Environmental and transcriptomic determinants of drought response in critically endangered Siamese rosewood

43 This work is supported by funding to T.H.H., P.C., B.T., R.J., I.T., J.J.M. by the National 44 Geographic Society (EC-95234R-22). We would like to thank Kate A. Hardwick from the 45 Royal Botanic Gardens, Kew, for assistance in supplying samples from Thailand by V.C. via 46 the Millenium Seed Bank Partnership. T.H.H. is supported with a Croucher Fellowship. 47 T.H.H. wishes to personally thank the anonymous host of B612 for providing him a writing 48 retreat and food during the winter of 2023 that initiates this manuscript. 49 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 Page 3 of 33

50 51 Rosewoods account for up to 40% of the global illegal wildlife trade, with Dalbergia 52 cochinchinensis (Siamese rosewood) being the most heavily exploited species in Southeast 53 Asia. Its survival is further threatened by intensifying drought linked to climate change and 54 hydrological alteration. Here we combine greenhouse drought experiments across six 55 provenances with full-length cDNA-seq to uncover how water-relations and carbon-use 56 strategies vary within this species. Multivariate trait analysis resolves a two-dimensional 57 isohydry space, in which a water-flux stringency axis ( gs–E) is largely orthogonal to a 58 carbon-economics axis (A–WUEi). Provenances differed strikingly, where two (KKH and DN) 59 showed a rare E↓ A↑ response, achieving high WUE i and maintaining growth under drought. 60 Contrary to expectation, precipitation of the wettest month, not the driest, predicted isohydry, 61 indicating that wet-season conditions set a developmental and hydrological floor for later 62 drought responses. We identified 76 drought-responsive genes and two genes associated with 63 isohydry axes, SEOR1 and a poorly characterised Notch-like protein AT4G14746. We also 64 detected provenance-specific isoform switches, where drought favoured a loss-of-function 65 PRX52 isoform lacking its signal peptide in the anisohydric provenance THB, and gain-of-66 function isoforms of ANN3 and LTPG5 in NP. These results reveal previously hidden 67 diversity in drought strategies, identify mechanism-related markers for screening, and provide 68 a simple climatic lever for climate-adjusted provenancing. We reveal post-transcriptional 69 regulation as a novel candidate substrate for local adaptation in a threatened tropical tree, 70 directly linking ecophysiology, climate, and genomics for conservation and restoration. 71 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 Page 4 of 33

72 Extreme droughts have emerged as a driver of accelerated forest mortality globally 1,2, 73 which threatens terrestrial biodiversity, climate forcing, and resource availability 3. 74 Mechanistically, tree mortality reflects interacting processes such as xylem hydraulic failure, 75 carbon depletion from prolonged stomatal closure, and heightened susceptibility to pests and 76 pathogens, whose prevalences are expected to increase as warming intensifies atmospheric 77 water demand 4. The IPCC’s Sixth Assessment Report concludes that many regions have 78 already experienced anthropogenically influenced increases in agricultural and ecological 79 drought, consistent with observed forest declines during recent heat-drought events 5. Many 80 tree species have evolved in response to spatial and temporal variability in water availability 81 through local adaptation and phenotypic plasticity6, but our understanding of these responses 82 must be improved to predict the impacts of environmental change and to safeguard 83 biodiversity. 84 Southeast Asia is a major biodiversity hotspot with disproportionally high levels of 85 endemism, which is experiencing unprecedented levels of drought threats due to climate and 86 land use changes. It also has the highest proportion of vascular plants, reptiles, birds, and 87 mammals classified as in the IUCN Red List7. Its main water body, the Mekong River Basin, 88 has had record low flows and severe multi-year drought in the recent decade, with 89 widespread soil-moisture deficits, delayed monsoons and saline intrusion in the delta 8. 90 Drought impacts are particularly amplified by strong climate variability, such as the 2015–91 2016 El Niño that brings exceptional heat and rainfall deficits, with recurrent moisture 92 shortfalls following the subsequent years 9. The extreme drought affected area has doubled 93 since 1950s and both the intensity and frequency of drought risks will continue to increase in 94 all scenarios except for the lowest emission pathway 10. Simultaneously, these climatic 95 stresses now interact with one of the world’s fastest hydropower build-outs11, with more than 96 1,000 dams already in place with further expansion planned 12. The large storage of water 97 behind dams during the wet season exacerbates the downstream impacts on water 98 availability8. Together these climate and infrastructure drivers are reshaping hydro-ecological 99 regimes across Southeast Asia and underlining the need to understand adaptive drought 100 responses for the region’s forests. 101 Dalbergia cochinchinensis Pierre (Siamese rosewood) produces extremely valuable 102 rosewood timber in the Mekong Region, and is endemic to Cambodia, Laos, Thailand, and 103 Vietnam. Rosewoods (Dalbergia spp.) amount to 30–40% of worldwide illegal wildlife trade, 104 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 Page 5 of 33 which is valued in total at USD 7–23 billion annually, making it the target of the world’s 105 largest wildlife crime 13. Siamese rosewood was assessed as ‘Critically Endangered’ in 2022 106 due to an estimated ≥ 90% population decline and population fragmentation caused by illegal 107 logging and habitat loss14. The species is on CITES Appendix II since 2013, but cross-border 108 trafficking has continued to be a significant threat, especially in Cambodia, Laos and 109 Vietnam14. In addition to overexploitation, habitat conversion, fire and overgrazing, 110 vulnerability analyses further indicate that a large share of the species range is exposed to 111 medium to very high combined pressures, and a measurable fraction to climate-change risk 112 by mid-century15. Conservation actions initiated in the early 2000s included in situ and ex situ 113 conservation stands and seed production areas, but they were limited in scale, usually with < 114 50 seed-producing trees per country 16,17. Renewed efforts to conserve the remnant 115 populations and their genetic diversity since the 2010s have involved the collection of genetic 116 materials, development of tree nurseries, and generation of value chains to incentivise local 117 livelihoods18,19. 118 With increasing drought risks in Mekong Region, our ability to safeguard the survival 119 and conservation tree species such as Siamese rosewood requires a better understanding of 120 the variation in drought resistance and growth performance. 121 A useful lens to help fill this gap is the isohydry-anisohydry continuum. Isohydric 122 plants stabilise leaf water potential ( Ψ leaf) by closing stomata early as soils dry, while 123 anisohydric plants allow Ψ leaf to decline to maintain carbon assimilation. We previously 124 reported that Siamese rosewood had an anisohydric response to short-term drought, which 125 maximised assimilation at the cost of water loss, whereas a isohydric response was found in 126 the sympatric species D. oliveri20. While this anisohydric response can sustain photosynthesis 127 during drought, it increases hydraulic risk and may lead to mortality under prolonged water 128 deficits21. 129 Identifying the molecular mechanisms and key genes in drought will provide 130 additional insights for the physiological responses. Recent population genetic studies have 131 shown that Siamese rosewood is predominantly outcrossing and thus retains high levels of 132 genetic diversity across much of its range despite severe demographic declines 22. Central 133 populations in Cambodia and eastern Thailand harbour the highest allelic richness, whereas 134 peripheral and heavily exploited stands in northeastern Thailand, Laos, and Vietnam tend to 135 exhibit reduced diversity and higher relatedness, consistent with historical bottlenecks, 136 habitat fragmentation, and local exploitation 23. In addition, our previous genomic scan has 137 identified substantial genetic differentiation in D. cochinchinensis driven by temperature- and 138 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 Page 6 of 33 precipitation- related environmental factors 24. Local adaptation is also found to be the 139 strongest in the periphery of the species range 24. Thus, it is very plausible that populations 140 will display differentiated response to drought. However, it has also been shown that 141 environment has a stronger effect on gene expression than standing genetic variation 25. The 142 gene expression or transcriptomic response can be considered an intermediate phenotype that 143 is informative of high-level physiological traits 26. It is well-established that plants remodel 144 their transcriptome under drought stress, which regulates stress sensors, signalling pathways, 145 and synthesis of hormones and enzymes27,28. However, fundamental research in forest trees is 146 still underrepresented compared to other plant crops 29, and systematic studies at population 147 level are scarce30. 