Phosphorus uptake, not carbon transfer, explains arbuscular mycorrhizal enhancement of Centaurea maculosa in the presence of native grassland species


† Author to whom correspondence should be addressed. E-mail:


  • 1Previous studies have shown that arbuscular mycorrhizas (AM) enhance the growth of the invasive forb Centaurea maculosa when growing with native grass species. Using 13CO2, we tested the hypothesis that this enhancement is explained by carbon transfer from native species to C. maculosa via mycorrhizal hyphal linkages.
  • 2A C. maculosa plant was paired with one of five native species – three grasses (Festuca idahoensis, Koeleria cristata and Pseudoroegneria spicata) and two forbs (Achillea millefolium and Gaillardia aristata) – in pots that separated the plants with either a mesh barrier (28 µm, excludes fine roots but not hyphae) or a membrane barrier (0·45 µm, excludes roots and hyphae).
  • 313CO2 was added to the atmosphere of either Centaurea or the native species after 20 weeks’ growth. A 25 min pulse application was followed by 7 days’ growth and subsequent harvest.
  • 4The biomass response of C. maculosa was consistent with previous experiments: C. maculosa was larger when growing in mesh barrier pots, when hyphae were able to access the opposite side of the pot; in mesh barrier pots only, biomass varied with neighbouring species. Native plant biomass did not vary between mesh- vs membrane-barrier pots.
  • 5There was no evidence of carbon transfer, either from the native plant to C. maculosa or in the reverse direction.
  • 6Centaurea maculosa contained significantly more phosphorus in mesh-divided pots, but this depended on the neighbouring plant. The P concentration in C. maculosa was significantly higher in mesh-divided pots when growing with a grass and not a forb. Native species contained more P in mesh-divided pots than membrane-divided pots, and P concentration differed between species (higher in forbs than grasses), but did not vary between mesh- and membrane-divided pots.
  • 7Our study suggests that C. maculosa is able to exploit its mycorrhizal symbiosis more effectively than the native grassland species. The mechanism for this appears to be luxury consumption of P through efficient utilization of extra-radical hyphae, but that effect is dependent on neighbouring species, and occurs when growing with a grass neighbour.
  • 8Although no single study can disprove the carbon-transfer hypothesis, our work suggests that AM-mediated neighbour effects are the result of mycorrhizal networks that increase species’ access to P. Whether the synergistic effects of neighbours are due to complementarity of AM fungal symbionts utilized by different plant species, or have to do with the structure of AM networks that develop more extensively with multiple host plants, remains to be investigated.


Invasive plant species can alter soil functions through shifts in nutrient cycling (Evans et al. 2001; Vitousek et al. 1987) and soil biota (Belnap & Phillips 2001), resulting in positive feedback that favours the growth of invasive species (Ehrenfeld, Kourtev & Huang 2001). Invasive species may also increase their competitiveness through exploitation of beneficial interactions with mutualists (Parker & Haubensak 2002; Richardson et al. 2000; Simberloff & Von Holle 1999). This study focuses on the effects of arbuscular mycorrhizas (AM), a plant–fungal symbiosis occurring in most species of herbaceous plants and some woody species (Smith & Read 1997). The fungal partner is dependent on the host plant to supply carbon (Pfeffer et al. 1999), and the host plant potentially receives multiple benefits including increased nutrient status, especially phosphorus (Koide 1991; Smith & Read 1997).

AM can influence plant community composition and the competitive ability of exotic species (Marler, Zabinski & Callaway 1999a) by altering the direction and strength of competitive interactions; by favouring mycotrophic over non-mycotrophic species (Allen & Allen 1984; Sanders & Koide 1994); or by favouring species with greater mycorrhizal dependency (Anderson, Hetrick & Wilson 1994; Hetrick et al. 1994; Smith, Hartnett & Wilson 1999). Additionally, AM fungi may affect species interactions through resource transfer between host plants via common hyphal networks (Newman 1988). Hyphal connections occur between plants (Heap & Newman 1980; Hirrel & Gerdemann 1979; Newman et al. 1994), and there is evidence that C or P can move between plants via these connections (Chiariello, Hickman & Mooney 1982; Francis & Read 1984; Grime et al. 1987; Heap & Newman 1980; Hirrel & Gerdemann 1979; Read, Francis & Finlay 1985; Whittingham & Read 1982).

