Conditional Stomatal Closure in a Fern Shares Molecular Features with Flowering Plant Active Stomatal Responses

Stomata evolved as plants transitioned from water to land, enabling carbon dioxide uptake and water loss to be controlled. In flowering plants, the most recently divergent land plant lineage, stomatal pores actively close in response to drought. In this response, the phytohormone abscisic acid (ABA) triggers signalling cascades that lead to ion and water loss in the guard cells of the stomatal complex, causing a reduction in turgor and pore closure. Whether this stimulus-response coupling pathway acts in other major land plant lineages is unclear, with some investigations reporting that stomatal closure involves ABA but others concluding that closure is passive. Here we show that in the model fern Ceratopteris richardii active stomatal closure is conditional on sensitisation by pre-exposure to either low humidity or exogenous ABA and is promoted by ABA. RNA-seq analysis and de novo transcriptome assembly reconstructed the protein coding complement of the C. richardii genome with coverage comparable to other plant models, enabling transcriptional signatures of stomatal sensitisation and closure to be identified. In both cases, changes in abundance of homologs of ABA, Ca2+ and ROS-related signalling components were observed, suggesting that the closure response pathway is conserved in ferns and flowering plants. These signatures further suggested that sensitisation is achieved by lowering the threshold required for a subsequent closure-inducing signal to trigger a response. We conclude that the canonical signalling network for active stomatal closure functioned in at least a rudimentary form in the stomata of the last common ancestor of ferns and flowering plants. Significance Statement Stomata are valve-like pores that control the uptake of CO2 and the loss of water vapour in almost all land plants. In flowering plants, stomatal opening and closure is actively regulated by a stimulus-response coupling network. Whether active stomatal responses are present in other land plant lineages such as ferns has been hotly debated. Here we show that stomatal responses in the fern Ceratopteris richardii are active but depend on their past growth environment, and demonstrate that fern stomatal closure and sensitisation are associated with the altered expression of genes whose homologs function in the canonical stomatal regulatory network of flowering plants. Genetic pathways for active stomatal regulation therefore most likely evolved before the divergence of ferns and flowering plants.


Introduction
Stomata are pores present on the surfaces of plant leaves that control the uptake of CO2 and the loss of water vapour. The acquisition of stomata was a key land plant adaptation, which together with the development of a waxy cuticle and vascular system allowed early plants to colonise the terrestrial environment [ 1 ]. The ability to control CO2 uptake is important in the context of photosynthesis, whereas the regulation of evapotranspiratory water loss impacts on water and mineral nutrient accumulation in the aerial parts of the plant, protects against short periods of reduced soil water availability and provides leaf cooling capacity. In addition, stomatal closure provides protection against invasion by some pathogens. The regulation of stomatal aperture by light, humidity, atmospheric CO2, and the plant hormone abscisic acid (ABA) has been extensively investigated in the model flowering plant Arabidopsis thaliana. These studies identified networks of intracellular signalling proteins and second messengers that act in the guard cells of the stomatal complex, to couple extracellular stimuli to an opening or closing response [ 2,3,4 ]. Although stomata are present in all land plant lineages except liverworts [ 5 ], in which they have likely been secondarily lost [ 6 ], the question of when active stomatal closure mechanisms evolved remains hotly debated.
Studies of stomatal responses in non-flowering plant species have led to conflicting interpretations of underlying mechanisms. Observations that stomata close in response to ABA and CO2 in two moss species [ 7 ], in silico evidence suggesting that stomata evolved only once [ 6 ] and the presence of genes encoding components of the Arabidopsis stomatal closure pathway in moss genomes [ 8,9 ] could indicate that active stomatal responses are ancient. However, reports from non-flowering vascular plants have been contradictory. Some suggest that in lycophytes and ferns stomatal closure is hydropassive and guard cells are insensitive to closure-inducing signals such as ABA and high CO2 levels [ 10,11,12,13,14 ] whilst in contrasting reports stomata in the lycophyte Selaginella uncinata [ 15 ] and a number of ferns[ 16,9,17,18,19,20 ] were shown to close in response to ABA and/or CO2. The observation that stomatal closure in response to exogenous ABA is conditional on growth conditions in some (but not all) species of fern [ 17 ] may highlight the confounding variable in previous reports, but the underlying molecular basis for fern stomatal closure and conditional responsiveness is unknown. 4 The application of modern genetic analysis approaches to the fern lineage remains challenging, with few techniques and resources available. To identify the pathways underlying fern stomatal closure mechanisms, we utilised RNA-seq technology and de novo transcript assembly to generate and compare transcriptome profiles associated with different stomatal responses in the fern Ceratopteris richardii. It has previously been suggested that stomatal closure in C. richardii is not activated by the canonical ABA-induced response pathway found in flowering plants, based on the observation that stomatal conductance is reduced in response to increased water vapour deficit even in a mutant lacking a functional homolog of a receptor kinase required for ABA-induced stomatal closure in Arabidopsis [ 21 ].
However, C. richardii is a neotropical semi-aquatic fern that is routinely grown at high humidity (i.e. low water vapour deficit) in the laboratory [ 22,23 ], and stomatal responses may have been masked by conditional behaviours now known from other ferns [ 17 ]. We demonstrate herein that the stomata of C.
richardii can actively close in response to both low humidity and ABA but that this response is dependent on previous sensitisation by either low humidity or ABA. Our RNA-seq analysis provides evidence that the signalling network controlling stomatal closure in ferns contains homologs of ABA transport, Ca 2+ signalling and ROS signalling components from canonical Arabidopsis stomatal closure pathways, and suggests that sensitisation potentiates these same pathways to respond to closure stimuli at a lower threshold. We conclude that a core of regulatory networks that control active stomatal closure is conserved between ferns and flowering plants.

