f u n g a l b i o l o g y 1 1 8 ( 2 0 1 4 ) 2 7 7e2 8 6

journa l homepage : www.e lsev ier . com/ loca te / funb io

Fungal endophytes associated with three SouthAmerican Myrtae (Myrtaceae) exhibit preferencesin the colonization at leaf level

Aline B. M. VAZa,b, Andre G. F. C. DA COSTAc, Luc�elia V. V. RAADd,Arist�oteles G�OES-NETOa,b,*aLaborat�orio de pesquisa em Microbiologia (LAPEM), Departamento de Ciencias Biol�ogicas,

Universidade Estadual de Feira de Santana, Feira de Santana, Bahia, BrazilbCentro de Pesquisas Ren�e Rachou (CPqRR), Fundac~ao Oswaldo Cruz (FIOCRUZ), Belo Horizonte, Minas Gerais, BrazilcDepartamento de Estat�ıstica, Universidade Federal de Minas Gerais, Belo Horizonte, Minas Gerais, BrazildInstituto de Economia, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Rio de Janeiro, Brazil

a r t i c l e i n f o

Article history:

Received 27 June 2013

Received in revised form

19 November 2013

Accepted 20 November 2013

Available online 3 December 2013

Corresponding Editor:

Andrew N. Miller

Keywords:

Alternating logistic regression

Clustered responses

Fungal distribution

Endophytic fungi

Statistical modelling

* Corresponding author. Laborat�orio de pesqude Feira de Santana, Feira de Santana, Bahia

E-mail addresses: arigoesneto@pq.cnpq.b1878-6146/$ e see front matter ª 2013 The Bhttp://dx.doi.org/10.1016/j.funbio.2013.11.010

a b s t r a c t

Fungal endophytes associated with Myrtaceae from Brazil and Argentina were isolated at

three levels of nesting: leaf, individual host trees, and site collection. The alternating logis-

tic regression (ALR) was used to model the data because it offers a computationally conve-

nient method for fitting regression structures involving large clusters. The objectives of this

study were to determine: (i) whether the colonization pattern is influenced by environmen-

tal variables, (ii) if there is some leaf part they prefer to colonize; (iii) if there is some fungal

endophyte aggregation between hierarchical levels; (iv) what the distance effect is on the

fungal association. The environmental variables were statistically significant only for Xy-

laria, i.e., when the elevation and water precipitation increase and the temperature de-

creases, the odds ratio of finding another fungal endophyte of that genus previously

found increases. Sordariomycetes, Xylariales, and Xylaria exhibited leaf fragment preference

to petiole and tip. Fungal endophytes showed association within leaf. The horizontal trans-

mission mode and the dispersal limitation may explain this association at the leaf level.

Moreover, our results suggest that when a fungal endophyte infects a leaf or host tree in-

dividual, the odds ratio of dispersal inside them is greater.

ª 2013 The British Mycological Society. Published by Elsevier Ltd. All rights reserved.

Introduction fungal endophytes can be divided in clavicipitaceous and

Fungal endophytes inhabit healthy plant tissues during at

least one stage of their life cycle without causing any appar-

ent symptoms of disease or negative effects on the hosts

(Petrini et al. 1992). According to Rodriguez et al. (2009) the

isa em Microbiologia (LAP44036-900, Brazil. Tel.: þr, arigoesneto@cebio.org,ritish Mycological Societ

nonclavicipitaceous; the first group infects some grasses

and the second one can be recovered from asymptomatic tis-

sues of nonvascular plants, ferns and allies, conifers, and

angiosperms. The nonclavicipitaceous endophytes can be

differentiated into three functional classes based on host

EM), Departamento de Ciencias Biol�ogicas, Universidade Estadual55 (75) 3161 8296; fax: þ55 (75) 3161 8132.arigoesneto@gmail.com (A. G�oes-Neto).y. Published by Elsevier Ltd. All rights reserved.

278 A. B. M. Vaz et al.

colonization patterns, mechanisms of transmission between

host generation, in planta biodiversity levels, and ecological

functions (Rodriguez et al. 2009). Endophytes occurring pri-

marily in aboveground tissues (Class 3 endophytes,

Rodriguez et al. 2009) are horizontally transmitted by spores

and hyphal fragments from plant to plant, by biotic or abi-

otic dispersion agents (Bernstein & Carroll 1977; Bertoni &

Cabral 1988; Saikkonen et al. 1998; Rodriguez et al. 2009).

Moreover, class 3 endophytes form highly localized infec-

tions, have the potential to confer benefits or costs on hosts

that are not necessarily habitat-specific, and are especially

notable for their high diversity within host tissues, plants,

and populations (Saikkonen et al. 1998; Rodriguez et al.

2009). This group is composed predominantly by ascomyco-

tan fungi, and Sordariomycetes, Dothideomycetes, Pezizomycetes,

Leotiomycetes, and Eurotiomycetes are the most frequently iso-

lated classes (Rodriguez et al. 2009). Most of the fungal endo-

phyte did not produce conidia or spores using the

conventional mycological media and the morphological

identification is not possible. For this reason, the internal

transcribed spacer (ITS1 and ITS2) and 5.8S regions of the

nuclear ribosomal repeat unit have been used as the primary

fungal barcode marker to species delimitation (Schoch et al.

2012).

The fungal endophyte composition reflects the interplay

of host species, geographic distance, and climatic factors

(U’ren et al. 2012), with temperature and moisture being

the most important variables for explaining fungal diversity

(Talley et al. 2002). Many host-associated microorganisms ex-

hibit patterns of genetic, morphological, and functional dif-

ferentiation that are related to the distribution of their

hosts (Papke & Ward 2004). Previous studies have shown dif-

ferent patterns on host plant and/or tissue preference by

fungal endophytes varying from high (McKenzie et al. 2000;

Su et al. 2010) to low tissue specificity (Cannon & Simmons

2002).

