Effect of forest fire prevention treatments on bacterial communities associated with productive Boletus edulis sites

Summary Cistus ladanifer scrublands, traditionally considered as unproductive, have nonetheless been observed to produce large quantities of king bolete (Boletus edulis) fruitbodies. These pyrophytic scrublands are prone to wildfires, which severely affect fungi, hence the need for fire prevention in producing C. ladanifer scrublands. In addition, B. edulis productions have severely decreased in the last years. A deeper understanding of the B. edulis life cycle and of biotic and abiotic factors influencing sporocarp formation is needed to implement management practices that facilitate B. edulis production. For example, some bacteria likely are involved in sporocarp production, representing a key part in the triple symbiosis (plant–fungus–bacteria). In this study, we used soil DNA metabarcoding in C. ladanifer scrublands to (i) assess the effect of site history and fire prevention treatment on bacterial richness and community composition; (ii) test if there was any correlation between various taxonomic groups of bacteria and mycelial biomass and sporocarp production of B. edulis; and to (iii) identify indicator bacteria associated with the most productive B. edulis sites. Our results show that site history drives bacterial richness and community composition, while fire prevention treatments have a weaker, but still detectable effect, particularly in the senescent plots. Sporocarp production correlated positively with genera in Verrucomicrobia. Several genera, e.g. Azospirillum and Gemmatimonas, were identified as indicators of the most productive sites, suggesting a potential biological role in B. edulis fructification. This study provides a better understanding of the triple symbiosis (plant–fungus–bacteria) involved in C. ladanifer–B. edulis systems.

However, these ecosystems, and the fungal communities associated with them, are frequently affected by wildfires (Mart ın-Pinto et al., 2006), which is detrimental to the majority of these fungi (Jim enez-Esquil ın et al., 2007). This fact has lead to the need for fire prevention treatments, which could improve mushroom production levels (Savoie and Largeteau, 2011).
In addition, due to the severe drought suffered in the Mediterranean basin, B. edulis production has plummeted since autumn 2014, with near zero production in some areas in Spain in the autumn of 2017 (O. Mediavilla, personal observation). This dramatic situation has highlighted the need for a deeper understanding of the B. edulis life cycle and the environmental factors and biological interactions that influence sporocarp formation.
Besides abiotic environmental factors, biotic factors, such as plant physiology and interactions with other soil fungi and bacteria, are also likely to be involved in sporocarp production (Antony-Babu et al., 2013). A greater understanding of bacterial communities in forest soils and the effect of different forest management options is essential because these microorganisms represent a key part in the triple symbiosis (plant-fungus-bacteria) (Barbieri et al., 2005;Bonfante and Anca, 2009), and they represent the most abundant microorganisms in forest soils (Hardoim et al., 2015;Baldrian, 2017). Despite this fact, to date, bacteria in forest soils have received less attention than fungi. Bacteria are ubiquitous microbes in many environments, where they may have several symbiotic functions in mushrooms. Some bacterial species play essential roles in nutrient cycling (Pent et al., 2017), mediating multiple critical steps in the nitrogen cycle, including N fixation (Llad o et al., 2017), which has a strong influence on the abundance of ectomycorrhizal fungi (Frey et al., 2004;Allison et al., 2007). Other bacterial species take part in C cycling processes (Fierer et al., 2013), increasing the availability of carbon for the plant and for the associated fungi that obtain their carbon source from other organisms (Honrubia, 2009). Some groups of bacteria can act also as mycorrhiza helper bacteria (MHB), which are able to stimulate the development of mycorrhiza, or to the contrary, act as pathogen inhibitors and antagonists (Frey-Klett et al., 2007). In this context, previous studies have focused on the role of bacteria in the production of fruiting bodies of other edible fungal species, such as truffles (Splivallo et al., 2015).
The symbiotic development of mycorrhizal fungi on plant roots has been reported to be influenced by bacteria present in the mycorrhizosphere (Barbieri et al., 2005). Although the role of bacteria in the formation of B. edulis mycorrhiza has already been demonstrated (Mediavilla et al., 2016), a deeper understanding of the effect of natural bacterial communities on the mycelium in the soil and on sporocarp production is needed.
