Received: 5 October 2020 | Accepted: 8 February 2021

DOI: 10.1111/ppa.13362

ORIGINAL ARTICLE

A statistical protocol to describe differences among nutrient

utilization patterns of Fusarium spp. and Trichoderma gamsii

Giovanna Jona Lasinio1 | Alessio Pollice2 | Livia Pappalettere3 |

Giovanni Vannacci3 | Sabrina Sarrocco3

1
Department of Statistical Sciences,
“Sapienza” University of Rome, Rome, Italy Abstract
2
Department of Economics and Finance, The Biolog phenotype microarrays (PM) system offers a simple and cheap tool to rap-
University of Bari Aldo Moro, Largo
Abbazia Santa Scolastica, Bari, Italy
idly provide a high throughput of information about the phenotypes of fungal isolates
3
Department of Agriculture, Food and in a short time. In order to improve the use of the PM system in fungal ecology stud-
Environment, University of Pisa, Pisa, Italy ies, the present work proposes a new statistical protocol based on two approaches,
Correspondence that is, a functional principal components analysis to describe similarity patterns of
Sabrina Sarrocco, Department of growth curves, and a Bayesian generalized additive model (GAM) to allow inferences
Agriculture, Food and Environment,
University of Pisa, Via del Borghetto 80, on specific growth features, in order to analyse nutrient fungal utilization in a model
56124 Pisa, Italy. system including four causal agents of Fusarium head blight, the natural competitor
Email: [email protected]
Fusarium oxysporum, and the beneficial isolate Trichoderma gamsii T6085. Analysis of
Alessio Pollice, Department of Economics
and Finance, Univeristy of Bari Aldo Moro,
data collected by the Biolog PM in our biological system showed a different nutri-
Largo Abbazia Santa Scolastica, Bari, Italy. tional competitive potential of the four pathogens, as well as an intermediate behav-
Email: [email protected]
iour of the natural competitor and of our biocontrol agent. This protocol, applicable to
different fungal phenotypical studies at both isolate and community level, allows a full
exploitation of data obtained by the PM system and provides important information
about the nutritional pattern of a single isolate compared to those of other fungi, a key
factor to be exploited in biocontrol strategies.

KEYWORDS
Bayesian generalized additive models, Biolog phenotype microarray system, functional
clustering, fungal nutrient utilization, Fusarium head blight

1 | I NTRO D U C TI O N with no physical interaction between them (Holomuzki et al., 2010).

Competition is an important threat in plant pathology. Intra-­and in-
Within an ecological community, populations can interact directly and terspecific competition among pathogens can occur across space and
indirectly either in an intraspecific (between individuals of the same time when spatial niche differentiation and/or temporal separation fail
species) or interspecific (between two or more species) way (Brooker (Fitt et al., 2006). In addition, competitive interaction within a coexist-
et al., 2009). Interspecific competition occurs when there is the simul- ing population is one of the mechanisms to control plant disease and to
taneous demand for the same resource that is in limiting supply, that is, reduce pathogen populations (Kinkel et al., 2014). An example of a dis-
not enough to meet the demands for all involved species. Two competi- ease where nutrient competition plays a crucial role for the survival and
tion strategies generally occur within fungal communities: interference development of the causal agents is Fusarium head blight (FHB) of small
competition—­monopolization of the habitat by antagonistic combat grain cereals. FHB, recognized as one of the most serious problems af-
(Sarrocco, 2016); and exploitation competition—­when one organism, fecting wheat all over the world, is caused by a complex of around 20
by exploiting the resource, reduces its availability to another organism fungal species, mostly belonging to Fusarium genus. Within this group,

1146 | © 2021 British Society for Plant Pathology wileyonlinelibrary.com/journal/ppa Plant Pathology. 2021;70:1146–1157.
LASINIO et al. | 1147

F. graminearum species complex (FGSC), as well as F. avenaceum, F. cul- agents of FHB on wheat (Baroncelli et al., 2016; Matarese
morum, and F. poae, are considered as the major species. From the per- et al., 2012). F. oxysporum 7121 belongs to a wider collection
spective of the global incidence of this disease, other species such as of F. oxysporum strains isolated from wheat straw buried in
F. acuminatum, F. cerealis, F. chlamydosporum, F. equiseti, F. langsethiae, soils collected close to Pisa (Italy) with a previous history of
F. sporotrichioides, and F. tricinctum are considered less important wheat cultivation. This isolate was selected because it is able
(Torres et al., 2019). From an ecological point of view, FHB causal to grow in the presence of 50 ppm deoxynivalenol (DON) and
agents seem to interact in a competitive rather than in a synergistic for its ability to colonize natural substrates (Sarrocco et al.,
way during pathogen/disease development (Xu & Nicholson, 2009). 2019).
During the disease cycle, plant debris is used by the pathogens F. graminearum ITEM 124 was isolated from Oryza sativa in Italy
to overwinter between two following cropping seasons, while spikes and its genome was recently sequenced, annotated, and released
at flowering are the most susceptible stage of wheat to infection. (Zapparata et al., 2017). F. culmorum ITEM 627 was isolated from
Application of beneficial fungi on cultural debris, as well as on spikes Triticum durum in former Yugoslavia, while F. langsethiae ITEM 11031
during anthesis, is based on the efficacy of competition for space was isolated from Zea mays in Italy. F. graminearum, F. culmorum, and
and nutrients as a mechanism to control FHB causal agents (Sarrocco F. langsethiae were kindly given by Antonio Moretti (ISPA-­CNR, Bari,
et al., 2019; Sarrocco & Vannacci, 2018). Italy; http://www.ispa.cnr.it/Colle​c tion). F. sporotrichioides 194, iso-
The knowledge of the nutritional requirements of fungi poten- lated from T. durum in Italy, was kindly given by Giovanni Beccari
tially involved in competitive relationships with pathogens is of (Department of Agricultural, Food and Environmental Sciences,
great interest in view of a future application in the field. The Biolog University of Perugia, Italy).
phenotype microarray (PM) system represents a valid, simple to All fungi, deposited in the Fungal Collection of the Plant
use, and relatively cheap tool to rapidly investigate niche overlap Pathology & Mycology Lab (DISAAA-­a, University of Pisa), were
and catabolic versatility of fungi by testing isolates against many maintained on potato dextrose agar (PDA; Difco) under mineral oil
different carbon sources one at a time (Pinzari et al., 2016). Until at 4 °C for long-­term storage and grown on PDA (T. gamsii, F. oxy​
now, data collected by Biolog analysis, consisting of the spectro- sporum, F. culmorum, F. langsethiae, and F. sporotrichioides) or oat
photometric reads at regular intervals of the multiwell plates (95 meal agar (OA; Difco) (F. graminearum) at 24 °C, under a 12 hr light/
wells containing different substrates plus one control well contain- dark photoperiod, when actively growing cultures were needed. All
ing water), have been analysed by different approaches. These in- pathogenic Fusarium isolates were regularly inoculated on the host
clude common analysis of single time-­point optical density values to maintain their virulence.
across substrates at a given reading point, as well as in model-­based
approaches developed in order to describe the kinetics of carbon
source utilization by fungi (Wirsel et al., 2002). Empirical models—­in 2.2 | Assessment of metabolic requirements
particular, semiparametric regression splines—­have now been im-
plemented to extrapolate growth curve parameters such as lag In order to set up a new statistical protocol to establish which
time, maximum rate of increase, and maximum optical density (Vaas substrates could potentially become a source of nutrient com-
et al., 2012). In general, data analysis is still a very controversial and petition, the four pathogens belonging to the FHB complex, the
complex aspect of using Biolog for fungal ecological studies. natural competitor F. oxysporum 7121, and the potential bio-
In order to improve the use of the Biolog system in fungal phe- control agent T. gamsii T6085 were analysed using the Biolog
notyping studies, the present work proposes a statistical protocol microbial identification system (https://www.biolog.com/
to analyse nutrient utilization in fungi based on all the information products-­p ortfolio-­o verview/microbial-­i dentification/). A total
included in each growth curve on all the substrates, by using—­as a of 100 μl of a water spore suspension (10 6 spores/ml) of each
model system—­four causal agents of FHB, a natural fungal competi- of the six fungal isolates was inoculated in each well of a Biolog
tor against these pathogens (Fusarium oxysporum), and a well-­known multiwell plate (FF, for filamentous fungi, MicroPlate) contain-
beneficial isolate of Trichoderma gamsii (T6085) showing promising ing water and 95 different carbon sources (https://www.biolog.
capabilities for use as a biocontrol agent for the management of FHB com/wp-­c ontent/uploads/2020/04/00A-­0 10rB-­F F-­S ell-­S heet-­
disease. Mar07.pdf) and incubated at 24 °C with a 12 hr light/dark pho-
toperiod. Two independent replicates were carried out for each
isolate. Fungal growth was spectrophotometrically measured as
2 | M ATE R I A L S A N D M E TH O DS optical density (OD) at 595 nm for 1 day, every 4 hr (12 hr dur-
ing the night). The OD values, normalized against those for A1
2.1 | Fungal isolates (control well), were used in the present work to set up a new
statistical protocol allowing a better exploitation of all the in-
T. gamsii T6085 was isolated in Crimea (Ukraine) from uncul- formation contained within each fungal growth curve (created
tivated soil and has been studied over a long period of time with OD values) on each substrate, as described in the follow-
for its ability to control F. graminearum, one of the main causal ing section.
1148 | LASINIO et al.

