Fungi are important actors in ecological processes and trophic webs in mangroves. Although saprophytic fungi occurring in the intertidal part of mangrove have been well studied, little is known about the diversity and structure of the fungal communities in this ecosystem or about the importance of functional groups like pathogens and mutualists. Using tag-encoded 454 pyrosequencing of the ITS1, ITS2, nu-ssu-V5 and nu-ssu-V7 regions, we studied and compared the fungal communities found on the marine and aerial parts of Avicennia marina and Rhizophora stylosa trees in a mangrove in New Caledonia. A total of 209 544 reads were analysed, corresponding to several thousand molecular operational taxonomic units (OTU). There is a marked zonation in the species distribution, with most of the OTU being found specifically in one of the microhabitat studied. Ascomycetes are the dominant phylum (82%), Basidiomycetes are very rare (3%), and 15% of the sequences correspond to unknown taxa. Our results indicate that host specificity is a key factor in the distribution of the highly diverse fungal communities, in both the aerial and intertidal parts of the trees. This study also validates the usefulness of multiple markers in tag-encoded pyrosequencing to consolidate and refine the assessment of the taxonomic diversity.
Mangroves are coastal biotopes developing within the intertidal zones of tropical and subtropical regions (Spalding et al., 1997). These forests feature unique plant species forming an interface between terrestrial, estuarine and near-shore marine ecosystems. Geochemical characteristics like salinity, soil humidity or nutrients concentrations are modified cyclically, on one side throughout the day with tides and on the other side with seasons, making the mangrove a very dynamic environment (Marchand et al., 2004). They cover approximately one-fourth of the world tropical coastline and are both ecologically and economically important as they protect coastal areas form erosion and storms, serve as a nursery for numerous exploited animal species (Nagelkerken et al., 2008) and are one of the major sources of terrestrial organic matter to the ocean (Kristensen et al., 2008).
Fungi are regarded as essential actors of the mangrove detritic food webs as they contribute to the degradation of the particular organic matter into dissolved organic matter (Hyde & Lee, 1995). Most investigators focused their attention on the taxonomic diversity of saprophytic fungi collected on the intertidal parts of mangrove trees, on floating or immersed wood debris (Newell et al., 1987; Sarma & Vittal, 2000; Sarma & Hyde, 2001; Fryar et al., 2004; Sridhar, 2004). Although a few articles provide data on arbuscular mycorrhizal fungi (Wang et al., 2010) and endophytes from mangrove trees (Kumaresan and Suryanarayanan, 2007; Gupta et al., 2009), the diversity of fungi occurring on living mangrove trees is poorly known. Yet, it has been shown that tree-associated fungi play a key role in forest ecosystems (Rodriguez et al., 2009) as they have profound impacts on plants, by increasing (symbionts) or lowering (parasites and pathogens) their fitness. It has also been shown that phyllosphere fungi are functionally relevant to the process occurring in the litter, as they play an important role in the early stages of the decomposition of organic matter (Osono, 2006). The mangrove trees are the main source of organic matter to the water and sediment phases through fallen leaves and branches (Marchand et al., 2004; Kristensen et al., 2008). Therefore, the fungal communities occurring on the living trees could play an important role in the mangrove ecosystem.
The mangrove-colonizing fungi identified to this day form a large ecological group with most of its members belonging to the Dikarya subkingdom (Sarma & Hyde, 2001). The studies conducted so far seem to indicate that the geographical situation of a mangrove (its latitude and oceanic basin) is one of the main factors determining the distribution of these fungi (Schmit & Shearer, 2004). The plant host specificity is also thought to be an important factor of the structure of fungal communities (Zhou & Hyde, 2001), although no specific study have been undertaken to evaluate this assumption. It is also noticeable that most of the previous work on mangrove fungi relies on traditional isolation- and cultivation-based methods, followed by taxonomic assignment using morphological characters and more recently molecular and phylogenetic analysis (Alias & Jones, 2000; Prasannarai & Sridhar, 2001). The main limits of this approach are the underestimation of the real fungal diversity (because only a small fraction of the microorganisms are readily cultivable) and the labour requirements (Arnold, 2007). To our knowledge, no metagenomic method has ever been used to assess the fungal diversity in mangrove forests. In the recent years, developments in sequencing technology allowed for the emergence of new cultivation-independent methods that provide novel insights into some of the ecological processes that affect the structure and diversity of fungal communities (Nilsson et al., 2009). For instance, tag-encoded 454 pyrosequencing of the nuclear ribosomal internal transcribed spacer 1 (ITS-1) has revealed unexpectedly high fungal diversities in forest soil or phyllosphere (Buee et al., 2009; Jumpponen & Jones, 2009).