Multivariate binary responses are common place in eco-

logical studies. Besides, the existence of multiple classes

and levels of nesting in these ecological data is also com-

mon. Repeated measurements within individuals tend to

be more similar than those taken among individuals, and

temporally or spatially proximate measurements are more

similar than temporally or spatially disparate ones (Fieberg

et al. 2009). Traditionally, many of the commonly used statis-

tics models in such studies (mainly regressions) assume in-

dependence among responses (Paradis & Claude 2002),

which is not valid in temporally or spatially autocorrelated

data sets. One example of correlated responses is on longitu-

dinal studies where each subject is followed over a period of

time in which repeated observations of the response and

covariates variables are recorded (Carey et al. 1993). Since re-

peated observations are made on the same subject, observed

responses are generally correlated (Pan & Connett 2002). In

our work, the fungal endophytes were isolated following a hi-

erarchical nesting: fungal endophytes occur in leaf frag-

ments of leaves, which are clustered within individual host

trees that in turn occur in a site collection (Fig 1). In this

case, as there is a hierarchical dependence among the vari-

ables, we can consider that they are correlated. This violates

the basic statistical assumption of independent observations

and requires the use of sophisticated approaches to statisti-

cal modelling.

Marginal models for temporally or spatially autocorre-

lated data sets model the marginal expectation of each bi-

nary variable separately as well as the association between

pairs of responses (Carey et al. 1993). To determine the rela-

tionship among responses with the explanatory variables,

Liang & Zeger (1986) proposed the first generalized estimat-

ing equations (GEE1) approach to model the marginal mean

(mean structure). Unfortunately, the dependence structure

was considered a nuisance (Liang & Zeger 1986; Prentice

1988). However, when the goal is estimating association pa-

rameters, the second-order generalized estimating equations

(GEE2) model gives more efficient results (Liang et al. 1992).

However, GEE2 methods become computationally impracti-

cal if the number of measures within the cluster is large,

e.g., higher than five (Carey et al. 1993). To solve these prob-

lems, the alternating logistic regression (ALR) procedure was

proposed for simultaneously regressing the response on ex-

planatory variables and modelling the association among re-

sponses in terms of the pairwise odds ratio (Carey et al. 1993;

Ananth & Kantor 2004). There may be two objectives for

modelling clustered data that include correlated responses

using ALR (Dobson & Barnett 2002). The first one is to model

the mean structure as a function of covariates, which can be

used for modelling the overall occurrence of fungal endo-

phytes. The second one is to model the dependence struc-

ture among pairs of binary responses, which can be used

for modelling the fungal endophyte association among clus-

tered data (Carey et al. 1993; Ananth & Kantor 2004). The ALR

is formulated based on the odds ratio, which is a particularly

straightforward measure to capture the association between

binary or categorical outcomes (Liang et al. 1992;

Molenberghs & Verbeke 2005).

In this paper, we studied the fungal endophytes atmultiple

hierarchical nesting levels using the ALR analysis. The study

objectives included the following: (i) to assess, using the

mean structure, whether the occurrence of fungal endophytes

at many distinct taxonomic levels is influenced by environ-

mental variables; (ii) to determine whether there is some

leaf part preference in the fungal endophyte colonization;

(iii) to evaluate, using the dependence structure, whether fun-

gal endophytes at many distinct taxonomic levels exhibit as-

sociations within a leaf, individual host tree, and collection

site; and (iv) to appraise whether increasing the distance

among different individual host trees in the same collection

site will decrease the association of distinct taxonomic fungal

groups.

Material and methods

Study areas

Three different sites were studied in Patagonia, Argentina, in

the Andean Patagonian region. In Argentina, Luma apiculata

(Myrtaceae) was collected at two different sites, Arrayanes for-

est (40�5104800S, 71�3605900W) and Puerto Blest (41�0103100S,71�4805800W), and Myrceugenia ovata var. nanophylla (Myrtaceae)

was collected at one site, Espejo lake (40�4101800S, 71�4200000W).

Fig 1 e Fungal endophyte isolation from leaf samples of Luma apiculata, Myrceugenia ovata var. nanophylla, and Eugenia

neomyrtifolia collected in Brazil and Argentina. The hierarchical nesting can be represented by the sets showed, where the

lowest categories are included in the upper category levels successively: {Leaf}3 {Individual host tree}3 {Site collection}. The

number of samples collected at each hierarchical level were: Site collection, Sc[ Sci1,., Sci5; Individual host tree, Ih[ Ihi1,.,

Ihi20; Leaf, L [ Ii1,., Ii5; Leaf fragment, Lf [ Lfi1,., Lfi6.

Preferences in the colonization at leaf level 279

These collection sites are situated in the Nahuel Huapi Na-

tional Park in the municipality of San Carlos de Bariloche,

Argentina. In Brazil, M. ovata var. nanophylla and Eugenia neo-

myrtifolia (Myrtaceae) were collected in two different sites of

an Atlantic rainforest area at the Centro de Pesquisas e Con-

servac~ao da Natureza Pr�o-Mata of PUCRS (29�2804400S,50�1002500W), in S~ao Francisco de Paula, Rio Grande do Sul

state, Brazil.

