Touch signaling and thigmomorphogenesis are regulated by complementary CAMTA3- and JA-dependent pathways

To obtain better insight into which genetic factors contribute to touch-induced transcriptional responses, we performed a survey to select promising candidate genes (Table 1). MYC2, MYC3, and MYC4 were previously identified as crucial regulators of the JA-dependent transcriptional responses to water spray and mechanical stimulation (18). Recently, a quadruple myc2 myc3 myc4 myc5 mutant was reported, which also has lost the function of the fourth member of this clade of MYC2-related bHLH transcription factors (19). WRKY40 and its close paralogs WRKY18 and WRKY60 have been implicated in a range of responses, e.g., to abscisic acid (ABA) (20) and retrograde signaling (21), but are also among the most touch responsive genes in Arabidopsis (18, 22, 23). Oxidative signal-inducible (OXI1) kinase has been implicated in mediating oxidative bursts to downstream signaling, making it of interest as touch responses are believed to involve early reactive oxygen species (ROS) signals (24). Touch-responsive phosphoprotein 1 (Treph-1) was picked up in a screen for touch-induced protein phosphorylation targets and was shown to be important for thigmomorphogenesis and regulating specific genes of the transcriptional touch response (25). In addition, Feronia receptor-like kinase, which binds to rapid alkalinization factors (RALFs), has been implicated in a wide range of plant responses, including immune signaling, development, flowering, cell wall integrity, pollen tube growth, and stomatal movement, as recently summarized (26). Moreover, it was suggested to mediate responses to osmotic swelling, which is considered as a type of mechanical stimulation (16). In particular, it was noted to be required for full induction of TCH2 and TCH4 touch marker genes after hypo-osmotic shock. Its homolog Theseus1 (THE1) is a receptor-like kinase that also binds RALF factors and is involved in cell wall integrity signaling (27, 28). The Jumonji domain protein ethylene insensitive 6 (EIN6) was picked up in a screen for ethylene insensitivity (29). It has also been reported to be required for induction of TCH3 by mechanical stimulation (30). The glutamate receptor-like (GLR) proteins are crucial receptors for the neurotransmitter glutamate in animal nervous cells (31). In plants, they are thought to operate as cation channels transporting Ca²⁺ across the plasma membrane, which may be activated by pH and glutamate released actively or from, e.g., wounded cells (32–34). In particular, GLR3.3 and GLR3.6 have been shown to mediate systemic signaling via electric potentials and Ca²⁺ waves, thereby affecting JA-induced gene expression (33, 34). Given the close similarity between the transcriptomes of wounding and touch signaling (23), and the involvement of JA and Ca²⁺ in touch responses, the glr3.3 glr3.6 double mutant was included.

Besides a literature search, we also studied a range of datasets to find clues to previously unidentified players in touch signaling. In one approach, we selected the top 50 most induced Arabidopsis genes 25 min after water spray in Col-0 (table S1) (18). This set of genes was searched for overrepresented transcription factor binding sites using the TF2Network online tool (35). In total, 108 motifs were identified as overrepresented (table S2). At the top of the list, we found three similar motifs that were predicted to be recognized by the calmodulin-binding transcription factor CAMTA3. CAMTA3 has been found to be involved in rapid stress responses, by activating the rapid stress response element in wound-responsive promoters (36). CAMTA3 is furthermore of interest given the role of Ca²⁺ in early touch signaling (24) and was thus selected for further characterization. Further down, the list was dominated by (putative or experimentally verified) motifs recognized by in total 51 unique WRKY transcription factors including AtWRKY40, supporting our choice to include the wrky18 wrky40 wrky60 mutant in our screen.

Previous studies could not detect changes in touch induced–gene expression in mutants in mechanosensitive channels such as mechanosensitive channel–like (MSL) multiple mutants (18) and Ca²⁺-permeable mechanosensitive channels mca1 mca2 mutants (23). A different class of MS channels that were initially characterized in animal systems is Piezo-type proteins (37). We searched the Arabidopsis genome and could only identify a single homolog (At2g48060; Piezo; PZO1), which at the start of this study had not been characterized. The Piezo protein is unusually large with 2462 amino acids, resulting in a molecular weight of 282 kDa. To study its function in mechanoperception, we obtained two transferred DNA (T-DNA) mutants in the At2g48060 locus and screened for homozygous mutants. The piezo-2 allele has previously been shown to be a null mutant (38). Quantification of the PIEZO transcript in piezo-3 (SALK_003004) mutants by quantitative real-time polymerase chain reaction (qRT-PCR) also confirmed a clear lack of mRNA (fig. S1). Both mutant alleles grew vigorously and did not show any obvious growth defects when grown on soil. In addition, when grown on standard horizontal MS agar plates, no clear differential phenotypes could be detected, suggesting that Piezo proteins are largely dispensable for normal growth in optimal conditions in plants. We were nevertheless interested in whether molecular phenotypes could be observed in the transcriptome after touch treatment.