2.3 | Statistical analysis protocol obtained by semiparametric generalized additive models (GAMs) and
were used to define and calculate the scores for each isolate on each
A detailed analysis of fungal phenotypes based on specific nutrient substrate, while accounting for their uncertainty.
utilization was obtained by a new statistical protocol (Figure 1) in
order to
2.4 | Characterization –­functional clustering
• cluster optical density growth curves into groups corresponding
to relevant growth features (characterization), and The main focus of conventional clustering algorithms is the defini-
• compare and order fungal isolates within substrates according to tion of homogeneous subgroups of individuals in a data set. Cluster
relevant growth features (assessment). analysis is also often applied to longitudinal data, such as growth
curves. In this case, a primary concern was to find curve patterns
We framed these two research questions into a preliminary ex- representing different shapes and variation. Recently, several clus-
ploratory (i.e., characterization in Figure 1) and a subsequent con- tering methods driven by the functional characteristics of the avail-
firmatory (i.e., assessment in Figure 1) statistical environment and able data have been proposed. The common rationale is to project
estimated the growth functions of the six isolates for all substrates by the observed growth curves to certain basis functions and to pro-
two approaches. To achieve the first target, growth curves of OD data cess curve projections by a functional mixed model.
for each isolate and substrate were projected onto an adequate basis, In this work functional clustering (FC) was performed by the k-­
and curve clusters were found by functional clustering. To achieve centres method described by Chiou and Li (2007) and implemented in
the second, Bayesian growth predictions and credible intervals were the funcit function (Funcy R package v. 0.8.6; Yassouridis et al., 2018).

F I G U R E 1 Analysis of fungal phenotypes based on nutrients utilization is obtained by exploratory clustering optical density growth
curves into groups corresponding to growth features (characterization): growth curves of optical density(OD) data for each isolate and
substrate are projected onto adequate basis and curve clusters are found by functional clustering; and confirmatory comparison and
ordering fungal isolates within substrates according to growth features (assessment): Bayesian growth predictions and credible intervals are
obtained by semiparametric generalized additive models (GAMs) and used to define and calculate scores for each isolate on each substrate,
accounting for their uncertainty [Colour figure can be viewed at wileyonlinelibrary.com]
LASINIO et al. | 1149

This method is a functional counterpart of the k-­means algorithm as • Maximum height A

cluster membership is predicted with a reclassification step: alternately, • Surface under the curve S
curves are assigned to classes, and classes are calculated anew depend-
ing on their assigned curves. Each curve is projected into all k eigenspa- For each parameter and fungal isolate we computed 95% cred-
ces generated by functional principal component analysis and then it is ible intervals within substrates. Uncertainty measures were then
assigned to one of them on the basis of a nonparametric random-­effect used to define growth scores that allow comparison and ordering of
model of the functional principal components, coupled with a nonpara- isolates within substrates. Growth scores are obtained as follows: let
metric iterative mean and covariance updating scheme. Optical density pisk be one of the four growth curve parameters for isolate i = 1, ⋯, 6,
growth data for each isolate and substrate were processed by the k-­ substrate s = 1, ⋯, 93 and k = 1, ⋯, 4 for latency 𝜆, slope 𝜇, maximum
centres semiparametric FC method, with k = 3. According to cluster A, and surface S.
membership, each substrate was then classified as large (fast growth),
medium (medium growth), or small (supporting poor growth). • Rescale all parameters to the range (0, 1) as follows:
p − mini (pisk )
%pisk = max iskp − min
i ( isk ) i (pisk )
• Only for the latency parameter use 1 − %𝜆isk instead of %𝜆isk to
2.5 | Assessment –­Bayesian generalized ease comparisons
additive models • Consider, for example, the fungal pair (i, j)such that %pisk > %pjsk in
substrate s
The behaviour of each isolate in each substrate was assessed by • To define the partial scores of the i-­th isolate within the s-­th sub-
an inferential procedure that allows a probabilistic comparison be- strate for the k-­th parameter check if the 95% credible intervals of
tween isolate features within substrates. We first searched for the %pisk and %pjskoverlap
“best” estimate of each growth curve and then computed summary a. YES → SCisk = SCjsk = %pisk + %pjsk ∕2
( )