The potential for resource movement between plants via hyphal networks has major implications for our understanding of plant interactions, but only if the amount of resource transferred is significant enough to alter host plant or donor plant fitness, which has yet to be explicitly shown for endomycorrhizal plant species (Ritz & Newman 1986). Fitter et al. (1998) and Graves et al. (1997) suggest that carbon transfer occurring between plants is retained in fungal tissue within plant roots and, while it could be important to fungal carbon budgets, probably has little effect on host plant fitness (Fitter et al. 1998).

Our study focused on the potential for carbon transfer via mycorrhizal hyphal linkages between Centaurea maculosa Lam. (Asteraceae) and species native to western US mountain grasslands. Centaurea maculosa is native to Eastern Europe, and was introduced into North America in the late 19th century. It was first recognized in Montana in the early 1900s, and now covers millions of acres of rangeland throughout the western USA and Canada (Jacobs, Carpinelli & Sheley 1996). The invasion of C. maculosa has decreased the diversity of native plant communities (Tyser & Key 1988), altering the habitat for native grazers and rangeland animals (Wright & Kelsey 1997). While C. maculosa is highly effective at invading disturbed rangelands, it also invades undisturbed pristine mountain grasslands (Tyser & Worley 1992).

In the field, C. maculosa is heavily mycorrhizal at low densities in recently invaded native grasslands or when dominant (Marler et al. 1999b). In the greenhouse, AM had no effects on C. maculosa biomass when grown alone, but AM altered competitive interactions between C. maculosa and the native grass Festuca idahoensis. When grown with AM fungi, either F. idahoensis was significantly smaller or C. maculosa was significantly larger, compared to interspecific pairs grown without AM fungi (Marler et al. 1999a). We hypothesized that carbon transfer was occurring via AM hyphal linkages, from the native grass to the invasive forb. This hypothesis is supported by previous research showing that carbon transfer occurred from Festuca ovina to Centaurea nigra when plants were mycorrhizal (Grime et al. 1987). In that case, resource transfer from Festuca to Centaurea was associated with enhanced species diversity. In our research, the potential resource transfer via mycorrhizal linkages may be a mechanism whereby C. maculosa gains extra advantage over native competitors, contributing to a decline in diversity of invaded grasslands.

The objective of this experiment was to assess whether C is transferred via AM between invasive species and native plants. We used 13C-labelled CO2 to quantify carbon transfer between species, either from the native plant to Centaurea, or from Centaurea to the native plant. Additionally, all plants were grown with mycorrhizal field soil in pots with a constructed barrier, to allow for a comparison of mycorrhizal plants whose external hyphae either had access to the other half of the pot, or not. We hypothesized that (1) C. maculosa biomass would be largest when growing in pots where hyphal links would be able to establish; (2) biomass patterns would correspond to patterns of 13C movement from native plants to C. maculosa; and (3) carbon transfer would not occur in measurable amounts from C. maculosa to the native species.


pot design and seedling establishment

To investigate C movement between plants, 4 l pots were divided vertically into halves with one of two barrier types. Mesh barriers composed of 28 µm nylon mesh (Sefar, Depew, NY, USA) allow movement of water, solutes and hyphae, but exclude fine roots. Membrane barriers (0·45 µm Magna nylon transfer membrane; Osmonics, Minnetonka, MN, USA) exclude fine roots and hyphae, but have minimal effects on water movement as measured with time domain reflectometry in pots watered on only one side of the barrier (S. Zimmerley and C.A.Z., unpublished results). Barriers were attached to the inside of the pots with rubber cement, duct tape to reinforce the cement, and silicone antimicrobial aquarium sealant applied along the points of contact between the barrier and the pot surface.

