The role of transcriptomes linked with responses to light environment ...

Journal of Ecology 2017, 105, 592–601

doi: 10.1111/1365-2745.12760

TRANSCRIPTOMIC AND GENOMIC ANALYSES OF COMMUNITIES

The role of transcriptomes linked with responses to light environment on seedling mortality in a subtropical forest, China ~ a3, Xiangcheng Mi1, Xiaojuan Liu1, Lei Chen1, Baocai Han1,2, Mar
ıa Natalia Uman Yunquan Wang1, Yu Liang1, Wei Wei1 and Keping Ma*,1,2 1

State Key Laboratory of Vegetation and Environmental Change, Institute of Botany, The Chinese Academy of Sciences, 20 Nanxincun, Xiangshan, Beijing 100093, China; 2University of Chinese Academy of Sciences, Beijing 100049, China; and 3Department of Biology, University of Maryland, College Park, MD 20742, USA

Summary 1. Differences in seedling survival in trees have a lasting imprint on seedling, juvenile and adult community structure. Identifying the drivers of these differences, therefore, is a critical research objective that ultimately requires knowledge regarding how organismal function interacts with the local environment to influence survival rates. 2. In tree communities, differences in light use strategies are frequently invoked to explain differences in seedling demographic performance through growth and survival trade-offs. For example, shade-tolerant species grow slowly and have higher survival rates, whereas shade-intolerant species grow quickly but have lower survival rates. Thus, functional traits related to photosynthesis should be strong predictors of demographic rates, but results in the literature are mixed indicating that additional or alternative information regarding organismal function should be considered. 3. Here, we provide a community-wide inventory of transcriptomes in a subtropical tree community. This information is utilized to determine the degree to which species share homologous genes related to gene ontologies for light use and harvesting. These species similarities are used in neighbourhood generalized linear mixed-effects models of seedling survival that evaluated seedling survival as a function of the transcriptomic, functional trait and phylogenetic composition of the local neighbourhood. The results show neighbourhood similarity in three of the 15 gene ontologies evaluated are significantly related to survival rates based on neighbourhood composition. For two of these ontologies, survival rates increase when neighbours are similar in their gene tree composition indicating the importance of abiotic filtering and performance hierarchies. 4. Synthesis. The present work takes a novel approach by sequencing the transcriptomes of naturally co-occurring tree species in a subtropical forest in China. The results show that the transcriptomic similarity of species is a significant predictor of differential survival. The study demonstrates that exploring the functional genomic similarity of non-model species in nature has the potential to increase the breadth and depth of our understanding of how gene function influences species cooccurrence and population dynamics in communities. Key-words: determinants of plant community diversity and structure, light availability, RNA-seq, seedlings dynamics, species coexistence, subtropical forest, transcriptomic similarity

Introduction Determining the mechanisms underlying species diversity and co-occurrence is considered one of the most challenging goals for community ecology (Sutherland et al. 2013). If all plant *Correspondence author. E-mail: [email protected]

species share similar major requirements, light, water and macronutrients, how we can explain species co-occurrence and diversity in plant communities particularly in subtropical and tropical regions where diversity reaches extraordinary levels? In other words, diverse plant communities seriously challenge classical and modern coexistence theory (e.g. MacArthur & Levins 1967; Chesson 2000). This has led to the development

© 2017 The Authors. Journal of Ecology © 2017 British Ecological Society

Transcriptomes and seedling mortality 593 of theoretical approaches that assume demographic equivalence and dispersal limitation to predict emergent community properties such as the species abundance distribution (Hubbell 2001). In response to Hubbell’s neutral theory, ecology redoubled its efforts to demonstrate non-random patterns of species distributions and individual demographic rates. Functional traits are usually utilized to determine whether species have different resource capture strategies (Keddy 1992; Cornelissen et al. 2003). There is now robust evidence that prove subtropical and tropical tree community structure and dynamics are decidedly non-neutral (e.g. Swenson & Enquist 2007, 2009; Swenson et al. 2012; Liu et al. 2013; Iida et al. 2014). However, these studies have focused on a relatively small number of functional traits that are indirectly related to resource acquisition and the statistical relationships reported are often weak. Thus, there is still a great deal unknown regarding how organismal functional diversity influences species diversity and co-occurrence in these forests. Approaches that supplement and expand upon the traditional functional trait-based approach may be required. Among all resources required by plants, light is perhaps one of the most limiting in subtropical and tropical forests (Augspurger 1984; Nicotra, Chazdon & Iriarte 1999). Light variation has the potential to be a major axis of niche differentiation for tree species and was considered to be important in defining guilds in early neutral theory where demographic neutrality occurred within but not among guilds (Hubbell & Foster 1986). Co-occurring tree species in forests often exhibit a broad range of light preferences that vary from shade tolerant to light demanding. Shade-tolerant species are often described as slow growing and conservative species. Conversely, light-demanding species tend to have rapid resource acquisition rates and fast growth. This variation among species along the light preference axis is thought to play a critical role in determining species co-occurrence in tree communities (Kobe 1999; Poorter 1999). For example, under shaded conditions, species are expected to exhibit similar ecological strategies along a light preference axis that allows them to tolerate limiting light. They therefore have similar growth rates and would be expected co-occur in similar light environments. On the other hand, co-occurring species might also show high dissimilarity in light capture strategies in the dissimilar light environment which would allow them to partition niche space and therefore co-occur. As noted above, the role of ecological similarity among species in community assembly has been explored by integrating information on functional traits related to resource acquisition (e.g. Weiher, Clarke & Keddy 1998; Stubbs, Wilson & Bastow Wilson 2004; Swenson & Enquist 2007, 2009). This functional approach provides a connection between species and their local environment (McGill et al. 2006). In the case of light environments, previous studies have utilized a series of easily measured functional traits to indicate where species fall along the global leaf economics spectrum (e.g. Reich, Walters & Ellsworth 1997; Wright et al. 2004). For example, photosynthesis increases with higher specific leaf area (SLA) and %N and %P content in the leaves, but this comes at the cost of lower leaf life spans.