Active stomatal closure in C. richardii is conditional on pre-sensitisation by low humidity or exogenous ABA
To test whether active stomatal closure in C. richardii is conditional on growth environment, stomatal responses were compared between two humidity pretreatments. Plants grown either in constant high relative humidity (94.9 ± 0.36%) (referred to hereafter as wet-grown) or periodically exposed to low (ambient) humidity (48.3 ± 0.69%) (referred to hereafter as dry-pretreated) for ten minutes (see Materials and Methods). After an interval of growth at high humidity to ensure all stomatal were open, both pretreatment groups were exposed to low humidity for 120 minutes. Stomata of fronds from drypretreated plants exhibited a significant (p < 0.05) and progressive reduction in stomatal pore area in 5 response to low humidity at both 60 and 120 minutes (Fig. 1A), with mean pore areas of 79.7% and 54.0% of the mean area measured at 0 minutes, respectively (Dataset S1). Simultaneous application of exogenous ABA significantly enhanced this response (p < 0.05) at both 60 (66.9%) and 120 minutes (40.0%) (Fig. 1A). In contrast, stomata of wet-grown plants showed a significant reduction in pore area (p < 0.05) only after 120 minutes exposure to low humidity and of much lesser magnitude than in drypretreated plants (89.0% of mean pore area at 0 minutes), with no response to ABA (p > 0.05) (Fig.   1A). Active stomatal closure in dry-pretreated plants is further supported by a specific reduction in pore aperture width and not length (SI Appendix, Fig. S1, S2), and with a greater resilience to wilting than seen in wet-grown plants (SI Appendix, Fig. S3), with no change in stomatal density (SI Appendix, Fig. S1). When assays were repeated under constant high humidity the responses of stomata from dry-pretreated plants were much reduced, with mean pore area at 60 and 120 minutes 81.6% and 76.2% that of 0 minutes respectively, no significant change between 60 and 120 minutes (p > 0.05) ( Fig.1B) and no enhancement of closure by exogenous ABA (Fig. 1B). Stomatal responses of wetgrown plants were similar between 120 minutes constant high humidity (86.1% of mean pore area at 0 minutes) (Fig. 1B) and low humidity. Dry pretreatment was thus necessary to sensitise stomata for active closure in response to a low humidity stimulus, and low humidity was required to see any effect of an ABA stimulus on closure.
To test whether stomata could also be sensitised solely by ABA, wet-grown plants were pretreated with periodic application of exogenous ABA or a mock control solution (see Materials and Methods). After subsequent exposure to a low humidity stimulus, stomata from ABA-pretreated plants showed a significant reduction in stomatal area compared to mock-pretreated controls (p < 0.05) at 60 and 120 minutes (Fig. 1C), achieving the same magnitude of response by 60 minutes (51.1% of mean pore area at 0 minutes) as seen in dry-pretreated plant by 120 minutes (Dataset S1). Exogenous ABA treatment in addition to the low humidity stimulus did not enhance this response (p > 0.05, Fig. 1C), presumably because maximum closure had already been reached by 60 minutes. Stomata from mockpretreated plants were essentially unresponsive to low humidity, with a transient reduction in aperture area at 60 minutes only (p < 0.05) to 87.7% of mean pore area at 0 minutes (Fig. 1C). As with drypretreated plants, ABA-pretreated plants showed greater resistance to wilting compared with controls 6 (SI Appendix, Fig. S3), with no change in stomatal density (SI Appendix, Fig. S1). When assayed under high humidity the closure responses of stomata from ABA-pretreated plants were lost (p > 0.05, Fig. 1D), but exogenous ABA did elicit a minor closure response (p < 0.05) after 120 minutes (87.0%) ( Fig. 1D; Dataset S1). Stomata from mock-pretreated plants exhibited a similar degree of responsiveness under both low and high humidity (Fig. 1C,D; Dataset S1). Together these results show C. richardii stomata can be sensitised by ABA without low humidity, enabling subsequent active closure in response to either low humidity or ABA.