growth parameters describing the isolate behaviour within the sub- b. NO → SCisk = %pisk and SCjsk = − %pjsk
strates. We searched for the best model in the semiparametric class • Define the average score of the i-­th isolate within the s-­th sub-
of GAMs. Let yist be the log-­OD of the i-­th isolate (i = 1, …, 6) in sub- strate as follows:
strate s (s = 1, …, 92) at time xt (t = 1, …, 17, xt = 0, 4, …, 96) and assume
4
it is expressed as follows: 1 ∑
SCis = SC
4 k = 1 isk
( ) ( )
yist = 𝜇 is + 𝛽 s xt + fs xt + 𝛽 is xt + fis xt + 𝜀ist
3 | R E S U LT S
with 𝜇is average growth of isolate i in substrate s; 𝛽 s growth rate in
substrate s; fs ( xt ) nonlinear growth effect in substrate s; 𝛽 is growth rate 3.1 | Characterization –­functional clustering
of isolate i in substrate s; fis ( xt ) nonlinear growth effect of isolate i in
substrate s; 𝜀ist random error. OD growth data for each isolate and substrate were processed by
Generalized additive models are estimated in the Bayesian infer- the k-­centres semiparametric FC method, with k = 3. In Figure 2,
ential framework making use of “spike and slab” prior distributions growth curves for the six isolates on 93 substrates (sebacic acid G2
that allow a “stochastic search” selection of relevant model terms and adenosine H10 were withdrawn from this analysis due to their
(Scheipl et al., 2012). In particular, this approach enables computa- unclear results; water A1 was used to normalize all OD data) are re-
tion of the posterior inclusion probability of each model component ported. The 93 × 6 = 558 growth curves have been clustered into
(for each isolate within substrates). Then, the “best” model is ob- three categories: large, medium, and small.
tained selecting model terms with large (>0.8) inclusion probabilities. Data that gave rise to Figure 2 were used to construct a con-
Bayesian estimation is carried out using Monte Carlo simulations: tingency table (Table 1) where, within each category, substrates
for each growth curve we generated a collection of (100) simulated have been grouped according to the isolates that metabolize them
curves that can be seen as samples from the curve posterior proba- (Table 1). Data presented in Table 1 indicate that, among Fusarium
bility distribution. These samples were used to build credible inter- isolates, F. graminearum (55, 23, and 15 substrates supporting large,
vals for the growth curve as well as for any summary of the curve medium, and small growth, respectively) had the best catabolic ver-
itself. In this work, 95% credible intervals were built using 0.025 and satility, followed by F. culmorum (7, 51, and 35 substrates supporting
0.975 percentiles of each curve (or summary parameter) distribution. large, medium, and small growth, respectively) and F. oxysporum (1,
To summarize the growth features of each isolate in each sub- 59, and 33 substrates supporting large, medium, and small growth,
strate, we computed the following summary parameters: respectively). In detail, F. graminearum showed the best saprotrophic
ability, whereas F. culmorum grew fast on seven substrates, five
• Latency λ of which are in common with F. graminearum, that is, quinic acid
• Growth rate μ computed as the slope of the steepest tangent to (F12), l-­alanine (G8), l-­asparagine (G10), l-­aspartic acid (G11), and
the curve l-­glutamic acid (G12). The only substrate where F. oxysporum grew
1150 | LASINIO et al.

(2.1 sugar acids, 2.2 hexosamines, 2.3 polyols); 3 = other sugars (3.1
polysaccharides, 3.2 oligosaccharides, 3.3 glucosides); 4 = nitrogen-­
containing compounds (4.1 peptides, 4.2 l-­amino acids, 4.3 biogene
and heterocyclic amines, 4.4 TCA cycle intermediates, 4.5 aliphatic
organic acids); 5 = others.
Cells in Table 2 are marked with different colours according to
their values, varying from 1.25 (blue) to −0.086 (red), to give a visual
assessment of the different catabolic activity of the four pathogens,
T. gamsii T6085, and F. oxysporum. The column for F. graminearum
scores shows the highest number of blue cells (Table 2). According
to score values, F. graminearum was able to grow the best on the ma-
jority of the tested substrates (74 out of 93 substrates), followed by
T. gamsii T6085, F. culmorum, and F. oxysporum, that showed similar
behaviour, showing the highest scores when grown on six, five, and
four out of the 93 substrates, respectively. However, F. graminearum
generally had a good catabolic versatility, showing an average score
of 1.012, whereas F. culmorum, T. gamsii T6085, and F. oxysporum
had an average score of 0.479, 0.334, and 0.513, respectively. Also,
in this second analysis, F. langsethiae and F. sporotrichioides were
F I G U R E 2 Exploratory clustering of optical density (OD) growth confirmed to have a scarce ability to grow on different substrates
curves into groups corresponding to growth features: growth because they showed, in both cases, a score value of 0.188. Despite
curves of OD data for each isolate and substrate are projected onto
the ability of F. graminearum to grow well on the majority of the
adequate basis and curve clusters are found by semiparametric
substrates, from Table 2 it is also possible to appreciate differences
k-­centres functional clustering. Each substrate is classified as large
(fast growth, blue line), medium (medium growth, grey-­black line), between its catabolic capacity and that of the nonpathogenic iso-
or small (poor growth, red line) [Colour figure can be viewed at lates T. gamsii T6085 and the natural competitor F. oxysporum for
wileyonlinelibrary.com] specific substrates, with sugars apparently playing an important
role. For example, within the group of polysaccharides (3.1), α-­ and
fast is l-­asparagine (G10), in common with both F. graminearum and β-­cyclodextrin (B1 and B2, respectively) were poorly assimilated by
F. culmorum. F. graminearum (score = 0.166 and = 0.046, respectively), whereas
With respect to T. gamsii T6085 (0, 29, and 64 substrates sup- T. gamsii was able to catabolize B1 (score = 1.230) and F. oxysporum
porting large, medium, and small growth, respectively), the antag- could catabolize B2 (score = 1.156).
onistic isolate collocated in an intermediate position. However, Among oligosaccharides (3.2), the pathogen seemed unable to
among those substrates supporting medium growth, d-­cellobiose catabolize C10 (maltitol), showing a score of 0.193, while T. gamsii
(A12), α-­d-­glucose (B12), and glycogen (C5) are in common with all grew very well, with a score of 0.817. Two other oligosaccharides,
the Fusarium isolates that had medium ability to metabolize such C8 (α-­d-­lactose) and C9 (lactulose), were assimilated well by T. gamsii
substrates. Finally, both F. langsethiae and F. sporotrichioides showed T6085 (both scores around 1.000), while the pathogen did not show
a very slow growth rate on all substrates, with their profiles charac- a good catabolic ability when inoculated on them (scores = 0.477 and
terized only by substrates supporting small growth. 0.134, respectively). Among monosaccharide-­related compounds,
the sugar acid d-­galacturonic acid (B8) was scarcely catabolized by
the pathogen (score = 0.342) whereas F. oxysporum seemed to use it
3.2 | Assessment –­Bayesian generalized well (score = 1.022).
additive models Finally, among monosaccharides, F. graminearum could not grow
well on the pentose d-­arabinose (A8), showing a score of 0.186,
The first step of this second analysis was to obtain the “best” esti- whereas F. oxysporum showed a score of 1.014. The pathogen grew
mate of each growth curve for each isolate. Then, the behaviour of on its stereoisomer l-­arabinose (A9), showing a score of 0.655, com-
each isolate in each substrate was assessed by an inferential proce- parable with that of F. oxysporum. The six isolates were ranked on
dure that allows a probabilistic comparison among isolates within each substrate and then the frequency distribution of ranks was ob-
substrates and the assignment of a score to each isolate valid for tained for each isolate (Table 3). F. graminearum showed the highest
comparison within each substrate. Growth scores, defined by un- frequency for rank 1 (76 out of 93), followed by T. gamsii T6085 (7),
certainty measures, are shown in Table 2. Substrates included in the and F. oxysporum and F. culmorum with very similar values (6 and 4,
Biolog PM FF microplates were ordered according to Atanasova and respectively). However, T. gamsii T6085 was positioned at rank 6 for
Druzhinina (2010), as follows: 1 = monosaccharides (1.1 heptoses, a quarter of the substrates. Finally, F. langsethiae and F. sporotrichioi-
1.2 hexoses, 1.3 pentoses); 2 = monosaccharide-­related compounds des showed a very high frequency distribution for rank 5 (33 and 32,
LASINIO et al. | 1151

