ArticlePDF Available

Three decades post reforestation has not led to the reassembly of arbuscular mycorrhizal fungal communities associated with remnant primary forests



The negative effects of deforestation can potentially be ameliorated through ecological restoration. However, reforestation alone may not reassemble the same ecological communities or functions as primary forests. In part, this failure may be owed to forest ecosystems inherently involving complex interactions among guilds of organisms. Plants, which structure forest food webs, rely on intimate associations with symbiotic microbes such as root-inhabiting mycorrhizal fungi. Here, we leverage a large-scale reforestation project on Hawai'i Island underway for over three decades to assess whether arbuscular mycorrhizal (AM) fungal communities have concurrently been restored. The reference ecosystem for this restoration project is a remnant montane native Hawaiian forest that provides critical habitat for endangered birds. We sampled soils from 12 plots within remnant and restored forest patches and characterized AM fungal communities using high throughput Illumina MiSeq DNA sequencing. While some AM fungal community metrics were comparable between remnant and restored forest (e.g., species richness), other key characteristics were not. Specifically, community membership and the identity of AM fungal keystone species differed between the two habitat types, as well as the primary environmental factors influencing community composition. Remnant forest AM fungal communities were strongly associated with soil chemical properties, especially pH, while restored forest communities were influenced by the spatial proximity to remnant forests. We posit that combined, these differences in soil AM fungal communities could be negatively affecting the recruitment of native plant hosts and that future restoration efforts should consider plant-microbe interactions as an important facet of forest health.
Molecular Ecology. 2020;00:1–14.
  1© 2020 John Wiley & Sons Ltd
Received: 22 January 2020 
  Revised: 21 August 2020 
  Accepted: 24 August 2020
DOI : 10.1111/me c.1 5624
Three decades post-reforestation has not led to the reassembly
of arbuscular mycorrhizal fungal communities associated with
remnant primary forests
Christopher B. Wall1| Cameron P. Egan1,2 | Sean I. O. Swift1| Nicole A. Hynson1
Christopher B . Wall and Cameron P. Egan contribute d equally.
1Pacific Biosciences Research Center,
University of Hawai‘i at Mānoa, Honolulu,
2Depar tment of Biolog y, Okanagan College,
Kelowna, BC, Canada
Christopher B. Wall, Pacific Biosciences
Research Center, Universit y of Hawai‘i at
Mānoa, Honolulu, HI, USA.
This study was supported by National
Science Foundation (award #1556856) and
the W.M. Keck Foundation.
The negative effects of deforestation can potentially be ameliorated through eco-
logical restoration. However, reforestation alone may not reassemble the same eco-
logical communities or functions as primary forests. In part, this failure may be owed
to forest ecosystems inherently involving complex interactions among guilds of or-
ganisms. Plants, which structure forest food webs, rely on intimate associations with
symbiotic microbes such as root-inhabiting mycorrhizal fungi. Here, we leverage a
large-scale reforestation project on Hawai‘i Island underway for over three decades
to assess whether arbuscular mycorrhizal (AM) fungal communities have concur-
rently been restored. The reference ecosystem for this restoration project is a rem-
nant montane native Hawaiian forest that provides critical habitat for endangered
birds. We sampled soils from 12 plots within remnant and restored forest patches
and characterized AM fungal communities using high-throughput amplicon sequenc-
ing. While some AM fungal community metrics were comparable between rem-
nant and restored forest (e.g. species richness), other key characteristics were not.
Specifically, community membership and the identity of AM fungal keystone species
differed between the two habitat types, as well as the primary environmental factors
influencing community composition. Remnant forest AM fungal communities were
strongly associated with soil chemical properties, especially pH, while restored forest
communities were influenced by the spatial proximity to remnant forests. We posit
that combined, these differences in soil AM fungal communities could be negatively
affecting the recruitment of native plant hosts and that future restoration efforts
should consider plant–microbe interactions as an important facet of forest health.
arbuscular mycorrhizal fungi, deforestation, ecological networks, ecological restoration,
microbial ecology, symbiosis
   WALL e t AL.
As the global rate of deforestation has increased precipitously over
the last 100 years, especially in the tropics, we now face the neg-
ative repercussions of these losses, such as changes in ecosystem
function, the loss of carbon storage, the deterioration of water and
nutrient cycles, and the loss of biodiversity (Lamb, 2005). While
major local and regional reforestation efforts are underway, it re-
mains unclear whether restored forests will ever regain the eco-
system functions of their primary forest predecessors (Crouzeilles
et al., 2016). Most terrestrial habitat restoration projects follow
a similar path: designation of an area for restoration based on its
potential to serve specific functions (e.g. critical habitat for endan-
gered species, freshwater regeneration), initial management of the
area to reduce non-native species, followed by outplanting of na-
tive vegetation, and, ideally, monitoring and assessment of recovery
(Suding, 2011). While this approach has been successful in some in-
stances (Nellemann & Corcoran, 2010), there are far more examples
of incomplete or stalled ecosystem recovery (Godefroid et al., 2011;
Maron et al., 2012). One potentially important factor often over-
looked in restoration practices is the role of symbiotic interactions
between plants and soil microbiota, such as plant-associated mycor-
rhizal fungi (Perring et al., 2015).
Mycorrhizal fungi are ubiquitous and critical root symbi-
onts of nearly all terrestrial plant species on earth (Brundrett &
Tedersoo, 2018). These biotrophic fungi serve a variety of roles—
from increasing host plants’ access to, and uptake of key nutrients,
to affecting plant water relations and providing protection from
herbivores and pathogens—all in exchange for plant-derived car-
bohydrates and lipids (Choi, Summers, & Paszkowski, 2018; Smith
& Read, 2008). At the scale of ecological communities, mycorrhi-
zal fungi can have significant effects on plant community compo-
sition and ecosystem function (van der Heijden, Martin, Selosse,
& Sanders, 2015). These effects are largely owed to changes in
the community composition of mycorrhizal fungi across habitats
(Steidinger et al., 2019) or over time (Martínez-García, Richardson,
Tylianakis, Peltzer, & Dickie, 2015). For instance, the establishment
of some plants is dependent on them forming a symbiosis with spe-
cific species or communities of mycorrhizal fungi (Bever, Schultz,
Pringle, & Morton, 2001).
Disturbances such as land conversion can have long-last-
ing effects belowground, such as altering soil chemistry (Tellen
& Yerima, 2018), seed banks (Weerasinghe, Ashton, Hooper, &
Singhakumara, 2019), and changing the composition of mycorrhizal
fungi communities (Oehl et al., 2010). These alterations can prevent
the reestablishment of historic native vegetation types and rein-
force extant, and often undesirable plants, such as invasive species
(Klironomos, 2002; Wubs, van der Putten, Bosch, & Bezemer, 2016).
This in turn can lead to positive feedbacks between existing vegeta-
tion, soil chemistry, and mycorrhizal fungi and can hinder the recruit-
ment of other plant or fungal species (Hausmann & Hawkes, 2010;
Kivlin & Hawkes, 2011). Even if host plants are generalists and
do not require specific mycorrhizal symbionts, not all symbionts
provide equal benefits to their hosts (Kiers, Lovelock, Krueger, &
Herre, 2000). Thus, when restoration effor ts fail to reestablish focal
plant species, it may often be owed to combinations of unamend-
able abiotic and biotic soil conditions (Heneghan et al., 2008; Wall
& Stevens, 2014).
To ameliorate the negative effects of land conversion, previous
studies have shown that incorporating mycorrhizal fungi from more
intact and native “reference” ecosystems into ecological restoration
practices, through either direct inoculations (Koziol & Bever, 2017)
or established plants facilitating mycorrhizal symbiosis in seedling
recruits (Dickie, Koide, & Steiner, 2002), can increase rates of na-
tive plant establishment, increase ecosystem diversity, and speed
community succession (Koziol & Bever, 2017). Similarly, studies from
other guilds of soil microbes have shown that spatial proximity plays a
role in determining microbial community membership, with a general
pattern of decreasing similarity in communities as distance increases
(Bahram et al., 2013; Horner-Devine, Claire Horner-Devine, Lage,
Hughes, & Bohannan, 2004; but see Meyer et al., 2018). However,
such studies are rare in tropical systems (Maltz & Treseder, 2015),
and it is unclear whether common practices such as outplanting
native plant species will effectively reassemble mycorrhizal fungal
communities similar to those found in reference habitats in the ab-
sence of any additional manipulations. It is also unknown whether
proximity to reference habitats influences AM fungal community
composition in neighbouring restoration sites. Therefore, a better
understanding of how common restoration practices affect mycor-
rhizal fungal communities over time and in relationship to geographic
proximity to reference communities is an important consideration
for habitat restoration.
Tropical oceanic islands, such as the Hawaiian archipelago, are
particularly susceptible to the negative effects of land conversion
(Harter et al., 2015). Tropical ecosystems have been disproportion-
ally affected by deforestation, which has led to increased carbon
emissions and decreases in biodiversity (Bastin et al., 2019; Haddad
et al., 2015; Keenan et al., 2015; Liang et al., 2016). These changes in
turn influence the magnitude and abundance of species interactions
or “networks” (Poisot, Stouffer, & Gravel, 2015; Tulloch, Chadès, &
Lindenmayer, 2018). Intact ecosystems, such as primary Hawaiian
forests, are made up of complex networks of interacting species
from numerous guilds (Bascompte & Jordano, 2007). The member-
ship and abundance of organisms within a guild can have strong
top-down or bottom-up effects on the function and diversity of the
ecological community as a whole (Tylianakis, Laliberte, Nielsen, &
Bascompte, 2010). Therefore, assessing within guild dynamics (e.g.
species diversity and interactions) can provide insight into broader
ecosystem properties (Bascompte & Jordano, 2013). A useful tool
for making these assessments, especially in the case of cryptic and
hyperdiverse communities of organisms such as soil microbes, is
through modelling one mode (co-occurrence) networks (Banerjee
et al., 2016).
On Hawaiʻi Island, the Hakalau Forest National Wildlife Refuge
(Hakalau) contains one of the largest and longest running reforesta-
tion projects in the State. In response to the need to protect some
of Hawaiʻi's most threatened and endangered bird species and their
native forest habitats, the Hakalau Refuge was established in 1985.
In this 32,733 acre refuge, there exists a mosaic of habitats, which
range from intact montane tropical forest that provide critical hab-
itat for endangered endemic bird species to 150-year-old and now
abandoned pasturelands (Jeffrey & Horiuchi, 2003). Since 1987, over
390,000 native Acacia koa (koa) trees have been outplanted into over
5,000 acres of abandoned pastureland on the refuge with the goals
of reducing grass cover, creating amenable conditions for native
canopy and understory plant species to reestablish, and providing
additional native bird habitat for endangered species (Yelenik, 2017).
Since the late 1990s, some of these previously afforested koa corri-
dors have undergone additional outplantings of native woody spe-
cies representative of the remnant forest communities including
Metrosideros polymorpha (‘ōhi‘a lehua), Cheirodendron trigynum (ʻō-
lapa), Myrsine lessertiana (kōlea), Coprosma rhynchocarpa (pilo), Rubus
hawaiensis (ʻākal a) , Leptecophylla tameiameia (pu kiawe) and Vaccinium
calycinum (ʻōhelo).
Despite the presence of thousands of mature outplants in the
restored forests (Yelenik, 2017) and the recruitment of diverse avian
groups that serve as pollinators and seed dispersers for native plants
(Paxton et al., 2018), native seedling recruitment remains low. In
restored forests, native plants are less than half the relative cover
compared to remnant forests, endangered bird species remain infre-
quent, and restored forest understories remain dominated by exotic
grasses (Paxton et al., 2018; Yelenik, 2017). Based on three previ-
ous studies the low prevalence of native understory plant species
appears to be owed, at least in part to: (a) the effects of the nitro-
gen-fixing koa forest canopy on soil chemistry, which benefits nitro-
philic grasses and limits native plant recruitment (Yelenik, 2017) and
(b) the absence of diverse ground cover types (i.e. nurse plants and
logs, bryophytes) that have positive correlations with native seed-
ling recruitment (Rehm, Thomas, Yelenik, Bouck, & D’Antonio, 2019;
Yelenik, Dimanno, & D’Antonio, 2015). Combined, these findings in-
dicate a diversity of factors are hindering native plant recruitment
and the restoration of habitats for endangered bird species. An ad-
ditional, and previously unexplored, consideration is how land use
history and subsequent reforestation has influenced soil microbiota,
such as arbuscular mycorrhizal (AM) fungi.