However, many functional leaf traits likely do not capture the complexity of species responses to light environments and richer assays of the functions underlying light capture might prove useful for understanding the drivers of species co-occurrence in shaded forest understories. One promising way to overcome the limitations of the traditional functional trait-based approach which relies on a few easily measured traits that loosely correlate with the physiology of interest is to incorporate information on functional genetic variation through the analysis of transcriptomes (Swenson 2012). Transcriptomes can now be generated for non-model species through the sequencing of mRNA in a tissue sample and by de novo assembly of the resulting reads. These assembled transcripts can be compared to the annotated genomes of model species to infer their putative functions. Functional phylogenomics approaches can then be utilized to cluster and align putative homologous genes across species to generate gene trees (Yang & Smith 2014; Smith et al. 2015; Yang et al. 2015). It is therefore now possible to quantify the number of putatively homologous genes shared among species in a community and to categorize them by their putative functions. In sum, it is possible, for example, to identify how many homologous genes are found between pairs of species for light harvesting responses, photosystem I and II assemblies, photoinhibition, and response to red light among many others. This approach has the potential to greatly enrich our understanding of the similarity of species in relation to light and photosynthesis. Therefore, the complex responses of species to light gradients that may drive their differential demography and co-occurrence could be further disentangled when information on functional genomic similarity is considered along with functional trait data. Here, we take a functional phylogenomic approach to tree community ecology. We test alternative hypotheses regarding the functional similarity of species and their local light environments and implications these have for their co-occurrence and performance. We generated transcriptomes for 101 nonmodel tree species that naturally co-occur in a long-term forest dynamics plot in subtropical China. The transcriptomes were used in a functional phylogenomics framework. Homologous gene trees related to light and photosynthesis were analysed to evaluate the demographic rates of naturally occurring seedling individuals in the plot. Specifically, we evaluated different models of survival rates for focal seedlings as a function of the neighbourhood species composition. Heterospecific neighbours similarity to focal species survival was weighted by the number of homologous genes they share, their functional trait similarity and their phylogenetic relatedness. The analyses were partitioned by individual gene ontology (GO) terms to determine which ontologies have the greatest predictive power. The specific questions we ask are: (i) What are the most important light-response GO terms influencing seedling survival in a subtropical forest in China? (ii) What is the effect of trascriptome similarity on seedling survival on shade-tolerant species and light-demanding species? (iii) What is the role of phylogenetic-, functional- and transcriptome-based similarity on predicting survival rates of seedlings?

© 2017 The Authors. Journal of Ecology © 2017 British Ecological Society, Journal of Ecology, 105, 592–601

594 B. Han et al.

Materials and methods STUDY SITE

The study site is the 24 ha (400 m 9 600 m) Gutianshan subtropical forest dynamics plot (GTS FDP) (29°150 6″–29°150 21″N, 118°070 1″– 118°070 24″E) located in the Gutianshan Forest Reserve, Kaihua County, Zhejiang Province, China. In the plot, all trees with a diameter at breast height ≥1 cm (diameter at 13 m, DBH) were tagged, measured, mapped and identified to species (Legendre et al. 2009). The plot is censused every 5 years and the data from the 2010 census were used in this study. The topography of the plot is rugged with attitude varying from 4463 to 7149 m above sea level. The region experiences wet seasons from March to June and in September and two dry seasons from July to August and October to February (Lai et al. 2009).

SEEDLING CENSUS

This study used 285 5 9 5 m seedling plots distributed in a regular pattern in each 20 9 20 m subplot in the 24 ha GTS plot in 2012. Every seedling plot was set in the northeast corner of a 20 9 20 m subplot in the GTS FDP and 2 m away from the subplot edges. The total area of the 5 9 5 m seedling plots accounts for 296% of the GTS 24 ha FDP. In each of the 285 seedling plots, all individuals with a DBH 1000 gene trees in rows and 101 species in the columns. Next, we selected from this matrix only those gene trees with GO annotations related to photosynthesis. This resulted in 135 rows (gene trees belonging to 47 light-related GO terms (Table S1, Supporting Information). Next, for each of the 47 GO terms, we selected the corresponding gene trees from the matrix and calculated the Euclidean distance between species. This distance indicates the frequency at which two species are not found in the homologous gene trees for the GO. In other words, lower values indicate higher similarity in the gene composition for a given GO term. For the presence–absence matrix, the distance simply indicates the number of unshared gene between species per GO. Given that large gene families are often subject of gene duplication and loss that could bias the results obtained in this study, we subsampled our gene data to use 15 out of 47 light-related GO terms where we were more confident that homologous were being identified.

© 2017 The Authors. Journal of Ecology © 2017 British Ecological Society, Journal of Ecology, 105, 592–601

Transcriptomes and seedling mortality 595 LIGHT DATA

Because the present work was interested in functions related to light, we quantified light availability at our study site. To do so, every 20 9 20 m subplot in the GTS FDP was divided into 4 10 9 10 m subplots. Hemispherical photographs were taken at the centre of each 10 9 10 m subplot within the seedling plot at 13 m above-ground with a Canon 7D camera (Kyushu, Japan) from May to July in 2013. The camera was equipped with Sigma 8 mm Fisheye Converter lens (Fukushima-ken, Japan) and arranged horizontally. Photographs were taken during either early dawn or late dusk to ensure the uniform overcast weather. Three replicate photos were taken and the photos showing the highest contrast between sky and foliage for each 10 9 10 m subplot were selected. Finally, the Gap Light Analyser software (version 2.0) (Frazer, Canham & Lertzman 1999) was used to convert photographs to a single canopy openness value, direct solar radiation and diffuse solar radiation (Beaudet & Messier 2002).

