Network formation by rhizomorphs of Armillaria lutea in natural soil: their description and ecological significance
Abstract
Armillaria lutea rhizomorphs in soil were mapped over areas of 25 m^{2} at a Pinus nigra (site I) and a Picea abies (site II) plantation. Rhizomorph density was 4.3 and 6.1 m m^{−2} soil surface with 84% and 48% of the total rhizomorph length in the mapped area interconnected in a network at site I and site II, respectively. At site I there were only two network attachments to Pinus stumps, but at site II many more to Picea roots and stumps. Anastomoses of rhizomorphs resulted in cyclic paths, parts of the network that start and end at the same point. Connections between different rhizomorph segments were shown to allow gaseous exchange. The network at site I consisted of 169 rhizomorphs (‘edges’), and 107 rhizomorph nodes (‘vertices’). Disruption of two critical edges (‘bridges’) would lead to large parts (13% and 11%) being disconnected from the remainder of the mapped network. There was a low probability that amputation of a randomly chosen edge would separate the network into two disconnected components. The high level of connectedness may enhance redistribution of nutrients and provide a robust rhizomorph structure, allowing Armillaria to respond opportunistically to spatially and temporally changing environments.
 Armillaria rhizomorphs
 network structure
 graph theory
 connectedness
 ecological persistence
 robustness
Introduction
The clonal dispersal of plant pathogenic Armillaria species occurs in temperate climatic zones by growth through soil of specialized strands, called rhizomorphs. These shoestringlike strands are 1–3 mm in diameter with a reddish brown to black outer cortex layer (Cairney et al., 1988) usually in the upper 30 cm soil layer (Redfern, 1973). Clones thus formed may persist over centuries and may be of impressive size (Smith et al., 1992; Ferguson et al., 2003) if there continue to be sufficient sources of nutrition for absorption (Rizzo et al., 1992) and translocation (Granlund et al., 1985; Cairney et al., 1988; Gray et al., 1996) under turgor pressure (Eamus & Jennings, 1984). Although rhizomorphs are in general insulated from the environment the peripheral hyphae may act as organs of nutrient uptake (Pareek et al., 2001) with oxygen diffusing through a central gasfilled cavity (Pareek et al., 2006). Contact of rhizomorphs with tree roots can result in treetotree spread of the fungus, even when direct contact between diseased and healthy roots is not made. In some species rhizomorphs grow epiphytically along roots (Baumgartner & Rizzo, 2001). Rhizomorphs are produced during the various stages of wood decay, but the extent of growth is speciesdependent and influenced by habitat and environmental conditions and the presence of secondary colonizers (e.g. Prospero et al., 2006).
Networks of fungal hyphae growing in pure culture (Mihail & Bruhn, 2005) and soil microcosms (Bolton & Boddy, 1993; Harris & Boddy, 2005) have been described in terms of nutrient translocation (Watkinson et al., 2005) and growth strategies in relation to grazing (Kampichler et al., 2004), but fungal networks in undisturbed ecosystems have been mapped only rarely (Thompson & Rayner, 1983). Although the rhizomorph growth habits of 15 species of Armillaria were described following the placement of inoculum segments in small volumes of soil in plastic bags (Morrison, 2004), the ecological relevance of Armillaria rhizomorph networks has only partially been appreciated. It is generally recognized that fungal networks occur commonly and their existence is of ecological relevance, e.g. for translocation of nutrients and carbon by mycorrhizal fungi (Leake et al., 2004), but quantitative tools to analyse fungal networks have not been developed. In this study, we mapped a rhizomorph network of Armillaria lutea in natural soil at two sites and analysed characteristics of the networks in terms of foraging strategies (Dowson et al., 1989) and the persistence of networks in time and space using graphtheoretical concepts (Harary, 1969; Wilson, 1979).
In graph theory, the term ‘network’ is used in a technical sense to mean a type of diagram, called a ‘graph’, to which the numerical values of some quantity are attached. A graph consists of a set of points or nodes, called ‘vertices’, and connections between them called ‘edges’. A graph is ‘connected’ if there is a ‘path’ (sequence of edges) that connects every vertex in the graph. It is ‘disconnected’ if there is no such path. A ‘cycle’ within a graph is a path that starts and ends at the same point. We identify the branching or fusion of rhizomorphs at a point in the network as a vertex, and a rhizomorph connecting two vertices as an edge. Transport of, for example, nutrients from one vertex to another is determined by the presence or absence of edges, as nutrients flow easily through the medulla of rhizomorphs (Granlund et al., 1984). If a rhizomorph is removed, transport of nutrients between two vertices is not prevented if these vertices are connected by more than one rhizomorph. The significance of vertices is that they bring flexibility to the rhizomorph system, as multiple edges result in more ways to transport nutrients.
