1. Introduction

Charles Darwin noted that the orchid family Orchidaceae is an ideal model for studying evolutionary processes and adaptations (Darwin, 1859, 1890; Govaerts et al., 2021). One key adaptation of orchids is lifeform diversification (Küper et al., 2004). These lifeforms--i.e., terrestrial or epiphytic--allow orchids to occupy diverse niches, from the forest understory to the canopy (Givnish, 2010; Collobert et al., 2023). Epiphytic orchids, which account for 67.6% of the diversity within Orchidaceae (Silvera et al., 2009; Zotz et al., 2021), grow on unique arboreal habitats that make them highly sensitive to environmental stresses, such as drought and forest fragmentation (Zotz et al., 2021). In a plant family that has the highest proportion of threatened or extinct species (Swarts and Dixon, 2009; Fay, 2018; Wraith and Pickering, 2018), epiphytic orchids often face the highest extinction risk (Carmona-Higuita et al., 2024). Therefore, identifying the factors (e.g., climate, pollinator traits) that drive lifeform evolution in orchids is essential for understanding both their diversification and conservation.

Previous studies have shown that orchid lifeform is closely linked to climate. Extensive species distribution surveys have shown that epiphytic orchids mainly inhabit tropical and subtropical regions, whereas terrestrial orchids are more prevalent in temperate regions (Zotz, 2005; Taylor et al., 2021). This difference in distribution may reflect climate differences; at continental to global scales, latitude serves as an effective proxy for joint variation in temperature and water (Kreft and Jetz, 2007; Whittaker et al., 2007). For instance, lower latitudes provide year-round warmth and frequently wetter environments (Bruijnzeel et al., 2011), which help sustain arboreal water balance and foster the development of host tree substrates on which many epiphytic orchids establish. Temperate climates, in contrast, impose freezing risk that limits persistent canopy occupancy by epiphytes (Zotz and Hietz, 2001; Rasmussen and Rasmussen, 2018). However, both lifeform and climate regions of many plants are determined mainly by species evolution history (Davis, 2005; Wang et al., 2022), the independent contribution of climate region to lifeform evolution requires detailed evaluation.

Another important factor that influences lifeform evolution is species evolution history, which is generally represented by species phylogeny. Epidendroideae, the largest subfamily within Orchidaceae, is predominantly composed of epiphytic species. This subfamily contains approximately 19, 000 orchid species, or 95% of all epiphytic orchid species (Zotz et al., 2021). Remarkably, about 95% of extant epiphytic orchids share a single common ancestor within the subfamily Epidendroideae, with additional independent origins in subfamily Cypripedioideae, Orchidoideae, and Vanilloideae (Zhang et al., 2023). This dominant single-origin event underscores the noticeable phylogenetic conservatism of epiphytism in orchids. If orchid lifeform evolution is determined mainly by species evolution history, then any factors driving orchid diversification are likely to influence lifeform evolution, such as coadaptation with the orchid mycorrhizal fungi (Yukawa et al., 2009; Martos et al., 2012). Thus, the evolution of lifeforms in orchids can be understood more clearly if we assess the role of species evolution on traits evolution. However, accurately disentangling the relative contributions of phylogeny from potential phylogeny-correlated traits is a challenge (Adams and Collyer, 2018; Ives, 2019; Wang et al., 2022).

Pollination-related traits, which are critical to orchid diversification, have also been shown to be associated with lifeform evolution. Orchids deploy both deceptive and rewarding pollinator attraction strategies (Jia and Huang, 2022; Steffelová et al., 2023). Epiphytic orchids are commonly pollinated by specialist pollinators, and frequently employ deceptive pollination strategies, attracting pollinators without providing rewards (Huda and Wilcock, 2008; Scopece et al., 2010). In contrast, terrestrial orchids reward generalist pollinators with nectar and resin (Tremblay et al., 2005; Papadopulos et al., 2013), or exhibit a greater propensity for abiotic pollination (Aguiar et al., 2012; Ackerman et al., 2023). However, while pollination traits closely link with orchid diversification, their independent contribution on lifeform evolution remains unclear.

