Gravitropic Gene Expression Divergence Associated With Adaptation to Contrasting Environments in an Australian Wildflower

ABSTRACT

Plants adapt to their local environment through complex interactions between genes, gene networks and hormones. Although the impact of gene expression on trait regulation and evolution has been recognised for many decades, its role in the evolution of adaptation is still a subject of intense exploration. We used a Multi‐parent Advanced Generation Inter‐Cross (MAGIC) population, which we derived from crossing multiple parents from two distinct coastal ecotypes of an Australia wildflower, Senecio lautus. We focused on studying the contrasting gravitropic behaviours of these ecotypes, which have evolved independently multiple times and show strong responses to natural selection in field experiments, emphasising the role of natural selection in their evolution. Here, we investigated how gene expression differences have contributed to the adaptive evolution of gravitropism. We studied gene expression in 60 pools at five time points (30, 60, 120, 240 and 480 min) after rotating half of the pools 90°. We found 428 genes with differential expression in response to the 90° rotation treatment. Of these, 81 genes (~19%) have predicted functions related to the plant hormones auxin and ethylene, which are crucial for the gravitropic response. By combining insights from Arabidopsis mutant studies and analysing our gene networks, we propose a preliminary model to explain the differences in gravitropism between ecotypes. This model suggests that the differences arise from changes in the transport and availability of the two hormones auxin and ethylene. Our findings indicate that the genetic basis of adaptation involves interconnected signalling pathways that work together to give rise to new ecotypes.

1. Introduction

As populations colonise and stabilise within new environments, they undergo physiological, morphological and behavioural evolution. These changes optimise survival and reproduction to the new conditions (Avital and Jablonka 2000; Carroll 2008; Hochachka and Somero 2002). However, connecting genetic variation to ecological diversity and elucidating the mechanisms underlying adaptation to environmental heterogeneity remains challenging (Feder and Mitchell‐Olds 2003; Hill, Vande Zande, and Wittkopp 2021). Gene expression acts as a pivotal bridge between genotype and phenotype, playing a crucial role in adaptive processes (Rockman 2005; Rockman and Kruglyak 2006). In response to their surroundings, organisms exhibit shifts in gene expression, enabling rapid physiological adjustments (Gracey 2007). Over time, these shifts drive long‐term evolutionary change (Khaitovich et al. 2006). Therefore, uncovering the patterns and drivers of gene expression divergence during adaptation is key for deepening our understanding of adaptive variation in natural populations (Hill, Vande Zande, and Wittkopp 2021; Rockman 2008; Triant, Nowick, and Shelest 2021). Plants are excellent models for studying adaptation because they continuously explore their surroundings through various physiological responses despite being sessile (Bhargava and Sawant 2013; Skriver and Mundy 1990). Tropisms, such as movements in response to light and gravity, are crucial for the survival and reproduction of plants (Gilroy 2008). Tropisms begin early in development and are controlled through hormonal gradients in various plant tissue (Esmon, Pedmale, and Liscum 2004). As a result, they enhance the match between phenotype and environment for most plants (Kiss 2006).

Phytohormones are central regulators of tropisms, mediating environmental cues into physiological responses and controlling growth and development through complex signalling networks (El Sabagh et al. 2022; Esmon, Pedmale, and Liscum 2004). For instance, the hormone auxin plays a vital role in many tropic responses, including phototropism and gravitropism, as asymmetric auxin distribution across cells initiates differential cell growth, which induces bending (Han et al. 2021; Muday 2001). Plants perceive gravity through starch‐filled amyloplasts within specialised statocytes, which sediment in the direction of gravity (Kolesnikov et al. 2016). The sedimentation of amyloplasts during gravitropic responses signals for auxin redistribution in response to the alignment of roots and shoot components with the direction of gravity (Joo, Bae, and Lee 2001; Nakamura, Nishimura, and Morita 2019), creating a concentration gradient that drives cell elongation on one side of the tissue. This results in the physical curvature of hypocotyls, stems, and branches to grow upwards against the gravity vector and roots to grow downwards into the soil (Rakusová et al. 2011).

The coordinated expression of numerous genes is crucial for the synthesis, transport and response to phytohormones (Dong et al. 2013; Rakusová et al. 2011). Moreover, induced gene expression changes can significantly affect the tropic responses of a plant across environments (Li, Hagen, and Guilfoyle 1991; McClure and Guilfoyle 1989). Work on tropic responses has found that plant hormones also play a critical role in stress responses, and the evolution of hormonal pathways is considered a key contributor to local adaptation in natural populations (reviewed in VanWallendael et al. 2019). For instance, different populations adapted to coastal, inland, and alpine environments often exhibit similar trait changes, such as variations in growth habit (Foster et al. 2007; Itoh 2021; Lowry, Rockwood, and Willis 2008; Walter et al. 2016), which likely stems from changes in key genes in the gibberellin (GA) and auxin pathways (Kou et al. 2021) as adaptations to their specific habitats (e.g., strong winds).

Studies suggest that evolutionary changes in hormonal pathways can simultaneously affect plant shape, function and fitness, influencing their ability to adapt to new environments (VanWallendael et al. 2020, 2019). Identifying the genetic changes that influence these pathways and understanding their impact on the transcriptional networks of plants will be useful for solving many fundamental questions in adaptation. In particular, the role of variation in hormone‐related gene expression during the evolution of adaptive traits in natural populations and its impact on the origin of new ecotypes and species requires further investigation (Brown and Kelly 2021; Hamann et al. 2021). Hormonal pathways play a critical role in local adaptation by driving pleiotropic effects on various traits. These pathways integrate environmental signals, leading to coordinated physiological and developmental responses. For instance, auxins and gibberellins are well‐known for their roles in plant growth and development, including responses to gravity and other environmental stimuli. These hormones can influence multiple traits simultaneously, facilitating adaptive changes in diverse environments. Studies have shown that similar hormonal pathways can lead to convergent evolution of traits in different species, underscoring their importance in local adaptation. In this study, we contribute to filling this gap by investigating how hormone‐related gene expression was affected during the evolution of an adaptive trait, gravitropism. To address this question, we examine gene expression in the Senecio lautus species complex.

