f u n g a l b i o l o g y r e v i e w s 2 6 ( 2 0 1 2 ) 1 2 e2 9

journal homepage: www.elsevier.com/locate/fbr

Review

Analysis of fungal networks

Luke HEATONa, Boguslaw OBARAb,c, Vincente GRAUb,c,d, Nick JONESa,e,

Toshiyuki NAKAGAKIf, Lynne BODDYg, Mark D. FRICKERh,*
a
Physics Department, Clarendon Laboratory, University of Oxford, Parks Road, Oxford OX1 3PU, UK
b
Oxford e-Research Centre, University of Oxford, 7 Keble Road, Oxford OX1 3QG, UK
c
Oxford Centre for Integrative Systems Biology, University of Oxford, South Parks Road, Oxford OX1 3QU, UK
d
Institute for Biomedical Engineering, University of Oxford, Headington, Oxford OX3 7DQ, UK
e
Department of Mathematics, Imperial College London, SW7 2AZ, UK
f
Department of Intelligent and Complex systems, Hakodate Future University, 116-2 Kamedanakano-cho, Hakodate, Hokkaido, Japan
g
Cardiff School of Biosciences, Cardiff University, Cardiff CF10 3US, UK
h
Department of Plant Sciences, University of Oxford, South Parks Road, Oxford OX1 3RB, UK

article info abstract

Article history: Mycelial fungi grow as indeterminate adaptive networks that have to forage for scarce
Received 4 February 2012 resources in a patchy and unpredictable environment under constant onslaught from
Accepted 7 February 2012 mycophagous animals. Development of contrast-independent network extraction algo-
rithms has dramatically improved our ability to characterise these dynamic macroscopic
Keywords: networks and promises to bridge the gap between experiments in realistic experimental
Fungal network microcosms and graph-theoretic network analysis, greatly facilitating quantitative
Mass-flow modelling description of their complex behaviour. Furthermore, using digitised networks as inputs,
Mycophagous invertebrate grazing empirically-based minimal biophysical mass-flow models already provide a high degree
Network resilience of explanation for patterns of long-distance radiolabel movement, and hint at global
Nutrient translocation control mechanisms emerging naturally as a consequence of the intrinsic hydraulic
Transport network connectivity. Network resilience is also critical to survival and can be explored both in silico
by removing links in the digitised networks according to particular rules, or in vivo by al-
lowing different mycophagous invertebrates to graze on the networks. Survival depends
on both the intrinsic architecture adopted by each species and the ability to reconnect
following damage. It is hoped that a comparative approach may yield useful insights
into not just fungal ecology, but also biologically inspired rules governing the combinatorial
trade-off between cost, transport efficiency, resilience and control complexity for self-
organised adaptive networks in other domains.
ª 2012 The British Mycological Society. Published by Elsevier Ltd. All rights reserved.

1. Network development in mycelial fungi responsive to local environmental conditions. Unlike other
biological transport networks such as plant or animal vascular
Mycelial fungi and plasmodial slime moulds (myxomycetes) systems, the network formed by these organisms is not part of
form elaborate interconnected networks that are highly the organism, it is the organism. These networks develop as

* Corresponding author.
E-mail address: [email protected] (M. D. Fricker).
1749-4613/$ e see front matter ª 2012 The British Mycological Society. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.fbr.2012.02.001
Analysis of fungal networks 13

the organism forages for new resources in a patchy resource microscopic mycelial network (Rayner et al., 1994, 1999; Read
environment and must both transport nutrients between et al., 2009, 2010).
spatially separated source and sink regions, and also maintain In the larger, more persistent saprotrophic and ectomycor-
the network integrity in the face of predation or random rhizal basidiomycetes that are able to grow out into soil from
damage (Selosse et al., 2006; Bebber et al., 2007b; Boddy and colonised food sources, the network architecture develops
Jones, 2007; Fricker et al., 2008a, 2009; Boddy et al., 2009). Fila- further with the formation of specialised high-conductivity
mentous fungi grow by apical extension of slender hyphae organs, termed cords or rhizomorphs (Cairney, 1992, 2005).
that then branch apically or sub-apically (Harris, 2008) to The extent to which mature cords have a clear structure varies
form a fractal, tree-like mycelium (Boddy and Donnelly, between species, but the internal anatomy has similar
2008). Ascomycetes have septa with pores which allow cyto- features across the basidiomycetes (Rayner et al., 1985;
plasmic continuity and organelle movement across the intact Cairney, 1992). The central inner medulla contains numerous
mycelium (Lew, 2011), but which can be quickly blocked by rigid, hollow vessel hyphae with diameters from 10 to 15 mm
Woronin bodies or organelles following damage (Jedd, 2011). and few septal pores that form a pathway for translocation
Likewise, septa in basidiomycetes are perforated by a central (Cairney, 1992, 2005). In contrast, the outer medulla contains
pore so the cytosol can flow between adjacent cells, but the loosely packed hyphae which may be as little as 2 mm wide.
dolipore structure prevents nuclear movement. Extensions Cords are covered by an outer rind that inhibits exchange
of the ER forming the septal pore cap (SPC) can cover the of water and solutes. As some regions of the mycelium expand
pore to restrict leakage following damage (Jedd, 2011). The and mature, other regions regress (Fig. 1), and the process
combination of tip growth and branching allows fungi to of autophagy appears to be critical to enable fungi to
explore complex physical environments using a range of effi- forage effectively through recycling redundant material to
cient, but species-specific space searching algorithms support new growth (Olsson, 2001; Falconer et al., 2005, 2007;
(Hanson et al., 2006; Held et al., 2011). In ascomycetes and Fricker et al., 2007). This is particularly noticeable when new
basidiomycetes, tangential hyphal fusions or anastomoses resources are encountered; cords interconnecting resources
occur behind the colony margin to form an interconnected thicken, whilst mycelium emanating in other directions

Fig. 1 e Development of Phallus impudicus growing from 2 3 2 3 1 cm beech wood block inocula across the surface of
non-sterile soil at 18  C at the times indicated. Initial colony growth forms a densely cross-linked network. Further
expansion is fuelled in part by recycling of redundant material, eventually thinning out to a minimal skeleton as the
resource is consumed. Scale bar [ 2 cm. Images courtesy of AD A’Bear.
14 L. Heaton et al.

from the original resource regresses, to an extent depending step in segmenting network structures is often to use matched
on the species and the relative sizes of the old and new filters based on the spatial properties of the object to locate its
resources (Boddy, 1999; Fricker et al., 2008b). The corded position from the ridges and ravines in the image intensity.
systems resulting from these different processes form visible The shape of the matched filter are often based on the second
networks interconnecting food resources on a scale of centi- derivative of a Gaussian through the calculation of the local
metres in laboratory microcosms to metres in undisturbed Hessian matrix. To detect features at different orientations,
woodland (Fricker et al., 2008a). Indeed, mycelial fungi form the filter is often rotated through a reasonable set of angles
the most extensive biological networks so far characterised, and the maximum response from the filter bank is calculated.
popularly known as the Wood Wide Web (Smith et al., 1992; In general, a tensor representation of an image, such as the
Simard and Durall, 2004; Southworth et al., 2005; Selosse Hessian, gives information about how much the image varies
et al., 2006; Lamour et al., 2007; Whitfield, 2007; Boddy and along and orthogonal to the dominant orientations within
Donnelly, 2008; Beiler et al., 2010; Simard et al., 2012). a certain neighbourhood (Knutsson et al., 2011). Inclusion of
different weightings for the orthogonal directions and inten-
sity components give measures such as vesselness (Frangi
2. Visualisation of network structure and et al., 1998) or neuriteness (Meijering et al., 2004), that have
network extraction been used to improve network-specific feature enhancement.
While tensor representations can be built on purely
Early measures of macroscopic mycelial organisation focussed intensity-based filters, these are sensitive to changes in absolute
on fractal dimension as a useful tool to capture aspects of the intensity and image contrast. To reduce this problem, methods
network structure as a metric (Boddy and Donnelly, 2008). based on local phase have been proposed as a contrast-
However, such a single summary statistic only expresses independent alternative to detect edges (Kovesi, 1999). Salient
a small fraction of the characteristics of the mycelial architec- features have similar values of local phase when observed at
ture. Tools from graph theory allow more detailed character- different scales, thus phase congruency values are high in the
isation of the network structure and dynamics further and direction perpendicular to the cords, while they remain close
allow exploration of the underlying mechanism leading to to zero in the direction parallel to the cords. More importantly,
network optimisation (see Section 3). This approach needs all the values of phase congruency are minimally affected by
branching points and linking cords to be digitised which, contrast changes. Thus the phase congruency tensor (PCT) can
because of the large number of interconnections, requires be used to improve significantly on intensity-based tensors to
(semi-)automated network extraction techniques. give PCT-vesselness or PCT-neuriteness enhancement algo-
Semi-automated image processing approaches that rithms (Obara et al., 2012, in press).
extract networks have been developed for mycelium grown on Following the selective enhancement arising from the PCT
substrates, such as cellophane (Crawford et al., 1993; Hitchcock approach (e.g. Fig. 2), the network can be segmented rapidly
et al., 1996; Ritz et al., 1996), or nitrocellulose (Barry et al., 2009) using a watershed transform. Normally the basins in the
to facilitate acquisition of high-contrast images. The subsequent watershed image represent the objects of interest. However,
processing procedure typically involves noise reduction, using in the case of fungal networks the watershed boundaries
median or band-pass filters, intensity-based thresholding, themselves delineate ridges in the image that correspond to
with manual or automated threshold selection, followed by the network branches of interest, and also guarantee extrac-
morphological processing to achieve an improved skeletonisa- tion of a fully-connected network as they form closed loops.
tion of the colony structure e.g. (Tucker et al., 1992; Barry et al., The resulting skeleton is then pruned on the basis of local
2009; Barry and Williams, 2011). However, these methods cost functions that incorporate both intensity and tensor
require rather specific culture conditions far removed from direction information (Fig. 2). A graph representation of the
those of the natural environment, and illumination regimes network is constructed from the pruned skeleton with link
that cannot readily be adapted for macroscopic networks grown weights based on the Euclidean lengths and the link diameter
in soil microcosms. Using the latter is crucial to explore the derived from sampling the local intensity, with appropriate
natural behavioural capability of these organisms and the full calibration, to give each link a weight that depends on its
range of their ecological responses (Boddy, 1999). In addition, length (l ) and cross-sectional area (a). Each cord is modelled
cords are multi-hyphal aggregates and show considerable vari- as a cylinder packed with identical hyphae, rather than
ation in contrast as the individual cords span several orders of a single tube that increases in diameter, although the internal
magnitude in diameter from mm to mm within a single colony, structure of cords can be much more complex (Rayner et al.,
and have to be segmented from the background of compressed 1985). This approach provides a rapid, robust and extremely
soil which is non-homogeneous in texture and reflectivity. To effective means to extract fungal networks with up to 106 links
date, delineation of the network architecture from the larger from even low-contrast, noisy images compared to <104 links
soil-based systems has only been possibly manually (Bebber possible with manual digitisation (Obara et al., submitted for
et al., 2007a; Fricker et al., 2007; Lamour et al., 2007; Rotheray publication) (Fig. 2).
et al., 2008; Boddy et al., 2010). As a result, the total number of
macroscopic fungal networks analysed so far is relatively low.
Recently, Obara et al. (2012, in press, submitted for 3. Characterisation of the network
publication) developed a high-throughput automated image
analysis approach to detect and characterise large complex Networks extracted using image analysis are strictly
fungal networks grown under realistic conditions. The first a subsample of the full network. The improved image analysis
Analysis of fungal networks 15