Fungal endophyte isolation

Five apparently healthy leaves were collected from each of

the 20 trees of all Myrtaceae species that occur in the studied

sites. The trees were spaced approximately 5 m apart. All the

leaves were stored in sterile plastic bags, and fungal isola-

tion was performed on the same day of the collection. The

leaves were surface-sterilized via successive dipping in

70 % ethanol (1 min) and 2 % sodium hypochlorite (3 min),

followed by washing with sterile distilled water (2 min). Af-

ter the leaf surface sterilization, six fragments (approxi-

mately 4 mm2) were cut from each leaf: one from the base

(C, near petiole), two from the middle vein (E and F), one

from the left margin (D), one from the right margin (B), and

one from the tip (A) (six leaf fragments/leaf; 30 leaf frag-

ments/tree; 600 leaf fragments/site; 3000 leaf fragments

overall) (Table 1, Fig 1). All the leaf fragments were plated

onto potato dextrose agar (PDA, Difco, USA) supplemented

with 100 mg ml�1 chloramphenicol (Collado et al. 1996). The

plates were incubated at 15 �C for up to 60 d. To test the ef-

fectiveness of the surface sterilization, 100 ml of the water

used during the final rinse was plated on the PDA to test

for epiphytic microbial contaminants. The binary responses

of the presence/absence of a fungal endophyte were consid-

ered for statistical analysis.

Name assignment of molecular operational taxonomic units(MOTUs)

Pure cultures of the fungal isolates were grouped based on the

following morphological characteristics: (i) form, (ii) size, (iii)

margin/border, (iv) surface, (v) colour, (vi) formation of aerial

mycelium, (vii) opacity, (viii) reverse colour. At least 50 % of

the fungal isolates of each morphospecies were identified by

directly extracting their total genomic DNA and sequencing

the ITS region of the rRNA cistron. The extraction of DNA

from filamentous fungi was performed according to Rosa

et al. (2009). The ITS domains of the rRNA cistron were ampli-

fied using the universal primers ITS1 (50-TCCGTAGGT-GAACCTGCGG-30) and ITS4 (50-TCCTCCGCTTATTGATATGC-30) as described byWhite et al. (1990). ITS amplification and se-

quencingwere performed as described by Vaz et al. (2009). The

ITS sequences obtained were analyzed in GenBank with

BLASTn to search for similarity with the sequences deposited.

MOTUs were defined using a 97 % ITS region identity thresh-

old (Edgar 2010; Sun et al. 2012).

The multivariate model

The environmental variables, by principal component analy-

sis (PCA), and the leaf fragments were considered in the

mean structure in the ALRmethod. A PCA of the environmen-

tal characteristics (elevation, water precipitation, and temper-

ature) at each site collection was performed (Table 2). Leaf

fragments were considered at the mean structure level to

evaluate whether there is some leaf part preference for the

fungal endophyte colonization. The leaf fragment was not

considered at the dependence structure because it was the

smallest sample unit. Thus it was not possible to make repli-

cates, then,when considering the odds ratio of finding a fungal

Table 1 e Description of variables used for theALR statistical analysis.

Covariates Values Description

Collection site 1e5 1 : Arrayanes forest

2 : Puerto Blest

3 : Espejo lake

4 : ProMata

5 : ProMata

Host tree species 1e3 Luma apiculata

Myrceugenia ovata

var. nanophylla

Eugenia neomyrtifolia

Individual host treea 1e20

Leafb 1e5

Fragmentc 1e6 C : Base (near petiole)

E : Middle vein upper

F : Middle vein lower

D : Left margin

B : Right margin

A : Tip

Fungal endophyte

numberdVaries

a Sampling was performed following a transect and each individ-

ual host tree was spaced approximately 5m apart. A total of 100 in-

dividuals from all host tree species was collected.

b Five apparently healthy leaves were sampled randomly from

each individual host tree resulting in a total of 500 leaves.

c Six fragments were cut from each leaf and it was obtained 3000

leaf fragments overall.

d The fungal endophyte number isolates vary among the different

species identified.

280 A. B. M. Vaz et al.

endophyte in one fragment, the comparison is performed

with the odds ratio in another leaf fragment.

In the dependence structure, the endophyte association at

three distinct levels of nesting was considered (from the least

inclusive until the most inclusive): leaf, individual host tree,

and site collection. Distance was also considered at the depen-

dence structure to represent how far each individual host tree

was from each other in a same collection site. These variables

were considered at the dependence structure because they

have dependence in their occurrence, i.e., one leaf was col-

lected fromone specific individual host tree in a specific collec-

tion site ({Leaf}3 {Individualhost tree}3 {Site collection}, Fig 1).

The firstmodel (1) was used to estimate the influence of en-

vironmental variables and leaf part preference in fungal endo-

phyte occurrence at the mean structure whereas the second

model (2) was used to estimate the association among the

levels of nesting.

(1) Logit Pr (Y ¼ 1) ¼ b0 þ b1I (PCA) þ b2I (leaf

fragment ¼ A) þ b3I (leaf fragment ¼ B) þ b4I (leaf

Table 2 e PCA among environmental characteristics of each si

Components Elevation Waterprecipitation

Tempera

PC1 0.709 0.623 �0.33

PC2 �0.007 0.475 0.88

PC3 �0.705 0.621 �0.34

fragment ¼ D) þ b5I (leaf fragment ¼ E) þ b6I (leaf

fragment ¼ F)

(The C fragment was considered at the intercept.)

(2) LogOR (Yj, Yk)¼a1I (collection site) þ a2 (distance jk)

(if j and k are different host trees in the same collec-

tion site)

a1I (collection site) þ a3I (individual host tree)

(if j and k are different leaves on the same host tree)

a1I (collection site) þ a3I (individual host tree) þ a4I (leaf)

(if j and k are different leaf fragments on the same

leaf)

The mean and dependence structures in our study were

modelled with the marginal odds ratio (Lipsitz et al. 1991).

