The ambrosia symbiosis is specific in some species and promiscuous in others

13 Pages·2014·1.62 MB·English

by  Martin Kostovcik

TheISMEJournal(2015)9,126–138
&2015InternationalSocietyforMicrobialEcology Allrightsreserved 1751-7362/15
www.nature.com/ismej
ORIGINAL ARTICLE
The ambrosia symbiosis is specific in some species
and promiscuous in others: evidence from
community pyrosequencing
Martin Kostovcik1,2,3, Craig C Bateman4, Miroslav Kolarik3,5, Lukasz L Stelinski6,
Bjarte H Jordal7 and Jiri Hulcr1,4
1School of Forest Resources and Conservation, University of Florida, Gainesville, FL, USA; 2Department of
Genetics and Microbiology, Faculty of Science, Charles University in Prague, Praha 2, Czech Republic;
3Institute of Microbiology AS CR, Praha 4, Czech Republic; 4Department of Entomology and Nematology,
University of Florida, Gainesville, FL, USA; 5Department of Botany, Faculty of Science, Charles University,
Praha 2, Czech Republic; 6Department of Entomology and Nematology, Citrus Research and Education
Center, University of Florida, Lake Alfred, FL, USA and 7Natural History Museum, University Museum of
Bergen, University of Bergen, Bergen, Norway
Symbioses are increasingly seen as dynamic ecosystems with multiple associates and varying
fidelity. Symbiont specificity remains elusive in one of the most ecologically successful and
economicallydamagingeukaryoticsymbioses:theambrosiasymbiosisofwood-boringbeetlesand
fungi. We used multiplexed pyrosequencing of amplified internal transcribed spacer II (ITS2)
ribosomal DNA (rDNA) libraries to document the communities of fungal associates and symbionts
inside the mycangia (fungus transfer organ) of three ambrosia beetle species, Xyleborus affinis,
Xyleborus ferrugineus and Xylosandrus crassiusculus. We processed 93 beetle samples from 5
locations across Florida, including reference communities. Fungal communities within mycangia
included 14–20 fungus species, many more than reported by culture-based studies. We recovered
previously known nutritional symbionts as members of the core community. We also detected
severalotherfungaltaxathatareequallyfrequentbutwhosefunctionisunknownandmanyother
transient species. The composition of fungal assemblages was significantly correlated with beetle
species but not with locality. The type of mycangium appears to determine specificity: two
Xyleborus with mandibular mycangia had multiple dominant associates with even abundances;
Xylosandrus crassiusculus (mesonotal mycangium) communities were dominated by a single
symbiont,Ambrosiellasp.Beetlemycangiaalsocarriedmanyfungifromtheenvironment,including
plant pathogens and endophytes. The ITS2 marker proved useful for ecological analyses, but the
taxonomicresolutionwaslimitedtofungalgenusorfamily,particularlyinOphiostomatales,which
areunder-representedinourampliconsaswellasinpublicdatabases.Thisinitialanalysisofthree
beetlespeciessuggeststhateachcladeofambrosiabeetlesandeachmycangiumtypemaysupport
afunctionally andtaxonomically distinct symbiosis.
The ISME Journal(2015)9, 126–138;doi:10.1038/ismej.2014.115;published online1August2014
Introduction and microbes (Gilbert et al., 2012). The improved
understandingispartlyaresultoftheavailabilityof
Some symbioses conform to the traditional one-on-
high-throughput,culture-independenttechniquesof
one mode of symbiotic relationships, but it is
communitysurveysandpartlyaresultofmaturation
increasingly understood that many symbioses are
of the field of symbiology. At the same time, it is
more open and consist of multiple associates with
becoming clear that not all microbes discoverable
varyingdegreeoffidelity(Calderaetal.,2009).This
withinanimalsandplantsbymoderntechniquesare
is particularly true for symbioses between animals
symbionts. Contemporary understanding of sym-
bioses emphasizes the shared evolutionary and
ecologicaldynamicsofcommunitieswherecoopera-
Correspondence: J Hulcr, School of Forest Resources and
Conservation and Department of Entomology and Nematology, tion among members prevails over antagonism
University of Florida, PO Box 110410, Gainesville, FL 32611, (Douglas, 2010).
USA. One of the most common symbioses in any forest
E-mail:[email protected]
ecosystem occurs between wood-boring insects and
Received3February2014;revised12April2014;accepted4June
2014;publishedonline1August2014 fungi. The two partners have complementary
Pyrosequencinganalysisofambrosiacommunities
MKostovciketal
127
qualities: the insects fly and can mechanically transcribedspacer(ITS)rDNAregion—isdifficultto
colonize wood, while their fungal partners concen- use in many ambrosia fungus groups. This region is
trate nutrients from the plant tissue and deliver extremely GC-rich and difficult to amplify and
them to the insect in the form of specialized sequence in many Ophiostomatales (Fraedrich
mycelium. Wood is one of the world’s largest stores et al., 2008). In other groups, such as Ambrosiella
of biomass, making this symbiosis one of the most (Microascales), ITS sequences are severely under-
ecologicallysuccessfulandeconomicallyimportant. represented or incorrectly identified in public
The most advanced stage of this relationship is the sequence databases, or ITS rDNA is highly con-
farming of ambrosia fungi by ambrosia beetles. This served and uninformative in species identification
obligate mutualism evolved multiple times between (e.g., Fusarium (Hypocreales) (O’Donnell, 2000)).
