Ancient bridges, modern threat: Conserving landscape heterogeneity ensures sustenance of living root-bridges in Meghalaya

17 The Anthropocene, marked by rapid biodiversity loss, has renewed attention on ‘nature’s 18 contributions to people’, which include ecological, cultural, and spiritual values. A striking 19 example of such nature –culture interactions are the living root -bridges of Meghalaya, India, 20 where Khasi people ingeniously train the aerial roots of Ficus elastica across river valleys to 21 form natural bridges. These structures, currently under consideration for UNESCO World 22 Heritage recognition, exemplify sustainable resource use in one of the wettest and most 23 landslide-prone regions of the world. However, increasing tourism and unregulated 24 construction is threatening to this nature culture relationship in a biodiversity hotspot. In such 25 situations, understanding the factors governing the geneflow is vital to the preservation of 26 interactions between the pollinating fig wasps, the fig tree and its seed dispersers. Despite its 27 ecological and cultural significance, little is known about the population biology or dispersal 28 potential of F. elastica. Here, we investigated dispersal potential of natural populations of F. 29 elastica across Meghalaya. Using ddRAD -SNP genotyping (~12K SNPs from 308 30 individuals), we detected four genetic clusters corresponding to E. Khasi, W. Khasi, W. Jaintia, 31 and Ribhoi hills. Redundancy analysis and resistance modelling revealed that wind regimes 32 and topography jointly structure these populations, enabling gene flow within but not across 33 river valleys. Analysis of spatial distribution of related individuals indicates short dispersal 34 distance of 1 -4km varying across different populations . Given F. elastica’s association with 35 riparian ecosystems, and with UNESCO recognition pending, our study underscores the need 36 for strict guidelines to curb habitat destruction and ensure the long -term survival of both the 37 species and its cultural legacy. 38 39

Fine-scale genetic structure, Ficus elastica, landscape heterogeneity, living root-40 bridges, dispersal potential 41 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 3, 2025. ; https://doi.org/10.1101/2025.09.01.673595doi: bioRxiv preprint 3

42 The current epoch, earlier called the Anthropocene, is characterised by biodiversity loss and 43 habitat fragmentation. While earlier frameworks valued nature primarily through ecosystem 44 services, the IPBES (Inter-governmental panel of biodiversity and ecosystem services) now 45 emphasizes the broader concept of ‘nature’s contributions to people’, encompassing 46 ecological, cultural, and spiritual values. This is especially relevant given that indigenous 47 communities manage nearly a quarter of Earth’s terrestrial habitats 48 (https://www.unep.org/news-and-stories/story/indigenous-peoples-and-nature-they-protect). 49 One famous example of such nature–culture interactions are the living root-bridges of 50 Meghalaya, India, where the Khasi people train aerial roots of Ficus elastica to form bridges 51 across steep, river-cut valleys. These living root-bridges or ‘jienkieng jri” are an example of 52 sustainable utilisation of natural resources by the local communities as means of 53 transportation across steep valleys with strongly-flowing streams (Ludwig et al. 2019). 54 Beside bridges, the species is also important for reducing the soil erosion in this landslide 55 prone area of Meghalaya, the wettest region in India, and the world. Given the ecological 56 importance and uniqueness of this practice there is an on-going process to declare the living 57 root-bridges as UNESCO world cultural heritage sites. However, the enhanced tourism to 58 these living root-bridge sites has also resulted in unchecked constructions across the hills, 59 causing major deforestation and landslide issues, threatening both the biodiversity and the 60 continuity of this traditional practice. 61 But what are the biological underpinnings of these root-bridges? These root-bridges are made 62 using Ficus elastica, a monoecious hemi-epiphytic strangler fig, naturally occurring in the 63 hilly terrain of India’s northeastern state Meghalaya. In the early 19th century, F. elastica or 64 the Indian rubber tree was widely cultivated across Asia for latex production resulting in 65 obscuring the geographical origin of this species. Till date, there are contradictions about the 66 global native range of this species, known to be occurring in mountainous limestone areas of 67 Northeast India, Myanmar and peninsular Malaysia (Corner 1985, Choudhary 2012, Harrison 68 et al., 2017). Particularly in India, the species was first documented in Meghalaya and then 69 subsequent surveys concluded a disproportionate abundance in southern Meghalaya, i.e. East 70 Khasi and West Jaintia hills, with sparse distribution in northern Meghalaya, Mizoram and 71 West Bengal (Choudhary 2012, Shukla et al., 2014, Singh et al., 2015). Similar to other hemi-72 epiphytes, its seedlings germinate in the upper canopy of a host tree with the mature sapling 73 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 3, 2025. ; https://doi.org/10.1101/2025.09.01.673595doi: bioRxiv preprint 4 gradually strangling the host tree and eventually setting down aerial and terrestrial root 74 systems. In the absence of host trees, it can also germinate and grow from cuttings on rocks, 75 boulders and cliffs, often common in such hilly riverine landscape. It is pollinated by a single 76 agaonid wasp species Platyscapa clavigera with seed dispersal possibly via birds and 77 mammals (Choudhary 2012, Nongbri et al. 2021). Although long distance seed dispersal 78 seems to be unlikely or rare (Heer et al., 2015, Krishnan & Borges 2018), it could happen via 79 planting of vegetative cuttings, given its use by local communities. 80 Till date, most of the studies on the natural population of F. elastica have been limited to the 81 mechanical properties of their roots and how the species can be incorporated into “green” 82 development (Choudhuri & Samal 2016; Ludwig et al., 2019). Some studies are available on 83 its germination, effective dispersal distance and ecology limited to its Southeast Asia’s 84 populations (Putz 1989, Harrison & Rasplus 2006, Hao et al. 2010), but nothing about its 85 genetic diversity. Within India, there is only one targeted study on its natural population of 86 Meghalaya giving insight to the breeding phenology of the species (Nongbri et al. 2021). 