148 The endangered status of and the drought threats facing Siamese rosewood together 149 highlight the need to improve our understanding of its drought tolerance variability and 150 adaptation across the remnant populations. The overarching aim of this study is to identify 151 the environmental, physiological, and transcriptomic determinants of the species-wide 152 heterogeneity in drought response, by building on the recent capacity in genomic research in 153 Siamese rosewood, including its high-quality reference genome and range-wide genomic 154 scan24,31. Seedling recruitment is a critical bottleneck in Siamese rosewood 32 and early 155 phenotyping can still provide valuable insights into drought adaptation. First, we characterise 156 the drought response in seedlings from six provenances by comparing various physiological 157 traits in a greenhouse drought experiment. Second, we identify genes that are differentially 158 expressed among different provenances under drought stress. Third, we analyse if different 159 provenances have differential transcript usage in response to drought. This study will 160 ultimately ensure drought resilience in germplasm of Siamese rosewood by integrating recent 161 developments in conservation and cutting-edge genomic technologies. 162 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 Page 7 of 33

163 Plant materials 164 Dried seeds of Dalbergia cochinchinensis were provided by the Institute of Forest and 165 Wildlife Research and Development, Cambodia; the National Agriculture and Forestry 166 Research Institute, Laos; and the Department of National Park, Wildlife and Plant 167 Conservation, Thailand in 2020 from six local seed sources across the species range ( Figure 168 1a and Supplementary Table 1 ). The seeds were scarified by placing them in 70°C distilled 169 water and left to cool to room temperature overnight. They were germinated on 1% agar in a 170 controlled greenhouse at 30°C and photoperiod 12L/12D. Germinants were first transferred 171 to 0.125L pots in a soil-perlite 3:1 (v:v) mixture and grown for three months. Healthy 172 seedlings were then transferred to 0.81 L pots with the same substrate and grown for further 173 two months. Throughout this period, plants were watered regularly to maintain at substrate 174 capacity and fertilized once a week using N /i5 P/i5 K 20:20:20 fertilizer (Chempak, Suffolk, 175 United Kingdom). 176 177 Experimental design 178 We used a split-plot design with 10 plants per provenance randomly distributed into 179 10 trays, making up to 60 plants in total. We randomly assigned half of the trays to either 180 well-watered control (C) or water-withholding treatment (D). The controls were watered 181 every other day to maintain substrate capacity at 40–55%, whereas the droughted were not 182 watered at all after the experiment started. We also randomly split the trays into two blocks 183 with the start of experiment staggered by one day. 184 There were two types of data collected: (1) continuous data of water relation and 185 photosynthetic measurements, and (2) end-point data of anatomical traits and biochemical 186 measurements. At the end of the experiment (day 14), three leaves were sampled from each 187 plant, snap-frozen in liquid nitrogen, and stored at –80°C. 188 189 Water relation and photosynthetic measurements 190 Measurements were taken between 10 am and 2 pm (at least two hours after sunrise 191 and two hours before sunset), when photosynthetic activity reached its plateau (i.e. point of 192 saturation). We measured soil water content ( SWC) by using a ML3 ThetaProbe Soil 193 Moisture Sensor (Delta-T Devices Ltd., Cambridge, England). We measured stomatal 194 conductance (g s), photosynthetic assimilation rate ( A), and transpiration rate ( E) using an 195 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 Page 8 of 33 infrared gas analyser LCpro T, Leaf Chamber & Soil Respiration System (SRS2000 T, ADC 196 BioScientific Limited, England) and its broad leaf chamber, with the light intensity set to 197 PAR 500 with equal parts of red, green, and blue lights. We calculated instantaneous water 198 use efficiency (WUEi) using the following equation33: 199 /g1849/g1847/g1831 /g3036/g3404 /g1827 /g1831 200 Anatomical traits 201 We recorded changes in height ( Δ Height), leaf number (Δ Leaf), and branch number 202 (Δ Branch) before and after the experiment. We measured leaf area using the equation for the 203 area of an ellipse ( S = ab π ). We measured fresh ( FW) and dry weights ( DW) of the leaves 204 before and after lyophilisation of two days in an Alpha 2-4LD-2 laboratory freeze-dryer 205 (Martin Christ GmbH, Germany). We calculated the leaf dry matter content ( LDMC) and 206 specific leaf area (SLA) using the following equations34: 207 /g1838/g1830/g1839/g1829 /g3404 /g1830/g1849 /g4666mg/g4667 /g1832/g1849 /g4666 mg/g4667 /g1845/g1838/g1827 /g3404 /g1845 /g4666c m /g2870/g4667 /g1830/g1849 /g4666mg/g4667 208 Pigment extraction and quantification 209 We ground ~25 mg of lyophilised leaves with a TissueLyzer (Retsch, Germany) at 210 25–1 s for 1 minute. We then extracted the pigments by adding an 80:20 mixture (v:v) of cold 211 acetone and 50 mM Tris buffer pH 8.0 and incubated for 72 hours, following Sims & 212 Gamon’s protocols35. After centrifugation, we transferred the supernatant to a 15ml Falcon 213 tube and then topped the extracts up with 6–12 ml of the acetone-Tris buffer to ensure that 214 the absorbances below were approximately in the range between 0 and 1. We used a 215 Nanodrop One (Thermo Fisher Scientific, Waltham, Massachusetts, USA) to measure the 216 absorbances of the extracts containing the leaves at 470, 537, 647, and 663 nm. We 217 determined the concentrations of anthocyanin (Ac ), chlorophyll a ( Chla) and b ( Chlb), and 218 carotenoids (Carot) with the following equations: 219 /g1827/g1855 /g3404 0.08173A /g2873/g2871/g2875/g3398 0.00697A /g2874/g2872/g2875/g3398 0.002228A /g2874/g2874/g2871 /g1829/g1860/g1864 /g3028/g3404 0.01373A /g2874/g2874/g2871/g3398 0.000897A /g2873/g2871/g2875/g3398 0.003046A /g2874/g2872/g2875 /g1829/g1860/g1864 /g3029/g3404 0.02405A /g2874/g2872/g2875/g3398 0.004305A /g2873/g2871/g2875/g3398 0.005507A /g2874/g2874/g2871 /g1829/g1853/g1870/g1867/g1872 /g3404 /g4670/g1827 /g2872/g2875/g2868/g3398 /g466617.1 /g3400 /g1829/g1860/g1864 /g3028/g3397/g1829 /g1860 /g1864 /g3029/g4667 /g3398 9.479 /g3400 /g1827/g1855/g4667/g4671 /g3400 119.26 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 Page 9 of 33 220 Sugar extraction and quantification 221 We heated ~25 mg of lyophilised leaves in 80% ethanol for one hour at 80°C. After 222 centrifugation, we determined glucose concentration from the supernatant using Osaki’s 223 anthrone method36. We mixed the supernatant with anthrone and sulfuric acid and heated it at 224 100°C for 10 minutes sharply. We cooled the mixture on ice and measured their absorbances 225 at 625 nm using a Genesys 150 UV-Visible spectrophotometer (Thermo Fisher Scientific, 226 Waltham, Massachusetts, USA). We calculated the concentration of total soluble sugars (TSS) 227 in the samples according to a D-glucose standard curve. 228 We washed the pellets in 80% ethanol to remove any glucose and then heated them at 229 100°C for 10 minutes with 30% perchloric acid to convert all starch to glucose. We again 230 measured absorbances of the extract at 625 nm again and used a D-glucose standard curve to 231 determine the concentration of starch (Starch). 232 233 Statistical analyses on traits 234 All statistical analyses were conducted in R 4.5.1. List of all traits included in this 235 analysis is presented as Supplementary Table 2. 236 We applied square-root transformation to g s, A, E, and WUE i and logarithmic 237 transformation to Ac, Chla, Chlb, and Carot to correct for normality. We assessed normality 238 based on the distribution of the residuals using a QQ plot. For continuous data, we conducted 239 two-way ANOVAs to examine the effects of SWC, provenance, and their interaction term on 240 gs, A, E, and WUE i. For end-point data, we conducted two-way ANOVAs to examine the 241 effects of treatment, provenance, and their interaction term anatomical and biomass traits ( Δ 242 Height, Δ Leaf, Δ Branch, LDMA, SLA), pigments (Ac , Chla, Chlb, and Carot), and sugars 243 (TSS and Starch). We initially included the blocks as a random variable in the ANOVAs, but 244 no difference was found thus the random variable was excluded from the models. 245 246 Quantification and climatic associations of isohydry 247 We quantified the provenance-level isohydry based on the principal component 248 analysis (PCA) of the coefficients of interaction effect of treatment × provenance on 249 physiological traits that show significant interactions. We then quantified individual-level 250 isohydry also based on physiological traits that show significant interactions, with gs, A, E, 251 and WUEi divided by SWC to obtain a single-value slope for each individual. Visual 252 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 Page 10 of 33 inspection confirmed that the loading of treatment aligned with the direction of second 253 principal component (PC2), thus the first principal component (PC1) was used as a composite 254 proxy for isohydry (denoted as Isohydry PC1 in this manuscript). 