Homologs of Arabidopsis regulators of stomatal aperture and ABA responses are differentially expressed during stomatal sensitisation in C. richardii.
To identify genetic components of the mechanisms underlying stomatal sensitisation in C. richardii, RNA-seq was used to establish genome-wide transcript profiles in fronds of wet-grown, dry-pretreated, and ABA-pretreated plants ( Fig. 2A). The completeness of the predicted proteome contained within the de novo transcriptome assembled was similar to that of complete plant genomes and substantially more complete than the published C. richardii partial genome assembly [ 24 ] (Fig. 2B). The majority of protein-coding transcripts (68.7%) had identifiable homology to Arabidopsis genes (see Materials and Methods), of which 49% were direct orthologs to Arabidopsis genes (Fig. 2C). The quality of our de novo proteome was thus sufficient to identify conserved genetic networks between Arabidopsis and C.

richardii.
A comparison of transcriptome profiles in wet-grown (non-sensitised) versus dry-pretreated or ABApretreated (sensitised) fronds revealed 67 transcripts with abundance levels that were significantly different (p < 0.05) in wet-grown versus dry-pretreated fronds (Fig. 3A) and 6919 transcripts with significantly different (p < 0.05) levels in wet-grown and ABA-pretreated fronds ( Fig. 3A; SI Appendix,   Fig. S4). The large number of genes responding to ABA pretreatment was anticipated because there were pleiotropic effects on frond morphology (SI Appendix, Fig. S1) and ABA is known to impact on other processes that are unrelated to stomatal physiology. However, just 47 transcripts displayed significant differences in abundance (42 at p<0.01; 5 at 0.01<p<0.05) in both wet-grown versus drypretreated and wet-grown versus ABA-pretreated comparisons (Fig. 3A). These 47 transcripts, 11 of which had significantly higher levels in sensitised fronds and 36 of which had lower levels (p < 0.05) ( Fig. 3B; Dataset S2), thus represent the stomatal sensitisation signature in C. richardii.
To assess the likely function of sensitisation signature transcripts, homologs were identified in Arabidopsis, similarity to known proteins was determined, and gene ontology (GO) terms were assigned. Eighteen of the 47 sequences have homology to one or more Arabidopsis genes, 10 show significant similarity to known proteins in BLAST searches, 12 could be assigned GO terms and 19 have no identifiable similarity to any other sequences (Dataset S2). Of those assigned GO terms for 'biological process' 46% were associated with responses to biotic or abiotic stimuli, of those assigned for 'cellular component' 43% were related to membranes, and of those assigned for 'molecular function', 25% were transporters (Dataset S2). The majority of transcripts in these categories accumulated to higher levels in wet-grown than in pretreated fronds, indicating down-regulation during sensitisation. Notably, two transcripts are homologous to Arabidopsis genes with annotations linked to ABA and stomata, respectively (SI Appendix, Fig. S4), both encoding plasma membrane ABC transporters that regulate stomatal aperture in Arabidopsis. One is homologous to an ABCG subfamily protein (AtPDR12) that transports ABA in stomatal guard cells during closure [ 25 ] and the second is homologous to an ABCC subfamily protein (AtMRP4) that regulates stomatal aperture either upstream or independently of ABA [ 26 ]. A third transcript within the sensitisation signature is homologous to AtGRDP1, a negative regulator of ABA responses [ 27 ]. In both dry-and ABA-pretreated C. richardii fronds, the AtPDR12 and AtGRDP1 homologs are both down-regulated whereas the AtMRP4 homolog is up-regulated (SI Appendix, Fig. S4; Dataset S2). Collectively these data suggest that although many transcripts in the sensitisation signature have no known homologs in other species and may have species-specific functions in stomatal regulation, there are recognisable components associated with known stimulus-response signalling mechanisms in flowering plants, including membrane transporters that regulate the inter-and intra-cellular gradients that are crucial for stomatal closure.