One of the most remarkable aspects of the transcriptional response to touching is not only its rapid induction but also its steep decline. The peak expression for most touch-responsive genes is around 22 to 25 min after treatment, yet the expression levels for many genes are down to nearly preinduction levels by 40 min (Fig. 1) (18, 23). This rapid decrease in mRNA levels within 10 to 15 min suggests an active mechanism of mRNA degradation, rather than a general unregulated decay. The half-life of some classic touch-inducible transcripts after inhibition of RNA polymerase by actinomycin D is much longer, e.g., 106 min for AtWRKY40 and 4.5 hours for TCH4 (39). We thus also looked for factors that may regulate this rapid mRNA decay. Two CCR4-associated factors Caf1a and Caf1b are consistently induced by touch treatments (18, 22). CCR4-CAF protein complexes are involved in mRNA destabilization by removing the poly-A tail of mRNAs and were reported to affect deadenylation of wound-induced transcripts (40). Therefore, the caf1a caf1b double mutant was obtained and included in this study.

Another mechanism of mRNA degradation is by removal of the protective 5′ 7-methyl guanosine mRNA cap (decapping). The scaffolding protein Varicose (VCS) facilitates mRNA decapping (41, 42), exposing the mRNA to 5′ to 3′ exonuclease activity by, e.g., exoribonuclease XRN4 (43), while the suppressor of VCS (SOV) protein degrades the RNAs in 3′ to 5′ direction (44). A recent study assessed the contribution of mRNA degradation enzymes VCS and SOV after inhibition of transcription with cordycepin (44). Col-0 has a natural null mutation in SOV, so the vcs mutant in Col-0 is effectively a sov vcs double mutant. We analyzed the RNA sequencing (RNA-seq) dataset comparing vcs and sov mutants to Col-0 (technically sov) and Col-0 with a restored SOV gene (“wild type”), specifically focusing on well-known touch-responsive transcripts (fig. S2 and table S3). Notably, the transcripts of, e.g., TCH2-4, JAZ8, WRKY40, and Caf1a were much more stable in the vcs and vcs sov mutants compared to wild type and sov. The mRNA degradation profiles of vcs and vcs sov were overall very similar, indicating that VCS is the main contributor. Thus, at least under controlled conditions and after inhibition of transcription, VCS seems to be important for efficient degradation of many core touch-responsive transcripts. Therefore, the vcs mutant, as well as the xrn4 mutant, were included in the screen.

The transcriptomic responses to touch were assessed in the collection of mutants selected above (Table 1). For touch treatment, plants were gently brushed with a paint brush. Twelve-day-old in vitro grown seedlings were mechanostimulated and samples were collected just before, or 22 or 40 min after touch treatment, and transcripts were measured using qRT-PCR. We collected samples at 22 min after touching, which is around the peak of expression for many touch-responsive genes (18), to assess mutants for their capacity to induce the transcriptional touch responses. We also collected samples at 40 min after touching, when transcript levels had nearly returned to basal levels, to assess whether mutants were affected in the rapid degradation of mRNA (Fig. 1 and figs. S3 to S5). A range of touch-inducible transcripts were selected for qRT-PCR analysis on the basis of previous studies.

The myc2 myc3 myc4 myc5 mutant showed reduced induction of known MYC2 target genes such as bHLH19, ERF109, and CML39 relative to wild type (Col-0) at 22 min, whereas TCH4 transcript levels were induced as in wild type (fig. S3). Overall, the quadruple mutant responded similarly to the triple myc2 myc3 myc4 mutant reported previously (18), indicating that MYC5 likely has a very minor contribution. In contrast, the wrky18 wrky40 wrky60 triple mutant showed a wild-type–like response for all the genes we tested, covering JA-dependent (e.g., bHLH19 and ERF109) and JA-independent transcripts (TCH3 and TCH4). Mutants in the glutamate-like receptors glr3.3 glr3.6 involved in systemic Ca²⁺ and JA signaling also showed a wild-type–like touch response (fig. S4). The ein6 mutant was previously reported to be deficient in TCH3 induction in response to mechanical stimulation (30); however, in our study and experimental conditions, the mutant actually responded slightly stronger to touch than wild type (Ler-0 in this case) at a transcript level for TCH3, while other tested touch marker genes bHLH19 and TCH2 responded like wild type. Mutants in TREPH1, a protein phosphorylated in response to mechanical stimulation, were previously reported to have impaired touch-induced gene expression of CML38, ERF11, and JAZ7 (25). Using the same treph1-1 T-DNA insertion mutant, we could not see significant alterations in touch-induced gene expression of ERF11, JAZ7, or CML38 in comparison with the touch-treated Col-0 under our growth conditions and using our experimental setup (fig. S5). In addition, treph1-1 plants showed wild-type–like responses for other touch-induced genes like TCH4, Caf1a, bHLH19, JAZ8, TCH2, and OM66.