87 Therefore, through this study we want to contribute to the existing knowledge of the breeding 88 biology of F. elastica by investigating the dispersal potential of the species and factors 89 governing it. By assessing the geneflow patterns, one can understand the intricate relation of 90 the fig, wasps and their surroundings, thereby identifying conservation priority areas and/or 91 proposing population management plans in the current Anthropocene era of climate change 92 and habitat alteration. 93 Gene flow, a cornerstone of eco-evolutionary studies, reveals how extrinsic and intrinsic 94 factors shape species’ genetic structure and biodiversity (Avise 2000; Hodel et al., 2018; 95 Medina et al., 2018; Fenderson et al., 2019). Since the inception of landscape genetics in the 96 1990s, such studies have informed biodiversity management at landscape scales and in 97 response to future climatic changes (van Strien et al., 2014; Bowman et al., 2016; Savary et 98 al., 2022). Although such studies are common for animals, there are only a handful for plants 99 especially at fine scale spatial level (Cruzan & Hendrickson 2020). In plants, gene flow 100 depends on pollen and seed movement driven by pollinators/seed dispersers and bioclimatic 101 conditions, mediated by reproductive traits (Nazareno et al., 2021). These dynamics are 102 particularly pronounced in the obligate Ficus–wasp mutualism, where pollinator (the 103 Agonidae wasps) reproduction is tightly coupled with pollen transfer (Wang et al., 2009; 104 Cruaud et al., 2012; Deng et al., 2020; Alamo-Herrera et al., 2022). These wasps are mobile 105 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 3, 2025. ; https://doi.org/10.1101/2025.09.01.673595doi: bioRxiv preprint 5 dispersers relying passively on wind regimes (strength and direction) for their movement 106 resulting in contrasting genetic patterns between two sexual system (monoecious vs 107 dioecious) in this genus (Backes and Jump 2011, Nazareno et al., 2013, Heer et al., 2015, 108 Cruzan & Hendrickson 2020, Borges 2021). For instance, dioecious species show restricted 109 pollen flow due to lower-canopy wind use (e.g., 200 m in F. hispida, F. exasperata), while 110 monoecious species-associated wasps exploit high-canopy winds for longer-distance 111 movement (Ahmed et al., 2009; Dev et al., 2011; Nazareno et al., 2013; Yang et al., 2015; 112 Kling & Ackerly 2021). Although these patterns broadly explain genus-level gene flow, 113 recent fine-scale studies in heterogeneous landscapes (such as mountainous regions or urban 114 landscapes) often deviate from expectations (Krishnan & Borges 2018; Nazareno et al., 2021; 115 Deng et al., 2025), underscoring the need to account for landscape heterogeneity beyond 116 simple geographic distance. In addition, seed-mediated dispersal also contributes to the 117 geneflow, but it is generally more spatially constrained than pollen flow, reflecting disperser-118 specific mobility (Petit et al., 2005; Lomáscolo et al., 2008; Yu et al., 2010; Heer et al., 119 2015). 120 Given that F. elastica is a wasp-pollinated species inhabiting riverine hill terrains of 121 Meghalaya, a correlation between wind connectivity and genetic variation is expected (Kling 122 & Ackerly 2021; Deng et al., 2025). However, in such landscapes, valleys and river 123 catchments can disrupt wind flow, creating complex interactions among topographic features 124 that shape genetic heterogeneity (Cruzan & Hendrickson 2020; Rojas‐Cortés et al., 2024; 125 Deng et al., 2025). We therefore hypothesize the presence of genetic lineages differentiated 126 by both wind and topography, and test whether observed genetic clusters arise from isolation 127 by distance and/or environmental factors such as valleys and wind regimes. To address this, 128 we combined population genetic structure analyses with genotype–environment association 129 tests and resistance modelling, and further assessed the spatial distribution of related 130 individuals to estimate potential dispersal distance. Using ddRAD-based SNP genotyping 131 with extensive sampling across Meghalaya, we not only fill a key knowledge gap in F. 132 elastica biology but also provide insights relevant to sustainable tourism and development, 133 particularly in light of the recent recognition of living root-bridges as UNESCO World 134 Heritage sites. 135

136 Sampling 137 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 3, 2025. ; https://doi.org/10.1101/2025.09.01.673595doi: bioRxiv preprint 6 F. elastica is found in the mountainous limestone rainforests of Meghalaya, with highest 138 density in East Khasi and West Jaintia hills. Reports of sparse distribution are also 139 documented from West Khasi, Southwest Khasi, East Jaintia and Ribhoi districts (Choudhary 140 2012, Shukla et al., 2014, Sharma et al., 2016). Given the lack of previous studies addressing 141 F. elastica natural populations in Meghalaya, we sampled thoroughly across valleys and 142 forested areas representing the elevation gradient of 50–1250 amsl. As shown in Figure 1, 143 southern populations of Meghalaya (East Khasi and West Jaintia) have more landscape 144 heterogeneity than the northern populations, therefore providing a natural system to test our 145 hypothesis. Overall, we collected 371 samples, one leaf from each individual tree along with 146 GPS coordinates, elevation, identity of streams and valleys. To avoid fungal growth, samples 147 were placed in silica gel and air dried before they were transferred to the lab in Bangalore for 148 long term storage at -20C. No natural population was found in Southwest Khasi and East 149 Jaintia hills. 150 DNA extraction and library preparation 151 Genomic DNA was isolated from 50 mg of leaf by disrupting the cell walls using liquid 152 nitrogen followed by Qiagen DNeasy plant pro kit according to the manufacturer’s 153 instructions. To avoid PCR failure due to secondary metabolites and latex-based inhibitors, 154 DNA extracts were purified using the magnetic bead-based method. For each set of 155 extractions, DNA was quantified along with a negative control sample to ensure no 156 contamination during the process. 157 ddRAD libraries were prepared using the protocol mentioned in Tyagi et al. (2024). Briefly, 158 200 ng of DNA was digested with enzyme pair Sph1 and Mluc1 (six and four base-pair 159 cutters respectively) followed by adaptor ligation to the cut ends. The unbound adapters are 160 removed by using 0.5X magnetic bead-based cleaning. Finally, PCR amplification was 161 performed with individual specific index primer pairs and the products are dual size-selected 162 for a targeted library of 200– 500 bp. The pooled library of 2nM was sequenced in 163 HiSeq2500 and NovaSeq6000 platforms. 