Here, we set out to examine changes in the diversity, community
composition, and spatial distribution of AM fungi ~30 years post-re-
forestation of a subtropical montane native Hawaiian ecosystem.
We leverage the use of species co-occurrence networks to evalu-
ate changes in AM fungal communities found in the soil of remnant
and restored forest habitats (Agler et al., 2016; Chagnon, Bradley,
Maherali, & Klironomos, 2013; Tulloch et al., 2016). We focused on
the communities of AM fungi present in the soils of remnant native
forest patches (the reference habitat for restoration efforts) and
compared these to soils from forests that have undergone extensive
refo rest ation and na tiv e pla nt sp e cie s out plantin g. We cho se to anal-
yse soil communities as these represent a mixture of the fungi ac-
tively colonizing hosts, as well as the inoculum potential from fungal
spores or other reproductive structures. We focused on AM fungi as
these are the dominant mycorrhizal symbionts associated with our
study system and are well known as important mutualists shaping
the performance of terrestrial plants (van der Heijden et al., 1998;
Wardle et al., 2004). Specifically, we asked whether there were sig-
nificant differences in the diversity and community composition of
soil AM fungal communities between replicate remnant and out-
planted (restored) forest plots. We then assessed the importance of
two main factors that could be influencing the diversity and compo-
sition of these fungal communities: (a) soil chemistry and (b) spatial
proximity of restored forest to remnant. Following these analyses,
we compared the fungal co-occurrence networks between remnant
and restored forests, assessed whether they differed in key network
characteristics, and identified differences in microbial keystone taxa.
Finally, we visualized the spatial aggregation of the restored forest
keystone species using kriging as a form of data interpolation to as-
sess how the distribution of keystone taxa varies across the forest
landscape and examined how spatial proximity of restored forests to
remnant forests affects AM fungal community composition.
2.1 | Study location
Soil samples were collected on the eastern slope of Mauna Kea within
the Hakalau Forest National Wildlife Refuge on the Island of Hawai‘i
(19°51′N, 155°18′W) (Figure S1). Hakalau has sustained close to
two centuries of habitat alteration, with large areas deforested and
converted to pasture for livestock grazing (upland forests) or agri-
culture (lowland forests). Intact remnant forest canopies of Hakalau
are dominated by A . koa and M. polymorpha and are vital habitats
for critically endangered native Hawaiian bird species. Restoration
of pastureland parcels began in 1987 with the planting of native A.
koa seedlings (~390,000) over a 3-year period, in additional to other
native woody species outplanted in the late-1990s, with the objec-
tive of restoring native forest vegetation and reestablishing habitat
for endangered native bird species (Jeffrey & Horiuchi, 2003). Today,
while the composition of the plant communities between remnant
and restored forests are similar (Figure S2), the age-stage demo-
graphics of the two habitats are different. Canopy tree diameter at
breast height (dbh) is larger in the restored forests (Figure S3), while
native understory development and seedling recruitment in the
restored forests is relatively low compared to the remnant forests
(Yelenik, 2017).
2.2 | Sample Collection
Soil samples were collected in July and August 2017 alongside host
plant species. Sampling occurred within two main habitats found
within Hakalau: (a) corridors of outplanted A. koa where additional
common native woody plants were also outplanted to mimic the
remnant forest communities, such as C. trigynum, M. lessertiana, C.
   WALL e t AL.
rhynchocarpa, R. hawaiensis, and V. calycinum but large patches of
pasture grass remain; and (b) remnant forest patches co-dominated
by A. koa and M. polymorpha, with sub-canopies consisting of native
trees, shrubs, ferns, and forbs where small patches of pasture grass
have invaded. We refer to these two habitats as either restored or
remnant forests. These habitat types were used to assess the com-
munities and networks of soil-borne AM fungal symbionts.
Twelve plots were selected along two transects within Hakalau
Forest, with three replicate plots for each habitat type (remnant
or restored forest) in each transect (Figure S1). Plot selection was
based on restored plots having a similar plant community compo-
sition to remnant forests (Figure S2). Within a single transect, rem-
nant forest plots were within 0.9–1.3 km (transect 8, south) and
0.2–1.0 km (transect 9, north) from neighbouring restored forest
plots (Figure S1), with distances between replicate plots ranging
from 96 to 491 m (remnant forest) and 39 to 364 m (restored for-
est). Irrespective of habitat type, the distance between plots ranged
from~40 m to 3 km, with the shortest between-plot distance being
beyond the distance for spatial autocorrelation in AM genets (>1 m)
(Mummey & Rillig, 2008). Within each plot (~12 m diameter) soil
samples were collected by removing the top organic matter (OM)
layer of recognizable duff and sampling the top 10 cm of mineral soil
using a sterilized bulb planter for a total volume of ~430 ml soil per
sample. Sampling at this depth captures the majority of AM fungi
(Davison et al., 2012). Samples were taken adjacent (<1 m) to up to
eight individuals of seven host plant species, each of which is known
to form symbiotic associations with AM fungi and had their locations
mapped, which allowed us to perform spatially explicit comparisons
of AM fungi within and among plots. This included six native host
species; A. koa, C. rhynchocarpa, C. trigynum, M. lessertiana, M. poly-
morpha, and R. hawaiensis, and non-native grasses dominated by P.
clandestinum (grasses were not identified to species). Where eight
individuals of each target species were not present, sampling inten-
sity was increased so that eight soil samples were collected from
each target species in each forest plot. In total, 56 soil samples were
taken in each plot, amounting to 672 soil samples collected across all
forest plots (n = 336 restored forest, n = 336 remnant forest). Soil
samples were bagged, stored on ice, and transferred to a 1°C cold
room where they were stored for up to one week until they were
further processed for DNA extraction.
2.3 | DNA Extraction and Sequencing
Soil samples were prepared for DNA extraction by first passing
samples through a 5 mm sieve to remove roots and larger debris.
To standardize the amount of soil material we were sampling from,
the remaining fine soil was homogenized for 1 min and a ~0.5 g sub-
sample was then taken and oven dried at 20°C for 72 hr. Drying soil
is a common practice in molecular studies of AM fungi and does not
result in selective degradation of fungal DNA (Janoušková, Püschel,
Hujslová, Slavíková, & Jansa, 2015). Initial tests revealed no change
in soil mass after 72 h of drying, indicating that all soil moisture had
been lost. After drying, samples were homogenized again for 1 min
and a 0.25 g subsample was transferred to 2 ml DNA extraction tubes.
DNA was extracted using the MP Bio FastDNA® spin kit for plant
and animal tissue (MP Biomedic als), following the manufacturer's in-
structions. A two-step PCR reaction was subsequently performed to
amplify the small subunit (SSU) of AM fungi's ribosomal RNA (rRNA)
and to adhere Illumina barcodes and adaptors to our amplicons. The
first PCR reaction targeted a 550 bp of AM fungal SSU rRNA using
the universal eukaryotic primer WANDA (Dumbrell et al., 2011) with
the Fluidigm tag CS1 attached to the 5′ end (Fluidigm Inc) and the
Glomeromycotina specific primer AML2 (Lee, Lee, & Young, 2008)
with the Fluidigm tag CS2 attached to the 3′ end. PCRs were per-
formed in 25 µl using 2.5 µl of DNA template, 7.8 µl of nuclease-free
H2O, 12.5 µl of 2× Kapa plant PCR buffer (containing 1.5 mM MgCl2
[1X] and 0.2 mM of each dNTP), 1.5 µl of 25 mM MgCl2, 0.5 µl of
50 μM AML2/CS2, 0.5 µl of 50 μM WANDA/CS1, and 0.2 µl 2.5 U/
µL Kapa3G DNA Polymerase (Kapa Biosystems). An Eppendorf
Mastercycler Nexus Thermal Cycler (Eppendorf North America) was
used for PCR reac tions, with thermo cycler se t at 95°C (2 min); 30 cy-
cles of 95°C (30 s), 54°C (40 s), 72°C (1 min), and then 72°C (10 min).
The second PCR primer complexes used the same Fluidigm tags
(CS1 or CS2) used in PCR1 primers, an 8 bp Illumina Nextera barcode
and an Illumina adapter (Egan et al. 2018). PCR2 reactions consisted
of 15-fold diluted amplicons from PCR1. PCRs in 20 µl were per-
formed using 1 µl of diluted PCR1 product, 6.89 µl of nuclease-free
H2O, 10 µl of 2× Kapa plant PCR buffer (containing 1.5 mM MgCl2
[1X] and 0.2 mM of each dNTP), 1.2 µl of 25 mM MgCl2, 0.75 µl of
2 μM Illumina adaptor/barcode primer complex, and 0.16 µl 2.5 U/
µL Kapa3G DNA Polymerase. Thermocycler conditions were: 95°C
(1 min), followed by 12 cycles of 95°C (30 s), 60°C (30 s), 68°C (1 min),
followed by 68°C (5 min). Amplicons were purified and normalized
to 25 ng using the Just-a-PlateTM PCR purification and normaliza-
tion plate (Charm Biotech), following the manufacturer's protocol.
Soil DNA amplicon pools from pooling 10 μl of eluted 25 ng DNA
from each sample were purified again and size selected for ampli-
cons ~650 bp in length using AMPure XP beads (Beckman Coulter).
Soil DNA libraries were sequenced on a single Illumina MiSeq run,
using 2 × 300 paired-end (PE) sequencing and 600 cycles (Illumina
Inc.) by the Genomics High Throughput Facility at the University of
California, Irvine (https://ghtf.bioch
2.4 | Bioinformatics
Four fastq files were generated from sequencing (forward and reverse
barcode reads and forward and reverse target reads). For each sample,
forward and reverse barcode reads were extracted and re-assembled
into a new fastq file representing each sample's unique barcode com-
bination using the command (https://gist. rst/6284164). Demultiplexing of forward and re-
verse reads was performed using the qiime demux emp-paired command
and resulted in 4,435,672 paired sequences being assigned to sample
barcodes. After demultiplexing, chimeras were removed, sequences
were quali ty ch ecked and the n as sembled into exa ct sequ ence variant s
(ESV) using DADA2 (Callahan et al., 2016) and the dada2 denoise-paired
command using QIIME default parameters. Of the initial 4,435,672 de-
multiplexed reads, 1,324,599 had a higher number of errors than ex-
pected, 743,188 did not merge with a respective pair-end, and 96,246
were identified as being chimeric, resulting in 2,271,639 sequences
being retained and assembled into 2,318 ESVs.
Quality filtered sequences were then identified to virtual taxa
(VT) in MaarjAM (Öpik, Metsis, Daniell, Zobel, & Moora, 2009) using
the BLAST + algorithm (Camacho et al., 2009), a sequence similarity
of ≥97%, sequence alignment ≥97%, and a BLASTe-value < 1e-50 .
Using these parameters, 359 ESVs were identified to VTs (repre-
senting 15% of ESVs that contained 544,786 sequences, ~24% of
sequences that passed initial quality filters). Reads that did not get
a hit with these criteria were subjected to a second BLAST + search
against the NCBI nucleotide database (accessed February 2019)
using a sequence similarity of ≥90%, and a BLAST e-value of <1e-5 0.
During this second identification round, a further 82 ESVs (repre-
senting 4% of total ESVs, that contained 10,991 sequences, ~0.5% of
sequences that passed initial quality filters) were identified. Because
our study system is relatively understudied in AM fungal research
(Öpik, Moora, Liira, & Zobel, 2006), we expected to detect a high
incidence of Glomeromycotina sequences not previously identified
and deposited into a curated database (Ohsowski, Zaitsoff, Opik, &
Hart, 2014). Therefore, we conducted a third identification round of
remaining unidentified ESVs using the MaarjAM database a second
time, and a relaxed sequence similarity of ≥80% (default in QIIME2).