NEIGHBOURHOOD MODEL ANALYSIS

A total of 85 species in our seedling plots had enough individuals to reliably model their survival rates. Below, we describe the models for 85 species (Table S2) where we had information regarding the phylogenetic, functional trait and transcriptomic dissimilarity between species. We performed the analyses on three groups of species: all species (85), shade-tolerant species (51 species) and light-demanding species (34 species). Shade-tolerant vs. light-demanding species were classified according to Song (2013) (Table S2). In total, we had 7413 total seedling individuals, 5648 shade tolerance species individuals and 1765 light-demanding seedling individuals. In order to assess whether the transcriptome, functional traits, phylogenetic and neighbour composition influences the seedling survival, we evaluated a series of 24 neighbourhood models of individual survival. For the first and most basic model, we examined the importance of neighbourhood conspecific and heterospecific neighbour densities on seedling survival. In this model, focal seedling survival was modelled as a function of conspecific seedling neighbour density (SCON), heterospecific seedling neighbour density (SHET) conspecific, conspecific adult neighbour densities (ACON) and heterospecific adult neighbour densities (AHET). The SCON and SHET variables simply counted the number of neighbouring individuals in the seedling plot (Bai et al. 2012). However, the ACON and AHET variables used adult individuals varying in size. Thus, we calculated these variables using the basal area (BA) of conspecific (ACON) and heterospecific (AHET) adults within 20 m from a focal seedling and dividing that value by the distance between the adult and the focal seedling (Canham, LePage & Coates 2004). Eqn (1) shows the formula used for this calculation: A¼

N X i

BAi Distancei

eqn 1

where i is an individual neighbour adult and N is the number of adult neighbours. To eliminate boundary effects, only the seedlings with a distance ≥20 m to the 24 ha plot edges were included. We refer to this model as ‘Biotic’ in the results. A second model considered the survival of a focal seedling given the canopy openness measured using the canopy photos. We refer to this model as ‘Canopy’ in the results. The Biotic and Canopy variables were used together in additional models. First, we combined only the Biotic and Canopy variables (i.e. 1 model). Then, we used

the Biotic and Canopy variables with the trait dissimilarity in the neighbourhood for one trait at a time (i.e. 5 models) or the phylogenetic dissimilarity in the neighbourhood (i.e. 1 model). The neighbourhood trait and phylogenetic dissimilarities were calculated using a null model approach to control for neighbourhood heterospecific densities and richness as well as neighbourhood conspecific density. Specifically, we calculated an observed mean pairwise distance between all individuals in the seedling neighbourhood using trait or phylogenetic distance matrices. Next, we generated a null distribution of mean pairwise distances by randomizing the names of species on the distance matrices 999 times and recalculating the neighbourhood mean pairwise distances each time (Swenson 2014). Finally, we calculated an effect size as the observed value minus the mean of the null distribution and a standardized effect size (SES) was calculated by dividing the effect size by the standard deviation of the null distribution and multiplying by negative one. Thus, positive SES values in this study indicate a neighbourhood of functionally or phylogenetically similar individuals and negative SES values indicate a dissimilar neighbourhood. Finally, we generated one neighbourhood survival model per GO term where we used the Biotic and Canopy variables with a GO SES value (i.e. 15 models). The GO SES value was calculated in the same manner as the trait and phylogenetic SES values with the exception that the Euclidean distance matrices between gene trees was utilized. All the models were assessed using generalized linear mixedeffects models (GLMMs) with binomial errors (Bolker et al. 2009). A logit transformation of seedling state: 1 (alive) or 0 (dead) was the response variable. Since the initial height of seedlings can affect the seedling survival significantly (Comita & Hubbell 2009), the logtransformed values of initial seedling height (logH) were included in all the models. Also, all variables were standardized by subtracting the mean value of the variable from the observed value and dividing by the standard deviation. Species and plots were included as intercept specific random effects. We used the Akaike information criterion (AIC) to compare the three models. The ΔAIC was calculated by subtracting the minimum value of AIC from each of the models’ AIC values. All the analyses were conducted in the R software, the GLMMs were conducted with LME4 package (Bates et al. 2016). Because multiple models were considered, we adjusted our P-values using a Holm correction.

Results NEIGHBOURHOOD MODEL RESULTS – ALL SPECIES

Models were compared using AIC and ΔAIC between values ≤2 indicated that models were indistinguishable. When considering all species in the same model, we found two model constructs that performed similarly well (Fig. 1; Tables 1 and 2). In both models, high conspecific seedling densities in the neighbourhood (SCON) negatively impacted focal individual survival. Further, both models found adult densities (ACON and AHET) were positively related to focal individual survival. Canopy openness did not significantly influence survival in either model. The difference between the models was that one included the LWR SES value and the other included the GO:0010201 SES value. Both of the SES variables were positive effects indicating that similarity in LWR or gene trees related to GO:0010201 (‘response to continuous far red light stimulus’) enhanced survival.

© 2017 The Authors. Journal of Ecology © 2017 British Ecological Society, Journal of Ecology, 105, 592–601

596 B. Han et al.

Fig. 1. Estimates of fixed effects (1 SE) for the best models for all seedling species. Filled dots represent significant values (P < 005) after a Holm correction for multiple tests. Open dots represent non-significant values. SCON is the conspecific seedling density, SHET is the heterospecific seedling density, ACON is the conspecific adult density, AHET is the heterospecific adult density, Canopy is canopy openness, LWR. sNRI is the standardized effect size of the leaf length-width ratio and GO.0010201.sNRI is the standardized effect size of GO:0010201. LWR, length to width ratio.

Table 1. Akaike information criterion (AIC) values of the generalized linear mixed models of seedling survival for all species combined, shade tolerance species and light-demanding species. Bold values indicate the models selected. These models used Euclidean distance matrices where we counted the number of tips per species in a gene tree

thereby attempting to avoid issues with paralogous, two additional models were equally supported. These included GO:0010196 (‘non-photochemical quenching’) or GO:000 9638 (‘phototropism’) with negative and positive effects respectively (Tables 3 and 4). The model with phylogenetic information was not selected in either case.

AIC

Candidate model

All species

Shade tolerance species

Null Biotic Canopy Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic +

619737 617822 619649 617822 617544 617868 618504 618510 618510 618161 617381 617620 617624 617789 617830 617917 618020 618067 618144 618153 618162 618165 618209 618214 618216

463833 462127 463935 462127 462297 462371 462735 462730 462725 462334 461670 461909 461912 461691 462202 462172 462294 462360 462332 462139 462353 462356 462454 462452 462406

Canopy Canopy Canopy Canopy Canopy Canopy Canopy Canopy Canopy Canopy Canopy Canopy Canopy Canopy Canopy Canopy Canopy Canopy Canopy Canopy Canopy Canopy

+ + + + + + + + + + + + + + + + + + + + +

lwr la sla fn fp phy GO.0010201 GO.0010362 GO.0009638 GO.0010196 GO.0009768 GO.0009785 GO.0043153 GO.0009643 GO.0010205 GO.0009585 GO.0010206 GO.0080005 GO.0009854 GO.0010117 GO.0009773

Lightdemanding species 156815 157022 156855 157022 157104 157300 157596 157609 157659 157355 157299 157330 157331 157299 157317 157331 157301 157284 157323 157216 157317 157344 157369 157362 157270

