Higher fungal diversity in dead wood is correlated with lower CO2 emissions in a natural forest

Wood decomposition releases almost as much CO2 to the atmosphere as does fossil-fuel combustion, so the factors regulating wood decomposition can affect global carbon cycling. We used metabarcoding to estimate the fungal species diversities of naturally colonized decomposing wood in subtropical China and, for the first time, compared them to concurrent measures of CO2 emissions. Wood hosting more diverse fungal communities emitted less CO2, with Shannon diversity explaining 26 to 44% of emissions variation. Community analysis supports a 'pure diversity' effect of fungi on decomposition rates and thus suggests that interference competition is an underlying mechanism. Our results are consistent with the theory of interference competition and with the results of published experiments using low-diversity, laboratory-inoculated wood, and we extend those results to a high-diversity, natural system for the first time. High levels of saprotrophic fungal biodiversity might be providing globally important ecosystem services by maintaining dead-wood habitats and by slowing atmospheric contribution of CO2 from the world's stock of decomposing wood.

Introduction much higher species diversity, more complex assembly histories, and selective faunal feeding on decomposers 14,15,16 . Thus, it is important to examine the relationships between fungal 55 diversity and decomposition rates in wood that is colonized and decomposing under natural conditions. Natural fungal communities can be characterized using metabarcoding 17  for CO 2 emission rates over three years 8 . We measured the extent to which natural fungal community variation can explain variation in decomposition rates.

70
Taxonomy results -Numbers of fungal OTUs ranged from 17 to 199 across wood pieces, tree species, and sampling dates, with means of 73. 8  41.1% of the 1,807 OTUs produced by uclust and 76.3% of the 1,565 OTUs produced by CROP were assigned to Fungi, and the proportions assigned to each fungal class were similar 75 across assignment methods (Table 1). Because we removed non-Fungi reads from the dataset before taxonomic assignment, we attribute the taxonomically unassigned OTUs to the still highly incomplete UNITE and Genbank databases used for taxonomic assignment. peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/051235 doi: bioRxiv preprint first posted online May. 1, 2016; In June 2012, CO 2 emissions from LC and SN also declined with fungal diversity, even though we used a fungal diversity estimate taken three months later (September 2012) and even after conservatively omitting an influential datum from SN (high CO 2 , low diversity) ( Fig. 1A, D). The third species (LX) did not return a significant regression, but its 85 CO 2 -diversity relationship was visually nearly indistinguishable from its congener LC, suggesting that wood species partly governs the emissions-diversity relationship. Variances explained (26% to 28%, Fig. 1A, D) were lower than in June 2013.
Finally, in September 2012, CO 2 emissions did not decline with higher fungal diversity ( Fig. 1B, E), which is consistent with the generally lower CO 2 emissions in September ( Fig. 1 90 top).
The above results were robust to two OTU-picking methods (CROP and uclust, Fig. 1), rarefaction (non-rarefied shown in Fig. 1; rarefied in Supporting Information S1), and two diversity estimates (Shannon in Fig. 1, Simpson in S1). Regressions using Simpson diversities were generally statistically more significant (S1). We also analyzed after omitting 95 single-read OTUs (which are more likely to be pipeline artefacts 23 ) and achieved the same results, except that the previously non-significant SN regressions in Sept 2012 (Fig. 1A, D) became statistically significant (authors' unpublished results). In short, the analyses presented in Fig. 1 are conservative.
Pure-diversity versus species-selection effects -Two general mechanisms could explain 100 the observed diversity-function relationships. The first is a 'pure diversity' effect where species identity does not matter, only that increased species richness and evenness per se is somehow responsible for slower wood decomposition. The second is a 'species-selection' effect where more diverse fungal communities might be more likely to contain particular species that cause slow decomposition and somehow also govern the overall decomposition 105 rate of the wood piece. To differentiate these two, we used a method devised by Sandau et al. 24 to generate a parameter λ for each regression in Fig. 1 (statistical details in Supporting Information S2). λ ranges between 0 and 1, with 0 indicating that variation in species composition does not account for variation in CO 2 emissions (i.e. the 'pure-diversity' effect).
For two tree species, LC and LX, λ always took values nearly equal to zero (Table 2). For the 110 third tree species SN, λ was also nearly zero in June 2012 but took intermediate values in peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/051235 doi: bioRxiv preprint first posted online May. 1, 2016; June 2013, suggesting that fungal composition in this tree species at this time had some explanatory power. The general failure to detect composition effects can be observed in the community ordinations (Fig. 2) by noting that the SN/June 2013 samples were also the only ones to line up along the CO 2 emissions gradient (except the lowest diversity samples). Not 115 surprisingly, conventional community-analysis tests returned the same conclusion: variation in community composition is not explained by CO 2 emissions (Supporting Information S3).

Discussion
We found that naturally colonised wood with more diverse fungal communities decomposes more slowly (Fig. 1), resulting in a negative relationship between fungal 120 biodiversity and the ecosystem function of decomposition. This result suggests positive relationships between fungal biodiversity and the ecosystem services of carbon storage and the provision of important decomposing-wood habitats in forests.