wood-boring beetles (Curculionidae, Scolytinae and Another obstacle, important for both culturing and
Platypodinae) and several groups of fungi, particu- culture-independent approaches, has been the spa-
larly Ophiostomatales, Microascales, Hypocreales tial distribution of fungal associates in and on the
and Saccharomycotina (Beaver, 1989; Farrell et al., beetle body. The beetles live in a very fungus-rich
2001; Massoumi Alamouti et al., 2009). environment, but the coevolved symbionts occur
The scolytine beetle–fungus symbiosis is receiv- almost exclusively within the beetle mycangia and
ing an increasing amount of research attention inpartsofthegalleriesmadebythebeetlesinwood
because of the massive impact, both ecological and and for only a limited part of the beetle life cycle
economic, that these organisms can cause. For (Kajimura and Hijii, 1992). Thus the traditional
example, the largest outbreak of forest pests in methodofanalyzingfungalcommunitiesfromentire
recorded history is currently occurring in the macerated beetle bodies has introduced a large
northern and western regions of North America, number of non-symbionts into the literature and a
causedbyseveralDendroctonusbarkbeetlesclosely large uncertainty regarding the true associates.
associated with symbiotic fungi. Also, new intro- In this work, we have carried out the first high-
ductions of ambrosia beetles and fungi into non- throughputsurveyofcommunitiesoffungiassociated
native regions are increasingly common; invasive with mycangia of three sympatric ambrosia beetles.
ambrosia insect–fungus couples are eradicating Beetle species were chosen to enable tests of several
naive, native tree species in many places around hypotheses. We included two clades with different
the world (Hulcr and Dunn, 2011). modes of fungus transportation (different mycangia),
Recent research on the ambrosia symbiosis illu- and species that commonly occupy the same dead
minated a relationship that is more species-rich, trees throughout the southeastern United States of
more ecologically dynamic and more evolutionarily America. To allow for strong statistical inference, the
promiscuousthananyotherinsect–fungussymbiosis. sampling included sufficient numbers of individuals
The obligate ambrosia symbiosis is now known and those were analyzed with a high-throughput
to have evolved at least eleven times in insects and sequence-based community survey method.
seven times in fungi (Van der Walt, 1972; Hulcr Specifically, we tested whether ambrosia beetle
etal.,2007;MassoumiAlamoutietal.,2009;Kasson mycangia contain one, or a very few, faithful fungal
et al., 2013) and involved origins of many different symbionts or whether they are more permissive. This
fungus transmission organs (mycangia). Still, one of test also implicitly addressed the question whether a
the essential characteristics of the symbiosis— culture-independent approach would corroborate the
specificity—remains controversial. It has been tra- taxonomic identities of ambrosia fungi previously
ditionally assumed that ambrosia beetles are mostly inferred from culture-based studies.
associatedwithasingledominantfungus(Hubbard, We also examined new questions that have not
1896, 1897; Doane and Gilliland, 1929; Leach et al., been addressed in previous publications. For exam-
1940; Francke-Grosmann, 1956; Batra, 1963), but an ple, the total diversity of the mycangial commu-
increasing number of studies suggests that some nities has not yet been quantified and may have
beetle species may have more than one associate been underestimated by culturing methods. Quanti-
(Batra, 1966; Norris, 1979; Ku¨hnholz, 2004; tative community data also allowed us to test
Gebhardt et al., 2004; Harrington et al., 2010; fundamentalsymbiologyquestions,suchaswhether
Carrillo et al., 2014). Recent experiments using the mycangial communities fit the pattern of high-
cultured ambrosia fungi suggested that sympatric fidelity exclusive symbioses, or the pattern of loose
communitiesofambrosiabeetlesmayactuallyshare assemblages of species with varying degrees of
a large pool of fungi and that horizontal symbiont association. Finally, we were able to test whether
exchange may be common (Carrillo et al., 2014). fungiarespecifictogeographiclocalitiesorwhether
The reasons for the unsettled understanding of they maintain specificity to their beetle species
beetle-fungus specificity have involved a range of across localities.
technological obstacles. Aside from the minute size To answer those questions, we provided the
andcrypticnature ofboththebeetlesandthefungi, first fine-scale and high-throughput data on
the main obstacle for culture-independent fungal communities in ambrosia beetle mycangia.
approaches has been that the most commonly used We comprehensively characterized communities of
marker for fungal community surveys—the internal fungal associates on multiple levels: within beetle
TheISMEJournal
Pyrosequencinganalysisofambrosiacommunities
MKostovciketal
128
specieswithsufficientlocalsampling,withinspecies
acrossalargerregionandalsobetweenthreespecies.
Materials and methods
Beetle species and locality selection
We sampled three ambrosia beetle species com-
monly found across the southeastern United States:
Xylosandrus crassiusculus, Xyleborus affinis, and
Xyleborus ferrugineus. This selection has multiple
advantages for this and future work. For example,
these species represent two types of mycangia
(Figure 1). The mandibular mycangium (X. affinis
and X. ferrugineus) has openings in the oral cavity
and is the ancestral and most widely distributed
mycangium type in the beetle tribe Xyleborini
(Hulcr and Cognato, 2010; Cognato et al., 2011).
The mesonotal mycangium (X. crassiusculus) is
located between the pronotum and the elytral basis;
it is larger and more recently evolved (Hulcr and
Cognato, 2010; Cognato et al., 2011). Another
advantage of this selection of taxa is that all three
speciesarenearlygloballydistributed,spreadpartly
through human trade (Wood and Bright, 1992).