To better understand the diversity of fungi encountered on mangrove trees and the factors that influence the structure of this colonization, we have undertaken a comparative analysis of the fungal communities on the intertidal and aerial parts of two mangrove trees species, Rhizophora stylosa and Avicennia marina, commonly found in New Caledonian mangroves. In this article, we present a study of the fungal diversity using tag-encoded pyrosequencing of four molecular markers (namely ITS1, ITS2, SSU rRNA V5 and V7). Our main goals are to evaluate (i) the diversity of the tree-colonizing fungi encountered in this mangrove, (ii) the influence of plant host specificity on the fungal species distribution in the aerial and intertidal parts of the trees and (iii) the effects of the four molecular regions studied on the fungal diversity estimation.
Materials and methods
Study site and sampling strategy
The studied mangrove is located in the Saint Vincent Bay, Southern Province, New Caledonia (coordinates: −21.934603, 166.078702). It spans 700 m of coastline and covers around 0.24 km2. The climate on site is subtropical with an average temperature of 22 °C and average monthly precipitation of 80 mm (annual maximum occurs in January with 26 °C and 130 mm). This site is representative of the mangroves of New Caledonia in terms of tree species zonation. A ribbon of A. marina colonizes the innermost part of the shore and surrounds a core of R. stylosa on the outer part and the waterfront (Marchand et al., 2011) (Supporting Information, Fig. S1). The samples used for this study were collected on 21 January 2010. Samples were collected from three sites located within 150 m from each other: α, coordinates: −21.936435,166.080052; β, coordinates: −21.93528,166.076855; θ, coordinates: −21.932722,166.075535. The three sites are exposed to similar environmental conditions in terms of water salinity (35 g L−1) and pH (8.0) and are considered as replicate. Each sampling site is located at the border between the A. marina and the R. stylosa stands, thus containing both tree species. For each site, samples were collected at low tide from a pool of five randomly picked full-grown R. stylosa and five full-grown A. marina. Each tree was divided in two parts: above maximum sea water level (aerial) and below maximum sea level (intertidal), thus dividing four microhabitat named AB, AH, RB and RH; where A = A. marina, R = R. stylosa, B = bottom level (immersed) and H = high level (emerged). Therefore, a total of 12 samples were collected. For each individual tree, samples from the aerial level consist of five leaves and five pieces of trunk and branch bark (5 × 1 cm on a 2 mm thickness), while samples from the intertidal level consist of five pieces of bark (10 × 1 cm on a 2 mm thickness). Samples were stored in liquid nitrogen for subsequent DNA extraction.
DNA extraction, PCR and pyrosequencing
Sample material was homogenized in liquid nitrogen using a mortar and pestle, and 500 mg of the resulting homogenate was used for genomic DNA extraction using the ‘NucleoSpin Plant II Kit’ (Macherey Nagel) according to the manufacturer's instructions. Two different amplicon libraries were constructed for every sample, the first one using the fungal primers ITS1F (5′-xxxxxCTTGGTCATTTAGAGGAAGTAA-3′) and ITS4 (5′-xxxxxTCCTCCGCTTATTGATATGC-3′) targeting the ITS region (White et al., 1990; Gardes & Bruns, 1993); the second one using the primer pair nu-SSU-0817-5′ (5′-xxxxxTTAGCATGGAATAATRRAATAGGA-3′) and nu-ssu-1536-3′ (5′-xxxxxATTGCAATGCYCTATCCCCA-3′) targeting the V5–V8 region of the SSU rRNA gene (Borneman & Hartin, 2000). Here, xxxxx represent the five nucleotides barcode used to identify each sample. Three independent PCRs were performed for each library under the following conditions: 5 μL of template DNA (diluted at 1 : 5, 1 : 25 and 1 : 100), 1 μM of each primer, 5% DMSO, 200 μM of each dNTP, 10 μL of Expand High Fidelity buffer and 2.5 unit of Expand High Fidelity enzyme mix (Roche Applied Science) in a total volume of 50 μL. Amplification was run on a Veriti Thermal Cycler (Applied Biosystems) following this programme: 95 °C for 10 min, 35 cycles of 30 s at 95 °C (denaturation), 52 °C for 1 min (annealing) and 72 °C for 40 s (extension), followed by 10 min at 72 °C. The triplicate PCR products were pooled and purified using the ‘Qiaquick PCR Purification kit’ (Qiagen). The 24 amplicon libraries were checked by agarose gel electrophoresis, and their concentration was estimated using a NanoDrop 1000 spectrophotometer (Thermo Scientific). An equimolar mix of amplicons was prepared for sequencing. The two 454 pyrosequencing primers A and B were added to the amplicons by random ligation, so that each amplicons library will generate reads from both the 5′ or 3′ ends (Fig. S2). The sequencing reaction on a Genome Sequencer FLX Titanium 454 System (454 Life Sciences/Roche Applied Biosystems) resulted in 294 102 reads.