Fig. 2 e Automated network extraction from a colony of Phallus impudicus grown on compressed soil from a wood-block
inoculum after 20 d. (a) Extracted network and the source of food overlaid on the input image with the centre line (red),
branching points (green), endpoints (blue), and food source boundary (magenta). (b) Network with w106 links pseudo-colour
coded for cord thickness. Scale bar corresponds to 2 cm.

techniques described above yield more complete sampling of values generally matching the observed qualitative level of
the network architecture down to the level of relatively minor cross-linking. However, this measure, like many other graph-
cords, but still cannot resolve individual fine hyphae. It is theoretic network metrics, can be heavily skewed by parts of
know that topological properties of networks and their the network that are perhaps less relevant when considering
samples can differ (Stumpf et al., 2005). There are geometrical the macroscopic architecture. Thus, the extent and level of
reasons to expect that subsamples of planar graphs will detail from the peripheral growing region or fine mycelium
inherit more of the properties of the full graph, nevertheless that is included in the analysis can affect a. Nevertheless,
potential sampling issues still have to be borne in mind values of a for fungi are similar to those for networks of tunnels
when considering the results described below. in ant galleries (Buhl et al., 2004), Physarum polycephalum (Beb-
The network topology is defined by classifying junctions ber, Nakagaki and Fricker, unpublished observations) and
(branch points, anastomoses and the food sources) as nodes, street networks in cities (Cardillo et al., 2006), suggesting that
and the cords between nodes as links. In general, during addition of up to 20 % of the maximum number of cross-links
foraging the number of nodes, number of links and the total into a planar network maybe sufficient to achieve desirable
material in the network, increase through time. However, the network properties in a range of different scenarios.
local scale network evolution is also characterised by selective Other topological network measures have not proved to be
loss of connections and thinning out of the fine mycelium and very informative for mycelia as they are heavily constrained
weaker cords (Fig. 1). Thus, fungal networks progress from by the developmental process and crowding effects restricting
a radial branching tree to a weakly connected lattice-like the maximum number of connections possible in a planar
network behind the growing margin, through a process of network (Hitchcock et al., 1996; Barrat et al., 2005; Bebber
fusion and reinforcement to form loops, and selective removal et al., 2007a; Fricker et al., 2007; Lamour et al., 2007). Thus,
and recycling of redundant material (Bebber et al., 2007a). the possible degree (k, Table 1) of each node is limited to 1
This shift can be quantified by the alpha coefficient (a) or for tips, 3 for branch points or fusion, or occasionally 4 for
meshedness (Haggett and Chorley, 1969; Buhl et al., 2004), initially overlapping cords that then fuse. Likewise, the
that gives the number of closed loops or cycles present as a frac- mean clustering coefficient, C, is of limited relevance for
tion of the maximum possible for a planar network with the fungal networks, as their growth habit effectively precludes
same number of nodes, according to Euler’s polyhedral formation of triads. The frequency distribution of node
formula (e.g. Barthelemy, 2010). The alpha coefficient strength (Table 1) shows more diversity than node degree
measured for Phanerochaete velutina tends to increase from alone, and follows an approximately log-normal distribution
near 0, as expected for a branching tree, to around 10e20 % as for P. velutina networks (Bebber et al., 2007a). However, we
systems become more cross-connected, depending on the have not found evidence for power law relationships that
resource environment (Bebber et al., 2007a). There is consider- have attracted so much attention in analyses of non-
able variation in the values of a for different species, with biological networks.
16 L. Heaton et al.

Table 1 e Definition of some network terms used

Topology Definition
measure

Node Branch points, fusions or tips

Link Connection between two nodes
Node degree The number of links attached to a node
Clustering A measure of the number of loops of
coefficient length 3 (i.e. a triangle). Values between 0 and 1
Node strength A measure of the importance of a node,
determined by summing the weight
of all links connected to the node
Link weight A measure reflecting the size (e.g. length,
cross-section area) or some other property
relative to other links

4. Predicted transport characteristics of the Fig. 3 e Theoretical optimal network structures that
mycelial network minimise route factor and total length. Networks can be
close to optimal in terms of both route factor and total
As fungal networks are embedded in Euclidean space, it is length. In Gastner and Newman’s model (Gastner and
straightforward to measure the Euclidean distance between Newman, 2006a) nodes are randomly distributed in two
the nodes and compare it with the shortest path through dimensional space with unit mean density, and node 0 is
the network between two nodes to give a route factor designated as the root or source for the network. The
(Gastner and Newman, 2006a). Good distribution networks network grows over time, and at each time step a new link
should be efficient in the sense that the paths from each is added, connecting a previously unconnected node i to
node to the centre ought to be relatively short and the sum some node j that was already part of the network. The
of the lengths of all links in the network should be low, so nodes i and j are chosen such that at each time step the cost
that the network is economical to build and maintain. These dij D blj0 is minimised, where dij is the Euclidean distance
two criteria are often at odds with each other, extra links may between i and j, lj0 is length through the network from node
be needed to reduce the route factor for some given node. j to the source node 0, and b is a tunable parameter. When
Nevertheless, Gastner and Newman (2006b) have shown b [ 0, the shortest possible link is added at every time step.
that there are solutions to the distribution problem that The resulting network will have the smallest possible mean
come remarkably close to being optimal in both senses. Inter- link length l, but it is likely that the mean route factor q will
estingly, these theoretically optimal networks bear striking be large. By increasing b the network no longer grows by
visual similarity to fungal networks in the absence of anasto- adding the shortest possible link, but the resulting
moses (Fig. 3). networks have route factors which are remarkably close to 1
In general, it is important to consider the weights of links (the minimum possible). Each data point on the graph
when considering transport efficiency. In most transport represents an example with 10,000 nodes, and the inset
networks, the flux through a given link reflects the diameter shows an example network with b [ 0.4. From Gastner and
of the pathway as well as just its length used in the calculation Newman (2006a) with permission.
of the route factor, as longer, thinner cords have greater resis-
tance to flow. In this case the diameter of a network can be
kept small regardless of the number of nodes that need to be
connected, or the Euclidean distance between them, as it is differential cord weighting improves the performance of the
always possible to increase the weight of some transport network. At present there are no suitable algorithms available
backbone to provide a ‘short’ route from one side of the to generate weighted planar networks with properties that
network to the other (Barrat et al., 2005; Barthelemy, 2010). mimic fungal networks, although some progress has been
Indeed, much of the literature on optimal transport networks made with models of related systems such as leaf veination
is concerned with the optimal distribution of weights for networks (Katifori et al., 2010; Katifori and Magnasco, 2011).
a given cost function (Murray, 1926; Sherman, 1981; Maritan In other areas of network theory, comparisons are typically
et al., 1996; West et al., 1997; Rinaldo et al., 2004; Durand, made with a reference network produced by random rewiring
2006, 2007; Bohn and Magnasco, 2007; Bernot et al., 2008; of the links or randomly reassigning the weights to different
Corson, 2010; Dodds, 2010; Katifori et al., 2010). links. However, neither approach gives an intuitively satis-
An overall measure of transport is the average network fying model to test performance against, as they have no bio-
efficiency (E ), defined as the mean of the reciprocal of shortest logical basis. We have previously used a two stage procedure
path lengths for transport through the network (Latora and to evaluate the performance of the fungal networks (Bebber
Marchiori, 2001, 2003). In isolation, E is not useful without et al., 2007a). In the first step, the Euclidean fungal network,
some frame of reference. However, it is not straightforward based on link length alone, is compared with model networks
to generate suitable reference models to test the extent that constructed using well defined neighbourhood graphs,
Analysis of fungal networks 17