Accurately disentangling the relative contributions of phylogeny from other potential phylogeny-correlated traits is a challenge, as it goes against the fundamental assumption of independence principle among predictor variables in conventional statistical models (Adams and Collyer, 2018; Ives, 2019; Wang et al., 2022). For example, the commonly used phylogenetic generalized least squares models (PGLS) can only disentangle the relative contributions of predictor variables to residual variance after removing the effect of phylogenetic covariance (Pagel, 1997; Housworth et al., 2004). Although nested ANOVAs and linear mixed models can disentangle the relative contributions among multiple predictors, including phylogeny, they do so at the cost of oversimplifying the phylogenetic information into discrete taxonomic levels (e.g., family and genus) as nested random effects (Chen and Wei, 2003; Rohlfs and Nielsen, 2015). Fortunately, partial R2 framework can overcome these limitations. This approach allows assessing the relative contribution of both covariates (e.g., phylogenetic or spatial autocorrelation) and other predictor variables simultaneously by asking how much of the explained variation is lost when each or group of predictor variables and the phylogeny covariance is removed separately (Ives and Garland, 2010; Ives, 2019).

In this study, we quantitatively assessed the impact of various factors on lifeform variation within Orchidaceae (orchid plants). Specifically, we assessed the independent explanatory power of phylogeny, climate (tropical, subtropical, temperate), pollination vectors (biotic or abiotic pollination), and pollinator attraction strategies (rewarding or deceptive pollination) on the lifeforms (terrestrial, epiphytic) of 2272 orchid species across all five subfamilies and 302 genera.

2. Materials and methods 2.1. Orchid trait collection

Our dataset was derived from a recently published comprehensive compilation of orchid traits (Ackerman et al., 2023) that includes data from 1211 publications from 1877 to 2020, covering 2921 orchid species across all five subfamilies and 416 genera globally. Given the completeness of the dataset and its biological relevance to lifeform, we primarily selected the following traits from 2467 species: lifeforms (epiphytic or terrestrial), climate regions (determined by latitude, with temperate zones defined as > 35.00°; subtropical zones as 23.27–35.00°; and tropical zones as < 23.27°), pollination vectors (biotic or abiotic), and pollinator attraction strategies (deceptive or rewarding) with a proportion of missing values lower than 23.80%. To avoid the oversimplified classification of traits affecting the explanatory power of these traits for orchid lifeforms, our dataset also classifies detailed pollination vectors (e.g., Hymenoptera, Lepidoptera, Diptera, Coleoptera, Aves, other unclassified pollinators, abiotic pollinators) and pollinator attraction strategies (e.g., fragrance reward, nectar reward, oil reward, pollen reward, sleeping site reward, trichome reward, other reward, food deceit, sex deceit, other deceit). The proportion of missing data is below 23.80% for pollinator attraction strategies. See the supplementary datasets for details.

2.2. Orchid phylogeny construction and lifeform evolution reconstruction

The orchid phylogeny of 2467 selected species was constructed with the 'U.PhyloMaker' package (Jin and Qian, 2023) in R software (R Core Team, 2023). The 'U.PhyloMaker' package provides a reliable framework for assembling large scale phylogenies (Jin and Qian, 2023). It constructs phylogeny of given species lists based on a backbone phylogeny, which was constructed with genes available in GenBank, including 72, 570 species from 482 families and 10, 581 genera of vascular plants (Zanne et al., 2014; Smith and Brown, 2018). Genera and species that were not included in the backbone phylogeny were added as polytomies within their families and genera separately. We then removed 194 species from the orchid phylogeny because of the uncertainty of their genera position. All following analyses were conducted based on the phylogeny and datasets of the remaining 2272 orchid species, covering all five subfamilies, 21 out of 22 tribes, and 302 out of 736 genera of the Orchidaceae (Table S1). The phylogenetic relationships of the five subfamilies and main tribes are supported by the recently published orchid phylogeny of 1921 species (Pérez-Escobar et al., 2024).