The association parameters were estimated by the odds ratio

of finding another fungal endophyte of the same taxonomic

level previously found at each one of the levels of nesting

(leaf, individual host tree, and site collection). A significantly

odds ratio suggests the presence of the aggregation of fungal

endophytes. The distance reflects the odds ratio of finding an-

other fungal endophyte of the same taxonomic level previ-

ously found when comparing two individual host trees

spaced 1 m from each other. Despite the distance between

each individual host tree species was approximately 5 m,

the basic unit of distance for this statistical analysis is 1 m.

The ALR analysis was performed using the ordgee function

from the geepack package (Yan 2002) and the PCA analysis us-

ing prcomp function. All the analyses were carried out using

the R program (R Development Core Team 2012) and the R

script is available in Supplementary material.

Results

A total of 939 fungal endophyte isolates were obtained from

3000 leaf fragments and were identified in 51 distinct MOTUs.

Almost all taxa belonged to Ascomycota, with only Trametes be-

longing to Basidiomycota. The most frequent fungal endo-

phytes recovered belonged to Sordariomycetes and

Dothideomycetes, accounting for 54.8 % and 38.9%, respectively.

Fungal endophytes belonging to the order Xylariales were the

only taxon isolated from all host trees (Table 3).

Figs 2 and 3 represent the probability of isolating one fun-

gal endophyte belonging to their respective taxonomic group

according to the environmental variables variation and leaf

fragment. In Fig 2, the ordinate axis represents the mean per-

centage of isolating a fungus belonging to the distinct fungal

taxonomic groups according to the first principal component

(PC1) from PCA analysis (abscissa axis). The environmental

variables used in this study (temperature, elevation, and

te collection.

ture Standarddeviation

Proportionof variance

Cumulativeproportion (%)

0 1.338 0.597 59.7

0 1.005 0.336 93.3

1 0.448 0.066 100

Table 3 e List of identified fungal endophytes from host tree species of Luma apiculata, Myrceugenia ovata var. nanophylla,and Eugenia neomyrtifolia from Andean Patagonian forest (Argentina) and Atlantic rainforest (Brazil).

Identification L1 L2 M1 M2 E1 Total

Sordariomycetes, Coniochaetales

Coniochaeta velutina [JQ346221] 1 1

Sordariomycetes, Diaporthales

Amphilogia sp. [JQ346197] 6 6

Diaporthe sp. 1 [JQ327869] 2 2 4

D. helianthi [JQ346194] 4 4

Diaporthe sp. 2 [JQ327871] 2 2

Diaporthe sp. 3 [JQ327870] 3 7 10

Diaporthe sp. 4 [JQ327872] 11 11

Diaporthe sp. 5 [JQ327871] 10 10

D. phaseolorum [JQ327873] 5 5

Diaporthe sp. 6 [JQ327874] 16 8 24

Diaporthe stewartii 6 6

Greeneria sp. 1 [JQ346195] 34 34

Sordariomycetes, Glomerellales

Colletotrichum sp. 1 [JQ346206] 23 23

Colletotrichum boninense [JQ346207] 5 5

Colletotrichum sp. 2 [JQ346208] 7 7

Colletotrichum sp. 3 [JQ346209] 9 9

Colletotrichum sp. 4 [JQ346210] 20 20

Colletotrichum sp. 5 [JQ346211] 5 5

Colletotrichum sp. 6 [JQ346212] 22 22

Colletotrichum sp. 7 [JQ346213] 12 12

Colletotrichum sp. 8 3 3

Sordariomycetes, Hypocreales

Cephalosporium sp. [JQ346222] 6 6

Sordariomycetes, Xylariales

Annulohypoxylon sp. 3 [JQ327866] 2 2

Annulohypoxylon sp. 1 [JQ327864] 3 3

Annulohypoxylon sp. 2 [JQ327865] 13 13

Annulohypoxylon sp. 4 [JQ327867] 1 1

Biscogniauxia sp. 1 [JQ327868] 1 1

Nemania sp. [JQ327862] 3 27 30

Xylaria berteri [JQ327861] 33 17 50

Xylaria castorea [JQ327858] 2 5 7

Xylaria sp. 1 [JQ327859] 9 112 11 132

Xylaria enteroleuca [JQ327860] 46 46

Dothideomycetes, Botryosphaeriales

Guinardia sp. 1 [JQ346219] 1 1

Guinardia mangiferae 37 21 58

Dothideomycetes, Dothideales

Dothiora cannabinae [JQ346227] 4 4

Dothideomycetes, Capnodiales

Cladosporium subtilissimum [JQ346203] 2 6 8

Cladosporium colombiae [JQ346204] 1 2 3

Mycosphaerella sp. [JQ346202] 192 192

Pseudocercospora basintrucata [JQ346205] 72 18 1 91

Dothideomycetes, Pleosporales

Camarosporium brabeji [JQ346215] 1 1

Didymella sp. [JQ346225] 1 1

Lewia infectoria [JQ346214] 8 8

Microsphaeropsis olivacea [JQ346217] 1 1

Paraconiothyrium sp. [JQ346216] 1 2 3

Leotiomycetes, Helotiales

Cryptosporiopsis actinidiae [JQ346199] 1 1

Mollisia cinerea [JQ346201] 1 1

Pezicula corylina [JQ346198] 34 34

(continued on next page)

Preferences in the colonization at leaf level 281

Table 3 e (continued )

Identification L1 L2 M1 M2 E1 Total

Eurotiomycetes, Eurotiales

Penicillium restrictum [JQ346224] 1 1

Pezizomycetes, Pezizales

Peziza sp. [JQ346218] 1 1

Fungi incertae sedis, Mortierellales

Mortierella sclerotiella [JQ346223] 1 1

Basidiomycota

Trametes hirsuta [JQ346220] 15 15

L1: Luma apiculata (Arrayanes forest, Argentina); L2: Luma apiculata (Puerto Blest, Argentina); M1: Myrceugenia ovata (Espejo lake, Argentina); M2:

Myrceugenia ovata (ProMata, Brazil); E1: Eugenia neomyrtifolia (ProMata, Brazil).