164 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 3, 2025. ; https://doi.org/10.1101/2025.09.01.673595doi: bioRxiv preprint 7 165 Figure1: Sampling location of Ficus elastica populations (represented by black triangles) across four 166 districts of Meghalaya along with highlighted steep valley areas (denoted by brown colour). 167 Variant calling and filtering 168 Raw reads were trimmed and barcodes were removed using custom script. Trimmomatic 169 v0.33 (Bolger et al., 2014) was used to check the quality control of raw reads by removing 170 reads with an average quality lower than 30 and shorter than 30 bp. All paired trimmed reads 171 were aligned with the Ficus microcarpa genome (phylogenetically closest species) (Rasplus 172 et al., 2018) using default settings of the BWA-MEM algorithm (Li & Durbin, 2009). Variant 173 calling was done using SAM and BCF tool pipelines (Li et al., 2009). Raw reads located in 174 repetitive regions with respect to the reference genome, <5 bp to indels and with lower 175 mapping quality (<30), were discarded. Further SNPs with sequencing depth <5, genotype 176 quality <30 and minor allele frequency <0.05 were filtered out. In addition, any sample with 177 more than 30% missing data was not used in downstream analysis. The filtered SNPs were 178 finally checked for Hardy-Weinberg Equilibrium. All the filtering steps were done using 179 VCFtools (Danecek et al., 2011). 180 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 3, 2025. ; https://doi.org/10.1101/2025.09.01.673595doi: bioRxiv preprint 8 Individual Identification and genetic relatedness 181 KING v2.3.1 (Manichaikul et al., 2020) was used to identify clones and calculate genetic 182 relatedness among the individuals as it is robust to the presence of population structure. The 183 estimates are based on kinship coefficient and identity by descent segments for all pairwise 184 relationships. The first order related individuals were removed from the subsequent 185 population clustering analysis to avoid over-estimation of ‘K’. 186 Assessing fine-scale genetic structure, diversity and dispersal potential 187 To identify the population structure of a perennial woody plant at a spatial extent of ~100km 188 (maximum linear distance between sampling points), we implemented an integrated 189 clustering approach. First, PCA was carried out using PLINK (Purcell et al., 2007) followed 190 by spatially explicit Bayesian clustering analyses using tess3R package (Caye et al. 2016). 191 This was done as PCA are known to be sensitive towards spatial heterogeneity and 192 autocorrelation thus resulting in underestimation of K’s especially at fine spatial scale. We 193 tested K values for 1–10 with 50 replicates and 10,000 iterations. True genetic clusters were 194 decided based on the cross-entropy criterion with 10% masked genotype. Subsequently, 195 genetic differentiation was estimated by using Weir and Cockerham’s pairwise Fst method 196 implemented in the hierfstat R package (Goudet 2005) with 10,000 iterations for 0.05 197 significance test. Population specific genetic diversity indices like heterozygosity and 198 inbreeding coefficient were calculated through the adegenet package (Jombart et al., 2018). 199 Finally, dispersal potential was estimated using Mantel correlogram in vegan package 200 (Oksanen et al., 2022) between genetic relatedness and geographic distance within each 201 genetic cluster. This was done using Spearman’s correlation test with 10,000 iterations. 202 Influence of space and environment on fine scale genetic structure 203 We tested both isolation by distance (IBD) and isolation by topography (IBT) to account for 204 topographical separation and impedance of gene flow, as may occur between river valleys. 205 For IBD, Mantel tests between individual pairwise Fst and geographic distance (km) were 206 performed. For topographical distance, topoDist package (Wang 2020) was used to calculate 207 paths along all 8 cardinal directions for each sampling point. Similar to IBD, Mantel tests 208 were also conducted between the topographic and genetic distance matrix. Both tests were 209 performed using the mantel.rtest function in the ade4 package (Dray and Dufour 2007). 210 Assuming such spatial autocorrelation may not be stationary (i.e linear relation with genetic 211 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 3, 2025. ; https://doi.org/10.1101/2025.09.01.673595doi: bioRxiv preprint 9 distance), the EEMS (estimation of effective migration surface) program was used as 212 exploratory tool to detect the regions of biogeographical barriers deviating the genetic 213 diversity patterns from the null expectation of IBD (Petkova 2016). We ran two independent 214 MCMC chains for 600 demes, 5 million MCMC runs with 50,0000 burnin and 5000 215 iterations to remove sample autocorrelations. Finally results from 2 chains were combined 216 and checked for convergence of log likelihoods using the EEMS plotting (rEEMSplot) 217 package. 218 Post barrier detection, relative contribution of environmental features (wind and topography) 219 on genomic variation was quantified using two approaches. First, we performed multivariate 220 redundancy analysis (RDA) to do genotype-environment association test (Capblancq & 221 Forester 2021). The genotypic matrix per individual was generated using the vcfR package as 222 the response variable (Knaus & Grünwald 2017). For environment variables, raster layers of 223 two wind parameter (u and v component) and three landscape features (topography (mTPI), 224 slope and hydrologically adjusted elevation (HAE)) were downloaded from Google Earth 225 Engine (GEE) (details provided in Supplementary table 1). mTPI (muti-scale topographical 226 position index) is a terrain index which can distinguish ridges from valley, thus providing 227 information about the physiogeography of an area. Similarly, to account for riverine valleys, 228 especially the downward direction of waterflow, we used hydrologically adjusted elevation 229 (HAE). As long-distance movement of Ficus pollinators is influenced by wind speed and 230 direction, we calculated these parameters for the study area using the u and v component 231 layers (details in supplementary R script). To mitigate overfitting, forward selections on 232 environmental were conducted using the ordiR2step function. Following this, significance 233 testing of selected variables (based on adjusted R2 values) for partitioning in RDA models 234 was done with 