To assess potential sampling bias of orchid species used in this study, we calculated Pielou's Evenness Index based on the total number of recorded species and the actual number of species used in each genus, resulting in a Pielou's Evenness Index of 0.827 (the value close to 1 indicating high diversity and high evenness). Additionally, the species accumulation curves show a similar trend between the total number of species and the actual number of species used (Fig. S1). This indicates that the distribution of species across genera in our sampling is relatively even, suggesting minimal bias in the dataset.

Ancestral state reconstruction of orchid lifeform was performed using the R package 'corHMM' (Boyko and Beaulieu, 2021), which applies a hidden Markov model to discrete traits while accounting for phylogenetic relationships. Three transition rate models, equal rates (ER), symmetrical rates (SYM), and all rates different (ARD), were fitted to the preprocessed Orchidaceae phylogeny and lifeform data. Model fit was assessed using the Akaike Information Criterion (AIC) and log-likelihood values. Under the best-fitted ARD model (Table S2), transition rate parameters and root priors were estimated with simulation of 500 stochastic character maps. Ancestral state probabilities for internal nodes, transition rate and times between terrestrial and epiphytic lifeforms were calculated (Table S3 and Fig. S2).

2.3. Phylogenetic conservatism calculation of orchid traits and their relative contribution to lifeform variation

To provide a foundational overview of the patterns within our dataset, we first summarized the overall distribution of lifeform, climate regions and pollination traits across the orchid phylogeny, and calculated the basic distributions of all factors across different lifeforms and their correlation, using the R packages 'ape' (Paradis and Schliep, 2019) and 'geiger' (Pennell et al., 2014).

To partition the relative contributions of phylogeny, climate region, pollination vector, and pollinator attraction strategy on lifeform evolution in orchids, we built a multi-factor model to calculate the Rlik2 of each factor or factor group on lifeform variation. First, we constructed a full model of phylogenetic logistic regression to regress lifeforms on the climate regions, pollination vectors, and pollinator attraction strategies, with orchid phylogeny as a covariate using the function phyloglm in the R package 'phylolm' (Ho and Ané, 2014). The Rlik2 for each factor or factor group were calculated by comparing the full model with the reduced models in which the factor or the factor group was removed, and measuring the consequent reduction in the likelihood, using the function R2_lik from the R package 'rr2' (Ives and Li, 2018). When evaluating the effect of phylogeny covariate, the reduced model is a generalized linear model that included all variables as full model but removed the phylogenetic covariate.

The granularity of trait classification (e.g., categorizing pollination vector into two broad types versus seven specific types), capturing different levels of biological information, may influence the relative contribution of each factor group (Rlik2) on lifeform evolution. To avoid potential bias via classification degree on evaluating the relative role of each factor on lifeform evolution, we additionally conducted similar analyses using more detailed classifications of pollination vectors (e.g., Hymenoptera, Lepidoptera, Diptera, Coleoptera, Aves, other unclassified pollinators, abiotic pollinators separately) and pollinator attraction strategies (e.g., fragrance reward, nectar reward, oil reward, pollen reward, sleeping site reward, trichome reward, other reward, food deceit, sex deceit, other deceit separately).