We maintained a BLAST e-value of <1e-50. We chose this method
because previously unidentified Glomeromycotina species may be
similar to taxa already present in databases, but will not be iden-
tified using strict search criteria. Using this method, an additional
1,171 ESVs (representing ~51% of total ESVs, containing 1,274,396
sequences, which were ~56% of sequences that passed initial qual-
ity filters) were taxonomically identified during this final round of
taxonomic assignment. Sequences were identified to the closest
VT in MaarjAM and newly identified ESVs were denoted by ap-
pending ‘_unidentified’ after the VT name (e.g. ESVs identified to
VTX00242 were identified as VTX00242_unidentified). Sequences
that remained unidentified, along with non-Glomeromycotina se-
quences, were filtered from downstream analyses using the qiime
taxa filter-table command. To account for differences in sequence
depth among samples, samples were rarefied to 1,500 sequences
per individual soil sample after removing all non-Glomeromycotina
and sequences present within PCR negative controls. The filtered
ESV table, taxonomy identification file, and metadata file were then
exported from QIIME for downstream analyses in R (version 3.6.1;
R Core Team, 2019).
2.5 | Soil Chemistry
A subset of soil samples (n = 20–28 per plot) were analysed for OM,
estimated total nitrogen (N), readily available phosphorus (P) using
weak Bray PI values, extractable cations (potassium [K], magnesium
[Mg], calcium [Ca], sodium [Na]), hydrogen (H), sulphate-S (S), pH and
cation exchange capacity (CEC). Soil chemical analyses were per-
formed following standard procedures by A & L Western Agricultural
Laboratories, Inc. Differences in soil chemistry parameters between
the two habitat types were assessed using Welch unequal variance
t tests (Welch, 1947).
2.6 | AM fungal diversity
AM fungal species richness was calculated as the sum of unique taxa
in each soil sample. The effect of habitat type on AM fungal species
richness was tested using a generalized linear model (GLM) with a
Poisson distribution and a log link function. Differences in AM fun-
gal community composition between habitats were examined using
the Bray–Curtis dissimilarity index (Bray & Curtis, 1957), with varia-
tion in AM fungal community composition visualized using nonmet-
ric multidimensional scaling (NMDS) ordination in three dimensions
(k = 3) using metaMDS in the package vegan (Oksanen et al., 2019).
Soil chemical properties were fitted to the NMDS ordination using
the envfit command in vegan. A permutational multivariate analysis
of variance using distance matrices (PERMANOVA; Anderson, 2001)
with 9,999 permutations was subsequently used to test for effects
of habitat type on AM fungal community composition using the
adonis function along with a test for multivariate homogeneity of
group dispersion using betadisper in the vegan package. To further
test the influence of environmental conditions on AM fungal com-
munities, all soil chemistry environmental traits were analysed in a
principal component analysis (PCA) and collapsed to a single prin-
cipal component explaining the largest proportion of variance (i.e.
PC1). The soil chemistry-PC1 and AM fungal community-NMDS1
were partitioned by habitat type and their statistical relationships
tested using a linear regression. Finally, spatial effects on community
dissimilarity were tested using distance between samples in metres
and the Bray–Curtis dissimilarity ordination of AM fungi community
through Mantel tests using Kendall rank correlation coefficient (r)
and 9,999 permutations using mantel. rtest in the ade4 package (Dray
& Dufour, 20 07). Mantel tests were performed at the level of the
entire study (pool ed across habit at types), within habitat types (rem-
nant or restored forests), and within transects with AM fungi com-
munities in restored forest plots subjected to pairwise tests based
on proximity to each of three remnant plots in the same transect.
2.7 | AM fungal co-occurrence networks
We examined AM fungal co-occurrence networks using probabilis-
tic graphical models and the Sparse Inverse Covariance Estimation
for Ecological Association Inference (SPIEC-EASI) package in R
(SpiecEasi, version 1.0.7; Kurtz, Mueller, Miraldi, & Bonneau, 2019)
using the spiec.easi function. Structure for the co-occurrence net-
work was selected using the Meinshausen and Bühlmann (MB)
   WALL e t AL.
algorithm (Meinshausen & Bühlmann, 2006) with co-occurrences of
virtual taxa (i.e. nodes) in the network normalized using regulariza-
tion parameter (“lambda”; Kurtz et al., 2019) estimated in Stability
Approach to Regularization Selection (StARS) bootstrapping (Liu,
Roeder, & Wasserman, 2010). The St ARS criterion is run through ex-
tensions of parallelized utilities for lambda selection with a regulari-
zation in the pulsar package in R (Müller, Bonneau, & Kurtz, 2016).
Our networks were assembled with lambda min/max ratio set to
0.01 and 20 and pulsar.params repetitions set to 100. Inferred co-
occurrence networks were converted to igraph objects using the
adj2igraph function, with node and network characteristics subse-
quently analysed using the igraph package (Csardi & Nepusz, 2006).
The observed networks were computed for the resident AM fungal
communities in the respective habitat types. Networks were then
reconstructed with a candidate keystone taxa computationally re-
moved (see below) to observe the effect of this hub on network to-
pology and structure (Agler et al., 2016).
AM fungal network characteristics between habitat types were
examined by comparing the overall network topologies by assem-
bling co-occurrence networks in each forest plot for each habitat
type. Our co-occurrence networks represent AM fungal taxa as
nodes, and if the presence or absence of a node is correlated with
another, these correlations are represented by edges (lines) be-
tween them. From these networks, we measured common network
metrics include node degree, betweenness centrality, and density
(Bascompte & Jordano, 2013). Nodes (or vertexes) in a network rep-
resent a taxonomic unit grouped at a specific level (in our case an
AM fungal VT). Edges are the lines connecting nodes in a network
and represent the correlations between nodes, with degrees being
the number of direct correlations for a node. Betweenness central-
ity is the proportion of times a node falls on the shor test network
path relative to all other nodes, and density is the obser ved network
edge s divid ed by all possib le edge s (Agler et al., 20 16). Bot h node de-
gree and betweenness centrality are used to assess a node's overall
connectedness within a network (Agler et al., 2016). We compared
all three metrics between the two habitat types using Welch unequal
variance t tests (Welch, 1947). Network “hubs” are defined as those
nodes with a high number of direct interactions that act to bridge
connections among other nodes within a network. As such, hubs
are important determinants for the assembly of the members of
the network with which they share connections (Agler et al., 2016;
Freeman, 1978). Therefore, we considered AM fungal VT (i.e. nodes)
with high connectedness (i.e. degrees) and betweenness centrality
scores as hubs.
Historically, keystone species have been characterized as those
occurring in relatively low abundance and having a disproportion-
ately large influence on their respective community or ecosystem
(Paine, 1969). Similar to previous studies of microbial networks, we
applied a parallel concept of a keystone species to identify potential
AM fungal keystone taxa (Agler et al., 2016). To assess which AM
fungal taxa may act as keystones, we evaluated each taxon in each
habitat's network for their relative abundance, betweenness cen-
trality, and node degree (Tipton et al., 2018). Then, node prevalence
scores were calculated by multiplying a taxon's relative abundance
in the network (i.e. proportion of soil samples where each taxon
was observed) by the relative abundance of the taxon within a soil
sample. Within each habitat's network, those taxa with low abun-
dance, but disproportionately large effects on determining network
architecture, were considered keystones (Agler et al., 2016; Berry &
Widder, 2014; Tipton et al., 2018). Candidate keystone AM fungal
taxa were subsequently identified as those taxa in the upper 75th
percentile for betweenness centrality, 95th percentile for node de-
gree and displaying the lowest prevalence score within this bin (i.e.
95th percentile), and being present in >1% of soil samples. The pro-
portion of total nodes each keystone interacted with were recorded,
and Welch unequal variance t tests were applied to each keystone
taxa to evaluate whether their distributions were influenced by hab-
itat types. As an additional measure to visualize the ubiquity and
abundance of AM fungi in soil samples, the propor tion of samples
where taxa were observed (i.e. ubiquity) was graphed against the
mean relative abundance for each VT (sensu Tipton et al., 2018).
We visualized the spatial distribution of candidate keystone taxa
across all plots using a heat map using the mean relative abundance
of a AM fungal keystone taxon in all soil samples in which they were
present. However, only the keystone taxon from the restored forest
was present in enough samples for spatial maps to be generated.
The purpose of the kriging map was to identify and visualize regions
of high and low mean relative abundance for this taxon. Kriging was
performed using an inverse distance weighted interpolation of the
keystone taxon's relative abundance (averaged across replicate soil
samples centred on a mapped individual host plant) to generate a
variogram model using krige in the R package gstat (G ler, Peb esma,
& Heuvelink, 2016; Pebesma, 2004). Predicted abundance values for
the restored forest keystone taxon were mapped in a spatial coor-
dinate object using the ssplot function in the sp package (Bivand,
Pebesma, & Gómex-Rubio, 2013).
Community membership of AM fungi identified in our soil samples
differed significantly between restored and remnant forest habitat
types (Figure 1a, Table 1, PERMANOVA: p < .001), and the variance
between AM fungal communities did not differ between forests
(p = .204). However, AM fungal species richness did not differ be-
tween forest habitat types (Figure S4, Table S1, p = .516). On aver-
age, 18 AM fungal V T were observed in each of the forest habitat
types (Figure S4). The factors influencing community composition
differed based on habitat type. In the remnant forests, AM fungal
community dissimilarity was influenced by changes in environment,
specifically soil chemistry (Figure 1b, Table 2); while restored for-
est communities were influenced by their overall proximity to other
restored forest plots (Figure S5c, Table S2) and the spatial proximity
to other remnant forest communities (Figure S6, Table S3, p < .001).
Overall, we observed the AM fungal communities pooled across
habitat types to become less similar as distance among sampling
points increased (Figure S5, Table S2, p < .001). Spatial patterns
within each habitat type revealed restored forest AM fungal com-
munities to also become less similar as distance between samples
increased (p < .001) but spatial proximity had no significant effect
on remnant forest communities (p = .108) (Figure S5b,c, Table S2).
Pairwise comparisons within each transect (n = 9 comparisons tran-
sect-1) revealed a significant positive distance decay relationships for
15 of 18 Mantel proximity tests (Table S3, p ≤ .037). Overall, AM
fungal communities in restored forests showed greater dissimilarity
as distance to neighbouring remnant forests increased (Figure S6).
Soil chemistry had a significant effect on remnant forest AM
fungi (Figure 1b), primarily pH, which explained nearly 26% of the
variation in community composition overall (Table S4). Applying a
principal component analysis to soil chemistry parameters produced
a single princi pal compo nent (P C1), which accounte d fo r 43% of va ri-
ation in soil chemistry (i.e. PC1-soil chemistry). Regressing PC1-soil
chemistry against the primary NMDS scores (i.e. NMDS1-AM fungi
[AMF] Bray–Curtis), a representation of beta diversity, allowed us
to assess the degree to which AM fungal communities between the
two forest habitats were changing in response to soil environmental
conditions. We observed PC1-soil chemistry to have a significant re-
lationship with AM fungal community NMDS1 in the remnant forest
(Table 2, p < .001), but this relationship was absent in the restored
forest (Figure 1b, p = .642). Soil chemistry differed between forest
habitat types, with the remnant forest plots having greater OM and
total nitrogen (p < .001), but lower potassium and sulphur (ppm)
(p ≤ .040) compared to the restored forest plots (Table S5). Remnant
forest soils were also more acidic (4.0 0 vs 4.78 pH, p < .001).
Conversely, restored forest soils had higher potassium, magnesium,
and calcium (Table S5, p ≤ .038). The NMDS analysis showed sig-
nificant soil chemistry vectors for total nitrogen, OM, CEC and H+,
which were all positively associated with the remnant forest and may
be driving differences in AM communities.
Co-occurrence network analyses revealed that the remnant for-
est harbours more connected (denser) networks than the restored
forest (Figure 2a,b, Table 3, p = .005), with a gre ate r number of ed ges
in the remnant forest compared to the restored forest (100 versus
79, respectively). However, the average betweenness centrality and
path length for networks in each plot was not affected by habitat
types (Table 3). Keystone species analysis detected two potential
keystone species, one from each habitat type (Figure 3). Remnant
forest candidate keystone taxon Claroideoglomus sp. VTX00225
had a betweenness centrality score of 193 and 8 node degrees,
being present in 2.8% of our remnant forest soil samples, but ab-
sent in the restored forest samples. The restored forest candidate
keystone taxon Acaulospora sp. VTX00272 had a betweenness
centrality score of 547 and 11 node degrees in the restored forest,
being present in 7.6% of samples (network prevalence 1.560e-04).