Pots were filled with an 8 : 1 silica sand : field soil mixture. Field soil was from a bluebunch wheatgrass (Pseudoroegneria spicata Pursh A. Love), rough fescue (Festuca scabrella Torrey ex Hook), and Idaho fescue (Festuca idahoensis Elmer)-dominated grassland in western Montana, which was invaded by C. maculosa. A 1/8-strength modified Hoagland's nutrient solution was applied at the rate of 125 ml per plant every 2 weeks. Phosphorus was excluded from the solution, unless plants showed signs of P-deficiency. At five application times (c. one per month) a 1/16-strength P addition was included in the nutrient solution, and added to all pots.

Centaurea maculosa was grown with one of five native species: Koeleria cristata L. Pers. (Junegrass), Pseudoroegneria spicata (bluebunch wheatgrass), Festuca idahoensis (Idaho Fescue), Gaillardia aristata Pursh (blanket flower), and Achillea millefolium L. (yarrow). The experimental design was a complete factorial, with five neighbouring species, two barrier types, and 10 replicates of each treatment, for a total of 100 pots. Approximately five of the 10 replicates were used for application of 13C to the native plant to assess C movement to C. maculosa, and five replicates were used for application of 13C to C. maculosa to assess C movement from the invasive forb to native species. Germinants were planted into pots in March 2000; pots were randomly located on greenhouse benches, and moved to a new position weekly to minimize the effects of environmental gradients. Reference plants were grown separately from spiked plants during the labelling and the chase period to measure natural abundances of 13C in each species. Additional plants were grown with spiked pots during the chase period, to measure atmospheric contamination due to respiration of 13C from spiked plants. The average day temperature in the greenhouse was 22 °C, with a night temperature of 18 °C, and a natural seasonal photoperiod.

Plants grew for 5 months prior to 13CO2 application. We used a pulse-chase application, the pulse being a high concentration added during a 30-min period, and the chase period the time after application when the isotope is incorporated into donor plant tissue and potentially transferred to the receiver plant. In a preliminary study with 27 replicates, the addition of 50 ml 13CO2 resulted in an increase of 1090 ± 45 µmol mol−1 CO2 in the sampling bag (Vaisala CO2 sensor GMM222, Helsinki, Finland). In the same study, we harvested spiked plants 3, 7 or 10 days post-application, and found that δ13C in spiked plants did not change significantly between chase periods, with an average of 48 ± 0·5% of the 13C taken up remaining in the plant tissues, 55% of that in root tissue (data not shown). The 7-day chase period was chosen to allow time for C movement between plants while C. maculosa was bolting, and to use a chase period similar to previous studies that demonstrated carbon transfer (Simard et al. 1997a; Simard et al. 1997b).

To separate the atmosphere of the spiked plant from the potential receiver plant, pots were fitted with a plastic collar just prior to 13CO2 application, which provided a rigid semicircular lip around half of the pot. A neoprene skirt was attached to the plastic collar, which partially covered the surface of the soil. Two layers of heavy-duty plastic wrap were used to cover the soil, and were taped to the neoprene to minimize diffusion of 13CO2 into the soil. A 10 l clear Tedlar polyvinyl fluoride gas sampling bag (Midan, Chino, CA, USA) was taped to the upper rim of the collar over the spiked plant. Then 50 ml 99%13CO2 (Isotec, Miamisburg, OH, USA) was injected using a gas syringe (Hamilton Samplelock syringe, Reno, NV, USA) through a Teflon septum on the gas-sampling bag. The shoot of the non-spiked plant was isolated inside a plastic container during the pulse period to minimize uptake of any leaked 13CO2.

Twenty-five minutes after addition of 13C, the gas sampling bags were removed in a separate room, in an airstream to move residual 13CO2 away from the non-spiked plants. Pots were then placed in a second greenhouse, separate from that where the labelling occurred, to grow for 7 days before harvest. Fans were used during the evenings of the 7-day chase period to diffuse any respired 13C. Carbon transfer was tested in both directions; from spiked C. maculosa (donor) to native plants (receiver), and from spiked native plants (donor) to C. maculosa (receiver).