We also assessed the touch-induced gene expression patterns in mutants of related receptor-like kinases Feronia (FER) and Theseus1 (THE1), and OXI1 kinase. Although the1 and oxi1 mutants displayed wild-type–like touch responses in bHLH19, TCH3, TCH4, and ERF109 expression (fig. S3), clear differences with Col-0 could be observed in fer-4 mutant plants (Fig. 1). Already in untouched control seedlings, we observed significantly higher transcript levels of touch marker genes including bHLH19, CML39, and JAZ8 in fer-4 compared to Col-0. Moreover, a hyperinduction 22 min after brushing treatment could be observed compared to touch-treated Col-0 for bHLH19 and CML39. This elevated expression compared to Col-0 continued 40 min after touching for bHLH19, CML39, JAZ7, JAZ8, and ERF11, while in Col-0 their expression had returned to near untouched control levels. However, the expression levels of TCH2 and TCH4 in fer-4 remained similar to Col-0 in untouched control, as well as after 22 and 40 min of induction (Fig. 1). These results indicate that Feronia kinase is a strong negative regulator of touch-inducible gene expression for several genes that have been associated with JA-responses like JAZ8, while JA-independently regulated genes (e.g., TCH2 and TCH4) do not appear to be affected directly.

Conversely, touch-induced gene expression of bHLH19 was strongly repressed in untouched control and at 22 min after touch treatment in mutants of the Ca²⁺-signaling related transcription factor CAMTA3 (Fig. 1). While most other touch-induced genes were not significantly affected, TCH2 induction was significantly lower at both 22 and 40 min after touching relative to touched Col-0. This suggests that CAMTA3 is affecting a separate branch of the touch-induced transcriptomic response. bHLH19 touch-induced expression also appeared suppressed at 22 min in the piezo mutant (Fig. 1), although the difference did not meet P value criteria in the two-way analysis of variance (ANOVA) analysis. The induction of CML39 appeared to be significantly repressed in the piezo mutant.

Last, we assessed the rapid reduction of mRNA levels observed at 40 min in Col-0 in the mutants with defects in mRNA decapping (vcs), deadenylation (caf1a caf1b), and mRNA degradation (xrn4). Despite the reported roles of these enzymes in mRNA stability, the mutants had a normal mRNA expression pattern for all tested touch-responsive genes at 40 min. As we had expected the expression of touch-induced genes to remain elevated for longer especially in the vcs-7 mutant based on the previous mRNA stability study (fig. S2) (44), we assessed a second allele vcs-8. In addition, in this mutant, mRNA levels declined as normal at 40 min (fig. S5).

As the impact of loss of Piezo expression on the overall transcriptome before and after touch is poorly understood, we performed a genome-wide RNA-seq transcriptomics study comparing Col-0 and piezo mutants before (i.e., untouched) and 22 min after touch (brushing) treatment (fig. S6 and table S4). Under untreated conditions, 160 genes were differentially expressed compared to Col-0, of which 82% was less expressed in the piezo (SALK_003004) mutant, including the Piezo gene itself. No significantly overrepresented Gene Ontology (GO) category could be identified and only a relatively small portion of the genes (22) were part of the previously defined 1671 “core touch-responsive” genes that were commonly differentially expressed in a wide range of mechanical stimulation–related whole transcriptomes (fig. S6A) (18). Two related MADS box transcription factors Sepallata1 (AtSEP1) and Sepallata3 (AtSEP3), which redundantly regulate flower development, were approximately five- to sevenfold down-regulated (fig. S6B) (45). In addition, Terminal Flower 1 (TFL1) involved in inflorescence identity determination (46) was approximately 2.5-fold down-regulated. Two ROXY-type glutaredoxins ROXY13 and ROXY14 were down-regulated and cysteine-rich receptor kinase 30 (CRK30). Four protein-encoding genes were more than twofold significantly up-regulated in piezo mutants, including expressed protein At3g01345 (153-fold), a nodulin transporter Umamit25 (6-fold), and a ROX1-like repressor (3-fold).

A total of 2265 genes were differentially expressed 22 min after brushing in Col-0 (padj < 0.05, 1.5× fold change). Of these, 70% were up-regulated and 30% were down-regulated (Fig. 2A). Nearly 50% of the differentially expressed genes (DEGs) after brushing in Col-0 were considered as commonly responsive to mechanical stimulation (core touch-responsive genes) as defined previously (18), demonstrating that a successful mechanical stimulation was achieved. We then looked for genes that showed altered touch-responsive gene expression in piezo mutants compared to Col-0 at 22 min after brushing (fig. S6C) (padj < 0.05). Only nine such genes could be identified, including bHLH19 as observed in the qRT-PCR experiment (Fig. 1) and the (mechanical) stress-activated transcription factor WRKY48, which has been shown to repress plant basal defense (47). Both transcription factor genes were less induced after touching in the piezo mutant compared to in Col-0. In conclusion, our findings indicate that the unique Piezo protein in the Arabidopsis genome does not appear to have a crucial role in regulation of mechanosensitive gene expression in seedlings.