FIGURE 1 AM fungal communities in remnant and restored
forest habitat types within Hakalau Forest National Wildlife
Refuge assessed by (a) nonmetric multidimensional scaling
(NMDS) ordination plot and (b) a regression of NMDS1 by principal
component (PC1) values derived by compressing soil chemistry
metrics to a single dimension (i.e. PC1-soil chemistry). NMDS
ordination is based on Bray–Curtis dissimilarity, with ellipses (top
panel) representing 95% standard deviation. Arrows indicate
significant correlations (p < .05) in soil chemistry parameters
(see Table S4); vector lengths indicate strength of significant
relationships with NMDS1 and NMDS2. Lines (bottom panel)
represent linear model fits and standard error (shaded areas),
with a significant relationship in the remnant forest (p < .0 01;
Table 2)
TABLE 1 Results of permutational multivariate analysis of variance (PERMANOVA) model testing habitat type effects on arbuscular
mycorrhizal (AM) fungal community beta diversity
df SS MS F-Model R2P(>F)
Habitat type 13.695 3.696 9.9 3 4 0.021 <.001
Residual 474 176.338 0.372 0 .979
Tot al 475 180.034 1.000
significant effects are in bold (p < .05).
   WALL e t AL.
Acaulospora sp. VTX00272 was also present in the remnant forest
(25.6% of samples, network prevalence 1.965e-03) and tended to
have a higher mean ubiquity (0.041 versus 0.012, p = .073) and mean
relative abundance (0.008 versus 0.002, p = .069) in re mnant forests
compared to restored forests (Figure S7). Due to a greater abun-
dance and lower connectivity in the remnant forest, Acaulospora sp.
VTX00272 was not characterized as a keystone species in this hab-
itat (betweenness centrality score of 203, 5 node degrees). Across
both habitat types, the most abundant and ubiquitous taxa were the
same three AM fungal VT—two in the genus Acaulospora and one
Glomus sp. (Figure S7).
The computational removal of Acaulospora sp. VT X00272 from
both remnant and restored forest co-occurrence networks desta-
bilized network topologies, notably with a loss of connections be-
tween taxa (Figure 2c,d). This was exemplified by a 48% (remnant
forest) and 38% (restored forest) reduction in network edges and
a reduction in the number of network hubs (i.e. nodes with high
betweenness centrality and degree) and increasing network dis-
connectedness for both habitat types (Figure 2c,d). In assessing
the number of nodes that had network scores >0, the removal of
Acaulospora sp. VTX00272 led to a ~25% (degrees) and ~43% (be-
tweenness centrality) reduction in nodes in both networks as well as
an overall lower node connectivity, with a 96% and 50% reduction
in the mean betweenness centrality of nodes in remnant and re-
stored forests, respectively. Visualization of the spatial distribution
of Acaulospora sp. V TX0 0272 us in g kriging showed hot spots relating
to high abundance of this taxon in all remnant plots, but pronounced
hot spots were centralized in two remnant forest plots in the south-
ern transect (Figure S8).
Our ability to restore habitats is informed by our basic understand-
ing of how ecological communities assemble. While much emphasis
in terrestrial habitat restoration is focused on reassembling plant
communities, which form the basis of ecological food webs, plant
success is inherently tied to their interactions with microbial symbi-
onts, such as mycorrhizal fungi (van der Heijden et al., 2015). Here
we examined whether ~30 years post-reforestation has led to the
reassembly of the AM fungal communities that support more intact
remnant native forest. We found that while AM fungal species rich-
ness and the domina nt taxa did not differ significant ly bet ween rem-
nant and restored forest plots (Figures S2 and S4), other important
community characteristics did. Specifically, community membership
differed significantly between the two habitat types as did the fac-
tors shaping these communities (Figure 1 and Figure S5b,c), including
the proximity of restored forest plots to neighbouring remnant for-
est (Figure S6). The interaction networks of AM fungi in the restored
and remnant forests also differed. Remnant forest networks were
denser than restored forest networks (Figure 2a,b), and each habitat
harboured a unique candidate keystone taxon (Figure 3). The rem-
nant forest keystone species (Claroideoglomus sp. VTX00225) was
absent from the restored forest, but the opposite was true for the
restored forest keystone species (Acaulospora sp. VT X00272), which
was four times more abundant in the remnant forest (Figure 3), the
possible ramifications of which are discussed in further detail below.
Species richness of AM fungi did not significantly differ among
remnant and restored forest plots (Figure S4). This finding was
somewhat surprising as disturbance, such as conversion of forests to
pastures, often leads to a decrease in the number of AM fungi (Hart,
Zaitsoff, van der Heyde, & Pither, 2016 and references therein, but
see Lekberg, Gibbons, Rosendahl, & Ramsey, 2013). Given that com-
munity composition differed significantly between these two habi-
tat types, it is possible that species richness in the restored forest
soils is inflated due to legacy effects of the historic pasture AM
fungal communities (Lankau, Bauer, Anderson, & Anderson, 2014).
However, the most abundant and ubiquitous taxa across both habi-
tat types were two VT in the genus Acaulospora, followed by Glomus
sp. VTX00067. In particular, Glomus sp. VTX00067 was 21% more
df SS MS F-value P (>F)
Full model
PC1 10. 417 0.417 54.967 <.001
Habitat Type 10.102 0.102 13.503 <.0 01
PC1 × Habitat
Typ e
10.103 0.103 13.548 <.001
Residual 291 2.208 0.008
Remnant forest
PC1 10. 247 0. 247 27.322 <.001
Residual 123 1.111 0.009
Restored forest
PC1 10.001 0.001 0.217 .642
Residual 168 1.098 0.007
aFull model intercept and slopes applied in Figure 1. Individual models show relationship of PC1
and NMDS1 in each forest, and significant effects are in bold (p < .05).
TABLE 2 Linear model results for
relationship between the soil chemistry
environmental parameters compressed
to a single principal component (PC1)
and forest habitat (remnant or restored)
with arbuscular mycorrhizal (AM) fungi
virtual taxa analysed in a nonmetric
multidimensional scaling (NMDS1)a
ubiquitous and 31% more abundant in the restored forest relative
to the remnant (Figure S7). The genus Acaulospora is considered a
highly stress-tolerant group that are uncommon in grasslands and
are generally thought to provide less benefit to their hosts than
species in Glomeraceae and Gigasporaceae (Chagnon et al., 2013;
Hart & Reader, 2002). However, the cost–benefit interactions of
FIGURE 2 Arbuscular mycorrhizal (AM) fungi co-occurrence networks for fungal communities in soil of (a, c) remnant and (b, d) restored
forests in Hakalau Forest National Wildlife Refuge. The observed intact networks (top row) represent complete co-occurrence networks in
each forest habitat. Reconstructed networks (bottom row) are the co-occurrence networks with the candidate keystone taxon identified in
the restored forest (Acaulospora sp. VTX00272) removed from both fungal networks, which resulted in a reduction in network edges and the
number of network hubs (i.e. nodes with high betweenness centrality and degree) and increasing network disconnectedness for both habitat
types. AM fungal networks are inferred from Meinshausen and Bühlmann algorithms (Meinshausen & Bühlmann, 2006). Nodes represent
AM fungal virtual taxa and are coloured by taxonomic family. Edges represent nonrandom interaction between nodes
(a) (b)
(c) (d)
TABLE 3 Network characteristicsa for soil arbuscular mycorrhizal (AM) fungi in remnant and restored habitats in Hakalau Forest National
Wildlife Refuge. Statistical tests of among habitat type effects were evaluated by Welch t tests
Remnant forest Restored forest tdf p
Observed intact network
Betweenness centrality 0.050 ± 0.003 0.037 ± 0.007 1.654 7. 2 .141
Connectedness 0.034 ± 0.001 0.027 ± 0.001 3.593 10.0 .005
Path length 4.880 ± 0.185 5.410 ± 0.440 −1. 117 6.7 .303
aNetwork characteristic are displayed as values are mean ± SE (n = 6), and significant effects are in bold (p < .05).
   WALL e t AL.
Acaulospora spp. with hosts have primarily been tested with forbs
and grasses, and their benefit to native Hawaiian woody species may
differ (Hart & Reader, 2002; Hart et al., 2016). Given that Acaulospora
spp. were among the most abundant and ubiquitous VT in our study
system overall (0.68–0.78 ubiquity), especially in the remnant for-
est (Figures S7 and S8), we suspect that these taxa play key roles
in the assembly of native forest communities, rather than being in-
dicative of historical pastureland communities. Conversely, fungi in
Glomeraceae tend to dominate disturbed habitats and are among
the most common and generalist species of AM fungi (Davison
et al., 2015; Öpik et al., 2009). Thus, it is not surprising that Glomus
sp. VTX00067 was more common in the restored forest plots, which
have experienced a greater degree of recent disturbance and occu-
pation by generalist grass species.
Numerous studies have found that proximity to reference ecosys-
tems is an important factor in community reestablishment post-distur-
bance (Crouzeilles et al., 2016). Indeed, we found additional support
for this relationship: independent of habitat type, as distance among
plots increased, so did AM fungal community dissimilarity (Figure S5a).
Specifically, when we compared the similarity of AM fungal commu-
nities in restored forest plots from each transect to remnant forest
plots on the same transect, those closer to remnant forests were
generally more similar than those further away (Figure S6). Therefore,
restoration sites closer to remnant forests harbour AM fungal commu-
nities that are more similar and may be the most success ful for habitat
restoration. However, when pooling AM fungal communities across
transects, the remnant forest plots showed no significant change in
AM fungal community composition as distance between plots in-
creased (Figure S5a). In older forests such as our remnant native for-
est plots, a relatively more homogenous community across plots may
be owed to environmental filtering selecting for specific community
members (Glassman, Wang, & Bruns, 2017; House & Bever, 2018).
While not tested here, the density of trees and their roots can also
play an important role in mycorrhizal communities. For instance, in
ectomycorrhizal forests, the composition of fungal communities may
be af fected by gr adient s in root densit y occurring across fore st edges,
within soil profiles, and during early succession, and these gradients
may select for different ectomycorrhizal species based on their func-
tional traits (Peay, Kennedy, & Bruns, 2011). Whether AM fungal
communities are similarly influenced by root density is unclear, but
deserves additional attention.
Focusing on the remnant forest mycorrhizal communities, which
are considered the reference for forest restoration, we found that
soil chemistry, specifically pH, was a significant predictor of AM
fungal community composition (Table S4). This finding has been
recapitulated in numerous studies of mycorrhizal fungi (Chagnon
FIGURE 3 Bubble plots of betweenness centrality plotted against node degrees used to identify candidate keystone arbuscular
mycorrhizal (AM) fungi taxa in (a) remnant and (b) restored forests of Hakalau Forest National Wildlife Refuge. Points are scaled to the
presence of AM fungi virtual taxa (i.e. nodes) and coloured by taxonomic family. Candidate keystone taxa in each habitat type are identified
with arrows. The remnant forest keystone taxon (Claroideoglomus sp. V TX00225) was absent from the restored forest habitat
(a) (b)
et al . , 2013; Johns on, Za k, Til m an, & Pf lege r, 1991; Oe h l et al., 20 10),
and while management of soil properties may not be a practical
solution to enhancing the reestablishment of specific mycorrhizal
fungi, it is worth explor ing thr ough pr act ices su ch as rem nant fore st
leaf litter transplants. Interestingly, the most abundant and ubiqui-
tous taxa in the remnant forest were in the genus Acaulospora which
are known to be common in low pH environments (Figure S7). The
remnant forest networks were denser than their restored forest
counterparts indicating that they are more susceptible to pertur-
bations such as disturbance (Campbell, Yang, Shea, & Alber t, 2012).