The S. lautus species complex offers a compelling system for investigating the relationship between local adaptation, gene expression divergence, and the origin of ecotypes and species. This species complex displays contrasting growth habits along the Australian coast, where populations inhabiting dunes grow erect, whilst those found in adjacent rocky headlands grow prostrate, forming mats on the ground (James, Arenas‐Castro, et al. 2021; James, Wilkinson, et al. 2021; Wilkinson et al. 2021). These forms have evolved repeatedly and independently many times, implicating a role for natural selection in their diversification (James, Arenas‐Castro, et al. 2021). Previous comparisons between Dune and Headland populations revealed the repeated differentiation of genes related to the transport and regulation of the phytohormone auxin (James, Arenas‐Castro, et al. 2021; Roda, Ambrose, et al. 2013; Roda, Liu, et al. 2013). This differentiation coincided with divergence in auxin‐controlled traits such as shoot height, shoot branching (Roda, Ambrose, et al. 2013; Roda et al. 2017), shoot gravitropism and reproductive barriers between Dune and Headland populations (Melo et al. 2014; Richards and Ortiz‐Barrientos 2016). Notably, previous field selection experiments revealed that plant height and shoot gravitropism could evolve in the natural habitats of the coastal populations in the direction of the local ecotype (Wilkinson et al. 2021). These results further reinforce the conclusion that natural selection has contributed to the origin of ecotypes in S. lautus and that plant height and shoot gravitropism are under selection in its coastal populations.

To understand the role of natural selection on gene expression during the evolution of an adaptive trait, we investigated gene expression variation during gravitropism. We used a previously created Multi‐parent Advanced Generation Inter‐Cross (MAGIC) population (Wilkinson et al. 2021, also known as a recombinant hybrid population) to parse out causal from incidental associations of expression with adaptive trait divergence (Figure 1). This population provides a diverse and randomised genetic background to reduce potential confounding effects from other traits (Kumar et al. 2023). This population was created from 23 Dune and 22 Headland individuals from the Lennox Head (NSW, Australia) population of S. lautus (Figure 1a). These individuals were randomly mated in three independent genetic lines for 10 generations. In such a population, trait correlations in the parental populations become disassociated, helping us track expression divergence one trait at a time. Because the MAGIC population consists of descendants from many founders, it carries large amounts of phenotypic and genetic diversity. Consequently, sub‐setting the population into gravitropic and agravitropic individuals leads to populations carrying similar allele frequencies at all loci, except for those linked to gravitropic differences. Here, we use the MAGIC population created in our laboratory and builds on the phenotypic and evolutionary data previously collected, to profile bulk‐segregant gene expression divergence during gravitropic responses that may be involved in adaptation to contrasting environmental condition. Under this experimental premise, we expected to find differences in genes related to hormonal pathways and their interconnectedness. With this approach, we illuminated the mechanisms by which natural selection influences complex cellular systems, paving the way to understand how they foster the origins of natural diversity.

FIGURE 1.

Experimental design of the MAGIC population and gravitropism RNA‐seq. Schematic diagram illustrating the derivation of the MAGIC population from ancestral populations and the setup of the factorial gravitropic experiment. (a) The creation of the MAGIC population. 23 Dune plants and 22 headland plants were collected from Lennox Head (NSW, Australia) for crossing. After 10 generations of equal contribution mating (including 3 generations as part of a selection experiment in the field), the plants were used to create the F11 generation used in this experiment. (b) Describes how many individuals that could be used for this experiment across Gravitropic and Agravitropic families. The mean angle for agravitropic plants once they had been turned was 17°, while the mean for gravitropic plants was 51°, based on the cut‐off we selected for the gravitropic responses in each tail. (c) Depicts the rotation response in gravitropic plants. (d) Shows the experiment that was conducted for RNA sequencing, where the plants were turned and after the allotted time point was completed, the hypocotyls were cut at the base of the cotyledons and beginning of the soil. These hypocotyls were placed into an Eppendorf tube and snap‐frozen in Liquid nitrogen. (e) Shows the different comparisons we could investigate with the RNA‐seq data. There are four responses that would give a biological understanding of gravitropism as described.

2. Methods

2.1. MAGIC Population

To study gene expression underlying divergent gravitropic responses in Senecio, we used the phenotypic data previously collected in the MAGIC population for gravitropism in each of its families (Wilkinson et al. 2021). The MAGIC population allows us to examine gene expression in a diverse and randomised genetic background, reducing potential confounding effects from other genetically unrelated traits. We used averaged gravitropism values for each F11 family that was measured in Wilkinson et al. (2021) and selected the 10 most (strongest shoot response to plant rotation) and least (weakest shoot response to plant rotation) gravitropic families (Figure 1b) for our RNA expression experiments. Gravitropic families rotated, on average, 51.05° ± 1.61° towards the gravity vector after rotation, whereas agravitropic families bent on average 17.65° ± 2.12° towards the gravity vector. By using individuals from the tails of the distribution of gravitropism in the F11 generation, we effectively reconstructed the trait of gravitropism and investigated the underlying genetic underpinnings that explain the phenotypic difference between gravitropic and agravitropic families. The randomised genetic background in the MAGIC population allows for a clearer association between gravitropism and gene expression differences given that any two random sets of individuals selected from the MAGIC population will have the same allele frequencies at all loci, barring from some chance differences. Our hypothesis is that the gravitropic and agravitropic sets of individuals will have allele frequencies beyond chance at a subset of loci, and these will in turn affect the expression profiles of the two groups. Studying gene expression in these contrasting families is crucial for identifying the specific genes and molecular pathways responsible for gravitropism. This knowledge could lead to a deeper understanding of plant responses to gravity, thus enhancing our understanding of adaptation and the formation of ecotypes.

2.2. Gravitropism Experiments

To extract RNA from the plants during gravitropic responses we designed an experiment to use as many individuals as possible (Figure 1c,d). We divided 600 F11 seeds into 60 pools of 10 seeds each (See Figure S1 for more information). Each pool consisted of 10 individuals from 10 different gravitropic or agravitropic families. Due to unexpected seedling deaths, not every pool consisted of 10 individuals, the accurate numbers can be found in Table S1 and is reflected in Figure 1b. Overall, 289 (out of 300) agravitropic individuals were used across the 30 agravitropic pools and 271 (out of 300) gravitropic individuals were used across the 30 gravitropic pools. This procedure led to the creation of 30 gravitropic and 30 agravitropic pools. We first scarified the seeds to induce germination and placed them on Petri dishes with filter paper. The emerging seedlings were grown in the dark for 3 days and then in the light for 4 days on a 12:12 light:dark cycle. At 1 week of age, we transplanted the seedlings into squat‐forestry tubes with UQ23 soil (with added osmocote). We watered daily for 2 weeks and then randomly allocated the pools into equally sized treatment and control groups (used as a fixed effect) over five time points (fixed effect) and three experimental batches. The control groups were not rotated, while the treatment groups were rotated 90° (called rotation below). We collected tissue from the hypocotyl at five different time points: 30, 60, 120, 240 and 480 min for pools allocated to control or rotation groups. For each unique combination of time point and experimental batch, there were four pools—gravitropic rotation, gravitropic control, agravitropic rotation, and agravitropic control. We treated the experimental batch as a random effect to control for variation between gravitropic and agravitropic groups arising from minor differences in the timing of the seeds being grown. We cut the hypocotyl at the base just above the soil and the base of the cotyledons. We then placed each hypocotyl into 2 mL Eppendorf tubes with two 2.8 mm stainless steel grinding balls and snap‐froze it in liquid nitrogen.