including the minimum spanning tree (MST) as a lower bound

for a low cost but vulnerable network, and the Delaunay trian- 5. Biological observations of network transport
gulation (DT), giving an upper bound for a well-connected,
robust, but expensive network (Buhl et al., 2004; Cardillo Many species of fungi forage in heterogeneous environments
et al., 2006; Gastner and Newman, 2006b). In the second step where they have to grow through nutrient deficient regions to
of analysis, the effect of including a fixed amount of material discover new resources and therefore require some kind of
in the network, equivalent to the total material in the real transport process to move the resources needed for growth
network, was examined. Thus, each link in the ‘uniform’ from the source of nutrients, to the growing margin. Whilst
fungal and model networks was allocated a constant weight, direct uptake and intra-hyphal nutrient diffusion may be
such that the total construction cost was the same to explore sufficient to sustain short-range local growth when resources
the consequences for transport if the fungus had allocated the are abundant (Olsson, 2001), long-distance translocation is
same amount of resource evenly over the existing or model required to deliver nutrients at a sufficient rate to growing
networks. This also allowed comparison with the real, differ- tips, particularly in fungi that grow out extensively from
entially weighted network as the network measures were in organic resources and are therefore too large to distribute
comparable units. nutrients through diffusion alone (Wells and Boddy, 1995;
The real weighted networks had much shorter physiolog- Wells et al., 1995; Davidson and Olsson, 2000; Boswell et al.,
ical paths, especially in the central region, than their corre- 2002, 2003, 2007). Remarkably little is known about the mech-
sponding uniform networks (Bebber et al., 2007a). More anism(s) underpinning such long-distance nutrient transloca-
surprisingly, the weighted fungal network efficiency was tion, but the mechanisms proposed include vesicles moved by
greater than the uniform DT and the uniform MST when motor proteins (Steinberg, 2007), contractile elements
the predicted transport from just the inoculum (root) to all (Jennings, 1987), diffusion through the vacuole system
other nodes was considered. Although very well connected, (Darrah et al., 2006), carefully regulated osmotic gradients
the DT performed poorly, as distributing material across the (Jennings, 1987; Cairney, 1992) and mass flows, discussed in
large number of links present gave each one low cross- detail below. There is also increasing evidence that a wide
sectional area and consequent high resistance. Conversely, range of macromolecules can be translocated within the
the MST performed better than the DT as it was populated mycelium, including quantum dots (Whiteside et al., 2009)
with few, but extremely thick, links. The uniform fungal and proteins (Woolston et al., 2011). Conversely, water films
networks were similar in performance to the MST, although on the surface of the network itself provide a highway for
they clearly have a different architecture, but the fully movement of bacteria, effectively bridging gaps between soil
weighted fungal network showed the best predicted trans- particles (Kohlmeier et al., 2005; Wick et al., 2010).
port behaviour. Thus, differential weighting of links in the At the microscopic scale GFP tagged proteins, chemical
real network gave a >4 fold improvement in local efficiency stains and quantum dots (Jennings, 1987; Cairney, 1992; Lew,
in comparison to a fully connected uniform network con- 2005; Whiteside et al., 2009) suggest that hyphae contain
structed with the same total cost (Bebber et al., 2007a). The both mass flow of fluid, and active transport mechanisms.
ability of fungal networks to modify link strengths in For example, there is evidence that motor proteins and the
a dynamic way is therefore crucial to achieve a high trans- cytoskeleton play a role in nuclear migration and positioning
port capacity. (Suelmann and Fischer, 2000), but live imaging of growing
Subtle shifts in the predicted transport performance of hyphae also indicates that mass flow of the cytoplasm is the
the network as it grows can be identified by which links dominant factor driving nuclear translocation (Lew, 2005;
are carrying the greatest number of shortest paths and Ramos-Garc
ıa et al., 2009). Additional direct evidence for
have a high shortest-path betweenness centrality (SPBC) mass flows within apical hyphae is through injection of oil
(Latora and Marchiori, 2007). The relative importance of droplets using a pressure probe to observe flow of the cyto-
particular links between the inoculum and added resource, plasm (Lew, 2005). These inert droplets move at the same
as judged by their SPBC, fluctuates in the early stage of rate (w0.2 mm s 1) as the other contents of the hyphae, indi-
growth with several cords competing before one thickens cating that a mass flow of fluid is responsible for much of
up sufficiently to achieve dominance (Fricker et al., 2008b, the observed motion, rather than motor proteins interacting
2009). Equally, one of the disadvantages of using shortest with the cytoskeleton or other specific molecular interactions.
path analysis is that comparable parallel pathways that are At a larger scale, radio-labelled carbon and phosphate have
only marginally longer do not feature in the analysis, but been observed to move over distances and time scales that
might be expected to participate in transport in a real cannot be explained by diffusion alone (Jennings, 1987;
system. This highlights one of the major problems with Cairney, 1992; Wells and Boddy, 1995; Wells et al., 1995;
using shortest path based summary statistics. Shortest Olsson and Gray, 1998; Lindahl et al., 2001; Fricker et al.,
path metrics such as diameter, efficiency, or SPBC, are rela- 2008b). For example, velocities in the range 4.4e9.8 mm s 1
tively easy to compute, and in each case there is a unique have been measured in Armillaria mellea (Jennings, 1987;
solution. However, such measures may be misleading, as Cairney, 1992), and velocities as large as 55 69 mm s 1 have
in actual transport networks material moves along parallel been measured in Serpula lacrymans (Jennings, 1987; Cairney,
flow paths, and not all of the material follows the shortest 1992). The extent of connectivity within the network has
path. It is therefore important to make sure the theoretical been demonstrated for rhizomorphs of Armillaria (Lamour
analyses mirror the actual transport processes occurring et al., 2007), but it is more difficult to establish whether the
within the fungal network itself. network can behave as a single contiguous entity for other
18 L. Heaton et al.

species. In addition, it is clear that the fungi have the potential

for sophisticated control of routing through regulated occlu-
sion of septal pores (Jedd, 2011).

6. Modelling mass flows

We infer that mass flow (advection) is likely to be an impor-

tant, if not the dominant, long-distance transport mechanism
in larger network forming fungi. However, mass flow in fungi
does not follow the paradigm for either plant or animal
vascular systems. Plants drive mass flows in the xylem by
transpiration from the leaves in an open system. They
also actively maintain osmotic gradients along the phloem,
inducing a flow of sap from sources, where water is drawn
from the surrounding tissue into the sieve-tubes of the
phloem, to sinks, where water leaves the phloem. Animals
use hearts or contractile regions to circulate blood through
hierarchical, fractal-like vascular systems that form a closed
system (Sherman, 1981). In fungi, cords tend to be insulated Fig. 4 e Physical principles of growth-induced mass flow.
from the environment by hydrophobic coatings and mass (a) Turgor pressure is induced by an osmotic gradient at the
flow can only take place when water is able to exit the trans- site of water uptake. Vesicles (circles) have to deliver material
location pathway through either localised exudation (e.g. Ser- to the tip and must move towards the tip faster than it grows,
pula lacrymans), evaporation, or by growth itself. As fungal while the cytosol behind the growing tip moves forward at
colonies form integrated hydraulic systems, the increase in the rate of tip growth (Lew, 2005). Conservation of volume
volume that results from hyphal growth requires an equiva- dictates that as the tip expands, fluid flows towards the tips
lent uptake of water, or a reduction in the volume of another from the site of water uptake creating a mass flow.
part of the mycelium that leads to growth-induced mass (b) Suppose that a fungus grows out of an inoculum (square)
flow (e.g. Fig. 4; Heaton et al., 2010). This latter phenomenon and into a region (oval). Some of the material that becomes
is particularly significant in experimental microcosms which part of the fungi may come from within the oval region. The
are run at high humidity thus restricting the potential for rest of the material must have travelled along the cords
any transpirational mass flows. (links) that cross the region’s boundary. If the volume of fungi
With the advent of a fully digitised time-series of weighted within the region increases by DV over a period of time t, and
networks (Bebber et al., 2007b; Fricker et al., 2007, 2008b, 2009; none of the material is drawn from within the region, it
Rotheray et al., 2008; Boddy et al., 2010) it is possible to follows that the average current flowing into the region is
calculate the magnitude of growth-induced mass flow for DV/t. Furthermore, if the total cross-sectional area of the
individual colonies (Heaton et al., 2010). In essence, the boundary crossing cords is a, the mean velocity of flow will
empirically-determined weighted network is recast as an elec- be DV/at (Heaton et al., 2010).
trical circuit analogue, with conductances dependent on the
cross-sectional area of the cords and inversely related to their
length. The total current flowing through the network is esti- data for radiolabel transport, and the pressure gradients
mated from change in the volume of all the cords in the needed to produce these flows are small. Furthermore, cords
network arising from thickening or thinning, or through new that were predicted to carry fast-moving or large currents
growth, between two time points (Fig. 4b). With the simpli- were significantly more likely to increase in size than cords
fying assumptions that the volume change represents water with slow-moving or small currents (Heaton et al., 2010).
movement, the inoculum is the source of water (although We do not yet have methods to measure water movement
the model accommodates uptake anywhere), and the flows directly in these networks. However, we can image nutrient
minimise the work required to overcome viscous drag, the movement by mapping the distribution of the amino-acid
current flowing through every link can be calculated (Fig. 5; analogue, 14C-amino isobutyrate (14C-AIB) using photon-
Heaton et al., 2010). If the total cross-sectional area of the counting scintillation imaging (Tlalka et al., 2002, 2007, 2008).
14
growing front is much larger than the cross-sectional area of C-AIB accumulates in the free amino acid pool and is not
the supporting mycelium, our analysis suggests that because metabolised in a range of woodland fungi so far examined, as
of the (effective) incompressibility of aqueous fluids, growth judged by the lack of incorporation of 14C in other metabolites
itself can drive mass flows which translocate resources or released as 14CO2, (Watkinson, 1984; Olsson and Gray,
towards the growing front at velocities that are much greater 1998). This allows it to be used as a proxy for nitrogen translo-
than the velocity of tip growth. The predicted distribution of cation (Watkinson, 1984) and provides an opportunity to
velocities was heavy tailed, with many links carrying low compare the predictions made by the theoretical network anal-
velocity mass flows, and a few links carrying high velocity ysis to the actual pattern of nutrient movement in the
mass flows. Nevertheless, the velocities predicted for the same microcosms (Fricker et al., 2008c, 2009; Heaton et al.,
major cords were in reasonable agreement with experimental submitted for publication). This requires a significant extension
Analysis of fungal networks 19

Fig. 5 e Network development and predicted currents in Phanerochaete velutina based on a growth-induced mass flow model.
Images (aec) show network development after 19 d, 25 d and 32 d respectively. The image intensity of cords was used to
estimate their thickness, enabling the production of weighted, digitised networks (def). These are colour coded to show the
estimated thicknesses of all sections of all links, while the white block represents the inoculum. Images (gei) are colour
coded according to the total volume that has passed through each cord, as calculated from a growth-induced mass flow
model (Heaton et al., 2010).

14
of the growth-induced mass flow model to incorporate the fate C-AIB in networks of P. velutina (Fig. 6; Heaton et al.,
of a nutrient loaded into the transport pathway which will also submitted for publication). This is a marked improvement
become dispersed by diffusion (Lew, 2011), and potentially over our previous analysis of 14C-AIB distribution patterns
taken out of the pathway during transit to maintain the based on the static network architecture or shortest-path
network itself (Heaton et al., submitted for publication). betweenness centrality, which only showed correlation in
The predictions made by such an Advection-Diffusion- regions of the network where transport actually occurred
Delivery (ADD) model provide an extraordinarily high level (Fricker et al., 2008b, 2009). The ADD model suggests that the
of explanatory power for the measured distribution of minimum currents consistent with the observed growth
20 L. Heaton et al.