1000 permutations using the anova.cca function. These procedures were 235 implemented using the vegan R package (Oksanen et al., 2022). Second, we estimated the 236 magnitude and direction of each environmental layer on geneflow by modelling conductance 237 using the radish package (Peterman & Pope 2021). Positive coefficient values indicate 238 increase in conductance (i.e. gene flow or migration) and vice versa by negative estimates at 239 higher values of the covariate. Proportion of shared alleles (DPS) was calculated by the poppr 240 package as the response variable. To estimate the dependency of genetic and geographic 241 distance, MLPE (maximum likelihood population effect) regression was used along with 242 loglinear conductance model. The landscape variables were downloaded from Google Earth 243 Engine (GEE) and wind speed layer of 50m height from ground (as monoecious wasp use 244 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 3, 2025. ; https://doi.org/10.1101/2025.09.01.673595doi: bioRxiv preprint 10 high canopy tree wind regime) was downloaded from Global Wind Atlas at 1km resolution. 245 Further, we tested the significance of landscape conductance over geographical distance in 246 explaining the genetic structure by fitting an IBD model followed by Anova test. All these 247 analyses were done in R v. 4.3.3 using raster, sp and sf packages (Hijmans et al.,2015, 248 Pebesma et al., 2012, Pebesma 2018). 249

250 Ficus elastica population genetics reveals fine scale genetic structure and varied 251 dispersal potential 252 ddRAD based SNP-genotyping resulted in 12,566 SNPs after variant filtering steps. Using 253 these SNPs, we identified 308 unique and 272 unrelated (excluding the first-order 254 relationships like parent-offspring and full-sibs) individuals. Recaptures were only found in 255 E. Khasi and W. Jaintia hills, where the practice of root-bridge construction persists. 256 Clustering analysis (based on 272 unrelated individuals) using TESS reveals three major 257 clusters corresponding to E. Khasi, W. Jaintia and Ribhoi-W. Khasi as one cluster (Fig. 2b). 258 Since E. Khasi was the most diverged cluster as seen in PCA (Fig. 2a), we re-ran TESS by 259 removing the E. Khasi individuals. This resulted in further division among Ribhoi and W. 260 Khasi populations (Fig 2c), with few admixed individuals in nearby lower elevation areas. 261 The genetic differentiation among the populations was low (0.02–0.05) but significant 262 (p<0.5) with E. Khasi having the most diverged population as seen in the earlier analysis 263 (Figure 2d). Overall, the average Ho and Fis across identified genetic clusters were 0.27 and -264 0.04 respectively. At the population level, these indices are very similar across populations, 265 except for E. Khasi showing the lowest Fis value of -0.005 (Table 1). 266 Table 1: Population wise details of sampling, individual identification and genetic diversity indices 267 Populations Samples collected Individuals identified HO FIS E. Khasi 236 189 0.11 -0.005 W. Khasi 17 15 0.10 -0.05 W. Jaintia 76 62 0.10 -0.02 Ribhoi 42 42 0.10 -0.03 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 3, 2025. ; https://doi.org/10.1101/2025.09.01.673595doi: bioRxiv preprint 11 Figure 2: Result of genetic structure analysis using integrated clustering approach as a) PCA was not 268 able to differentiate any population other than the E. Khasi. To account for spatial heterogeneity and 269 autocorrelation at fine spatial scale, tess3R was implemented which resulted in b) three major clusters 270 corresponding to E. Khasi, W. Jaintia along with Ribhoi and W. Khasi as one. Given, E. Khasi is the 271 most differentiated population, c) one more tess3R run was conducted after removing the population, 272 resulting in further differentiation of W. Khasi and Ribhoi population. Overall low genetic 273 differentiation is observed between populations with significant pairwise FST values (p < 0.05) 274 denoted with *. 275 The correlogram results of all populations show that relatedness is significantly negatively 276 correlated at shorter geographic distance of 1–4 km (varies across population) indicating 277 short distance gene flow. However, in E. Khasi hills, positive correlation was also present, 278 indicating long distance gene flow within a range of 13–18km (Figure 3a–d). 279 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 3, 2025. ; https://doi.org/10.1101/2025.09.01.673595doi: bioRxiv preprint 12 280 Figure 3: Mantel correlogram to understand the correlation between pairwise relatedness and 281 geographic distance in a) E. Khasi, b) W. Jaintia, c) Ribhoi and d) W. Khasi. Filled shape indicates 282 the significant correlation at p-value < 0.05. The results show short distance dispersal for all the 283 populations (negative correlation) along with long distance dispersal only for E. Khasi population 284 (positive correlation). 285 Fine scale genetic structure as a response to environmental heterogeneity 286 The influence of isolation by distance and topography on genetic diversity seems to be weak 287 but significant across the landscape (r=0.3, p=0.001). Similar uneven patterns were also 288 observed in the migration surface of the EEMS output highlighting valleys and river 289 catchment areas as barriers to gene flow, especially in the southern region of the F. elastica 290 distribution (Figure 4b). This was further corroborated by RDA bi-plots where the northern 291 population clusters (Ribhoi and W. Khasi) were associated with wind parameters whereas the 292 southern populations (E. Khasi and W. Jaintia) are affected by landscape heterogeneity 293 (Figure 4c). Overall, variance in data is mostly explained by wind speed followed by wind 294 direction, mTPI, slope and HAE in descending order (Table 2). The coefficient estimates of 295 conductance analyses show slope to be insignificant in contributing to gene flow. However, 296 the other three variables were significant with mTPI and HAE showing the highest negative 297 coefficient estimate and wind speed with a positive effect (Table 2 and Figure 4d). Finally, 298 the ANOVA test also supports the hypothesis of environmental variables having a significant 299 effect on genomic variation rather than just geographic distance. 300 301 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 3, 2025. ; https://doi.org/10.1101/2025.09.01.673595doi: bioRxiv preprint 13 Table 2: Details of respective variable contribution towards population structuring estimated by 302 redundancy and geneflow facilitation assessed by conductance analysis. 