2.3. RNA Extraction and Library Preparation

We extracted and sequenced total RNA from 60 pooled samples. First, we ground each frozen hypocotyl sample using a Geno/Grinder (SPEX Sample Prep) and stainless‐steel grinding balls. After the grinding was completed, the samples were pooled for RNA extraction. We then used the Bioline Isolate II mini kit (https://www.bioline.com/mwdownloads/‐download/link/id/1209/) for RNA extraction. We assessed the quality of the RNA using a 1.5% agarose gel, run at 120 V for 25 min, and RNA yields using a Biochrom NanoVue with 1 μL of the sample (Figure S2). We sent the 60 pooled RNA samples to Beijing Genomics Institute (BGI) Australia located at Queensland Institute for Medical Research (QIMR) Berghofer (Brisbane, Australia) for library preparation using the BGIEasy RNA Library Prep Set and sequencing using DNA nanoball sequencing (DNBseq) (BGI 2018). This method uses adaptors ligated to fragmented mRNA to produce libraries containing single‐stranded, circularised DNA with an adaptor in the middle (Li et al. 2019). The library was amplified with phi29 (a DNA polymerase) to make DNA nanoballs, which were loaded into the patterned microarray, and the paired‐end 150 reads were sequenced by synthesis (BGI 2018). Any further necessary steps followed the instruction from the manufacturer.

2.4. Quality Control and Read Mapping

We inspected the raw reads for all 60 samples using FASTQC V0.12.0 (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/) to assess the overall quality of the sequences and determine the extent of base trimming required to eliminate low‐quality bases. Based on this inspection, we identified the first 10 base pairs (bp) at the start of each read as regions of low quality and thus trimmed these nucleotides using the Trimmomatic (Bolger, Lohse, and Usadel 2014) tool (http://www.usadellab.org/cms/?page‐=trimmomatic).

Following quality control, we assembled and mapped trimmed RNA‐Seq reads onto a S. lautus reference transcriptome (V1.4.1 Wilkinson et al. 2021) using HISAT2 (https://ccb.jhu.edu/software/hisat2/‐index.shtml), an efficient and accurate tool for aligning sequencing reads to a large reference genome (Kim, Langmead, and Salzberg 2015). We chose HISAT2 due to its ability to handle spliced alignments and its compatibility with downstream tools. The output from HISAT2 was in SAM (Sequence Alignment/Map) format, which we then converted to binary format (BAM) using Samtools (http://samtools.sourceforge.net) for easier handling and reduced storage requirements (Li et al. 2009). We then processed the BAM files with a TPM (Transcripts Per Million) calculator (https://github.com/ncbi/TPMCalculator) to normalise read counts and allow comparisons across different genes and samples (Vera Alvarez et al. 2018) by accounting for sequencing lane, transcript length, sequencing bias, and total number of reads (Conesa et al. 2016). We chose TPM normalisation as it significantly outperforms other methods of normalisation, such as CPM and FKPM (Johnson and Krishnan 2022; Zhao et al. 2021). Finally, to prepare our data for subsequent analysis, we used GMAP/GSNAP (http://www.gene.com/share/gmap) to translate the aligned reads into GFF3 format.

2.4.1. Data Filtering

We began with 573,226 transcripts. For the quality filtering of our dataset, we applied several filtering criteria. (i) We eliminated 176,287 transcripts with missing read counts across 60 pools, which would obstruct downstream statistical analyses. Next, (ii) we set a read count threshold, filtering out transcripts with fewer than 10 reads in any pool to ensure robust differential gene expression identification. Additionally, (iii) we recognised the possibility of alternative splicing, where a single gene can give rise to multiple transcript variants. To manage this, we identified transcript sequences with a high degree of overlap (> 95% sequence intersection) and clustered them together. This step reduced redundancy and helped to ensure that our analysis focused on expression differences between genes rather than differences between transcript variants of the same gene. Following these data filtering steps, we were left with 269,209 gene sequences that we could confidently use for further data analysis (see Table S3 for details).

2.5. Gene Expression Quantification and Analysis

We employed bulk‐segregant analysis to profile gene expression divergence in gravitropic and agravitropic pools of recombinant individuals from the MAGIC population. This approach allows us to identify average population responses in gene expression resulting from allelic frequency differences between gravitropic and agravitropic families. We employed multiple methods to identify differentially expressed (DE) genes across various time points, which we summarise next.

We identified DE genes as those showing significant changes in expression at various time points using Cuffdiff2 (https://bioinformaticshome.com/tools/rna‐seq/descriptions/Cufflinks.html). We used a false discovery rate (FDR) of 0.05 to account for multiple testing in combination with a twofold change cutoff. We further identified DE genes using the limma Trend and Voom functions (https://bioconductor.org/packages/release/bioc/html/limma.html). Limma is an R‐Bioconductor software package that provides an integrated solution for differential expression analyses of data from gene expression experiments (Smyth, Thorne, and Wettenhall 2003). Limma Trend and Voom offer a robust framework for identifying genes that display significant expression changes under experimental conditions. Previous studies have also shown that it performs well when studying time series data and provides low FDRs (Corchete et al. 2020; Smyth 2005). Using limma, we fitted two models to identify genes that are either DE between two groups across all time points, or alternatively between two groups at a single specified time point where we treated all variables, including time, as discrete factors (see the Supporting Information for more information about how this model was fitted). We used average gene expression to manage the constraints of using batches (further information can be found in the Supporting Information). Having identified the DE genes, we next sought to understand their functional significance through gene set and gene ontology (GO) enrichment analyses.

2.5.1. Gene Set and Gene Ontology Enrichment

To glean more insights into the predicted biological context of the identified genes from S. lautus, we performed orthologue mapping against the A. thaliana genome. GO term enrichment is a technique used in the functional analysis of large‐scale genomic or transcriptomic data. The approach aims to determine if certain biological processes, cellular components, or molecular functions—collectively referred to as GO terms—are over‐represented (enriched) among a set of genes of interest, more than would be expected by chance (Khatri and Drăghici 2005). We primarily focused on biological function GO terms. We compared the frequency of GO terms associated with a given set of genes (like those DE in our experiment) against the frequency of GO terms in A. thaliana, where we chose to focus only A. thaliana genes that had orthologues in the S. lautus transcriptome. Certain GO terms will be more prevalent in our gene set than the background set under control conditions, indicating that these biological processes or functions might be particularly relevant or altered under experimental conditions.

