outcrossing distance (Price & Waser, 1979; Waddington, 1983; Waser, 1993), at least for the B. ilicifolia study population, lies at or beyond the genetic distance to the pollen source population 30 km away.

Increased ecological amplitude following wide-outcrossing

In the current study, we have extended the assessment of inbreeding depression on a common or local substrate to also assess the relative performance of the products of wide outcrossing to narrow outcrossing on nonlocal substrates and stress conditions. Here, we demonstrate that the outcrossed products of mating between individuals within a typically small and relatively isolated population of B. ilicifolia show a significant decline in vigour when grown on non-native lateritic soils, with biomass approximately half that of seedlings grown on their native sandy soils. In contrast, the outcrossed products of mating between individuals from populations 30 km apart showed no decline in biomass when grown on non-native lateritic soils compared with their performance on native sandy soils. This is despite the observation that the growth of all seedlings in laterite, irrespective of pollen source, was inhibited when younger, possibly due to reduced root growth due to high soil bulk density. By harvest though (when seedlings were 46 weeks old), wide-outcrossed seedlings had recovered, possibly due to increased rhizosphere carboxylate exudation, whereas narrow outcrossed seedlings had not.

These results suggest that even complete outcrossing between individuals within a typical population of B. ilicifolia, reinforces narrow ecological amplitude, whereas heterosis, following wide outcrossing between individuals from genetically differentiated populations, can increase ecological amplitude. Our results suggest that, in the absence of wide outcrossing, there is an interaction between narrow ecological amplitude, small relatively isolated populations and mating that acts to prevent the wider distribution of this species, and may contribute to local endemism more generally. Although wide outcrossing has been shown to play an important part in the ‘genetic rescue’ of locally inbred populations (Richards, 2000; Ingvarsson, 2001; Tallmon et al., 2004), we suggest that wide outcrossing, facilitated by inter-population dispersal, may provide a natural mechanism for increasing ecological amplitude, and extending the distribution beyond current environmental constraints, of species. An ability to increase the range or ecological amplitude of a species will depend on a multitude of factors, including the frequency of wide outcrossing, the strength of heterosis in the products of wide outcrossing and a balance between population genetic differentiation and inbreeding within small fragmented populations. A capacity for long-distance dispersal of pollen and seed, thereby facilitating wide outcrossing, has been demonstrated in some Banksia species (Coates & Sokolowski, 1992; He et al., 2004)

Although widely outcrossed seedlings tended to outperform narrowly outcrossed seedlings, there was no effect of native and non-native sandy soil on the performance of seedlings from within the same progeny class. This result, the measured physical and chemical soil properties, and similar patterns of rhizosphere carboxylate exudation, suggest that, in terms of plant performance, the native and non-native sandy soils are similar, and reinforces the observation that depth to groundwater is the key limiting factor affecting the distribution of B. ilicifolia (Groom et al., 2001; Groom, 2004). Declining B. ilicifolia populations have been linked to groundwater drawdown and/or below-average annual rainfall (Groom et al., 2000). For example, up to 80% reduction in B. ilicifolia tree numbers has been observed when groundwater levels fell by 2 m between two consecutive summers, in conjunction with extremes in summer temperature (Groom et al., 2000). Thus, heterosis following wide outcrossing is unlikely to translate to significant range expansion where access to groundwater is limiting. Genetic rescue, though, may play an important role in increasing the resilience of inbred populations to harsh summer conditions. Field trials over many years with relatively inbred and outbred progeny are required to assess this suggestion. However, the ideal situation of conducting these experiments over two or three generations (Fenster & Galloway, 2000) is unlikely to be achievable in this long-lived, slow-growing species.