X affinis and X. ferrugineus are native in warm Figure 1 (a) Sampling location in Florida, US. Pie size
and humid regions throughout the Americas, but correspondstotherelativenumberofbeetlespecimenscollected
both have now colonized the tropics and subtropics from each location; color sections indicate ratios of three beetle
species. (b) Sampling locations on the beetles: Digital cross-
of other continents. X. crassiusculus is native to
sectionsofXyleborusferrugineusandXylosandruscrassiusculus
eastern Asia and is aggressively colonizing the rest showing the two different types of mycangia, mandibular and
of the world, frequently becoming a dominant mesonotal,respectively.
ambrosia beetle in new regions. This history of
range expansion makes these species ideal candi-
datesforfutureanalysesofsymbiontfidelityduring
biological invasions. DNA extraction
All three species were sampled from various Beforetheextractionoffungifrommycangia,beetles
habitats across Florida (Figure 1, Supplementary were surface-washed by 15-s vortex in sterile
Table S1). Effort was made to collect all three phosphate buffer saline with 0.1% Tween, followed
species from most of the locations, which enabled by 5s vortex in 40% ethanol. After drying, beetles
us to compare the effect of locality as well as beetle were either stored in 95% ethanol at (cid:2)801C or
species on the fungal community. Altogether, 93 immediately processed for mycangial extraction.
samples were processed, including 35 X. crassius- The contents of the mesonotal mycangia of X.
culusfrom4locations,38X.affinisfrom2locations crassiusculus were extracted from beetles stretched
and 20 X. ferrugineus from 3 locations. To test for with 0.15-mm minuten pins on sterilized paraffin
contamination, PCR biases and sequencing biases, underadissectingmicroscope.Thefungalmasswas
several control samples were included: a negative removed with a pin and immediately placed in
control (a blank sample passed through all steps of extractionbuffer.InX.affinisandX.ferrugineus,the
library preparation from DNA extraction to sequen- content of the mandibular mycangia cannot be
cing) and several reference communities of pure accessed directly; instead the frontal part of the
culture fungal strains previously isolated from the head that contains the paired mycangium was cut
same beetle species from the same locations off, avoiding the esophagus, and crushed. Fungal
(Supplementary Table S1). DNA was extracted with the UltraClean Microbial
Beetles were collected during their dispersal DNA (MoBio, Carlsbad, CA, USA) isolation kit
flight, which is when the mycangium contains the according to manufacturer’s instructions. DNA
largestamountofsymbiontinoculum(Kajimuraand concentrations were assessed using the PicoGreen
Hijii, 1992). Beetles were collected using plastic fluorescent dye in the Quant-iT kit (Life
2-liter bottles with a cut-out window, filled with Technologies, Carlsbad, CA, USA) on a Realplex
approximately 50ml of ethanol. Ethanol served as cycler (Eppendorf, Hauppauge, NY, USA) and
both the bait and the preservation medium. Traps adjusted to the concentration of 10ngml(cid:2)1. The
were checked daily, and all trapped beetles were presence of amplifiable DNA in all the samples was
stored in 95% ethanol and at (cid:2)801C. confirmed by a test amplification.
TheISMEJournal
Pyrosequencinganalysisofambrosiacommunities
MKostovciketal
129
Marker selection amplicon was subjected to a 10-cycle amplification
The marker used for the fungal community survey with fusion primers. The forward fusion primer
wastheITS2ofthenuclearrDNA,onaverage180bp gITS7 consisted of the template primer, the pyrose-
long. The ITS2 marker provides several benefits: quencing adapter A and a 10-bp sample-specific
(1)TheITS1-5.8S-ITS2partofthenuclearribosomal barcode. Barcodes were designed, and fusion
DNA (ITS) has been proposed as the universal primers were tested using EDITTAG (Faircloth and
‘barcode’ for identification of fungi (Schoch et al., Glenn, 2012). The ‘edit distance’ metric was
2012); (2) As the fungal ‘barcode’, ITS as a whole conservatively set to five to minimize sample
andITS2inparticularisthebestrepresentedfungal assignment errors due to sequencing, replication
genomicelementinpublicdatabases(Nilssonetal., and oligonucleotide synthesis errors. An edit
2009);(3)ThelengthofITS2islessvariablethanthe distance of 5 means that at least 5 insertions,
whole ITS region. This is important, as length substitutions or deletions would have to occur in
variation can bias community pyrosequencing one tag to convert it into another tag and therefore
toward shorter amplicons (Ihrmark et al., 2012); confuse demultiplexing. For barcode sequences of
(4) Using just the ITS2 region instead of the entire all samples, see Supplementary Table S1. The
ITS region eliminates the formation of chimeric reverseITS4fusionprimerconsistedofthetemplate
amplicons of heterospecific ITS1 and ITS2 regions, primer and the pyrosequencing adapter B.
induced bytheannealingofthehighlyconservative This nested PCR process reduces amplification
5.8S regions from different species (Quince et al., biascausedbyapplyinglongfusionprimersdirectly
2009; Bellemain et al., 2010). To further increase to the template DNA and results in the recovery of
taxonomic coverage of the primers, a recently higher taxon diversity (Berry et al., 2011). Amplifi-
developed degenerate forward primer gITS7 was cation was conducted on Mastercycler Nexus
used (50-GTGARTCATCGARTCTTTG-30) (Ihrmark (Eppendorf) in 25-ml reactions with proofreading
et al., 2012) in combination with a universal ITS4 Q5High-Fidelitypolymerase(NewEnglandBiolabs)
reverse primer (50-TCCTCCGCTTATTGATATGC-30) at 0.02Uml(cid:2)1 in Q5 reaction buffer, 2mM MgCl2,
(White et al., 1990). 200mM dNTPs, 500nM gITS7 primer, 300nM ITS4
Using ITS as a marker for an ambrosia community primer and 10ng of template. Cycling conditions
survey alsohasits problems. It is known that several were as follows: initial denaturation at 951C for
traditionallyusedprimersfortheITSregioninfungal 4min followed by 25 cycles of (50s at 941C, 50s at
community surveys (that is, ITS1f and ITS4; 561C, 45s at 721C) and a final elongation step for
Jumpponen and Jones, 2009; Amend et al., 2010; 10min at 721C.