303 # non-significant 304 NA- Not applicable 305 306 Figure 4: The figure highlights the contribution of environmental heterogeneity in explaining the fine 307 scale genetic structure and short distance dispersal pattern observed in the natural population of Ficus 308 elastica in Meghalaya. Both the (b) EEMS and (d) Conductance analysis concludes the presence of 309 valleys and river catchments as barriers to geneflow resulting in stark population differentiation (a) 310 between geographically closed populations (E. Khasi and W. Jaintia). Additionally, (c) RDA results 311 corroborate with conductance analysis where wind parameters define the genetic composition of 312 northern populations (Ribhoi and W. Khasi) with higher conductance rate in those populations (d). 313 314 315 Redundancy results Conductance results Variables Variation Explained p value Coefficient Estimates p value HAE 3.83 0.03 -0.20 0.001 mTPI 6.34 0.001 -0.22 0.001 Slope 4.3 0.001 0.04 0.192# Wind Speed 17.26 0.001 0.08 0.01 Wind Direction 7.24 0.001 NA NA preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 3, 2025. ; https://doi.org/10.1101/2025.09.01.673595doi: bioRxiv preprint 14

316 Despite global recognition of living root-bridges as remarkable examples of sustainable 317 livelihood solutions and green architecture, their underlying biological and ecological context 318 has remained little understood. To contribute to the existing knowledge, field surveys were 319 conducted across Meghalaya forests with reported (anecdotal reports were also considered) 320 natural population of F. elastica and root-bridges. Using ddRAD based SNP genotyping 321 accompanied with landscape genetic analyses, our study contributes to recent developments 322 in plant population genetics emphasizing the importance of assessing fsgs while 323 simultaneously accounting for environmental heterogeneity. 324 Short dispersal potential of F. elastica 325 As expected, we found genetic clusters within the small spatial scale of our study area 326 defined by landscape features. These clusters corresponded to E. Khasi, W. Jaintia, W. Khasi 327 and Ribhoi districts with no highly related individuals identified between two populations 328 (Supplementary figure 1). Interestingly, all the recaptures were limited to the area near root-329 bridge villages. This could be possibly due to the use of vegetative cuttings at the sites 330 selected for construction of bridges. Similar to perennial species, moderate Ho and negative 331 values of Fis were observed within each genetic subpopulation (Guiller et al., 2023, Huang et 332 al., 2023, Niskanen et al., 2024, Rojas‐Cortés et al., 2024). However, such patterns could also 333 arise due to random mating facilitated by the frequent reproductive asynchrony occurring at 334 shorter distances as indicated in our Mantel correlogram results. The negative correlation 335 between genetic relatedness and geographic distance shows that individuals at distances of 1–336 4km (varies across population) do not exchange genes. These results therefore suggest short 337 gene flow distance (via wasp pollination or seed disperser) resulting in observed fine scale 338 genetic structure. Such inferences were also reported in other fsgs studies on Ficus species 339 occurring in heterogeneous landscapes (Wang et al., 2009, Krishnan & Borges et al., 2018, 340 Deng et al., 2020, Nazareno et al., 2021). 341 Isolation by distance is not enough to explain fsgs in heterogenous landscape 342 Supplementing the population genomic data with landscape analyses showed how genetic 343 differentiation among these populations is a result of differential environmental response. For 344 example, the RDA results showed that wind parameters have more effect on overall genetic 345 variation whereas the conductance analysis highlighted landscape variables (i.e mTPI and 346 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 3, 2025. ; https://doi.org/10.1101/2025.09.01.673595doi: bioRxiv preprint 15 HAE representing valley and river catchments respectively) as barriers to the gene flow. 347 These complementary results follow the general pattern observed in gene flow studies 348 performed at larger landscape scales inferring the role of wind regime in Ficus reproductive 349 biology, as well as it resonates with the findings of habitat-specific studies dealing with 350 localised bioclimatic and landscape features (Robledo‐Arnuncio 2011, Cruzan & 351 Hendrickson 2020, Quinteros et al., 2024). However, it is important to address the presence 352 of a substantial amount of unexplained variance in our data which is often found in 353 genotype–environment association studies using neutral loci data (similar results were found 354 in other studies such as Gibson & Moyle 2020, Liu et al. 2023, Rojas‐Cortés et al., 2024). 355 This could also be the reason for observing significant IBD and IBT tests indicating spatial 356 autocorrelation as a key contributor to neutral genomic variation. 357 Landscape heterogeneity is the key for high genetic diversity 358 Interestingly, in all the analyses, E. Khasi consistently showed stark differences from other 359 areas. It is the most differentiated population (according to PCA, TESS and pairwise Fst 360 results) with the lowest Fis of -0.005 and signatures of both short- and long-distance dispersal. 361 Such contrasting results can be attributed to the landscape heterogeneity of this region, 362 especially higher occurrence of riverine valleys. For instance, in the RDA analysis the E. 363 Khasi samples are spread across two axes explained by slope and HAE along with mTPI. 364 Further, the conductance layer and EEMS highlight the riverine valleys as barriers to gene 365 flow within the E. Khasi area. To validate these outcomes, we did another TESS run (as 366 mentioned in the methods) within the Khasi population and found that it is further divided 367 into three lineages corresponding to the three river catchments in this region (Suuplementary 368 figure 2). This corroborates the significant positive Mantel correlogram results which suggest 369 long distance gene flow within the riverine valleys, but not across the valleys. Finally, to 370 confirm that such patterns are only confined to E. Khasi, we performed a TESS run within W. 371 Jaintia samples also because of similar topography features. However, the cross-entropy 372 criteria could not identify a true K for W. Jaintia, probably because the distribution of F. 373 elastica here is mostly found along a single river catchment (maximum elevation difference 374 of 300amsl). This is unlike E. Khasi where the landscape is divided into three river 375 catchments with elevational difference of 500-1000amsl, thus confirming our inference. 