255 Provenance-level PCR is useful in quick visualisation and quantification of isohydry 256 at whole-provenance level, which is more spatially explicit and useful in conservation context. 257 However, individual-level PCR reveals a more substantial variation in isohydry among 258 individual trees that are not attributable to provenances, and thus is more suitable for 259 subsequent analyses on environmental and transcriptomic associations. 260 We tested the effect of 19 climatic variables that were biologically meaningful, from 261 the WorldClim 2 database ( bio_1–bio_19). To correct for the multicollinearity inherent in 262 climatic data, we used a stepwise approach to remove climatic variables that have a variance 263 inflation factor (VIF) > 10. We then examined the effects of treatment, these climatic 264 variables, and their interactions on both Isohydry PC1 and PC2, using multivariate analysis of 265 variance (MANOVA). 266 267 RNA extraction, full-length cDNA library construction, and sequencing 268 We isolated total RNA from leaves using the Monarch Total RNA Miniprep Kit (New 269 England Biolabs, United Kingdom). We determined their quantity on a Qubit 4 Fluorometer 270 (Thermo Fisher Scientific, United Kingdom), assessed their purity using a NanoDrop One 271 Spectrophotometer (Thermo Fisher Scientific), with A260/280 and A260/230 above 1.80, and 272 verified their integrity on a 1% bleach-agarose gel. 273 The 6-µl reverse transcription reaction contained 3 µl of ~200 ng total RNA, 2 µl of 274 10 µM Reverse Transcription Primer (5’–AGCAGTGGTATCAACGCAGAGTAC(T) 30V–3’) 275 and 1 µl of 10 mM dNTP. The reaction was incubated at 70°C for 5 min and held on ice to 276 allow the primer to anneal to mRNA with poly(A) tail. cDNA synthesis was performed by 277 adding 2.5 µl of Template Switching RT Buffer, 1 µl of Template Switching RT Enzyme 278 Mix (#M0466, New England Biolabs, United Kingdom), and 0.5 µl of 75 μ M Template 279 Switching Oligo (TSO) (5’–280 GCTAATCATTGCAAGCAGTGGTATCAACGCAGAGTACATrGrGrG–3’) to the 6-µl 281 primer-annealed reaction. The reaction was incubated at 42°C for 90 min, 85°C for 5 min, 282 and held at 4°C. Full-length cDNA (fl-cDNA) was amplified by adding 12.5 µl of Q5 Hot-283 Start High-Fidelity 2X Master Mix (New England Biolabs, United Kingdom), 10 µl of 284 nuclease-free water, and 1 µl of 10 µM cDNA PCR Primer (5’–285 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 Page 11 of 33 AAGCAGTGGTATCAACGCAGAGT–3’) to the 10 µl cDNA product. The thermal cycling 286 profile was 98°C 45 s, 20× [98°C 10 s, 62°C 15 s, 72°C 3 min], 72°C 5 min, for yielding ~1 287 µg fl-cDNA product. We cleaned up the fl-cDNA using 1.2× AMPure XP (Beckman Coulter, 288 United States). 289 Nanopore libraries were constructed with the ligation sequencing chemistry using 290 ~200 fmol pooled library (~250 ng for 2,000 bp c DNA). Nanopore libraries were then 291 sequenced and basecalled using the super-accuracy model on a GridION system (Oxford 292 Nanopore Technologies, United Kingdom) at the Department of Biology, University of 293 Oxford. 294 Basecalled reads were trimmed for Nanopore adaptors, the primer sequences, and 295 split for chimeras using dorado 0.6.2+14a7067. We then identified and quantified known and 296 novel transcripts using IsoQuant 3.4.137 and minimap 2.28-r1209, supplied with the reference 297 genome and gene annotation of D. cochinchinensis (Dacoc 1.2). We also extracted the 298 transcript sequences from the transcript model generated by IsoQuant with gffread 0.12.7. 299 300 Differential gene expression analysis 301 We imported the gene-level count data from IsoQuant into R 4.4.1 and performed the 302 analysis with DESeq2 1.44.0 38. We removed low-expression genes where there were less 303 than 6 samples (the size of the smallest experimental unit) with normalised counts greater 304 than or equal to 10. We conducted likelihood ratio tests to compare a full model which 305 accounts for provenance, ~ Treatment + Provenance + PC1 + PC2 + Treatment:PC1 + 306 Treatment:PC2, with a reduced model in which the effect of interest is removed. First, we 307 tested the main effect of drought treatment on gene expression ( the drought effect ) with a 308 reduced model of ~ Provenance + PC1 + PC2 + Treatment:PC1 + Treatment:PC2. Second, we 309 tested the differential effect of treatment on gene expression in isohydry (the isohydry effect), 310 as represented by the two principal axes (PC1 and PC2) using two LRTs with reduced models 311 of ~ Treatment + Provenance + PC1 + PC2 + Treatment:PC2 and ~ Treatment + Provenance 312 + PC1 + PC2 + Treatment:PC1 for PC1 and PC2 respectively. For both tests, we applied 313 independent hypothesis weighting, which could increase detection power in genome-scale 314 multiple testing 39, with an FDR threshold of 0.05 to discover significantly differentially 315 expressed genes (DEG). 316 We conducted gene set enrichment analyses (GSEA) on the effect size, which is the 317 χ 2-statistics in LRT, to search for Gene Ontology (GO) terms and Kyoto Encyclopedia of 318 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 Page 12 of 33 Genes and Genomes (KEGG) pathways that are significantly enriched using clusterProfiler 319 4.12.040 and fgsea 1.30.041. 320 321 Differential transcript usage, annotations, and functional consequences 322 We analysed differential transcript usage with the pipeline IsoformSwitchAnalyzeR 323 2.8.042, which incorporated DEXSeq 1.55.1 43. First, we assessed isoform switching between 324 control and drought-treated individuals, and set provenances as the batch effect ( the overall 325 drought effect). Second, we assessed isoform switching between control and drought-treated 326 individuals for each provenance ( the provenance effect ). We determined significant 327 differential transcript usage as those with a difference in isoform usage (dIF) > 0.01 and an 328 isoform switch Q-value < 0.05. Individual-level isohydry could be not used as a factor, like in 329 the case of differential gene expression analysis, because continuous variables are not 330 compatible with existing differential transcript usage analysis pipelines. 331 We annotated isoforms exhibiting significant differential transcript usage using a suite 332 of complementary tools. Coding potential was assessed with CPC2 1.0.144. Conserved protein 333 domains were identified using pfam_scan.pl on Pfam database version 38.045. Signal peptides 334 were predicted with SignalP 6.0 46. Subcellular localisation was inferred using DeepLoc 2.0 47 335 in Accurate mode. Transmembrane helices were predicted with DeepTMHMM 1.0.44 48. 336 Intrinsically disordered protein regions were identified using AIUPred v2.1.2 49 (aka IUPred 337 3). These annotations were integrated within the IsoformSwitchAnalyzeR framework to 338 evaluate the functional consequences of isoform switching events, including potential 339 changes in coding potential, protein domain architecture, secretion signals, membrane 340 localisation, and structural disorder. 341 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 Page 13 of 33

342 Variable responses in water relation and photosynthetic traits 343 All four water relation and photosynthetic traits varied significantly in response to the 344 changes in soil water content (SWC) caused by the drought treatment. We also observed 345 provenances effects and provenance x SWC interactions for several of the same traits. 346 Stomatal conductance (g s) decreased with declining SWC by a coefficient of –8.063e-05 347 (F1,408 = 4.11, p = 0.043), with significant variation among provenances (F 5,408 = 2.93, p = 348 0.013) and a significant SWC × provenance interaction (F 5,408 = 2.74, p = 0.019). Positive 349 slopes (higher gs with higher SWC) were observed in THA (1.20e-03), NP (1.03e-03), and PT 350 (1.01e-03) , while negative slopes (lower gs with higher SWC) were found in KKH (–5.52e-351 05), DN (–8.06e-05), and THB (–7.73e-04) (Figure 3a–b). Photosynthetic assimilation rate (A) 352 also showed significant effects of SWC (F 1,408 = 5.39, p = 0.021) and provenance (F 5,408 = 353 4.47, p < 0.001), as well as a strong SWC × provenance interaction (F 5,408 = 4.97, p < 0.001). 354 THA (0.0037) and NP (0.00051) had positive slopes, while PT (–0.0014), THB (–0.0018), 355 DN (–0.0038), and KKH (–0.010) had negative slopes ( Figure 3c–d). Transpiration rate ( E) 356 was strongly influenced by SWC (F 1,408 = 14.52, p < 0.001) and provenance (F 5,408 = 4.14, p 357 = 0.0011), while the SWC × provenance interaction was nearly significant (F 5,408 = 2.06, p = 358 0.069). Five of the provenances produced positive slopes PT (0.0060), THA (0.0059), NP 359 (0.0051), KKH (0.0020), and DN (0.0014) had and a negative slope was only observed in 360 THB (–0.0014) ( Figure 3 e–f). Instantaneous water-use efficiency (WUE i) increased 361 significantly under lower SWC (F 1,408 = 19.82, p < 0.001), with a significant SWC × 362 provenance interaction (F 5,408 = 4.23, p < 0.001) but no main effect of provenance. 363 Accordingly, most of the provenances had negative slopes, while THA (–0.0012), NP (–364 0.0029), PT (–0.0057), DN (–0.0065), and KKH (–0.015), while a positive slope was found 365 for THB (0.00057) (Figure 3g–h). 