To date, there is no information about bacterial communities present in the mycorrhizosphere of the C. ladanifer-B. edulis symbiosis and the possible influences of management treatments. We hypothesize that bacterial communities are affected by site history and fire prevention treatments in these ecosystems. The bacteria observed could play an active role in the presence and fructification of B. edulis, and some of them may even act as indicator species at the most productive B. edulis sites. Therefore, the aim of this study was to analyse the diversity of bacterial communities and their interaction with B. edulis and their potential role in B. edulis sporocarp production under different forest fire prevention treatments. Our specific aims were (i) to investigate the effects of different fire prevention treatments and site history on the bacterial richness, abundance and composition; (ii) to determine whether there was a correlation between bacterial communities and B. edulis sporocarp production and mycelium present in the soil; and (iii) to identify bacterial indicator species associated with the most productive B. edulis sites.

Taxonomic composition of the microbial community
The 2165 bacterial OTUs were classified into 24 phyla and 381 genera. Most of the OTUs were identified at the genus level; however, identification down to the species level was impossible owing to database limitations. The soil microbial community was dominated by Proteobacteria, which represented 25% of the bacterial community, followed by Actinobacteria (14%), Acidobacteria (11%), Planctomycetes (10%), Bacteroidetes (9%), Firmicutes (8%), Chloroflexi (5%) and Verrucomicrobia (4%). Less representative phyla (up to 9% of the OTUs) had a lower dominance; 4% of OTUs were not identified. The total number and proportional distribution of OTUs among different phyla are shown in Fig. 1. Within the phylum Proteobacteria, up to 44% belonged to the class Alphaproteobacteria, 23% were Gammaproteobacteria, 21% were Deltaproteobacteria and 12% belonged to the class Betaproteobacteria (Fig. 1). Genera with the highest OTU richness are presented in Table 1.
Relationship between the abundance of bacterial communities and B. edulis sporocarp production and extraradical B. edulis mycelium in the soil To determine relationships between bacterial abundance and B. edulis sporocarp production and extraradical mycelium in the soil, linear regressions were carried out for every phylum. For sporocarp production, we observed a statistically significant, but weak positive correlation for the relative abundance of the phylum Verrucomicrobia (R 2 = 0.20, P = 0.01), and weak negative correlations for the phyla Actinobacteria (R 2 = 0.20, P = 0.01) and Planctomycetes (R 2 = 0.14, P = 0.03).
However, the relative abundance levels of the bacterial phyla did not correspond significantly to the amount of extraradical B. edulis mycelium in the soil. Only the abundance of Planctomycetes showed a nearly significant negative correlation with the amount of extraradical B. edulis mycelium in the soil (R 2 = 0.13, P = 0.068).

Impact of site history and fire prevention treatments on bacterial communities
The bacterial richness was significantly affected by the site history of the stand (P = 0.039). Mean OTU richness was highest in the senescent plots (371) and lowest in the young burnt plots (328), and the difference between them was significant (P = 0.002). There was also a significant difference between the richness in the young burnt plots and the young cleared plots (361) (P = 0.013). The different fire prevention treatments only had a significant effect on bacterial richness in the senescent plots, where bacterial richness was significantly higher in plots with total clearing than in the control (P = 0.008). Analysing the effects of the different treatments on the richness of each phylum revealed substantial differences regarding their responses to treatments at sites with different site histories. However, the richness of almost all phyla tended to increase in senescent plots after either total clearing or controlled burning treatments. The richness of Acidobacteria (P = 0.032) and Proteobacteria (P = 0.040) was significantly affected by the treatments (Fig. 2). Both these phyla showed a peak in OTU richness in senescent plots that had been totally cleared.
The treatments significantly affected the relative abundance levels of Planctomycetes (P = 0.003). The only significant difference was in the senescent plot where a  controlled burning had been performed, and these plots showed the highest abundance (Fig. 3). The NMDS analysis resulted in a two-dimensional solution with a final stress value of 0.0693. The ordination plot revealed a strong structuring of bacterial communities according to the site history and, to a lesser extent, to the treatment (Fig. 4). MRPP confirmed the importance of the site history in shaping bacterial community composition (effect size A = 0.0907, probability P < 0.00001). Similarly, PERMANOVA indicated that bacterial community composition differed significantly among the different site histories (P = 0.0002).