376 Scope of improvement 377 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 3, 2025. ; https://doi.org/10.1101/2025.09.01.673595doi: bioRxiv preprint 16 Finally, we believe integration of local Ficus demography and wasp associated genetic data 378 would have resulted in resolving the observed patterns into pollen vs seed mediated gene 379 flow. Although, seed dispersal distance is expected to be spatially limited from pollen 380 mediated ones, in case of hemi-epiphytes one can expect an opposite result owing to long 381 movement range of dispersers. Subsequently incorporation of age-class based sampling 382 (sapling vs strangler vs adult), would have given better understanding of diversity drivers by 383 examining the role of species ecology and its habitat on intra-species genetic variability. 384 Conservation Implication 385 Overall, our study supports the notion of environmental complexity facilitating the genetic 386 heterogeneity in Ficus spp. Our results clearly indicate that within Meghalaya, the high 387 genetically diverse population resides in E. Khasi district characterised by topographical 388 complexity. However, recent developmental projects in this district to facilitate high tourist 389 pressure to the living root-bridges has resulted in loss of forest cover in valleys. These 390 developments pose a direct threat to the biodiversity of this area. Considering that F. elastica 391 of this region is primarily associated with the riparian ecosystem, we believe that our 392 inference could be used to push towards better sustainable development plans. Since, the 393 living root-bridges are in process of being declared UNESCO world heritage site, our insight 394 into the importance of landscape heterogeneity as a key to ensure the viable population of F. 395 elastica in this region, would ensure some strict guidelines against rampant habitat 396 destruction. Particularly, we would like to propose that only the root-bridges present at the 397 southern border of Meghalaya should be considered for tourism as those areas are 398 economically developed. Also, the results of short dispersal distance could be used as a 399 guiding parameter to make green spaces around the living root-bridges (allowed for tourism) 400 facilitating gene flow. Finally, we believe our methodology can be implemented globally for 401 any plant taxon to investigate the genetic diversity patterns in context of its habitat. Although 402 inferences of such studies may be applicable at localised scale, in today’s world with climate 403 change and habitat conversion, the site-specific information is critical in guiding the 404 biodiversity management plans. 405 Funding 406 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 3, 2025. ; https://doi.org/10.1101/2025.09.01.673595doi: bioRxiv preprint 17 This work is funded by the Department of Biotechnology, Ministry of Science and 407 Technology Government of India under the project titled “The ecology and genetics of the 408 living root-bridges of Meghalaya [ BT/PR40532/FCB/125/78/2020]. 409 Acknowledgments 410 We are grateful to Mr. Morningstar Khongthaw, founder ofF Living bridge foundation (LBF) 411 for facilitating the field work as it requires permission from the headman of each village to 412 enter the adjoining forested area. We are also thankful to Syrwet u barim mariang jingkieng 413 jri cooperative federation ltd. (SUBMJJCFL), led by Kong Iora Dkhar and her team for 414 introducing us with the village representatives working towards conservation of root-bridges 415 in their respective areas. We thank Riban War, Lea Kuttkat, Jylliew and Bah Shekhon for all 416 their support during sample collection; Prof. Krishnan Upadhyay for helping in field logistics 417 arrangement. We thank P. Praveen and Mayuresh Gangal for their guidance in ddRAD 418 library preparation and data generation; Divyashree Rana for providing feedback and 419 suggestion in landscape analysis. We also acknowledge the support of NCBS NGS facility in 420 standardising the library preparation protocol. 421 Conflict of Interests 422 The authors share no conflict of interests. 423 Author Contribution 424 TG designed the research with input from RB. Sampling, data generation and analysis were 425 done by TG with inputs and discussions from RB and UR. The manuscript is written by TG 426 and majorly revised by RB with inputs taken fr om UR . RB and UR were responsible for 427 acquiring funding and logistics to ensure completion of the work. 428 Data Availability Statement 429 All the supplementary information (Table S1 and Figure S1) along with R code s is uploaded 430 in Zenodo (DOI: 10.5281/zenodo.17032732). The raw sequence data will be made publicly 431 accessible after the manuscript is accepted for publication. 432

433 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 3, 2025. ; https://doi.org/10.1101/2025.09.01.673595doi: bioRxiv preprint 18 Ahmed, Sophia, Stephen G. Compton, Roger K. Butlin, and Philip M. Gilmartin. "Wind-434 borne insects mediate directional pollen transfer between desert fig trees 160 kilometers 435 apart." Proceedings of the National Academy of Sciences 106, no. 48 (2009): 20342-20347. 436 Avise, J. C. 2000. Phylogeography: The History and Formation of Species. Harvard 437 University Press. 438 Bacles, C. F. E., and A. S. Jump. 2011. Taking a Tree's Perspective on Forest Fragmentation 439 Genetics. Trends in Plant Science 16, 13–18. 440 Bolger, A.M., Lohse, M. and Usadel, B., 2014. Trimmomatic: a flexible trimmer for Illumina 441 sequence data. Bioinformatics, 30(15), pp.2114-2120. 442 Bowman, J., Greenhorn, J.E., Marrotte, R.R., McKay, M.M., Morris, K.Y., Prentice, M.B. 443 and Wehtje, M., 2016. On applications of landscape genetics. Conservation Genetics, 17(4), 444 pp.753-760. 445 Capblancq, T. and Forester, B.R., 2021. Redundancy analysis: A Swiss Army Knife for 446 landscape genomics. Methods in Ecology and Evolution, 12(12), pp.2298-2309. 447 Caye, K., Deist, T.M., Martins, H., Michel, O. and François, O., 2016. TESS3: fast inference 448 of spatial population structure and genome scans for selection. Molecular Ecology Resources, 449 16(2), pp.540-548. 450 Chaudhuri, P., Bhattacharyya, S. and Samal, A.C., 2016. Living Root Bridge: A potential no 451 cost eco-technology for mitigating rural communication problems. Int. J. Exp. Res. Rev, 5, 452 pp.33-35. 453 Chaudhary, L.B., Sudhakar, J.V., Kumar, A., Bajpai, O., Tiwari, R. and Murthy, G.V.S., 454 2012. Synopsis of the genus Ficus L.