Fig. 6 e Measured and predicted patterns of 14C-AIB distribution in networks of Phanerochaete velutina. (a) Mycelial network of
Phanerochaete velutina, photographed just before 14C-AIB was added to the inoculum. (b) Photon-counting scintillation image
of 14C-AIB distribution integrated over 32 h. The brightness of the image reflects the total number of photons emitted from
each region. (c) Measurement of 14C-AIB label from (b) using a superimposed manually digitised network, coloured to
indicate the photon count. Links that were not covered by the scintillation screen are coloured black. (d) Predicted 14C-AIB
intensity from the ADD model (Heaton et al., submitted for publication), with the assumption that AIB enters the network at
the inoculum at a constant rate, each link in the final network continues to grow (or shrink) at the same rate that was
observed over the final time step, and 10 % of each cord is occupied by transport vessels. (e) Predicted intensity under the
same assumptions as (d), except that in this case 20 % of each cord is assumed to be occupied by transport vessels.

would effectively transport resource from the inoculum to the the plausible assumption that P. velutina has evolved to reduce
growing tips over the timescale of growth. Nevertheless, the work needed to overcome viscous drag, as significantly
whilst advective mass flows carry resource over long greater energy savings can be made by preferentially thick-
distances from the inoculum out towards the growing tips ening the high current cords. The speeds predicted by the
(Jennings, 1987; Cairney, 1992; Olsson and Gray, 1998), diffu- ADD model are consistent with experimental data. For
sion and active transport mechanisms may be essential near example, a radio-labelled source of carbon has been measured
the sites where the cell wall is expanding. This follows moving at a velocity 7 mm s 1 away from the inoculum. This is
because the cytosol within the apical hyphae moves forward the same order of magnitude that the ADD model predicts for
at the same rate as the growing tips (Lew, 2005), but to trans- a major cord. The pressure gradients required to produce the
port resource from the base of these hyphae to the growing predicted flows are very modest and unlike previous analyses
tips, the resource has to move faster than the rate of growth. (Jennings, 1987; Rayner, 1991; Lew et al., 2004), it was suggested
that intrahyphal concentration gradients are not strictly
necessary for the production of mass flows.
7. Biophysical consequences of mass flows In other vascular systems with nutrient distribution
involving mass flows, a local adaptive response to wall shear
The incompressibility of the fluids within fungi ensures that stress is a key mechanism that enables the optimisation of
there is a rapid global response to local fluid movements. the network (Kamiya et al., 1984; Pries et al., 2009). Flow veloc-
Furthermore, the velocity of fluid flow is a local signal that ities are of the order 100 1000 mm s 1 in these systems and
can convey quasi-global information about the role of a cord induce wall shear stresses of the order 0.1e1 Nm 2. Flows in
within the mycelium. There was a correlation between the fungi are much lower than this, nevertheless, the wall shear
thickening of cords and the speeds or flux densities predicted stress will be greater in the septal pores, because at that point
by the ADD model (Heaton et al., submitted for publication). the same current that is passing through the hyphae or trans-
Similarly, there was a positive correlation between predicted port vessels must pass through a smaller channel, which
current and the thickening of cords. This is consistent with means the local velocity of flow must be greater. If the mass
Analysis of fungal networks 21

flows also pass through the much smaller vessels that are in a linear network, switching from water uptake at one end
found within cords, or if the shear wall stress is detected at to water uptake throughout the link will halve the scale of
the septal pores, fungi could plausibly detect velocities of mass-flow. In the case of a branching network the reduction
the order 10 mm s 1. It is less likely that fungi can detect the in advection can be much greater. For example, if water
difference between currents with a mean velocity much uptake occurs throughout a branching network of unit
smaller than this, as the corresponding changes in wall shear cross-sectional area the velocity of mass flow would equal
stress would be very small, even in the vessels whose diam- the velocity of tip growth v in every link, regardless of its
eter is only 2 mm. generation (Fig. 7). This contrasts dramatically with the case
As well as experiencing wall shear stress, the vessels where all the water uptake occurs at the inoculum. In the
within pressure driven vascular systems also experience case where there are n generations in the tree and the link
intramural stress. This force per unit area is experienced in question is part of generation i, the mean velocity is v2n i
throughout the vessel wall (and not just on the inside surface), rather than v (Fig. 7).
as the vessel must resist the tendency to expand or burst. The ADD model also helps to explain previously chal-
Hyphae and transport vessels are subject to several atmo- lenging experimental observations, most notably, the occur-
spheres of pressure (about 4e5 bar) (Money, 1990, 1997; Lew rence of sudden route switching (Fricker et al., 2008b), now
et al., 2004; Lew, 2005, 2011; Money, 2008) and the fluids within interpreted as a change in relative distal growth, and the
these structures flow at a very modest rate (Wells et al., 1995; colony-wide coupling of polarised transport and growth
Olsson and Gray, 1998; Tlalka et al., 2002, 2008; Lew, 2005; arrest at sites remote from encounter with a new food
Fricker et al., 2008b). Consequently, the intramural stress will resource (Tlalka et al., 2008). As growth, mass-flow and
be orders of magnitude greater than the shear wall stress. nutrient transport are coupled, there may be an interesting
Nevertheless, this does not mean that it is implausible that interaction between nutrient availability, control of branch-
fungi are sensitive to changes in wall shear stress. The cell ing and nutrient transport. It is well known that the rate of
wall must be rigid enough to withstand the intramural stress hyphal branching increases when tips encounter resource
(Money, 1997, 2008; Harold, 2002; Lew, 2011), but proteins rich environments (Boswell et al., 2003, 2007; Falconer et al.,
embedded in the lipid membranes of tethered organelles, or 2005; Tlalka et al., 2008) which will give an increase in flux
in the septal pores (where the velocity of flow and the wall density in this region. If the rate of input from the inoculum
shear stress will be greatest) may be sensitive to the scale of remains similar, there has to be a concomitant reduction in
flows within a fungal network. fluxes elsewhere, in line with experimental observations
The curvature of vessels can have important effects on the (Tlalka et al., 2008).
fluid flows within them (Truskey et al., 2010), particularly if By contrast, it is difficult to reconcile growth-induced mass
flow rates are high, the vessel is large and the radius of curva- flows, which might be expected to be directed apically, with
ture tight. Whilst this is an important issue for the major observations of bi-directional radiolabel movement (Lindahl
curved arteries of the human vascular system, in the case of et al., 2001; Tlalka et al., 2007, 2008). At the growing margin,
fungal networks we estimate the effect of curvature on fluid one of the potential roles of the tubular vacuolar system
flows is negligible. may be to provide an alternative pathway from the cytoplasm
to allow basally directed diffusive transport of acquired
The importance of localising water uptake solutes (Darrah et al., 2006). In regions more distal from the
tip, bi-directional movement might involve establishing an
The efficiency of growth-induced mass flows as a means of anti-parallel circulation system within individual cords
transport requires that water uptake and growth are spatially (Fricker et al., 2007). However, there is currently no direct
separated. If water uptake occurs throughout the network, evidence for such a system and it is still challenging to
growth-induced mass flows will still occur, but the scale of conceive how such loops would be able to re-configure them-
advection will be significantly reduced (Fig. 7). For example, selves as the network architecture is remodelled.

Fig. 7 e Velocities in a branching network with different patterns of water uptake. (a) If water uptake only occurs in the first
link, the velocity of mass flow halves from one generation to the next. (b) If water uptake occurs evenly throughout the
network, the velocity may be constant throughout.
22 L. Heaton et al.

such as the DT or MST (e.g. Fig. 8; Bebber et al., 2007a; Lamour

8. Network robustness et al., 2007). Having a large number of alternate pathways is
important in this context, and the differential strengthening
High transport capacity and low construction cost could have of links not only imparts high transport capacity but also
come at the expense of other network properties, such as robustness to damage (Bebber et al., 2007a).
robustness to damage, as there is no a priori reason why link A static analysis of the network represents a minimum
weight allocation for one feature necessarily enhances estimate of the real network resilience in nature, as the
another. Robustness to damage from physical breakage or network is also able to respond to local damage, by modifica-
grazing by invertebrates (Harold et al., 2005; Bretherton et al., tion of adjacent link strength, and to regrow and reconnect.
2006; Wood et al., 2006; Boddy and Jones, 2007), is of major Thus, for example, local mechanical damage to a small region
significance to long-lived mycelial systems. of the network promotes strengthening of distal circumferen-
This can be appreciated by examining the effects of tial connections (Fricker et al., 2009), whilst continuous
breaking links in models of the fungal networks and assessing collembolan grazing trims the network back to the reinforced
the impact on transport or overall connectivity (Bebber et al., core (Fig. 9; Rotheray et al., 2008; Boddy et al., 2010), in support
2007a; Lamour et al., 2007; Rotheray et al., 2008; Boddy et al., of the in silico predictions, but also promotes an increase in
2010). We note that cords (links) are the biologically relevant tangential connections making the network more resilient,
target for attack rather than nodes in non-biological systems. at the cost of a reduction in exploration (Fricker et al., 2007,
In natural systems, which links are broken depends on the 2009; Rotheray et al., 2008; Boddy et al., 2010). Since different
agent causing damage. With invertebrate grazing, for example, species have different mycelial architecture, not surprisingly
different species graze in different ways: collembolan often they have different resilience to damage depending on the
target fine mycelium, millipedes graze arcs at growing fronts, extent of connectivity. In general, a high degree of connec-
and woodlice often devour mycelium in long straight paths tivity confers greater resilience, but this comes at an increased
(Crowther et al., 2011a, b, c, in press). These different grazing cost in terms of more material in the network or a reduction in
patterns have not yet been mimicked in artificial experiments exploratory growth (e.g. Fig. 9).
on digitised networks in silico. Rather, links have been broken
at random or by targeting critical connections (Fig. 8; Lamour
et al., 2007), or in an order assuming that the probability of 9. Comparison with transport networks in
breakage increased with length and decreased with the thick- other domains
ness of the link (Bebber et al., 2007a). Robustness was measured
by the size of the connected components remaining or the The challenges that balancing the competing demands of cost,
transport efficiency, and compared to standard networks efficiency, resilience and control complexity place on the

Fig. 8 e Rhizomorph network of Armillaria lutea growing over an area of 25 m2 in a Pinus nigra plantation. (a) A manually
extracted planar graph in which the 107 vertices (nodes) and 169 edges (links) have been numbered. (b) The minimal
spanning tree for the same node positions as (a).
Disruption of two critical links (78 or 81) would lead to large parts (13 % and 11 %) being disconnected from the remainder of
the mapped network. However, there is a low probability that amputation of a randomly chosen link 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. From Lamour et al. (2007) with permission.
Analysis of fungal networks 23

Fig. 9 e Link evolution in colonies of Phanerochaete velutina in response to grazing. Mycelial systems of P. velutina grown from
beech wood blocks in trays (57 3 57 cm) of compressed non-sterile soil. For display, images were processed by background
subtraction, contrast limited histogram equalisation, contrast stretching and look-up table inversion to give black-on-white
representations of the colony morphology. Superimposed pseudo-colour display of the evolution of each link in networks
with new resources added at 36 d (R) for three replicate ungrazed colonies (aec) and three colonies with grazing Collembola
added at 49 d (def). Link evolution was calculated as the ratio between the sum of the differences in link diameter between
successive time-points and the maximum difference over the whole time period. Continuous growth is indicated by red,
continuous regression of cords by blue and cords that remain constant throughout are indicated by green. The position
of Perspex lid supports are indicated by dotted outlines. I [ inoculum, scale bar [ 10 cm. From Boddy et al. (2010) with
permission.

network organisation have strong parallels with those faced the process of Darwinian natural selection based on variation,
in the design of anthropogenic infrastructure networks. We competition and survival has explored a significant range of
note, however, that the fluidic character of fungal networks possible network organisations and the resulting systems
makes them unusual since it allows information to spread are likely to be well-adapted to survive and reproduce under
across the system on very fast timescales. The solutions that particular biotic and abiotic conditions to solve certain ecolog-
fungi achieve may represent good compromises to such ical problems (Fricker et al., 2009). A range of network architec-
a combinatorial optimisation problem, and may yield useful tures, development and dynamics can be found within the
insights into the design of delocalised, robust infrastructure fungi and myxomycetes, suggesting a comparative approach
networks that operate without central control. This presumes may be instructive. However, the constraints imposed by the
that solutions adopted by biological networks will exemplify components used to construct the network (i.e. branching
useful generic theoretical principles, such as persistence, tubes) may have a profound effect on the possible network
robustness, error-handling or appropriate redundancy, as organisation and dynamics, so that any result can only be
they have been honed by evolution. The expectation is that generalised to a very limited set of real-world problems.
24 L. Heaton et al.