Arfi et al., 2012) amplify poorly or not atall in many The second-step PCR products were individually
Ophiostomatales, including several clades of ambro- cleaned up (UltraClean PCR clean-up kit, MoBio),
sia fungi (Fraedrich et al., 2008). The amplification their concentrations measured using the PicoGreen-
failure is likely due to Ophiostomatales-specific based Quant-iT kit (Life Technologies), and pooled
formation of secondary structures in the GC-rich in equimolar proportions to create an amplicon
region at the conservative end of the 18S nrDNA library at a final concentration of 10ngml(cid:2)1. The
whereITS1fandother50 primersanneal(Whiteetal., pooledampliconlibrarywascleanedupagainusing
1990; Gardes and Bruns, 1993). Fortunately, during UltraClean PCR clean-up kit (MoBio). The library
our trials with a high-performance polymerase (Q5 wassequencedontwo1/8thregionsofasequencing
High-FidelityDNApolymerase,NewEnglandBiolabs, plate on GS-FLX Titanium (454 Life Sciences,
Ipswich, MA, USA), the amplification of ITS2 was Branford, CT, USA) at the Interdisciplinary Center
reliable (see below). With the two above-mentioned for Biotechnology Research at the University of
primers, we have successfully amplified ITS2 from Florida. We expected to obtain an average sequen-
DNA extracts from pure cultures that included repre- cing depth of 2000 reads per sample provided that
sentatives of three clades of common ambrosia fungi, standard yield per region is 100000 reads. Raw
such as several Raffaelea spp. (Ophiostomatales). demultiplexed sequencing data with sample anno-
tations are available at the Short Read Archive
(http://www.ncbi.nlm.nih.gov/Traces/sra/) under the
PCR amplification and library preparation study accession number SRX348622.
Themarkerwasamplifiedinallsamplesfollowinga
nested-PCR approach. First, the total mycangial
DNA was subjected to amplification with the Quality control and reference communities
template-specific primers gITS7 and ITS4 for 25 To confirm that our PCR-based results do not suffer
cycles. The number of cycles was lower than usual from major biases and that they reflect the fungal
to prevent depletion of some of the forward primers communities within the beetles, we included three
in the degenerate mixture. All samples were ampli- quality-control measures.
fied in triplicate to decrease stochastic variability First,weincorporatedtworeferencecommunities
between reactions (Schmidt et al., 2013). Pooled of known composition into the sequencing library.
PCR reactions were cleaned by UltraClean PCR One was composed of several representatives of
clean-up kit (MoBio). Second, a 1-ml aliquot of this genus Raffaelea and the second one was composed
TheISMEJournal
Pyrosequencinganalysisofambrosiacommunities
MKostovciketal
130
of representatives of clades previously reported 2010), with 97% similarity threshold to reference
as ambrosia fungi, including Ophiostomatales, sequences. As the reference database, we used the
Microascales and a basidiomycetous yeast. UNITEþINSDC data sets, accessed 26 August 2013
Second,toassurethatmajorculturablesymbionts (Abarenkov et al., 2010). The reference sequences
withknownPCRbiasesarenotmissed(forexample, representedspecieshypotheses(Koljalgetal.,2013)
Raffaelea, Fraedrich et al., 2008), the sequencing defined by dynamic clustering with a floating level
output was compared with a reference culture data of similarity recommended by specialists on parti-
set. Relative abundances of dominant fungal taxa cular lineages of fungi. Reads that failed to hit an
were estimated by dilution-to-extinction plating existing reference sequence were clustered de novo.
of mycangia on Potato Dextrose Agar media Singletons (clusters containing only one read) were
from mycangia of the same beetle species discarded as they are indistinguishable from
(Supplementary Table S6). sequencing errors (Kunin et al., 2010). For each
Third, we exhaustively reviewed primary litera- OTU,themostabundantreadwasdesignatedasthe
ture and extracted records of fungal associates representative sequence. OTUs were identified
isolated from X. affinis, X. ferrugineus and taxonomically using RDP classifier (minimum
X. crassiusculus (Table 1). Due to taxonomic confidence set to 0.8) (Wang et al., 2007) against
uncertainties in older literature, we only surveyed the UNITEþINSDC data set.
reports published after the first reclassification of
ambrosia fungi by Batra (1967).