(Moraceae) in India. Taiwania, 57(2), pp.193-216. 455 Corner, E.J.H., 1985. Ficus (Moraceae) and Hymenoptera (Chalcidoidea): figs and their 456 pollinators. Biological Journal of the Linnean Society, 25(2), pp.187-195. 457 Cruaud, A., Rønsted, N., Chantarasuwan, B., Chou, L.S., Clement, W.L., Couloux, A., 458 Cousins, B., Genson, G., Harrison, R.D., Hanson, P.E. and Hossaert-Mckey, M., 2012. An 459 extreme case of plant–insect codiversification: figs and fig-pollinating wasps. Systematic 460 biology, 61(6), pp.1029-1047. 461 Cruzan, M.B. and Hendrickson, E.C., 2020. Landscape genetics of plants: challenges and 462 opportunities. Plant Communications, 1(6). 463 Danecek, P., Auton, A., Abecasis, G., Albers, C.A., Banks, E., DePristo, M.A., Handsaker, 464 R.E., Lunter, G., Marth, G.T., Sherry, S.T. and McVean, G., 2011. The variant call format 465 and VCFtools. Bioinformatics, 27(15), pp.2156-2158. 466 Deng, J.Y., Van Noort, S., Compton, S.G., Chen, Y. and Greeff, J.M., 2020. Conservation 467 implications of fine scale population genetic structure of Ficus species in South African 468 forests. Forest Ecology and Management, 474, p.118387. 469 Deng, J.Y., Fu, R.H., Compton, S.G., Yuan, C., Kling, M.M., Chen, X.Y., Song, Y.B., Jiang, 470 K., Liu, M., Greeff, J.M. and Chen, Y., 2025. Wind Direction and Strength Determine the 471 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 3, 2025. ; https://doi.org/10.1101/2025.09.01.673595doi: bioRxiv preprint 19 Genetic Structure of an Insect‐Pollinated Plant Across Heterogeneous Landscape. Journal of 472 Biogeography, 52(6), p.e15119. 473 Dev, S.A., Kjellberg, F., Hossaert‐McKey, M. and Borges, R.M., 2011. Fine‐scale population 474 genetic structure of two dioecious Indian keystone species, Ficus hispida and Ficus 475 exasperata (Moraceae). Biotropica, 43(3), pp.309-316. 476 Dick, C. W., Bermingham, E., Lemes, M. R., & Gribel, R. 2007. Extreme long-distance 477 dispersal of the lowland tropical rainforest tree Ceiba pentandra L. (Malvaceae) in Africa and 478 the Neotropics. Molecular Ecology, 16, 3039–3049. 479 Dray, S. and Dufour, A.B., 2007. The ade4 package: implementing the duality diagram for 480 ecologists. Journal of statistical software, 22, pp.1-20. 481 Fenderson, L. E., A. I. Kovach, and B. Llamas. 2019. Spatiotemporal Landscape Genetics: 482 Investigating Ecology and Evolution Through Space and Time. Molecular Ecology 29, 218–483 246. 484 Gibson, M.J. and Moyle, L.C., 2020. Regional differences in the abiotic environment 485 contribute to genomic divergence within a wild tomato species. Molecular Ecology, 29(12), 486 pp.2204-2217. 487 Goudet, J., 2005. Hierfstat, a package for R to compute and test hierarchical F‐statistics. 488 Molecular ecology notes, 5(1), pp.184-186. 489 Guiller, A., Decocq, G., Kichey, T., Poli, P., Vandepitte, K., Dubois, F., Honnay, O. and 490 Closset-Kopp, D., 2023. Spatial genetic structure of two forest plant metapopulations in 491 dynamic agricultural landscapes. Landscape and Urban Planning, 231, p.104648. 492 Hamrick, J. L., Godt, M. J. W., & Sherman-Broyles, S. L. (1992). Factors influencing levels 493 of genetic diversity in woody plant species. New Forests, 6, 95–124. 494 Hao, G.Y., Sack, L., Wang, A.Y., Cao, K.F. and Goldstein, G., 2010. Differentiation of leaf 495 water flux and drought tolerance traits in hemiepiphytic and non‐hemiepiphytic Ficus tree 496 species. Functional Ecology, 24(4), pp.731-740. 497 Harrison, R. D. & Rasplus, J.-Y. Dispersal of fig pollinators in Asian tropical rain forests. 498 Journal of Tropical Ecology 22, 631–639 (2006). 499 Heer, K., Kalko, E.K., Albrecht, L., Garcia-Villacorta, R., Staeps, F.C., Herre, E.A. and Dick, 500 C.W., 2015. Spatial scales of genetic structure in free-standing and strangler figs (Ficus, 501 Moraceae) inhabiting Neotropical forests. PLoS One, 10(7), p.e0133581. 502 Hijmans, R.J., Van Etten, J., Cheng, J., Mattiuzzi, M., Sumner, M., Greenberg, J.A., 503 Lamigueiro, O.P., Bevan, A., Racine, E.B., Shortridge, A. and Hijmans, M.R.J., 2015. 504 Package ‘raster’. R package, 734, p.473. 505 Hodel, R. G. J., Knowles, L. L., Dunaway, J. F., Payton, A. C.,McDaniel, S. F., Soltis, P. S., 506 et al. 2018. Terrestrial species adapted to sea dispersal: differences in propagule dispersal of 507 two Caribbean mangroves. Molecular Ecology 27, 4612–4626. 508 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 3, 2025. ; https://doi.org/10.1101/2025.09.01.673595doi: bioRxiv preprint 20 Huang, J.F., Li, S.Q., Xu, R. and Peng, Y.Q., 2023. East‒West genetic differentiation across 509 the Indo-Burma hotspot: evidence from two closely related dioecious figs. BMC Plant 510 Biology, 23(1), p.321. 511 Janzen, D. H. (1979). How to be a fig. Annual Review of Ecology and Systematics, 10, 13–512 51. 513 Jombart, T., Kamvar, Z.N., Collins, C., Lustrik, R., Beugin, M.P., Knaus, B.J. and Jombart, 514 M.T., 2018. Package ‘adegenet’. Github Repository. 515 Kling, M. M., and D. D. Ackerly. 2021. “Global Wind Patterns Shape Genetic 516 Differentiation, Asymmetric Gene Flow, and Genetic Diversity in Trees.” Proceedings of the 517 National Academy of Sciences of the United States of America 118, no. 17: e2017317118. 518 Knaus, B.J. and Grünwald, N.J., 2017. vcfr: a package to manipulate and visualize variant 519 call format data in R. Molecular ecology resources, 17(1), pp.44-53. 520 Krishnan, A. and Borges, R.M., 2018. A fig tree in a concrete jungle: fine-scale population 521 genetic structure of the cluster fig Ficus racemosa in an urban environment. Urban 522 Ecosystems, 21(1), pp.171-181. 523 Li, H. and Durbin, R., 2009. Fast and accurate short read alignment with Burrows–Wheeler 524 transform. bioinformatics, 25(14), pp.1754-1760. 525 Li, H., Handsaker, B., Wysoker, A., Fennell, T., Ruan, J., Homer, N., Marth, G., Abecasis, 526 G., Durbin, R. and 1000 Genome Project Data Processing Subgroup, 2009. The sequence 527 alignment/map format and SAMtools. bioinformatics, 25(16), pp.2078-2079. 528 Liu, H.J., Li, M.H., Wang, Z., Wang, T. and Su, Y.J., 2024. Local adaptation and 529 demographic history of vulnerable conifer Cephalotaxus oliveri in southern China. Journal of 530 Systematics and Evolution, 62(3), pp.457-474. 531 Lomáscolo, S.B., Levey, D.J., Kimball, R.T., Bolker, B.M. and Alborn, H.T., 2010. 532 Dispersers shape fruit diversity in Ficus (Moraceae). Proceedings of the National Academy of 533 Sciences, 107(33), pp.14668-14672. 534 Ludwig, F., Middleton, W., Gallenmüller, F., Rogers, P. and Speck, T., 2019. Living bridges 535 using aerial roots of ficus elastica–an interdisciplinary perspective. Scientific reports, 9(1), 536 p.12226. 