The above results used the number of times a species was represented in a particular gene tree. When we considered simply the presence or absence of species from gene trees,

NEIGHBOURHOOD MODEL RESULTS – SHADETOLERANT SPECIES

Models considering only the 51 shade-tolerant species were considered next. Two models were indistinguishable with both including a GO SES variable (Fig. 2; Tables 1 and 2). As in the models that considered all species, conspecific seedling density negatively influenced focal individual survival, while adult con- and heterospecific densities positively influenced survival. Canopy openness had a positive, but not significant, influence on survival rates. In one model, the SES for GO:0010196 (‘non-photochemical quenching’) negatively impacted survival, while in the other model the SES for GO:0010201 (‘response to continuous far red light stimulus’) positively impacted survival (Fig. 2; Tables 1 and 2). When considering only the presence or absence of species in gene trees instead of their number of tips, the same models were supported (Tables 3 and 4). In all cases, the phylogenetic and trait information was not important.

NEIGHBOURHOOD MODEL RESULTS – LIGHTDEMANDING SPECIES

Models that considered only the 34 light-demanding species in our study were considered next. A total of 24 models were considered and the model including only seedling height and canopy openness (i.e. the ‘Canopy’ model) was the best supported (Tables 1 and 3) with trait, phylogenetic and transcriptomic SES values not being important (Tables 1–4).

© 2017 The Authors. Journal of Ecology © 2017 British Ecological Society, Journal of Ecology, 105, 592–601

Transcriptomes and seedling mortality 597 Table 2. The standardized effect size type and their effects on seedling survival in mixed models. The Estimate, P-value and adjusted P-value calculated using a Holm correction (P.adjust) by false discovery rate for total species, shade tolerance species and light-demanding species. These models used Euclidean distance matrices where we counted the number of tips per species in a gene tree. Bold P.adjust values are corresponding to the best models showed in Table 1, respectively Total species

Shade tolerance species

Light-demanding species

SES type

Estimate

P-value

P.adjust

Estimate

P-value

P.adjust

Estimate

P-value

P.adjust

LN LP LA LWR SLA Phylogeny GO:0009585 GO:0009638 GO:0009643 GO:0009768 GO:0009773 GO:0009785 GO:0009854 GO:0010117 GO:0010196 GO:0010201 GO:0010205 GO:0010206 GO:0010362 GO:0043153 GO:0080005

00058 00056 01092 01361 00131 00295 0036 00956 00529 00687 00086 00631 0013 00097 00981 013 00309 00268 00959 00678 00304

08942 08974 00096 00028 0775 04395 04108 00136 02114 00467 08671 00789 07479 08147 00373 00035 03747 04433 00133 01498 04528

08974 08974 00571 00368 08974 06339 06339 00571 04439 01401 08974 02071 08974 08974 01306 00368 06339 06339 00571 03495 06339

00174 00205 01005 01187 00124 00483 00917 0105 00498 0064 0041 00689 00081 00111 01517 01455 00455 00418 01053 00716 00476

07512 06933 00499 00334 08138 02596 00713 00181 03203 01067 04755 0088 08604 08172 00057 00046 02582 0301 00176 01908 03086

08581 08564 01747 01403 08581 04484 02139 00950 04484 02490 06241 02310 08604 08581 00599 00599 04484 04484 00950 04007 04484

00575 00036 01526 01932 00754 00371 01127 00536 00761 00547 01073 0053 00146 00257 00833 008 00508 00542 00545 00801 00441

04752 09644 00485 00166 04244 06758 02051 0517 0344 0454 02964 052 08726 07559 03872 03883 04799 04537 051 03926 05937

06825 09644 05093 03486 06825 07884 06825 06825 06825 06825 06825 06825 09162 08355 06825 06825 06825 06825 06825 06825 07334

LN, leaf nitrogen content; LP, leaf phosphorus content; LA, leaf area; LWR, leaf length to width ratio; SLA, specific leaf area.

Discussion A key goal in community ecology is to determine the functional basis for species performance, abundance and co-occurrence. In other words, community ecologists are still trying to establish the function–performance relationships that should lead to more robust predictions (McGill et al. 2006; Swenson 2013). A possible reason for previously reported weak or non-existent relationships between individual performance and traits (e.g. Iida et al. 2014; Paine et al. 2015) is that trait information on the individual level is needed for strong inference (e.g. Liu et al. 2016). An additional reason may be that the traits frequently measured in trait-based community ecology do not provide enough detail regarding the functional differences between species (Swenson 2012). Here, we explored an alternative approach to trait and phylogenetically based community ecology that integrates transcriptomic information on genes involved in photosynthesis. We evaluated how the survival of focal seedlings is related to the dissimilarity between species in their homologous gene content for 15 GO terms involved in species responses to light. We compared these models to models that included traditional trait and phylogenetic information. We were able to identify GO terms where having similar neighbours was positively related to survival as would be expected under an abiotic filtering or performance differences model of species co-occurrence. We describe our results in detail below.

CONSPECIFIC AND HEREROSPECIFIC DENSITY DEPENDENCE

Conspecific negative density dependence is considered one of the most important forces operating in natural seedling communities (Janzen 1970; Comita & Hubbell 2009; Lebrija-Trejos et al. 2014; Umana et al. 2016). Specifically, strong conspecific negative density dependence will regulate populations and prevent a single species from becoming dominant and reducing local species richness. Our results support the finding that the density of conspecific seedlings consistently negatively impacted the survival of focal individuals (Figs 1 and 2; Tables 1 and 2). Conversely, we found no evidence of heterospecific negative density dependence in the seedling layer. Thus, negative conspecific effects greatly outweighed heterospecific seedling effects, which should indicate large niche differences (Chesson 2000) and promote species coexistence. We did find a strong positive effect of con- and heterospecific densities when considering adult neighbours. This may suggest that the negative density dependence detected in the seedling layer results from seedling competition or Janzen–Connell effects (Janzen 1970; Connell 1971) unrelated to adults. We also note that the data used in this study follows a large ice/snow storm that occurred in 2008 (Man, Mi & Ma 2011) where many adult trees died. It is possible that our results reflect a large recruitment of seedlings under adults after the canopy was opened by adult mortality occurring due to the storm. However, we caution that more