Transport costs and optimal transport networks vascular networks represent an optimal or near optimal
compromise between conflicting costs, as the work required
The design of optimal distribution networks for water, elec- to overcome viscous drag is much smaller in large vessels,
tricity, telephone signals, etc is of great practical import in but this benefit of thickening is offset by a cost, and as a simpli-
urban planning, and consequently aspects of this family of fying assumption it is reasonable to assume that the metabolic
problems have been studied since antiquity. Over the last cost of building and maintaining a vessel is proportional to its
15 y, a string of papers has explored the topic of transport volume. The optimal arrangement is when the cube of the
networks that are optimal in some given sense. Some authors radius of the parent vessel equals the sum of the cubes of the
have pursued highly abstract models which assess the cost of radii of the daughters (Murray’s law). Actual vascular systems
different ways of connecting source nodes to sinks (with approximately follow Murray’s law (Sherman, 1981; Sherman
a variety ways of assessing the cost of any given pattern of et al., 1989; McCulloh et al., 2003; Kassab, 2006; Savage et al.,
flow) (Maritan et al., 1996; Banavar et al., 2000a, b, 2002, 2010; 2008), but it is worth noting that significant deviations from
Dreyer, 2001; Dreyer and Puzio, 2001; Bohn and Magnasco, Murray’s law can result in only modest increase in the amount
2007; Bernot et al., 2008; Corson, 2010; Dodds, 2010), of energy required to overcome viscous drag (Sherman et al.,
others have focussed on the general properties of networks 1989; Truskey et al., 2010).
with flows driven by potential differences (Sherman, 1981;
Durand, 2006, 2007; Katifori et al., 2010), while West and Fluctuating demand, robustness and loops
colleagues consider the optimum network given a number of
biologically inspired assumptions (West et al., 1997, 1999a, Although vascular networks are frequently depicted as
2002; West and Brown, 2004; Savage et al., 2008). The papers branching trees (Bernot et al., 2008; Pries et al., 2009), many
by Banavar and West are particularly notable, as they played natural and almost all man made networks contain loops.
a key role in the development of the theory of allometric Street plans are full of loops, most fungal networks contain
scaling, which attempts to explain the relationship between loops (Bebber et al., 2007a; Fricker et al., 2007, 2008a, 2009;
body size and metabolic rate. Boddy et al., 2009, 2010), as do retinal vasculatures (Fruttiger,
Different studies use different definitions of a network, and 2002), and the veins of many leaves contain recursively nested
the authors optimise different cost functions. For example, sets of loops (Roth-Nebelsick et al., 2001; Durand, 2006; Sack
Durand (Durand, 2006, 2007) considers the optimal geometry and Holbrook, 2006; Corson et al., 2009; Corson, 2010; Katifori
and the relationship between the local geometry and the local et al., 2010). Recent theoretical analyses indicate that if
topology of hydraulic networks whose currents derive from a network is attacked, or subject to fluctuating loads, the
a potential, explicitly analogous to electrical networks, that optimal form is no longer a branching tree but will contain
are embedded in an ambient space. In contrast, Banavar loops (Corson et al., 2009; Corson, 2010; Katifori et al., 2010;
et al. (2000a, b, 2002, 2010) propose a more abstract model Katifori and Magnasco, 2011).
where the graph is not assumed to be embedded in a target Optimisation under damage to links implies the formation
space, and the currents through the nodes are not explicitly of loops almost by definition, as if there were only one route
constrained to derive from a potential difference. connecting the source and a given sink, an infinite amount
A very general observation at the heart of the analysis by of power would be dissipated when that route is cut. Robust-
Banavar and colleagues is that any efficient flow pattern ness to damage can be conferred by a (topologically
must be such that at every point, the flow moves materials minimum) ring joining the outermost nodes (Roth-Nebelsick
away from the source. In other words, although there may et al., 2001), and more generally it is well known that redun-
be loops in the networks they consider, optimally efficient dant links can confer a degree of robustness to a transport
flows always transport materials in a directed manner. This network (Roth-Nebelsick et al., 2001; Rinaldo et al., 2004; Sack
feature of ‘efficient’ material flows is common to many defini- and Holbrook, 2006; Fricker et al., 2007, 2009; Barthelemy,
tions of ‘efficient’ (Sherman, 1981; Maritan et al., 1996; West 2010). Plants and fungi are under constant attack, from the
et al., 1997, 1999a, 2002; Banavar et al., 1999, 2000a, b, 2002, elements as well as pathogens and a wide variety of grazing
2010; Dreyer, 2001; Dreyer and Puzio, 2001; Colizza et al., animals (Roth-Nebelsick et al., 2001; Sack and Holbrook,
2004; Rinaldo et al., 2004; West and Brown, 2004; Durand, 2006; Rotheray et al., 2008; Boddy et al., 2010; Crowther et al.,
2006, 2007; Bernot et al., 2008; Savage et al., 2008; Corson, 2011a, b, c, in press). If fungal networks were branching trees,
2010; Dodds, 2010; Katifori et al., 2010), and it is inevitable severing any branch would disconnect the network. Likewise,
when the flows are driven by differences in potential. if the leaf vascular network was treelike, damage to any vein
Banavar and colleagues abstract approach enables the would result in the death of all the leaf sections downstream
construction of formal proofs concerning optimal networks, from that vein. The value of redundant links is therefore quite
but the proper physical interpretation of those results is some- clear, but optimality under robustness to damage can produce
what elusive (Bohn and Magnasco, 2007). A different approach, hierarchical, recursively nested loops (Katifori et al., 2010;
taken by West and colleagues builds on the legacy of Cecil D. Katifori and Magnasco, 2011), as can be found in actual leaf
Murray, and the principle of minimum work (Murray, 1926; venation networks (Roth-Nebelsick et al., 2001; Sack and
Sherman, 1981; West et al., 1997, 1999b, 2002; West and Holbrook, 2006).
Brown, 2004; Savage et al., 2008). Fluid flow is essential to Katifori et al. (2010, 2011) and Corson et al. (2009, 2010) also
most biological transport networks, and so cost functions show the importance of fluctuating demand for optimal trans-
that reflect the energy required to overcome viscous drag are port networks. Many definitions of optimality yield treelike
of particular biological significance. Murray observed that structures, with a single path connecting any two points
Analysis of fungal networks 25

(Sherman, 1981; Banavar et al., 1999, 2000a, b; Durand, 2006, optimisation” (Hanson et al., 2006; Xu et al., 2009) algorithms
2007; Bohn and Magnasco, 2007; Bernot et al., 2008). In this similar to those that have evolved from the study of ant colony
light, the many loops found in leaf venation networks, fungal foraging patterns (Dorigo et al., 1999) or based on Physarum
networks or vascular networks could be interpreted as (Tero et al., 2006, 2007, 2008; Nakagaki et al., 2007; Watanabe
a compromise between transport efficiency and other require- et al., 2011).
ments, such as tolerance to damage. However, Katifori et al. Even at this stage, some common features of biological
and Corson et al. show that the optimality of branching trees network formation seem to emerge. Fungal networks are con-
is contingent on the assumption of a stationary flow through structed by local iterative developmental processes rather
the network. By contrast, they found that optimisation under than predetermined blueprints or centralised control, with
a varying load leads can lead to the formation of dense, recur- growth involving over-production of links and nodes, fol-
sively looped structures (Corson et al., 2009, 2010; Katifori et al., lowed by selective pruning of some links and reinforcement
2010; Katifori and Magnasco, 2011). More specifically, when of others. Such a process mimics the process of Darwinian
time variations or fluctuations are allowed for, the resulting evolution in which natural selection removes less fit offspring.
optimal structures contain loops, although they do share the This ‘Darwinian network model’ may be applicable to other
hierarchical organisation that is characteristic of treelike biological systems, including foraging ant trails, Physarum,
optimal networks. axon development and angiogenesis, and may represent a
This result is biologically very significant, as in general generalised model for growth of physical biological networks.
biological transport networks operate in the face of consider- Based on the ant colony and Physarum models, we might
able spatiotemporal irregularities. As fungal networks grow expect the generic ingredients in such a model will involve
and the supporting mycelium matures, different parts of a non-linear positive reinforcement term related to the local
the growing margin will need to be supplied with resources flux and a linear decay term. Notably this model differs from
at different times. If fungal networks have indeed evolved other models of weighted network evolution that only incor-
to maximise transport efficiency, they should be optimal in porate differential strengthening of links, i.e., ‘the busiest
relation to a history of transport demands, and we should get busier’, rather than additional differential weakening
not necessarily expect that at any instant the network is and loss that is the hallmark of evolution by natural selection.
the best way of linking the sites of resource uptake with However, the model has parallels with the selective link
the sites of resource demand. Nevertheless, we note that removal model proposed for unweighted networks (Salathe
fungal transport networks have unusual features compared et al., 2005). In infrastructure networks where costs are associ-
to the preceding models of networks for the delivery of nutri- ated with creation and maintenance of links, where links
ents. The fungal network is both a transport system and the differ in some measure of fitness, and where material can be
organism itself. Fungal networks are vastly more dynamic recycled, such a Darwinian model may be applicable. In prac-
than vascular networks in plants and humans and it seems tical terms such a process may also be witnessed in the evolu-
that fungi use their networked state to process their environ- tion of real infrastructure networks, such as British railways
ment, possibly using their fluid based nature to rapidly following the Beeching reviews in the early 60’s (British
couple behaviour even in disparate parts of themselves. Transport Commission, 1963; British Railways Board, 1965).
This type of analysis has strong parallels with the conceptual In these reviews, the flux along various routes was measured
ideas set out by Rayner two decades ago (Rayner, 1991; and routes with too low a level of traffic, mainly branch lines,
Rayner et al., 1994, 1999). were targeted for closure. At the same time, major routes were
strengthened to cope with the expected sourceesink relation-
ships for both passenger and freight traffic. Interestingly, the
10. Future developments reports focussed on efficiency rather than any explicit consid-
eration of resilience, which may explain the sensitivity of the
Characterisation of mycelial networks is still in its infancy. current network to disruption.
However, the network approach provides a way of quantifying A second feature of interest emerging, particularly through
and analysing complex fungal systems in detail, and also consideration of the Physarum and fungal networks, is the
makes it possible to link measurements in laboratory micro- extent that coupled flows may contain global information.
cosms to observations of networks in the field. The simplest Networks involving physical flows obey continuity equations
models predicting transport through such networks are based and are therefore intrinsically coupled across the network.
on shortest path considerations through the weighted This automatically means that increasing the flow in one
network and are relatively straightforward to calculate, but part of the network will lead to reductions elsewhere, even
only capture part of the experimentally determined transport though the local conditions in the distal region remain the
behaviour. Models that include parallel-flow pathways same. Thus each part of the network is influenced by and
coupled to the empirically-measured growth of the network can influence the whole network, but without any global
improve the match between simulation and experiment assessment of behaviour. Useful properties of the network
significantly and explain several previously challenging may emerge from the interaction between the local update
empirical observations. The next conceptual advance will be rules governing topology and flows without the need for
to identify the rules that allow the network iteratively to refine long-distance communication or calculation of aggregate
its structure and transport behaviour to yield the network properties of the network. It is this coupling in the Physarum
architectures observed. It is conceivable that the identification model that allows the network to resolve from a fine mesh
of such rules will allow development of generic “fungal colony into a quasi-optimal solution (Tero et al., 2006, 2007, 2008;
26 L. Heaton et al.