Statistical analysis
Statistical analysis was conducted in QIIME 1.7.0
Data processing (Caporaso et al., 2010) unless noted otherwise. For
Sequencing data were processed using QIIME 1.7.0 most analyses, samples were rarefied to the number
(Caporaso et al., 2010). Reads o150bp, 4600bp, of reads in the least numerous sample, which had
with 46 ambiguous bases and with homopolymers 670 reads. The completeness of fungal taxon
410bp, were discarded. The minimal quality score sampling was estimated by rarefaction curves.
was set to 25, and no mismatches in the primer Alpha diversity of fungal communities in each of
sequencewereallowed.Dataweredenoisedaccord- the three beetle species was estimated by species
ing to Reeder and Knight (2010). Chimeric accumulation curves, the Chao1 estimator of total
sequences were removed using a combination of diversity,theDominanceindex(1(cid:2)Simpsonindex)
reference-basedanddenovochimeradetectionwith and the Shannon index. To quantify differences in
USEARCH 6.1 (Edgar et al., 2011); only reads fungal communities between samples (beta diver-
detected as chimeric with both approaches were sity), a matrix of pairwise Bray–Curtis dissimila-
discarded. For the subsequent analyses, the ITS2 rities was computed. The non-phylogenetic Bray–
region was extracted from each sequence using Curtis metric was used, as the ITS2 region is
Fungal ITS extractor (Nilsson et al., 2010), as the unalignable between different fungal genera. Sam-
conserved flanking regions are known to distort ple similarities were visualized using Principal
similarity searches, taxonomic assignments and Coordinate Analysis. The effect of beetle species,
clustering results (Bruns and Shefferson, 2004). location and their interaction on the fungal compo-
The resulting ITS2 reads were clustered to opera- sition of samples was tested using two comple-
tional taxonomic units (OTUs) by UCLUST (Edgar, mentary nonparametric approaches: analysis of
Table1 FungireportedfromX.affinis,X.ferrugineusandX.crassiusculusintheliterature
Beetlespecies Isolatedfungi Source
Allthree Raffaelealauricola Ploetzetal.,2011
Ceratocystisfimbriata HerreraandGrillo(1989)
X.affinis Cephalosporiumpallidum Verrall,1943
Ambrosiellataxonomicspp.1and2(P) Batra,1967
Cephalosporiumpallidum(adaptedfromVerrall,1943) Biedermannetal.,2009
Fusariumsp.,Raffaeleasp.,unspecifiedyeast Roeper,1996
X.crassiusculus Raffaelealauricola Harringtonetal.,2010
Ambrosiellasp. HarringtonandFraedrich,2010
Ambrosiellasp.(closetoA.xylebori) Roeper,1996
X.ferrugineus Fusariumsolani,Cephalosporiumsp.,Graphiumsp. BakerandNorris,1968
Fusariumsolani,Cephalosporiumsp. Norris,1979
Ceratocystisfimbriata SaundersandKnoke,1967b
Fusariumsp. SaundersandKnoke,1967a
Fusariumsolani WeberandMcPherson(1984)
Fusariumsolani ZimmermannandButin(1973)
CephalosporiumiscurrentlyclassifiedinthegenusAcremonium.
TheISMEJournal
Pyrosequencinganalysisofambrosiacommunities
MKostovciketal
131
similarities (ANOSIM; Clarke, 1993) and permuta- We used both incidence (presence/absence) and
tional multivariate analysis of variance (PERMA- abundance of reads. PCR-based community surveys
NOVA;Anderson,2001).Statisticalsignificancewas aresometimessummarizedusingonlyincidencedata
determined with 999 permutations. The effect of to circumvent biases in DNA extraction and PCR
locality and the interaction between locality and (Bellemain et al., 2010). On the other hand, inci-
beetle species was tested using the two largest dence-only data may overemphasize environmental
samples (Colt Creek and Lake Wales) owing to contaminants that are frequent but have minor
unbalanced species representation in other abundance and have no role in the symbiosis. Also,
locations. we used amplification of short fragments and
To test if any particular OTU was significantly degenerate primers with large taxonomic coverage,
associated with any particular beetle species, bothofwhichwereshowntoimprovereconstruction
ANOVA and the G-test of independence were used. of true community structure (Ihrmark et al., 2012).
The G-test is used for presence/absence data,
whereas ANOVA is used for relative abundance
Results and discussion
data.Bothtestsemployed100denovorarefiedOTU
tables so the resulting P-values are the averages of
Data
P-valuesfrom singletests.Thesignificancelevelfor
In total, 197672 reads were generated. After quality
ANOVAwas corrected for multiple comparisons by
filtering, denoising and chimera removal, 148923
the false discovery rate correction (Benjamini and
reads were retained for further analysis. After
Hochberg, 1995).
demultiplexing (assignment of reads to beetle
samples), the average read count per sample was
1551. The sequencing depth was relatively even
Identification of core associates
across samples: the poorest sample contained 670
The fungal community of each beetle species was
reads and the richest sample contained 4000 reads.
partitioned into core OTUs (persistent, abundant,
The 97% sequence similarity cutoff produced 402
non-randomly associated and likely biologically
clusters representing OTUs with at least two reads.
relevant taxa) and occasional or rare OTUs (oppor-
This sequencing depth was sufficient for recovering
tunist, contaminants or other non-symbionts). Com-
most of the diversity present in each beetle species
pletely sampled biological communities typically
given the similarity threshold, as indicated by
follow log-normal species abundance distribution
the flattening of species accumulation curves
(SAD) (Magurran and Henderson, 2003). Culture-
(Supplementary Figure S1).