366 367 Anatomical and biochemical traits 368 Height change ( Δ Height) showed a significant treatment × provenance interaction 369 (F5,48 = 3.06, p = 0.0177), although treatment (F1,48 = 1.47, p = 0.23) and provenance (F5,48 = 370 1.34, p = 0.26) alone were not significant. Final height was higher in drought-treated 371 individuals in KKH (+1.2 cm), DN (+0.8), and NP (+0.2), and lower in drought-treated 372 individuals in THB (–0.8), PT (–1.2), and THA (–2.6) ( Figure 4 a–b). Leaf ( Δ Leaf) and 373 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 Page 14 of 33 branch ( Δ Branch) number changes were unaffected by treatment, provenance, or their 374 interaction (p > 0.05) (Supplementary Figure 1a–b). 375 Chlorophyll a content (Chla) showed significant effects of provenance (F5,48 = 2.65, p 376 = 0.0342) and a significant treatment × provenance interaction (F5,48 = 3.14, p = 0.0156). Chla 377 was higher in drought-treated individuals in KKH (+1.94), DN (+1.50), THB (+1.48), and 378 lower in THA (–0.20), NP (–0.73), and PT (–0.91) ( Figure 4c–d). No significant effects were 379 detected for anthocyanin (Ac ), chlorophyll b ( Chlb), or carotenoids (Carot ), although the 380 provenance effect on carotenoids was marginal (F 5,48 = 1.99, p = 0.0966) ( Supplementary 381 Figure 1c–e). 382 Neither leaf dry matter content (LDMC) nor specific leaf area (SLA) was significantly 383 affected by treatment, provenance, or their interaction, although SLA showed a marginal 384 treatment effect (F1,48 = 2.99, p = 0.0904) (Supplementary Figure 1f–g). 385 Neither total soluble sugars ( TSS) nor starch ( Starch) concentrations differed 386 significantly among treatments, provenances, or their interactions ( p > 0.05) (Supplementary 387 Figure 1h–i). 388 All ANOVA tables were summarised in Supplementary Table 3. 389 390 Characterisation and environmental association of isohydry 391 We characterised provenance- and individual-level isohydry using the first two 392 principal components (PC1 and PC2) that summarise the water relations and photosynthetic 393 traits. The provenance-level two-dimensional isohydry space captured 67.58 + 32.05 = 394 99.63% of the variation in the coefficients of effect of SWC × Provenance on the water 395 relations and photosynthetic traits (Figure 1b). PC1 co-varied largely with A and WUEi, while 396 PC2 co-varied largely with gs and E. The co-direction of loadings suggested that gs and E 397 were positively correlated, and the same between A and WUEi. However, the orthogonality 398 suggested that gs–E and A–WUEi were largely independent from each other. 399 The individual-level two-dimensional isohydry space captured 55.18 + 38.94 = 400 94.12% of the variation in water relations and photosynthetic traits (Figure 1c). PC1 co-varied 401 largely with gs and E, while PC2 co-varied largely with A and WUEi. Similarly, gs–E and A–402 WUEi were largely independent from each other. 403 Bioclimatic variables were largely inter-correlated and after filtering, only four 404 variables were retained in the model (VIF < 10), namely precipitation of wettest month 405 (bio_13), precipitation of driest month ( bio_14), and precipitation seasonality (bio_15). Only 406 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 Page 15 of 33 precipitation of wettest month (bio_13) showed a significant treatment × isohydry interaction 407 (F2,51 = 3.20, p = 0.049) ( Figure 1d–e ). Higher precipitation of the wettest month led to a 408 higher isohydry score. 409 410 Differentially expressed genes for drought response and isohydry 411 Full-length transcriptome sequencing yielded an average of 2.40 Gb (SD ± 0.85) for 412 36 individuals. The mean read length was 729.20 bp and N50 was 952.33 bp. 413 For the drought effect , we detected 76 genes that were significantly differentially 414 expressed ( Figure 5a and b and Supplementary Table 4 ). The most statistically significant, 415 annotated genes were GASA14 (Dacoc22547), CYP77A4 (Dacoc12206), PME1 416 (Dacoc00037), DEG11 (Dacoc04747), CXE20 (Dacoc12653), RDUF2 (Dacoc26635), XTH9 417 (Dacoc27022), and THE1 (Dacoc27311). There were 21 drought-response genes that were 418 unannotated. Five gene ontology terms were enriched, including Golgi cis cisterna 419 (GO:0000137) (q = 0.0017), cell wall organization or biogenesis (GO:0071554) ( q = 0.016), 420 small molecule biosynthetic process (GO:0044283) ( q = 0.016), external encapsulating 421 structure organization (GO:0045229) ( q = 0.019), and cell wall organization (GO:0071555) 422 (q = 0.025) (Supplementary Table 5). 423 For the isohydry effect , we only detected two genes that were significantly 424 differentially expressed. One strongly associated with PC1 ( gs–E) was Dacoc11700 425 (AT4G14746) ( q = 0.025, Figure 5c ), that encodes a neurogenic locus notch-like protein. 426 Higher expression of Dacoc11700 led to a higher PC1 score, which implies steeper gs–E 427 slopes and thus early stomatal closure, a traditional isohydric response ( Figure 5d ). 428 Dacoc11700 was predicted to contain a signal peptide and locate on the outside of cell 429 membrane. The gene strongly associated with PC2 (A–WUE i) was Dacoc11936 (SEOR1, 430 Sieve Element Occlusion Related 1) (q = 0.00092, Figure 5 e), that encodes a protein 431 primarily located within the sieve elements in the phloem. Similarly, higher expression of 432 SEOR1 led to a higher PC2 score (Figure 5f), which implies steeper A–WUEi slopes and thus 433 decreased photosynthetic efficiency, also a traditional isohydric response. 434 435 Differential transcript usage 436 We detected significant provenance-specific isoform switches between well-watered 437 controls and drought-treated individuals for two provenances NP and THB, but not any 438 overall effect of drought treatment or provenance alone. 439 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 Page 16 of 33 For THB, we detected 2 isoform switches. Dacoc21458 (PRX52, Peroxidase 52) 440 showed a significant isoform switch (q = 0.020) between a full functional isoform RB and an 441 truncated isoform transcript7028 which had an alternative transcription start site with a 442 missing signal peptide ( Figure 6a ). The isoform fraction of RB decreased from 98.89% to 443 41.60% between control and drought, whereas that of transcript7028 increased from 1.11% to 444 58.40%, constituting a change of 57.29% ( Figure 6b ). We also discovered a novel and 445 previously unannotated gene 16703 on chromosome 7 that showed a significant isoform 446 switch (q = 4.72e-15) (Supplementary Figure 3a), which had 4 non-coding isoforms of varied 447 lengths and no domain or topology could be predicted. All 4 isoforms had relatively similar 448 abundance in control, but the longest isoform transcript16686 dominated 94.30% of the 449 abundance in drought, constituting a change in isoform fraction of 80.06% ( Supplementary 450 Figure 3b). 451 For NP, we detected 2 further isoform switches. Dacoc26112 (ANN3, Annexin 3) 452 showed a significant switch (q = 0.0056) between a full functional isoform RA and three 453 other alternative isoforms ( Supplementary Figure 4a ). The isoform fraction of RA increased 454 from 10.22% to 45.23% between control and drought, whereas the non-coding isoform 455 transcript41999 decreased from 67.18% to 8.03% ( Supplementary Figure 4b ). Another gene 456 Dacoc06197 (LTPG5, Non-specific lipid transfer protein GPI-anchored 5) switched between 457 a full functional isoform RA and an alternative isoform transcript43524 with an earlier start 458 site ( q = 0.034) ( Supplementary Figure 5a ). The isoform fraction of RA increased from 459 55.62% to 94.59% between control and drought, whereas that of transcript43524 decreased 460 from 44.38% to 5.41%, constituting a change of 38.98% (Supplementary Figure 5b). 461 However most importantly, these genes have no significant difference in gene-level 462 expression between control and treatment, and would not have been detected in the 463 differential gene expression analysis ( Figure 6c , Supplementary Figure 3c , Supplementary 464 Figure 4c, and Supplementary Figure 5c ). Therefore, isoform switches are the sole mechanism 465 that regulates the response to drought in these genes. 466 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 Page 17 of 33