Nakagaki et al., 2007). Furthermore, the computational over- Bebber, D.P., Hynes, J., Darrah, P.R., Boddy, L., Fricker, M.D., 2007a.
head for such self-organised networks scales well with the Biological solutions to transport network design. Proc. R. Soc.
number of additional nodes. B 274, 2307e2315.
Bebber, D.P., Tlalka, M., Hynes, J., Darrah, P.R., Ashford, A.,
Set against this progress are some equally difficult chal-
Watkinson, S.C., Boddy, L., Fricker, M.D., 2007b. Imaging
lenges. The transition to three dimensions, particularly in complex nutrient dynamics in mycelial networks. In: Gadd, G.,
realistic soil or wood microcosms, is immensely problematic Watkinson, S.C., Dyer, P. (Eds.), Fungi in the Environment.
for the vast majority of imaging approaches. There is potential Cambridge University Press, pp. 3e21.
to record the micro-structure of a porous soil system using Beiler, K.J., Durall, D.M., Simard, S.W., Maxwell, S.A.,
high-resolution X-ray tomography and then predict the Kretzer, A.M., 2010. Architecture of the wood-wide web: Rhi-
zopogon spp. genets link multiple Douglas-fir cohorts. New
behaviour of different fungi using modelling approaches
Phytol. 185, 543e553.
(Blair et al., 2007; Pajor et al., 2010), but there is currently not
Bernot, M., Caselles, V., Morel, J.M., 2008. Optimal Transportation
sufficient contrast to resolve the actual fungal distribution Networks: Models and Theory. Springer-Verlag, Berlin.
within the soil. Likewise, the observation that fungi can take Blair, J.M., Falconer, R.E., Milne, A.C., Young, I.M., Crawford, J.W.,
up and translocate quantum dots (Whiteside et al., 2009), 2007. Modeling three-dimensional microstructure in hetero-
and the increasing number of species that can now express geneous media. Soil Sci. Soc. Am. J. 71, 1807e1812.
fluorescent proteins (Lorang et al., 2001; Czymmek et al., Boddy, L., 1999. Saprotrophic cord-forming fungi: meeting the
challenge of heterogeneous environments. Mycologia 91, 13e32.
2004; Leroch et al., 2011) coupled with confocal or multi-
Boddy, L., Donnelly, D.P., 2008. Fractal geometry and microor-
photon imaging facilitate 3-D or 4-D data collection from
ganisms in the environment. In: Senesi, N., Wilkinson, K.J.
living systems (Czymmek, 2005), but only in relatively translu- (Eds.), Biophysical Chemistry of Fractal Structures and
cent media to a depth of around 100 mm. Processes in Environmental Systems. John Wiley, Chichester,
UK, pp. 239e272.
Boddy, L., Hynes, J., Bebber, D.P., Fricker, M.D., 2009. Saprotrophic
cord systems: dispersal mechanisms in space and time.
Acknowledgements Mycoscience 50, 9e19.
Boddy, L., Jones, T.H., 2007. Mycelial responses in heterogeneous
environments: parallels with macroorganisms. In: Gadd, G.,
Research in the authors’ laboratories has been supported by
Watkinson, S.C., Dyer, P. (Eds.), Fungi in the Environment.
BBSRC (43/P19284), NERC (GR3/12946 & NER/A/S/2002/882),
Cambridge University Press, pp. 112e158.
EPSRC (GR/S63090/01), EU Framework 6 (STREP No. 12999), Boddy, L., Wood, J., Redman, E., Hynes, J., Fricker, M.D., 2010.
Oxford University Research Infrastructure Fund and the Fungal network responses to grazing. Fungal Genet. Biol. 47,
University Dunston Bequest. We thank A. Ashford, K. Burton, 522e530.
P.R. Darrah, D.P. Donnelly, D. Eastwood, J. Efstathiou, J. Hynes, Bohn, S., Magnasco, M.O., 2007. Structure, scaling, and phase
N. Johnson, E. Lopez, P. Maini, F. Reed-Tsochas, M. Tlalka, G.M. transition in the optimal transport network. Phys. Rev. Lett.
98, 088702.
Tordoff, S.C. Watkinson and members of CABDyN for stimu-
Boswell, G.P., Jacobs, H., Davidson, F.A., Gadd, G.M., Ritz, K., 2002.
lating discussions. Functional consequences of nutrient translocation in mycelial
fungi. J. Theor. Biol. 217, 459e477.
Boswell, G.P., Jacobs, H., Davidson, F.A., Gadd, G.M., Ritz, K., 2003.
references Growth and function of fungal mycelia in heterogeneous
environments. Bull. Math. Biol. 65, 447e477.
Boswell, G.P., Jacobs, H., Ritz, K., Gadd, G.M., Davidson, F.A., 2007.
Banavar, J.R., Colaiori, F., Flammini, A., Maritan, A., Rinaldo, A., The development of fungal networks in complex environ-
2000a. Topology of the fittest transportation network. Phys. ments. Bull. Math. Biol. 69, 605e634.
Rev. Lett. 84, 4745e4748. Bretherton, S., Tordoff, G.M., Jones, T.H., Boddy, L., 2006.
Banavar, J.R., Damuth, J., Maritan, A., Rinaldo, A., 2002. Supply- Compensatory growth of Phanerochaete velutina mycelial
demand balance and metabolic scaling. PNAS 99, systems grazed by Folsomia candida (Collembola). FEMS
10506e10509. Microbiol. Ecol. 58, 33e40.
Banavar, J.R., Maritan, A., Rinaldo, A., 1999. Size and form in British Railways Board, 1965. The Development of The Major
efficient transportation networks. Nature 399, 130e132. Railway Trunk Routes. British Railways Board.
Banavar, J.R., Maritan, A., Rinaldo, A., 2000b. Scaling e rivers, British Transport Commission, 1963. The Reshaping of British
blood and transportation networks e reply. Nature 408, 160. Railways e Part 1: Report. Her Majesty’s Stationery Office.
Banavar, J.R., Moses, M.E., Brown, J.H., Damuth, J., Rinaldo, A., Buhl, J., Gautrais, J., Sole, R.V., Kuntz, P., Valverde, S.,
Sibly, R.M., Maritan, A., 2010. A general basis for quarter- Deneubourg, J.L., Theraulaz, G., 2004. Efficiency and robust-
power scaling in animals. PNAS 107, 15816e15820. ness in ant networks of galleries. Eur. Phys. J. B 42, 123e129.
Barrat, A., Barthelemy, M., Vespignani, A., 2005. The effects of Cairney, J.W.G., 1992. Translocation of solutes in ectomycorrhizal
spatial constraints on the evolution of weighted complex and saprotrophic rhizomorphs. Mycol. Res. 96, 135e141.
networks. J. Stat. Mech. P05003. Cairney, J.W.G., 2005. Basidiomycete mycelia in forest soils:
Barry, D., Chan, C., Williams, G., 2009. Morphological quantifica- dimensions, dynamics and roles in nutrient distribution. My-
tion of filamentous fungal development using membrane col. Res. 109, 7e20.
immobilization and automatic image analysis. J. Ind. Micro- Cardillo, A., Scellato, S., Latora, V., Porta, S., 2006. Structural
biol. Biotechnol. 36, 787e800. properties of planar graphs of urban street patterns. Phys. Rev.
Barry, D.J., Williams, G.A., 2011. Microscopic characterisation of E 73, 066107.
filamentous microbes: towards fully automated morphological Colizza, V., Banavar, J.R., Maritan, A., Rinaldo, A., 2004. Network
quantification through image analysis. J. Microsc. 244, 1e20. structures from selection principles. Phys. Rev. Lett. 92,
Barthelemy, M., 2010. Spatial networks. Phys. Rep. 499, 1e101. 198701.
Analysis of fungal networks 27