Along these lines, the absence of the remnant forest keystone taxon
Claroideoglomus sp. VTX00225 in the restored forest may be owed
to the sensitivity of this genus to abiotic stress, and its affiliation
with undisturbed forest habitats (Lenoir, Fontaine, & Lounès-Hadj
Sahraoui, 2016). Due to the high network connectivity of this taxon,
its absence in the restored forest may be affecting assembly pro-
cesses in these communities and in turn, their function (Banerjee,
Schlaeppi, & van der Heijden, 2018; Duhamel et al., 2019; Tulloch
et al., 2018). However, additional experimental approaches that
specifically test effects of this taxon on AM fungal community as-
sembly and function are needed.
Our finding that the keystone taxon of the restored forest network
is present in significantly greater abundance in the remnant forests is
interesting and could be interpreted in multiple ways: it could indicate
that this species is a central player in native forest AM fungal commu-
nities, or conversely, that it has invaded the remnant forest from the
historic pastures. However, we suspect the former since Acaulospora
spp. are not common components of grassland communities and lack
traits associated with more ruderal species such as fast establish-
ment in soils and roots (Chagnon et al., 2013; Hart & Reader, 2002).
Under this scenario, it is encouraging that Acaulospora sp. VTX00272
is acting as a keystone species in the restored forests. This may indi-
cate that despite its lower abundance relative to the remnant forest,
Acaulospora sp. VTX00272 has maintained (or regained) its central-
ized role in facilitating the reassembly of AM fungal communities, and
overtime may contribute to restored forest AM fungal communities
becoming more similar to those in remnant forests. However, this hy-
pothesis requires further exploration through experimentation.
To test the prediction that Acaulospora sp. VTX00272 was im-
portant for AM fungal community assembly in both remnant and re-
stored forests, we removed it from both habitat's communities and
rebuilt our interaction networks. This resulted in significantly less
connected networks with fewer hubs (Figure 2c,d). The absence of
this taxon therefore destabilizes the network topology in both rem-
nant forests and restored forests. However, the greater connected-
ness and node degree of Acaulospora sp. VTX00272 in the restored
forests (Figure 2a,b) indicates this species is performing a dispro-
portionately greater role in the network topology in this habitat.
Together, these results demonstrate that regardless of abundance,
Acaulospora sp. V TX00272 is acting to knit together the AM fungal
communities and, based on its keystone status, appears to be of par-
ticular importance in the restored forest communities.
In summary, 30 years post-reforestation has led to the reestab-
lishment of some community properties of AM fungi, but the reas-
sembly processes remain incomplete or possibly stalled, which in
turn could be hindering the restoration of these forests. We suggest
the following for future forest restoration efforts: first, targeting
revegetating sites in close proximity to remnant forest communi-
ties may increase the likelihood of colonization by remnant forest
fungal communities; second, pre-inoculation of hosts with rem-
nant communities prior to outplanting may facilitate the success-
ful reestablishment of native plants with their mycorrhizal fungal
partners found in native forest soils, which has been shown previ-
ously to increase plant restoration success (Malz & Treseder, 2015;
Neuenkamp, Prober, Price, Zobel, & Standish, 2019); finally, the
functional roles of potential keystone, abundant and ubiquitous AM
fungal taxa should be further explored through manipulative exper-
iments. Overall, consideration of the symbiotic landscape should be
taken into account when attempting to restore forest communities.
Tropical forests are made up of complex interactions across guilds of
organisms, and restoration goals are unlikely to be met by focusing
on single groups in isolation.
We would like to thank Corbin Amend, Dave Bouck, Ken Davidson,
Sarah Schoepflin and Stephanie Yelenik for help with sampling,
and the Hakalau Forest National Wildlife Refuge for site access.
Additionally, we would like to thank James Downey, Kacie Kajihara,
Terrence McDermott and Danyel Yogi for their dedication and long-
hours processing samples for molecular and chemical analyses. We
also wish to thank Anthony Amend, Carla D'Antonio, Pierre-Luc
Chagnon, Evan Rehm, Laura Tipton and Stephanie Yelenik for valu-
able discussions relating to this study.
The authors declare no personal or commercial conflicts of interest.
C.P.E and N.A.H. designed the experiment. C.P.E., S.I.O.S and N.A.H.
collected the data. C.P.E. performed laboratory analyses and bioin-
formatics. C.B.W. and C.P.E. analysed the data. C.B.W., C.P.E. and
N.A.H. wrote the paper.
All data and code for analyses are archived at Github (ht tps://github.
com/cbwal l/Hakal au-metac ommun ity-AMF/relea ses/tag/vpub) and
Zenodo (Wall, 2020). DNA sequences and project metadata are ar-
chived in the NCBI Sequence Read Archive (SRA) database under
BioProject accession no. PRJNA644447.
Christopher B. Wall
Cameron P. Egan
Nicole A. Hynson
   WALL e t AL.
Agler, M. T., Ruhe, J., Kroll, S., Morhenn, C., Kim, S.-T., Weigel, D., &
Kemen, E. M. (2016). Microbial hub taxa link host and abiotic fac-
tors to plant microbiome variation. PLOS Biology, 14(1), e1002352. al.pbio.1002352
Anderson, M. J. (2001). A new method for non-parametric multivariate
analysis of variance. Austral Ecolog y, 26, 32–46. htt ps://
Bahram, M., Kõljalg, U., Cour ty, P.-E., Diédhiou, A. G., Kjøller, R.,
Põlme, S., … Tedersoo, L. (2013). The distance decay of similarity
in communities of ectomycorrhizal fungi in different ecosystems
and scales. The Journal of Ecology, 101(5), 1335–1344. https://doi.
org /10.1111/1365-2745.12120
Banerjee, S., Kirkby, C. A., Schmutter, D., Bissett, A., Kirkegaard, J. A.,
& Richardson, A. E. (2016). Network analysis reveals functional
redundancy and keystone taxa amongst bacterial and fungal com-
munities during organic matter decomposition in an arable soil. Soil
Biology & Biochemistry, 97, 188–198.
Banerjee, S., Schlaeppi, K., & van der Heijden, M. G. A . (2018). Keystone
taxa as drivers of microbiome structure and functioning. Nature
Reviews Microbiology, 16(9), 567–576.
Bascompte, J., & Jordano, P. (2007). Plant-animal mutualistic networks:
The architecture of biodiversity. Annual Review of Ecology Evolution
and Systematics, 38, 567–593. ev.ecols
Bascompte, J., & Jordano, P. (2013). Mutualistic networks. Princeton,
CAPrinceton Universit y Press.
Bastin, J.-F., Finegold, Y., Garcia, C., Mollicone, D., Rezende, M., Routh,
D., … Crowther, T. W. (2019). The global tree restoration poten-
tial. Science, 365(6448), 76–79.
Berry, D., & Widder, S. (2014). Deciphering microbial interactions and
detecting keystone species with co-occurrence networks. Frontiers
in Microbiology, 5, 219.
Bever, J. D., Schultz, P. A., Pringle, A., & Morton, J. B. (2001). Arbuscular
mycorrhizal fungi: More diverse than meet s the eye, and the eco-
logical tale of why the high diversity of ecologically distinct species
of arbuscular mycorrhizal fungi within a single community has broad
implications for plant ecology. BioScience, 51 (11), 923–931. https://[0923:AMFMD T]2.0.CO;2
Bivand, R. S., Pebesma, E., & Gómez-Rubio, V. (2013). Applied spa-
tial data analysis with R . New York, NY: Springer. ht tps://doi.
Bray, J. R., & Curtis, J. T. (1957). An ordination of the upland forest com-
munities of southern Wisconsin. Ecological Monographs, 27(4), 325–
Brundrett, M. C., & Tedersoo, L. (2018). Evolutionary history of mycor-
rhizal symbioses and global host plant diversity. New Phytologist,
220(4), 1108–1115. .1111/nph.14976
Callahan, B. J., McMurdie, P. J., Rosen, M. J., Han, A. W., Johnson, A.
J. A ., & Holmes, S. P. (2016). DADA2: High-resolution sample infer-
ence from Illumina amplicon data. Nature Methods, 13(7), 581–583.
Camacho, C., Coulouris, G., Avagyan, V., Ma, N., Papadopoulos, J., Bealer,
K., & Madden, T. L. (2009). BLAST+: Architecture and applications. BMC
Bioinformatics, 10, 421.
Cam pbel l, C., Yang, S. , Shea , K., & Albe rt , R. (201 2). Top olo gy of pl ant-p ol-
linator networks that are vulnerable to collapse from species ex-
tinction. Physical Review. E, 86 (2), 021924.
PhysR evE.86.021924
Chagnon, P.-L., Bradley, R. L., Maherali, H., & Klironomos, J. N. (2013). A
trait-based framework to understand life history of mycorrhizal fungi.
Trends in Plant Science, 18 (9), 484–491.
tplan ts.2013.05.001
Choi, J., Summers, W., & Paszkowski, U. (2018). Mechanisms underlying
establishment of arbuscular mycorrhizal symbioses. Annual Review
of Phytopathology, 56(1), 135–160.
Crouzeilles, R ., Curran, M., Ferreira, M. S., Lindenmayer, D. B., Grelle, C.
E. V., & Rey Benayas, J. M. (2016). A global meta-analysis on the eco-
logical drivers of forest restoration success. Nature Communications,
7(1), 1–8. s11666
Csardi, G., & Nepusz, T. (2006). The igraph software package for com-
plex network research. InterJournal, Complex Systems. 1695–1704.
Davison, J., Moora, M., Opik, M., Adholeya, A., Ainsaar, L., Ba, A., …
Zobel, M. (2015). Global assessment of arbuscular mycorrhizal fun-
gus diversity reveals very low endemism. Science, 349 (6251), 970–
973. ce.aab1161
Davison, J., Öpik, M., Zobel, M., Vasar, M., Metsis, M., & Moora, M.
(2012). Communities of arbuscular mycorrhizal fungi detected in for-
est soil are spatially heterogeneous but do not var y throughout the
growing season. PLoS One, 7(8), e41938.
journ al.pone.0041938
Dickie, I. A ., Koide, R . T., & Steiner, K. C. (2002). Influence of established
trees on mycorrhizas, nutrition, and growth of Quercus rubra seed-
lings. Ecological Monographs, 72, 505–521. ht tps://
012-9615(2002)072[0505:ioeto m];2
Dray, S., & Dufour, A.-B. (2007). The ade4 Package: Implementing the
duality diagram for ecologists. Journal of Statistical Software, Articles,
22(4), 1–20. jss.v022.i04
Duhamel, M., Wan, J., Bogar, L. M., Segnitz, R. M., Duncritts, N. E., &
Peay, K . G. (2018). Plant selection initiates alternative successional
trajectories in the soil microbial community after disturbance.
Ecological Monographs, 83, e01367.
Dumbrell, A . J., Ashton, P. D., Aziz, N., Feng, G. U., Nelson, M., Dytham,
C., … Helgason, T. (2011). Distinct seasonal assemblages of ar-
buscular mycorrhizal fungi revealed by massively parallel pyrose-
quencing. The New Phytologist, 190(3), 794–804. https://doi.
org /10.1111/j .1469-8137.2010.036 36.x
Egan, C. P., Rummel, A., Kokkoris, V., Klironomos, J., Lekberg, Y., & Hart,
M. (2018). Using mock communities of arbuscular mycorrhizal fungi
to evaluate fidelity associated with Illumina sequencing. Fungal
Ecology, 33, 52–64.
Freeman, L. C. (1978). Centrality in social networks concep-
tual clarification. Social Networks, 1(3), 215–239. https://doi.
org/10.1016/0 378-8733 (78)90 021-7
Glassman, S. I., Wang, I. J., & Bruns, T. D. (2017). Environmental filter-
ing by pH and soil nutrients drives community assembly in fungi at
fine spatial scales. Molecular Ecology, 26(24), 6960–6973. ht tps://doi.
org /10.1111/me c.14414
Godefroid, S., Piazza, C., Rossi, G., Buord, S., Stevens, A.-D., Aguraiuja,
R., Vanderborght, T. (2011). How successful are plant species re-
introductions? Biological Conservation, 144(2), 672–682. https://doi.
org/10.1016/j.biocon. 2010 .10.003
Gräler, B., Pebesma, E., & Heuvelink, G. (2016). Spatio-temporal
Interpolation using gstat. The R Journal, 8(1), 20 4–218. https://doi.
org/10.32614/ RJ-2016-014
Haddad, N. M., Brudvig, L. A., Clobert, J., Davies, K. F., Gonzalez, A., Holt,
R. D., … Townshend, J. R. (2015). Habitat fragmentation and its last-
ing impact on Earth’s ecosystems. Science Advances, 1(2), e1500052.