467 Drought response and isohydry in Siamese rosewood 468 Decreased soil water content (SWC) imposes significant constraints on gs in Siamese 469 rosewood, but the magnitude and direction of responses are provenance-dependent, as shown 470 by the SWC × provenance interaction, which was significant for gs and A, and marginal for E. 471 The near-orthogonality of gs–E and A–WUEi implies an independent carbon-economics 472 dimension, in which adjustments in stomatal water-flux is decoupled from photosynthetic 473 performance. This observation means that different provenances adopt distinct combinations 474 of flux control and carbon maintenance in response to drought stress. Some provenances, 475 such as THB exhibited sharp reductions in E with relatively modest declines in A. In contrast, 476 individuals from KKH and DN decreased E in response to declining water availability but 477 increased A, thus increasing WUEi even when the level of drought increases. 478 The E–A decoupling has been observed when a change in stomatal flux does not 479 always correspond to a change in photosynthetic capacity. In some cases of extreme heat, 480 transpiration continues even as photosynthesis declines to near zero ( E↑ A↓ )50, which may 481 stem from a passive physical effect of high temperatures that increases the fluidity of water 482 and the permeability of guard cells 51. This response also has an adaptive advantage where 483 sacrificing water is necessary for cooling by transpiration to enhance the survival of leaves 484 but ultimately will deplete water stores 52. However, Siamese rosewood provenances KKH 485 and DN display a rare, opposite relationship, where transpiration decreases and 486 photosynthesis continues to increase (E↓ A↑ ). There may be a completely different mechanism 487 via mesophyll CO 2 conductance, thus maintaining comparatively high CO 2 supply to 488 chloroplasts even as stomata are tightly closed53. Aquaporin-mediated CO2 diffusion has been 489 found to modulate mesophyll conductance and to respond to stress and hormones, such as 490 drought and ABA priming 54–56. Mesophyll conductance is not measured in our study and its 491 occurrence in Siamese rosewoods is speculative, however, the observation of elevated WUE i 492 during drought for KKH and DN may be linked to improved sustainability and growth 493 outcomes of trees57, especially for a species such as Siamese rosewood, which is fast growing, 494 ecologically pioneering species with high water consumption20. 495 Pigment and growth responses indicate functional acclimation without large structural 496 shifts over our experimental timeframe. We observed that Δ Height and Chl a exhibit 497 treatment × provenance interactions, whereas SLA, LDMC, TSS, and Starch show no 498 detectable treatment or interaction effects. Such patterns are broadly consistent with the 499 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 Page 18 of 33 notion that short-term drought primarily elicits adjustments in gas exchange, pigments, and 500 hormonal signalling, while structural reconfiguration often requires longer or more severe 501 stress and can vary with species and context 58. The high-WUEi provenances, KKH and DN, 502 continue to reflect the adaptive benefits of E–A decoupling, where in both provenances Chla 503 and Δ Height has increased in drought-treated samples. 504 Characterising isohydry in a gs–E/A–WUEi framework as developed in the present 505 study depends on its definition. If isohydry is defined solely by the stomatal closure in 506 declining water availability, then judging by the direction of gs–E in the provenance-PCA, PT 507 and THA would be the most isohydric, and THB would be the most anisohydric at whole-508 provenance levels. If the definition of isohydry also considers the assimilation and water use 509 efficiency, then judging by the direction of PC2 in the provenance-PCA, THA would be the 510 most isohydric and KKH would be the most anisohydric at whole-provenance levels. Our 511 study highlights the heterogeneity in drought response in line with contemporary views of the 512 isohydry-anisohydry continuum, which emphasises that water-potential regulation emerges 513 from multiple, only partly covarying stomatal, hydraulic, osmotic, and metabolic processes, 514 rather than a single trait or score59,60. 515 516 Wettest months, not dry, determine the isohydry-anisohydry continuum 517 Precipitation in wettest months is found to be the single determinant factor for the 518 isohydry-anisohydry continuum in Siamese rosewood, with more isohydric the provenance 519 associated with higher the precipitation. This may be explained by the high wet-season 520 precipitation acting as a trait construction window which promotes anatomical features, such 521 as larger, more conductive earlywood vessels 61. However, larger vessels often track greater 522 vulnerability to cavitation, that is, higher P5062, and thus requires stronger stomatal control to 523 protect the vulnerable xylem63. Moreover, stomatal density and size and cuticle properties are 524 gradually set towards leaf maturation 64, which may peak after the wettest month. Higher 525 precipitation is reported to increase stomatal density across plant communities 65, and may 526 also reduce cuticle thickness 66. Thus, the development and growth in wettest months are 527 likely to determine anatomical adaptation that predicts drought response. 528 An additional effect of higher precipitation in wettest months may be to reduce the 529 carbon penalty commonly associated with isohydry during drought. Wet-season carbon gain 530 can build up larger non-structural carbohydrate (NSC) pools, which are used for maintenance 531 respiration, osmotic adjustment, and defence 67. NSC might persist into the dry season such 532 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 Page 19 of 33 that carry-over carbon can buffer the risk of carbon starvation under stomatal closure 68. A 533 wetter peak month also boosts soil and plant water stores, and thus higher capacitance 69. 534 Stomata spend less total time closed to maintain the same hydraulic safety, thus the area 535 under the photosynthetic curve can be larger even under isohydry 70,71. In contrast, drought 536 stresses in drier months only act on the anatomical and hydraulic settings after they are built 537 in wetter months, and thus stomatal and hydraulic limits are unlikely to be constrained by the 538 dry season. 539 540 Differential gene expression reveals different pathways 541 Most of our understanding of water status regulation along an isohydry-anisohydry 542 continuum in plants is at the level of hydraulics and stomatal physiology while very little is 543 known of its genetic control. Comparative omics has begun to reveal drought-response 544 modules in model and crop plants, and a few tree systems, but explicit searches for putative 545 regulators of isohydry, beyond generic drought or ABA pathways, remain scarce and largely 546 indirect72,73. Here we take a different tack by mapping transcript abundance onto composite 547 axes of water-use and carbon strategies, thereby nominating candidate regulators of these key 548 traits. 549 The Dacoc11936 gene, which is homologous to SEOR1, is the only gene found to be 550 associated with the phenotypic variation along the A–WUEi axis. It encodes a structural 551 phloem P-protein that polymerises and occludes sieve elements upon wounding 74. Although 552 live-imaging studies show that SEOR1 agglomerations do not significantly alter phloem flow 553 under non-wounding conditions in Arabidopsis thaliana75, phloem protection could represent 554 an adaptive response to recurrent mechanical stress in Siamese rosewood due to seasonally 555 dry conditions of southeast Asian forests. This prediction is plausible given that assimilation 556 can be feedback-limited by sink unloading and transport capacity as phloem unloading is 557 largely convective with a diffusion component 76. Therefore, it is plausible that SEOR1-558 medicated occlusion stabilises phloem transport, buffering carbon allocation and thus A–559 WUEi. 560 Dacoc11700 (AT4G14746) is the only gene found to be associated with the 561 phenotypic variation along the gs–E axis. It is annotated as a “neurogenic locus notch-like 562 protein” and remains poorly experimentally characterised in Arabidopsis, but it is predicted 563 to be extracellular 77. It contains a signal peptide and thus could hypothetically contribute to 564 cell-cell or apoplastic signal perception of cues such as ABA. It represents a candidate 565 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 Page 20 of 33 regulator for isohydry that requires functional validation, such as a CRISPR/Cas9 knock-out 566 or overexpression complemented by sub-cellular localisation analyses to verify its apoplastic 567 targeting. 568 We have identified many drought-responsive genes that have been validated in 569 previous research. For example, GASA14 is a small, cysteine-rich, secreted apoplastic 570 peptide that integrates GA and ABA responses and modulates reactive oxygen species (ROS) 571 accumulation78. DEG11 is a chloroplastic protease that helps degrade photodamaged 572 photosystem during photoinhibition, which is important for limiting photooxidative damage 573 with drought-induced stomatal closure and excess excitation 79. CXE20 binds strigolactones 574 (SLs) which impact root system architecture and mycorrhizal signalling, such that altered SL 575 availability can shift water foraging and carbon allocation under drought stress 80. PME1, 576 XTH9, and THE1 all contribute to the modulation of cell wall organisation: PME1 is 577 responsible for pectin de-esterification of homogalacturonan in the cell wall that changes 578 pectin stiffness and porosity 81; XTH9 remodels hemicellulose and drives xylem cell 579 expansion82; THE1 responses to cell wall damage and transduces into hormonal response 580 such as abscisic acid (ABA) production and wall remodelling 83. The associated matrix 581 polysaccharides are synthesised or methyl-esterified in the Golgi cis-cisterna before secretion 582 and thus can adjust wall composition during drought84. 