We used ShinyGO v0.77 (http://bioinformatics.sdstate.edu/go/) for GO term enrichment and visualisation, setting a FDR threshold of 0.1 (Ge, Jung, and Yao 2019). This software creates networks of GO terms, where nodes represent the enriched GO terms, and node size corresponds to the number of genes within that term. The edges (lines between nodes) indicate connections between GO terms, reflecting the multifunctional nature of many genes. The weight of the edge demonstrates the number of genes involved in both connected GO terms. We proceeded to select the top 10 GO terms for further in‐depth analysis. Further to these unsupervised analyses, we used a list of over 300 curated genes (Table S3) known to be involved in the gravitropic response in plants (Wilkinson et al. 2021; James et al. 2023) to directly explore the role of various known functions of the auxin pathway—biosynthesis, conjugation, transport, and signalling.

2.5.2. Weighted Gene Co‐Expression Network Analysis

We employed Weighted Gene Co‐expression Network Analysis (WGCNA) (http://www.genetics.ucla.edu/labs/horvath/CoexpressionNetwork/Rpackages/WGCNA) to identify modules of co‐expressed genes associated with gravitropism (Langfelder and Horvath 2008). For this, we used the gravitropic control vs. gravitropic rotated DE gene subset, as changes in the expression of these genes are expected to indicate a functional response to gravitropism. The aim was to observe alterations in co‐expression patterns transitioning from functional responses, as typical of gravitropic plants, to non‐functional or absent responses, as seen in agravitropic and un‐rotated control plants respectively. To this end, we constructed a co‐expression network on the selected genes for each of the four groups. WGCNA first constructs a similarity co‐expression matrix using Pearson's correlations, called the correlation matrix (Columns E–H of Table S4). Then an adjacency matrix is created (Columns A–D of Table S4). Next, the topological overlap matrix (TOM) is constructed from the adjacency matrix to evaluate the shared connections in the network (Columns R–U of Table S4), offering a robust method to discern biologically meaningful clusters of co‐expressed genes (Langfelder and Horvath 2008).

Co‐expression analysis was based on variance stabilised expression using the deseq2 library (varianceStabilizingTransformation function). Genes with > 0.99 correlation across all the groups were merged, reducing the count from 517 to 428 genes. We constructed unsigned networks (where we do not consider the sign, positive or negative, of the correlation between genes) with power 12, which results in a more stringent network, where only strong correlations are considered significant. In our WGCNA analysis, we tested the robustness of the co‐expression networks using a range of correlation cutoffs. This approach ensures that the identified gene modules are stable and not artefacts of a specific cutoff threshold. The selection of the final network was based on a comprehensive analysis of multiple cutoff values to ensure the most biologically relevant and statistically robust network was chosen. Using Cytoscape version 3.8.2 (http://cytoscape.org), we visualised the resulting networks by exporting the edge and node results from WGCNA through the ExportNetworkToCytoscape function. In our study, Cytoscape facilitated the creation and interpretation of gene interaction networks, showcasing the relationships among DE genes, GO‐enriched genes, and those detected via WGCNA.

2.6. Identification of Interactions Between Genes

To understand how differential gene expression may contribute to a complex trait like gravitropism, we sought to identify linked genes, defining linked genes as those that have a known interaction where the proteins are physically interacting, or those that act as intermediate links between multiple DE genes (Table 1 and Table S5). Using the previously identified DE genes and those found via WGCNA and GO term enrichment, we utilised the BioGrid version 4.4.223 (https://thebiogrid.org/) and IntAct (https://www.ebi.ac.uk/intact/home) databases to identify predicted genetic interactions between the genes of interest. We recorded all detected interactions (Table S3) and visualised them in a comprehensive network using the Cytoscape software. The number of interactions and their predicted functions can be found in Table S2. This approach allowed us to hypothesise a potential network, further elucidating the complex interplay between genes influencing gravitropism.

TABLE 1.

Functional insights and possible gene interactions during the gravitropic response in the S. lautus MAGIC population from Lennox Head, NSW.

Gene family/function	Key DE genes identified	Known interactions	Notable observations/comments
Auxin‐related	SAUR 14, SAUR21, SAUR23, SAUR 50, AUX/IAA, etc.	ARF19 with AUX/IAA (Heterodimer formation)	Major early response at 30 min in gravitropic plants.
Ethylene‐related	ERF gene family, Ethylene Insensitive 3	ERF12, JAZ1, WRKY41, MYB54, RVE2, IAA16, IAA20, IAA34	Down‐regulated in both gravitropic and agravitropic plants under rotation.
Gibberellin‐regulated	Gibberellin regulated protein	UBQ3, MAD2	Identified in gravitropic plants.
Histone modification	H3.2, H2A, H3	ARF19 and ARF7	A significant early response, possibly (mediating histone modification)
Ubiquitin	UBQ3	ARF19, ARF7, SAUR14, MYC‐TF
GDSL esterase	GDSL esterase	CYTOSOLIC ABA RECEPTOR KINASE 11, RELA/SPOT HOMOLOGUE 1, RSH1
Myc‐TF	BHLH 157, BHLH149, IBH1

3. Results

3.1. High‐Quality Transcriptome Data From Pooled RNA‐Seq in the F11 MAGIC Population of S. lautus

The quality and reproducibility of RNA‐seq results are contingent upon obtaining high‐quality sequence data and addressing potential batch effects. We initially refined the raw sequence reads to ensure robust results, removing adapters and low‐quality sequences (refer to Methods for details). Our experimental design involved conducting three separate gravitropism experiments under uniform environmental conditions (Figure S1). RNA was extracted from 556 individuals for each of these experiments, subsequently evaluated at five distinct time points. These individuals were grouped into 60 pools containing representatives from 10 gravitropic or 10 agravitropic hybrid families. Illumina‐based pooled sequencing yielded 94 million high‐quality sequence reads per pool. Mapping these sequences to the S. lautus reference genome yielded 573,226 unique transcripts. We eliminated 176,287 transcripts with missing read counts across 60 pools, which would obstruct downstream statistical analyses following data filtering steps (see methods for details), we were left with 269,209 gene sequences that we could confidently use for further data analysis (Tables S1 and S2).