Our results are consistent with five published experiments using laboratory-inoculated wood, which have all found negative relationships between fungal diversity and 125 decomposition rates 10,11,12,13,25 30 found no relationship between wood loss and fungal OTU diversity after 12 years of decomposition but also reported that the least-decayed logs had the highest community diversities, again consistent with our results. All three studies found peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/051235 doi: bioRxiv preprint first posted online May. 1, 2016; community-composition differences for logs that differed in remaining undecomposed 140 weights. Importantly, none of those three field studies measured fungal diversities and CO 2 emission rates concurrently, as we did here (Fig. 1). We observe that relationships between CO 2 emission rate and fungal diversity varied from month to month and across tree species ( Fig. 1, Table 2, S1), suggesting that fungal activity and composition are dynamic and 145 environmentally responsive. Thus, that fungal community composition measured after years of decomposition might not reflect the communities that were active during decomposition, obscuring any relationship between mass loss and fungal diversity.
Our study helps to reconcile the differing results found in the published laboratory and field studies, by making concurrent measurements of emissions and fungal diversity in a field Mechanisms. -In contrast to wood, microbial diversity is reported to accelerate the decomposition of soil organic matter 31,32 , and this is thought to represent a general pattern 33,34 (but see Creed et al. 35 for leaf litter). We hypothesise that because soil organic matter presents 160 a much higher diversity of resources than does dead wood, niche complementarity amongst decomposer species drives positive relationships between diversity and soil organic matter decomposition.
A plausible biological mechanism explains why wood decomposition should slow with fungal diversity. Interference competition has long been predicted to evolve when niche 165 overlap is high and the disputed resource is valuable 36 . In forests, decomposing wood resources are available to many fungal species, and aggressive interactions are indeed observed among these fungi 37 . Elsewhere, it has been shown that interference competition reduces virulence (= host consumption rate) in endosymbioses 38.39,40 and productivity in peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/051235 doi: bioRxiv preprint first posted online May. 1, 2016; bacterial communities 41,42 . Consistent with those findings in other contexts, interference 170 competition can also explain why fungal biomass has been found to explain variance in wood mass loss 9 . When a piece of wood is colonized by many fungal species, the hypothesized higher levels of interference competition would result in less wood converted into fungal biomass (or CO 2 ). High niche overlap is also consistent with the observed 'pure diversity' effect of fungal diversity on emissions ( Possible errors and future experiments. -One potential source of error is that metabarcoding provides only approximate estimates of species frequencies 22, 23 , yet Shannon and Simpson indices, both of which discount rare species, were able to explain variation in CO 2 emissions here. There are two likely and non-exclusive explanations. (1) Low-read 185 ('rare') OTUs were more likely to be sequence artifacts from the metabarcoding pipeline and thus should be discounted, and (2) in a Norway spruce forest, Ovaskainen et al. 21 found that abundances of fungal fruiting bodies and OTU read numbers from metabarcoding were positively correlated, suggesting that low-read OTUs represent low-biomass species, which could have little influence on decomposition rates. 190 Another possible source of error is that we did not experimentally control for the age of the wood pieces, and thus an alternative explanation is that the observed correlation between fungal diversity and CO 2 emissions rate ( Fig. 1) might be caused by sampling along a successional gradient in which older wood pieces have less remaining wood to decompose (and thus lower emissions) and have also accumulated more fungal species. However, we 195 found no relationship between decay class and emissions rates (Methods: Experimental setup and Statistical analyses) in our dataset, nor did we in the 320-piece superset from which our samples were drawn 8 , whereas this alternative explanation predicts that the least-decayed wood pieces should show the highest emissions. Also, we found mostly 'pure-diversity' peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/051235 doi: bioRxiv preprint first posted online May. 1, 2016; effects of fungal communities on emissions (Table 2), whereas this alternative explanation 200 invokes a successional sere and thus predicts compositional effects. We suggest other long-term experiments in which sterilized and even-aged wood pieces are allowed to be colonized and sampled for CO 2 emissions rates and fungal diversity over many years in the field.  Experimental setup. -At our site, most woody debris comes from Lithocarpus chintungensis, (LC), Lithocarpus xylocarpus (LX), and Schima noronhae (SN), so we only examined those three species. In early 2010, branches from these three species, already decomposing on the forest floor, were identified to species by a botanist from the Xishuangbanna Tropical Botanical Garden, collected, and cut into a total of 320 wood pieces, 235 sized to fit a field-respiration chamber (ca. 10 cm diameter and 20 to 30 cm length), tagged, weighed, and measured for size and decay class (further details in Liu et al. 8 ). The three decay classes were DKC1 = a knife could not penetrate, DKC2 = knife could slightly penetrate with appreciable resistance, DKC3 = knife could deeply penetrate with little resistance 51 . We used similar-sized pieces to control for potential effects of wood size on 240 fungal communities 52 . The pieces were placed on the forest floor within a 60 x 3 m belt transect following an elevation contour. We collected source wood from within 500 m of this transect, utilizing about 5% of downed woody debris from these species in these decay classes, potentially arising from about 6000 source trees (D.A. Schaefer, unpublished data).