© 2017 The Authors. Journal of Ecology © 2017 British Ecological Society, Journal of Ecology, 105, 592–601

598 B. Han et al. Table 3. Akaike information criterion (AIC) values of the generalized linear mixed models of seedling survival for all species combined, shade tolerance species and light-demanding species. Bold values indicate the models selected. These models used Euclidean distance matrices where we counted the presence or absences of species in a gene tree AIC

Candidate model

All species

Shade tolerance species

Null Biotic Canopy Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic + Biotic +

619737 617822 619649 617822 617544 617868 618504 61851 61851 618161 617482 617642 617661 617669 617799 617856 618001 618022 6181 618148 618153 618157 618173 618175 618216

46383 46212 46393 46212 46229 46237 46273 46273 46272 46233 46153 46148 46191 46192 46223 46125 46213 46229 46229 46244 46239 46234 46236 46244 4624

Canopy Canopy Canopy Canopy Canopy Canopy Canopy Canopy Canopy Canopy Canopy Canopy Canopy Canopy Canopy Canopy Canopy Canopy Canopy Canopy Canopy Canopy

+ + + + + + + + + + + + + + + + + + + + +

lwr la sla fn fp phy GO:0010201 GO:0010196 GO:0009638 GO:0010362 GO:0009768 GO:0009585 GO:0009785 GO:0043153 GO:0009854 GO:0009643 GO:0080005 GO:0010205 GO:0010206 GO:0010117 GO:0009773

Lightdemanding species 156815 157022 156855 157022 157104 1573 157596 157609 157659 157355 1573709 157348 1573225 1573225 1572646 1569794 1573712 1573027 157371 1572494 1573704 1573222 1573128 1573618 1572757

detailed experimental analyses would be required to fully consider these alternative hypotheses. TRANSCRIPTOMIC, FUNCTIONAL TRAIT AND PHYLOGENETIC COMPOSITION OF NEIGHBOURHOODS

The GO terms that had consistently significant effects on seedling survival were related to photosynthetic responses to different light wavelengths. For example, responses to red-far red light (GO:0010201) was positively correlated with the seedling survival where species with different sets of genes involved in the responses to similar types of light tend to exhibit higher survival (Figs 1 and 2; Tables 1–4). This result suggests that the spectral composition of the light, also known as light quality, plays a critical role in determining photosynthetic responses in tree seedlings and ultimately species co-occurrence by selecting for species with similar photosynthetic responses and performance and excluding dissimilar species. Light quality has been frequently ignored in ecological studies with much more attention being paid to light quantity (e.g. Kobe 1999; Poorter 1999). However, the canopy interference with incoming light affects not only light quantity, but also quality. As a result, light in the understory is mainly

composed of wavelengths falling within the red and green spectrum. To our knowledge, only one study has evaluated the effect of light quality on tropical or subtropical seedling species (Lee et al. 1996). Their findings are in agreement with our results suggesting that beyond light intensity, light quality also affects seedling development (Lee et al. 1996). Combined, these results may indicate why canopy openness is not always an important predictor of seedling demographic performance, as in this study, and measures of light quality should be considered more frequently. We also found that neighbourhood similarity in GO:0009638 (phototropism) is positively related to survival (Tables 1–4). In other words, species with similar homolog composition for phototropism tend to survive at higher rates in the same location. As with the red-far red results, from this we infer that abiotic filtering and or the exclusion of dissimilar species due to performance differences (Chesson 2000) drive the phototropism result. Lastly, we found that similarity in the GO term related to non-photochemical quenching (GO:0010196) was negatively related to survival (Fig. 1; Tables 1 and 2). Photoprotective energy quenching can limit the production of active oxygen species and non-photochemical quenching is vital in this process (Khan, Rai & Tripathi 1986; Horton, Ruban & Walters 1994; Holt, Fleming & Niyogi 2004). It is possible that this result is indicative of niche differentiation in non-photochemical quenching and there is some evidence for niche differentiation related to response to changes light in conditions in the seedling layer in other systems (Scholes, Press & Zipperlen 1996). However, additional analyses and experimentation would be needed to address this hypothesis. Across the different traits analysed in this study, the LWR was the only trait that showed a significant relationship with seedling survival. Co-occurring species with similar LWR values exhibited higher survival (Figs 1 and 2; Tables 1–4). Previous work has shown that LWR is related to the light quality (far red or red:far red ratios) (Kasperbauer 1987). Thus, it may not be surprising that the LWR results were in the same direction as the GO:0010201 (response to red-far red) results. The lack of significant results for the other traits (e.g. leaf %N, leaf %P and SLA) suggests that many of the plant traits frequently measured by community ecologists fail to predict differential demography (Paine et al. 2015) and a broader assay of plant function is likely needed (Swenson 2012, 2013). The phylogenetic relatedness of neighbouring seedling species had no effect on focal seedling survival in any of our analyses. There are two possible reasons for this outcome. One is that there are no traits with phylogenetic signal that influence the population dynamics of species in this forest (Swenson 2013). Previous work has shown that phylogenetic information does not predict spatial patterns of trait diversity in this forest (Liu et al. 2013). A second possibility is that there are important functions that are correlated with phylogenetic relatedness, but that other functions more strongly predict survival than phylogenetic relatedness alone. We propose that it is likely the former than the latter possibility.

© 2017 The Authors. Journal of Ecology © 2017 British Ecological Society, Journal of Ecology, 105, 592–601

Transcriptomes and seedling mortality 599 Table 4. The standardized effect size type and their effects on seedling survival in mixed models. The Estimate, P-value and adjusted P-value calculated using a Holm correction (P.adjust) by false discovery rate for total species, shade tolerance species and light-demanding species. These models used Euclidean distance matrices where we counted the presence or absence of species in a gene tree. Bold P.adjust values are corresponding to the best models showed in Table 3, respectively Total species

Shade tolerance species

Light-demanding species

SES type

Estimate

P-value

P.adjust

Estimate

P-value

P.adjust

Estimate

P-value

P.adjust

LN LP LA LWR SLA Phylogeny GO:0009585 GO:0009638 GO:0009643 GO:0009768 GO:0009773 GO:0009785 GO:0009854 GO:0010117 GO:0010196 GO:0010201 GO:0010205 GO:0010206 GO:0010362 GO:0043153 GO:0080005