In contrast to B. ilicifolia, two congeneric species, both widespread co-dominant overstorey trees, Banksia attenuata and B. menziesii, occur in very large, more or less continuous populations, showing wide ecological amplitude by inhabiting a wide range of topographical positions within the landscape from dune crest to low-lying areas. The impact of groundwater drawdown and harsh summer conditions on these two species is far less than that for B. ilicifolia (Groom et al., 2001). Although further testing is needed, these observations suggest that these robust plants are possibly relatively outbred, with large populations avoiding the genetic problems associated with small isolated populations, emphasizing the relationship between genetic variation, inbreeding, population size and isolation, and ecological amplitude. Some populations of B. attenuata and B. menziesii are known to be completely outcrossing (Scott, 1980), and B. menziesii is self-incompatible (Ramsey & Vaughton, 1991). Although we do not yet have data on genetic variation for B. ilicifolia and its more widespread congeners, widespread plant species generally show higher genetic variation than their rare congeners (Gitzendanner & Soltis, 2000). However, many exceptions to this trend are known, requiring empirical studies of individual taxa to be sure about population genetic architecture.

In the present study, seedling growth differed in response to soils of different texture and chemistry. Decreased growth of local outcrossed seedlings in lateritic soil compared with native sandy soils, was associated with reduced size of the root system, possibly associated with a high soil bulk density or aluminium toxicity. Although the seedlings on lateritic soil had a lower shoot P concentrations than those on sandy soils, it is unlikely that this caused the decrease in growth, because these P concentrations are in the range considered normal for Proteaceae (Foulds, 1993). Also, local outcrossed seedlings had similar P concentrations as nonlocal outcrossed plants, but showed less growth. Interestingly, roots of nonlocal outcrossed plants produced much larger amounts of carboxylates than local-outcrossed seedlings, but this was not in response to low leaf P concentrations and did not lead to increased P uptake, because local-outcrossed plants had the same P concentration at much lower carboxylate exudation. Increased carboxylate exudation may have been a response to high Al concentrations in the rhizosphere (Delhaize et al., 1993; Lambers et al., 2002; Shane et al., 2003). The detoxification of Al by outcrossed plants may have allowed for better growth compared with that of selfed plants (Delhaize et al., 1993; Lambers et al., 2002; Shane et al., 2003).

Waterlogging and water stress had relatively small effects on growth, despite relatively large effects on shoot P concentration. This supports the conclusion that for the duration of the experiment, decreased P uptake was not the most important growth-limiting factor. The 6-week experiment did not reveal significant differences between local and nonlocal outcrossed plants in terms of their tolerance to waterlogging or water stress, but differences might become apparent during longer-term experiments or at different stress levels.

Concluding remarks

Our results demonstrate that wide outcrossing (over 30 km) leads to heterosis in B. ilicifolia and improved performance on nonlocal soils, facilitated in part by an improved capacity to overcome soil barriers through greater root carboxylate exudation. Soil type greatly influenced seedling growth, with weaker root development and lower total plant P uptake on lateritic soils compared with that on the two sandy soils. Extreme water conditions reduced growth, cluster-root production, root exudation and total plant P uptake. These observations suggest that the interaction of narrow ecological amplitude and the genetic consequences of small naturally fragmented populations may partly explain the narrow range of local endemics. On the contrary, natural phenomena such as metapopulation dynamics (for example, rare long-distance pollen dispersal) may provide opportunities for increased genetic variation and increased ecological amplitude. Further research is needed to assess patterns of mating and vigour of seedlings from different-sized populations in order to study the relationship between individual fitness, population density, breeding system and genetic variation.


We gratefully acknowledge the Wesfarmers CSBP Ltd. Agricultural Lab for analysis of soil. Dr Jamie O'Shea and Bob Cooper are thanked for permission to work in the Harry Waring Marsupial Reserve. Bambang Heliyanto thanks AusAid for providing a scholarship to undertake postgraduate research at the University of Western Australia. We are grateful to Kevin Murray for statistical advice, Chris Szota for collection of lateritic soil and many other people, especially Mohammad Nuruzzaman, Stuart Pearse, Sharmin Islam, Djajadi, Drs Matt Denton, Imran Malik, Danica Godgin and Qifu Ma for their help during the course of experimentation, and Prof. Steve Hopper and two anonymous reviewers for their helpful comments that improved an earlier draft of this paper.