282 A. B. M. Vaz et al.

water precipitation) did not vary inside each site collection,

and thus each point in the figure represents the PCA rotation

mean values obtained for each site collection: In Argentina;

Luma apiculata in Arrayanes forest (�0.63, first point), L. apicu-

lata Puerto Blest (0.36, fourth point), Myrceugenia ovata var.

nanophylla Espejo lake (�0.59, second point) and in Brazil: M.

ovata var. nanophylla ProMata (2.36, fifth point), Eugenia neomyr-

tifolia ProMata (�0.49, third point). The comparison at fungal

class level showed that there was no overlapping in their oc-

currence, with Sordariomycetes more frequently associated

with the tropical host tree species E. neomyrtifolia andM. ovata

Fig 2 e The graphs represent the probability of isolating one

fungal endophyte belonging to its respective taxonomic

group (Sordariomycetes, Dothideomycetes, Xylariales, Capno-

diales or Xylaria) considering the first principal components

(PC1) of the PCA of the environmental variables. Each point

refers to an each site collection: In Argentina; Luma apiculata

in Arrayanes forest (L0.63, first point), Luma apiculata Pu-

erto Blest (0.36, fourth point), Myrceugenia ovata var. nano-

phylla Espejo lake (L0.59, second point) and in Brazil:

Myrceugenia ovata var. nanophylla ProMata (2.36, fifth point),

Eugenia neomyrtifolia ProMata (L0.49, third point).

var. nanophylla from Brazil, and Dothideomycetes more fre-

quently associated with the temperate host tree species L. api-

culata from Puerto Blest and Arrayanes forest. This same

pattern maintained when considered the lowest taxonomic

level order and genus. The only exception happened with M.

ovata var. nanophylla from Espejo lake (Argentina), which pre-

sented similar occurrence of Xylariales, Xylaria, and Dothideo-

mycetes fungi.

The ordinate axis in Fig 3 represents the mean percentage

of isolating a fungus belonging to the distinct fungal taxo-

nomic groups according to the leaf fragment (abscissa axis).

The leaf fragments A, C, and F exhibited the highest percent-

age of fungal endophytes belonged to the Sordariomycetes,

Xylariales, and Xylaria. This pattern, however, was not ob-

served for Dothideomycetes and Capnodiales.

Fig 3 e The graphs represent the probability of isolating one

fungal endophyte belonged to its respective taxonomic

group (Sordariomycetes, Dothideomycetes, Xylariales, Capno-

diales or Xylaria) considering the leaf fragment: near petiole

(C), middle vein (E and F), left margin (D), right margin (B),

and tip (A).

Fig 4 e The barplots represent the probability (ordinate axis) of isolating two fungal endophytes belonging to their respective

taxonomic group (Sordariomycetes, Dothideomycetes, Xylariales, Capnodiales or Xylaria) in the same individual host tree and

leaf (abscissas axis).

Fig 5 e The graphs represent the probability (ordinate axis)

of isolating two fungal endophytes belonging to their re-

spective taxonomic group (Sordariomycetes, Dothideomycetes,

Xylariales, Capnodiales or Xylaria) with increasing distance

between two individual host trees.

Preferences in the colonization at leaf level 283

Figs 4 and 5 represent the dependence structure and

depict presence/absence of fungal endophyte obtained from

the same site collection. In Fig 4, the ordinate axis represents

the probability of finding or not (yes or no) another fungal en-

dophyte of the same taxonomic level previously found when

comparing the same individual host tree and leaf (abscissa

axis). The probability was higher in the same leaf than in

the same individual host tree. Fig 5 shows the probability of

finding another fungal endophyte of the same taxonomic level

previously found when increasing the distance among differ-

ent individual host trees in the same site collection. When in-

creasing the distance between two host trees at the same site

collection, there is no significant decrease or increase in the

probability (Pr [Yj ¼ 1; Yk ¼ 1]).

The ALR models were analyzed at all fungal taxonomic

levels, by species, genus, family, order, class and the models

converged to Sordariomycetes, Xylariales, Xylaria, Dothideomy-

cetes, and Capnodiales (Table 4), whichmeans that the program

generated results only for these fungal groups. Xylariawas the

only genus isolated from all the host plants studied. Many

models that consider species, genus, family, order, and class

do not converge because the fungal endophytes belonging to

those taxonomic levels were isolated at only one collection

site and/or at low frequency.

Only the first PCA was considered in the mean structure,

explaining nearly 60 % of the variance (Table 3), and it was

Table

4e

TheALRstatisticalanalysisco

nsideringth

efu

ngalendophyte

levels

ofgenus,

ord

er,andclass

isolatedfrom

Lumaapiculata,M

yrceu

genia

ovata

var.nanop

hylla,

andEugen

ianeo

myrtifoliafrom

theAndeanPatagonianforest

(Arg

entina)andth

eAtlanticra

inforest

(Bra

zil).

Sordariom

ycetes

Xylariales

Xylaria

Dothideomycetes

Capn

odiales

P-value

O.R

I.L.

S.L.

P-value

O.R

I.L.

S.L.

P-value

O.R

I.L.

S.L.

P-value

O.R

I.L.

S.L.

P-value

O.R

I.L.

S.L.