RNA-seq analysis revealed strong effects on gene expression in the fer-4 mutant compared to Col-0 under untreated conditions. In total, 1035 genes were differentially expressed, with 456 genes more highly expressed and 579 genes with reduced expression (Fig. 2B). As observed in the qPCR, 164 of these were core touch-responsive genes such as JAZ8, CML39, lipoxygenase 4 (first step in JA biosynthesis), and ERF11, with around two-thirds showing higher basal expression in fer-4 mutants (table S5). GO signatures of wounding, hypoxic, and biotic responses were up-regulated among these touch-responsive genes, while response to salicylic acid (SA) appeared down-regulated. However, TCH2/3/4 gene expression was not significantly affected. Among the non–core touch-responsive genes that were up-regulated in fer-4, we found an up-regulation of iron starvation and iron response–related genes, suggesting that the lack of Feronia triggers an iron starvation–like response. Among non–core touch-regulated genes, we found that genes involved in cell wall organization, GA response, and gravitropism were down-regulated.

In total, 411 genes had an altered touch response in the fer-4 mutant (Fig. 2, C and D). Most genes showed a stronger touch responsiveness. Among these hypertouch-sensitive genes, we observed a wide range of JA biosynthesis components such as Lipoxygenases LOX3 and LOX4, allene oxide cyclase AOC3, and 12-oxophytodienoate reductase OPR3 (Fig. 2E). In addition, key JA signaling components such as MYC2 and 5 JAZ/TIFY genes were much more induced after touching in fer mutants. In agreement, GO analysis indicated an overrepresentation of JA response, response to fungus, wounding, and (a)biotic stress responses. This indicates that Feronia plays a significant role in touch-responsive gene expression by repressing at least part of the JA-response, rather than by activating touch response signaling as proposed previously for mechanical stimulation by hypo-osmotic swelling (16). In our hands and growth conditions, and using whole seedlings, we did not observe the reduced expression of, e.g., TCH2, CML49, and WRKY33 in fer-4 before or after touching as reported previously for osmotic swelling in root tissue (16).

The qRT-PCR analysis of selected touch marker genes revealed that the transcription factor CAMTA3 is required for full touch-induced expression of bHLH19 and TCH2 (Fig. 1). The contribution of CAMTA3 in genome-wide transcriptional response to touching was further assessed using RNA-seq analysis (Fig. 3 and table S6). Under untreated conditions, we observed that 516 genes were differentially expressed in camta3-1 mutants compared to Col-0 (padj < 0.05, 2× fold change), and most of which (80%) were lower expressed in camta3-1 than in wild type (Fig. 3A), including CAMTA3 itself. A GO analysis revealed that genes involved in defense to fungi were particularly overrepresented among down-regulated genes in camta3 mutants (fig. S7). Twelve genes were more than 10-fold down-regulated in camta3 indicating a strong positive contribution of CAMTA3 to their expression. Approximately 20% of DEGs in untreated camta3 plants were part of the previously defined 1671 core touch-responsive genes (18). This is much higher than expected by chance (5%), indicating a role for CAMTA3 in the regulation of touch-responsive genes.

To assess the contribution of CAMTA3 to touch-responsive gene expression, we searched the 2265 genes that were touch-responsive in Col-0 for genes that were differentially expressed in camta3 at 22 min versus Col-0 at 22 min (padj < 0.05; Fig. 3, B and C). In total, 79 genes showed an altered touch response in camta3-1 mutants (Fig. 3, B and C), with around half of the genes (54%) showing an attenuated touch responsiveness. This was as opposite to what was observed in the fer-4 mutant (Fig. 2), where most genes showed an increased touch responsiveness. A total of 63% of these genes were considered “core” touch-responsive genes, showing that CAMTA3 has a predominantly positive role in gene expression regulation of touch responsive genes. Among these genes, we confirmed bHLH19 as being less induced in camta3-1 mutants as observed previously by qRT-PCR (Fig. 1). Transcript levels of TCH2 were induced 8-fold in Col-0 and only 5.6-fold in camta3-1, although the significant P value of 0.02 fell below the cutoff after the stringent multiple testing correction (padj < 0.05). This is likely explained by the lower number of biological replications used for the RNA-seq (three) versus qRT-PCR (five) for the camta3-1 22-min sample group. In addition, calmodulin-like 23 (CML23), the closest homolog of TCH2/CML24, was significantly less induced in camta3-1 mutants after touching compared to in Col-0 (Fig. 3C). GO analysis further showed that genes involved in defense, cell wall modification, SA response, abscisic acid response, ROS, hypoxia, regulation of flower development, and vegetative to reproductive phase transition of meristem were affected (table S7). To support that CAMTA3 is a regulator of these 79 genes, we used the TF2Network tool to search for transcription factor binding. Consistently, overrepresented sites were found for CAMTA3 and related proteins CAMTA1 and CAMTA2 (table S7).