Bioinformatics, OTU designation and statistical analysis
First, the raw fasta file were parsed using the trim2.pl and barcode.pl scripts from the pangea pipeline (Giongo et al., 2010) to only keep the reads that met all the following criteria: (i) minimum sequence length of 200 bp; (ii) minimum Phred quality score of 20; (iii) presence of a correct barcode at the beginning of the sequence. The reads that passed this quality check had their primers and barcode sequences trimmed. The header of each read was also modified to only contain the barcode and sequencing primer information and a unique identification number. Then, the reads were sorted by their sequencing primers generating four files (ITS1F, ITS4, nu-ssu-0817 and nu-ssu-1536), with each file containing reads from the 12 samples. To eliminate the nonfungal-related reads, the data were analysed using NCBI blastn against the nonredundant GenBank database. Overall, only 751 reads were removed, 64% of them being identified as R. stylosa and 21% as A. marina, indicating the high specificity of the PCR primers used. Potential chimeric sequences were detected and removed by the Chimera-test tool provided in the Fungal Metagenomics Project pipeline (https://biotech.inbre.alaska.edu/fungal_portal/). To cluster the reads in operational taxonomic unit (OTU), each file was run through CD-HIT-EST (Li & Godzik, 2006). The clusters were generated at several identity thresholds: 90–96% in 2 % increment and 97–100% in 1% increment, with the ‘-g 1’ option that allow to cluster a new read with the best possible pre-existing cluster (instead of the first cluster in chronological order). As the reads analysed in this study are composed of both variable and conserved regions (Fig. S2), applying an arbitrarily chosen threshold based on known values of sequence variability seems inadequate. A more empirical methodology was used, similar to the one devised by Jumpponen & Jones (2009). Briefly, clustering is done at increasingly stringent sequence identity, until a certain sequence identity value is reached for which the number of OTU suddenly increases. Based on this method, a sequence identity of 98% is chosen for each data set (Fig. S3). CD-HIT-EST output files are treated using custom-made Perl and Python scripts to generate OTU-distribution matrix. These matrices indicate for each OTU the number of reads found for the 12 site-tree-level combinations. The singletons were then removed from the data sets. Briefly, an OTU was considered a singleton if it appeared only once in the CD-HIT-EST output. Therefore, an OTU can be found only once in one of the 12 samples, but it is also found in at least one other sample. Thus, the diversity and richness indicators can still be computed for each sample.
The OTU-distribution matrixes were used to perform rarefaction analysis and calculate diversity (Shannon) and richness (ACE and Chao1) indices with the EstimateS software package. Using the Past Software (http://folk.uio.no/ohammer/past/), nonmetric multidimensional Scaling (nMDS), analysis of similarities (anosim) and nonparametric multivariate analysis of variance (npmanova) were performed on truncated matrices (containing only the OTUs supported by at least five reads) to limit the effects of small-size OTUs.
Phylogenetic assignment of the reads
The longest read of each cluster is considered as its representative. To evaluate the taxonomic composition of the samples, the representative reads were analysed using NCBI blastn against two fungal-specific databases. Representatives from the ITS1F and ITS4 data were aligned against the Fungal RSyst database, which contains 23 390 ITS sequences of the major fungal groups. It is composed of curated data from well-identified fungi collected in the UNITE database and in the GenBank database (downloaded on February 2011, see Buee et al., 2009). Representative reads from the nu-SSU-0817-5′ and nu-SSU-1536-3′ were analysed against the phymyco database (downloaded on February 2011, see Le Calvez et al., 2009), which contains 9775 SSU rRNA sequences automatically extracted from the GenBank/EMBL/DDBJ databases and curated to remove the short or low-quality sequences. The reads were assigned to a taxon based on their best blastn result using the megan software (Huson et al., 2007) employing the NCBI taxonomy. The lowest common ancestor algorithm parameters were set as follows: ‘min support = 1’ so that a taxon can be supported by a single read; ‘min score = 200’ to filter out the short-length, high-identity fragments; the other parameters were set to ‘default’.