Mea

nstru

cture

Intercept(b)

0.001

1e

e<0.001

1e

e<0.001

1e

e<0.001

1e

e<0.001

1e

e

PCA1(b

1)

0.091

1.34

0.95

1.87

0.363

0.88

0.66

1.16

0.003

1.50

1.15

1.95

0.942

1.02

0.64

1.62

0.798

1.08

0.60

1.95

Fr¼

A0.082

0.60

0.34

1.07

0.000

0.64

0.23

0.80

0.000

0.61

0.48

0.79

0.592

1.25

0.56

2.79

0.193

1.65

0.78

3.50

Fr¼

B0.000

0.40

0.26

0.61

0.000

0.45

0.00

0.64

0.000

0.46

0.32

0.66

0.839

1.09

0.48

2.47

0.141

1.54

0.87

2.75

Fr¼

D0.000

0.31

0.21

0.44

0.000

0.37

0.00

0.53

0.000

0.36

0.24

0.54

0.364

1.43

0.66

3.13

0.073

1.95

0.94

4.03

Fr¼

E0.000

0.49

0.36

0.66

0.000

0.50

0.00

0.73

0.001

0.46

0.30

0.72

0.452

1.29

0.66

2.53

0.121

1.62

0.88

2.98

Fr¼

F0.006

0.50

0.31

0.82

0.000

0.71

0.00

0.83

0.000

0.66

0.53

0.83

0.637

1.25

0.50

3.12

0.161

1.76

0.80

3.88

Dep

enden

cestru

cture

Intercept

(Siteco

llectiona1)

0.245

1.81

0.67

4.93

0.848

1.40

0.04

44.98

0.875

1.25

0.07

21.22

0.589

2.01

0.16

25.42

0.564

2.73

0.09

83.30

Distance

(a2)

0.262

0.99

0.98

1.00

0.444

1.01

0.99

1.03

0.915

1.00

0.99

1.01

0.389

1.00

1.00

1.01

0.311

1.01

1.00

1.02

Individualhost

(a3)

0.009

1.94

1.18

3.20

0.289

5.59

0.23

134.48

0.136

4.80

0.61

37.71

0.129

3.52

0.69

17.92

0.155

3.90

0.60

25.42

Leaf(a

4)

0.000

2.45

2.04

2.93

0.108

2.64

0.81

8.63

0.003

2.66

1.41

5.03

0.002

3.07

1.52

6.22

0.008

3.46

1.38

8.67

O.R.:Oddsratio;I.L.:Inferiorlimit;S.L.:Superiorlimit;Fr:

Leaffragment.Significa

ntvaluesin

bold.

284 A. B. M. Vaz et al.

statistically significant only forXylaria (odds ratio 1.5, P¼ 0.003;

Table 4). In other words, when the elevation and water precip-

itation increase and the temperature decreases, the odds ratio

of finding another fungal endophyte of the genus Xylaria previ-

ously found increases. The second variable considered in the

mean structure was the leaf fragment. The odds ratios of the

leaf fragment were statistically significant for Sordariomycetes,

Xylariales, and Xylaria and ranged from 0.31 (0.21e0.44) to 0.71

(0.00e0.83).However, nodifferences in thedistributionofDothi-

deomycetes and Capnodiales were found. At the dependence

structure, the odds ratios, in general, were statistically signifi-

cant at the leaf level formostof thegroups, except forXylariales,

and the individualhost tree levelwassignificantonly forSordar-

iomycetes. The site collection and distance among host trees

were not statistically significant for any group (Table 4).

Discussion

Fifty-five distinct MOTUs were identified in this study based

on sequencing of the ITS region of rRNA,which is the accepted

fungal DNA barcode region (Schoch et al. 2012). Many works

have reported that fungal endophytes within Ascomycota en-

compass the majority of foliar fungal endophyte species

(Arnold 2007; Arnold & Lutzoni 2007; Johnston et al. 2012;

Vaz et al. 2012; Langenfeld et al. 2013). Sordariomycetes and

Xylariales are considered the dominant groups associated

with tropical hosts and the secondmost frequently fungal en-

dophyte group recovered belonged to Dothideomycetes (Arnold

& Lutzoni 2007; Higgins et al. 2007). This same pattern was ob-

served in our study since the fungal endophytes belonging to

those groups were the only ones isolated from all host tree

species in all collection sites. The affinity between fungal en-

dophyte and hosts varies according to the substrate chemis-

try, which influences the outcomes of interactions

(biochemical inhibition, competition, avoidance, no response)

(Arnold et al. 2003; Rajala et al. 2013). There is no overlapping in

the curves representing the probability of isolating one fungal

endophyte belonging to its respective taxonomic group (Fig 2),

suggesting some host specificity in the fungal endophyte col-

onization. This host specificity may be caused by the host leaf

chemistry and the biological interactions.

The pattern of finding a higher probability of infection near

the petiole than in more distal leaf fragments has previously

been demonstrated (Bernstein & Carroll 1977; Wilson &

Carroll 1994; Cannon & Simmons 2002). According to Wilson

& Carroll (1994), the petiole end of the leaf expands less than

the more distal two-thirds of the leaf. Consequently, infec-

tions that are established in the more distal leaf parts before

the leaf is fully expanded will become ‘diluted’ as leaf blade

expands more compared with the petiole segments. The re-

sult is a leaf with more dense infections at the petiole end

(Wilson & Carroll 1994). Considering our results, we suggest

that Sordariomycetes, Xylariales, and Xylaria most likely

exhibited some preference for leaf tissue colonization (Fig 3,

Table 4). Probably, the highest odds radios found near the pet-

iole and gradually decreasing towards the tip can be explained

by the leaf developmental pattern and reflect the ‘dilution’ of

the fungal endophyte community. The somewhat greater col-

onization of the leaf tip region compared with the midrib and

Preferences in the colonization at leaf level 285

margin fragments could possibly be explained by the leaf tip

being more prone to colonization, as rainwater draining off

at the apex would tend to wash unestablished fungal propa-

gules on the leaf surface towards the leaf tips (Wilson &

Carroll 1994). However, this cannot be considered to explain

the Dothideomycetes and Capnodiales distributions because

there is no statistically significant leaf fragment colonization

pattern, suggesting that apparently there is no preference in

the leaf fragment colonization for these groups.