Each wood piece was initially weighed with a GLL portable electronic balance (accuracy 245 0.5 g) and for moisture content with an Extech MO210 moisture meter (calibrated as shown in Liu et al. 8 ). Their volumes were calculated as cylinders, based on length and the average of 5 circumferential measurements along their lengths. From those, initial weight, volume, and density were all calculated. Oven drying of these wood pieces was not done, because it would have altered microbial communities and wood chemistry. 250 CO 2 emissions rate measurements. -Individual wood-piece CO 2 release rates were measured in the field in a closed, ventilated chamber (10 L) connected to an infrared gas analyzer (Licor 820, Lincoln, NE, USA). After chamber closure and initial stabilization, linear CO 2 concentration increase rates were logged for at least 5 min. Pieces remained in the field for CO 2 measurements (within 5 m) and were handled carefully to limit fragmentation. 255 Temperature and moisture were measured for each sample at each sampling time. The wood-piece CO 2 release rate (R WD , μmol C g -1 h -1 ) was calculated as follows: peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/051235 doi: bioRxiv preprint first posted online May. 1, 2016; R WD = (1000*∆CO 2 *P*(V-V s )) / (24*R*(T s +273)*W C ) (1) Where ∆CO 2 represents the measured CO 2 concentration increase (ppm day -1 ), P is the internal pressure (kPa; measured by the Licor 820), V is the volume of the system (10.08 L, 260 including the chamber volume and tubing volume and Licor optical path), V s is the volume of the wood piece (L), R is the gas constant (8.314 L -1 kPa -1 K -1 mol -1 ), T s is the wood temperature (ºC), and WC is the carbon weight of each piece (g; 47% of its dry weight).
These measurements were made eight times from September 2010 to June 2013, approximately every four months (Fig. 1). 265 Sampling for genetic analysis. In the laboratory, these wood pieces were peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/051235 doi: bioRxiv preprint first posted online May. 1, 2016; re-measured for volume (as above), and twenty-two wood pieces exhibiting >15% volumetric weight loss since the start of the experiment, indicating substantial fragmentation in the field, were excluded from the analysis. The remaining wood pieces were dried at 70 ºC to constant 290 weights and then individually weighed to the nearest 0.1 g on an electronic balance. Denoising and chimera removal: Denoiser in QIIME 55 was used to remove characteristic 454 sequencing errors. Next, ITSx 1.0.3 56 was used to extract the variable ITS2 region from the whole reads (i.e. conserved 5.8S and LSU flanking sequences were stripped) and to remove 335 non-fungal-ITS reads. The extracted sequences were clustered at 99% similarity with USEARCH v7.0.1090 57 to remove replicate sequences and chimeras. OTU-picking and taxonomic assignment: We used two methods to cluster the reads into OTUs. First, we used a reference-based method in QIIME (pick_open_reference_otus.py: max_accepts 20 max_rejects 500 stepwords 20 length 12 -suppress_align_and_tree) in which reads were first 340 clustered by matching at 97% similarity to the UNITE 12_11 fungal database 58 , which itself had previously been clustered at 97% similarity for use within QIIME. Unassigned reads (the vast majority) were then clustered de novo using the uclust option at 97% similarity, producing 1,807 OTUs in total. For these latter OTUs, we attempted to assign taxonomies using QIIME's assign_taxonomy.py against the UNITE database. Second, we performed de 345 novo 97%-similarity clustering with CROP 1.33 59 , producing 1,565 OTUs. We assigned taxonomies against Genbank using the NNCauto and QCauto methods in Claident 60 . peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/051235 doi: bioRxiv preprint first posted online May. 1, 2016; Sequence data will be deposited at datadryad.org (doi: to be assigned) and in GENBANK's Short Read Archive (Accession number: PRJNA252416). An example bioinformatic script will also be deposited at datadryad.org (doi: to be assigned). because the residuals suggested accelerating CO 2 emissions at the lowest fungal diversity, making our results conservative. In trial models, we also tested for significant effects of wood surface temperature and decay class (see Experimental setup), but they did not interact significantly with the fungal diversity term and mostly did not enter significantly as additive terms, and so have been omitted here for simplicity. Liu et al. 8 also did not find a correlation 370 between decay class and CO 2 emissions rates for these wood pieces.
To test whether a 'pure diversity' effect is sufficient to explain the observed diversity-function relationships, we use a new method by Sandau et al. 24 where community similarities (1-Jaccard binary) between all wood pieces are used to create a variance-covariance matrix that is then included in the linear regressions, thus taking into 375 account potential non-independence of wood pieces due to the fact that some communities are similar to each other (Supporting Information S2). In Supporting Information S3, we also peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/051235 doi: bioRxiv preprint first posted online May. 1, 2016; use conventional community analyses to test for an effect of community composition on CO 2 emissions. We limit our tests to June 2012 and 2013, as only these exhibited significant declines in emissions with fungal-species diversity (Fig. 1A, C, D, F).