3.2. Rotating Gravitropic Plants Leads to Early and Late‐Stage Changes in Gene Expression

We performed time course analysis to explore the dynamic changes in gene expression between gravitropic and agravitropic conditions, and the significant differences that emerged at distinct time points during the experiment. At the 30‐min time point, gene expression exhibited pronounced differences between both rotated and control conditions: particularly, significant variation was observed between the agravitropic rotation conditions compared to the other three treatment groups (Figure 2). These differences in expression, however, stabilised by the 60‐min interval, at which there were no significant differences between groups. As the experiment progressed to 120 min, a significant increase in gene expression appeared in the agravitropic control group (AC vs. GC: F‐value 13.7, p < 0.001, n = 20). Specifically, we detected differences between agravitropic control and agravitropic rotated pools with the agravitropic control group showing higher average gene expression (for statistical differences between each comparison see Table S2). A downward trend in gravitropic rotated gene expression was identified at the 240‐min point, with significant differences between gravitropic rotated pools and the other three conditions. By the 480‐min measurement, the decline became pronounced across both the control groups (agravitropic and gravitropic controls); however, gene expression substantially increased in the gravitropic rotated group (Figure 2). Interestingly, the combined influence of these conditions culminated in a substantial increase in gene expression by the end of the experiment. These differences can also be observed when dimension reduction tools are used, as a Principal Component Analysis (PCA) shows the structure of the data which indicates that each time point groups together (Figure S2). We examined the gene expression at 480 min post‐rotation where we observed striking differences between agravitropic and gravitropic pools (across both control and rotated) and identified 2200 DE transcript sequences. Of these, 1567 exhibited expression differences between the gravitropic and agravitropic control groups, highlighting the inherent differences between the two types of families within the F11 MAGIC population. Among the genes showing significant temporal expression changes, those related to the auxin and ethylene pathways were particularly notable.

FIGURE 2.

Gene expression dynamics over time in the MAGIC population of S. lautus from Lennox Head, NSW. The plot depicts average gene expression levels (Log₂ TPM) across five different time points. Each line represents a unique combination of gravitropic conditions and rotation; Agravitropic‐Control (Green‐Solid), Agravitropic‐Rotated (Green‐Dashed), Gravitropic‐Control (Orange‐Solid), and Gravitropic‐Rotated (Orange‐Dashed). Standard error was calculated. Dynamic changes in average gene expression over an eight‐hour period following rotation. This figure highlights the temporal fluctuations in the transcriptome, with significant expression changes observed at specific time points. Additional analyses identify key gene groups contributing to these changes, providing insights into the molecular basis of the gravitropic response.

We also investigated multiple comparisons to find shared DE genes (Figure 3). We found that there was a large amount of overlap for DE genes between the control pools and rotated pools (1144 genes), indicating that there are still many similarities between the gravitropic and agravitropic pools that need to be removed to find the genes contributing to gravitropism alone. On the other hand, a much smaller number of genes were found to be shared between the gravitropic (control and rotated) and agravitropic (control and rotated) pools (177 genes). These genes show that there is still some overlap within the gravitropic responses of both the gravitropic and agravitropic pools. We removed these 177 genes from the dataset when investigating the differences in expression underlying gravitropic responses. These findings show that while there are many differences in gene expression to be investigated further, many of these genes are not comparable as they are DE across both groups. These genes may be related to shared functions such as circadian rhythm or responses to dark.

FIGURE 3.

Summary of shared differential gene expression in S. lautus MAGIC population. This figure shows the number of DE genes identified between key comparisons, indicated by connecting arrows. Gravitropic control (GC) versus agravitropic control (AC) contrasts (a) highlight inherent gene expression variation potentially linked to gravitropism‐related genetic differences. Such differences may underpin distinct gravitropic and agravitropic phenotypes during development. The gravitropic rotated (GR) versus agravitropic rotated (AR) comparison highlights differences between the responses to changes in the direction of gravity. The 1144 shared differentially expressed genes between the AC‐GC contrast and AR‐GR contrast are genes that are differentially expressed regardless of the gravitropic treatment. The agravitropic control versus agravitropic rotation contrast gauges expression shifts in the agravitropic group under rotation (b). Comparing this with the agravitropic control versus gravitropic control contrast reveals if such shifts are rotation‐specific or are intrinsic between phenotypic groups. Similarly, the gravitropic control vs. gravitropic rotated contrast assesses the response to rotation of the gravitropic group. Contrasting this with agravitropic control versus gravitropic control determines rotation‐specificity or inherent group differences. Comparing agravitropic control versus agravitropic rotated and gravitropic control versus gravitropic rotated evaluates potential interaction effects between phenotype and treatment on expression. Varied responses between gravitropic and agravitropic groups to rotation indicate phenotype‐dependent treatment effects, shedding light on molecular mechanisms underpinning phenotypic disparities.

3.3. Differential Expression of Auxin and Ethylene‐Related Genes During the Gravitropic Response in S. lautus

We identified a range of early and late‐stage changes in expression and found many DE genes underlying histone modification and within the auxin pathway, particularly concerning auxin conjugation and transport (Table S2 shows RNA expression statistics for auxin‐related genes from S. lautus). These observations provide a foundation for understanding the molecular mechanisms behind gravitropism. A co‐expression cluster of auxin‐related genes, mainly from the SAUR family, was prominent at the 30‐min mark. Ethylene‐related transcripts, predominantly from the ERF gene family, were down‐regulated in gravitropic and agravitropic control groups but upregulated in the rotated groups (Figures 4 and 6b). Furthermore, WGCNA highlighted connections between specific auxin and ethylene genes in gravitropic networks, which were absent in agravitropic networks (Figure 5a,b). Notably, the agravitropic control group displayed expression differences in genes such as GDSL esterase and histone modification genes compared to the gravitropic control group, even without rotation, as depicted in Figure 4. Through WGCNA, we identified specific gene modules that are differentially co‐expressed between the gravitropic and agravitropic conditions. These modules highlight key functional groups and pathways involved in the gravitropic response. For instance, genes related to auxin signalling and response were found to be more strongly co‐expressed in the gravitropic network compared to the agravitropic network (Figure 5b).

FIGURE 4.

Gene expression for candidate pathways over time. Based on previous knowledge of gravitropism, some hormonal pathways were of interest for differential expression analysis. Auxin (a), ethylene (b) and gibberellins (c) are all well‐known to be involved in gravitropic processes and were found to be differentiated over time. Histone modifications (d) are also important in controlling many developmental processes but were not found to be significantly differentiated at any time point between the four treatment groups. Panels show the average expression of 73 auxin‐, six ethylene‐, three gibberellin‐, and three histone‐related genes. AC is agravitropic control, AR is agravitropic rotation, GC is gravitropic control, and GR is gravitropic response treatment.

FIGURE 6.