We found opposite traits relationship pattern when comparing the common correlation analysis with partial R2 framework. This discrepancy may result from the differences in phylogenetic conservatism among traits. Therefore, we assessed phylogenetic conservatism of each orchid trait (lifeform, climate region, pollinator attraction strategy and pollination vector) separately, using two metrics. First, we calculated a common phylogenetic signal index, the alpha (α) value, for discrete variables, using the penalized likelihood estimation method through the phyloglm function in the R package 'phylolm' (Ives and Garland, 2010). This parameter quantifies the phylogenetic signal strength of binary traits: α = 0 indicates trait evolution follows Brownian motion (strong phylogenetic signal), while α > 0 corresponds to an Ornstein-Uhlenbeck process, with larger values indicating weaker phylogenetic signal. Second, we assessed phylogenetic conservatism of each orchid trait by calculating the Rlik2 of orchid phylogeny for each trait using similar method shown above. The Rlik2 value is calculated by comparing a full phylogenetic logistic regression model, which regresses the trait on the phylogenetic covariance matrix against a reduced, non-phylogenetic model containing only an intercept (Ives, 2019).

3. Results

Overall, 50.3% (1144/2272) of orchid species are epiphytic, while 49.7% (1128) are terrestrial. Ancestral lifeform reconstruction indicated that epiphytism was derived from terrestrial ancestors, with 13 independent origins and 98 reversions, with the majority (99.39%) of epiphytes originating from a single radiation within subfamily Epidendroideae (Figs. 1A and S2; Table S3). Kendall correlation analysis showed that the epiphytic lifeform was strongly correlated with tropical regions (Kendall's τ = 0.70, P < 0.01), whereas orchid lifeform was only weakly correlated with both pollination vector and abiotic pollinators pollinator attraction strategy (i.e., absolute values lower than 0.06) (Fig. 1B–E).

When combining phylogeny, climate region and pollination traits together to explain lifeform evolution in orchid, the Rlik2 results support the dominant role of phylogeny. Orchid phylogeny alone explained 62% variation of orchid lifeform, followed by pollinator attraction strategy, which explained 24% of the variation independently. Climate region contributed the least, explaining only an additional 3.4% variation of lifeforms (Fig. 2A and Table S4). Similar results were obtained when pollination traits were classified in detail: orchid phylogeny remained the factor with the highest explanatory power (56.1%), while the explanatory power of pollinator attraction strategy and climate region (26.9% and 5.1%, respectively) increased slightly (Fig. 2B and Table S5).

Phylogenetic signal and phylogeny-only Rlik2 calculations indicated that phylogenetic conservatism of orchid traits varied substantially among orchid species. Orchid lifeforms exhibited the strongest phylogenetic conservatism, as indicated by the lowest α value (0.61); tropical and temperate climate regions showed relative weaker phylogenetic conservatism, with moderate α values (1.15 and 2.07, respectively). In contrast, phylogenetic conservatism was markedly weaker for pollination-related traits, as indicated by their higher α values: 3.37 for pollinator attraction strategy and 5.62 for pollination vector (Fig. 2C; Table S6). A similar pattern of phylogenetic conservatism on orchid traits was also detected by the Rlik2 calculation. When excluding phylogeny-dependent traits such as climate region and pollination, orchid phylogeny alone accounted for 86.3% of the variation in lifeform. The distribution of species in tropical (73.2%) or temperate regions (67.6%) was also explained largely by orchid phylogeny. In contrast, pollinator attraction strategies (45.2%) and pollination vectors (17.2%) showed weaker phylogenetic influence (Fig. 2D and Table S6). In addition, species distribution in subtropical region also showed weak phylogenetic conservatism indicated by both a high α value and a low Rlik2 value (Fig. 2C and D; Table S6).

4. Discussion

Epiphytic orchids exhibit greater species diversity and inhabit more extreme physical environments than do terrestrial orchids (Gravendeel et al., 2004). Understanding evolutionary adaptations in orchids requires identifying the factors that drive variation in orchid lifeform. Here, we quantified the relative contribution of phylogeny, climate, and pollination traits on orchid lifeform variation, thereby revealed the dominant role of species evolution process (phylogeny) on lifeform evolution in orchids.