independent high-throughput data sets, however,
are frequently negatively skewed, which is thought
to be caused by low-frequency extraneous species Symbiont community diversity
(Unterseher et al., 2011). We identified core fungal Our results show an order of magnitude greater
taxa within each beetle species based on disconti- diversity of fungal species occurring in beetle
nuity in a log-normal SAD using persistence- mycangia than previously reported from culturing
abundance plots. First, OTUs were plotted on a studies (Table 1). The estimated total numbers of
persistence-abundance plot (Magurran and associates within each beetle species were similar:
Henderson, 2003) so that the number of samples in the Chao1 estimate of total diversity ranged from
which each taxon was observed was plotted against 14.71to20.06(Table2).Theseestimateswerelikely
the maximum abundance of each taxon in any one approaching the true diversity, as our sampling was
sample (Unterseher et al., 2011). The plots provide near exhaustive (Supplementary Figure S1). The
means for visual detection of a discontinuity; OTUs estimates may have been slightly underestimated as
areiterativelyseparatedintwoandseparatelytested singleton reads were removed during quality con-
for fit with the log-normal SAD (core species, right) trol, but singletons are probably not relevant in the
and log-series SAD (rare or transient species, left). symbiosis.
Table2 Diversityindicesformycangialfungalcommunitiesofthethreeambrosiabeetlespecies
Species Observed Observed Chao1 Chao1,s.e. Shannon Shannon Dominance Dominance
species species,s.e. estimate index index,s.e. Index index,s.e.
(1(cid:2)Simpson)
X.affinis 16.59 4.33 17.62 5.35 2.77 0.52 0.23 0.09
X.ferrugineus 14.18 3.46 14.71 3.79 2.88 0.39 0.19 0.08
X.crassiusculus 17.62 6.40 20.06 8.22 1.83 0.54 0.49 0.14
Allthreespecieshaveacomparabletotaldiversityofassociates.ThecommunityinXylosandruscontainsthehighestdiversity(highChao1)but
withhighlyskewedabundances(lowShannonindex),sotheapparentdiversityislikelydrivenbytransientspecies.ThetwoXyleborusspp.
displaytheoppositepattern:lowertotalspeciesrichnessbutmoreevenlyrepresented.Allsampleswererarefiedto670sequences.
Topbarsdisplayaveragecommunitiesacrossspecies,bottombarsareconcatenatedcommunitiesfromindividualbeetles.
TheISMEJournal
Pyrosequencinganalysisofambrosiacommunities
MKostovciketal
132
Even though ambrosia beetle genera have similar S2, Supplementary Figure S1). The two representa-
overall diversity of associated fungi, it appears tives of the beetle genus Xyleborus with mandibular
that each genus, characterized by a unique type mycangia (X. affinis and X. ferrugineus) had fungal
of mycangium, hosts communities of different communities similar in their diversity and evenness
taxonomic composition and different diversity (communitydominanceindex¼0.19and0.23,respec-
structure (Figure 2, Table 2, Supplementary Table tively; Figure 2, Table 2). On the other hand, fungal
communities of Xylosandrus crassiusculus were
heavilydominated(Figure2,dominanceindex¼0.49)
by a single species, Ambrosiella sp., and also
contained more incidental transient species.
This study analyzed only a single representative
of the mesonotal mycangium; however, the differ-
ences were significantand also reflect our reference
culturesanddataintheliterature(Table1).Further,
the community of Xylosandrus crassiusculus may
be unique partly, because the species was intro-
duced into the United States only several decades
ago (Anderson, 1974). Non-native populations that
originate from small initial populations may lose
heritable associates during theestablishmentbottle-
neck(Reyetal.,2013).ThefactthatX.crassiusculus
is dominated by Ambrosiella sp. is likely a true
pattern, as Ambrosiella sp. has been reported as a
dominant associate of the beetle in its native East
Asia (Kinuura, 1995) and of other species from the
same phylogeneticclade(that is,Cnestus;Six etal.,
2009). Studies of additional beetle species are
needed to test the pattern of correlation between
fungal community structure and mycangium type.
The core associates
Fungal OTUs were divided into core species and
transient species based on the position of each OTU
within the SAD (Figures 3 and 4, Supplementary
Figure 2 Stack plot showing species richness and evenness of TablesS5andS8).OTUsetsthatfollowedlog-normal
mycangialfungalcommunitiesofthethreebeetlespecies.Upper distribution were considered core associates. Incon-
stacks report averages within each species, lower stacks report sistent OTUs that fit a log-series distribution of
communities in each individual. Each color band represents
abundances were considered transient. The core
differentfungalOTU;thebandwidthcorrespondstotherelative
communityconsistedof8fungalOTUs(48.2%OTUs,
abundanceofreadsoftheOTUinthedataset.Arrowspointout
thetwodifferenttypesofmycangia. 78% of sequences) for X. affinis and 6 OTUs (42.3%
Figure3 Separatingcoreandtransientcommunitymembersbasedondiscontinuityinpersistence/abundancedistributionoffungal
OTUs. Dashed lines are segregating OTU sets that fit the log-normal abundance distribution (core symbionts, right) and log-series
distribution(transientassociates,left).
TheISMEJournal
Pyrosequencinganalysisofambrosiacommunities
MKostovciketal
133
Figure4 Distributionofabundanceclassesofcorespecies(transient,incidentalspeciesexcluded,seeMaterialsandmethodssection
for details). Red curve shows expected log-normal distribution of abundances. From left to right: X. affinis, X. ferrugineus,
X.crassiusculus.CommunitiesinbothXyleborusaremoreevenlyabundantthanthatinXylosandruscrassiusculus.