Transcripts per million (TPM) over time for four genes identified in the WGCNA that show differential expression across the treatment groups from the gravitropism experiment. The green lines represent the agravitropic groups, while the orange lines represent the gravitropic groups. The solid lines represent rotated groups, while the dashed lines are the control group. Time is represented on the x‐axis showing expression at each of the five time points (30, 60, 120, 240, and 480 minutes). Each panel shows the expression as TPM of an individual gene with the name of the gene and the name of the gene as it was identified in the S. lautus transcriptome (in TrN2TFQ format). (a) shows the expression of AUX/IAA over time; (b) shows the expression of Ethylene‐insensitive 3; (c) shows the expression of Histone H3; and (d) shows the expression of the Auxin Response Factor ARF19.

FIGURE 5.

Co‐expression analyses of differentially expressed genes between Gravitropic hybrids from the MAGIC population of S. lautus at Lennox Head, NSW, exposed to Control or Rotated conditions. (a) Each panel shows the output of the WGCNA visualised in Cytoscape, where the clustering of the same subset of DE genes is compared across the two groups of pools. The colouring of genes is based on the clustering method used in WGCNA. The number of edges between genes (connections between genes based on the similarity of expression over time) and the size of the nodes (the number of connections a gene has) in each network changes between plants exposed to control (no rotation) or treatment (rotation) conditions. (b) Expression correlations among candidate genes in the MAGIC population of S. lautus that show differential gene expression in gravitropism experiments. Connectedness among genes varies according to the gene function and the treatment (control and rotation treatment) for a group of hybrids from the MAGIC population with different levels of gravitropic response (agravitropic and gravitropic). Standard error was calculated using the connectivity values from each of the 15 pools for each treatment group.

The increased expression of gibberellin‐regulated proteins suggests that gibberellins play a crucial role in modulating the gravitropic response, influencing processes such as cell elongation and differentiation. Histone modification genes point to the involvement of epigenetic mechanisms, which may enable rapid and reversible changes in gene expression in response to gravitational stimuli. GDSL esterases likely contribute to modifying cell wall components and signalling pathways, facilitating the mechanical and signalling aspects of gravitropic bending. These combined pathways illustrate the complexity of the genetic network underlying the adaptive response to gravitropism. These findings underscore the complex interplay between auxin and ethylene signalling pathways in the gravitropic response of S. lautus, giving us insights into the regulatory networks that govern plant adaptation to gravitational changes.

3.4. Co‐Expression Networks Reveal Candidate Auxin‐Related Genes Interacting During the Gravitropic Response

Our analysis of DE genes reveals potential co‐expression networks that may play a role in gravitropic responses. Of the 428 genes that had a twofold change, a statistical cutoff of at a significance level of 5% and which appeared in the WGCNA networks (Figure 5a,b), 68 genes have one or more known genetic interactions based on IntAct and BioGrid databases. Among these 68 genes, 16 had known interactions with other genes that were identified within the co‐expression network (Table S2). These known interactions from the literature indicate the potential for these genes to interact during gravitropic responses (Figure 7). For example, ARF19 and AUX/IAA were found to be DE in the gravitropic plants. UBQ3 gene expression was significantly under‐expressed in the gravitropic rotated group compared to the other three treatment groups (Figure 6b and Table 1). These co‐expression networks highlight the elaborate interactions between different genes, which we explore further in the discussion to understand their role in adaptive evolution.

FIGURE 7.

Proposed molecular/cellular mechanism underlying differences between gravitropic and agravitropic phenotypes. In gravitropic plants (above), the TIR complex interacts with AUX/IAA, facilitating its transfer to UBQ3 for ubiquitination and subsequent degradation. This prevents AUX/IAA from inhibiting auxin response factors, thereby promoting the activation of auxin response elements crucial for gravitropic reactions. Conversely, in agravitropic plants (below), the AUX/IAA is not targeted for degradation by UBQ3. This enables it to bind and inhibit ARFs, suppressing the AuxREs' activation and hindering gravitropic responses. Created with BioRender.com.

4. Discussion

Our findings underscore the significant role of hormonal pathways, particularly auxin, ethylene and gibberellin, in the gravitropic response of S. lautus. In this study we have shown how natural selection can influence hormonal pathways during ecotypic divergence, revealing a diverse functional landscape linked to gravitropism in the coastal ecotypes of the S. lautus species complex. Functional analysis of the MAGIC population directly led us to identify candidate genes and biological pathways with distinct expression patterns in gravitropic versus agravitropic plants. Overall, our data showcases that differences in gravitropic responses are not merely the result of a few genes working in isolation but are instead orchestrated by a network of genes operating with one another.

While gene expression studies have inherent complexities in inferring phenotypic changes, our approach offers robust insights into the genetic basis of gravitropism. We focused on a developmental trait with clear candidate genes and performed control experiments to account for potential intrinsic differences like circadian rhythms. This strategy allows us to confidently interpret the gene expression differences we observed, even considering factors like alternative splicing and post‐translational modifications that can influence expression levels. Overall, our results highlight the complexity of adaptive evolution and underscore the importance of hormonal pathways in this process, providing a solid foundation for future studies on plant adaptation.

4.1. The Role of Hormones During Adaptive Evolution

A key finding of our study is the significant role of hormonal pathways, especially auxin, ethylene and gibberellin, in modulating gravitropic responses in S. lautus. The differential expression of auxin‐related genes such as ARF19, SAUR23, and IAA29 underscores auxin's pivotal role in regulating gravitropism in S. lautus. This is not surprising given the role of auxin in cell elongation and differentiation across plants (Aryal et al. 2020; Ma and Li 2019; Rayle and Cleland 1992), which is the underlying process driving asymmetrical cell elongation of the basal layer leading to gravitropic rotation of the hypocotyls. In model systems like A. thaliana, ARF19 and IAA29 form a heterodimer in the AUX/IAA PIB1 domain (Guilfoyle 2015) which are suggested to be ubiquitinated, releasing ARFs and regulating the expression of auxin‐responsive genes (Jiang et al. 2017) (Figure 7). In gravitropic S. lautus plants, ARF19 and AUX/IAA were found to be DE compared to the agravitropic plants at the comparative time points. ARF19 and ARF7 are transcriptional activators of auxin‐responsive genes (Okushima et al. 2005) through mediating histone modifications (Ito et al. 2016). Pleiotropic effects of hormonal pathways can, however, play an important role in driving local adaptation. Auxins, for example, not only regulate tropic responses but also modulate root architecture, flowering time, and stress responses (Aloni et al. 2006; Jing et al. 2023; Ke et al. 2018). This multifaceted influence makes hormonal pathways key drivers of adaptation, allowing plants to optimise growth and development in varying environments.