TA B L E 1 Substrates included in the Biolog phenotype microarray FF microplates grouped by semiparametric functional clustering
according to their ability to be metabolized by each isolate

Isolate Small Medium Large

Fusarium culmorum A2 A3 A4 A8 B1 B2 B4 B6 B8 B11 C2 C4 A5 A6 A7 A9 A10 A11 A12 B3 B5 B7 B9 B10 D8 F12 G8 G10 G11
C6 C8 C9 C10 D1 D2 D6 D7 D10 D12 B12 C1 C3 C5 C7 C11 C12 D3 D4 D5 D9 G12 H3
E3 E8 F2 F5 F6 F7 F8 G6 G7 G9 H1 D11 E1 E2 E4 E5 E6 E7 E9 E10 E11 E12 F1
H8 H12 F3 F4 F9 F10 F11 G1 G3 G4 G5 H2 H4 H5
H6 H7 H9 H11
Fusarium A3 A8 B1 B2 B6 B8 C1 C2 C8 C9 C10 D6 A2 A5 A9 A12 B4 B7 B9 B11 B12 C5 C6 D2 D7 A4 A6 A7 A10 A11 B3
graminearum D10 E3 F6 D9 D12 E1 F2 F8 G3 G6 H3 H5 H8 B5 B10 C3 C4 C7
C11 C12 D1 D3 D4
D5 D8 D11 E2 E4
E5 E6 E7 E8 E9 E10
E11 E12 F1 F3 F4
F5 F7 F9 F10 F11
F12 G1 G4 G5 G7
G8 G9 G10 G11
G12 H1 H2 H4 H6
H7 H9 H11 H12
Fusarium langsethiae A2 A3 A4 A5 A6 A7 A8 A9 A10 A11 A12
B1 B2 B3 B4 B5 B6 B7 B8 B9 B10 B11
B12 C1 C2 C3 C4 C5 C6 C7 C8 C9 C10
C11 C12 D1 D2 D3 D4 D5 D6 D7 D8
D9 D10 D11 D12 E1 E2 E3 E4 E5 E6 E7
E8 E9 E10 E11 E12 F1 F2 F3 F4 F5 F6
F7 F8 F9 F10 F11 F12 G1 G3 G4 G5 G6
G7 G8 G9 G10 G11 G12 H1 H2 H3 H4
H5 H6 H7 H8 H9 H11 H12
Fusarium oxysporum A3 A5 A6 B1 B2 B4 B6 B8 B11 C1 C2 C6 A2 A4 A7 A8 A9 A10 A11 A12 B3 B5 B7 B9 G10
C8 C9 C10 D6 D7 D10 E2 E3 E8 F2 F5 B10 B12 C3 C4 C5 C7 C11 C12 D1 D2 D3
F7 F8 G6 G7 H1 H3 H7 H11 H12 D4 D5 D8 D9 D11 D12 E1 E4 E5 E6 E7 E9
E10 E11 E12 F1 F3 F4 F6 F9 F10 F11 F12
G1 G3 G4 G5 G8 G9 G11 G12 H2 H4 H5
H6 H8 H9
Fusarium A2 A3 A4 A5 A6 A7 A8 A9 A10 A11 A12
sporotrichioides B1 B2 B3 B4 B5 B6 B7 B8 B9 B10 B11
B12 C1 C2 C3 C4 C5 C6 C7 C8 C9 C10
C11 C12 D1 D2 D3 D4 D5 D6 D7 D8
D9 D10 D11 D12 E1 E2 E3 E4 E5 E6 E7
E8 E9 E10 E11 E12 F1 F2 F3 F4 F5 F6
F7 F8 F9 F10 F11 F12 G1 G3 G4 G5 G6
G7 G8 G9 G10 G11 G12 H1 H2 H3 H4
H5 H6 H7 H8 H9 H11 H12
Trichoderma gamsii A3 A4 A5 A8 A9 B1 B2 B6 B8 B10 B11 C1 A2 A6 A7 A10 A11 A12 B3 B4 B5 B7 B9 B12
T6085 C2 C3 C4 C6 C7 C9 C10 C11 D3 D5 C5 C8 C12 D1 D2 D4 D11 E2 E6 E7 E9 E10
D6 D7 D8 D9 D10 D12 E1 E3 E4 E5 E8 E12 G8 G10 G11 G12
E11 F1 F2 F3 F4 F5 F6 F7 F8 F9 F10
F11 F12 G1 G3 G4 G5 G6 G7 G9 H1 H2
H3 H4 H5 H6 H7 H8 H9 H11 H12

Abbreviations: Small, poor growth; medium, medium growth; large, fast growth.

respectively) and rank 6 (30 and 31, respectively), corresponding to and potential biocontrol agents is a fundamental prerequisite in
approximately a third of the substrates. order to define the best strategy for disease management (Sarrocco
& Vannacci, 2018). In this context, substrate colonization is the first
step in the cascade of events that lead to the use of different bio-
4 | DISCUSSION logical control mechanisms such as antibiosis, mycoparasitism, and
induction of resistance in host plants (Jaroszuk-­Scise et al., 2019).
Within the scenario of a complex disease such as FHB of wheat, The knowledge of comparative nutritional requirements among the
knowledge of the nutrient requirements of the main causal agents main actors of debris colonization (either natural or deliberately
1152 | LASINIO et al.

TA B L E 2 Growth scores, defined by uncertainty measures, assigned to each isolate for each substrate included in the Biolog phenotype
microarray FF microplates