Table3 CoreOTUsofthethreebeetlespeciesandtheiridentityasaconsensusofsearchesintheUNITEþINSDC,GenBankBLASTn
andamatchwithourreferencesequencedcommunitiesandreferencecultures
Species1 OTUcode Consensusidentification
X.crassiusculus,X.ferrugineus,X.affinis New.ReferenceOTU2 Microascales(Ambrosiella)
JN882303 Cryptococcussp./Bandoniozymaglucofermentans
GU117066 Malasseziarestricta
GU721625 Cladosporiumsp.
HM211210 Fusariumsp.
New.ReferenceOTU4 Xylarialessp.
X.crassiusculus,X.affinis GU721291 Epicoccumnigrum
JN693502 Diplodiacorticola
New.ReferenceOTU1 Candidasp.
X.crassiusculus,X.ferrugineus NocoreOTUshared NA
X.affinis,X.ferrugineus NocoreOTUshared NA
X.affinis New.ReferenceOTU0 Candidalaemsonensis
EU568927 Pichiajadinii
FJ381698 Candidamycetangii
X.ferrugineus GQ850385 Strelitzianaafricana
X.crassiusculus New.ReferenceOTU146 Ceratocystissp.
New.ReferenceOTU206 Ceratocystissp.
New.ReferenceOTU49 Sugiyamaellasmithiae
Abbreviation:OTU,operationaltaxonomicunit.
of OTUs, 64% of reads) for X. ferrugineus. the genus Ambrosiella (Microascales; Table 3,
In X. crassiusculus, the core consisted of 11 OTUs Supplementary Table S8). Identification of this
(62.4%OTUs,86%ofreads),butmostofthemwereof genuswaspossible onlythrough acomparison with
low abundance, as 65% of all reads were represented the reference communities included in the sequen-
by one dominant symbiont (Figures 2 and 4). cing run. Identification directly by BLAST or RDP
In summary, each beetle species appears to have was impossible due to the weak representation of
multiple core associates. However, communities in Ambrosiella spp. in public databases, unsettled
both Xyleborus spp. have much more even abun- nomenclature and the conserved sequence of the
dances than those in Xylosandrus, where the ITS2 marker, as discussed above.
dominant species (Ambrosiella sp.) is 410(cid:3) more For example, the OTU HQ538467, which was
abundant than most core species. (Supplementary abundant and frequent in both Xyleborus sp., did
Table S4). This is consistent with our semi-quanti- not have a match in taxonomic databases. It is most
tative reference cultures (Supplementary Tables S6 likelyRaffaeleasp.asitmatchedourRaffaelea-only
and S7). reference community. The Raffaelea-like OTU was
rare in X. crassiusculus, but an order of magnitude
morecommoninbothXyleborusspecies.Thegenus
Taxonomic identity of the fungal community Raffaelea appears to be the primary symbiont of
Wehaverecoveredgenerathathavebeenreportedas Xyleborusspp.basedonliterature(Harringtonetal.,
the dominant mutualistic symbionts, primarily 2010)andonourreferencecultures(Supplementary
TheISMEJournal
Pyrosequencinganalysisofambrosiacommunities
MKostovciketal
134
Tables S6 and S7). Its closest match in taxonomic Many other fungi were present whose genera,
databases was an undetermined Ceratocystis sp. identified by BLAST, suggest diverse life strategies
UndersamplingofRaffaeleasspp.andotherophios- not related to beetles. These include plant patho-
tomatoidassociatesinourdatasetwaslikelydueto gens, epiphytes, endophytes, soil saprophytes and
amplificationinefficiencyandtemplatecompetition common molds, such as Cladosporium sp.
in complex fungal communities. ITS sequences of Culture-independent assessments of community
these taxa are also rare in public databases, further structure are becoming essential for describing
illustrating the problematic amplification in such symbioses, be it the ambrosia beetle–fungus system
taxa (for example, Fraedrich et al., 2008). or other fungus-farming systems or any other
In addition to the known mutualists, we also
detected many other fungi that are sufficiently
frequenttoqualifyforthestatisticalcorecommunity
butthathaveneverbeenreportedfromtheambrosia
symbiosis (Table 1, Supplementary Table S4).
The ecological function of these fungi and their
role in the ambrosia consortium cannot be inferred
from sequence data. However, the genera to which
these fungi belong do have well-characterized
ecologies, which allows us to predict potential
function.
(cid:4) One of the most frequently detected and well-
identified species was Diplodia corticola—a tree
pathogen that has recently invaded the United
States and is increasingly damaging to economic-
ally important trees (U´rbez-Torres et al., 2010;
Dreaden et al., 2011; Lynch et al., 2013). This
indicates that not all fungi transported by ambro-
sia beetles are involved in the symbiosis and that
ambrosia beetles may frequently transport tree
pathogens.
(cid:4) Another example of a high-frequency associate
with unknown functions was Fusarium. The ITS
marker is not suitable for identification of fusaria
to the species level (O’Donnell, 2000; Balajee
etal.,2009).Fusariaarefrequentlyrecoveredfrom
wood-boring beetles as common phoretic con-
taminants (Batra, 1963, 1967). However, several
clades in the Fusarium solani complex are also
intimately associated with at least three clades of
ambrosia beetles (Norris, 1979; Kasson et al.,
2013). We do not know whether the common
Fusarium OTU detected within the mycangia of
beetles investigated here was symbiotic or phore-
tic and whether it is one species or several.