583 584 Differential transcript usage is a potential mechanism of local adaptation 585 Alternative transcript usage can enable rapid loss- or gain-of-function under drought, 586 by toggling coding potential, reshaping domains, exposing transcripts to nonsense-mediated 587 decay, or adding or removing signal peptides that redirect secretion and subcellular 588 localization, thereby tuning processes from guard-cell signalling to cell-wall and cuticle 589 biogenesis85–87. With most plant drought transcriptomics aggregating data to the gene level, 590 isoform diversity and its functional consequences are obscured 88, and therefore population-591 specific or locally adapted mechanisms to distinct hydroclimates are missed. 592 We detected two isoform switches between control and drought treatment specific to 593 the provenance THB, which was identified as the most anisohydric considering gs–E axis. 594 We observed that PRX52 switched to a drought-dominant, loss-of-function isoform with a 595 missing signal peptide. PRX52 is a class III apoplastic peroxidase involved in the synthesis of 596 S units in interfascicular fibres during lignification 89. Class III peroxidases characteristically 597 possess an N-terminal signal peptide for entry into the secretory pathway 90, and co-localise 598 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 Page 21 of 33 with laccases to lignifying wall domains during secondary cell-wall deposition 91. Therefore, 599 the drought-dominant isoform of PRX52 is likely to result in reduced lignin synthesis and 600 higher cell extensibility, which might facilitate stomatal opening and mesophyll expansion at 601 lower turgour92. Class III peroxidases also broadly modulate reactive oxygen species (ROS) 602 homeostasis during stress responses 93. Reduced secretion of PRX52 may dampen the 603 apoplastic hydrogen peroxide (H2O2) burst, which is a proximate trigger of stomatal closure, 604 and thus sustain a higher gs. Other class III peroxidases PRX4, PRX33, PRX34, and PRX71 605 have been shown in guard cells 94. This discovery of reduced secretion in PRX52 regulating 606 anisohydry opens novel avenues for drought research, which may be validated in future 607 studies with ROS staining and imaging in guard cells. 608 We detected a gain-of-function phenomenon in the provenance NP, where both 609 ANN3 and LTPG5 switched from non-coding or non-native isoforms, respectively, to the 610 functional coding isoform during drought. Annexins are Ca² /i5 -dependent phospholipid-611 binding proteins that relocalise to the plasma membrane when cytosolic Ca² /i5 rises, 612 participate in vesicle trafficking and exocytosis, and modulate ROS–Ca² /i5 signalling in 613 stress95. ANN3 specifically has been linked to eATP-regulated growth and vesicle polarity 96. 614 The drought-biased ATTS switch may thus upgreulate ANN3 protein to stabilise both the 615 plasma membrane and exocytosis that maintains cell wall supply and repair 97. On the other 616 hand, GPI-anchored LTPGs localise to the outer leaflet of the plasma membrane and are 617 required for exporting cuticular lipids and depositing suberin to the plant surface 98. LTPG5 618 has confirmed roles in defence signalling, typical of many GPI-APs that interface with RLKs 619 at the cell surface 99. The drought-responsive LTPG5 full-length isoform may act by 620 reinforcing the cuticle and reducing non-stomatal water loss. 621 622 Implications for conservation and drought research 623 This study provides spatially explicit evidence for drought-responsive conservation of 624 Siamese rosewood, which is crucial in times of intensifying drought and extreme weather 625 conditions in their habitat. It enables climate-adjusted provenancing and assisted gene flow, 626 such as matching seed sources to planting sites using a dual filter that combines each 627 provenance’s position in the isohydry space with wettest-month precipitation of both source 628 and target sites. High- WUE/i1, growth-maintaining provenances, such as KKH and DN, are 629 prime candidates for dry, variable environments, but should be deployed in diverse, mosaic 630 mixes to hedge risk and preserve genetic diversity. We present marker panels supported by 631 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 Page 22 of 33 understanding of underlying mechanism useful for drought tolerance screening and breeding, 632 such as common variants responsible for cell wall, but other key drought-responsive genes 633 and isoform switches also offer novel avenues in engineering drought-tolerant plants. Future 634 drought research in Siamese rosewood and at large may include common garden validation of 635 wettest-month precipitation in shaping isohydry behaviour. Our work could support several 636

(1) extend measurements into mesophyll conductance and test aquaporin 637 involvement in improving water-use efficiency; (2) integrate our genomic markers into 638 genomic selection for water-efficient growth; and (3) run operational pilots that compare 639 survival, growth, and hydrological impacts of provenance mixes under real reforestation 640 settings. Siamese rosewood is a promising species that has pioneering ability suitable for 641 forest landscape restoration. Conserving this critically endangered species will continue to 642 realise its ecological and socioeconomic value and may set a valuable model for other 643 threatened tropical tree species in Southeast Asia. 644 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 Figures 645 646 647 Figure 1. (a) Six local seed provenances of Dalbergia cochinchinensis in this study. The map is colour coded with the 648 precipitation of wettest month (mm). The colour scheme of the provenances is consistent in all figures throughout this 649 manuscript. (b) Provenance-level PCA on the coefficients of effect of Treatment × Provenance on physiological traits that 650 show significant interactions. (c) Individual-level PCA on physiological traits that show significant interactions. (d) The 651 interaction effect of precipitation of wettest month and treatment on the isohydry score (which is a composite score derived 652 from the first principal component (PC1) of the individual-level PCA in Figure 1c, see Methods for details). (e) 653 Corresponding interaction plot, which shows model-predicted relationships and the fitted values between precipitation of 654 wettest month and isohydry score. 655 656 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 657 658 Figure 2. Experimental design of the greenhouse experiment. We studied 10 plants per provenance and randomly assigned 659 half of the plants to either well-watered control (C) or water-withholding treatment (D). 660 661 662 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 663 Figure 3. (a) Stomatal conductance ( gs), (c) assimilation rate ( A), (e) transpiration rate ( E), and (g) water use efficiency 664 (WUEi) along the gradient of soil water content (SWC). All traits were square-root transformed to correct for normality. (b), 665 (d), (f), and (h) Corresponding interaction plots of g s, A, E, and WUE i, which show model-predicted relationships and the 666 fitted values between SWC and the corresponding trait. 667 668 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 669 670 671 Figure 4. (a) Change in height ( Δ Height) and (c) chlorophyll a content ( Chla) between control (C) and drought (D) among 672 six provenances. (b) and (d) are the corresponding interaction plots, which show model-predicted relationships and the fitted 673 values between treatment and the corresponding traits. Chla was log-transformed to correct for normality in the model in 674 Figure 3d, but raw values were visualised in Figure 3c. 675 676 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 677 678 Figure 5. Volcano plots of –log 10 q-values against log2 fold change for (a) the drought effect, (c) the isohydry effect (using 679 PC1), and (e) the isohydry effect (using PC2) using likelihood ratio test (LRT) between a full model and a reduced model 680 without the effect of concern. (b) Manhattan plot of –log 10 q-values against chromosome position for the drought effect. (d) 681 VST-transformed gene expression of Isohydry PC1-associated gene Dacoc11700 (AT4G14746). (f) VST-transformed gene 682 expression of Isohydry PC2-associated gene Dacoc11936 ( AtSEOR1) C and D denote well-watered control and drought 683 treatment respectively. 684 685 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 686 Figure 6. Isoform switch of Dacoc21458 (PRX52) between control and drought treatment in provenance THB. (a) 687 Structures and annotations of the two isoforms. (b) Gene expression between control (C) and drought (D) conditions. (c) 688 Isoform fraction of the two isoforms between control and drought conditions. (c) Gene expression between control (C) and 689 drought (D) conditions. 690 691 692 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 Page 29 of 33