00058 00056 01092 01361 00131 00295 0086 00877 00372 00735 00093 00536 00455 00305 01162 01288 00277 00241 00871 00676 00336

08942 08974 00096 00028 07750 04395 00548 00169 03928 00379 08557 01355 02698 04999 00157 00059 04260 04905 00176 01501 04093

08974 08974 00616 00588 08974 06153 01439 00616 06153 01137 08974 03152 05151 06175 00616 00616 06153 06175 00616 03152 06153

00174 00205 01005 01187 00124 00483 01822 01007 00187 00612 00414 00747 00600 00152 01775 01670 00427 00393 00998 00710 00391

07512 06933 00499 00334 08138 02596 500E-04 00184 07163 01355 04711 00680 01961 07701 00020 00020 02904 03317 00194 01934 04080

08086 08086 01497 01169 08138 04543 00105 00815 08086 03162 06183 01785 03744 08086 00140 00140 04691 04976 00815 03744 05712

00575 00036 01526 01932 00754 00371 01798 00534 00929 00767 01025 00042 00062 00289 00459 00076 00510 00561 00534 00793 00087

04752 09644 00485 00166 04244 06758 00446 04771 02630 02945 03097 09579 09471 07524 06221 09366 04757 04364 04771 03969 09145

07707 09644 03395 03395 07707 09461 03395 07707 07707 07707 07707 09644 09644 09644 09332 09644 07707 07707 07707 07707 09644

LN, leaf nitrogen content; LP, leaf phosphorus content; LA, leaf area; LWR, leaf length to width ratio; SLA, specific leaf area.

Fig. 2. Estimates of fixed effects (1 SE) for the best models for the 51 shade-tolerant seedling species. Filled dots represent significant values (P < 005) after a Holm correction for multiple tests. Open dots represent non-significant values. Grey dots represent the marginally significant effects (005 < P.adjust < 01). SCON is the conspecific seedling density, SHET is the heterospecific seedling density, ACON is the conspecific adult density, AHET is the heterospecific adult density, Canopy is canopy openness, GO.0010201.sNRI is the standardized effect size of GO:0010201 and GO.0010196.sNRI is the standardized effect size of GO:0010196. SHADE-TOLERANT AND LIGHT-DEMANDING SPECIES

The results from our models including all species were similar to those including only the shade-tolerant species (Figs 1 and 2; Tables 1–4). This may not be surprising given that 51 of the 85 species analysed were shade tolerant. The results for the models that included only light-demanding species, however, deviated from the shade-tolerant results. In particular, trait and transcriptomic information did not predict seedling

survival in light-demanding species. Rather, the best models of light-demanding seedling survival included only initial seedling height and canopy openness information (Tables 1– 4). While increased light is expected to positively affect seedling growth, the survival rates of shade-tolerant species likely will not be as influenced by light. Light-demanding species, however, do not tolerate low light levels and it is expected that their survival should be related to light quantity. The

© 2017 The Authors. Journal of Ecology © 2017 British Ecological Society, Journal of Ecology, 105, 592–601

600 B. Han et al. results therefore suggest that if transcriptomic and trait similarity does effect seedling survival, those effects are far outweighed by light quantity. CAVEATS

The present work offers an alternative approach for examining the role of neighbourhood similarity on the survival of seedlings using transcriptomic information on homologous genes across co-occurring species. While the results are potentially interesting and they indicate that transcriptomic data in some cases provides detail that cannot be obtained from functional trait or phylogenetic data, there are clear limitations to our approach that should be noted. First, the analyses utilized GO annotations that were originally generated for model species and used to annotate the transcriptomes of non-model species in this study. Thus, we acknowledge that some annotations may be flawed as the genes in the nonmodel species may not have the same function as those in the model species (Pavlidis et al. 2012). Furthermore, even in the best-case scenario, where there are no annotation flaws, it is possible to over-extend inferences from GO results (Pavlidis et al. 2012). In this work, we discussed where the GO output aligns with previous physiological literature for tree species, but we have tried to restrict our discussion for this reason. Lastly, our approach quantifies the presence and absence of species in putatively homologous gene trees and calculates a dissimilarity value between all species. This approach could be biased due to gene duplications and the presence of paralogous. Future research should aim to refine this approach by considering gene tree topology and branch lengths in more detail (Swenson 2012; Smith et al. 2015).

Conclusions Community structure is the outcome of different demographic processes operating across individuals and species. Determining how organismal function predicts demography is therefore a major research objective. In recent years, functional traits have been popularized as a means to accomplish this linkage, but results have been mixed. Here, we have presented an alternative transcriptome-based approach that yields deeper and broader information regarding organismal function than functional traits and phylogeny. We applied this approach to a tree seedling community in subtropical China in order to determine whether the transcriptomic similarity of neighbouring individuals influences focal individual survival rates. We found that only a few gene ontologies related to differential survival rates in seedlings. This indicates that transcriptomic information may be useful in community ecology, but there is still a great deal of unexplained variation. Future studies should consider other aspects of the (or the entire) transcriptome and should potentially also consider differential gene expression for photosynthesis-related genes. Such work was not possible presently due to computational and financial barriers, but we expect these barriers will continue to be reduced allowing ecologists to leverage the full power of transcriptomes in the near future.

Author’s contributions K.P.M., X.C.M., W.W., B.C.H. and M.N.U. conceived the ideas and designed methodology; B.C.H., L.C., X.J.L., Y.L. and Y.Q.W. collected the data; B.C.H., M.N.U. and Y.Q.W. analysed the data; B.C.H. and M.N.U. led the writing of the manuscript. All authors contributed critically to the drafts and gave final approval for publication.

Acknowledgements We thank Dr. Kequan Pei, Dr. Yan Zhu, Jianyu Yang, Yin Guo, Wubing Xu, NingNing Wang and Yinan Liu for their helpful discussions and kind assistance for data collections. The research was financially supported by the National Natural Science Foundation of China (no. 31261120579) and M.N.U. was funded with US-China Dimensions of Biodiversity grant (DEB-1241136).

Data accessibility Data available from the Dryad Digital Repository https://doi.org/10.5061/dryad. 2cf46 (Han et al. 2017).