Our analysis to identify bacterial indicator species associated with the most productive B. edulis sites revealed 18 OTUs that were characteristic of plots with high B. edulis sporocarp production (Table 2), including members of the Azospirillum, Gemmatimonas and Opitutus genera.

Discussion
The phyla Proteobacteria, Actinobacteria and Acidobacteria were the most abundant, which support the findings reported in previous studies of forest soils (Deveau et al., 2016;Sun et al., 2017). These phyla seem to be abundant in most soils (Lauber et al., 2009), which appears to indicate their functional importance (Llad o et al., 2017). Alphaproteobacteria represented the most prevalent class within Proteobacteria. This finding could be explained by the low pH of the soils at the study site. pH is an important driver of bacterial community composition (Lauber et al., 2009), and acidic soils are usually dominated by Alphaproteobacteria and Acidobacteria (Felske et al., 1999;Kaiser et al., 2001;Baldrian et al., 2012;Shen et al., 2013). Some members of the class Alphaproteobacteria mediate N 2 fixation (Llad o et al., 2017), strongly influencing the abundance of ectomycorrhizal fungi (Lauber et al., 2008). The Alphaproteobacterial genera Sinorhizobium and Rhizobium, present in the studied soil, both harbour multiple nitrogen fixer species (Barbieri et al., 2005).
Regarding the correlation analysis between bacterial abundance and B. edulis sporocarp production, some previous studies have shown that although interactions between bacteria and fungi can be crucial (Boer et al., 2005), these interactions can be positive, negative or neutral, i.e. they can involve pathogenic or helper bacteria as well as mutualists (Bonfante and Anca, 2009;de Boer, 2017). In our study, the phylum Verrucomicrobia showed a weak positive correlation with the production of sporocarps. Buckley and Schmidt (2001) suggested that Verrucomicrobia abundance may increase with soil moisture content. At our study sites, an increase in shrub cover increased soil moisture in the soil and was associated with increased B. edulis sporocarp production. There was a suggested negative correlation for Actinobacteria which could be explained by its heliophilous behaviour. The abundance of these bacteria tends to increase in the absence of vegetation, a trend that we observed in the treatments where the host plants were removed. This response was observed by Zhang et al. (2016) who reported a significant increase in these bacteria after forest clear-cutting. By contrast, the abundance of B. edulis mycelium in the soil, and  consequently sporocarp production, tends to decrease when the host is eliminated (Mediavilla et al., 2017). When analysing the impact of fire prevention treatments on bacterial communities, most of the phyla did not show a significant response in terms of richness and abundance. This suggests a certain resilience of the bacterial communities after forest management treatments and supports the results of Smith et al. (2008) who highlighted the important role of bacteria in the resilience of forests to disturbances and in the regeneration processes. Other forest management practices have also been reported to have a minor impact on soil bacterial community structure and diversity (Nacke et al., 2011). The dominant vegetation and land use were reported to affect fungi more strongly than bacteria, especially the ectomycorrhizal taxa because of their dependence on the host plant (Bu ee et al., 2009;Zinger et al., 2011).
Both total clearing and controlled burning increased bacterial richness, which may be explained by the increase in stand heterogeneity. Senescent C. ladanifer stands normally have a high density of the plants reaching almost total cover (Hern andez-Rodr ıguez et al., 2013), which provides a rather homogeneous structure with a low diversity of niches. Both the total clearing and the controlled burning treatments essentially removes most of the closed scrublands, creating new available niches microorganisms. At this point, any perturbation would be desirable in order to increase the bacterial richness (Santill an et al., 2018).
In terms of proportional abundance, the only bacterial phylum that was affected by the fire prevention treatments was the Planctomycetes, which appeared to be more abundant in the senescent plots where the controlled burning was performed. It was previously observed that some years after prescribed burning, nutrient levels and mineralization rates decrease in the affected soils (Choromanska and DeLuca, 2001;Reich et al., 2001;Carter and Foster, 2004). In these conditions, Planctomycetes could survive since their members are typically slow-acting decomposers (Dedysh and Kulichevskaya, 2014).