Although the mathematical concepts of graph theory have been widely applied, there have been few examples of applications in population biology until relatively recently. Network theory has been applied to networks at the gene and protein level and increasingly existing techniques are being applied to ecological systems (Proulx et al., 2005). A particular theoretical question concerns the architecture of biological networks. Recently, Southworth et al. (2005) analysed Quercus garryana–mycorrhizal associations (20 trees/40 fungal morphotypes network) using graphtheoretic concepts. They concluded that all trees had about the same linking to fungal morphotypes in the network, but that certain morphotypes, e.g. Cenococcum geophilum, had more links to trees than did other morphotypes. However, the authors had no direct evidence of physical sharing of resources through these links. In this study we apply graph theory to an analysis of the rhizomorph network and discuss an ecological interpretation.
Materials and methods
Mapping of rhizomorph networks
In two c. 40yearold tree plantations (site I: Pinus nigra ssp. maritima; site II: Picea abies) near Wageningen, the Netherlands, we prepared maps of rhizomorphs of A. lutea in soil over a plot area of 25 m^{2}. The plantation size at both sites was c. 1 ha at an elevation of 30 m above sea level, with horizontal aspect. The Pinus site had a dense shrub layer consisting entirely of Prunus serotina. In the c. 10 years before the study was undertaken both shrubs and pine trees were heavily attacked by Armillaria, but in the sampling year this was less so for Pr. serotina. Furthermore the Pinus site had a moderate dense herb layer of Deschampsia flexuosa. In the Picea plantation there was neither a shrub nor a herb layer. Although both sites were situated on Pleistocene moraine sand, the Picea plot was slightly podzolic, whereas the Pinus plot was not. At each site the soil and surface litter was handremoved up to c. 25 cm depth and rhizomorphs were located in 1 m^{2} grids and drawn on a twodimensional map at a scale of 1 : 10. The depth was not recorded because this was small compared with the surface dimensions. Isolates of the rhizomorphs from both sites were identified by A. PérezSierra (Royal Horticultural Society, Wisley, UK) as representing A. lutea Gillet [=Armillaria gallica Marxm. and Romagn.=Armillaria bulbosa (Barla) Kile and Watl.] with PCR–restriction fragment length polymorphism of the IGSregion of the rRNA gene using speciesspecific primers (Anderson & Stasovski, 1992; Chillali et al., 1997).
Observation of internal connectedness of rhizomorph anastomoses
Anastomoses of rhizomorphs were frequently observed. To investigate whether fused rhizomorph segments were internally connected or not, air was forced through waterimmersed rhizomorphs at one end and the occurrence of air bubbles was observed distally beyond the point of fusion. For Xray microscopy (Skyscan1072 desktop Xray microtomograph), two rhizomorph segments fused by anastomosis were cut with a sharp blade to a length of c. 2 mm. The combination of Xray transmission technique with tomographical reconstruction gave threedimensional information about the internal microstructure, constructed as a set of flat crosssections. Photographs were taken at 21 heights (steps of 0.091 mm), starting above the point where the rhizomorph segments were fused, and ending below this point.
Decay time of dead rhizomorphs
The decay time of dead rhizomorphs was monitored to determine whether dead remnants of Armillaria would be present and mapped at the two sites. Rhizomorphs from site I were killed by gamma irradiation (25 kGy), and 10 cm pieces were incubated in pots containing forest soil of low (–24.6 kPa) or high (−3.9 kPa) water potential. Prior to incubation, soil was airdried for 1 week and sieved through a 1.0 mm mesh. The fresh weight of the rhizomorph pieces per pot was recorded after washing them with water and drying between filter paper. To estimate their dry weight, the water content of additional fresh rhizomorph pieces was determined. The incubation temperatures were high (20°C) or low (10°C), roughly encompassing the range of soil temperatures over the spring to autumn period. A soil temperature typical of winter was not used because of low microbial activity and decay rates. At four harvest times (4, 10, 16 and 30 weeks), the soil was sieved through a 1.0 mm mesh under tap water. The dry weight of the remaining rhizomorph pieces was determined after 24 h at 105°C, and the percentage dry weight loss was used as a measure of the state of decay (three replicates per treatment). Similarly, in December 1997 two dead 10 cm rhizomorph pieces were put in each of 24 nets (mesh size of 1.1 mm) containing sieved forest soil. The nets were buried in the forest soil at a depth of 5–10 cm and after 4, 10, 16 and 30 weeks the nets were recovered and the dry weight of the remaining rhizomorph pieces determined (six replicate nets on each occasion).