Homologs of genes required for stomatal closure in Arabidopsis are associated with stomatal closure in C. richardii 8 To identify genes underlying fern stomatal closure mechanisms, transcript accumulation profiles were generated from fronds with different stomatal sensitivities, before and 60 minutes after exposure to a range of treatments ( Fig. 2A; SI Appendix, Fig. S5). Four treatments were carried out with drypretreated (sensitised) fronds: maintenance at high humidity with and without an ABA stimulus (no closure, Fig. 1B), exposure to a low humidity stimulus (closure, Fig. 1A), and simultaneous exposure to low humidity and ABA stimuli (enhanced closure, Fig. 1A). Transcriptomes were also generated from wet-grown (non-sensitised) and ABA-pretreated (sensitised) plants treated with a low humidity stimulus (wet-grown -no closure, Fig 1A; ABA-pretreated -closure, Fig 1C). For each of the six assays, transcripts were identified that were present at significantly different levels (p <0.01) before and after treatment (Dataset S3). Notably, fewer transcripts showed significantly different levels in the assay with ABA-pretreated fronds than in the other assays, possibly reflecting an altered baseline of expression at timepoint 0 as a result of ABA pretreatment, and/or more rapid closure than seen in drypretreated stomata (Fig. 1C) precluding the capture of at least some changes underlying the closure mechanism at the sampled 60 minute timepoint. Transcripts specifically associated with stomatal closure were identified by comparing transcript profiles between pairs of assays where stomata closed in one but remained open in the other (SI Appendix, Fig. S5), using multiple assay-pairs to filter out other experimental or environmental responses. A total of 1858 transcripts that showed a significant change in levels (p < 0.01) specifically upon stomatal closure were identified from four such paired datasets, with the majority being up-regulated on closure (SI Appendix, Fig. S5). Four of the 1858 transcripts were detected in all four paired datasets, 141 were present in at least three of the datasets and 877 were found in at least two of the datasets (Fig. 4A). Hereafter we refer to the 877 transcripts as stomatal closure-associated transcripts, with the 141 transcripts from within that set that are present in at least three datasets referred to as the stomatal closure signature.
To determine whether any of the closure associated transcripts are homologous to genes involved in Arabidopsis stomatal closure networks, stomatal and ABA annotations were mapped onto the whole transcriptome assembly (see Materials and Methods). Fifty-one of the 877 transcripts had homology to Arabidopsis genes with stomatal or ABA annotations, of which five were closure signature transcripts (Fig. 4B). To further assess possible gene function, the 51 transcripts were screened for homology to 9 14 Arabidopsis genes with proven roles in stomatal closure [ 2,28 ], all 14 of which had identifiable homologs within the whole transcriptome assembly (Dataset S4). These comparisons revealed that one of the five annotated signature transcripts was homologous to the SHAKER family of gated potassium ion channels that function in the guard cell membrane [ 29 ] (Fig. 4C). Homologs of aquaporin, CPK calcium-dependent protein kinase, respiratory burst oxidase homolog (RBOH) and annexin gene families were present within the remaining group of 46 annotated closure-associated transcripts ( Fig.   4C; Dataset S3). Although no canonical components of the ABA signal transduction pathway were identified, two families of ABA influx channels involved in stomatal closure were also represented (Fig.   4C). Collectively, these data show that homologs of genes in many of the signalling pathways that mediate stomatal closure in Arabidopsis are associated with stomatal closure in C. richardii, and thus suggest that closure mechanisms are at least partially conserved in ferns and angiosperms.