Well FG TG FC FL FS FO Substrate Group

E3 0.813 0.762 0.812 0.456 0.778 0.706 Sedoheptulosan 1.1

B12 1.030 0.818 0.579 0.261 0.242 0.878 α-­d-­Glucose 1.2
B5 0.998 0.644 0.275 0.082 0.004 0.335 d-­Fructose 1.2
B7 0.958 0.693 0.314 0.109 −0.078 0.705 d-­G alactose 1.2
D2 0.902 0.936 0.434 0.206 0.214 0.678 d-­Mannose 1.2
E8 1.079 0.231 0.253 0.287 0.251 0.156 d-­Tagatose 1.2
B6 1.163 0.186 0.314 0.157 0.238 0.453 l-­Fucose 1.2
D12 0.991 0.113 0.248 0.190 0.194 0.573 l-­R hamnose 1.2
E5 1.146 0.309 0.338 0.312 0.180 0.672 l-­S orbose 1.2
A8 0.186 0.120 0.245 0.157 0.272 1.014 d-­A rabinose 1.3
D10 1.216 0.172 0.303 0.176 0.185 0.299 d-­Psicose 1.3
E1 0.855 0.081 0.965 0.046 0.161 0.937 d-­R ibose 1.3
E12 1.087 0.657 0.790 0.315 0.173 0.686 d-­Xylose 1.3
A9 0.655 0.060 0.833 0.051 0.178 0.919 l-­A rabinose 1.3
C7 1.229 0.059 0.437 0.038 −0.011 0.630 2-­Keto-­d-­gluconic acid 2.1
B8 0.342 0.100 0.643 0.199 0.187 1.022 d-­G alacturonic acid 2.1
B10 1.187 0.155 0.460 0.131 0.117 0.521 d-­G luconic acid 2.1
C3 1.210 0.102 0.510 0.027 −0.051 0.604 d-­G lucuronic acid 2.1
G1 1.190 0.159 0.553 0.128 0.089 0.571 d-­S accharic acid 2.1
C2 1.169 0.110 0.325 0.139 0.142 0.177 Glucuronamide 2.1
B11 1.067 0.065 0.166 0.002 −0.001 0.501 d-­G lucosamine 2.2
A3 0.623 0.998 0.205 0.083 0.203 0.127 N-­Acetyl-­d-­galactosamine 2.2
A4 1.172 0.142 0.147 0.130 0.165 0.801 N-­Acetyl-­d-­glucosamine 2.2
A5 0.630 0.208 1.022 0.053 0.159 0.133 N-­Acetyl-­d-­mannosamine 2.2
A6 1.213 0.446 0.359 0.092 0.148 0.127 Adonitol 2.3
A10 1.174 0.574 0.268 0.131 0.200 0.741 d-­A rabitol 2.3
D1 1.229 0.676 0.278 0.100 0.064 0.502 d-­Mannitol 2.3
E4 1.167 0.157 0.596 0.269 0.238 0.694 d-­S orbitol 2.3
C4 1.219 0.028 0.219 0.092 0.047 0.284 Glycerol 2.3
B4 1.249 0.549 0.188 0.081 0.069 0.066 i-­Erythritol 2.3
C6 1.026 0.119 0.355 0.007 −0.026 0.683 m-­Inositol 2.3
E11 1.153 0.200 0.488 0.241 0.258 0.555 Xylitol 2.3
B1 0.166 1.230 0.353 0.166 0.138 0.540 α-­Cyclodextrin 3.1
B2 0.046 0.240 0.735 0.148 0.025 1.156 β-­Cyclodextrin 3.1
B3 0.987 0.671 0.798 0.282 0.345 0.550 Dextrin 3.1
C5 0.887 0.991 0.786 0.279 0.525 0.714 Glycogen 3.1
C8 0.477 1.028 0.230 0.209 0.215 0.102 α-­d-­L actose 3.2
A12 1.088 0.809 0.431 0.269 0.275 0.802 d-­Cellobiose 3.2
D3 1.158 0.081 0.521 0.219 0.065 0.417 d-­Melezitose 3.2
D4 1.171 0.339 0.372 0.098 0.135 0.509 d-­Melibiose 3.2
D11 1.220 0.147 0.695 0.234 0.215 0.603 d-­R affinose 3.2
B9 1.139 0.565 0.371 0.105 0.130 0.555 Gentiobiose 3.2
C9 0.134 1.011 0.237 0.141 0.243 0.225 Lactulose 3.2
C10 0.193 0.817 0.460 0.175 0.378 0.282 Maltitol 3.2
C11 1.209 0.122 0.304 0.249 0.242 0.613 Maltose 3.2

(Continues)
LASINIO et al. | 1153

TABLE 2 (Continued)

Well FG TG FC FL FS FO Substrate Group

C12 1.197 0.444 0.495 0.237 0.262 0.728 Maltotriose 3.2

D9 1.219 0.121 0.603 0.210 0.062 0.584 Palatinose 3.2
D5 1.225 0.209 0.488 0.078 0.062 0.632 α-­Methyl-­d-­galactoside 3.3
D7 1.236 0.093 0.273 0.199 0.094 0.188 α-­Methyl-­d-­glucoside 3.3
A7 1.227 0.433 0.353 0.036 0.143 0.702 Amygdalin 3.3
A11 1.098 0.361 0.344 0.248 0.313 0.420 Arbutin 3.3
D6 0.441 0.092 0.225 0.067 0.201 1.031 β-­Methyl-­d-­galactoside 3.3
D8 1.194 0.200 0.790 0.097 0.196 0.495 β-­Methyl-­d-­glucoside 3.3
E9 1.085 0.599 0.614 0.221 0.139 0.491 d-­Trehalose 3.3
E2 1.229 0.393 0.544 0.305 0.139 0.209 Salicin 3.3
E6 1.187 0.352 0.620 0.283 0.270 0.485 Stachyose 3.3
E7 1.072 0.631 0.615 0.283 0.379 0.478 Sucrose 3.3
E10 1.112 0.620 0.531 0.278 0.238 0.713 Turanose 3.3
H1 1.018 0.137 0.142 0.215 0.114 0.107 Glycyl-­l-­glutamic acid 4.1
G9 1.177 0.271 0.275 0.272 0.286 0.463 l-­A lanyl-­glycine 4.1
F1 1.138 0.031 0.422 0.220 0.093 0.719 γ-­Aminobutyric acid 4.2
G8 1.057 0.218 0.689 0.269 0.259 0.611 l-­A lanine 4.2
G10 1.091 0.391 0.845 0.184 0.297 0.577 l-­A sparagine 4.2
G11 1.071 0.352 0.661 0.331 0.398 0.616 l-­A spartic acid 4.2
G12 1.071 0.349 0.650 0.341 0.243 0.596 l-­G lutamic acid 4.2
H2 1.174 0.159 0.286 0.272 0.258 0.658 l-­O rnithine 4.2
H3 0.982 0.136 1.121 0.263 0.222 0.086 l-­Phenylalanine 4.2
H4 1.125 0.142 0.642 0.272 0.260 0.627 l-­Proline 4.2
H5 0.992 0.462 0.969 0.265 0.251 0.864 l-­P yroglutamic acid 4.2
H6 1.013 0.371 0.629 0.360 0.280 0.633 l-­S erine 4.2
H7 0.984 0.135 0.478 0.197 0.100 0.109 l-­T hreonine 4.2
G6 0.897 0.225 0.154 0.334 0.306 0.222 N-­Acetyl-­l-­glutamic acid 4.2
H8 1.153 0.466 0.621 0.234 0.319 0.918 2-­Amino ethanol 4.3
H9 1.109 0.126 0.882 0.214 0.241 0.569 Putrescine 4.3
H11 1.007 0.145 0.221 0.211 0.096 0.102 Uridine 4.3
F7 1.239 0.202 0.269 0.170 0.197 0.365 α-­Ketoglutaric acid 4.4
F10 1.204 0.052 0.380 0.131 0.137 0.239 d-­Malic acid 4.4
F3 1.243 0.166 0.555 0.109 0.125 0.361 Fumaric acid 4.4
F11 1.231 0.237 0.469 0.165 0.086 0.424 l-­Malic acid 4.4
G4 1.185 0.230 0.533 0.264 0.111 0.419 Succinic acid 4.4
F4 1.215 0.098 0.343 0.134 0.189 0.463 β-­Hydroxybutyric acid 4.5
F2 1.151 0.096 0.491 0.046 0.227 0.172 Bromosuccinic acid 4.5
F5 1.250 0.208 0.291 0.183 0.193 0.296 γ-­Hydroxybutyric acid 4.5
F9 1.197 0.067 0.422 0.139 0.131 0.348 l-­L actic acid 4.5
H12 0.999 0.153 0.143 0.236 0.146 0.141 Adenosine-­5′-­ 5
monophosphate
F8 1.220 0.134 0.202 0.162 0.169 0.250 d-­L actic acid methyl ester 5
C1 0.430 0.408 1.049 0.135 −0.005 −0.086 Glucose-­1-­phosphate 5
G7 1.026 0.151 0.134 0.252 0.154 0.159 l-­A laninamide 5
F6 0.549 0.158 0.487 0.076 0.295 1.112 p-­Hydroxyphenylacetic acid 5
F12 1.208 0.191 0.544 0.267 0.252 0.444 Quinic acid 5

(Continues)
1154 | LASINIO et al.