The nearly opposite traits association patterns revealed by the common correlation analysis and the partial R2 framework highlight the nonnegligible role of species evolution in lifeform evolution. Basic correlation analysis indicated that orchid lifeform is strongly associated with climate and minimally associated with pollination traits (Fig. 1E). However, when considering orchids phylogeny, we found that the apparently strong correlation between climate and lifeform was largely determined by the shared evolutionary history across orchid diversification (Fig. 2A and B). For instance, the subfamily Epidendroideae comprises ~71% of orchid species, most of which are epiphytic and distributed in tropical regions, whereas species from another large clade, the subfamily Orchidoideae, are mainly terrestrial and distributed in temperate regions (Chase et al., 2015; Zhang et al., 2023). This leads to highly similar phylogenetic distribution patterns between lifeform and climates, thereby creating an apparently strong association. Thus, when excluding covariate effect of orchid phylogeny, climate alone explained only a small amount of lifeform variation (3.4%) (Fig. 2A). Notably, the limited effect of climate on lifeform evolution pertains to macro-evolutionary or large-geographical scales. Our finding does not refute the influence of micro-climate factors, such as humidity or temperature on the distribution and abundance of orchids with different lifeforms in local communities, especially in vertical gradients of forests, from ground to canopy (Gotsch et al., 2017; Miller et al., 2021).

The strong link between pollination traits and lifeform in orchid evolution was also obscured by evolutionary history (Fig. 1E). In contrast to both lifeform and climate region, pollination traits exhibit quite low phylogenetic conservatism (Fig. 2A and B). The relatively high contribution (23.9%) of pollinator attraction strategies on lifeform variation only emerged when the effect of phylogeny was removed. The association between lifeform and pollinator traits may have been established during species diversification and/or as a result of population dynamics, in which different orchid lifeforms may require diverse pollinator attraction strategies (Neiland and Wilcock, 1998; Huda and Wilcock, 2008; Johnson, 2010). Both epiphytic and terrestrial orchids adapt their pollination strategies or lifeforms to optimize fitness under varying environmental conditions, thereby enhancing the evolutionary association along the species evolution (Van der Niet et al., 2014). For instance, epiphytic orchids often exhibit higher sympatric species diversity but have relatively small population sizes, due to space constraints and the dry, strong-light conditions of their arboreal habitat (Phillips et al., 2020). Thus, the deceptive pollination strategy might be more advantageous in enhancing genetic novelty to adapt to more heterogeneous habitats by improving outcrossing among long-distance populations via deceived pollinators (Huda and Wilcock, 2008). In contrast, terrestrial orchids may often have relatively lower sympatric species diversity. The population size of terrestrial orchid species benefits from rich soil but suffers from stronger competition with other plants. Thus, the rewarding pollination strategy may be preferred to enhance fruit setting rates and rapidly increase population sizes in resource-abundant environments (Tremblay et al., 2005).

Our findings highlight the dominant role of species evolution in lifeform evolution, suggesting that any factors that influence orchid diversification can influence lifeform evolution, either directly or indirectly. For example, beyond pollination traits, micro-climate factors and mycorrhizal fungi may also influence lifeform evolution via their influence on species evolution of orchids. Epiphytic and terrestrial orchids show clear differences in the micro-climate preference (Zhang et al., 2015; Lima et al., 2023) and mycorrhizal associations (Martos et al., 2012; Johnson et al., 2021). Variation in micro-climates and mycorrhizal associations among local communities can influence population fitness of orchids with different lifeforms, and may contribute to the divergent selection among populations or species with varied lifeform (Jacquemyn et al., 2014; McCormick et al., 2018). The ecological selection and genetic basis of lifeform evolution along the species evolution of orchids represents another interesting topic in orchid research, which may further improve our understanding of trait evolution via species diversification. Similarly, more factors or processes that may influence species diversification, such as genome polyploidization (Moraes et al., 2022) and herbivore pressure (Spicer and Woods, 2022) could affect lifeform evolution by influencing species diversification. Recognizing the dominant role of species evolution in trait evolution provides a more comprehensive framework for understanding the drivers of trait evolution.