693 694 1. Martínez-Vilalta, J., Lloret, F. & Breshears, D. D. Drought-induced forest decline: causes, 695 scope and implications. Biol Lett 8, 689–691 (2012). 696 2. Hammond, W. M. et al. Global field observations of tree die-off reveal hotter-drought 697 fingerprint for Earth’s forests. Nat Commun 13, 1–11 (2022). 698 3. Gazol, A., Pizarro, M., Hammond, W. M., Allen, C. D. & Camarero, J. J. Droughts preceding 699 tree mortality events have increased in duration and intensity, especially in dry biomes. Nature 700 Communications 16, 1–11 (2025). 701 4. McDowell, N. G. et al. Mechanisms of woody-plant mortality under rising drought, CO2 and 702 vapour pressure deficit. Nat Rev Earth Environ 3, 294–308 (2022). 703 5. Calvin, K. et al. IPCC, 2023: Climate Change 2023: Synthesis Report, Summary for 704 Policymakers. Contribution of Working Groups I, II and III to the Sixth Assessment Report of 705 the Intergovernmental Panel on Climate Change [Core Writing Team, H. Lee and J. Romero 706 (Eds.)]. IPCC, Geneva, Switzerland. IPCC, 2023: Climate Change 2023: Synthesis Report. 707 Contribution of Working Groups I, II and III to the Sixth Assessment Report of the 708 Intergovernmental Panel on Climate Change [Core Writing Team, H. Lee and J. Romero 709 (eds.)]. IPCC, Geneva, Switzerland. (2023). doi:10.59327/ipcc/ar6-9789291691647.001. 710 6. Alberto, F. J. et al. Potential for evolutionary responses to climate change – evidence from tree 711 populations. Glob Chang Biol 19, 1645 (2013). 712 7. Brook, B. W., Sodhi, N. S. & Bradshaw, C. J. A. Synergies among extinction drivers under 713 global change. Trends Ecol Evol 23, 453–460 (2008). 714 8. Mekong River Commission. Mekong Low Flow and Drought Conditions in 2019–2021: 715 Hydrological Conditions in the Lower Mekong River Basin. https://www.mrcmekong.org/wp-716 content/uploads/2024/08/LowFlowReport20192021.pdf (2022). 717 9. Thirumalai, K., DInezio, P. N., Okumura, Y. & Deser, C. Extreme temperatures in Southeast 718 Asia caused by El Ninõ and worsened by global warming. Nat Commun 8, 1–8 (2017). 719 10. Zhang, L., Chen, Z. & Zhou, T. Human Influence on the Increasing Drought Risk Over 720 Southeast Asian Monsoon Region. Geophys Res Lett 48, e2021GL093777 (2021). 721 11. Hughes, A. C. Understanding the drivers of Southeast Asian biodiversity loss. Ecosphere 8, 722 e01624 (2017). 723 12. Ang, W. J., Park, E., Pokhrel, Y., Tran, D. D. & Loc, H. H. Dams in the Mekong: a 724 comprehensive database, spatiotemporal distribution, and hydropower potentials. Earth Syst 725 Sci Data 16, 1209–1228 (2024). 726 13. UNODC. World Wildlife Crime Report: Trafficking in Protected Species. (United Nations 727 Publication, 2016). 728 14. Barstow, M. et al. Dalbergia cochinchinensis. The IUCN Red List of Threatened Species 2022 729 https://www.iucnredlist.org/species/215342548/2822125 (2022). 730 15. Gaisberger, H. et al. Range-wide priority setting for the conservation and restoration of Asian 731 rosewood species accounting for multiple threats and ecogeographic diversity. Biol Conserv 732 270, 109560 (2022). 733 16. CDRI. Community forestry for sustainable forest management and livelihhods: a case study of 734 Osoam Community FOrest users group. Cambodia Development Review 18, (2014). 735 17. Luoma-aho, T., Hong, L. T., Ramanatha Rao, V. & Sim, H. C. Forest Genetic Resources 736 Conservation and Management: Proceedings of the Asia Pacific Forest Genetic Resources 737 Programme (APFORGEN) Inception Workshop. (2003). 738 18. APFORGEN. Conserving Rosewood Genetic Resources for Resilient Livelihoods in the 739 Mekong - Project Inception Workshop Report. (2018). 740 19. Jalonen, R. et al. Effective Conservation of Asian Rosewoods (Dalbergias) and Their Genetic 741 Resources Through Regional Collaboration. in Handbook of Asian Rosewoods (eds. Warrier, R. 742 R., Joshi, G. & Arunkumar, A. N.) 461–471 (Springer, Singapore, 2025). doi:10.1007/978-743 981-96-9459-4_32/TABLES/1. 744 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 Page 30 of 33 20. Hung, T. H. et al. Physiological responses of rosewoods Dalbergia cochinchinensis and D. 745 oliveri under drought and heat stresses. Ecol Evol https://doi.org/10.1002/ece3.6744 (2020) 746 doi:10.1002/ece3.6744. 747 21. McDowell, N. et al. Mechanisms of plant survival and mortality during drought: why do some 748 plants survive while others succumb to drought? New Phytologist 178, 719–739 (2008). 749 22. Hartvig, I. et al. Population genetic structure of the endemic rosewoods Dalbergia 750 cochinchinensis and D. oliveri at a regional scale reflects the Indochinese landscape and life-751 history traits. Ecol Evol 8, 530–545 (2018). 752 23. Hartvig, I. et al. Conservation genetics of the critically endangered Siamese rosewood 753 (Dalbergia cochinchinensis): recommendations for management and sustainable use. 754 Conservation Genetics 1–16 (2020) doi:10.1007/s10592-020-01279-1. 755 24. Hung, T. H. et al. Range-wide differential adaptation and genomic offset in critically 756 endangered Asian rosewoods. Proc Natl Acad Sci U S A 120, e2301603120 (2023). 757 25. Gibson, G. The environmental contribution to gene expression profiles. Nature Reviews 758 Genetics 2008 9:8 9, 575–581 (2008). 759 26. Ackermann, M., Sikora-Wohlfeld, W. & Beyer, A. Impact of Natural Genetic Variation on 760 Gene Expression Dynamics. PLoS Genet 9, e1003514 (2013). 761 27. Benny, J., Pisciotta, A., Caruso, T. & Martinelli, F. Identification of key genes and its 762 chromosome regions linked to drought responses in leaves across different crops through 763 meta-analysis of RNA-Seq data. BMC Plant Biol 19, 1–18 (2019). 764 28. Benny, J. et al. Gaining Insight into Exclusive and Common Transcriptomic Features Linked 765 to Drought and Salinity Responses across Fruit Tree Crops. Plants 9, 1–20 (2020). 766 29. Castillejo, M. A., Pascual, J., Jorrín-Novo, J. V. & Balbuena, T. S. Proteomics research in 767 forest trees: A 2012-2022 update. Front Plant Sci 14, (2023). 768 30. Song, F. et al. Transcriptome and association mapping revealed functional genes respond to 769 drought stress in Populus. Front Plant Sci 13, 829888 (2022). 770 31. Hung, T. H. et al. Reference transcriptomes and comparative analyses of six species in the 771 thr eatened rosewood genus Dalbergia. Sci Rep 10, 17749 (2020). 772 32. Canham, C. D. & Murphy, L. The demography of tree species response to climate: seedling 773 recruitment and survival. Ecosphere 7, e01424 (2016). 774 33. Hatfield, J. L. & Dold, C. Water-use efficiency: Advances and challenges in a changing 775 climate. Front Plant Sci 10, 429990 (2019). 776 34. Pérez-Harguindeguy, N. et al. New handbook for standardised measurement of plant 777 functional traits worldwide. Aust J Bot 61, 167–234 (2013). 778 35. Sims, D. A. & Gamon, J. A. Relationships between leaf pigment content and spectral 779 reflectance across a wide range of species, leaf structures and developmental stages. Remote 780 Sens Environ 81, 337–354 (2002). 781 36. Osaki, M., Shinano, T. & Tadano, T. Redistribution of carbon and nitrogen compounds from 782 the shoot to the harvesting organs during maturation in field crops. Soil Sci Plant Nutr 37, 783 117–128 (1991). 784 37. Prjibelski, A. D. et al. Accurate isoform discovery with IsoQuant using long reads. Nature 785 Biotechnology 2023 41:7 41, 915–918 (2023). 786 38. Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for 787 RNA-seq data with DESeq2. Genome Biol 15, 1–21 (2014). 788 39. Ignatiadis, N., Klaus, B., Zaugg, J. B. & Huber, W. Data-driven hypothesis weighting 789 increases detection power in genome-scale multiple testing. Nature Methods 2016 13:7 13, 790 577–580 (2016). 791 40. Wu, T. et al. clusterProfiler 4.0: A universal enrichment tool for interpreting omics data. The 792 Innovation 2, 100141 (2021). 793 41. Korotkevich, G. et al. Fast gene set enrichment analysis. bioRxiv 060012 (2021) 794 doi:10.1101/060012. 795 42. Vitting-Seerup, K., Sandelin, A. & Berger, B. IsoformSwitchAnalyzeR: analysis of changes in 796 genome-wide patterns of alternative splicing and its functional consequences. Bioinformatics 797 35, 4469–4471 (2019). 798 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 Page 31 of 33 43. Anders, S., Reyes, A. & Huber, W. Detecting differential usage of exons from RNA-seq data. 799 Genome Res 22, 2008 (2012). 800 44. Kang, Y. J. et al. CPC2: a fast and accurate coding potential calculator based on sequence 801 intrinsic features. Nucleic Acids Res 45, W12–W16 (2017). 802 45. El-Gebali, S. et al. The Pfam protein families database in 2019. Nucleic Acids Res 47, D427–803 D432 (2019). 804 46. Nielsen, H. Practical Applications of Language Models in Protein Sorting Prediction: SignalP 805 6.0, DeepLoc 2.1, and DeepLocPro 1.0. Methods in Molecular Biology 2941, 153–175 (2025). 806 47. Thumuluri, V., Almagro Armenteros, J. J., Johansen, A. R., Nielsen, H. & Winther, O. 807 DeepLoc 2.0: multi-label subcellular localization prediction using protein language models. 808 Nucleic Acids Res 50, W228–W234 (2022). 809 48. Hallgren, J. et al. DeepTMHMM predicts alpha and beta transmembrane proteins using deep 810 neural networks. bioRxiv 2022.04.08.487609 (2022) doi:10.1101/2022.04.08.487609. 811 49. Erdos, G. & Dosztányi, Z. AIUPred: combining energy estimation with deep learning for the 812 enhanced prediction of protein disorder. Nucleic Acids Res 52, W176–W181 (2024). 813 50. Urban, J., Ingwers, M. W., McGuire, M. A. & Teskey, R. O. Increase in leaf temperature 814 opens stomata and decouples net photosynthesis from stomatal conductance in Pinus taeda and 815 Populus deltoides x nigra. J Exp Bot 68, 1757–1767 (2017). 816 51. Marchin, R. M., Medlyn, B. E., Tjoelker, M. G. & Ellsworth, D. S. Decoupling between 817 stomatal conductance and photosynthesis occurs under extreme heat in broadleaf tree species 818 regardless of water access. Glob Chang Biol 29, 6319–6335 (2023). 