Fungal diversity in the samples
In all, 294 102 reads were generated by the 454 pyrosequencing. A total of 209 544 reads (71.25%) passed the quality control and chimera detection steps, distributed in four files as follows: ITS1F = 62 485 reads, ITS4 = 82 097 reads, nu-ssu-0817 = 33 746 reads and nu-ssu-1536 = 31 216 reads. The average read length is 318 bp (standard deviation: 63; median: 323; minimum: 172; maximum: 546). From these files, OTUs were generated by clustering the reads at 98% sequence identity for each of the four data sets. It is noticeable that an important fraction of these OTUs are singletons: on average, they account for 54 % (standard deviation: ± 2%) of all OTUs, but they only represent 5.5 % (± 0.5%) of the reads (data not shown). We acknowledge that singletons are considered to be mostly artefacts (up to 75% according to Tedersoo et al., 2010) and can lead to an overestimation of the diversity; therefore, we removed them from the analysis.
The overall numbers of nonsingleton OTUs generated for each molecular marker are very high. ITS1F = 2877, ITS4 = 3506, nu-ssu-0817 = 1683, nu-ssu-1536 = 1455. The number of OTUs generated for each host-microhabitat combination is also surprisingly high (Table 1). For instance, the poorest sample (nu-ssu-1536 θ RB) contains 271 OTUs, while in the richest sample (ITS4 β RH), a total of 1001 OTUs were identified. The average number of OTUs for each tree-level-marker combination is 522. The number of OTUs ranged from 401 to 856 in the ITS1F data set and 464–1001 in the ITS4 data set. As expected, the nu-ssu-0817 and nu-ssu-1536 data sets showed less OTUs (respectively ranging from 353–564 and 271–386) probably because of the shallower sequencing depth. None of the rarefaction curves reached a plateau (Fig. 1), and Chao1 richness estimators are higher than the numbers of observed OTUs (40–100% higher), indicating an incomplete coverage of the fungal diversity despite the important depth of the sequencing. The Shannon diversity index computed for every host–microhabitat–marker combination ranged from 3.18 to 5.34 suggesting that the fungal communities are highly diversified. It is noticeable that for each sampling site and markers, the lowest diversity and evenness indices are found in the immersed A. marina samples and the highest indices are observed in the emerged R. stylosa. However, no specific pattern is identifiable for the emerged A. marina or the immersed R. stylosa, with contradicting indications from the different markers.
Table 1. Number of reads analysed, number of observed species, Chao1 richness estimator, Shannon diversity estimator and evenness calculated at 98% sequence identity threshold for the 12 samples and 4 molecular markers
No. of reads
No. of OTUs
No. of reads
No. of OTUs
No. of reads
No. of OTUs
No. of reads
No. of OTUs
Distribution of the OTUs in the samples
To compare the OTU composition of the fungal communities in each sample, a nMDS was performed based on the Bray–Curtis dissimilarity (Fig. 2). For the four different markers, this ordination method indicates that the samples collected from the same host microhabitat harbour similar communities and that communities found in the four host microhabitats are different from one another. To validate this result, we used two statistical tests: the anosim, a distribution-independent analogue of one-way anova; and the npmanova. Both tests require to establish a priori groups of data and use permutation (n = 10 000) to assess the statistical significance of the results. Given the indication of the nMDS, the data from the same host microhabitat were grouped together (i.e. groups are αAB-βAB-θAB; αAH-βAH-θAH; αRB-βRB-θRB; αRH-βRH-θRH). anosim results indicate a strong similarity of the data within the groups (ITS1F: R =0.972 P <0.0001; ITS4: R =0.969 P =0.0002; nu-ssu-0817: R =0.821 P =0.0002; nu-ssu-1536: R =0.887 P =0.0002) and npmanova results indicate that there is a statistically significant difference between each group (ITS1F: P < 0.001; ITS4: P < 0.001; nu-ssu-0817: P < 0.001; nu-ssu-1536: P < 0.001). This analysis indicates that there are highly significant differences in the composition of the fungal community found on both trees, strongly suggesting that the plant host specificity is an important structuring factor in the fungal colonization of mangrove trees. We also note that the host specificity has a similar effect on the structure of the communities in both the aerial and intertidal parts of the trees. This analysis also points out that the different environmental conditions encountered at the three sampling sites have a much smaller impact on the fungal community's composition, although the variations in the diversity indices denote a spatial patchiness (Table 1). To check the effect of low-abundant OTUs on these results, the OTUs distribution in the complete matrices (containing all OTUs) was analysed and provided similar results (Fig. S4).