Changes in transcript abundance during stomatal closure and sensitisation in C. richardii are largely ABA-independent
To further dissect ABA-related mechanisms underlying stomatal closure in C. richardii, we exploited the fact that stomata in dry-pretreated plants only close in response to ABA when the ABA treatment is carried out in conditions of low humidity (Fig. 1A,B). ABA-related changes in transcript abundance that are specific to stomatal closure should thus be observed following ABA treatment under low humidity but not following treatment under high humidity. Transcript levels that changed in response to ABA under low and high humidity conditions were identified by comparing transcriptomes of fronds that were exposed to either a mock or ABA treatment solution (SI Appendix, Fig. S6). The abundance of 578 and 607 transcripts was significantly altered (p < 0.01) in response to ABA under low and high humidity conditions, respectively. Levels of two of the 141 closure signature transcripts changed after ABA treatment under high humidity (representing an unknown protein and a UDP-glucose 6dehydrogenase, Dataset S5), but none changed after treatment under low humidity (Fig. 5A).
Abundance of a single closure-associated transcript (representing a polyphenol oxidase) was influenced by ABA under both humidity conditions but levels changed in opposite directions in the two conditions (SI Appendix, Fig. S6; Dataset S5). Levels of just 13 closure associated transcripts changed in response to ABA treatment specifically under low humidity (SI Appendix, Fig. S6; Dataset S5). These results suggest that ABA-induced closure of fern stomata (Fig. 1) is mediated primarily via mechanisms operating at translational or post-translational levels.

Transcriptome profiles during stomatal sensitisation and closure in C. richardii suggest that sensitisation is associated with the down-regulation of closure-associated genes
To determine whether stomatal sensitisation and closure mechanisms share any common features, the identities of sensitisation signature transcripts were compared to those of closure associated transcripts. Twenty four of the 47 sensitisation signature transcripts were present in the closure associated transcript dataset, with two also being present in the much smaller closure signature dataset ( Fig. 5B). Fourteen of these 24 common transcripts have similarity to known proteins (Dataset S3), including homologs of the ABCG transporter AtPDR12 (Fig. 4C), AtGRDP1, a peroxidase superfamily protein, and two calcium-binding EF-hand family proteins. Notably, all 24 transcripts decreased in abundance during sensitisation and increased in abundance during closure (Fig. 5C), returning to a similar level as seen in untreated wet-grown plants (p > 0.05; SI Appendix, Fig. S7; Dataset S3). The fact that more than 50% of the sensitisation signature transcripts are also stomatal closure associated transcripts suggests that stomatal sensitisation acts by priming the closure mechanism.