Results
Mapping of rhizomorph networks
Total rhizomorph length in the observed area was 109 (site I) and 152 m (site II). At several places interconnected rhizomorphs (black lines in Fig. 1) crossed the boundary of the mapped area, indicating that the rhizomorph system extended beyond the observed areas. Cyclic paths, parts of the network that start and end at the same point, were observed as the result of branching and subsequent anastomoses between rhizomorph segments. In many cases, larger cycles were embracing or closely connected to one or more smaller cycles (e.g. the larger cycle A connected to the smaller cycle B in Fig. 1a). Also, many small cycles were produced at this finer scale (Fig. 2a), giving rise to a complex network structure.
The largest connected component can readily be visualized by reducing the mapped rhizomorph system to the cyclic paths (Fig. 1, red lines). In a number of cases several small cycles occurred closely together within the rhizomorph system (e.g. at C in Fig. 1a) but in other cases cycles were larger and simpler in form (e.g. at D in Fig. 1a). At site I, the largest connected component within the mapped area was attached only twice to a dead stump of P. serotina, although attachments may occur outside this area. At site II, many more attachments to Picea stumps and roots were observed −73 in total. Also, at site II there were 253 rhizomorphs not being part of the largest connected component, which were attached to tree roots and less than a few centimetres in length.
Observation of internal connectedness of rhizomorph anastomoses
Forcing air through one end of the rhizomorph segment showed air bubbles at the distal end beyond the point of fusion (Fig. 2b), indicating a continuity of air space between the rhizomorphs. This was confirmed by Xray crosssection analysis of two fused rhizomorph segments (Fig. 2c). Of the 21 images taken at decreasing heights, the middle one demonstrated clearly the presence of a continuum between the two segments (Fig. 2d).
Decay time of dead rhizomorphs
Dead rhizomorph remnants that had decayed under controlled conditions, or when buried in the forest soil, for 30 weeks were reduced to many small brittle pieces, which were hollow or reduced to spiral melanin sheaths. Such deteriorated pieces of rhizomorph were never encountered in the network mapping at either site. Even by 16 weeks rhizomorph remnants occasionally had broken down into smaller parts, but these retained some physical integrity and were recorded. High temperature and high soil water potential appeared to favour decay of dead rhizomorphs, but only soil water potential affected weight loss (Fig. 3) significantly (P<0.01), and then only at 4 and 30 weeks.
Analysis of Rhizomorph networks
The mapping of the rhizomorph network in the observed area at site I (Fig. 1a), here referred to as graph G_{1}, is most likely a ‘directed graph’, i.e. a graph in which arrows, single or in both directions, can be assigned to the edges, if this information is available. Flow of nutrients through rhizomorphs and mycelial cord systems can be simultaneously bidirectional (Granlund et al., 1985; Cairney, 1992, 2005; Gray et al., 1996; Olsson & Gray, 1998). However, for a given nutrient at a given time and place in a rhizomorph network the net flow may be onedirectional, and indeed switch in response to nutritional or other environmental cues. G_{1} is not a connected graph but it can be expressed as the union of connected graphs, each of which is defined to be a ‘component’ of G_{1}. Thus, in G_{1} (Fig. 1a) the blue edges are not connected to the red or black edges. The largest connected component of G_{1} forms 84% of the total graph. Focusing now only on this largest connected component (red line in Fig. 1a), a topologically equivalent ‘planar’ graph G_{2} (Fig. 4) can be constructed for ease of visualization in which edges do not cross. G_{2} lies entirely within the boundary of the mapped area.