harvest and tissue analysis

Plants were removed from the soil, washed, separated into root and shoot tissue, and dried for 48 h at 65 °C. Biomass was recorded, and a subset of root samples set aside for analysis of mycorrhizal colonization. The remaining tissue was ground into fine powder in a cyclone sample mill (Udy Corporation, Fort Collins, CO, USA). Root and shoot tissues from non-spiked plants were ground separately, while root and shoot tissues from each spiked plant were combined. The University of California, Davis, Stable Isotope Laboratory analysed plant material for total C and 13C content, using an isotope ratio mass spectrometer. δ13C is based on the ratio of 13C to 12C of the tissue analysed, and is calculated as:

13C = 1000[(Rsample/RPDB) − 1](eqn 1)

where RPDB = 0·0112372, the ratio of 13C to 12C in the Pee Dee Belemnite standard. To account for differences between species in 13C due to water-use efficiency, shifts in δ13C were calculated as:

δ 13Ctissue  −  δ13Creference plants(eqn 2)

Root segments for mycorrhizal analysis were rehydrated, cleared for 48 h in 10% KOH at room temperature, acidified for 12 h in 3% HCl, and stained with a Trypan blue/lactoglycerol stain (Phillips & Hayman 1970). Root segments were scored for mycorrhizal colonization using a magnified intersections method (McGonigle et al. 1990), and data reported as percentage of intersections observed that contain AM structures.

phosphorus analysis

Dried plant tissue was prepared for P analysis by incinerating at 500 °C for 5 h, and digesting the resulting ash in 2 m HCl. Phosphorus quantities were measured on digests by inductively coupled plasma atomic emission spectrometry analysis by the Soil, Plant and Water Analytical Laboratory (Montana State University).


plant biomass

Biomass of C. maculosa was significantly affected by the identity of neighbour (two-way Type III anova; F4,83 = 4·59, P = 0·002) and the barrier type (F1,83 = 5·37, P = 0·02), with an increase in biomass when growing in mesh-divided as compared to membrane-divided pots, and greatest biomass when growing with F. idahoensis or G. aristata (Fig. 1a). Considering only plants growing in membrane-divided pots, there was no significant effect of neighbour on C. maculosa biomass (one-way anova; F4,43 = 1·38, P = 0·26); however, in mesh-divided pots neighbour identity had a significant effect on C. maculosa biomass (one-way anova; F4,40 = 4·19, P = 0·006). Native plant biomass varied significantly between species (Fig. 1b; F4,83 = 22·78, P < 0·001), but was not affected by barrier type (F1,83 = 2·03; P = 0·16), and the species–barrier interaction was not significant (F4,83 = 1·60, P = 0·18).

Figure 1.

(a) Biomass of Centaurea maculosa when grown with one of five neighbours (P.spic = Pseudoroegneria spicata, F.ida = Festuca idahoensis, K.cris = Koeleria cristata, G.aris =Gaillardia aristata, and = Achillea millefolium), in pots with either a mesh or membrane divider. (b) Corresponding biomass of the native plants when growing with C. maculosa. Error bars represent SEM. Lower-case letters indicate statistically different averages for species biomass combined across barrier types.

mycorrhizal colonization

Mycorrhizal colonization differed between species (two-way anova: F5,24 = 5·02, P = 0·003) but not between barrier types (F1,2 = 0·02, P = 0·88), and the species–barrier interaction was not significant (F5,24 = 1·11, P = 0·38). Percentage colonization was 64% for C. maculosa, 62% for P. spicata, 37% for F. idahoensis, 54% for K. cristata, 55% for G. aristata and 83% for A. millefolium.

carbon transfer

  • 13C in plants adjacent to pots that had received 13CO2 spiking was compared to the δ13C of reference plants that were not spiked, and were grown separately from spiked pots during labelling and the 7-day chase period. More 13C in the former set of plants would indicate that a measurable amount of 13CO2 was respired from spiked plants and fixed by adjacent unlabelled plants. There was no significant difference between reference plants and plants used to test for 13C leakage (δ13C root tissue: −28·1 ± 0·2 vs−28·4 ± 0·4‰, t24 = 0·60, P = 0·55; shoot tissue: −29·8 ± 0·5 vs−29·4 ± 0·3‰, t24 = 0·84, P = 0·41). Spiked plants, 7 days post-labelling, showed an average shift in δ13C of 87·9 ± 4·8‰.