819 52. Aparecido, L. M. T., Woo, S., Suazo, C., Hultine, K. R. & Blonder, B. High water use in 820 desert plants exposed to extreme heat. Ecol Lett 23, 1189–1200 (2020). 821 53. Théroux-Rancourt, G., Éthier, G. & Pepin, S. Threshold response of mesophyll CO2 822 conductance to leaf hydraulics in highly transpiring hybrid poplar clones exposed to soil 823 drying. J Exp Bot 65, 741–753 (2014). 824 54. Perez-Martin, A. et al. Regulation of photosynthesis and stomatal and mesophyll conductance 825 under water stress and recovery in olive trees: correlation with gene expression of carbonic 826 anhydrase and aquaporins. J Exp Bot 65, 3143–3156 (2014). 827 55. Flexas, J. et al. Tobacco aquaporin NtAQP1 is involved in mesophyll conductance to 828 CO2in vivo. The Plant Journal 48, 427–439 (2006). 829 56. Chen, J. et al. Aquaporins and CO2 diffusion across biological membrane. Front Physiol 14, 830 1205290 (2023). 831 57. Rancourt, G. T., Éthier, G. & Pepin, S. Greater efficiency of water use in poplar clones having 832 a delayed response of mesophyll conductance to drought. Tree Physiol 35, 172–184 (2015). 833 58. Hancock, R. D. et al. Physiological, biochemical and molecular responses of the potato 834 (Solanum tuberosumL.) plant to moderately elevated temperature. Plant Cell Environ 37, 439–835 450 (2014). 836 59. Martínez-Vilalta, J. & Garcia-Forner, N. Water potential regulation, stomatal behaviour and 837 hydraulic transport under drought: deconstructing the iso/anisohydric concept. Plant Cell 838 Environ 40, 962–976 (2017). 839 60. Hochberg, U., Rockwell, F. E., Holbrook, N. M. & Cochard, H. Iso/Anisohydry: A Plant–840 Environment Interaction Rather Than a Simple Hydraulic Trait. Trends Plant Sci 23, 112–120 841 (2018). 842 61. Schreiber, S. G., Hacke, U. G. & Hamann, A. Variation of xylem vessel diameters across a 843 climate gradient: insight from a reciprocal transplant experiment with a widespread boreal tree. 844 Funct Ecol 29, 1392–1401 (2015). 845 62. Lobo, A. et al. Assessing inter- and intraspecific variability of xylem vulnerability to 846 embolism in oaks. For Ecol Manage 424, 53 (2018). 847 63. Salvi, A. M. et al. Hydroscapes, hydroscape plasticity and relationships to functional traits and 848 mesophyll photosynthetic sensitivity to leaf water potential in Eucalyptus species. Plant Cell 849 Environ 45, 2573–2588 (2022). 850 64. Yang, K. tong, Chen, G. peng & Xian, J. ren. Stomatal distribution pattern for 90 species in 851 Loess Plateau – Based on replicated spatial analysis. Ecol Indic 148, 110120 (2023). 852 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 Page 32 of 33 65. Liu, C. et al. Relationships of stomatal morphology to the environment across plant 853 communities. Nat Commun 14, 1–11 (2023). 854 66. Li, X. et al. A thinner jacket for frosty and windy climates? Global patterns in leaf cuticle 855 thickness and its environmental associations. New Phytologist 248, 107–124 (2025). 856 67. Hartmann, H. & Trumbore, S. Understanding the roles of nonstructural carbohydrates in forest 857 trees - from what we can measure to what we want to know. New Phytol 211, 386–403 (2016). 858 68. Scartazza, A., Moscatello, S., Matteucci, G., Battistelli, A. & Brugnoli, E. Seasonal and inter-859 annual dynamics of growth, non-structural carbohydrates and C stable isotopes in a 860 Mediterranean beech forest. Tree Physiol 33, 730–742 (2013). 861 69. O’Grady, A. P., Eamus, D. & Hutley, L. B. Transpiration increases during the dry season: 862 patterns of tree water use in eucalypt open-forests of northern Australia. Tree Physiol 19, 591–863 597 (1999). 864 70. Matheny, A. M. et al. Observations of stem water storage in trees of opposing hydraulic 865 strategies. Ecosphere 6, 1–13 (2015). 866 71. Zhang, Y. J., Meinzer, F. C., Qi, J. H., Goldstein, G. & Cao, K. F. Midday stomatal 867 conductance is more related to stem rather than leaf water status in subtropical deciduous and 868 evergreen broadleaf trees. Plant Cell Environ 36, 149–158 (2013). 869 72. Song, F. et al. Transcriptome and association mapping revealed functional genes respond to 870 drought stress in Populus. Front Plant Sci 13, 829888 (2022). 871 73. Dal Santo, S. et al. Distinct transcriptome responses to water limitation in isohydric and 872 anisohydric grapevine cultivars. BMC Genomics 2016 17:1 17, 1–19 (2016). 873 74. Ernst, A. M. et al. Sieve element occlusion (SEO) genes encode structural phloem proteins 874 involved in wound sealing of the phloem. Proceedings of the National Academy of Sciences 875 109, E1980–E1989 (2012). 876 75. Froelich, D. R. et al. Phloem ultrastructure and pressure flow: Sieve-Element-Occlusion-877 Related agglomerations do not affect translocation. Plant Cell 23, 4428–4445 (2011). 878 76. Ross-Elliott, T. J. et al. Phloem unloading in arabidopsis roots is convective and regulated by 879 the phloempole pericycle. Elife 6, (2017). 880 7 7. Hooper, C. M., Castleden, I. R., Tanz, S. K., Aryamanesh, N. & Millar, A. H. SUBA4: The 881 interactive data analysis centre for Arabidopsis subcellular protein locations. Nucleic Acids Res 882 45, D1064–D1074 (2017). 883 78. Sun, S. et al. GASA14 regulates leaf expansion and abiotic stress resistance by modulating 884 reactive oxygen species accumulation. J Exp Bot 64, 1637–1647 (2013). 885 79. Kapri-Pardes, E., Naveh, L. & Adam, Z. The Thylakoid Lumen Protease Deg1 Is Involved in 886 the Repair of Photosystem II from Photoinhibition in Arabidopsis. Plant Cell 19, 1039 (2007). 887 80. Roesler, K. et al. Arabidopsis Carboxylesterase 20 Binds Strigolactone and Increases Branches 888 and Tillers When Ectopically Expressed in Arabidopsis and Maize. Front Plant Sci 12, 639401 889 (2021). 890 81. Sénéchal, F. et al. Tuning of Pectin Methylesterification: PECTIN METHYLESTERASE 891 INHIBITOR 7 MODULATES THE PROCESSIVE ACTIVITY OF CO-EXPRESSED 892 PECTIN METHYLESTERASE 3 IN A pH-DEPENDENT MANNER. Journal of Biological 893 Chemistry 290, 23320–23335 (2015). 894 82. Kushwah, S. et al. Arabidopsis XTH4 and XTH9 Contribute to Wood Cell Expansion and 895 Secondary Wall Formation. Plant Physiol 182, 1946–1965 (2020). 896 83. Bacete, L. et al. THESEUS1 modulates cell wall stiffness and abscisic acid production in 897 Arabidopsis thaliana. Proc Natl Acad Sci U S A 119, e2119258119 (2022). 898 84. Harholt, J., Suttangkakul, A. & Scheller, H. V. Biosynthesis of Pectin. Plant Physiol 153, 384 899 (2010). 900 85. Tognacca, R. S. et al. Alternative splicing in plants: current knowledge and future directions 901 for assessing the biological relevance of splice variants. J Exp Bot 74, 2251–2272 (2023). 902 86. Lam, P. Y., Wang, L., Lo, C. & Zhu, F. Y. Alternative Splicing and Its Roles in Plant 903 Metabolism. International Journal of Molecular Sciences 2022, Vol. 23, Page 7355 23, 7355 904 (2022). 905 87. Laloum, T., Martín, G. & Duque, P. Alternative Splicing Control of Abiotic Stress Responses. 906 Trends Plant Sci 23, 140–150 (2018). 907 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint Drought response of Siamese rosewood Hung et al. 2025 Page 33 of 33 88. Sarantopoulou, D. et al. Comparative evaluation of full-length isoform quantification from 908 RNA-Seq. BMC Bioinformatics 2021 22:1 22, 1–24 (2021). 909 89. Fernández-Pérez, F., Pomar, F., Pedreño, M. A. & Novo-Uzal, E. The suppression of AtPrx52 910 affects fibers but not xylem lignification in Arabidopsis by altering the proportion of syringyl 911 units. Physiol Plant 154, 395–406 (2015). 912 90. Lüthje, S. & Martinez-Cortes, T. Membrane-Bound Class III Peroxidases: Unexpected 913 Enzymes with Exciting Functions. International Journal of Molecular Sciences 2018, Vol. 19, 914 Page 2876 19, 2876 (2018). 915 91. Hoffmann, N., Benske, A., Betz, H., Schuetz, M. & Lacey Samuels, A. Laccases and 916 Peroxidases Co-Localize in Lignified Secondary Cell Walls throughout Stem Development. 917 Plant Physiol 184, 806–822 (2020). 918 92. Lima, T. R. A. et al. Lignin composition is related to xylem embolism resistance and leaf life 919 span in trees in a tropical semiarid climate. New Phytologist 219, 1252–1262 (2018). 920 93. Li, S., Zheng, H., Sui, N. & Zhang, F. Class III peroxidase: An essential enzyme for enhancing 921 plant physiological and developmental process by maintaining the ROS level: A review. Int J 922 Biol Macromol 283, 137331 (2024). 923 94. Arnaud, D. et al. Cytokinin-Mediated Regulation of Reactive Oxygen Species Homeostasis 924 Modulates Stomatal Immunity in Arabidopsis. Plant Cell 29, 543–559 (2017). 925 95. Davies, J. M. Annexin-Mediated Calcium Signalling in Plants. Plants 2014, Vol. 3, Pages 128-926 140 3, 128–140 (2014). 927 96. Xu, J. et al. ATANN3 Is Involved in Extracellular ATP-Regulated Auxin Distribution in 928 Arabidopsis thaliana Seedlings. Plants 12, 330 (2023). 929 97. Koerdt, S. N., Ashraf, A. P. K. & Gerke, V. Annexins and plasma membrane repair. Curr Top 930 Membr 84, 43–65 (2019). 931 98. DeBono, A. et al. Arabidopsis LTPG Is a Glycosylphosphatidylinositol-Anchored Lipid 932 Transfer Protein Required for Export of Lipids to the Plant Surface. Plant Cell 21, 1230 (2009). 933 99. Ali, M. A., Abbas, A., Azeem, F., Shahzadi, M. & Bohlmann, H. The Arabidopsis GPI-934 Anchored LTPg5 Encoded by At3g22600 Has a Role in Resistance against a Diverse Range of 935 Pathogens. Int J Mol Sci 21, 1774 (2020). 936 937 938 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.14.692107doi: bioRxiv preprint