Corson, F., 2010. Fluctuations and redundancy in optimal trans- complex nutrient dynamics in mycelial networks. J. Microsc.
port networks. Phys. Rev. Lett. 104, 048703. 231, 317e331.
Corson, F., Adda-Bedia, M., Boudaoud, A., 2009. In silico leaf Fricker, M.D., Lee, J.A., Boddy, L., Bebber, D.P., 2008c. The interplay
venation networks: growth and reorganization driven by between structure and function in fungal networks.
mechanical forces. J. Theor. Biol. 259, 440e448. Topologica 1.
Crawford, J.W., Ritz, K., Young, I.M., 1993. Quantification of fungal Fruttiger, M.A., 2002. Development of the mouse retinal vascula-
morphology, gaseous transport and microbial dynamics in ture: angiogenesis versus vasculogenesis. Invest. Ophthalmol.
soil e an integrated framework utilizing fractal geometry. Vis. Sci. 43, 522e527.
Geoderma 56, 157e172. Gastner, M.T., Newman, M.E.J., 2006a. Optimal design of spatial
Crowther, T.W., Boddy, L., Jones, T.H., 2011a. Outcomes of fungal distribution networks. Phys. Rev. E 74, 016117.
interactions are determined by soil invertebrate grazers. Ecol. Gastner, M.T., Newman, M.E.J., 2006b. Shape and efficiency in
Lett. 14, 1134e1142. spatial distribution networks. J. Stat. Mech., P01015.
Crowther, T.W., Boddy, L., Jones, T.H., 2011b. Species-specific Haggett, P., Chorley, R.J., 1969. Network Analysis in Geography.
effects of soil fauna on fungal foraging and decomposition. Arnold, London.
Oecologia 167, 535e545. Hanson, K.L., Nicolau, D.V., Filipponi, L., Wang, L., Lee, A.P., 2006.
Crowther, T.W., Jones, T.H., Boddy, L., 2011c. Species-specific Fungi use efficient algorithms for the exploration of micro-
effects of grazing invertebrates on mycelial emergence and fluidic networks. Small 2, 1212e1220.
growth from woody resources into soil. Fungal Ecol. 4, 333e341. Harold, F.M., 2002. Force and compliance: rethinking morpho-
Crowther T.W., Jones T.H., Boddy L. Interactions between sapro- genesis in walled cells. Fungal Genet. Biol. 37, 271e282.
trophic basidiomycete mycelia and mycophagous soil fauna. Harold, S., Tordoff, G.M., Jones, T.H., Boddy, L., 2005. Mycelial
Mycology, in press, doi:10.1080/21501203.2012.656723. responses of Hypholoma fasciculare to collembola grazing:
Czymmek, K.J., 2005. Exploring fungal activity with confocal and effect of inoculum age, nutrient status and resource quality.
multiphoton microscopy. In: Dighton, J., White, J.F., Mycol. Res. 109, 927e935.
Oudemans, P. (Eds.), The Fungal Community: Its Organisation Harris, S.D., 2008. Branching of fungal hyphae: regulation,
and Role in the Ecosystem, pp. 307e330. mechanisms and comparison with other branching systems.
Czymmek, K.J., Bourett, T.M., Howard, R.J., 2004. Fluorescent Mycologia 100, 823e832.
protein probes in fungi. Methods Microbiol. 34, 27e62. Heaton, L.L.M., Lo
pez, E., Maini, P.K., Fricker, M.D., Jones, N.S.,
Darrah, P.R., Tlalka, M., Ashford, A., Watkinson, S.C., 2010. Growth-induced mass flows in fungal networks. Proc. R.
Fricker, M.D., 2006. The vacuole system is a significant intra- Soc. B 277, 3265e3274.
cellular pathway for longitudinal solute transport in basidio- Heaton L.L.M., Lo
pez E., Maini P.K., Fricker M.D., Jones N.S.
mycete fungi. Eukaryot. Cell 5, 1111e1125. Advection, diffusion and delivery over a network. Phys. Rev. E,
Davidson, F.A., Olsson, S., 2000. Translocation induced outgrowth submitted for publication.
of fungi in nutrient-free environments. J. Theor. Biol. 205, Held, M., Edwards, C., Nicolau, D.V., 2011. Probing the growth
73e84. dynamics of Neurospora crassa with microfluidic structures.
Dodds, P.S., 2010. Optimal form of branching supply and collec- Fungal Biol. 115, 493e505.
tion networks. Phys. Rev. Lett. 104, 048702. Hitchcock, D., Glasbey, C.A., Ritz, K., 1996. Image analysis of
Dorigo, M., Di Caro, G., Gambardella, L.M., 1999. Ant algorithms space-filling by networks: application to a fungal mycelium.
for discrete optimization. Artif. Life 5, 137e172. Biotechnol. Tech. 10, 205e210.
Dreyer, O., 2001. Allometric scaling and central source systems. Jedd, G., 2011. Fungal evoedevo: organelles and multicellular
Phys. Rev. Lett. 8703, 038101. complexity. Trends Cell Biol. 21, 12e19.
Dreyer, O., Puzio, R., 2001. Allometric scaling in animals and Jennings, D.H., 1987. Translocation of solutes in fungi. Biol. Rev.
plants. J. Math. Biol. 43, 144e156. 62, 215e243.
Durand, M., 2006. Architecture of optimal transport networks. Kamiya, A., Bukhari, R., Togawa, T., 1984. Adaptive regulation of
Phys. Rev. E 73, 016116. wall shear stress optimizing vascular tree function. Bull. Math.
Durand, M., 2007. Structure of optimal transport networks subject Biol. 46, H14eH21.
to a global constraint. Phys. Rev. Lett. 98, 088701. Kassab, G.S., 2006. Scaling laws of vascular trees: of form and
Falconer, R.E., Bown, J.L., White, N.A., Crawford, J.W., 2005. function. Am. J. Physiol. Heart Circ. Physiol. 290, H894eH903.
Biomass recycling and the origin of phenotype in fungal Katifori, E., Magnasco, M.O., 2011. Quantifying Loopy Network
mycelia. Proc. R. Soc. B 272, 1727e1734. Architecture. arXiv:1110.1412v1.
Falconer, R.E., Bown, J.L., White, N.A., Crawford, J.W., 2007. Katifori, E., Szollosi, G.J., Magnasco, M.O., 2010. Damage and
Biomass recycling: a key to efficient foraging by fungal fluctuations induce loops in optimal transport networks. Phys.
colonies. Oikos 116, 1558e1568. Rev. Lett. 104, 048704.
Frangi, A., Niessen, W., Vincken, K., Viergever, M., 1998. Multiscale Knutsson, H., Westin, C.-F., Andersson, M., 2011. Representing
vessel enhancement filtering. In: Wells, W., Colchester, A., local structure using tensors II. In: Heyden, A., Kahl, F. (Eds.),
Delp, S. (Eds.), Medical Image Computing and Computer- Image Analysis. Lecture Notes in Computer Science, vol. 6688
assisted Intervention. Springer, Berlin/Heidelberg, pp. 130e137. Springer, Berlin, Heidelberg, pp. 545e556.
Fricker, M., Boddy, L., Nakagaki, T., Bebber, D.P., 2009. Adaptive Kohlmeier, S., Smits, T.H.M., Ford, R.M., Keel, C., Harms, H.,
biological networks. In: Gross, T., Sayama, H. (Eds.), Adaptive Wick, L.Y., 2005. Taking the fungal highway: Mobilization of
Networks: Theory, Models and Applications. Springer, pp. 51e70. pollutant-degrading bacteria by fungi. Environ. Sci. Technol.
Fricker, M.D., Bebber, D.P., Boddy, L., 2008a. Mycelial networks: 39, 4640e4646.
structure and dynamics. In: Boddy, L., Franklin, J.C., van Kovesi, P.D., 1999. Image features from phase congruency. Videre
West, P. (Eds.), Ecology of Saprotrophic Basidiomycetes. 1, 1e26.
Academic Press, Amsterdam, pp. 3e18. Lamour, A., Termorshuizen, A.J., Volker, D., Jeger, M.J., 2007.
Fricker, M.D., Boddy, L., Bebber, D.P., 2007. Network organisation Network formation by rhizomorphs of Armillaria lutea in
of mycelial fungi. In: Howard, R.J., Gow, N.A.R. (Eds.), The natural soil: their description and ecological significance.
Mycota. Springer-Verlag, Berlin, pp. 309e330. FEMS Microbiol. Ecol. 62, 222e232.
Fricker, M.D., Lee, J.A., Bebber, D.P., Tlalka, M., Hynes, J., Latora, V., Marchiori, M., 2001. Efficient behavior of small-world
Darrah, P.R., Watkinson, S.C., Boddy, L., 2008b. Imaging networks. Phys. Rev. Lett. 87, 198701.
28 L. Heaton et al.

Latora, V., Marchiori, M., 2003. Economic small-world behavior in Papin, J.A., 2009. Structural adaptation and heterogeneity of
weighted networks. Eur. Phys. J. B 32, 249e263. normal and tumor microvascular networks. PLoS Comput.
Latora, V., Marchiori, M., 2007. A measure of centrality based on Biol. 5, e1000394.
network efficiency. New J. Phys. 9, 188. Ramos-Garc
ıa, S.L., Roberson, R.W., Freitag, M., Bartnicki-
Leroch, M., Mernke, D., Koppenhoefer, D., Schneider, P., Garc
ıa, S., Mourin ~ o-Pe