Now, the best-defined component of touch-responsive transcriptional regulation involves the JA-dependent pathway, which to a large extent is regulated by the MYC2 MYC3 MYC4 transcription factors (18). Therefore, we performed a comparative analysis to evaluate the relative contribution of CAMTA3, MYC2 MYC3 MYC4, and Feronia to touch signaling (Fig. 4). Previously, 357 genes were identified to have a differential touch response in myc2 myc3 myc4 at 25 min after water spray compared to Col-0 (18), which is not an identical but for this context comparable experimental setup. A comparison of fer-4 and myc234 mutants revealed that 50 genes showed an altered touch response in both mutants relative to wild type (Fig. 4, A and B). This list was very strongly overrepresented in genes related to JA biosynthesis (e.g., LOX3/4, OPR3, and AOC3) and signaling (JAZ/TIFY proteins and MYC2 itself). In agreement, the GO categories of JA and wounding responses were overrepresented among these genes (table S8). To get a better overview of how these common genes are regulated in fer-4 and myc234 mutants, the expression values were plotted in heatmaps (Fig. 4B). This demonstrated very that fer-4 and myc234 showed almost completely opposite responses: While the common genes were more strongly up- or down-regulated in fer-4 compared to Col-0, they showed attenuated responses in myc234. Hence, Feronia is an important negative regulator of a large part of the JA-response positively regulated by MYC2 MYC3 MYC4. The responsive genes are, however, not fully overlapping, so we were interested to see whether any specific differences occurred between myc234 and fer-4. Therefore, we performed a GO analysis of the 361 fer-4–specific and 307 myc234-specific genes (Fig. 4A and table S9). Twenty-one of 50 GO categories at P < 0.05 present among the fer-4–specific genes were also overrepresented among myc234-specific genes, indicating that generally similar cellular processes are affected by both regulators. Specifically to Feronia, we observed a clear overrepresentation of cell wall components and metal ion transport, which was not observed in the myc234-regulated genes. Conversely, GO categories related to programmed cell death, calcium-ion binding, and defense to fungi were only observed among myc234-regulated genes.

Of the 79 genes with differential touch response at 22 min in camta3-1 (Fig. 3), 10 were overlapping with the myc234 357 DEGs (Fig. 4C and table S10), including bHLH19, UGT74E1, CA²⁺-DEPENDENT MODULATOR OF ICR1, and ethylene response factors ERF6/98/104. Twenty-nine genes were overlapping between the 79 camta3-1 and 411 fer-4 touch DEGs (Fig. 4D). The CAMTA3 transcript itself was 2.6× fold-induced by touching at 22 min in Col-0, with no significant differences in myc234 or fer-4 compared to Col-0. Three of these genes overlapping between myc234 and camta3-1 (Fig. 4A) have been defined as part of the core “MYC2 regulon,” based on direct promoter binding and transcriptional regulation (18). Cell wall–related genes, including expansin like EXLA1, EXLA3, and EXL1, and cell wall-associated kinases WAK1 and WAK2 were present in the 69 camta3-specific genes. Other transcription factors such as WRKY48 and WRKY25, pathogenesis-related proteins, and defense-related secreted peptides STMP4 and STMP5 were generally positively regulated by CAMTA3. Three genes involved in vegetative stage to flowering transition were less induced after touch in the camta3-1 mutant including calmodulin-like 23, laccase 8, and far-red-elongated hypocotyl-like fhl. In contrast, no clear signs of JA-related signaling could be observed. To verify the JA-(in)dependent expression of the CAMTA3-regulated genes, we assessed the expression of the 79 CAMTA3-regulated touch-responsive genes in a detailed RNA-seq analysis looking at Arabidopsis seedlings dipped into MeJA or mock solutions (48). Only bHLH19 showed an early and consistent up-regulation by dipping into MeJA solution from 15 min to 8 hours of treatment (fig. S8). Jacalin-related lectin 33 (JAL33) was induced by JA after approximately 2 to 3 hours only. We also checked their expression in a recent RNA-seq transcriptomics series that uses volatile methyl-jasmonate application (MeJA), thereby avoiding touch-induced effects (49). All three genes (bHLH19, Ca²⁺-dependent modulator of ICR1 CMI1, and ERF104) that were common to the MYC2 regulon and camta3-1–regulated touch-responsive genes (Fig. 4A) were induced by MeJA within 15 min (table S11). Only 3 of 69 camta3-1–specific genes (EXLA3, AT2G36650, and AT2G01300) were induced after 15 min of MeJA, further indicating that CAMTA3 is part of the JA-independent touch response network.