TABLE 2 (Continued)

Well FG TG FC FL FS FO Substrate Group

G3 0.863 0.127 0.757 0.218 0.196 0.720 Succinamic acid 5

G5 1.171 0.201 0.573 0.250 0.245 0.615 Succinic acid monomethyl 5
ester
A2 0.939 0.802 0.522 0.245 0.157 0.822 Tween 80 5

Note: The red to blue colour scale corresponds to increasing values of the growth scores. Carbon substrates found in the FF plates ordered according
to the groups proposed by Atanasova and Druzhinina (2010): 1: monosaccharides: 1.1 heptoses, 1.2 hexoses, 1.3 pentoses; 2: monosaccharide-­
related compounds: 2.1 sugar acids, 2.2 hexosamines, 2.3 polyols; 3: other sugars: 3.1 polysaccharides, 3.2 oligosaccharides, 3.3 glucosides; 4:
nitrogen-­containing compounds: 4.1 peptides, 4.2 L-­amino acids, 4.3 biogene and heterocyclic amines, 4.4 TCA-­c ycle intermediates, 4.5 aliphatic
organic acids; 5: others.
Abbreviations: FG, Fusarium graminearum; TG, Trichoderma gamsii T6085; FC, Fusarium culmorum; FL, Fusarium langsethiae; FS, Fusarium
sporotrichioides; FO, Fusarium oxysporum.

TA B L E 3 Frequency distribution of ranks and isolates for all for competition, the availability of a system that includes the max-
substrates imum number of nutrients in a single assay would be an ideal tool

Isolate to measure the metabolic diversity of these organisms. It is not

easy to imagine a single assay that allows the simultaneous testing
Rank FG TG FC FL FS FO Total
of hundreds of carbon and nitrogen sources, as well as sulphur and
1 76 7 4 0 0 6 93 phosphorus ones, and also including several important environmen-
2 9 7 34 5 2 36 93 tal conditions such as pH, light, and oxygen availability (Atanasova
3 3 14 38 4 6 28 93 et al., 2010). However, the Biolog PM system seems to be capable of
4 2 26 14 21 22 8 93 doing this, providing a phenotypic characterization of several fungal
5 2 16 1 33 32 9 93 isolates in a short time.

6 1 23 2 30 31 6 93 Although it is accepted that the PM system can be used to

obtain information on the catabolic phenotypes of fungi, both
Total 93 93 93 93 93 93
at isolate and community level (Bochner et al., 2011), much more
Abbreviations: FG, Fusarium graminearum; TG, Trichoderma gamsii
controversial and complex is the issue of data analysis. Analytical
T6085; FC, Fusarium culmorum; FL, Fusarium langsethiae; FS, Fusarium
sporotrichioides; FO, Fusarium oxysporum. methods used and developed until now have been extensively
reviewed by Pinzari et al. (2016). However, even if complex ap-
introduced) can help to rank nutrient competition as one of the proaches such as kinetic or multivariate analysis are described,
major (or minor, as seems to be the case) mechanisms of action of single time-­p oint estimation is still the most frequently used tool
a successful antagonist. As a side effect, comparative nutritional for elaborating this kind of data (Oszust et al., 2020). The proto-
analysis offers the opportunity to identify nutrients solely exploited col adopted in the present study can be applied to two or more
by the antagonist, suggesting the combined use of the antagonist fungi that require comparison according to their nutritional re-
and such nutrient(s) in field applications could offer an ecological quirements. A very promising field of application is the study of
advantage to the beneficial fungus. synthetic microbial consortia (Kong et al., 2018), where single
F. graminearum and F. culmorum are recognized worldwide as the components should occupy different ecological niches to avoid
main causal agents of FHB whereas, more recently, F. langsethiae has possible competition among members of the consortium, but
been frequently isolated from wheat kernels in Europe. There has must share at least part of their nutritional requirements with the
been significant interest in the latter pathogen, together with F. spo- pathogen, to boost competition for resources and restrict or sup-
rotrichioides, due to their quite new mycotoxigenic arsenal (Nazari press its activity (Niu et al., 2020).
et al., 2019). These four pathogens were chosen in the present work, The statistical protocol set up in the present work can be used in
together with the beneficial isolate T. gamsii T6085 (Matarese et al., two different ways. Using the functional principal component anal-
2012; Sarrocco et al., 2019, 2020; Vicente et al., 2020), studied for ysis, it is possible to effectively describe global features of isolates,
its potential as a biocontrol agent of FHB, and F. oxysporum 7,121 determine their general preferences in terms of substrate, and en-
(Sarrocco et al., 2012), well known for its ability to compete for cul- visage similarities in terms of growth. Based on functional principal
tural debris, as model organisms for setting up a new statistical pro- component analysis, the characterization approach allows for adap-
tocol to better exploit data derived by the PM system and therefore tive estimation of growth curves but prevents a proper uncertainty
to fully understand their catabolic ability in the perspective of bio- assessment. Relying on the Bayesian GAMs instead, it is possible to
logical control of FHB on wheat. rigorously compare growth behaviours in terms of specific growth
Because it is generally accepted that fungal phenotype is parameters, and rank isolates in terms of summary scores, account-
mainly determined by macronutrients that could act as territory ing for estimation uncertainty. In summary, functional principal
LASINIO et al. | 1155