Our study also offers novel insights into orchid conservation by supplying a deep understanding of the evolutionary forces shaping their adaptive traits (Hansen, 1997; Labra et al., 2009). By uncovering the substantial variation of phylogenetic conservatism on different traits, our study reveals a major difference in adaptive potential of different traits. Traits with high phylogenetic conservatism are stable and unlikely to adapt to changing environment (Losos, 2008; Cooper et al., 2010). For example, the endangered status of some orchids is constrained by traits with high phylogenetic conservatism, e.g., epiphytic lifeform via habitat fragmentation or canopy destruction (Wiens et al., 2010). For those species, the extinction risk is difficult to mitigate by evolutionary adaptation; thus, artificial assistance, such as habitat protection and ex-situ conservation, will be required (Wiens, 2004; Broennimann et al., 2007; Crisp and Cook, 2012). In contrast, traits with low phylogenetic conservatism indicate high evolutionary plasticity and adaptive potential. For instance, if the endangered status of some orchids is primarily linked to pollination traits, the extinction risk may be mitigated by modifying their non-adaptive traits, such as shifting to alternative pollinators or transitioning to self-pollination in response to environmental change. Previous studies have also demonstrated rapid plant evolution in pollination systems under changed biotic environments (Ramos and Schiestl, 2019; Dorey and Schiestl, 2024).

Therefore, different conservation strategies may be required for endangered species depending on the phylogenetic conservatism of their non-adaptive traits. For example, in-situ conservation may be required for species reliant on highly conservative traits, e.g., lifeforms or climate regions, whereas maintaining dynamic populations under changing environment; while ex-situ conservation may be more suitable for endangered species limited currently by low-conservative traits, e.g., pollination traits in orchids. This is especially important for orchids conservation, as there is a notable imbalance between significant conservation gaps and limited conservation capacity (Veach et al., 2017; Shrestha et al., 2019), which necessitates optimized resource allocation (Ackerly et al., 2000; Ackerly, 2009; Sanchez-Martinez et al., 2024).

In conclusion, our study uncovered the often overlooked yet disproportionately important role of species evolution in trait evolution. Accordingly, we propose an evolutionary view of orchid conservation that considers the adaptive potential of critical traits of endangered taxa.

Acknowledgments

We would like to thank Yanbao Ma and Ying Feng for their contributions to improving the clarity and quality of the manuscript. This work was supported by the Yunnan Revitalization Talent Support Program (XDYC-QNRC-20230573), the National Natural Science Foundation of China grant (32371701), the 14th Five-Year Plan of the Xishuangbanna Tropical Botanical Garden, CAS (E3ZKFF3B), Xizang Yarlung Zangbu Grand Canyon National Nature Reserve expenditure project of forestry and grassland ecological protection and restoration funding (GZFCG2023-14256), and construction and management of the research center for the protection and utilization of orchids in Motuo, Xizang Autonomous Region, China (KH230350A).

Ethics

This work did not require ethical approval from a human subject or animal welfare committee.

Data availability statement

Raw data and custom scripts are available in the Data Storage Community of Plant Diversity at Science Data Bank (https://doi.org/10.57760/sciencedb.j00143.00145) and Zenodo (https://doi.org/10.5281/zenodo.17528133).

CRediT authorship contribution statement

Tianwen Zhang: Conceptualization, Methodology, Investigation, Data curation, Formal analysis, Visualization, Writing – original draft, Writing – review & editing. Jun-Wen Zhai: Conceptualization, Writing – review & editing, Supervision, Project administration, Funding acquisition. Gang Wang: Conceptualization, Methodology, Writing – review & editing, Supervision, Project administration, Funding acquisition.

Declaration of generative AI and AI-assisted technologies in the manuscript preparation process

During the preparation of this work the authors used ChatGPT in order to improve the English expression. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the published article.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.pld.2026.01.005.