Hart, M. M., & Reader, R. J. (20 02). Host plant benefit from association
with arbuscular mycorrhizal fungi: Variation due to differences in size
of mycelium. Biology and Fertility of Soils, 36(5), 357–366. https://doi.
Hart, M. M., Zaitsoff, P. D., van der Heyde, M., & Pither, J. (2016). Testing
life history and trait-based predictions of AM fungal community
assembly. Pedobiologia, 59(4), 203–213. ht tps://
Harter, D. E. V., Irl, S. D. H., Seo, B., Steinbauer, M. J., Gillespie, R.,
Triantis, K. A. , … Bei erkuh nlein, C. (2 01 5). Impac ts of globa l climate
change on the floras of oceanic islands – Projections, implications
and current knowledge. Perspec tives in Pla nt Ecology, Evolution
and Systematics, 17(2), 160–183.
Hausmann, N. T., & Hawkes, C. V. (2010). Order of plant host establish-
ment alters the composition of arbuscular mycorrhizal communities.
Ecology, 91(8), 2333–2343.
Heneghan, L., Miller, S. P., Baer, S., Callaham, M. A., Montgomery, J.,
Pavao-Zuckerman, M., … Richardson, S. (2008). Integrating soil eco-
logical knowledge into restoration management. Res toration Ecology,
16(4), 608–617.
Horner-Devine, M. C ., Claire Horner-Devine, M., Lage, M., Hughes, J. B.,
& Bohannan, B. J. M. (2004). A taxa–area relationship for bacteria.
Nature, 432, 750–753. e 03073
House, G. L., & Bever, J. D. (2018). Disturbance reduces the differentia-
tion of mycorrhizal fungal communities in grasslands along a precip-
itation gradient. Ecological Applications, 28(3), 736–748. https://doi.
Janoušková, M., Püschel, D., Hujslová, M., Slavíková, R., & Jansa, J.
(2015). Quantification of arbuscular mycorrhizal fungal DNA in roots:
How impor tant is material preser vation? Mycorrhiza, 25(3), 205–214.
Jeffrey, J., & Horiuchi, B. (2003). Tree planting at hakalau forest national
wildlife refuge—The right tool for the right stock type. Native Plants
Journal, 4(1), 30–31.
Johnson, N. C., Zak, D. R., Tilman, D., & Pfleger, F. L . (1991). Dynamics
of vesicular-arbuscular mycorrhizae during old field succession.
Oecologia, 86(3), 349–358. 17600
Keenan, R. J., Reams, G. A., Achard, F., de Freitas, J. V., Grainger, A.,
& Lindquist, E. (2015). Dynamics of global forest area: Results
from the FAO global forest resources assessment 2015. Fores t
Ecology and Management, 3 52, 9–20.
Kiers, E. T., Lovelock, C. E., Krueger, E. L., & Herre, E. A. (2000).
Differential effects of tropical arbuscular mycorrhizal fungal inoc-
ula on root colonization and tree seedling growth: Implications for
tropical forest diversity. Ecology Letters, 3(2), 106–113. https://doi.
Kivlin, S. N., & Hawkes, C . V. (2011). Differentiating between ef fects of
invasion and diversity: Impacts of aboveground plant communities
on belowground fungal communities. New Phytologist, 189(2), 526–
Klironomos, J. N. (2002). Feedback with soil biota contributes to plant
rarity and invasiveness in communities. Nature, 417(6884), 67–70.
htt ps:// /10.1038/417067a
Koziol, L., & Bever, J. D. (2017). The missing link in grassland restoration:
Arbuscular mycorrhizal fungi inoculation increases plant diversity
and accelerates succession. The Journal of Applied Ecology, 54(5),
1301–1309. https://doi .org/10.1111/1365-2664.128 43
Kurtz, Z., Mueller, C., Miraldi, E., & Bonneau, R. (2019). SpiecEasi: Sparse
inverse covariance for ecological statistical inference. R package ver-
sion 1.0.7.
Lamb, D. (2005). Restoration of degraded tropical forest landscapes.
Science, 310(5754), 1628–1632.
ce.11117 73
Lankau, R. A., Bauer, J. T., Anderson, M. R., & Anderson, R. C. (2014).
Long-term legacies and partial recovery of mycorrhizal communities
after invasive plant removal. Biological Invasions, 16(9), 1979–1990.
Lee, J., Lee, S., & Young, J. P. W. (2008). Improved PCR primers
for the detection and identification of arbuscular mycorrhizal
fungi. FEMS Microbiology Ecology, 65(2), 339–349. https://doi.
org /10.1111/j .1574-6941.2008.0 0531. x
Lekberg, Y., Gibbons, S. M., Rosendahl, S., & Ramsey, P. W. (2013). Severe
plant invasions can increase mycorrhizal fungal abundance and di-
ver si ty. ISME Journal, 7(7 ), 1424–1433. ht tps://doi.o rg /10.103 8/
Lenoir, I., Fontaine, J., & Lounès-Hadj Sahraoui, A. (2016). Arbuscular
mycorrhizal fungal responses to abiotic stresses: A review.
Phytochemistry, 123, 4–15.
Liang, J., Crowther, T. W., Picard, N., Wiser, S., Zhou, M., Alberti, G., …
Reich, P. B. (2016). Positive biodiversity-productivity relationship
predominant in global forests. Science, 354(6309), aaf8957–aaf8957. ce.aaf8957
Liu, H., Roeder, K., & Wasserman, L. (2010). Stability approach to reg-
ularization selection (StARS) for high dimensional graphical mod-
els. Proceedings of the 23rd International Conference on Neural
Information Processing Systems - Volume 2, 1432–1440. Retrieved
Maltz, M. R., & Treseder, K. K. (2015). Sources of inocula influence my-
cor rhizal colonization of pla nt s in res toration projec ts: A meta-analy-
sis: Mycorrhizal inoculation in restoration. Restoration Ecolog y, 23(5),
625–634. htt ps://doi.or g/10.1111/r ec .12231
Maron, M., Hobbs, R. J., Moilanen, A., Matthews, J. W., Christie,
K., Gardner, T. A., … McAlpine, C. A. (2012). Faustian bargains?
Restoration realities in the context of biodiversity offset policies.
Biological Conservation, 155, 141–148 . ht tps://
Martínez-García, L. B., Richardson, S. J., Tylianakis, J. M., Peltzer, D. A., &
Dickie, I. A. (2015). Host identity is a dominant driver of mycorrhizal
fungal community composition during ecosystem development. New
Phytologist, 205(4), 1565–1576.
Meinshausen, N., & Bühlmann, P. (2006). High-dimensional graphs and
variable selection with the Lasso. Annals of Statistics, 34(3), 1436–
1462. 36060 00000281
Meyer, K. M., Memiaghe, H., Korte, L., Kenfack, D., Alonso, A., &
Bohannan, B. J. M. (2018). Why do microbes exhibit weak biogeo-
graphic patterns? The ISME Journal, 12(6), 1404–1413. https://doi.
org/10.103 8/s41396 -018-0103-3
Müller, C. L., Bonneau, R., & Kurtz, Z. (2016). Generalized stability
approach for regularized graphical models. arXiv. 1605.07072.
Retrieved from
Mummey, D. L., & Rillig, M. C. (2008). Spatial characterization of arbus-
cular mycorrhizal fungal molecular diversity at the submetre scale in
a temperate grassland. FEMS Microbiology Ecology, 64(2), 260–270.
https://doi.or g/10.1111/ j.1574-6941. 20 08.00 475.x
Nellemann, C., & Corcoran, E. (2010). Dead planet, living planet: bio-
diversity and ecosystem restoration for sustainable develop-
ment. Retrieved from https://www.cabdi rect/abstr
act/20103 196279.
Neuenkamp, L., Prober, S. M., Price, J. N., Zobel, M., & Standish, R. J.
(2019). Benefits of mycorrhizal inoculation to ecological restoration
depend on plant functional type, restoration context and time. Fungal
Ecology, 40, 140–149.
Oehl, F., Laczko, E., Bogenrieder, A., Stahr, K., Bösch, R., van der Heijden,
M., & Sieverding, E. (2010). Soil type and land use intensit y determine
the composition of arbuscular mycorrhizal fungal communities. Soil
Biology and Biochemistry, 42(5), 724–738.
soilb io.2010.01.006
Ohsowski, B. M., Zaitsoff, P. D., Opik, M., & Hart, M. M. (2014). Where
the wild things are: Looking for uncultured Glomeromycota.
The New Phytologist, 204(1), 171–179. /10.1111/
   WALL e t AL.
Oksanen, J. F., Blanchet, G., Friendly, M., Kindt, R., Legendre, P., McGlinn,
D., …Wagner, H. (2019). veg an: Commu nity Ecol og y Package. R pack-
age version 2.5-6. Retrieved from https://CRAN.R-proje
p a c k a g e =vegan
Öpik, M., Metsis, M., Daniell, T. J., Zobel, M., & Moora, M. (20 09).
Large-scale parallel 454 sequencing reveals host ecological
group specificity of arbuscular mycorrhizal fungi in a boreon-
emoral forest. New Phytologist, 184 (2), 424–437. https://doi.
org /10.1111/j .1469-8137.20 09.02920.x
Öp ik, M. , Moor a, M., Li ira, J. , & Zob el, M. (20 06) . Comp osi t ion of root-c ol-
onizing arbuscular mycorrhizal fungal communities in different eco-
systems around the globe: Arbuscular mycorrhizal fungal communi-
ties around the globe. The Jou rnal of Ecology, 94(4), 778–790. https:// 20 06.01136 .x
Paine, R. T. (1969). The Pisaster-Tegula interaction: Prey patches, pred-
ator food preference, and intertidal community structure. Ecology,
50(6), 950–961.
Paxton, E. H., Yelenik, S. G., Borneman, T. E., Rose, E. T., Camp, R. J.,
& Kendall, S. J. (2018). Rapid colonization of a Hawaiian restoration
forest by a diverse avian community. Restoration Ecology, 26 (1), 165–
173. https://doi.or g/10.1111/rec .1254 0
Peay, K. G., Kennedy, P. G., & Bruns, T. D. (2011). Rethinking ectomycor-
rhiz al succes si on: Are roo t densi ty an d hy phal ex plo ration typ es driv -
ers of spatial and temporal zonation? Fungal Ecology, 4(3), 233–240.
Pebesma, E. J. (2004). Multivariable geostatistics in S: The gstat package.
Computers & Geosciences, 30(7), 683–691.
Perring, M. P., Standish, R. J., Price, J. N., Craig, M. D., Erickson, T. E.,
Ruthrof, K. X., … Hobbs, R . J. (2015). Advances in restoration ecol-
ogy: Rising to the challenges of the coming decades. Ecosphere, 6(8),
Poisot, T., Stouffer, D. B., & Gravel, D. (2015). Beyond species: Why
ecological interaction networks vary through space and time. Oikos,
124(3), 243–251.
R Core Team (2019). R: A Language and Environment for Statistical
Computing. R Foundation for Statistical Computing, ht tps://w w-
Rehm, E. M., Thomas, M. K., Yelenik, S. G., Bouck, D. L., & D’Antonio,
C. M. (2019). Bryophyte abundance, composition and importance to
woody plant recruitment in natural and restoration forests. Forest
Ecology and Management, 444, 405–413.
Smith, S. E., & Read, D. J. (2008). Mycorrhizal symbiosis, 3rd edn.
Amsterdam, The Netherlands; Boston, MAAcademic Press.
Steidinger, B. S., Crowther, T. W., Liang, J., Van Nuland, M. E., Werner,
G. D. A., Reich, P. B., … Peay, K. G. (2019). Climatic controls of
decomposition drive the global biogeography of forest-tree sym-
bioses. Nature, 569( 7756), 404–4 08.