Graph G_{2} (Fig. 4) consists of 107 vertices (n) and 169 edges (m). The degree of a vertex, d(v), is the number of edges incident with this vertex, which mostly equals 3 (83% of vertices) as a result of simple branching, but 17 vertices (16%) have degree 4 (i.e. vertices 16, 20, 28, 30, 31, 41, 48, 59, 61–64, 74, 76, 77, 85 and 106). Although degree4 vertices may be the result of multiple branching at one node, our observations point to the possibility of anastomosis of two crossing rhizomorphs. A ‘simple’ graph is a graph without multiple edges between vertices and without loops, where a ‘loop’ is an edge from a single vertex to itself. G_{2} is not a simple graph, as it has 23 double edges, for example between vertices 83 and 84, and five loops, for example at vertex 107 (Fig. 4).
Connectedness is a basic characteristic of a network. The degree to which the rhizomorph network is connected may have ecological implications: a strongly connected network will experience only small consequences from the amputation of rhizomorphs. The connectedness of a network can be determined by calculating the number of edges that must be removed in order to disconnect the graph. If removal of one edge results in a disconnected graph then such an edge is called a ‘bridge’ and, by inspection, occurs in G_{2} nine times, namely edges 62, 65, 68, 73, 78, 81, 155, 163 and 168. Disruption of bridges connecting two large parts of the network has a major impact on the whole network. So, amputation of rhizomorph 78 would disconnect 13% of the rhizomorph system, based on number of vertices. For rhizomorph 81 this percentage is similar (11%), but disruption of one of the other seven bridges would disconnect only 1–4% of the network. Thus, disruption of a bridge in the original connected graph G_{2} gives rise to two connected graphs (components), each with their own network properties. Depending on the location of resources, if a network is separated by disruption (either of a bridge or a set of edges) into two components of the same size, both components are likely to persist, but if one is small there will only be minor benefits of nutrient redistribution and the small component may have little chance to persist. In G_{2} (n=107) the removal of edges 26, 150, 148 and 112 disconnects the graph into two components of about the same size, containing 52 and 55 vertices.
In a planar graph the twodimensional regions bounded by the edges in the graph are called ‘faces’. Euler's formula (Wilson, 1979) states that for a connected graph n−m+f=2, where f is the number of faces. In G_{2} the number of faces equals m−n+2=169–107+2=64. The degree of a face, d(F), is defined as the number of edges on the boundary of that face. A large facedegree may relate to a large region where Armillaria has not foraged. A few faces have a very large degree, but the median value is 3 (Fig. 5). The frequency distribution of vertexdegrees has been used to characterize different types of networks and has been applied to a mycorrhizal association network (Southworth et al., 2005). However, for the Armillaria rhizomorph networks the vertex degrees were either 3 or 4, which precludes this form of analysis for vertices.
For a cycle in a connected graph, removal of any one edge will still result in a connected graph. If this procedure is then repeated with one of the remaining cycles, continuing until there are no cycles left, then the graph that remains is still one that connects all the vertices. This graph is called a ‘spanning tree’. The ‘matrixtree theorem’ (Harary, 1969) can be used to calculate the number of spanning trees in any connected simple graph, obtained by removing any multiple edges and loops. The number of spanning trees in G_{2} equals 5.6 × 10^{21} and one spanning tree G_{3} is shown in Fig. 6. Due to this high number of possible spanning trees, the probability that removal of a randomly chosen edge disconnects the network into two components is infinitesimally small. The number of spanning trees of a network may be interpreted as a measure of the robustness of the network, which in turn could be related to ecological persistence, the ability of the network to respond to spatially and temporally changing environments.
In the rhizomorph mapping (Fig. 1a) a rhizomorph was attached in only two instances to a root or stump, serving as a nutrient source. The first source (S_{1}) is attached to vertex 81 and the second source (S_{2}) is attached to edge 44. Thus, G_{2}, the largest component of the rhizomorph network, is attached to both nutrient sources, but disruption of these two rhizomorphs would be sufficient to remove G_{2} from the sources within the observed area. Disruption of rhizomorphs which cross the boundary of the observed area would remove the mapped system from sources outside the mapped area. The geographical distance to the furthest vertex in G_{2} (vertex 62) measures 4.19 m from S_{1} and 4.56 m from S_{2}; the distance between S_{1} and S_{2} is 2.2 m. Assuming that all parts of the rhizomorph network need access to nutrients, the distance over which the nutrients and water have to be transported in relation to a source is important and related to network structure. In a connected graph, the ‘distance’d(v_{i}, v_{j}) from vertex i to vertex j is the length of a ‘shortest path’ from vertex i to vertex j. Here, length is expressed in number of edges traversed rather than the physical length (m) of edges. For example, vertex 81 is connected to source S_{1} in graph G_{2} (Fig. 4); the number of edges traversed in a shortest path from vertex 81 to vertex 62 is 15. There are 16 such paths, one of which, as an example, follows the vertex sequence: 81→89→90→91→92→93→94→95→66→51→52→53→54→65→63→62.