component analysis has higher flexibility in describing similarity pat- in order to evaluate the possibility of using this substrate for formula-
terns of growth curves, while Bayesian GAMs, and the subsequent tion of biologically active chemicals in solid state dispersion and liquid
computation of growth parameters, allow inferences on specific state. The authors indicated that Trichoderma species preferred the
growth features. Both approaches involve the semiparametric esti- α-­ and γ-­cyclodextrin to β-­cyclodextrin, which is in line with what we
mation of growth curves and are complementary and comparable in observed in the T. gamsii isolate here that grew well on α-­cyclodextrin
terms of computational complexity. and poorly on β-­cyclodextrin. In addition, they suggested that this
Results obtained by the functional clustering presented here information may contribute to promote the elimination of this class
gave a general overview of the different behaviour of the four of compounds from any organic and inorganic matrices (Oros et al.,
pathogens and the two nonpathogenic isolates. From this analysis, 2020). At the same time, the scarce growth of F. graminearum and the
F. graminearum seemed to potentially be the most competitive iso- other pathogenic Fusarium spp. on both α-­ and β-­cyclodextrin may
late in terms of nutrient assimilation, being able to quickly grow on be of particular relevance in view of the control of these pathogens,
the majority of tested substrates. This is in agreement with Leplat because this compound is already used with tebuconazole for the
et al. (2012) that, among all Fusarium species involved in the FHB management of foot and crown rot of wheat caused by F. culmorum
of wheat, F. graminearum is a good saprotroph thanks to its wide (Balmas et al., 2006).
enzymatic arsenal that allows a rapid growth when fresh matter is Another substrate differentially used by F. graminearum (slowly)
available. Data shown in Table 2 can be of further help in choosing and T. gamsii T6985 (rapidly) was maltitol. The assimilation of this
substrates that could put the nonpathogenic isolate at a competitive specific polyol, already observed in several Trichoderma spp., may
advantage towards F. graminearum. have a role on the ecology of T. gamsii T6085, because it could be
From our analysis, F. culmorum was the second isolate, in order used as an indicator of dehydrogenase activity potentially involved
of growth rate, as it developed fast on seven substrates, six of which in survival when drought conditions occur (Hoyos-­C arvajal & Bissett,
were in common with F. graminearum. This apparent niche overlap 2011).
supports Xu and Nicholson’s (2009) explanation when describing More difficult to explain is the different behaviour of T. gamsii
the community ecology of fungal pathogens causing FHB. Disease T6085 and F. graminearum on both α-­d-­lactose and lactulose. This is
development is the result of the interactions of pathogen compo- especially due to the fact that metabolism of both these substrates
nents, where F. graminearum is the most competitive, while the com- is involved in the intracellular galactoglycome in Trichoderma reesei
petitiveness of F. culmorum varies with the competing species (Xu & for the production of cellulases and hemicellulases (Karaffa et al.,
Nicholson, 2009). T. gamsii T6085 and F. oxysporum 7,121 seemed to 2013).
occupy an intermediate position, thus confirming on the one hand Even if referring to a narrow group of isolates, this work shows a
the metabolic versatility of Trichoderma species to adapt to several new protocol for the analysis of data collected by the PM system, al-
environments (Kubicek et al., 2019), and on the other hand, the po- lowing full exploitation of data obtained by this phenotypic approach
tential nutritional competitive ability of isolates belonging to F. oxys- and, at the same time, providing ecological information about what
porum species that are well known to have a saprotrophic capacity, one isolate prefers compared to the others. Many other diseases rely
allowing them to outcompete other organisms on natural substrates on plant debris colonization by the pathogen(s) to guarantee their
such as crop residues (Sarrocco et al., 2019). survival, and residue conservation increases the risk of epidemics
The other two pathogens, F. langsethiae and F. sporotrichioides, for cereal, and specifically wheat, diseases (Kerdraon et al., 2019).
were functionally clustered in the small category, because they were In the context of a biological control strategy, these substrates (pre-
never able to grow fast on any of the substrates included in the FF ferred by the antagonist and not by the pathogens) could be evalu-
microplates. This appears to be in contrast with what has been re- ated as additives to Trichoderma biopreparations in order to improve
ported in the literature, where F. sporotrichioides is described as one competitiveness in the targeted pathogen community (Oszust et al.,
of the other Fusarium species associated with FHB that show a bet- 2020). However, it is fundamental to keep in mind that substrate
ter saprotrophic capacity in soil than F. graminearum (Pereyra and competition is only a part of a much more complex interaction oc-
Dill-­Macky, 2008). curring among organisms that involves many other environmental
Information collected by the second analysis here gave a statis- factors, such as temperature, pH, water availability, as well as other
tically supported comparison of the catabolic ability of each isolate. mechanisms based on the physical and chemical interactions among
Scores calculated here allow for a better appreciation and comparison competitors, as has been demonstrated for T. gamsii T6085 (Sarrocco
of the behaviour of the six fungi on each substrate. This approach also et al., 2019).
gives a tool to rapidly detect how many and which substrates are dif-
ferentially used. The first evident difference in the catabolic ability of C O N FL I C T O F I N T E R E S T
F. graminearum and T. gamsii T6085 can be found for α-­cyclodextrin, None of the authors have present or potential conflicts of interest,
which was well used by the antagonist and scarcely exploited by the nor financial, personal, or other relationships with other persons or
pathogen. Oros et al. (2020) published an investigation of the growth organizations that might inappropriately influence or be perceived
and development of 17 Trichoderma species on α-­ and β-­cyclodextrin, to influence their work.
1156 | LASINIO et al.