To investigate whether constitutive overexpression of CAMTA3 could affect JA-dependent and JA-independent touch response marker gene expression, we obtained two p35S:CAMTA3-YFP (OX) lines (Table 1) (50). We also included camta3-1 and a second T-DNA allele (camta3-2) as controls (Table 1) (50). Already under untreated conditions (Fig. 5), JA-independent touch marker genes TCH2 and CRK41 were significantly higher expressed in both CAMTA3 OX lines. JA marker gene JAZ8 showed normal expression in the OX lines, however. After touching, TCH2, CRK41, and bHLH19 showed significantly less induction in both camta3-1 and camta3-2 mutants, confirming the role of CAMTA3 in their induction. TCH2, CRK41, bHLH19, and PP2A5, however, showed a higher touch-induced expression in yellow fluorescent protein (YFP)–CAMTA3 overexpression plants. TCH4 showed a slightly higher basal expression in the CAMTA3 OX lines, while camta3-2 showed slightly less touch responsiveness for TCH4, although these changes did not meet the statistics cutoffs (Fig. 5). In contrast, genes representing JA-responsive genes such as JAZ8 and CML39 showed a wild-type–like touch response in both CAMTA3 loss- and gain-of-function lines.

Although many mutants have been assessed for deregulation of touch-responsive gene expression, only very few have been reproducibly shown to have significant effects on the transcriptional response (our data and studies cited above). The strongest effects found so far in, e.g., myc234, jar1, and coi1 mutants were on genes that were induced by the JA-dependent pathway. In contrast, these mutants showed no effect on the “JA-independent” touch-responsive TCH2-4 genes (18). Moreover, mutants previously suggested to have reduced induction of TCH2-4 (fer-4 and ein6) (16, 30) showed wild-type–like responses to touching of green tissues in our study (Fig. 1 and fig. S4). Our findings suggested that in camta3-1 mutants TCH2 expression was less induced after touching than in wild-type plants (Fig. 1). As CAMTA3 is part of a small gene family in Arabidopsis, we hypothesized that CAMTA3 may cooperate with related factors to regulate gene expression of TCH genes. We therefore obtained a set of multiple mutants for CAMTA1, CAMTA2, and CAMTA3. The developmental effects in single camta3 mutants, including smaller rosette size and faster senescence, were more pronounced than in camta1 camta2 mutants, suggesting that CAMTA3 has the largest individual contribution as observed previously (fig. S9) (51). Double camta1 camta3 mutants looked relatively similar to camta3-1 mutants, while camta2 camta3 (camta2/3) mutants showed a stronger phenotype than single camta3-1 mutants. Last, the camta1 camta2 camta3 (camta1/2/3) mutant initially germinated and established itself relatively normally but then further growth became strongly reduced, resulting in severe dwarfism (fig. S9). As camta2/3 and camta1/2/3 mutants showed the strongest developmental effects, we assessed their capacity to induce touch-responsive gene expression of JA-dependent and JA-independent genes (Fig. 6). Strikingly, the camta1/2/3 mutants showed a significant inhibition of touch-induced expression of bHLH19, CML23, CRK41, WRKY48, PP2A5, EXLA3, TCH2, and TCH4 (Fig. 6), while camta2/3 and camta3-1 mutants showed an intermediate but also significantly diminished touch responsiveness for most of these genes. Thus, the loss of multiple CAMTA1/2/3 members resulted in a progressively stronger loss of touch responsiveness for JA-independent touch marker genes. In contrast, genes representing JA-dependent touch responses such as JAZ8 and CML39 showed a wild-type–like touch response in the camta3-1, camta2/3, and camta1/2/3 mutants. TCH3 and WRKY33 actually showed a higher basal expression in the camta2/3 and camta1/2/3 mutants and an enhanced touch responsiveness, suggesting that they may also be regulated by a third pathway.

Detailed analysis of the promoter regions of TCH2, TCH4, bHLH19, and EXLA3 showed the presence of CGCGT-related CAMTA1-3 binding motifs (52), suggesting direct interaction of CAMTA1-3 with their promoters (table S12). To assess whether CAMTA3-related transcription factors directly regulate the expression of TCH2, TCH4, bHLH19, and EXLA3, we assessed their binding using a DNA affinity purification sequencing (DAP-seq) dataset (53). Although CAMTA3 and CAMTA2 were not assessed in the DAP-seq study, CAMTA1 and related CAMTA5 were included in the assays, showing a clear and significant enrichment for CAMTA1 binding in the TCH2, TCH4, EXLA3, and bHLH19 promoters (Fig. 7A). Binding by CAMTA1 to the bHLH19 promoter appeared to be relatively weaker compared to TCH2 and TCH4. CAMTA5 binding could only be observed to the TCH2 promoter.