rez, R.R., 2009. Cytoplasmic bulk flow
Mosbach, A., Doehlemann, G., Hahn, M., 2011. Living colors in propels nuclei in mature hyphae of Neurospora crassa. Eukar-
the Gray mold pathogen Botrytis cinerea: codon-optimized yot. Cell 8, 1880e1890.
genes encoding green fluorescent protein and mCherry, which Rayner, A.D.M., 1991. The challenge of the individualistic
exhibit bright fluorescence. Appl. Environ. Microbiol. 77, mycelium. Mycologia 83, 48e71.
2887e2897. Rayner, A.D.M., Griffith, G.S., Ainsworth, A.M., 1994. Mycelial
Lew, R., Levina, N., Walker, S., Garrill, A., 2004. Turgor regulation interconnectedness. In: Gow, N.A.R., Gadd, G.M. (Eds.), The
in hyphal organisms. Fungal Genet. Biol. 41, 1007e1015. Growing Fungus. Chapman and Hall, London, pp. 21e40.
Lew, R.R., 2005. Mass flow and pressure-driven hyphal extension Rayner, A.D.M., Powell, K.A., Thompson, W., Jennings, D.H., 1985.
in Neurospora crassa. Microbiology 151, 2685e2692. Morphogenesis of Vegetative Organs, Developmental Biology
Lew, R.R., 2011. How does a hypha grow? The biophysics of of Higher Fungi. Cambridge University Press, Cambridge.
pressurized growth in fungi. Nature Rev. Microbiol. 9, 509e518. 249e279.
Lindahl, B., Finlay, R., Olsson, S., 2001. Simultaneous, bidirec- Rayner, A.D.M., Watkins, Z.R., Beeching, J.R., 1999. Self-integra-
tional translocation of 32P and 33P between wood blocks tion e an emerging concept from the fungal mycelium. In:
connected by mycelial cords of Hypholoma fasciculare. New Gow, N.A.R., Robson, G.D., Gadd, G.M. (Eds.), The Fungal
Phytol. 150, 189e194. Colony. Cambridge University Press, Cambridge, pp. 1e24.
Lorang, J.M., Tuori, R.P., Martinez, J.P., Sawyer, T.L., Redman, R.S., Read, N.D., Fleißner, A., Roca, G.M., Glass, N.L., 2010. Hyphal
Rollins, J.A., Wolpert, T.J., Johnson, K.B., Rodriguez, R.J., fusion. In: Borkovich, K.A., Ebbole, D. (Eds.), Cellular and
Dickman, M.B., Ciuffetti, L.M., 2001. Green fluorescent protein Molecular Biology of Filamentous Fungi. American Society of
is lighting up fungal biology. Appl. Environ. Microbiol. 67, Microbiology, pp. 260e273.
1987e1994. Read, N.D., Lichius, A., Shoji, J.Y., Goryachev, A.B., 2009. Self-sig-
Maritan, A., Colaiori, F., Flammini, A., Cieplak, M., Banavar, J.R., nalling and self-fusion in filamentous fungi. Curr. Opin. Mi-
1996. Universality classes of optimal channel networks. crobiol. 12, 608e615.
Science 272, 984e986. Rinaldo, A., Banavar, J.R., Colizza, V., Maritan, A., 2004. On
McCulloh, K.A., Sperry, J.S., Adler, F.R., 2003. Water transport in network form and function. Physica A 340, 749e755.
plants obeys Murray’s law. Nature 421, 939e942. Ritz, K., Millar, S.M., Crawford, J.W., 1996. Detailed visualisation of
Meijering, E., Jacob, M., Sarria, J.C.F., Steiner, P., Hirling, H., hyphal distribution in fungal mycelia growing in heteroge-
Unser, M., 2004. Design and validation of a tool for neurite neous nutritional environments. J. Microbiol. Methods 25,
tracing and analysis in fluorescence microscopy images. Cy- 23e28.
tometry 58A, 167e176. Roth-Nebelsick, A., Uhl, D., Mosbrugger, V., Kerp, H., 2001.
Money, N.P., 1990. Measurement of hyphal turgor. Exp. Mycol. 14, Evolution and function of leaf venation architecture: a review.
416e425. Ann. Bot. 87, 553e566.
Money, N.P., 1997. Wishful thinking of turgor revisited: the Rotheray, T.D., Jones, T.H., Fricker, M.D., Boddy, L., 2008. Grazing
mechanics of fungal growth. Fungal Genet. Biol. 21, 173e187. alters network architecture during interspecific mycelial
Money, N.P., 2008. Insights on the mechanics of hyphal growth. interactions. Fungal Ecol. 1, 124e132.
Fungal Biol. Rev. 22, 71e76. Sack, L., Holbrook, N.M., 2006. Leaf hydraulics. Annu. Rev. Plant
Murray, C.D., 1926. The physiological principle of minimum work. Biol. 57, 361e381.
I. The vascular system and the cost of blood volume. PNAS 12, Salathe, M., May, R.M., Bonhoeffer, S., 2005. The evolution of
207e214. network topology by selective removal. J. R. Soc. Interface 2,
Nakagaki, T., Iima, M., Ueda, T., Nishiura, Y., Saigusa, T., Tero, A., 533e536.
Kobayashi, R., Showalter, K., 2007. Minimum-risk path finding Savage, V.M., Deeds, E.J., Fontana, W., 2008. Sizing up allometric
by an adaptive amoebal network. Phys. Rev. Lett. 99, 068104. scaling theory. PLoS Comput. Biol. 4, e1000171.
Obara, B., Fricker, M.D., Gavaghan, D., Grau, V., 2012 .Jan 27.. Selosse, M.A., Richard, F., He, X.H., Simard, S.W., 2006. Mycor-
Contrast-independent curvilinear structure detection in rhizal networks: des liaisons dangereuses? Trends Ecol. Evol.
biomedical images. IEEE Trans Image Process [Epub ahead of 21, 621e628.
print]. Sherman, T.F., 1981. On connecting large vessels to small: the
Obara, B., Fricker, M.D., Gavaghan, D., Grau, V., 2012. Contrast- meaning of Murray’s law. J. Gen. Physiol. 78, 431e453.
independent Junction Detection of Branching Points in Sherman, T.F., Popel, A.S., Koller, A., Johnson, P.C., 1989. The cost
Network-like Structures. SPIE Medical Imaging, San Diego, CA. of departure from optimal radii in microvascular networks. J.
Obara B., Grau V., Fricker M.D. Bioimage informatics approaches Theor. Biol. 136, 245e265.
for extraction and analysis of fungal networks. Bioinformatics, Simard, S.W., Beiler, K.J., Bingham, M.A., Deslippe, J.R., Philip, L.J.,
submitted for publication. Teste, F.P., 2012. Mycorrhizal networks: mechanisms, ecology
Olsson, S., 2001. Colonial growth of fungi. In: Howard, R.J., and modelling. Fungal Biol. Rev. 26.
Gow, N.A.R. (Eds.), Biology of the Fungal Cell. Springer, Simard, S.W., Durall, D.M., 2004. Mycorrhizal networks: a review
pp. 125e141. of their extent, function, and importance. Can. J. Bot. 82,
Olsson, S., Gray, S.N., 1998. Patterns and dynamics of 32P-phos- 1140e1165.
phate and labelled 2-aminoisobutyric acid (14C-AIB) translo- Smith, M.L., Bruhn, J.N., Anderson, J.B., 1992. The fungus Armil-
cation in intact basidiomycete mycelia. FEMS Microbiol. Ecol. laria bulbosa is among the largest and oldest living organisms.
26, 109e120. Nature 356, 428e431.
Pajor, R., Falconer, R., Hapca, S., Otten, W., 2010. Modelling and Southworth, D., He, X.H., Swenson, W., Bledsoe, C.S.,
quantifying the effect of heterogeneity in soil physical condi- Horwath, W.R., 2005. Application of network theory to poten-
tions on fungal growth. Biogeosciences 7, 3731e3740. tial mycorrhizal networks. Mycorrhiza 15, 589e595.
Pries, A.R., Cornelissen, A.J.M., Sloot, A.A., Hinkeldey, M., Steinberg, G., 2007. Hyphal growth: a tale of motors, lipids, and
Dreher, M.R., Ho € pfner, M., Dewhirst, M.W., Secomb, T.W., the Spitzenkorper. Eukaryot. Cell 6, 351e360.
Analysis of fungal networks 29

Stumpf, M.P.H., Wiuf, C., May, R.M., 2005. Subnets of scale-free Wells, J.M., Boddy, L., Evans, R., 1995. Carbon translocation in
networks are not scale-free: sampling properties of networks. mycelial cord systems of Phanerochaete velutina (Dc, Pers) Par-
PNAS 102, 4221e4224. masto. New Phytol. 129, 467e476.
Suelmann, R., Fischer, R., 2000. Nuclear migration in fungi e West, G.B., Brown, J.H., 2004. Life’s universal scaling laws. Phys.
different motors at work. Res. Microbiol. 151, 247e254. Today 57, 36e42.
Tero, A., Kobayashi, R., Nakagaki, T., 2006. Physarum solver: West, G.B., Brown, J.H., Enquist, B.J., 1997. A general model for the
a biologically inspired method of road-network navigation. origin of allometric scaling laws in biology. Science 276,
Physica A 363, 115e119. 122e126.
Tero, A., Kobayashi, R., Nakagaki, T., 2007. A mathematical model West, G.B., Brown, J.H., Enquist, B.J., 1999a. The fourth dimension
for adaptive transport network in path finding by true slime of life: fractal geometry and allometric scaling of organisms.
mold. J. Theor. Biol. 244, 553e564. Science 284, 1677e1679.
Tero, A., Yumiki, K., Kobayashi, R., Saigusa, T., Nakagaki, T., 2008. West, G.B., Brown, J.H., Enquist, B.J., 1999b. A general model for
Flow-network adaptation in Physarum amoebae. Theory Biosci. the structure and allometry of plant vascular systems. Nature
127, 89e94. 400, 664e667.
Tlalka, M., Bebber, D., Darrah, P.R., Watkinson, S.C., Fricker, M.D., West, G.B., Woodruff, W.H., Brown, J.H., 2002. Allometric scaling
2007. Emergence of self-organised oscillatory domains in of metabolic rate from molecules and mitochondria to cells
fungal mycelia. Fungal Genet. Biol. 44, 1085e1095. and mammals. PNAS 99, 2473e2478.
Tlalka, M., Bebber, D.P., Darrah, P.R., Watkinson, S.C., Whiteside, M.D., Treseder, K.K., Atsatt, P.R., 2009. The brighter
Fricker, M.D., 2008. Quantifying dynamic resource allocation side of soils: quantum dots track organic nitrogen through
illuminates foraging strategy in Phanerochaete velutina. Fungal fungi and plants. Ecology 90, 100e108.
Genet. Biol. 45, 1111e1121. Whitfield, J., 2007. Fungal roles in soil ecology: underground
Tlalka, M., Watkinson, S.C., Darrah, P.R., Fricker, M.D., 2002. networking. Nature 449, 136e138.
Continuous imaging of amino-acid translocation in intact Wick, L.Y., Furuno, S., Harms, H., 2010. Fungi as transport
mycelia of Phanerochaete velutina reveals rapid, pulsatile fluxes. vectors for contaminants and contaminant-degrading
New Phytol. 153, 173e184. bacteria. In: Timmis, K.N. (Ed.), Handbook of Hydrocarbon
Truskey, G.A., Yuan, F., Katz, D.F., 2010. Transport Phenomena in and Lipid Microbiology. Springer Berlin, Heidelberg,
Biological Systems. Pearson Prentice Hall, New Jersey. pp. 1555e1561.
Tucker, K.G., Kelly, T., Delgrazia, P., Thomas, C.R., 1992. Fully- Wood, J., Tordoff, G.M., Jones, T.H., Boddy, L., 2006. Reorganiza-
automatic measurement of mycelial morphology by image tion of mycelial networks of Phanerochaete velutina in response
analysis. Biotechnol. Prog. 8, 353e359. to new woody resources and collembola (Folsomia candida)
Watanabe, S., Tero, A., Takamatsu, A., Nakagaki, T., 2011. Traffic grazing. Mycol. Res. 110, 985e993.
optimization in railroad networks using an algorithm Woolston, B.M., Schlagnhaufer, C., Wilkinson, J., Larsen, J., Shi, Z.,
mimicking an amoeba-like organism, Physarum plasmodium. Mayer, K.M., Walters, D.S., Curtis, W.R., Romaine, C.P., 2011.
Biosystems 105, 225e232. Long-distance translocation of protein during morphogenesis
Watkinson, S.C., 1984. Inhibition of growth and development of of the fruiting body in the filamentous fungus, Agaricus bispo-
Serpula lacrymans by the non-metabolized amino-acid analog rus. PLoS ONE 6, e28412.
alpha-aminoisobutyric-acid. FEMS Microbiol. Lett. 24, 247e250. Xu, H., Falconer, R., Bradley, D., Crawford, J., 2009. FUNNet e
Wells, J.M., Boddy, L., 1995. Effect of temperature on wood decay a novel biologically-inspired routing algorithm based on fungi.
and translocation of soil-derived phosphorus in mycelial cord In: Communication Theory, Reliability, and Quality of Service,
systems. New Phytol. 129, 289e297. pp. 97e102.