Taxonomic composition of the fungal communities
To evaluate the taxonomic composition of each tree-level sample, the representative reads were compared, using blastn, with two databases containing only fungal sequences and then classified using megan and the NCBI taxonomy. The taxonomic assignment at the specie level for the 25 most abundant OTUs in each sample is given in Table S1. Across the whole data set, the reads were classified into 129 orders (Fig. S5). The taxonomic compositions of the host-microhabitat samples from the three sampling sites are similar, only small variations exist. Most of the taxonomic orders identified contain only a small percentage of the diversity of a given sample (< 0.1% of the reads for each order, Fig. S5). All four markers indicate that six orders dominate the samples. The Capnodiales and Pleosporales have a broad distribution pattern, with no apparent specificity for one of the microhabitat. Three orders exhibit strong tree-level specificity: the Lecanorales are found mainly on the emerged R. stylosa (RH), the Dothideales on the emerged A. marina (AH) and the Xylariales on the immersed A. marina (AB). The Diaporthales exhibit the most singular distribution pattern, being mostly found on both the emerged A. marina (AH) and the immersed R. stylosa (RB). However, even if the four data sets provide similar taxonomic compositions, some differences are observable: the Eurotiales and Magnaporthales found preferentially in the inundated samples are absent in the SSU rRNA data sets while the Arthoniales, Botrysphaeriales, Chaetothyriales and Helothiales are much less abundant in the ITS data sets. At the class level, variations between the sampling sites and between the markers fade, but marked differences persist between the four microhabitats studied (Fig. 3). The Dothideomycetes are prominent in all the samples and represent from 25% (± 4.8%) to 55% (± 3%) of the reads. Lecanoromycetes are the dominant class in the emerged R. stylosa (RH) and are five times more abundant in this microhabitat than in any other. Sordariomycetes also exhibit a preferential distribution pattern, as they are up to three times more abundant in the immersed microhabitat than in the emerged one. Finally, analysis at the phylum level reveal that the relative abundances of the Ascomycetes and Basidiomycetes in the four data sets are somewhat similar to those previously established in the literature, with 82% (± 5%) of the reads assigned to the Ascomycetes and 2% (± 1%) of the reads to Basidiomycetes (data not shown). The remaining 16% correspond to reads not assigned to any taxon. Comparative analysis of the four microhabitats reveals that the percentage of reads assigned to the Basidiomycetes is slightly higher in the emerged samples at 3% (± 1.5%) of the reads and lower in the immersed samples with only 0.5% (± 0.3%).
To our knowledge, this is the first study of the fungal diversity in mangroves using next-generation sequencing of the rDNA. A culture-independent approach was chosen to obtain a broader view of the different fungal communities colonizing mangrove trees. We focused our interest on the fungal communities occurring on trees, as they are vastly understudied despite the fact that they could play an important role in the mangrove ecosystem. To make a step towards a more complete view of mangrove fungi, we studied both the immersed and emerged parts of two of the most commonly encountered mangrove trees.
The tag-encoded 454 pyrosequencing approach used in this study was shown to be extremely powerful to study the fungal diversity (Buee et al., 2009; Jumpponen & Jones, 2009), but it relies on a critical clustering step. Although the clustering will determine the number of OTU and therefore all the diversity indicators classically used, there is currently no consensus over which clustering algorithm is the most suitable for this kind of approach. During a preliminary study (data not shown), we compared three clustering algorithms on one of our data set: CD-HIT-EST, blastclust (Altschul et al., 1990) and uclust (Edgar, 2010). Our choice of using CD-HIT-EST was based on the facts that it generated 50% less OTU and it is orders of magnitude faster than blastclust and also tends to generate fewer singletons than blastclust and uclust (respectively 35% and 15% less). Our choice of using CD-HIT-EST was also shared by Schloss & Westcott (2011) who compared the performance of these three clustering tools and demonstrated that CD-HIT-EST was a good heuristic solution to cluster reads to an OTU. The 98% sequence identity threshold chosen for clustering is consistent with the 2.51% intraspecific variability of the fungal ITS (weighted average) or with the 1.96% intraspecific variability for the Ascomycota ITS calculated by Nilsson et al. (2008). Le Calvez et al. (2009) also used 98% sequence identity to cluster fungal SSU rRNA reads. We also note that the number of OTUs generated is closely related with the number of reads in each file, with an R2 = 0.9999 (data not shown). This suggests that for each molecular marker, similar number of OTUs could have been generated for similar depth of sequencing.