Suding, K. N. (2011). Toward an era of restoration in ecology: Successes,
failures, and opportunities ahead. Annual Revi ew of Ecology, Evolution,
and Systematics, 42(1), 465–487. ev-
ec ols ys-102710 -14511 5
Tellen, V. A., & Yerima, B. P. K. (2018). Effects of land use change on
soil physicochemical properties in selec ted areas in the North West
region of Cameroon. Environmental Systems Research, 7(1), 3. https://
Tipton, L., Müller, C. L., Kurtz, Z. D., Huang, L., Kleerup, E., Morris, A ., …
Ghedin, E. (2018). Fungi stabilize connectivity in the lung and skin
microbial ecosystems. Microbiome, 6(1), 12 .
Tulloch, A. I. T., Chadès, I., Dujardin, Y., Westgate, M. J., Lane, P. W., &
Lindenmayer, D. (2016). Dynamic species co-occurrence networks
require dynamic biodiversity surrogates. Ecography, 39(12), 1185–
1196. htt ps:// .1111/ecog .02143
Tulloch , A . I. T., Cha dès, I., & Li nde nmayer, D. B. (2 018). Sp ecies co-occur-
rence analysis predicts management outcomes for multiple threats.
Nature Ecology & Evolution, 2(3), 465–474.
s41 559- 017-0457-3
Tylianakis, J. M., Laliberte, E., Nielsen, A., & Bascompte, J. (2010).
Conservation of species interaction networks. Biological
Conservation, 143(10), 2270–2279.
van der Heijden, M. G. A., Klironomos, J. N., Ursic, M., Moutoglis, P.,
Streitwolf-Engel, R., Boller, T., … Sanders, I. R. (1998). Mycorrhizal
fungal diversity determines plant biodiversity, ecosystem vari-
ability and productivity. Nature, 396(6706), 69–72. https://doi.
org/10.103 8/23932
van der Heijden, M. G. A., Martin, F. M., Selosse, M.-A., & Sanders, I.
R. (2015). Mycorrhizal ecology and evolution: The past, the pres-
ent, and the future. New Phytologist, 205(4), 14 06–1423. https://doi.
org /10.1111/np h.13288
Wall, C. (2020). cbwall/Hakalau-metacommunity-AMF: Post-
disturbance reassembly of arbuscular mycorrhizal fungal com-
munities (Version vpub). Zenodo.
Wall, C. B., & Stevens, K. J. (2014). Assessing wetland mitigation ef forts
using standing vegetation and seed bank community structure in
neighboring natural and compensatory wetlands in north-central
Tex as. Wetlands Ecology and Management, 23 (2), 149–166. https://
Wardle, D. A., Bardgett, R . D., Klironomos, J. N., Setälä, H., van
der Putten, W. H., & Wall, D. H. (2004). Ecological linkages be-
tween aboveground and belowground biota. Science, 304(5677),
Weerasinghe, M., Ashton, M. S., Hooper, E. R., & Singhakumara, B. M.
P. (2019). Floristics of soil seed banks on agricultural and disturbed
land cleared of tropical forests: Seed banks in disturbed tropical
lands. Restoration Ecology, 27(1), 138–147. .1111/
rec .12711
Welch, B. L. (1947). The generalization of “student”s’ problem when sev-
eral different population variances are involved. Biometrika, 34(12),
28–35. t/34.1-2.28
Wubs, E. R. J., van der Putten, W. H., Bosch, M., & Bezemer, T. M. (2016).
Soil inoculation steers restoration of terrestrial ecosystems. Nature
Plants, 2, 16107. ts.2016.107
Yelenik, S. G. (2017). Linking dominant Hawaiian tree species to under-
story development in recovering pastures via impact s on soils and
litter: Canopy tree effects on understory regeneration. Restoration
Ecology, 25(1), 42–52.
Yelenik, S. G., DiManno, N., & D’Antonio, C. M. (2015). Evaluating nurse
plants for restoring native woody species to degraded subtropi-
cal woodlands. Ecology and Evol ution, 5(2), 300–313. https://doi.
Additional supporting information may be found online in the
Supporting Information section.
How to cite this article: Wall CB, Egan CP, Swift SIO, Hynson
NA. Three decades post-reforestation has not led to the
reassembly of arbuscular mycorrhizal fungal communities
associated with remnant primary forests. Mol Ecol.
2020;00:1–14. .1111/mec .15624
... Previous work has shown that generalist AM fungi are favoured in disturbed habitats compared to those found with specific hosts (Helgason et al., 2007;Bennett et al., 2013), but that host-specific AM fungi may be necessary to restore latesuccessional plant communities in degraded ecosystems (Koziol & Bever, 2017). However, the presence of these host-specific fungi is often diminished by disturbance and the introduction of novel invasive plants (Moora et al., 2011) and may be especially difficult to maintain if restoration sites are distant in time or space from their source pools (Wall et al., 2020). Biogeography is a particularly important consideration for AM fungal community dynamics as many host-specific taxa are also generally considered poor dispersers (Egan et al., 2014), while a few generalists are globally ubiquitous (Davison et al., 2015). ...
... Roots were sampled from up to eight individuals of each host species by tracing from large to fine roots for a total of 569 root samples across 12 plots. Soil was sampled by taking the top 10 cm of mineral soil directly adjacent to where roots were sampled for a total of 583 soil samples including samples taken for the soil chemistry analyses in Wall et al. (2020). If eight individuals of each host species were not present in a given plot, we increased sampling intensity so that each plot yielded the same number of root and soil samples per host. ...
... These types of disturbances can lead to significant alterations in edaphic factors such as soil chemistry (Wubs et al., 2016). In a previous study, Wall et al. (2020) found that despite the similar climate, slope, aspect, elevation and soil parent material, the soil chemical properties of reforested (restored) pasture sites differed from remnant forest patches and that these differences affect soil AM fungal community membership. In the present study, our aim was to examine the interactions between soil pools of AM fungi, geography and habitat type on host-associated communities. ...
Habitat restoration may depend on the recovery of plant microbial symbionts such as arbuscular mycorrhizal (AM) fungi, but this requires a better understanding of the rules that govern their community assembly. We examined the interactions of soil and host‐associated AM fungal communities between remnant and restored patches of subtropical montane forests. While AM fungal richness did not differ between habitat types, community membership did and was influenced by geography, habitat and host. These differences were largely driven by rare host‐specific AM fungi that displayed near‐complete turnover between forest types, while core AM fungal taxa were highly abundant and ubiquitous. Remnant forest bipartite networks were more specialized and hosts more specific than in the restored forest. Host‐associated AM fungal communities nested within soil communities in both habitats, but only significantly so in the restored forest. Our results provide evidence that restored and remnant forests harbor the same core fungal symbionts, while rare host‐specific taxa differ, and that geography, host identity and taxonomic resolution strongly affect the observed distribution patterns of these fungi. We suggest that host‐specific interactions with AM fungi as well as spatial processes should be explicitly considered to effectively reestablish target host and symbiont communities.
... As the global deforestation rate has increased sharply over the last 100 years (Wall et al., 2020), innumerable natural forests were converted to agricultural lands. Further, agricultural land is increasingly being abandoned in recent decades, which has emerged as a global issue (Khorchani et al., 2022). ...
Full-text available
Natural revegetation has been widely confirmed to be an effective strategy for the restoration of degraded lands, particularly in terms of rehabilitating ecosystem productivity and soil nutrients. Yet the mechanisms of how natural revegetation influences the variabilities and drivers of soil residing fungal communities, and its downstream effects on ecosystem nutrient cycling are not well understood. For this study, we investigated changes in soil fungal communities along with ~160 years of natural revegetation in the Loess Plateau of China, employing Illumina MiSeq DNA sequencing analyses. Our results revealed that the soil fungal abundance was greatly enhanced during the later stages of revegetation. As revegetation progresses, soil fungal richness appeared first to rise and then decline at the climax Quercus liaotungensis forest stage. The fungal Shannon and Simpson diversity indexes were the lowest and highest at the climax forest stage among revegetation stages, respectively. Principal component analysis, Bray–Curtis similarity indices, and FUNGuild function prediction suggested that the composition, trophic modes, and functional groups for soil fungal communities gradually shifted along with natural revegetation. Specifically, the relative abundances of Basidiomycota, Agaricomycetes , Eurotiomycetes , and ectomycorrhizal fungi progressively increased, while that of Ascomycota, Sordariomycetes , Dothideomycetes , Tremellomycetes , saprotrophic, pathotrophic, arbuscular mycorrhizal fungi, and endophyte fungi gradually decreased along with natural revegetation, respectively. The most enriched members of Basidiomycota (e.g., Agaricomycetes , Agaricales , Cortinariaceae , Cortinarius , Sebacinales , Sebacinaceae , Tricholomataceae , Tricholoma , Russulales, and Russulaceae ) were found at the climax forest stage. As important carbon (C) sources, the most enriched symbiotic fungi (particularly ectomycorrhizal fungi containing more recalcitrant compounds) can promote organic C and nitrogen (N) accumulation in soils of climax forest. However, the most abundant of saprotrophic fungi in the early stages of revegetation decreased soil organic C and N accumulation by expediting the decomposition of soil organic matter. Our results suggest that natural revegetation can effectively restore soil fungal abundance, and modify soil fungal diversity, community composition, trophic modes, and functional groups by altering plant properties (e.g., plant species richness, diversity, evenness, litter quantity and quality), quantity and quality of soil nutrient substrates, soil moisture and pH. These changes in soil fungal communities, particularly their trophic modes and functional groups along with natural revegetation, impact the accumulation and decomposition of soil C and N and potentially affect ecosystem C and N cycling in the Loess Plateau of China.
... A neutral view of assembly considers interactions among individuals are essentially random and it is relative species abundances that determine network patterns (Canard et al., 2012). However, deterministic factors such as the mismatch of environmental tolerances of plants and fungi (Arraiano-Castilho et al., 2021) and human effects (Wall et al., 2020) may regulate the establishment of interaction relationships by imposing spatial constraints on encounter probabilities. For instance, structures of aboveground networks are best predicted by a combination of species abundance and other deterministic factors such as spatial overlap (Vázquez et al., 2009;Sáyago et al., 2013), phylogeny (Cagnolo et al., 2011) and trait matching (Olito and Fox, 2015). ...
Full-text available
Plant and root fungal interactions are among the most important belowground ecological interactions, however, the mechanisms underlying pairwise interactions and network patterns of rhizosphere fungi and host plants remain unknown. We tested whether neutral process or spatial constraints individually or jointly best explained quantitative plant–ectomycorrhizal fungal network assembly in a subtropical forest in southern China. Results showed that the observed plant–ectomycorrhizal fungal network had low connectivity, high interaction evenness, and an intermediate level of specialization, with nestedness and modularity both greater than random expectation. Incorporating information on the relative abundance and spatial overlap of plants and fungi well predicted network nestedness and connectance, but not necessarily explained other network metrics such as specificity. Spatial overlap better predicted pairwise species interactions of plants and ectomycorrhizal fungi than species abundance or a combination of species abundance and spatial overlap. There was a significant phylogenetic signal on species degree and interaction strength for ectomycorrhizal fungal but not for plant species. Our study suggests that neutral processes (species abundance matching) and niche/dispersal-related processes (implied by spatial overlap and phylogeny) jointly drive the shaping of a plant-ectomycorrhizal fungal network.
... The OTUs in the present study were assigned to its closest related VT in MaarjAM database if sequence similarity ≥97%, and a BLAST e-value ≤1e-50, query coverage ≥80% (Öpik et al., 2010;Merckx et al., 2012). Since several OTUs could not be assigned to existing VTXs using the criteria, the closest VTXs were identified based on the relaxed criteria using a sequence similarity of ≥90%, and a BLAST e-value of ≤1e-50, query coverage ≥80% (Wall et al., 2020). We note that both OTUs in our study and VTXs were clustered at a 97% sequence similarity threshold for partial SSU rDNA (Öpik et al., 2010). ...