Higher 13C abundances in receiver plants growing in mesh-barrier as compared to membrane-barrier pots would be evidence for carbon transfer between plants via mycorrhizal hyphal linkages. Analysis of receiver plant tissues showed that shifts in δ13C for C. maculosa did not vary between barrier types for either root tissue (Table 1 two-way anova ln-transformed data; F1,45 = 0·23, P = 0·64) or shoot tissue (two-way anova ln-transformed data; F1,45 = 0·14, P = 0·72). Neither root nor shoot tissue δ13C in C. maculosa varied between species of spiked plant (root tissue: F4,45 = 2·14, P = 0·10; shoot tissue: F4,45 = 1·59, P = 0·20), or with a barrier–spiked species interaction (root tissue: F4,45 = 1·16, P = 0·19; shoot tissue: F4,45 = 0·95, P = 0·44).

Table 1.  Shifts in δ13C (‰) ± SE (sample size in parentheses) of Centaurea maculosa receiver plants and spiked native plants
Plant materialSpiked speciesAchillea millefolium
Pseudoroegneria spicataFestuca idahoensisKoeleria cristataGaillardia aristata
Shoot  2·2 ± 0·8 (6)  1·6 ± 0·8 (5)  1·2 ± 0·7 (5)    0·7 ± 0·5 (4)    3·3 ± 1·1 (5)
Root  1·5 ± 0·5 (6)  1·7 ± 0·4 (5)  1·8 ± 0·4 (5)    0·8 ± 0·3 (4)    2·8 ± 0·7 (5)
Spiked plant82·1 ± 10·5 (6)75·8 ± 13·1 (3)77·0 ± 16·7 (2)129·8 ± 30·9 (4)105·7 ± 3·7 (5)
Shoot  4·5 ± 1·7 (4)  0·9 ± 0·5 (3)  1·2 ± 0·3 (5)    1·4 ± 0·5 (4)    6·1 ± 4·5 (4)
Root  4·2 ± 2·4 (4)  0·9 ± 0·4 (3)  1·2 ± 0·2 (5)    1·2 ± 0·4 (4)    4·3 ± 2·5 (4)
Spiked plant79·5 ± 11·1 (5)63·2 ± 25·2 (3)77·9 ± 7·4 (3)119·2 ± 23·1 (4)  64·1 ± 19·2 (3)

When 13CO2 was applied to C. maculosa to determine whether C moved from Centaurea to native species, there was a significant difference between species in δ13C of shoot tissue (two-way anova, F4,42 = 5·16, P = 0·002), but not root tissue (two-way anova: F4,42 = 0·71, P = 0·78). Shifts in P. spicata and A. millefolium were significantly greater than shifts in F. idahoensis or K. cristata (post hoc Bonferonni tests, α = 0·05; Table 2). Mesh- vs membrane-barrier pots did not result in a significant shift in δ13C in either root (F1,42 = 2·65, P = 0·11) or shoot tissue (F1,42 = 1·76, P = 0·19), and the barrier–species interaction was not significant (root tissue: F4,42 = 0·71, P = 0·59; shoot tissue: F4,45 = 1·49, P = 0·23).

Table 2.  Shifts in δ13C (‰) ± SE (sample size in parentheses) of native species receiver plants and Centaurea maculosa spiked plants
Plant materialSpeciesAchillea millefolium
Pseudoroegneria spicataFestuca idahoensisKoeleria cristataGaillardia aristata
Shoot1·6 ± 0·3 (5)0·6 ± 0·4 (4)  0·5 ± 0·4 (4)0·8 ± 0·2 (4)2·3 ± 0·9 (5)
Root1·2 ± 0·3 (5)0·9 ± 0·5 (4)  1·6 ± 0·2 (4)0·7 ± 0·4 (4)1·5 ± 0·6 (5)
Spiked C. maculosa  86·6 ± 9·7 (16)  
Shoot2·9 ± 0·5 (5)0·4 ± 0·3 (4)  0·5 ± 0·2 (5)2·3 ± 0·8 (3)1·9 ± 0·7 (4)
Root2·2 ± 0·3 (5)1·4 ± 0·4 (4)  1·5 ± 0·5 (5)1·8 ± 0·1 (3)1·4 ± 0·6 (4)
Spiked C. maculosa  69·9 ± 13·7 (7)  