In this study, we used two different pairs of primers to amplify the ITS and SSU rRNA V5–V7 region of the rDNA, to circumvent the biases caused by primer specificity. Indeed, Hong et al. (2009) demonstrated that different primer pairs could recover different fractions of a microbial community. We also decided to sequence the amplicons libraries from both their 5′ and 3′ ends, to obtain the ITS1 and ITS2 regions of the ITS amplicons and the V5 and V7 regions of the SSU rRNA amplicon without introducing the bias of a new primer. An equimolar mix of all amplicons was prepared before sequencing, but despite a careful estimation of DNA concentration, very disparate numbers of reads were obtained. The nu-ssu-0817 and nu-ssu-1536 reads were twice less numerous than the ITS1F reads and three times less numerous than the ITS4 reads. We hypothesize that these variations arose from a inaccurate estimation if the amplicons average length (this value was used to convert the DNA concentration into a number of amplicons per volume unit) and to a difference in specificity during the A and B 454 linkers ligation biased towards the ITS4 end of the ITS amplicons.
The taxonomic composition of the communities given by the four molecular markers, although convergent, showed noticeable differences, with several abundant taxonomic orders differing from one data set to the other. Part of this difference could be linked to the use of two different databases containing different diversity, but we noted that these orders have similar representations in both databases (ratio of sequence in the Fungal R-Syst and the phymyco databases: Arthoniales: 10, Botrysphaeriales: 0.35, Chaetothyriales: 0.95, Helothiales: 1.1). This result suggests that using multiple primers is a way to enhance the quality of the description of the taxonomic composition of communities. Our results also indicate that all four markers give very similar values of fungal diversity; therefore, using any of these markers appears to be sufficiently robust to assess the alpha and beta diversity. A standardization of the molecular marker studied and the PCR primers used would greatly benefit to the scientific community, as it could allow comparing the diversity observed in various studies. The ITS-1 region amplified with the ITS1–ITS4 primers seems to be the best candidate for such standardization as it is currently the most used in fungal diversity studies. On the other hand, if the aim of the study is to precisely describe the taxonomic composition of a community, it seems that the use of a single marker leads to poor results, because of the primer specificity and the short length of the reads. Primer specificity issues can be mitigated using multiple primer pairs targeting one or several molecular markers. Addressing the short reads length could be done by sequencing the amplicons from both 5′ and 3′, and assembling the reads might provide longer reads and thus more reliable taxonomic assignments. But it could also generate chimeras for certain well-used markers (i.e. assembling the ITS-1 and ITS-2 reads would be based on the 5.8S sequences, which are highly conserved thus leading to potentially false results). Other molecular regions, in particular phylogenetic markers like β-tubulin, EF1 alpha or RPB1/2 genes, could benefit from this amplicon-assembling methodology, but it would require the creation of consequent quality databases to correctly assign the reads to a taxon.
Community structure and OTUs distribution
Using the tag-encoded 454 pyrosequencing approach, we described extremely diverse fungal communities, with several hundreds OTUs generated in each sample. Such values were not anticipated, as they are orders of magnitude higher than any of fungal species richness reported previously from mangroves. For instance, Schmit & Shearer (2004) identified in a meta-analysis 75 fungal species on A. marina and 32 on R. stylosa and an overall fungal diversity of 163 species collected on 16 tree host species. In a similar fashion, based on intensive search of literature records of xylophilous Basidiomycetes in mangroves, a list with 112 species were presented by Baltazar et al. (2009), which is less than the 278 OTUs assigned to the Basidiomycota phylum in the ITS1F data set (ITS4: 227 OTUs; nu-ssu-0817: 227 OTUs; nu-ssu-1536: 187 OTUs). Furthermore, the latest review of the described marine fungi (Jones et al., 2009) report 530 species, compared with 2048 OTUs we described in the inundated part of the mangrove (ITS4 data set, data not shown), indicating a large gap in our knowledge of the mangrove colonization by fungi.