The endophytes of woody plants are horizontally trans-

mitted by hyphal fragmentation and/or spores from plant

to plant (Faeth & Hammon 1997; Arnold et al. 2000; Arnold

2005) and may be released passively by herbivores or physical

agents such as wind or rain (Rodriguez et al. 2009). Thus, the

fungal endophyte colonization depends on the availability

and viability of fungal propagules in the surrounding envi-

ronment (Schulz & Boyle 2005). This mode of transmission

may explain the association observed at the leaf and individ-

ual host tree levels for some taxonomic groups (Fig 4). How-

ever, it was observed that the distance between host trees

inside each site collection was not statistically significant

(Fig 5 and Table 4). This result corroborates that the dispersal

limitation is an important factor in explaining the biogeo-

graphic pattern of fungal endophytes (Hubbell 2001; Martiny

et al. 2011).

Moreover, we observed that the fungal colonization

exhibited a statistically significant association only at the

leaf and individual host tree level (only for Sordariomycetes).

Because one fungal endophyte can infect a leaf/individual,

the odds ratio of dispersal inside the same leaf/individual is

greater.We applied the ALR analysis , which offers a computa-

tionally convenient method for modelling dependence clus-

tered data. To date, this is the first work that uses ALR for

analyzing fungal endophyte communities. The occurrence of

Xylaria was influenced by elevation, temperature, and water

precipitation. Sordariomycetes, Xylariales, and Xylaria exhibited

leaf fragment preference to petiole and tip. Moreover, there is

a fungal endophyte association mainly within a leaf and no

significant association of the distinct taxonomic fungal groups

increasing the distance between host trees in each site collec-

tion. These results suggest that when a fungal endophyte in-

fects a leaf or individual host tree, the odds ratio of dispersal

inside them is greater. The next steps in our research program

will be to elucidate the fungal endophyte coevolution with

host trees by analyzing and comparingmolecular phylogenies

from studied fungal groups and their corresponding host trees

species.

Acknowledgement

Aline B.M.Vaz received a postdoctoral scholarship from CNPq

(the Conselho Nacional de Desenvolvimento Cient�ıfico e Tec-

nol�ogico of Brazil).

Appendix A. Supplementary data

Supplementary data related to this article can be found at

http://dx.doi.org/10.1016/j.funbio.2013.11.010.

r e f e r e n c e s

Ananth CV, Kantor ML, 2004. Modeling multivariate binary re-sponses with multiple levels of nesting based on alternatinglogistic regressions: an application to caries aggregation.Journal of Dental Research 83: 766e781.

Arnold EA, 2005. Diversity and ecology of fungal endophytes intropical forests. In: Deshmukh S (ed.), Current Trends in Myco-logical Research. Oxford & IBH Publishing Co. Pvt. Ltd., NewDelhi, pp. 49e68.

Arnold EA, 2007. Understanding the diversity of foliar endophyticfungi: progress, challenges, and frontiers. Fungal Biology Review21: 51e66.

Arnold AE, Maynard Z, Gilbert GS, Coley PD, Kursar TA, 2000. Aretropical endophytes hyperdiverse? Ecology Letters 3: 267e274.

Arnold AE, Mej�ıa LC, Kyllo D, Rojas EI, Maynard Z, Robbins N,Herre EA, 2003. Fungal endophytes limit pathogen damage ina tropical tree. Proceedings of the National Academy of Sciences ofthe United States of America 26: 15649e15654.

Arnold EA, Lutzoni F, 2007. Diversity and host range of foliarfungal endophytes: are tropical leaves biodiversity hotspots?Ecology 88: 541e549.

Bernstein ME, Carroll GC, 1977. Internal fungi in old-growthDouglas fir foliage. Canadian Journal of Botany 55: 644e653.

Bertoni MD, Cabral D, 1988. Phyllosphere of Eucalyptus viminalis. IIDistribution of endophytes. Nova Hedwigia 46: 491e502.

Cannon PF, Simmons CM, 2002. Diversity and host preference ofleaf endophytic fungi in the Iwokrama Forest Reserve,Guyana. Mycologia 94: 210e220.

Carey V, Zeger SL, Diggle P, 1993. Modelling multivariate binarydata with alternating logistic regressions. Biometrika 80:517e526.

Collado J, Platas G, Gonz�alez I, Pel�aez F, 1996. Geographical andseasonal influences on the distribution of fungal endophytesin Quercus ilex. New Phytologist 144: 525e532.

Dobson AJ, Barnett A, 2002. An Introduction to Generalized LinearModels, 3rd edn. Chapman & Hall/CRC, London.

Edgar RC, 2010. Search and clustering orders of magnitude fasterthan BLAST. Bioinformatics 26: 2460e2461.

Faeth SH, Hammon KE, 1997. Fungal endophytes in oak trees:long-term patterns of abundance and associations with leaf-miners. Ecology 78: 810e819.

Fieberg J, Rieger RH, Zicus MC, Schildcrout JS, 2009. Regressionmodeling of correlated data in ecology: subject-specific andpopulation averaged response patterns. Journal of AppliedEcology 46: 1018e1025.

Higgins KL, Arnold AE, Miadlikowska J, Sarvate SD, Lutzoni F, 2007.Phylogenetic relationships, host affinity, and geographicstructure of boreal and arctic endophytes from three majorplant lineages.Molecular Phylogenetics and Evolution 42: 543e555.