Discussion
Stomata facilitated the adaptation of plants to land and with the exception of liverworts, all extant land plant lineages rely on stomatal function to modulate the exchange of air and water with the surrounding environment. Despite the likely monophyletic origin of stomata and their conserved function across these extant groups, the evolution of their regulatory networks is disputed, with discussions focused on whether the ABA-dependent active stomatal closure mechanisms found in flowering plants [ 2,30 ] are conserved in non-flowering plant lineages. We show that stomatal closure in the model fern C. richardii is active, but is conditional on sensitisation by prior exposure to either low humidity or ABA. To identify the mechanisms underlying these processes, we utilised RNA-seq to compare transcriptomes of whole fronds that were exposed to different conditions under which stomata did or did not close, generating a high-quality de novo assembly from which we could discriminate signatures associated with stomatal responses. These show that both stomatal sensitisation and closure in C. richardii include changes in the expression of genes required for active stomatal closure in the flowering plant Arabidopsis, suggesting that guard cell signalling mechanisms are conserved. Active stomatal closure responses are thus likely to have evolved in the last common ancestor of land plants.

Conserved active stomatal closure mechanisms in ferns and flowering plants
The requirement for sensitisation to reveal active stomatal closure responses, shown here in C.

Stomatal sensitisation in C. richardii shares features with stomatal acclimation in flowering plants
Both dry and ABA pretreatment are sufficient to condition stomatal sensitivity in C. richardii, with both treatments inducing similar genome-wide changes in transcript abundance that enable an active closure response on subsequent exposure to low humidity. The stomata of flowering plants can also be sensitised or de-sensitised to closure-inducing signals, through the process of acclimation. For example, newly-developed Arabidopsis stomata become increasingly sensitive to closure-promoting environmental stimuli during leaf maturation in an ABA-dependent manner [ 42 ] whereas species transferred to high humidity conditions show impaired stomatal closure when subsequently exposed to lower humidity, despite accumulation of endogenous or exogenous ABA [ 43,44 ]. In flowering plants, the mechanism underlying altered guard cell sensitivity to ABA is thought to involve altered abundance of the ABA receptors PYL/PYR/RCAR[ 30 ]. Although there was no evidence of increased ABA receptor abundance in sensitised C. richardii fronds, the possibility that the whole-frond dataset was not sufficiently sensitive to detect all guard cell-specific changes cannot be excluded. However, reduced abundance of an AtGRDP1 homolog [ 27 ] was detected in sensitised fronds. Given that AtGRDP1 negatively regulates the expression of the ABA response regulators ABI3 [ 45 ] and ABI5 [ 46 ], and that loss of function leads to ABA hypersensitivity [ 27 ] it is thus possible that stomatal sensitisation in C. richardii 13 results from increased sensitivity to ABA. As such, conditional stomatal sensitivity in ferns and stomatal acclimation in flowering plants may both act by altering the threshold at which subsequent inductive signals, including ABA, can trigger stomatal closure.
Consistent with the idea that stomatal sensitisation lowers the threshold for induction of stomatal closure, the abundance of 24 closure-associated transcripts is reduced during sensitisation, with all 24 returning to the same levels seen prior to sensitisation upon subsequent induction of closure. This set of 24 includes all transcripts in the sensitisation signature that have calcium-binding and ROS-related annotations, plus homologs of AtGRDP1 and AtPDR12, which is an ABCG transporter that increases guard cell uptake of ABA [ 25 ]. On the basis of these observations, we speculate that sensitisation results when Ca 2+ , ROS and/or ABA signalling cascades within the guard cell are potentiated to reduce the threshold required for a stimulus to trigger closure. Alongside this down-regulation of closureassociated genes, it is notable that a putative homolog of AtMRP4 is conversely up-regulated during sensitisation. AtMRP4 is a guard cell-expressed ABCC transporter that negatively regulates stomatal opening in an ABA-independent manner [ 26 ]. Collectively, our results support a scenario whereby fern stomata are sensitised by altering the intracellular physiological status of guard cells, enhancing the ability to respond to canonical closure signals, and attenuating competing opening responses, using mechanisms conserved with flowering plants.