Previous studies of candidate genes in S. lautus have identified a range of hormone‐related genes during gravitropism. ENODL1 and ABA3 are likely contributing to a large percentage of the divergence in gravitropism between the Dune and Headland populations at Lennox Head (Wilkinson et al. 2021). In our study, ABA3 was upregulated in the F11 gravitropic population, indicating that this gene may also be DE as well as containing genetic differences, showing that this gene may be vital for differential gravitropic responses. Another study by James et al. (2023) looked more specifically at auxin‐related genes throughout the natural populations of S. lautus. This study found 50 auxin gene regions associated with gravitropism that have evolved repeatedly and independently, many of which overlap with the DE genes identified in our study, such as PINs, SAURs and WAT1. James et al. (2023) also found that auxin transport genes were enriched relative to other auxin‐related functions. Our results suggest that auxin‐related genes form a foundational mechanism guiding gravitropic responses in S. lautus. The repeated involvement of these pathways in adaptive traits across different species highlights their evolutionary significance. Future transplant experiments into the native Dune and Headland environments could elucidate allelic effects on transcription and physiology throughout the life cycle. In addition, our DEG analysis should be integrated with genetic variants through expression QTL mapping (Nica and Dermitzakis 2013; Zhabotynsky et al. 2022) or Transcriptome Wide Association Studies (TWAS). This would provide compelling evidence that our genes of interest are linked with the adaptive trait.

4.2. Interactions Between Hormonal Pathways During Adaptive Evolution

Our data directly supports the interconnected roles of auxin, ethylene and gibberellin pathways in gravitropic responses. This is evidenced by the coordinated upregulation in expression patterns we observed in our rotated plants compared to the control (Figure 4). Previous studies have also suggested an interplay between ethylene and auxin in gravitropic responses, which our data corroborates (Stepanova and Alonso 2019; Wheeler and Salisbury 1980). Ethylene is known to interact with auxin during gravitropic processes within both the roots and shoots of many plant species, such as A. thaliana (Harrison 2006), tomato species (Madlung et al. 1999). A range of auxin‐resistant mutants are also resistant to ethylene in both tomato (Kelly and Bradford 1990) and A. thaliana (Bennett et al. 1996), however it is contentious whether ethylene acts downstream of auxin (Madlung 1999), modulates auxin (Buer et al. 2006; Kong et al. 2024), or has multiple effects (Li 2008) during gravitropic processes. We found that the gene Ethylene Insensitive 3 (EIN3) was DE and highly connected with auxin‐related genes in the WGCNA networks (Figures 5a and 7). EIN3 is known to interact with MYC transcription factors and represses ABA responses during early seedling growth. In Brassicaceae, EIN3 is a family of genes with five homologues (EIL1–EIL5) which participate in a range of functions such as hypocotyl elongation, flowering time and many other developmental processes (Houben et al. 2022). Understanding these interactions is crucial for comprehending the broader regulatory networks at play.

While perhaps less central than auxin and ethylene, the involvement of gibberellin in gravitropism must be considered as it is well known to contribute to agravitropic phenotypes (Löfke et al. 2013; Ross and Wolbang 2008). As we found that putative gibberellin genes are DE and are involved in the WGCNA network, it hints at its potential role in amplifying gravitropic responses, possibly by modulating auxin transport or influencing cell growth. In rice, it was found that gibberellin promotes cell growth and main stem elongation by destabilising DELLA proteins, which inhibit cell growth (Ross and Wolbang 2008). Dwarfism often observed in coastal and alpine populations might be driven by divergence in key genes in the gibberellin (GA) pathway (Barboza‐Barquero et al. 2015; de Souza and MacAdam 2001; Liu et al. 2017), which are likely to be adaptations to their specific habitats (e.g., strong winds) and key to the origin of many plant ecotypes (Foster et al. 2007; Hesp 1991; Lowry 2012; Lowry, Rockwood, and Willis 2008). Additionally, in A. thaliana, asymmetric distribution of gibberellin was found to regulate PIN auxin transporters, which allows for further asymmetric auxin distribution underlying gravitropic responses (Löfke et al. 2013; Luo et al. 2015). These results suggest that interactions between hormonal pathways could be driving a suite of adaptive phenotypes across plants. However, the co‐expression analysis, while informative, cannot confirm direct interactions between genes in the coastal environments of S. lautus. These interactions need to be further confirmed with other experiments such as Yeast 2 hybrid experiments for detecting protein–protein interactions (Brückner et al. 2009).

Beyond individual genes and gene pathways, our study emphasises the role of gene co‐regulation in gravitropism. The observed co‐expression networks, particularly between auxin‐related and other hormone‐related genes, suggest a tight‐knit regulatory network underpinning the gravitropic response in S. lautus. These networks not only highlight the complexity of the genetic control of gravitropism but also underscore the potential for multiple genes to coordinate and drive adaptive responses. Plant genomes often contain gene families or duplicates that can perform similar functions (Flagel and Wendel 2009; Panchy, Lehti‐Shiu, and Shiu 2016). In gravitropism, even if gene expression is altered in one gene, other redundant genes might compensate for its function (Huo et al. 2018). This has been found in A. thaliana within the LAZY family of genes, which is vital for asymmetric auxin distribution during gravitropic responses (Jiao et al. 2020). Six LAZY genes have been identified that were functionally redundant. In comparison, in the soybean Glycine max, 15 genes could be identified in the LAZY family, with much of the variability between gene homologues related to where and when the genes are expressed (Jiao et al. 2020). This redundancy can add a layer of robustness to plant responses (Huo et al. 2018), ensuring that minor genetic perturbations don't disrupt critical processes like gravitropism or pleiotropic responses relying on hormonal signalling pathways often recruited during multiple stages of plant development (Panchy, Lehti‐Shiu, and Shiu 2016). Such co‐regulation highlights the importance of gene networks in maintaining functional responses under varying conditions.

4.3. Stress Responses and Their Potential Influence on Gravitropism

Our results align with recent reviews suggesting that complex gene networks, often mediated by hormonal pathways, control plant adaptation (VanWallendael et al. 2019). VanWallendael et al. 2019 also emphasise the significance of hormones, such as ABA, ethylene and jasmonates, in diverse stress responses suggesting that such hormones could also play a role in environmental adaptations in a range of plant species. The gene networks responding to environmental stresses, as highlighted by VanWallendael et al. (2019), suggest that the gene expression changes observed during gravitropism in S. lautus might be influenced by underlying stress responses or adaptive strategies. This may be intertwined with its gravitropic responses and possibly with stressors like salt, a common element in the rocky headland but not in dune soils (Roda, Ambrose, et al. 2013). Salt adaptation is likely to be an important trait in the local adaptation of S. lautus to the two environments and could potentially even be driving the evolution of gravitropism. On the headlands, salt spray from the ocean is salinising the soil where S. lautus grows. This extra salt is likely affecting cell elongation by affecting the ability of cells to absorb water (Colin et al. 2023), a vital step for gravitropic responses, as the stem is unable to bend without the asymmetric elongation of cells on the basal layer. Gravitropism in S. lautus may have even evolved through epistasis or as a trade‐off from salt tolerance.