Using the line intersection method developed for measuring the length of root in a sample (Newman, 1966; Marsh, 1971), the length of each rhizomorph in G_{1} (the red lines in Fig. 1a) was measured and assigned to graph G_{2} (Fig. 7). Such a graph is then formally a ‘network’, and the number assigned to each edge e is the ‘weight’ of e, in this case length in metres. The shortest path from vertex i to vertex j is then the path with minimum total weight. A frequently used algorithm is the one from Dijkstra (1959). The shortest path from vertex 81 to vertex 62 measures 7.29 m and has edge sequence: 126 (or 127)→125→124→120→117→115→114→112→110→109→105→103→102→78→80→81→101→98→96. This procedure can be applied to each path in a spanning tree. The spanning tree that minimizes the total length of the edges is termed a ‘minimal’ spanning tree. To construct a spanning tree with this property the algorithm of Prim (1957) may be used. Using this algorithm the spanning tree G_{3} (Fig. 6) is obtained as the minimal spanning tree with a length of 24.98 m, which is 58% of the total length of G_{2}.
Discussion
Rhizomorphs of A. lutea at two locations in a natural forest soil occurred in the form of networks. Outgrowth of isolations made for identification of the Armillaria species indicated that most rhizomorphs were viable. Accompanying experiments indicated that the mapped rhizomorphs were either alive or had died within a 16–30week period before the time of observations, and we are confident that intact but dead rhizomorph remnants were not observed extensively in the mapping of the networks.
In Armillaria, explorative growth is accomplished through the formation of rhizomorphs, which are well suited to this because of their insulation from the environment (Rayner et al., 1994). The ability of Armillaria to form rhizomorph networks through anastomoses, as demonstrated in our study, also reflects a strategy of persistence (Reaves et al., 1993). Persistence is achieved through features such as the number of spanning trees conferring robustness to the network structure in relation to disturbance. Rhizomorph networks enable the search for nutrients in time as well as exploration in space and their robustness would be of particular value for Armillaria at the soil leaf–litter layer interface, where rhizomorphs are frequently found and which is often disturbed. The high frequency of cyclic paths may limit the effects of unsuccessful exploration by enabling the redistribution of nutrients within the network and reducing the chance of a disconnected system when a rhizomorph is amputated. Thus, the occurrence of cyclic paths may explain in part the high age of some clones of A. lutea (Smith et al., 1992). A robust network is one that is not substantially weakened when one edge is disconnected from part of the network. Graph G_{2} (Fig. 4) contains nine bridges, of which amputation of any one of two of them (edge 78 or 81) would disconnect a considerable part of the rhizomorph system (based on the number of vertices). Amputation of one of the other seven bridges would disconnect only 1–4% of the original network (G_{2}); if these disconnected components were explorative parts of the network they may not subsequently persist. The impact of such disruption is lower on a densely connected network, for which the number of possible spanning trees is high, as in G_{2}, than in a sparsely connected one.
The substrate of A. lutea, a weak pathogen and saprotroph (Rishbeth, 1982; Thompson & Boddy, 1983; Luisi et al., 1996), probably consists of coarse and weakened woody root material. At site I, both nutrient sources were attached to the largest component of the rhizomorph network, but disruption of the two rhizomorphs attached to these nutrient sources would be sufficient to remove the network from the sources, although there may be attachments to other sources outside the mapped area. The maintenance costs for the network are probably lower than the costs for production of new rhizomorphs. We often observed rhizomorphic growth against woody roots without any sign of infection. This behaviour may be very similar to the behaviour of Armillaria reported in tropical Africa (Leach, 1939; Swift, 1972), where quiescent lesions on woody host roots are common, ‘waiting’ until circumstances are suitable for infection. Although Armillaria is able to form basidiocarps, the success rate of basidiospores in colonizing new substrate, e.g. freshly cut stumps, is extremely low (Rishbeth, 1970; Termorshuizen, 2000), and therefore basidiospores contribute little to persistence.