References Augspurger, C.K. (1984) Light requirements of neotropical tree seedlings: a comparative study of growth and survival. Journal of Ecology, 72, 777– 795. Bai, X.J., Queenborough, S.A., Wang, X.G. et al. (2012) Effects of local biotic neighbors and habitat heterogeneity on tree and shrub seedling survival in an old-growth temperate forest. Oecologia, 170, 755–765. Bates, D., Maechler, M., Bolker, B. & Walker, S. 2016. lme4:Linear mixedeffects models using Eigen and S4. R package version 1.1-12. Available at: http://CRAN.R-project.org/package=lme4 (16 April 2016). Beaudet, M. & Messier, C. (2002) Variation in canopy openness and light transmission following selection cutting in northern hardwood stands: an assessment based on hemispherical photographs. Agricultural and Forest Meteorology, 110, 217–228. Bolger, A.M., Lohse, M. & Usadel, B. (2014) Trimmomatic: a flexible trimmer for illumina sequence data. Bioinformatics, 30, 2114–2120. Bolker, B.M., Brooks, M.E., Clark, C.J., Geange, S.W., Poulsen, J.R., Stevens, M.H.H. & White, J.S.S. (2009) Generalized linear mixed models: a practical guide for ecology and evolution. Trends in Ecology & Evolution, 24, 127– 135. Canham, C.D., LePage, P.T. & Coates, K.D. (2004) A neighborhood analysis of canopy tree competition: effects of shading versus crowding. Canadian Journal of Forest Research, 34, 778–787. Chen, L., Mi, X.C., Comita, L.S., Zhang, L.W., Ren, H.B. & Ma, K.P. (2010) Community-level consequences of density dependence and habitat association in a subtropical broad-leaved forest. Ecology Letters, 13, 695–704. Chesson, P. (2000) Mechanisms of maintenance of species diversity. Annual Review of Ecology and Systematics, 31, 343–358. Comita, L.S. & Hubbell, S.P. (2009) Local neighborhood and species’ shade tolerance influence survival in a diverse seedling bank. Ecology, 90, 328–334. Connell, J.H. (1971) On the role of natural enemies in preventing competitive exclusion in some marine animals and in rain forest trees. In: Dynamics of Populations (eds P.J. den Boer & G.R. Gradwell), pp. 298– 312. Centre for Agricultural Publishing and Documentations, Wageningen, The Netherlands. Cornelissen, J., Lavorel, S., Garnier, E. et al. (2003) A handbook of protocols for standardised and easy measurement of plant functional traits worldwide. Australian Journal of Botany, 51, 335–380. van Dongen, S. (2000) Graph clustering by flow simulation. PhD dissertation, University of Utrecht, Utrecht, The Netherlands. van Dongen, S. & Abreu-Goodger, C. (2012) Using MCL to extract clusters from networks. Methods in Molecular Biology, 804, 281–295. Enright, A., van Dongen, S. & Ouzounis, C. (2002) An efficient algorithm for largescale detection of protein families. Nucleic Acids Research, 30, 1575–1584. Frazer, G.W., Canham, C.D. & Lertzman, K.P. (1999) Gap Light Analyser (GLA), version 2.0: imaging software to extract canopy structure and gap light indices from true-colour fisheye photographs. Simon Fraser University, Burnaby, BC, Canada; Institute of Ecosystem Studies, Millbrook, NY, USA.

© 2017 The Authors. Journal of Ecology © 2017 British Ecological Society, Journal of Ecology, 105, 592–601

Transcriptomes and seedling mortality 601 Grabherr, M.G., Haas, B.J., Yassour, M. et al. (2011) Full-length transcriptome assembly from RNA-Seq data without a reference genome. Nature Biotechnology, 29, 644–652. Haas, B.J., Papanicolaou, A., Yassour, M. et al. (2013) De novo transcript sequence reconstruction from RNA-seq using the Trinity platform for reference generation and analysis. Nature Protocols, 8, 1494–1512. Han, B., Uma~ na, M.N., Mi, X., Liu, X., Chen, L., Wang, Y., Liang, Y., Wei, W. & Ma, K. (2017) Data from: The role of transcriptomes linked with responses to light environment on seedling mortality in a subtropical forest, China. Dryad Digital Repository, https://doi.org/10.5061/dryad.2cf46 Holt, N.E., Fleming, G.R. & Niyogi, K.K. (2004) Toward an understanding of the mechanism of nonphotochemical quenching in green plants. Biochemistry, 43, 8281–8289. Horton, P., Ruban, A.V. & Walters, R.G. (1994) Regulation of light harvesting in green plants (indication by nonphotochemical quenching of chlorophyll fluorescence). Plant Physiology, 106, 415. Hubbell, S.P. (2001) The Unified Neutral Theory of Biodiversity and Biogeography. Princeton University Press, Princeton, NJ, USA. Hubbell, S.P. & Foster, R.B. (1986) Biology, chance, and history and the structure of tropical rain forest tree communities. Community Ecology (eds J. Diamond & T.J. Case), pp. 314–329. Harper and Row, New York, NY, USA. Iida, Y., Kohyama, T.S., Swenson, N.G., Su, S.H., Chen, C.T., Chiang, J.M. & Sun, I.-F. (2014) Linking functional traits and demographic rates in a subtropical tree community: the importance of size-dependency. Journal of Ecology, 102, 641–650. Janzen, D.H. (1970) Herbivores and the number of tree species in tropical forests. American Naturalist, 97, 501–528. Kasperbauer, M.J. (1987) Far-red light reflection from green leaves and effects on phytochrome-mediated assimilate partitioning under field conditions. Plant Physiology, 85, 350–354. Keddy, P.A. (1992) Assembly and response rules: two goals for predictive ecology. Journal of Vegetation Science, 3, 157–164. Khan, M.L., Rai, J.P.N. & Tripathi, R.S. (1986) Regeneration and survival of tree seedlings and sprouts in tropical deciduous and sub-tropical forests of Meghalaya. India Forest Ecology and Management, 14, 293–304. Kobe, R.K. (1999) Light gradient partitioning among tropical tree species through differential seedling mortality and growth. Ecology, 80, 187–201. Lai, J.S., Mi, X.C., Ren, H.B. & Ma, K.P. (2009) Species-habitat associations change in a subtropical forest of China. Journal of Vegetation Science, 20, 415–423. Lebrija-Trejos, E., Wright, S.J., Hern
andez, A. & Reich, P.B. (2014) Does relatedness matter? Phylogenetic density-dependent survival of seedlings in a tropical forest. Ecology, 95, 940–951. Lee, D.W., Baskaran, K., Mansor, M., Mohamad, H. & Yap, S.K. (1996) Irradiance and spectral quality affect Asian tropical rain forest tree seedling development. Ecology, 77, 568–580. Legendre, P., Mi, X.C., Ren, H.B., Ma, K.P., Yu, M.J., Sun, I.F. & He, F.L. (2009) Partitioning beta diversity in a subtropical broad-leaved forest of China. Ecology, 90, 663–674. Liu, X.J., Swenson, N.G., Zhang, J.L. & Ma, K.P. (2013) The environment and space, not phylogeny, determine trait dispersion in a subtropical forest. Functional Ecology, 27, 264–272. Liu, X.J., Swenson, N.G., Lin, D.M., Mi, X.C., Umana, M.N., Schmid, B. & Ma, K.P. (2016) Linking individual-level functional traits to tree growth in a subtropical forest. Ecology, 97, 2396–2405. MacArthur, R. & Levins, R. (1967) The limiting similarity, convergence, and divergence of coexisting species. American Naturalist, 101, 377–385. Man, X.X., Mi, X.C. & Ma, K.P. (2011) Effects of an ice storm on community structure of an evergreen broadleaved forest in Gutianshan National Nature Reserve, Zhejiang Province. Biodiversity Science, 19, 197–205. McGill, B.J., Enquist, B.J., Weiher, E. & Westoby, M. (2006) Rebuilding community ecology from functional traits. Trends in Ecology & Evolution, 21, 178–185. Nicotra, A.B., Chazdon, R.L. & Iriarte, S.V.B. (1999) Spatial heterogeneity of light and woody seedling regeneration in tropical wet forests. Ecology, 80, 1908–1926. Paine, C.E.T., Amissah, L., Auge, H. et al. (2015) Globally, functional traits are weak predictors of juvenile tree growth, and we do not know why. Journal of Ecology, 103, 978–989. Pavlidis, P., Jensen, J.D., Stephan, W. & Stamatakis, A. (2012) A critical assessment of storytelling: gene ontology categories and the importance of validating genomic scans. Molecular Biology and Evolution, 29, 3237–3248. Poorter, L. (1999) Growth responses of 15 rain-forest tree species to a light gradient: the relative importance of morphological and physiological traits. Functional Ecology, 13, 396–410.