The young burnt plots displayed the lowest bacterial richness. This could be because most of the phyla detected at the study sites prefer acid soils. However, the pH in the young burnt plots was likely higher than in the other plots because fire tends to increase soil pH (Shen et al., 2016) due to the deposition of ash (Certini, 2005). Our findings agree with those of Smith et al. (2008), who reported that bacterial communities tend to recover faster in cleared sites than in burned ones.
When analysing bacterial composition, we emphasize the strong influence of the stand site history in our study. Bonfante and Anca (2009) also reported that bacterial community composition is determined by site history. This also supports the findings of Smith et al. (2008), who reported the effect of harvesting and fires on the composition of forest soil microbial communities. Fires drive a shift in the bacterial community structure (Rousk et al., 2010) due to the increase of pH. Baath et al. (1995) also reported that an increase in soil pH due to fire had a noteworthy effect on bacterial community composition. Specifically, we observed a higher abundance of Firmicutes in the young plots, which were most  Smith et al. (2008), who also reported that Firmicutes were characteristic in post-fire soils. This phylum is able to survive under extreme conditions and may be favoured by wildfires. Our results revealed some bacterial OTUs that could be considered as indicator species of the most productive B. edulis sites. Of these, the genus Azospirillum represents the best-characterized genus containing plant growth-promoting rhizobacteria (Steenhoudt and Vanderleyden, 2000). Its members possess a number of beneficial properties, including nitrogen fixation, production of indole-3-acetic acid (IAA) and deamination of the ethylene precursor 1-aminocyclopropane-1-carboxylate (ACC) (Creus et al., 2005;Blaha et al., 2006). Furthermore, plants associated with Azospirillum communities develop an increased number of lateral roots and root hairs, which will enhance not only the amount of root surface area available for nutrient uptake (Steenhoudt and Vanderleyden, 2000) but also the number of mycorrhizal roots. We also identified the genus Gemmatimonas as an indicator species, which is considered to be a phosphate-solubilizing bacterium (Yang et al., 2017) and has also been listed among N 2 -fixing bacteria (Krieg et al., 2010). The genus Opitutus was the most predominant among the indicator species. This genus belongs to the phylum Verrucomicrobia, from which we obtained a positive correlation with B. edulis sporocarp production, which could be due to their role in C cycling (Fierer et al., 2013).
Based on the results obtained in this study, it is possible to suggest some management directions. Because senescent stands tend to have lower bacterial richness, perturbation resulting in the rejuvenation of these areas are desirable, to trigger an increase in bacterial richness. This fact will also allow for an increase of B. edulis sporocarps production and for a decrease in potential fire severity by reducing the amount of available fuel. In this context, some mechanical management (i.e. clearing) seems to be the best option, because its effect on bacterial communities is less severe than in the case of burning, and this treatment showed favourable effects on bacterial richness in our study. However, it is desirable to maintain habitat heterogeneity at the landscape scale by leaving patches of senescent scrublands intact to maximize overall bacterial richness in the area.
Here we present insights into the bacterial communities associated with highly productive B. edulis sites in C. ladanifer scrublands under different fire prevention treatments, using a metagenomics approach. We demonstrated that the site history influenced the bacterial composition and identified bacterial species that were indicators of highly productive B. edulis sites, noting the biological role played by some bacterial phyla.
Despite these findings, there is still much uncertainty about the association between individual bacterial species and fungal fruiting. Although further research is needed in this field, this study provides a better understanding of the triple symbiosis (plant-fungus-bacteria) involved in C. ladanifer-B. edulis systems.