Plant material and growth conditions
All experiments were performed using Ceratopteris richardii wild-type laboratory strain Hn-n [ 47 ]. Sporophytes were grown to 46-50 days old before sampling fronds for both stomatal response assays and RNA-seq. Plants were pretreated with either periodic low humidity or exogenous ABA application for a period of three weeks after transplant, then grown without further pretreatment for two weeks.
During pretreatment, humidity conditions within magenta boxes were measured across three replicates

Stomatal response assay
The responses of stomata from humidity-and ABA-pretreated plants were tested in separate assays.
Plants undergoing stomata response assays were treated with 100uM ABA or mock solutions (formulation as pretreatment) at the start of the assay (0 minutes), applied by foliar spray in a design orthogonal to humidity or ABA pretreatments such that each pretreatment+treatment combination was represented by three plants per assay. Plants were either immediately exposed to low humidity by removal from Magenta boxes into ambient air (48.3±0.69%), or in control experiments maintained at high humidity by immediately resealing Magenta boxes after ABA/mock treatment. At three timepoints (0, 60 and 120 minutes from the assay start) a single frond was sampled from one plant for each pretreatment+treatment combination and photographed to record stomata morphology, sampling an independent plant at each timepoint within each pretreatment+treatment combination, selecting fronds at shoot position 7-9, bearing 3 lobes (SI Appendix, Fig. S1). Stomata morphology was recorded by imaging was under brightfield microscopy at 10x magnification using an Olympus BX51 microscope (Olympus Corporation, Hamburg, Germany) with attached Qimaging MP3.3-RTV-R-CLR-10-C MicroPublisher camera and QImage Pro software package (Teledyne Photometrics, Tuscon, USA).
Photography at each timepoint was completed within 15 minutes, and pretreatment+treatment combinations were imaged in the same order between timepoints. Three separate regions were photographed on each frond (one within each lobe), and 20 stomata within each region were selected systematically for measurement, totally 60 stomata per plant per timepoint. Stomatal measurements were made from scaled images using the FIJI software package [ 50 ]. Assays were repeated three times for each combination of pretreatments and low or high humidity stimulus. Whole-plant photography was performed using a Cybershot DSC-HX7V digital camera (Sony, Tokyo, Japan). During figure preparation images were adjusted for brightness and contrast using Photoshop 2019 (Adobe, San Jose, USA).

RNA extraction and sequencing
The sampling of frond tissue for RNA-seq was performed by snap-freezing single fronds in liquid nitrogen from specific pretreatment+treatment combinations at 0 or 60 minutes ( Fig. 2A). Three biological replicates were harvested for each pretreatment+treatment combination, taken from three independently-performed assays. RNA was extracted using the RNeasy Plant mini kit with on-column DNase treatment (QIAgen, Hilden, Germany). RNA quality was validated by Agilent 2100 Bioanalyser RNA 6000 Pico assay (Agilent, Santa Clara, USA). Libraries were prepared from single frond samples using 1µg starting RNA using the TruSeq RNA library prep kit V2 (Illumina Systems, San Diego, USA) and quantified by KAPA library quantification (Roche, Basel, Switzerland) on a CFX96 qPCR machine (Bio-Rad, Hercules, USA). Paired-end sequencing was performed by HiSeq2000 sequencing system using a NextSeq 500/550 High Output kit (150 cycles) (Illumina Systems, San Diego, USA).