Kong, Z., Hart, M. & Hongguang, L. (2018) Paving the way from the lab
DATA AVA I L A B I L I T Y S TAT E M E N T to the field: using synthetic microbial consortia to produce high-­
The data that support the findings of this study are available from quality crop. Frontiers in Plant Science, 9, 1467.
the corresponding author upon reasonable request. Kubicek, C.P., Steindorff, A.S., Chenthamara, K., Manganiello, G.,
Henrissat, B., Zhang, J. et al. (2019) Evolution and comparative ge-
nomics of the most common Trichoderma species. BMC Genomics,
ORCID
20, 485.
Giovanna Jona Lasinio https://orcid.org/0000-0001-8912-5018 Leplat, J., Friberg, H., Abid, M. & Steinberg, C. (2012) Survival of Fusarium
Alessio Pollice https://orcid.org/0000-0002-2818-9373 graminearum, the causal agent of Fusarium head blight. A review.
Livia Pappalettere https://orcid.org/0000-0002-8803-3557 Agronomy for Sustainable Development, 33, 97–­111.
Matarese, F., Sarrocco, S., Gruber, S., Seidl-­Seiboth, V. & Vannacci, G.
Giovanni Vannacci https://orcid.org/0000-0001-8218-6029
(2012) Biocontrol of Fusarium head blight: interactions between
Sabrina Sarrocco https://orcid.org/0000-0002-7080-8369 Trichoderma and mycotoxigenic Fusarium. Microbiology, 158,
98–­106.
REFERENCES Nazari, L., Pattori, E., Somma, S., Manstretta, V., Waalwijk, C., Moretti, A.
et al. (2019) Infection incidence, kernel colonisation, and mycotoxin
Atanasova, L. & Druzhinina, I.S. (2010) Global nutrient profiling by phe-
accumulation in durum wheat inoculated with Fusarium sporotrichi-
notype MicroArrays: a tool complementing genomic and proteomic
oides, F. langsethiae or F. poae at different growth stages. European
studies in conidial fungi. Journal of Zhejiang University Science B, 11,
Journal Plant Pathology, 153, 715–­729.
151–­168.
Niu, B., Wang, W., Yuan, Z., Sederoff, R.R., Sederoff, H., Chiang, V.L. et al.
Balmas, V., Delogu, G., Sposito, S., Rau, D. & Migheli, Q. (2006)
(2020) Microbial interactions within multiple-­strain biological con-
Use of a complexation of tebuconazole with β-­c yclodextrin
trol agents impact soil-­borne plant disease. Frontiers in Microbiology,
for controlling foot and crown rot of durum wheat incited by
11, 2452.
Fusarium culmorum. Journal of Agricultural and Food Chemistry,
Oros, G., Jakab, A. & Cserhati, T. (2020) Utilization of cyclodextrins
54, ​4 80–­4 84.
by Trichoderma species. Open Chemical and Biomedical Methods
Baroncelli, R., Zapparata, A., Piaggeschi, G., Sarrocco, S. & Vannacci, G.
Journal, 3, 90–­97.
(2016) Draft whole-­genome sequence of Trichoderma gamsii T6085,
Oszust, K., Cybulska, J. & Frąc, M. (2020) How do Trichoderma genus fungi
a promising biocontrol agent of Fusarium head blight on wheat.
win a nutritional competition battle against soft fruit pathogens? A
Genome Announcements, 4, e01747-­15.
report on niche overlap nutritional potentiates. International Journal
Bochner, B.R., Gadzinski, P. & Panomitros, E. (2011) Phenotype microAr-
Molecular Science, 21, 4235.
rays for high-­ throughput phenotypic testing and assay of gene
Pereyra, S.A. & Dill-­Macky, R. (2008) Colonization of the residues of
function. Genome Resources, 11, 1246–­1255.
diverse plant species by Gibberella zeae and their contribution to
Brooker, R.W., Callaway, R.M., Cavieres, L.A., Kikvidze, Z., Lortie, C.,
Fusarium head blight inoculum. Plant Disease, 92, 800–­8 07.
Michalet, R. et al. (2009) Don’t diss integration: a comment on
Pinzari, F., Ceci, A., Abu-­Samra, N., Canfora, L., Maggi, O. & Persiani, A.
Ricklefs’s disintegrating communities. The American Naturalist, 174,
(2016) Phenotype MicroArrayTM system in the study of fungal func-
919–­927.
tional diversity and catabolic versatility. Research in Microbiology,
Chiou, J.M. & Li, P.L. (2007) Functional clustering and identifying sub-
167, 710–­722.
structures of longitudinal data. Journal of the Royal Statistical
Sarrocco, S. (2016) Dung-­inhabiting fungi: a potential reservoir of novel
Society: Series B, 69, 679–­699.
secondary metabolites for the control of plant pathogens. Pest
Fitt, B.D., Huang, Y.J., van den Bosch, F. & West, J.S. (2006) Coexistence
Management Science, 72, 643–­652.
of related pathogen species on arable crops in space and time.
Sarrocco, S., Esteban, P., Vicente, I., Bernardi, R., Plainchamp, T.,
Annual Review of Phytopathology, 44, 163–­182.
Domenichini, S. et al. (2020) Straw competition and wheat root
Holomuzki, J.R., Feminella, J.W. & Power, M.E. (2010) Biotic interac-
endophytism of Trichoderma gamsii T6085 as useful traits in the
tions in freshwater benthic habitats. Journal of the North American
biocontrol of Fusarium head blight. Phytopathology, https://doi.
Benthological Society, 29, 220–­244.
org/10.1094/PHYTO​-­09-­20-­0 441-­R
Hoyos-­C arvajal, L. & Bissett, J. (2011) Biodiversity of Trichoderma in
Sarrocco, S., Matarese, F., Moretti, A., Haidukowski, M. & Vannacci,
Neotropics. In: Grillo, O. and Venora, G. (Eds.) The Dynamical
G. (2012) DON on wheat crop residues: effects on mycobi-
Processes of Biodiversity –­Case Studies of Evolution and Spatial
ota as a source of potential antagonists of Fusarium culmorum.
Distribution. InTech, pp. 303–­320.
Phytopathologia Mediterranea, 51, 225–­235.
Jaroszuk-­S cise, J., Tyskiewicz, R., Nowak, A., Ozimek, E., Majewska,
Sarrocco, S., Valenti, F., Manfredini, S., Esteban, P., Bernardi, R., Puntoni,
M., Hanaka, A. et al. (2019) Phytohormones (auxin, gibberellin)
G. et al. (2019) Is exploitation competition involved in a multitrophic
and ACC deaminase in vitro synthesized by the mycoparasitic
strategy for the biocontrol of Fusarium head blight? Phytopathology,
Trichoderma DEMTkZ3A0 strain and changes in the level of auxin
109, 560–­570.
and plant resistance markers in wheat seedlings inoculated with
Sarrocco, S. & Vannacci, G. (2018) Preharvest application of beneficial
this strain conidia. International Journal of Molecular Sciences, 20,
fungi as a strategy to prevent postharvest mycotoxin contamina-
4923.
tion: a review. Crop Protection, 110, 160–­170.
Karaffa, L., Coulier, L., Fekete, E., Overkamp, K.M., Druzhinina, I.S., Mikus,
Scheipl, F., Fahrmeir, L. & Kneib, T. (2012) Spike-­and-­slab priors for func-
M. et al. (2013) The intracellular galactoglycome in Trichoderma re-
tion selection in structured additive regression models. Journal of
esei during growth on lactose. Applied Microbial Biotechnology, 97,
the American Statistical Association, 107, 1518–­1532.
5447–­5456.
Torres, A.M., Palacios, S.A., Yerkovich, N., Palazzini, J.M., Battilani, P.,
Kerdraon, L., Laval, V. & Suffert, F. (2019) Microbiomes and patho-
Leslie, J.F. et al. (2019) Fusarium head blight and mycotoxins in
gen survival in crop residues, an ecotone between plant and soil.
wheat: prevention and control strategies across the food chain.
Phytobiomes Journal, 3, 246–­255.
World Mycotoxins Journal, 12, 333–­335.
Kinkel, L., Schlatter, D., Xiao, K. & Baines, A.D. (2014) Sympatric inhi-
Vaas, L.A.I., Sikorski, J., Michael, V., Goker, M. & Klenk, H.P. (2012)
bition and niche differentiation suggest alternative coevolutionary
Visualization and curve-­ parameter estimation strategies for
trajectories among Streptomycetes. ISME Journal, 8, 249–­256.
LASINIO et al. | 1157

efficient exploration of phenotype microarray kinetics. PLoS One, Zapparata, A., Da Lio, D., Somma, S., Vicente Muñoz, I., Malfatti, L.,
7, e34846. Vannacci, G. et al. (2017) Genome sequence of Fusarium gramin-
Vicente, I., Baroncelli, R., Moran-­Diez, M.E., Bernardi, R., Puntoni, G., earum ITEM 124 (ATCC 56091), a mycotoxigenic plant pathogen.
Hermosa, R. et al. (2020) Combined comparative genomics and Genome Announcements, 5, e01209-­17.
gene expression analyses provide insights into the terpene syn-
thases inventory in Trichoderma. Microorganisms, 8, 1603.
Wirsel, S.G., Runge-­Frobose, C., Ahren, D.G.G., Kemen, E., Oliver, R.P. &
How to cite this article: Lasinio GJ, Pollice A, Pappalettere L,
Mendgen, K.W. (2002) Four or more species of Cladosporium sympatri-
Vannacci G, Sarrocco S. A statistical protocol to describe
cally colonize Phragmites australis. Fungal Genetics and Biology, 35, 99–­113.
Xu, X. & Nicholson, P. (2009) Community ecology of fungal pathogens caus- differences among nutrient utilization patterns of Fusarium
ing wheat head blight. Annual Review of Phytopathology, 47, 83–­103. spp. and Trichoderma gamsii. Plant Pathol. 2021;70:1146–
Yassouridis, C., Ernst, D. & Leisch, F. (2018) Generalization, combination 1157. https://doi.org/10.1111/ppa.13362
and extension of functional clustering algorithms: the R package
funcy. Journal of Statistical Software, 85, 1–­25.