Study site
The study area was located in Zamora province, north-western Spain (0730462-0731929 longitude-UTM, 4619644-4621757 latitude-UTM 29T Grid), 750-780 m above sea level. Twenty-seven plots (2 9 50 m) were established at the study site. The characteristic natural community is dominated exclusively by C. ladanifer, with ca. 80% canopy coverage. This study site has been widely used as an experimental area to study the effects of fuel reduction treatments on fungal communities (Hern andez-Rodr ıguez et al., 2013, 2015b. Based on the soil characteristics, the soil in this zone is classified as Inceptisol suborder Xerept (Soil Survey Staff, 2010) and characterized by stoniness, acidity (pH 5.0-5.5), and lack of calcium (1.5-4.0 meq 100 g À1 , atomic absorption method) and phosphorus (< 4.0 mg kg À1 , Olsen method). The availability of nitrogen (0.09-0.18 g 100 g À1 , modified Kjeldahl method) and potassium (71-128 mg kg À1 , atomic emission method) is variable, with a good level of humification. This region is characterized by a sub-Mediterranean climate: mean temperatures range from 14.5 to 15.8°C, and annual precipitation ranges from 450 to 700 mm. Climatic data were provided by the closest meteorological station (Alcañices, 0724617 Longitude-UTM, 4618218 Latitude-UTM, 29T Grid and 806 m above sea level, Spanish Meteorological Agency).

Fuel reduction treatments
The study site comprised three areas with different ages and site histories: (a) an eight-year-old stand regenerating from a wildfire (young burnt); (b) an eight-year-old stand following total clearing of the previous stand (young cleared); and (c) a 20-year-old senescent stand following a natural wildfire (senescent). Because of the short cycle life of C. ladanifer, a stand of 20 years of age is considered as senescent or declining (Oria- de-Rueda et al., 2008). These three stands are located in the same area of C. ladanifer scrubland and share similar mesoclimatic and edaphic conditions. Specifically, young burnt and young cleared areas are around two hundred metres apart, while the senescent area is a kilometre away. More information on the experimental setup can be found in Mediavilla et al. (2017).
Forest fire prevention treatments were applied depending on their feasibility in accordance with the age of the stands and vegetation characteristics. In the eightyear-old stands (i) and (ii), the fuel reduction treatments were as follows: (1) uncleared, which acted as the control; (2) 50% cleared; and (3) total clearing, i.e. all C. ladanifer plants were removed. In the 20-year-old stand (iii), the treatments were as follows: (1) uncleared, which acted as the control; (2) total clearing; and (3) controlled burning (Table 3).
Total clearing was performed using a tractor with a brush thrasher mower; whereas, the 50% cleared treatment was carried out manually by removing half of the plants with a brush cutter. Both clearing treatments were carried out in spring 2010. Controlled burning was performed with the help of Zamora EPRIF (Integral Fire Prevention Team) (Ministry of Agriculture, Food and Environment) in October 2010 under favourable weather conditions that allowed ignition without the risk of uncontrolled fire. For each treatment, we collected samples from three different plots. In summary, we studied three different stands with three different forest fire prevention treatments per stand, with three plots per treatment, resulting in a total of 27 plots sampled.

Sampling and molecular work
At each site, in December 2013, five soil cores, 3.5 cm in diameter and 26 cm in length, were extracted at 5 m intervals along the longitudinal axis of the plot (Taylor, 2002). The soil was dried at room temperature, and coarse elements were discarded. The five cores were pooled resulting in a composite soil sample per site. DNA was extracted from 0.25 g of dry soil per sample using a PowerSoil â DNA Isolation Kit (MoBio Laboratories, Carlsbad, CA, USA) and according to manufacturer's instructions. The primers 515F and 806R-trP1 (Caporaso et al., 2011(Caporaso et al., , 2012 were used to amplify a portion of ca. 255 bp of bacterial 16S rDNA. The forward primer was labelled with sample-specific multiplex identification DNA-tags (MIDs). For each of the 27 samples, the following PCR protocol was used for three positive and one negative reaction: one cycle of 95°C for 5 min, then 35 cycles of 95°C for 20 s, 54°C for 30 s, and 72°C for 1.5 min, ending with one cycle of 72°C for 7 min. Negative PCR reactions were performed for each primer pair and contained elution buffer instead of DNA. PCR products were checked for DNA concentrations using QIAxcel Advanced System (Qiagen, Venlo, The Netherlands). Emulsion PCR and Ion Torrent sequencing was carried out at the Naturalis Biodiversity Center (Leiden, The Netherlands). We used the sequencing Ion 318 TM Chip to allow for the highest possible sequencing coverage.