Reich, P.B., Walters, M.B. & Ellsworth, D.S. (1997) From tropics to tundra: global convergence in plant function. Proceedings of the National Academy of Sciences United States of America, 94, 13730–13734. Scholes, J., Press, M. & Zipperlen, S. (1996) Differences in light energy utilisation and dissipation between dipterocarp rain forest tree seedlings. Oecologia, 109, 41–48. Smith, S.A., Moore, M.J., Brown, J. & Yang, Y.A. (2015) Analysis of phylogenomic datasets reveals conflict, concordance, and gene duplications with examples from animals and plants. BMC Evolutionary Biology, 15, 150. Song, Y.C. (2013) Evergreen Broad-leaved Forests in China, ClassificationEcology-Conservation. Science Press, Beijing, China. Stamatakis, A. (2006) RAxML-VI-HPC: maximum likelihood-based phylogenetic analyses with thousands of taxa and mixed models. Bioinformatics, 22, 2688–2690. Stubbs, W.J., Wilson, J.B. & Bastow Wilson, J. (2004) Evidence for limiting similarity in a sand dune community. Journal of Ecology, 92, 557–567. Sutherland, W.J., Freckleton, R.P., Godfray, H.C.J. et al. (2013) Identification of 100 fundamental ecological questions. Journal of Ecology, 101, 58–67. Swenson, N.G. (2012) The functional ecology and diversity of tropical tree assemblages through space and time: from local to regional and from traits to transcriptomes. ISRN Forestry, 2012, 743617. Swenson, N.G. (2013) The assembly of tropical tree communities–the advances and shortcomings of phylogenetic and functional trait analyses. Ecography, 36, 264–276. Swenson, N.G. (2014) Functional and Phylogenetic Ecology in R. Springer UseR! Series. Springer, New York, NY, USA. Swenson, N.G. & Enquist, B.J. (2007) Ecological and evolutionary determinants of a key plant functional trait: wood density and its community-wide variation across latitude and elevation. American Journal of Botany, 91, 451–459. Swenson, N.G. & Enquist, B.J. (2009) Opposing assembly mechanisms in a Neotropical dry forest: implications for phylogenetic and functional community ecology. Ecology, 90, 2161–2170. Swenson, N.G., Stegen, J.C., Davies, S.J. et al. (2012) Temporal turnover in the composition of tropical tree communities: functional determinism and phylogenetic stochasticity. Ecology, 93, 490–499. Umana, M.N., Forero-Montana, J., Muscarella, R., Nytch, C.J., Thompson, J., Uriarte, M., Zimmerman, J.K. & Swenson, N.G. (2016) Inter-specific functional convergence and divergence and intra-specific negative density dependence underlie the seed to seedling transition in tropical trees. American Naturalist, 187, 99–109. Weiher, E., Clarke, G.D.P. & Keddy, P.A. (1998) Community assembly rules, morphological dispersion, and the coexistence of plant species. Oikos, 81, 309–322. Wright, I.J., Reich, P.B., Westoby, M. et al. (2004) The worldwide leaf economics spectrum. Nature, 428, 821–827. Yang, Y. & Smith, S.A. (2014) Orthology inference in non-model organisms using transcriptomes and low-coverage genomes: improving accuracy and matrix occupancy for phylogenomics. Molecular Biology and Evolution, 31, 3081–3092. Yang, Y., Moore, M.J., Brockington, S.F. et al. (2015) Dissecting molecular evolution in the highly diverse plant clade Caryophyllales using transcriptome sequencing. Molecular Biology and Evolution, 32, 2001–2014. Received 17 October 2016; accepted 14 February 2017 Handling Editor: F. Andrew Jones

Supporting Information Details of electronic Supporting Information are provided below. Table S1. The 47 light-related gene ontology (GO) terms. Bold GO terms were used in this study. Table S2. Species used in the analyses. Species code is a unique code with six letters assigned to each species in GTS FDP.

© 2017 The Authors. Journal of Ecology © 2017 British Ecological Society, Journal of Ecology, 105, 592–601