Our observations also indicate that only a very small number of fungal OTU dominate each sample, while the majority of the diversity is present in minute quantities (Table S1). This seems to confirm previous analysis concerning microbial eukaryote species, revealing that numerous environments contain a few functionally active species and a large ‘seed bank’ of species able to survive under different conditions (Finlay, 2002), corresponding to a potential functional reserve essential for the ecosystem's resilience in response to environmental disturbances.
The analysis of the OTUs spatial distribution reveals that the fungal communities found on A. marina and R. stylosa are markedly different. The vast majority of the OTUs are found almost exclusively in a single microhabitat (AB, AH, RB or RH). When considering the dominant OTUs, we note that none of them is ubiquitous, 80% are found majoritarily in a single microhabitat and 20% of them are found in similar abundance in two microhabitats. Interestingly, these two microhabitats always share a common tree host or a common sea level. These data suggest a strong host specificity of the fungal colonization of both the aerial and intertidal parts of the mangrove. This is relatively surprising, as we assumed that the presence of seawater would create a continuous, homogeneous media, recycled cyclically by the tides, in which fungal dispersion and propagation between trees would be promoted. Host specificity in the aerial part was expected, as it has been shown to happen in plant pathogens fungi, mycorrhizas, endophytes and saprobes (Zhou & Hyde, 2001). Several studies have described this effect in terrestrial ecosystems, in soils and in phyllospheres where it occurs to varying degrees depending on the climate and the fungal lineage considered (Ferrer & Gilbert, 2003; Buée et al., 2007; Tedersoo et al., 2008). However, there was little evidence of this phenomenon in marine fungi or in intertidal mangrove fungi. Jones (2000) pointed out that some fungi occurred more readily on test blocks of one wood type than any another. Meanwhile, another research group (Hyde, 2007) indicated that certain fungal species were only associated with a host species or genera. In the same way, some authors reported that fungal endophytes show some degree of host specificity at least for families of host trees and that such specificity may influence endophytes distribution more than geographical location of the host plants (Petrini & Carroll, 1981; Suryanarayanan and Kumaresan, 2004). On the other hand, Schmit & Shearer (2004) revealed through a meta-analysis of published data that mangrove trees that are close phylogenetically do not necessarily harbour microfungal communities that are distinctly different from less closely related hosts. In our study, given the spatial proximity of the A. marina and R. stylosa sampled at each site (less than a few metres), it is unlikely that the variations observed are linked to differing environmental conditions. It is probable that the host specificity observed in this study is enhanced by the high phylogenetic differences of the two host species studied, respectively, refereed to the asterids and rosids clades. We also note that the variability of the fungal community between the three sampling sites was limited. This is probably caused by the very small variations of micro-environmental conditions between these locations. Indeed, despite being over 150 m apart from each other, the sampling sites are exposed to similar temperatures, humidity, pH and tide level. The most noticeable difference is that the α site is more exposed to waves and spray than β or θ.
Furthermore, given the fact that the samples were collected in the transition zone between the two tree stands, it is probable that our data are not totally representative of the fungal diversity in this mangrove. The environmental conditions encountered in the A. marina zone (higher salinity, shorter exposure to seawater, higher exposure to wind and higher light exposure) and the R. stylosa zone (higher tree density, higher tide level and lower exposure to light) mean that further fungal diversity could be retrieved from these loci.
Taxonomic diversity and functional groups
The taxonomic assignation of OTUs confirms that the mangrove is an ecosystem largely dominated by the Ascomycetes in both its aerial and its submerged parts and that Basidiomycetes are extremely rare, especially in the marine compartments. They are slightly more frequent in the phyllosphere, where they account for 3% of the diversity and are evenly distributed between the A. marina and R. stylosa. These Ascomycetes/Basidiomycetes ratios were expected, as they comparable to those found by Jumpponen & Jones (2009) in the phyllosphere of Quercus macrocarpa (continental climate) and El-Said (2001) in the phyllosphere of bananas (tropical climate) and to the value reported by Jones et al. (2009) in a review of marine fungi.
The low abundance of Basidiomycetes in this mangrove is probably coupled to the fact that most Basidiomycetes are saprophytes (Agrios, 2005; Mohapatra, 2008) unable to thrive on the tissues we sampled from living trees. Indeed, in this mangrove, we had the opportunity to study the fungal diversity in a sediment stratum containing dead R. stylosa tissue (Arfi et al., 2011). We observed a clear shift in the Ascomycetes/Basidiomycetes ratio, which was established at 1, and the most abundant OTU was affiliated to a Basidiomycetes.