Quantification of B. edulis mycelium in the soil and sporocarp production levels Boletus edulis DNA was amplified using real-time PCR (qPCR) in an Applied Biosystems â 7500 Real-Time PCR System (Applied Biosystems, Mannheim, Germany), using the kit qPCR Boletus edulis-VK provided by Vacunek S.L. (Bilbao, Spain), and according to the manufacturer's instructions and a final reaction volume of 25 ll. The kit provides the necessary reagents and enzymes, mixed in a single master mix. The kit uses a Taqman probe marked with FAM-BHQ1 â and also a positive internal control with primers and a Taqman probe marked with JOE-BHQ1 â , which allows for the detection of false negatives caused by inhibition. Real-time PCR cycling conditions were 10 min at 95°C, 45 cycles at 95°C for 15 s and 60°for 60 s. Three replicates of each sample were included in the analysis, as well as a negative control (using deionized water instead of DNA template). A standard curve with five points and three replicates per point using known amounts of mycelium in the soil was generated. This curve was used to translate the outputs from the qPCR System into amount of mycelium in soil. Ten-fold serial dilutions were performed from 11 400 000 to 1140 ng of mycelium g of soil À1 . See Mediavilla et al. (2017) for more detailed information. To quantify the production of B. edulis sporocarps, all B. edulis sporocarps were collected and weighed on a weekly basis during the mushroom season from 2010 to 2013. The sporocarps were harvested, transported to the laboratory and stored at 4°C. Fresh weight was measured (kg fw ha À1 year À1 ) within 24 h after collection (Hern andez-Rodr ıguez et al., 2015b).

Bioinformatic analysis
The initial clean-up of the raw sequence data was carried out using the online platform Galaxy (https://main. g2.bx.psu.edu/root), in which the sequences were sorted into samples, and identification tags were removed. Poor-quality ends were trimmed off based on a 0.02 error probability limit in GENEIOUS PRO 5.6.1 (BioMatters, Auckland, New Zealand). Subsequently, sequences were filtered using USEARCH V.8.0 (Edgar, 2010) based on the following settings: all sequences were truncated to 200 bp and sequences with an expected error of > 1 were discarded. For each sample, sequences were collapsed into unique sequence types while preserving their counts and excluding singletons. The quality-filtered sequences from all samples were grouped into operational taxonomic units (OTUs) at 97% sequence similarity and putative chimeric sequences were removed using USEARCH. We assigned sequences to taxonomic groups of bacteria based on pairwise similarity searches against the curated Gold database of the Ribosomal Database Project (Cole et al., 2009). The resulting matrix containing only OTUs with at least 70% sequence similarity to a bacterial reference sequence had 105 553 quality-filtered sequences. We rarefied the OUT matrix to the smallest sequence count per sample (2342 sequences), and the resulting matrix of 2165 OTUs was used for all subsequent statistical analyses. Representative sequences of bacterial OTUs were submitted to GenBank with the accession numbers MK323080 -MK325185.

Statistical analysis
We compared the richness and abundance of bacterial phyla among the sampling sites using ANOVA and Tukey's HSD test in R (R Development Core Team 2016). Furthermore, linear regression analyses in R were used to examine relationships between the abundance of bacterial phyla and sporocarp production, as well as the extraradical B. edulis mycelium in the soil variable (logarithm of B. edulis mycelium concentration) as quantified by real-time PCR (Mediavilla et al., 2017). To compare bacterial community composition across the studied stands, we ran non-metric multidimensional scaling (NMDS) analysis on the Hellinger-transformed OTU table. Two ordinations were carried out, one based on abundance and another based on presence/absence. Data were subjected to 999 iterations per run using the Sørensen similarity (Bray-Curtis index) and a random number to start. In addition, we tested whether bacterial communities were statistically different across forest types using the multiple-response permutation procedure (MRPP) and permutation-based nonparametric MANOVA (PERMANOVA) (Anderson, 2001). Finally, we tested whether individual OTUs showed a significant association with the sites with the highest B. edulis sporocarp production levels using indicator species analyses (Dufrêne and Legendre, 1997) in PC-ORD. For this purpose, the plots were classified into two categories according to their B. edulis sporocarp production values: plots producing 0-20 kg ha À1 were considered to be poorly productive and those producing > 20 kg ha À1 were considered to be highly productive (Hern andez-Rodr ıguez et al., 2015a).