Full-text available
Plants producing dust seeds often meet their carbon demands by exploiting fungi at the seedling stage. This germination strategy (i.e., mycoheterotrophic germination) has been investigated among orchidaceous and ericaceous plants exploiting Ascomycota or Basidiomycota. Although several other angiosperm lineages have evolved fully mycoheterotrophic relationships with Glomeromycota, the fungal identities involved in mycoheterotrophic germination remain largely unknown. Here, we conducted in situ seed baiting and high‐throughput DNA barcoding to identify mycobionts associated with seedlings of Burmannia championii (Burmanniaceae: Dioscoreales) and Sciaphila megastyla (Triuridaceae: Pandanales), which have independently evolved full mycoheterotrophy. Subsequently, we revealed that both seedlings and adults in B. championii and S. megastyla predominantly associate with Glomeraceae. However, mycorrhizal communities are somewhat distinct between seedling and adult stages, particularly in S. megastyla. Notably, the dissimilarity of mycorrhizal communities between S. megastyla adult samples and S. megastyla seedling samples is significantly higher than that between B. championi adult samples and S. megastyla adult samples, based on some indices. This pattern is possibly due to both mycorrhizal shifts during ontogenetic development and convergent recruitment of cheating‐susceptible fungi. The extensive fungal overlap in two unrelated mycoheterotrophic plants indicates that both species convergently exploit specific AM fungal phylotypes. This article is protected by copyright. All rights reserved.
... carbon) while building less costly roots (due to the reduced cell wall construction required for rhizonodes; Dickie & Holdaway, 2011;Kiers et al., 2011). AMF have particularly slow rates of recovery from forest disturbance, even after the employment of restoration practices (Wall et al., 2020); thus, compartmentalisation may play a critical role in the maintenance of redwood forests. Furthermore, the common presence of similar nodule-like structures on both extinct and extant AMF-associated tree species ( Figure S7) provides an exciting opportunity to explore the role of these structures in a broad range of hosts from both ecological and evolutionary perspectives (Beslow et al., 1970;Dickie & Holdaway, 2011;Duhoux et al., 2001;Grand, 1969;Russell et al., 2002). ...
Full-text available
Root‐associated fungal communities modify the climatic niches and even the competitive ability of their hosts, yet how the different components of the root microbiome are modified by habitat loss remains a key knowledge gap. Using principles of landscape ecology, we tested how free‐living versus host‐associated microbes differ in their response to landscape heterogeneity. Further, we explore how compartmentalisation of microbes into specialised root structures filters for key fungal symbionts. Our study demonstrates that free‐living fungal community structure correlates with landscape heterogeneity, but that host‐associated fungal communities depart from these patterns. Specifically, biotic filtering in roots, especially via compartmentalisation within specialised root structures, decouples the biogeographic patterns of host‐associated fungal communities from the soil community. In this way, even as habitat loss and fragmentation threaten fungal diversity in the soils, plant hosts exert biotic controls to ensure associations with critical mutualists, helping to preserve the root mycobiome. Using the principles of landscape ecology, we tested how free‐living versus host‐associated microbes might differ in their response to landscape heterogeneity. Furthermore, we describe how compartmentalisation of microbes into specialised root structures filters for key fungal symbionts (arbuscular mycorrhizal fungi). Even as habitat loss and fragmentation threaten fungal diversity in the soils, we found that plant hosts exert biotic controls to ensure associations with critical mutualists, helping to preserve the root mycobiome. Photo courtesy of Anthony Ambrose.
Full-text available
The restoration of trees remains among the most effective strategies for climate change mitigation. We mapped the global potential tree coverage to show that 4.4 billion hectares of canopy cover could exist under the current climate. Excluding existing trees and agricultural and urban areas, we found that there is room for an extra 0.9 billion hectares of canopy cover, which could store 205 gigatonnes of carbon in areas that would naturally support woodlands and forests. This highlights global tree restoration as our most effective climate change solution to date. However, climate change will alter this potential tree coverage. We estimate that if we cannot deviate from the current trajectory, the global potential canopy cover may shrink by ~223 million hectares by 2050, with the vast majority of losses occurring in the tropics. Our results highlight the opportunity of climate change mitigation through global tree restoration but also the urgent need for action.
Full-text available
A spatially explicit global map of tree symbioses with nitrogen-fixing bacteria and mycorrhizal fungi reveals that climate variables are the primary drivers of the distribution of different types of symbiosis.
Full-text available
Analysis of patterns in the distribution of taxa can provide important insights into ecological and evolutionary processes. Microbial biogeographic patterns almost always appear to be weaker than those reported for plant and animal taxa. It is as yet unclear why this is the case. Some argue that microbial diversity scales differently over space because microbial taxa are fundamentally different in their abundance, longevity and dispersal abilities. Others have argued that differences in scaling are an artifact of how we assess microbial biogeography, driven, for example, by differences in taxonomic resolution, spatial scale, sampling effort or community activity/dormancy. We tested these alternative explanations by comparing bacterial biogeographic patterns in soil to those of trees found in a forest in Gabon. Altering taxonomic resolution, excluding inactive individuals, or adjusting for differences in spatial scale were insufficient to change the rate of microbial taxonomic turnover. In contrast, we account for the differences in spatial turnover between these groups by equalizing sampling extent. Our results suggest that spatial scaling differences between microbial and plant diversity are likely not due to fundamental differences in biology, and that sampling extent should be taken into account when comparing the biogeographic patterns of microorganisms and larger organisms.
Full-text available
Characterization of arbuscular mycorrhizal (AM) fungal communities increasingly relies on high throughput sequencing (HTS) datasets, but whether sequence data accurately depict AM fungal communities is unknown. We sequenced mock communities of 16 AM fungal morphospecies from six families that varied in relative abundance. To assess sequence variation within fungal individuals, we sequenced single spores of Rhizophagus irregularis. We observed that the relative abundance and taxonomic identity of operational taxonomic units (OTUs) within AM fungal families closely matched expected values, but this decreased at lower taxonomic levels. Multiple OTUs were observed within single spores, suggesting that using OTUs to estimate species richness may inflate richness estimates and reflect sequence variation within individuals. While HTS may introduce some bias in relative abundance estimates and taxonomic identification, we observed high consistency among replicate samples from the same mock community, indicating that these data can inform ecological patterns.
Restoration of tropical forests can lead to enhanced ecosystem services and increases in native biodiversity. Bryophytes may be an integral part of the forest restoration process and can serve a critical role in forest functioning. However, the recovery of bryophytes and their ability to facilitate woody plant establishment during restoration remains poorly studied, especially in the tropics. We investigate how bryophyte abundance and community composition, as well as woody plant seedling associations with bryophyte mats and other ground cover types change from under the canopy of intact forest to under trees in restoration corridor plantings in Hawaii. Restoration corridors consisted of corridors of koa (Acacia koa)trees that were planted roughly 30 years ago. Some corridors were planted around remnant ʻōhiʻa trees (Metrosideros polymorpha)that can be several hundred years old. We sampled under ʻōhiʻa in intact forest and both koa and ʻōhiʻa trees in restoration corridors. In restoration corridors, bryophyte abundance was low relative to intact forest and species diversity was a subset of that found in intact forest despite restoration corridors being several decades old. Seedlings strongly associated with bryophytes across all habitats suggesting that bryophytes may significantly enhance forest seedling establishment when present in restoration corridors. Other ground cover types like woody litter and nurse logs also had a positive association with forest seedlings but were rare in restoration corridors. Grass remained a dominant ground cover type in restoration corridors under koa and remnant ʻōhiʻa trees and only a single seedling was ever found growing in this grass. Enhancing bryophyte growth and recovery within restoration plantings through the reduction of grass cover could facilitate native plant establishment.
Because interactions between plants and microbial organisms can influence species diversity and rates of nutrient cycling, how plants shape microbial communities is fundamental to understanding the structure of ecosystems. Despite this, the spatial and temporal scales over which plants influence microbial communities is poorly understood, particularly whether past abiotic or biotic legacies strongly constrain microbial community development. We examined biogeochemical cycling and microbial community structure in a coastal landscape where historical patterns of vegetation transition after a large fire in 1995 are well known, allowing us to account for past abiotic and biotic conditions. We found that alternative states in microbial community structure and ecosystem processes emerged under different plant species, regardless of past conditions. Greenhouse studies further demonstrated that these differences arise from direct plant selection of microbes, with selection stronger in roots compared with soils, especially for bacteria. Correlation of microbial community structure with seedling growth rates was also stronger for fungi compared to bacteria. Despite these effects, minimal overlap between seedling and field microbial communities indicates that the effects of initial plant selection are not stable, rather plant selection initiated alternative successional trajectories after the fire. Using data from a guild where we have abundant natural history information ‐ ectomycorrhizal fungi ‐ we show that greenhouse communities are dominated by ruderal taxa that are also common in the field after the fire, and that these ruderal fungi strongly alter spatial patterns in plant‐soil feedback, enabling invasion and transformation of soils previously occupied by heterospecific plants, thus potentially acting as keystone mutualists. This article is protected by copyright. All rights reserved.
Mycorrhizal inoculation can enhance outcomes of ecological restoration, but the benefits may be context-dependent. Here, we performed a meta-analysis of field studies to elucidate conditions in which adding mycorrhizal fungi enhances restoration success. We found inoculation increased plant biomass by an average effect size of 1.7 in 70 independent comparisons from 26 field-based studies, with the largest increases to N-fixing woody plants, C4-grasses and plants growing in soils with low plant-available P. Growth responses to inoculation increased with time for the first 3 yr after inoculation, especially for N-fixing woody plants and plants growing in severely altered soils. We found that mycorrhizal inoculation increased species richness of restored plant communities by 30%, promoted establishment of target species, and enhanced similarity of restored to reference communities. We conclude that the addition of mycorrhizal fungi to restoration sites can facilitate rapid establishment of vegetation cover, and restoration of diverse plant communities more akin to reference sites.
Most land plants engage in mutually beneficial interactions with arbuscular mycorrhizal (AM) fungi, the fungus providing phosphate and nitrogen in exchange for fixed carbon. During presymbiosis, both organisms communicate via oligosaccharides and butenolides. The requirement for a rice chitin receptor in symbiosis-induced lateral root development suggests that cell division programs operate in inner root tissues during both AM and nodule symbioses. Furthermore, the identification of transcription factors underpinning arbuscule development and degeneration reemphasized the plant's regulatory dominance in AM symbiosis, with degeneration temporally linked to ceasing fungal phosphate delivery. Finally, the finding that AM fungi, as lipid auxotrophs, depend on plant fatty acids (FAs) to complete their asexual life cycle revealed the basis of fungal biotrophy. Intriguingly, lipid metabolism is also central for asexual reproduction and interaction of the fungal sister clade, the Mucoromycotina, with endobacteria, indicative of an evolutionarily ancient role for lipids in fungal mutualism. Expected final online publication date for the Annual Review of Phytopathology Volume 56 is August 25, 2018. Please see for revised estimates.
Microorganisms have a pivotal role in the functioning of ecosystems. Recent studies have shown that microbial communities harbour keystone taxa, which drive community composition and function irrespective of their abundance. In this Opinion article, we propose a definition of keystone taxa in microbial ecology and summarize over 200 microbial keystone taxa that have been identified in soil, plant and marine ecosystems, as well as in the human microbiome. We explore the importance of keystone taxa and keystone guilds for microbiome structure and functioning and discuss the factors that determine their distribution and activities.
Little is known about how soil seed banks vary in germination, composition and density under different land uses after tropical forest conversion. Seed banks can potentially act as one source of regeneration for reforestation of old agricultural lands. Our study documents the composition and density of germinants in soil seed banks from four land use surrounding the Sinharaja forest in southwest Sri Lanka. These include: (1) kekilla fern lands, (2) pine plantations, and (3) tea. These were compared to the adjacent (4) mature rainforest. During the six‐month period of monitoring, we recorded 1674 germinants (0.036 germinants cm–3 soil), representing 46 species. Germinants of tree and shrub species were restricted to the pine and rain forest soils and all of them are considered pioneers. The soils of the rain forest had the lowest species richness, density and diversity of germinants; tea lands comprised much higher richness, Shannon diversity and density. However, almost all germinants in tea were grasses and herbs as compared with other land uses. A multivariate analysis of the germinants of soil seed banks revealed that the four land use types comprise very different compositions and abundances, some of which can be associated to differences in growth habit (trees, shrubs, vines, herbs, grasses). Our results suggest that pine plantations may facilitate some tree and shrub regeneration. However, the seed banks beneath tea and kekilla fern land do not comprise any woody plant species. This may explain why agricultural lands such as tea do not revert back to forest easily.