Dense rhizomorphic networks may also occur at sites where wood is or has been present, resulting in anastomosis by rhizomorph encounters. For many Armillaria spp. there is a good correlation between ectotrophic rhizomorph abundance on root collars and wood and frequency in soil (Marçais & Wargo, 2000; Lygis et al., 2005). Dense networks of rhizomorphs of A. lutea, without differentiation into individual hyphae, have often been observed on aboveground parts of decomposing wood at these sites (A.J. Termorshuizen, pers. commun.) and could occur below ground for this saprotrophic species. In this type of network formation, the length of the edges enclosing the faces would be similar to the dimensions of host material, i.e. coarse woody roots. Formation of anastomoses, and thus cyclic paths, are more likely with growth of rhizomorphs over living or dead roots, where there is a greater chance of meeting other rhizomorphs, rather than by random encounters only. This would be true especially for very small cycles (e.g. at C in Fig. 1a), which may arise from rhizomorphic colonization of a single piece of root. However, these roots were observed only in a few cases due possibly to differences in rates of decay and persistence of roots and rhizomorphs. Additionally, rhizomorphs that are not successful in attaining food substrate and that are not part of cyclic paths would be amputated relatively soon (Rayner, 1991).
The networks at site I and II showed some similarities, including the total length of the cyclic paths, as well as some differences (Table 1). The total rhizomorph system at site II is 1.5 times longer than at site I, but the largest connected component at site II is a twofold lower fraction of the total rhizomorph system. We propose that the rhizomorph system at site II is relatively young, showing many attachments to Picea stumps and tree roots. The largest connected component is characterized by many attachments to nutrient sources. The rhizomorphs attaching the network to the sources are assumed to be amputated when the food source is exhausted. Also, young rhizomorphs with only a few centimetres outgrowth from a nutritional source may have little chance to persist if they fuse with a rhizomorph network, by the time that the source is exhausted. Therefore, an older network, as is assumed to occur at site I, is likely to have fewer attachments to Pinus stumps and will have a larger connected component and more cyclic paths.
Network characteristic  Site I  Site II 
Total rhizomorph length  109 m (4.3 ± 0.4 m m^{−2})  152 m (6.1 ± 0.8 m m^{−2}) 
No. of short (<5 cm) protrusions^{*}  417 (16.7 ± 2.2 m^{−2})  491(19.6 ± 4.0 m^{−2}) 
No. of long (>5 cm) protrusions  127 (5.1 ± 0.5 m^{−2})  268 (10.7 ± 3.5 m^{−2}) 
No. of cases that rhizomorphs crossed each other without anastomosis  90 (3.6 ± 0.7 m^{−2})  176 (7.0 ± 2.2 m^{−2}) 
Largest connected component  84%  48% 
No. of cases that interconnected rhizomorphs left the mapped area  25  21 
Total length of the cyclic paths  43 m  37 m 
No. of cyclic paths  63  50 
No. of connections of the largest connected component to stumps/roots^{†}  2  73 
In this study we have explored different possibilities for graphtheoretic concepts to describe a rhizomorph network and assist in the interpretation of growth strategies of Armillaria. Our approach may be more generally valid for those soilborne mycelial fungi characterized by their ability to form networks and provides a new tool to connect network structure with function (Morris & Robertson, 2005). The introduction of graphtheoretic properties to describe fungal growth may lead to new insights in ecological understanding of Armillaria rhizomorph networks in relatively undisturbed environments. These networks provide a balance between persistence and opportunism, a ‘waitandsee’ strategy, in relation to a changing biotic and abiotic environment.
Acknowledgements
We thank A. PérezSierra (Royal Horticultural Society, Wisley, UK) for identification of the Armillaria isolates, Dr H.J. Broersma (University of Twente, Enschede, The Netherlands) for his helpful suggestions and comments in the development of this paper, and anonymous reviewers for their helpful comments on an earlier version of the manuscript.
Footnotes

Editor: Jim Prosser
 © 2007 Federation of European Microbiological Societies