It appears from our results that the Basidiomycetes fungi do not play any significant role in a healthy mangrove. On the other hand, the vast majority of the most abundant, functionally relevant OTUs are assigned to Ascomycetes species. The intertidal samples harbour widespread plant pathogens (Eutypella, Phaeophleospora and Phaeosphaeria), root pathogens (Gaeumannomyces, Cytospora, Magnaporthe and Pyricularia) and root endophytes (Leptodontidium), while lichens (Pyrrhospora) and saprobes (Plicaria) are less abundant. The importance of pathogens in the inundated compartments is probably the main reason to the strong host specificity observed as these fungi generally able to colonize a limited number of plant species (Agrios, 2005). In the aerial samples, plant pathogens (Phaeoramularia and Mycosphaerella) and leaves pathogens (Diaporthe and Ramulispora) are also well represented. We note that the most prominent amphibious OTUs (i e. found in both the aerial and intertidal samples) correspond to saprophytic fungi (Rhytidhysteron, Alternaria). This is coherent with the results of Osono (2006), which indicate that some phyllosphere fungi of forest trees are primarily saprobic, being specifically adapted to colonize and utilize dead host tissue. We also note that the R. stylosa aerial parts are dominated by lichens from the Lecanorales order (Amandinea and Calicium), while there are almost absent from the A. marina trees. This specificity might derive from the physiology of the mangrove trees: the Avicenniae species have the unique feature of excreting salt by foliar glands located on the lower surface of leaves (Osborne & Berjak, 1997; Gilbert et al., 2002) and are therefore coated with salt, while the Rhizophoracea exhibit a more classical physiology. The literature indicates that halophytic Lecanorales (and generally Lecanoromycetes) are rare with only three marine species in one genus (Dactylospora) identified. We also note that Hortaea werneckii and Aurebasidium pullulans, which are known to colonize hypersaline environments (Kogej et al., 2005), are predominantly found in the AH samples. It is also interesting to note that our approach detected more than 50 Lecanoromycetes genera in the immersed samples, which highlights a gap in current knowledge on marine fungi. The most prominent genera were Rhizocarpon, Lecanora, Pleopsidium, Rhizoplaca and Rinodia with corresponding bit scores supporting these assignments similar to those encountered across the whole data set (generally between 250 and 350).
We are relatively surprised by the fact that the cosmopolitan Xylariales species were predominantly found on the submerged A. marina, as there is no evidence of such specificity in previous work and they are usually very well represented in tropical host plants where many occur as endophytes (Davis et al., 2003). We also take note that the estimation of the number of Xylariales OTUs generated in this study could be overestimated, as Nilsson et al. (2008) showed that some fungal species belonging to this order could have very important intraspecific variability of their ITS (up to 24.2% for Xylaria hypoxylon).
To conclude, our study revealed that mangrove tree harbours fungal communities of previously unforeseen richness and diversity in both their aerial and intertidal parts. These communities are mainly composed of plant pathogens, while saprophytic fungi are extremely rare. The plant host specificity is one of the major factors influencing the distribution of the fungal species in this ecosystem, a phenomenon that had never been studied in mangroves using a global approach. This strong zonation is a probably a result of the very different physiology exhibited by the two mangrove tree species studied. In future studies of the mangrove fungal communities, it would be interesting to focus on fungi presenting an amphibious distribution to improve their taxonomic assignation to deepen our understanding of the origins of marine fungi (Hibbett & Binder, 2001) and the role of mangroves as an interface between marine and terrestrial ecosystems. On the other hand, isolation of mangrove fungi should continue to enrich the databases with new species and to further study the potential of these organisms in biotechnology (Raghukumar, 2008).
The authors thank Thierry Dostes (CNRS-IMM, Marseille) for his advice and technical IT support, Dr Martine Rodier (IRD UR 167 CyRoCo, Nouméa) and Dr Jonathan Deborde (IRD UR206, IMPMC, Nouméa) for their support in New Caledonia, Dr Francis Martin (INRA UMR 1136 IAM, Nancy) for the access to the UMR 1136 IAM computing resources and Dr R. Henrik Nilsson (Department of Plant and Environmental Sciences, University of Gothenburg) for the scientific discussion on clustering algorithms. The corresponding author is indebted to Audrey Finot for her critical assistance with the PCR experiments and to Jedd Ansell-Milla for his friendly review of an early draft of this manuscript. Sampling authorization was granted by the Direction de l'Environnement (DENV) of the South Province. This work was partly funded by the INRA AIP.