537 Manichaikul, A., Mychaleckyj, J.C., Rich, S.S., Daly, K., Sale, M. and Chen, W.M., 2010. 538 Robust relationship inference in genome-wide association studies. Bioinformatics, 26(22), 539 pp.2867-2873. 540 Mazal, L., Corenblit, D., Barsoum, N., Steiger, J., Skot, L., Fumanal, B. and Till-Bottraud, I., 541 2021. Fine-scale spatial genetic structure and intra-specific interactions of Populus nigra 542 within a natural river corridor along the lower Allier River (France). Flora, 275, p.151763. 543 Medina, I., Cooke, G. M., and Ord, T. J. (2018). Walk, swim or fly? Locomotor mode 544 predicts genetic differentiation in vertebrates. Ecology Letters 21, 638–645. 545 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 3, 2025. ; https://doi.org/10.1101/2025.09.01.673595doi: bioRxiv preprint 21 Nazareno, A. G., Alzate-Marin, A. L., & Pereira, R. A. S. (2013). Dioecy, more than 546 monoecy, affects plant spatial genetic structure: The case study of Ficus. Ecology and 547 Evolution, 10, 3495–3508. 548 Nazareno, A.G., Knowles, L.L., Dick, C.W. and Lohmann, L.G., 2021. By animal, water, or 549 wind: can dispersal mode predict genetic connectivity in riverine plant species?. Frontiers in 550 Plant Science, 12, p.626405. 551 Niskanen, A.K., Kujala, S.T., Kärkkäinen, K., Savolainen, O. and Pyhäjärvi, T., 2024. Does 552 the seed fall far from the tree? Weak fine-scale genetic structure in a continuous Scots pine 553 population. Peer Community Journal, 4. 554 Nongbri, L.B., Daniel, J.A. and Borges, R.M., 2021. Pollinating and non-pollinating fig 555 wasps (Hymenoptera: Chalcidoidea) from Ficus elastica, the living root bridge tree of 556 Meghalaya in northeast India. Current Science, 121(8), pp.1099-1106. 557 Oksanen, J., Simpson, G., Blanchet, F., Kindt, R., Legendre, P., Minchin, P., O’Hara, R., 558 Solymos, P., Stevens, M., Szoecs, E., Wagner, H., Barbour, M., Bedward, M., Bolker, B., 559 Borcard, D., Carvalho, G.,Chirico, M., De Caceres, M., Durand, D., … Weedon, J., 2022. 560 vegan: Community ecology package. R package version 2.6-4. https:// cran.r- project.org/ 561 package= vegan. 562 Pebesma, E., Bivand, R., Pebesma, M.E., RColorBrewer, S. and Collate, A.A.A., 2012. 563 Package ‘sp’. The Comprehensive R Archive Network, 9. 564 Pebesma, E., 2018. Simple features for R: standardized support for spatial vector data. 565 Peterman, W.E. and Pope, N.S., 2021. The use and misuse of regression models in landscape 566 genetic analyses. Molecular ecology, 30, pp. 37–47. 567 Petit, R.J., Duminil, J., Fineschi, S., Hampe, A., Salvini, D. and Vendramin, G.G., 2005. 568 Invited review: comparative organization of chloroplast, mitochondrial and nuclear diversity 569 in plant populations. Molecular ecology, 14(3), pp.689-701. 570 Petkova, D., Novembre, J. and Stephens, M., 2016. Visualizing spatial population structure 571 with estimated effective migration surfaces. Nature genetics, 48(1), pp.94-100. 572 Purcell, S., Neale, B., Todd-Brown, K., Thomas, L., Ferreira, M.A., Bender, D., Maller, J., 573 Sklar, P., De Bakker, P.I., Daly, M.J. and Sham, P.C., 2007. PLINK: a tool set for whole-574 genome association and population-based linkage analyses. The American journal of human 575 genetics, 81(3), pp.559-575. 576 Putz, F. E. & Holbrook, N. M. Strangler fig rooting habits and nutrient relations in the llanos 577 of Venezuela. American Journal of Botany 76, 781–788 (1989). 578 Quinteros, K., Piatscheck, F., Satler, J., Heath, T. and Nason, J., 2024. Comparative 579 Phylogeography of Mexican Rock Fig (Ficus petiolaris) and its Fig Wasp Pollinator: Effects 580 of Abiotic Versus Biotic Factors. Authorea Preprints. 581 Rasplus, J.Y., Rodriguez, L.J., Tollon-Cordet, C. and Cruaud, A., 2018. Revisiting the 582 phylogeny of Ficus (Moraceae): When Next Generation Sequencing corroborates Past 583 Generation Botanists. bioRxiv, p.340463. 584 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 3, 2025. ; https://doi.org/10.1101/2025.09.01.673595doi: bioRxiv preprint 22 Rojas‐Cortés, Á.P., Gasca‐Pineda, J., González‐Rodríguez, A. and Ibarra‐Manríquez, G., 585 2024. Genomic diversity and structure of a Neotropical microendemic fig tree. Ecology and 586 Evolution, 14(3), p.e11178. 587 Robledo‐Arnuncio, J.J., 2011. Wind pollination over mesoscale distances: an investigation 588 with Scots pine. New Phytologist, 190(1), pp.222-233. 589 Savary, P., Foltête, J.C., Moal, H. and Garnier, S., 2022. Combining landscape and genetic 590 graphs to address key issues in landscape genetics. Landscape Ecology, 37(9), pp.2293-2309. 591 Sharma, S.C. and Parthankar, N.C., 2016. Effect of Hill-cutting on plant community: A case 592 study in the subtropical forest of Raid Marwet region, Ri-Bhoi district of Meghalaya, India. 593 Journal of Environment and Natural Sciences, 10, pp.1-8. 594 Shukla, G., Biswas, R., Das, A.P. and Chakravarty, S., 2014. Plant diversity at Chilapatta 595 Reserve Forest of Terai Duars in sub-humid tropical foothills of Indian Eastern Himalayas. 596 Journal of Forestry Research, 25(3), pp.591-596. 597 Singh, S.B., Mishra, B.P. and Tripathi, S.K., 2015. Recovery of plant diversity and soil 598 nutrients during stand development in subtropical forests of Mizoram, Northeast India. 599 Biodiversitas Journal of Biological Diversity, 16(2). 600 van Strien, M.J., Keller, D., Holderegger, R., Ghazoul, J., Kienast, F. and Bolliger, J., 2014. 601 Landscape genetics as a tool for conservation planning: predicting the effects of landscape 602 change on gene flow. Ecological Applications, 24(2), pp.327-339. 603 Wang, R., Ai, B., Gao, B.Q., Yu, S., Li, Y.Y. and Chen, X.Y., 2009. Spatial genetic structure 604 and restricted gene flow in a functionally dioecious fig, Ficus pumila L. var. pumila 605 (Moraceae). Population Ecology, 51(2), pp.307-315. 606 Wang, I.J., 2020. Topographic path analysis for modelling dispersal and functional 607 connectivity: Calculating topographic distances using the topoDistance r package. Methods in 608 Ecology and Evolution, 11(2), pp.265-272. 609 Yang, L. Y., Machado, C. A., Dang, X. D., Peng, Y. Q., Yang, D. R., Zhang, D. Y., & Liao, 610 W. J. (2015). The incidence and pattern of copollinator diversification in dioecious and 611 monoecious figs. Evolution, 69, 294–304. 612 Yu, H.U.I., Nason, J.D., Ge, X. and Zeng, J., 2010. Slatkin’s Paradox: when direct 613 observation and realized gene flow disagree. A case study in Ficus. Molecular Ecology, 614 19(20), pp.4441-4453. 615 616 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted September 3, 2025. ; https://doi.org/10.1101/2025.09.01.673595doi: bioRxiv preprint