(cid:4) Numerous OTUs represented different lineages of
yeasts or yeast-like fungi, both ascomycetous
(Saccharomycotina) and basidiomycetous. An
example is an OTU identified as either Cryptococ-
cus sp. or Bandoniozyma glucofermentas, which
was frequent in all three beetle species, and most
abundant in the more permissive mandibular
mycangiaofXyleborus.Yeastsarecommonassoci-
ates of many wood-boring beetles (Ganter, 2006;
Rivera etal.,2009),not only ambrosia beetles,and
appear to be specialized to the habitat rather than
to the vector (Calderon and Berkov, 2012). Yeast Figure5 PrincipalCoordinateAnalysisplotsoffungalcommu-
may compete with beetles for some of the same nities in mycangia of Xyleborus affinis, X. ferrugineus and
Xylosandrus crassiusculus based on Bray–Curtis dissimilarities.
nutrients (that is, glucose in B. glucofermentas
(a) Samples labeled according to the ambrosia beetle species of
(Valenteetal.,2012)),buttheymayalsobeasource
origin.(b)Sampleslabeledaccordingtothegeographiclocationof
of vitamins and other essential metabolites. origin.
TheISMEJournal
Pyrosequencinganalysisofambrosiacommunities
MKostovciketal
135
Table4 Theeffectofbeetleidentityandcollectionlocalityonfungalcommunitycompositionevaluatedbytwocomplementary
nonparametrictests
Category PERMANOVA(Pseudo-Fstatistic) PERMANOVA(P-value) ANOSIM(Rstatistic) ANOSIM(P-value)
Species 12.4569 0.001* 0.6923 0.001*
Locality 4.0961 0.006* 0.0588 0.091
Speciesþlocality 9.9311 0.001* 0.5051 0.001*
Abbreviations:ANOSIM,analysisofsimilarities;PERMANOVA,permutationalmultivariateanalysisofvariance.Beetlespeciesisthestrongest
predictoroffungalcommunitycomposition;localityandtheinteractionbetweenspeciesandlocalityhavelesssignificanteffects.Significant
P-valuesaremarkedwithanasterisk.
multi-member symbiotic complex. However, quan- possibly as spores, but appears to better restrict
tities of microbial members are only an approxima- their proliferation.
tion of their importance for the macrosymbiont. Statistical differences between fungal commu-
In our case, many of the taxa that dominated the nities were driven by many fungal species, more
sequencing output were indeed the coevolved than just by the known primary symbionts
nutritional symbionts, but we know that only (Supplementary Table S8). As many as 10 different
because of external experimental evidence. Other OTUs were significantly associated with only 1or 2
taxa in our output were equally abundant, yet we beetle species (ANOVA and G-test; Supplementary
cannot conclude whether that is due to their TablesS4andS8).Thesearestatisticalassociations,
ecological significance or due to incidental causes, andtheirecologicalrelevanceremainstobestudied.
such asprolific spore-formingor amplification bias.
We caution against overinterpreting next-generation
Conclusions
sequencing data as a reflection of ecological
functions in symbiology.
Ambrosia beetle mycangia contain one or, in
some beetle species, a few of the most coevolved
nutritional symbionts, but the fungal community
as a whole is more diverse and more promiscuous
Factors that shape the fungal community composition
than previously documented. The promiscuity
Beetlespecieswasamuchstrongerpredictoroffungus
appears to differ significantly between beetle
community composition than locality (Figure 5). This
lineages. Each lineage, equipped with a distinct
is suggested by both nonparametric tests (ANOSIM
organ for fungus transmission, carries an over-
and PERMANOVA, Table 4), as well as by the
lapping but different set of associates and contami-
Principal Coordinate Analysis, in which samples
nants, each with their own degree of fidelity and
labeled by beetle species show distinct clustering,
community structure. Horizontal exchange of fungi
whereassampleslabeledbylocationdonot(Figure5).
may be common (Carillo et al., 2013), but each
However,theeffectoflocalitymayhavebeenreduced
mycangiumtypesupportstheassemblyofadistinct
bythelimitedgeographicrangeofoursamples,which
community.
is in contrast with the nearly global distribution of all
Our results also corroborate some of the beetle–
three beetle species. There was also a significant
fungus associations reported in the literature (the
influence of the interaction between beetle species
second hypothesis, see Table 1). However, publica-
andlocality;however,itsimplicationsareunclear,and
tionsfromthepremoleculareravarysignificantlyin
it may be a result of uneven sampling.
identification accuracy and report only the most
The differences between fungal communities in
common and easily culturable fungi, which repre-
the two beetle genera may be mediated by the
sent only a fraction of the actual diversity discov-
differenttypesofmycangia.Severalgroupsoffungi,
ered here.
whose function in the beetle is unknown, were
To avoid further spread of the uncertainty about
significantly more abundant in the mandibular
beetle–fungus associations that has plagued the
mycangia of the two Xyleborus species than in the
ambrosia symbiosis literature, we recommend that
mesonotalmycangiumofXylosandrus.Xylosandrus
future surveys of this unusually diverse system
contained a greater number of species in total,
include careful extraction techniques, sufficient
including numerous transient fungi, but the major-
sample sizes, phylogeny-informed species selection
ityofthemrepresentedonlyfractionsofabundance.
and a combination of culture-dependent and
This pattern supports the hypothesis that the
independent methods.
mandibular mycangium, although much smaller
andconcealed,providesamorepermissiveenviron-
ment for a broader array of symbionts, as well as,
Conflict of Interest
nonspecific fungi. The thoracic mycangium of
Xylosandrus also collects nonspecific fungi, The authors declare no conflict of interest.
TheISMEJournal