To validate these interactions, we recombinantly expressed and purified a truncated 6xHis-tagged version of CAMTA3, containing the DNA binding domain present in the first 153 amino acids [CAMTA3 (1–153)], as the full-length 116-kDa protein was previously found not to be suitable for electromobility shift assays (EMSAs) (52). The binding of CAMTA3 (1–153) to the putative CGCGT binding sites present within the promoter regions that showed CAMTA1/5 binding in the DAP-seq data was assessed by EMSAs. For this purpose, 40–base pair (bp) ³²P-radiolabeled probes were created containing the wild-type sequences or with specific mutations in the CGCGT-like motifs (Fig. 7B and table S12). The TCH2 promoter contained two adjacent motifs that were contained within the same probe, while the EXLA3 and bHLH19 promoters contained two and four CGCGT-like motifs, respectively. Strong binding of CAMTA3 (1–153) could be observed to probes for TCH2, TCH4, and EXLA3 (probe EXLA3-1 only). Two shifted bands of slightly different size could be observed in the TCH2 promoter, suggesting that both CGCGT motifs can be bound by CAMTA3 (1–153). The binding to the radiolabeled probes could be strongly reduced by addition of excess unlabeled probe. Specific mutation of the CGCGT sites in the probes completely abolished binding by CAMTA3 (1–153), further confirming the interaction is sequence-specific. Unexpectedly, no clear binding of CAMTA3 (1–153) to four different probes containing CGCGT-like sites in the bHLH19 promoter could be observed (Fig. 7B and fig. S10). Touch treatment did not affect the subcellular localization of CAMTA3-YFP, with a clear nuclear localization in both untouched seedlings and seedlings 22 min after touching (fig. S11). In conclusion, our results show that CAMTA1/3 can directly bind the promoters of TCH2, TCH4, and EXLA3 and perhaps weakly to the promoter of bHLH19.

Regular touch treatment has been shown to result in reduced rosette size and delayed flowering in plants (7). As our transcriptomic analyses indicated altered touch responsivity at the gene expression level, we assessed the impact of fer-4, camta3, and piezo mutations on thigmomorphogenesis (Fig. 8 and figs. S12 and S13). Two-week-old soil-grown seedlings were exposed to mild touching with a brush twice daily until all plants had bolted. Regularly touched Col-0 plants showed a clear delay in bolting time, with the average untouched plant flowering around 27 days after sowing, while touched Col-0 plants only bolted around 31.5 days (Fig. 8, A to C). In addition, rosette size at day 35 after sowing was reduced by more than 40% in touched compared to untouched Col-0 (Fig. 8, D and E). A similar effect of regular touching was observed in piezo mutants (fig. S12), further confirming that Piezo mechanosensitive channels do not contribute strongly to touch responses in aboveground tissues. As a control for JA signaling-defective mutants, myc234 mutants were grown alongside camta3-1 mutants and showed a clear insensitivity to touch treatment, with no significant difference in bolting time in touched compared to untouched myc234 mutants, as previously reported (fig. S13) (18). The fer-4 mutant showed a strong reduction in overall growth compared to Col-0 even under untreated conditions as previously reported (54), with a three times smaller rosette area on day 35 under untouched conditions (fig. S12). Despite the general growth retardation, fer-4 plants were still very responsive to touch-induced delays in flowering and rosette size (fig. S12). In addition, camta3-1 plants showed a moderate reduction in rosette size under untreated conditions compared to Col-0 but flowered around the same time as Col-0 under untouched conditions, 27 days after sowing (Fig. 8, A to E). In contrast, camta3-1 mutants showed an almost complete insensitivity to regular touching, with no significant difference in bolting time and rosette size in touched compared to untouched camta3-1 mutants (Fig. 8, A to E). We also performed a similar experiment on the camta3-2 mutant, confirming the loss of thigmomorphogenesis was comparable to that observed in the camta3-1 mutant (fig. S14). In conclusion, CAMTA3 plays a critical role in thigmomorphogenesis in plants.

To study touch-mediated molecular responses, Arabidopsis thaliana wild-type (Col-0 and Ler-0) and mutant (listed Table 1) plants were grown as follows. Arabidopsis seeds were surface sterilized with chlorine gas. The seeds were sown in plates containing 1× MS with vitamins, 1% sucrose, 1% phytoagar, MES (0.5 g/liter), and pH adjusted to 5.8. The plates were stratified at 4°C for 3 days before transferring to standard growth condition (21°C, 16-hour light/8-hour dark cycle, 100 μmol m⁻² s⁻¹). To study the touch-mediated molecular responses, 12-day-old wild-type and mutant seedlings were gently brushed with a soft paint brush for 10 times, taking about 30 s per plate (see movie S1). For transcript analysis, samples were collected 0, 22, and 40 min after touch treatment and then snap-frozen in liquid N₂ and stored at −80°C until use. Approximately 10 to 12 seedlings were harvested to represent one biological replication. Four to five biological replicates were used for statistical analysis unless otherwise stated. The list of mutants used in this study is available in Table 1. Mutants were either obtained from the Nottingham Arabidopsis Stock Centre (NASC) or requested from other laboratories, and then, genotypes were confirmed using PCR. Myc2 myc3 myc4 myc5 was obtained from G. Howe (Michigan State University); caf1a caf1b and camta3-1 were obtained from K. Dehesh (University of California Riverside); fer-4 and the1 were obtained from B. Dotson (Lund University); vcs-7 was obtained from L. Sieburth (University of Utah); vcs-8 and xrn4 were gifted by C. Bailly (Sorbonne Université); glr3.3 glr3.6 was gifted by E. Farmer (University of Lausanne); wrky18 wrky40 wrky60 were gifted by D.-P. Zhang (Tsinghua University); camta1/2/3 multiple mutants were gifted by M. Thomashow (Michigan State University); and camta3-2, p35s::CAMTA3-YFP lines; 19-5 (OX1) and 32-2 (OX2) were gifted by J. Lee (Leibniz Institute of Plant Biochemistry).