phosphorus in plant tissues

Plant phosphorus was calculated as total P content (mg P per plant) and P concentration (mg P g−1 plant biomass). There was significantly more total P in C. maculosa when growing in mesh-divided than in membrane-divided pots (two-way anova; F1,91 = 13·96, P < 0·001). Neighbour also had a significant effect (F4,91 = 3·96, P = 0·005), with highest P content when growing with F. idahoensis or G. aristata (Fig. 2a). Post hoc comparisons showed that the P content of C. maculosa was greater in mesh-divided than in membrane-divided pots when growing with F. idahoensis (t16 = 2·9, P = 0·01) or K. cristata (t15 = 2·20, P = 0·04).

Figure 2.

(a,b) Phosphorus content and (c,d) P concentration in Centaurea maculosa and native species growing in mesh- or membrane-divided pots. Letters represent significant (P < 0·05) differences between neighbouring species; * represents differences between mesh- and membrane-barrier pots. Error bars represent SEM.

Total P content in native species varied between barrier types (two-way anova, ln-transformed data; F1,89 = 4·89, P = 0·03), and species (F4,89 = 13·21, P < 0·001), and there was a significant barrier–species interaction (F4,89 = 2·98, P = 0·02). Most P was accumulated by A. millefolium, followed by K. cristata and P. spicata (Fig. 2b). Phosphorus content in native species differed significantly between barrier types for K. cristata (t16 = 2·46, P = 0·03) and A. millefolium (t18 = 2·79, P = 0·01).

Combined P content in both plants in a pot was 8·6 ± 0·3 mg per pot in mesh-barrier pots and 6·9 ± 0·3 mg per pot in membrane-barrier pots. Barrier type significantly affected the quantity of P taken up (two-way anova; F1,88 = 14·06, P < 0·001), species composition was not significant (F4,88 = 1·47, P = 0·22), and the barrier–species interaction was close to significant (F4,88 = 2·24, P = 0·07).

Tissue P concentration in C. maculosa was affected by barrier (two-way anova, arcsine-transformed data; F1,91 = 1·18, P < 0·001), but not by neighbour (F4,91 = 1·17, P = 0·33), and there was no significant barrier–neighbour interaction (F4,91 = 1·80, P = 0·14). The P concentration in C. maculosa was higher in mesh-divided pots only when growing with one of the three grasses (Fig. 2c; P. spicata as neighbour: t18 = 3·78, P = 0·001; F. idahoensis: t16 = 2·95, P = 0·009; K. cristata: t15 = 2·02, P = 0·06).

Native plant P concentration did not vary between mesh- and membrane-divided pots (two-way anova, square root arcsine-transformed data: F1,89 = 0·67, P = 0·42), but there was a significant difference between species (F4,89 = 24·97, P < 0·001), and a species–barrier interaction (F4,89 = 2·66, P = 0·04). Forbs had a higher P concentration than the grasses (Fig. 2d), and the only close-to-significant difference for individual species between mesh- and membrane-divided pots was for G. aristata, which showed a decrease in P concentration when growing in mesh-divided as compared to membrane-divided pots (t13 = 1·95, P = 0·07).


Many species of Centaurea are very invasive in western North American grasslands, and LeJeune & Seastedt (2001) have proposed that Centaurea's agressiveness is partly because it competes more strongly than native species for P. This study underscores the importance of AM for increasing the competitive ability of C. maculosa. While all plants in this study were mycorrhizal, plants growing in mesh-divided pots had access via mycorrhizal hyphae to their neighbour's rooting zone. Of the six species used, only C. maculosa and A. millefolium were larger in mesh-divided than in membrane-divided pots. The biomass of the other native species did not differ with barrier type, suggesting that C. maculosa exploited its mycorrhizal symbiosis more effectively than most of the native species. Mycorrhizal effects on species interactions were also evident in that total biomass of C. maculosa was affected by neighbouring species only in mesh-divided pots.