Transcriptome assembly and analysis
Paired end reads were processed using rCorrector [ 51 ] with default parameters to correct likely erroneous k-mers. The Harvard FAS Informatics "FilterUncorrectabledPEfastq" script (https://github.com/harvardinformatics/TranscriptomeAssemblyTools, accessed 13 January 2020) was used to discard read pairs for which one of the reads was deemed unfixable. TrimGalore (https://www.bioinformatics.babraham.ac.uk/projects/trim_galore) was used with default settings to trim adapter sequences and low-quality bases identified using the tools Cutadapt [ 52 ] and FastQC [ 53 ].
Reads were mapped to the SILVA rRNA SSUParc and LSUParc databases, release 132[ 54 ] using Bowtie 2[ 55 ] with option "--very-sensitive-local" and those pairs for which neither read was mapped to the rRNA database were retained. Analysis using FastQC showed that the retained reads had per base sequence content deviations greater the 5% from the mean for the first 7 and last 3 bases of the reads and so these were clipped using TrimGalore. The transcriptome was assembled using Trinity[ 56 ] with default settings. Transcript abundance estimation was performed using Salmon [ 57 ] within the Trinity pipeline. Differential expression testing was performed using DESeq2 [ 58 ] (options "-min_reps_min_cpm 2,1") also with the Trinity pipeline. All differential expression comparisons between sample pairs are given in Datasets S2, S3 and S5. Venn diagrams were generated using the InteractiVenn web tool [ 59 ].
A proteome was constructed by analysing the transcriptome using Transdecoder (http://transdecoder.sf.net) "LongOrfs" and "Predict" with the Trinity gene-to-transcript map supplied as input and with default parameters. Amino acid sequences were additionally identified for sequences that were expressed above the minimum threshold for the differential expression analysis (> 1 CPM in at least two biological replicates) but which had not yet had an amino acid sequence identified. This was achieved using "TransDecoder.LongOrfs" with option "-m 10" to set a minimum length of 10 [ 61 ]. Each orthogroup inherited the annotations of the Arabidopsis genes that it contained. The same reference proteomes were used to assess completeness of the predicted C. richardii proteome using BUSCO analysis [ 79 ].
To identify homologs of the annotated protein coding genes a series of searches were carried out, terminating at the first success for each gene. These searches were for: an ortholog in Arabidopsis identified by OrthoFinder, an Arabidopsis homolog from within the same OrthoFinder orthogroup, any significant DIAMOND [ 80 ] hit to an Arabidopsis homolog (e < 10 -3 ), and significant hit for a homolog within the 12 plant species above, any significant hit within the BLAST nr database [ 81 ].

Data availability
The sequencing reads obtained from each library have been deposited in the NCBI Sequence Read      Dataset S1. C. richardii stomatal response measurements.
Dataset S5. C. richardii transcripts responsive to ABA under either low or high humidity.   A. Venn comparison of genes for which transcript abundance is significantly different between wetgrown and dry-pretreated fronds (green) or wet-grown and ABA-pretreated fronds (yellow) at two levels of stringency (p < 0.01 and 0.01 > p < 0.05, as shown).

B.
Venn comparison between the abundance responses for each transcript identified with a significant change in transcript abundance (p < 0.05) within each pretreatment in (A). The central value in grey represents all remaining transcripts within the assembly that do not change abundance in response to either pretreatment.
28 A. Venn comparison of transcripts accumulating to significantly different levels (p < 0.01) in association with stomatal closure between four different closure datasets (as shown). Within each dataset, transcripts specific to stomatal closure were identified by comparison with similar experimental 29 conditions in which stomata did not close (SI Appendix, Fig. S6). A total of 877 transcripts are associated with closure in at least two datasets (bold), 141 of which are present in at least three datasets and are thus referred to hereafter as closure signature transcripts (bold, underlined).  A. Venn comparison between stomatal closure signature transcripts (bold, underlined, see Fig. 4A), all transcripts with annotations relating to ABA and stomata functions, and all transcripts found to significantly change abundance (p < 0.01) in response to ABA treatment under low or high humidity.

B. Venn comparison between the 877 closure-associated transcripts (labelled in bold in
Transcript levels that changed specifically in response to ABA were identified by comparing abundance at 0 and 60 minutes between mock and ABA treatments (SI Appendix, Fig. S6). No overlap between the stomatal closure signature and ABA-responsive transcripts was found.

B.
Venn comparison between transcripts identified as stomatal closure-associated transcripts (bold) and the stomatal sensitisation signature (see Fig. 3A). Underlined numbers given in brackets denote closure signature transcripts from within the closure-associated transcript dataset.
C. Venn comparison between the directional change in level of stomatal closure-associated (bold) and sensitisation signature transcripts during stomatal closure and stomatal sensitisation, respectively.
Underlined numbers given in brackets denote closure signature transcripts within the closure-31 associated transcript dataset. Transcripts common to both sensitisation and closure datasets are all down-regulated in sensitised fronds relative to wet-grown controls and are all up-regulated during closure.