Hubbell SP, 2001. The Unified Neutral Theory of Biodiversity and Bio-geography. Princeton University Press, NJ, USA.

Johnston PR, Johansen RB, Williams AFR, Paula Wikie J, Park D,2012. Patterns of fungal diversity in New Zealand Nothofagusforests. Fungal Biology 116: 401e412.

Langenfeld A, Prado S, Nay B, Cruaud C, Lacoste S, Bury E,Hachette F, Hosoya T, Dupont J, 2013. Geographic localitygreatly influences fungal endophyte communities in Cephalo-taxus harringtonia. Fungal Biology 117: 124e136.

Liang KY, Zeger SL, 1986. Longitudinal data analysis using gen-eralized linear models. Biometrika 73: 13e22.

Liang KY, Zeger SL, Qaqish BF, 1992. Multivariate regressionanalyses for categorical data. Journal of the Royal Statistical So-ciety 54: 3e40.

Lipsitz SR, Laird NM, Harrington DP, 1991. Generalized estimatingequations for correlated binary data: using the odds ratio asa measure of association. Biometrika 78: 153e160.

286 A. B. M. Vaz et al.

McKenzie EHC, Buchanan OK, Johnston PR, 2000. Checklist offungi on Nothofagus species in New Zeland. New Zeland Journalof Botany 38: 635e720.

Martiny JBH, Eisen JA, Penn K, Allison SD, Horner-Devine C, 2011.Drivers of bacterial b-diversity depend on spatial scales. Pro-ceedings of the National Academy of Sciences of the United States ofAmerica 108: 7850e7854.

Molenberghs G, Verbeke G, 2005. Model for Discrete LongitudinalData, 2nd edn. Springer Science, USA.

Pan W, Connett JE, 2002. Selecting the working correlationstructure in generalized estimating equations with applicationto the lung health study. Statistica Sinica 12: 475e490.

Papke RT, Ward DM, 2004. The importance of physical isolation tomicrobial diversification. FEMSMicrobiology Ecology48: 293e303.

Paradis E, Claude J, 2002. Analysis of comparative data usinggeneralized estimating equations. Journal of Theoretical Biology218: 175e185.

Petrini O, Sieber TN, Toti L, Viret O, 1992. Ecology, metaboliteproduction, and substrate utilization in endophytic fungi.Natural Toxins 1: 185e196.

Prentice RL, 1988. Correlated binary regression with covariatesspecific to each binary observation. Biometrics 44: 33e48.

R Development Core Team, 2012. R: a language and environment forstatistical computing.

Rajala T, Velmala SM, Tuomivirta T, Haapanen M, M€uller M,Pennanen T, 2013. Endophyte communities vary in the nee-dles of Norway spruce clones. Fungal Biology 117: 182e190.

Rosa LH, Vaz ABM, Caligiorne RB, Campolina S, Rosa CA, 2009.Endophytic fungi associated with the Antarctic Grass De-schampsia antarctica Desv. (Poaceae). Polar Biology 32: 161e167.

Rodriguez RJ, White Jr JF, Arnold EA, Redman RS, 2009. Fungalendophytes: diversity and functional roles. New Phytologist182: 314e330.

Saikkonen K, Faeth SH, Helander M, Sullivan TJ, 1998. Fungalendophytes: a continuum of interactions with host plants.Annual Review of Ecology and Systematics 29: 319e343.

Schoch CL, Seifert KA, Huhndorf S, Robert V, Spouge JL,Levesque CA, Chen WFungal Barcoding Consortium, 2012.

Nuclear ribosomal internal transcribed spacer (ITS) region asa universal DNA barcode marker for Fungi. Proceedings of theNational Academy of Sciences of the United States of America 109:6241e6246.

Schulz B, Boyle C, 2005. The endophytic continuum. MycologicalResearch 109: 661e686.

Su YY, Guo LD, Hyde KD, 2010. Response of endophytic fungiof Stipa grandis to experimental plant function group re-moval in Inner Mongolia steppe, China. Fungal Diversity 43:93e101.

Sun X, Ding Q, Hyde KD, Guo LD, 2012. Community structure andpreference of endophytic fungi of three woody plants ina mixed forest. Fungal Ecology 5: 624e632.

Talley SM, Coley PD, Kursar TA, 2002. The effects of weather onfungal abundance and richness among 25 communities in theIntermountain West. BMC Ecology 2: 1e7.

U’ren JM, Lutzoni F, Miadlikowska J, Laetsch AD, Arnold AE, 2012.Host and geographic structure of endophytic and endolichenicfungi at a continental scale. American Journal of Botany 99:898e914.

Vaz ABM, Brand~ao LR, Vieira MLA, Pimenta RS, Morais PB,Sobral MEG, Rosa LH, Rosa CA, 2012. Diversity and antimi-crobial activity of fungal endophyte communities associatedwith plants of Brazilian savanna ecosystems. African Journal ofMicrobiology Research 6: 3173e3185.

Vaz ABM, Mota RC, BomfimMRQ, Zani CL, Rosa CA, Rosa LH, 2009.Antimicrobial activity of endophytic fungi associated withOrchidaceae in Brazil. Canadian Journal of Microbiology 55:1381e1391.

White TJ, Bruns T, Lee S, Taylor J, 1990. Amplification and directsequencing of fungal ribosomal RNA genes for phylogenetics.In: Innis MA, Gelfand DH, Sninsky JJ, White TJ (eds), PCR Pro-tocols: a guide to methods and applications. Academic Press, SanDiego, CA, USA, pp. 315e322.

Wilson D, Carroll GC, 1994. Infection studies of Discula quercina, anendophyte of Quercus garryana. Mycologia 86: 635e647.

Yan J, 2002. Geepack: yet another package for generalized esti-mating equations. R News 2: 2e14.