Inspired by research suggesting the potential for carbon transfer between species via mycorrhizal networks (Simard et al. 1997b), this study was designed to quantify C movement between native grassland species and an invasive forb. However, our data showed no evidence of carbon transfer between plants via hyphal linkages, despite biomass patterns consistent with the carbon-transfer hypothesis. The pulse-chase application of isotope can document carbon transfer only during the chase period (Simard, Durall & Jones 1997a) which, in this study, was 7 days of a 5-month experiment. To maximize the likelihood of documenting carbon transfer, we applied 13C just as C. maculosa was beginning to bolt. We cannot rule out the possibility that carbon transfer occurs between plants at the seedling establishment stage or under different environmental conditions.

An alternative mechanism for mycorrhizal enhancement of C. maculosa growth is increased P nutrition. Phosphate ions diffuse slowly in soils and, after nitrogen, P is usually the most limiting nutrient for plant growth. Plant P content, a measure of how much P is extracted from the soil, was significantly greater in both C. maculosa and native species growing in mesh-divided pots, suggesting that plants in mesh-barrier pots accessed P from the opposite side of the pot.

In contrast, only C. maculosa had a higher tissue P concentration when growing in pots with hyphal access to both sides, suggesting that it used its increased access to P in two ways – plants produced more biomass, and accumulated P to concentrations in excess of those necessary for that production – that is, luxury consumption. However, our results also show that luxury consumption in C. maculosa is dependent on neighbouring species, occurring with grass, but not forb, neighbours (Fig. 2c). Plant nutrient concentration is correlated with a capacity for rapid growth under favourable conditions (Aerts & Chapin 2000). It also provides the plant with nutrients that can be moved into reproductive tissues, which can influence root development and subsequent P uptake of offspring (Zhu & Smith 2001).

The potential for neighbouring species to influence luxury consumption of P might be the result of fungal community attributes that vary with host plant. For example, production of extra-radical hyphae may be related to root morphology, with an increase in external hyphae as roots become coarser (Miller, Reinhardt & Jastrow 1995). The fine rooting structure of grass species relative to forbs suggests the possibility that, if grasses produce fewer external hyphae than forbs, there are potentially more small soil pores available for AM hyphae associated with C. maculosa to exploit when growing with a grass. Alternatively, a shift in fungal species associated with different host plants could change the potential to access P at different distances from the roots (Jakobsen, Abbott & Robson 1992; Jakobsen, Gazey & Abbott 2001). In both scenarios, C. maculosa would have greater access to P from its neighbour's rooting zone.

Mycorrhizal hyphal networks have been envisioned as a conduit for movement of resources between plants (Perry 1998; Wilkinson 1998). However, that plants in this study showed combined greater P content when growing together in pots where hyphal networks were able to form suggests increased hyphal development with multiple host plants. Instead of resources moving between plants, plants may be accessing resources from a hyphal network that functions as an additional pool of resources. The increased access to nutrients resulting from networks connecting two plants may be the result of complementarity of fungal symbionts or synergistic effects of host plants on hyphal network development.

Fitter (1977) suggested that AM may be a disadvantage to a host plant even in a low-P environment, if neighbouring species gained more of an advantage from AM. Our research demonstrates the importance of AM in affecting the competitive ability of an invasive forb with native grassland species, and suggests that part of C. maculosa's ability to invade grasslands depends on its mycorrhizal symbiont to increase tissue P concentration, but not, apparently, for interspecific transfer of C. The mechanisms for neighbour effects on mycorrhizal function of C. maculosa remain speculative. The role of multiple host plants in contributing to hyphal networks and increasing access of host plants to limiting resources needs to be further investigated.


R. Bunn, J. Calabrese, S. Lewellen and S. Zimmerley assisted with greenhouse and laboratory work. Thanks to S. Simard for technical advice regarding isotope application, and to Montana State University Technical Services for the design of collars for attaching gas-sampling bags to pots. M. Marler and two anonymous reviewers provided helpful comments on the manuscript. This research was funded by NSF grant DEB-9726829 to Ragan Callaway and Cathy Zabinski, and a grant to Ragan Callaway from the Andrew W. Mellon Foundation.