Our study provides evidence for specific genes and pathways involved in gravitropism in S. lautus, offering insights into broader plant adaptations. Gravitropism is integral to how plants interact with their environment, influencing aspects like light capture, nutrient uptake, and even reproductive success (Bastien et al. 2013; Kawamoto and Morita 2022; Morita 2010). Our findings, which shed light on the genes and pathways driving this response, pave the way for a deeper understanding of how plants like S. lautus adapt and thrive in varying coastal environments. In the case of S. lautus, differences in gravitropic responses could directly impact how the plant accesses resources. If gravitropism also affects root growth directionality in S. lautus (Su et al. 2017), a common finding across seed plants (Zhang et al. 2019), it could influence water and nutrient uptake in specific environments (Bailey, Currey, and Fitter 2002), such as the sandy soils of dunes versus the rocky terrain of headlands. Root gravitropism has also been found to be divergent between some natural populations of S. lautus (Wilkinson 2019). The genetic underpinnings of gravitropism in S. lautus might have significant evolutionary consequences. If certain gene expression profiles confer advantages in particular habitats, these genetic variants could be positively selected over time. Recognising the evolutionary basis of expression divergence in future analyses will help us appreciate the adaptive significance of gravitropic responses in different habitats.

4.4. Interactions Between Genes Form a Model for Divergence Between Gravitropic Responses

As previously described, gravitropism ensures plants orient themselves in response to gravity through complex molecular pathways. Drawing from our differential expression data (Figure 4), we present a model that aligns with the current literature and observable phenotypic patterns from the MAGIC population. Our findings highlight the formation of an AUX/IAA gradient in gravity‐responsive cells driven by transporters (Chen et al. 1999; Rodrigo et al. 2011). TIR/AFB F‐box proteins are crucial in regions with high auxin concentrations within the SCF E3 ubiquitin ligase complex (Jiang et al. 2017). This complex then collaborates with AUX/IAA proteins to create a co‐receptor, leading to the breakdown of the repressor protein of AUX/IAA (Figure 7). This critical step activates ARF transcription factors, facilitating the transcription of auxin‐regulated genes. Additionally, our data points to the protein UBQ3—especially pronounced in rotated specimens—as pivotal in this process. Previous studies have shown that UBQ3 interacts with the 26S proteasome, promoting the degradation of Aux/IAA repressor proteins and enabling the expression of auxin‐responsive elements (AuxREs) (Jiang et al. 2017). Supporting this are our observations of DE genes such as AuxREs, SAURs, GH3.2 and AIPs (Table 1). Interestingly, agravitropic hybrids exhibit diminished UBQ3 levels, hinting at a potential disruption or absence of the ubiquitination essential for gravitropism. UBQ3 interacts with eight other genes in the co‐expression network, the highest number of within‐network interactions. However, functional validation of these interactions remains a subject for future study. While our model is rooted in prior studies and bridges molecular interactions with observable traits, its reliance on gene expression data can sometimes limit its functional implications. Hence, a crucial next step would be to test the applicability of this model in natural populations, offering insights into its broader relevance and any environment‐specific nuances. Although the MAGIC population harbours diversity from a coastal population pair in S. lautus, it only represents a fraction of natural genetic diversity in the species complex. Future work on MAGIC populations of Dune‐Headland pairs from other localities could reveal hormonal or functional convergence and divergence levels during adaptation to contrasting environments.

This study will be informative for causal network construction as networks can link genetic variants to variations in gene expression, shedding light on the directional network of adaptive phenotypes (Quiver and Lachance 2022). Finally, functional validation of our results is key, yet transformation efforts are incipient in S. lautus, suggesting that this standard of evidence will become available only in the future (He et al. 2013; Li et al. 2020). However, we anticipate that studying the effects of broader genomic features like epigenetic modifications or non‐coding RNAs in gravitropism remains within our grasp (Kamal et al. 2020; Yamamuro et al. 2016). Overall, the model we present offers a structured approach for future investigations into the genetic mechanisms of gravitropism, and as a byproduct, of adaptation.

5. Conclusion

In this study, we investigated the genetic mechanisms underlying adaptive divergence in gravitropic and agravitropic populations of S. lautus. Our findings reveal complex gene expression patterns associated with these adaptive differences, particularly in auxin and ethylene hormonal pathways. We demonstrated that adaptation in S. lautus is not driven by single genes acting independently, but rather by a coordinated network of multiple genetic elements. This aligns with previous studies in this system that have suggested a polygenic basis for adaptation. Based on our analysis of DE genes, we developed a molecular model of gravitropism in S. lautus involving the activation of auxin‐response factors. This model provides a framework for future studies to test its applicability in natural populations and explore how phenotypically similar populations may employ common or distinct functional solutions. Our results contribute significantly to understanding the role of gene expression in local adaptation of S. lautus and highlight the importance of interconnected hormonal pathways in adaptation to contrasting environments. To further advance our knowledge, future studies should focus on validating our findings in natural populations, investigating the potential influence of stress responses and other adaptive strategies on gravitropism, and exploring the role of epigenetic modifications and non‐coding RNAs. By integrating these approaches, we can gain a more comprehensive understanding of the complex genetic mechanisms underlying adaptive divergence in S. lautus and other plant systems. In summary, our study offers new insights into the genetic basis of adaptive divergence in S. lautus, highlighting the critical role of multiple interacting genetic elements and pathways in shaping plant adaptation to varied environments. This comprehensive understanding is essential for advancing our knowledge of plant adaptation.

Author Contributions

D.O.‐B. and Z.B. conceived and designed the experiment with input from F.B., and M.J.W., Z.B., M.J.W., and F.B. conducted the gravitropism experiments. Z.B. performed RNA extractions with the guidance of F.B.; H.J. conducted the transcriptome read mapping and Z.B. conducted RNA expression analyses with the advice and help of J.L., D.D., S.B., and D.O.‐B.; Z.B. wrote the first draft and completed the final version with input from M.J.W. and D.O.‐B. All authors gave meaningful feedback to later versions of the manuscript.

Conflicts of Interest

The authors declare no conflicts of interest.

Open Research Badges

This article has earned Open Data, Open Materials and Preregistered Research Design badges. Data, materials and the preregistered design and analysis plan are available at https://doi.org/10.48610/eaea206.

Supporting information

Data S1.

Data Availability Statement

The data and code underlying the methods and approaches used in this paper are available on GitHub, at https://github.com/ZoeBr/Broad‐et‐al.‐2023/tree/main.