RNA extraction, cDNA synthesis, and RT-qPCR were performed as previously described by (65). Primers are shown in table S12. Briefly, frozen tissue samples were ground in a fine powder using a Qiagen Tissuelyser II bead mill, and total RNA was isolated using a Spectrum Plant Total RNA kit (Sigma-Aldrich, STRN250-1KT) with On-Column deoxyribonuclease (DNase) treatment (Sigma-Aldrich, DNASE70) according to the manufacturer’s instructions. A total of 500 ng of total RNA was used for cDNA synthesis using an iScript cDNA synthesis kit (Bio-Rad, Hercules, CA, USA). RT-qPCR was performed using diluted cDNA (1:7) and SYBR green fluorescent dye (SsoAdvanced Universal Sybr green mix, Bio-Rad). A list of primers used for the RT-qPCR is available in table S12. For normalization, ubiquitin was used as internal control gene. The relative transcript level of each transcript was calculated by comparing touched versus control plants according to (66). Data were analyzed using two-way ANOVA (genotype × treatment) with post hoc Tukey tests (letters show significant difference between all genotypes in the different time points at P < 0.05). Because of the dynamic range of touch induced gene expression, the values were not always normally distributed, as indicated by the Kolmogorov-Smirnov test. For those genes, the expression values were log₂-transformed to obtain normal distribution before ANOVA analysis.

For RNA-seq library preparation, total RNA was treated with Ambion Turbo DNase (Thermo Fisher Scientific, AM1907) and quantified using the Qubit RNA BR Assay Kit (Invitrogen, Q10210). Four biological replicates (n = 4) were used for 0-min samples of Col-0, camta3-1, fer-4, and piezo and 22-min samples of Col-0 and piezo. Three biological replicates (n = 3) were used for 22-min samples of camta3-1 and fer-4. Then, 500 ng of RNA was used for library preparation using Illumina TruSeq mRNA (poly-A selection) and TruSeq RNA UD Indexes for up to 96 samples (Illumina, 20022371). Samples were pooled and sequenced on a half Illumina NovaSeq6000 S4 lane, 2 × 150 bp reads, incl Xp kit. Data were processed using demultiplexing and quality controlled with FastQC. Alignment of reads was performed against the TAIR11 annotation using STAR (67). On average, 32 million reads per sample were generated. Counts were assigned to genes using featureCounts (68), and analysis of DEGs was performed with DeSEQ2 (69). Transcripts were considered differentially expressed if padj < 0.05 and fold change ≥ 1.5 or fold change ≤ −1.5. Raw RNA-seq data files are available on ArrayExpress (accession number E-MTAB- 10920). GO enrichment analysis was performed using TF2Network and GOrilla (70).

Seeds were stratified at 4°C for 2 days before grown in soil mixture (soil, perlite, and vermiculate 4:1:1) under the standard growth conditions (22°C, 16-hour light/8-hour dark, 100 μmol m⁻² s⁻¹). Two-week-old seedlings were gently brushed twice daily approximately 10 times with a soft art paint brush. Touch treatment was applied for 18 to 26 days until all plants had bolted depending on the genotypes. Plants were considered to have bolted when the inflorescence stock reached to 1-cm height, and percentage of bolting was calculated as previously described (25). The rosette area of 31- to 35-day-old plants was measured using ImageJ.

The DNA binding domain of CAMTA3 (amino acids 1 to 153) was cloned into pDEST17 6xHisfusion vector and recombinantly expressed in Rosetta 2 (DE3) cells as previously described (71). In brief, a 50-ml culture was grown overnight at 28°C. The culture to was diluted in 1 liter of LB media to an optical density at 600 nm (OD₆₀₀) of 0.1 and subsequently grown to OD₆₀₀ 0.6 at 28°C. Protein expression was induced by addition of 1 mM isopropyl-β-d-thiogalactopyranoside and incubated at 28°C for 6 hours. Cells were pelleted by centrifugation and resuspended in 2 × 10 ml of lysis buffer containing one tablet of Roche cOmplete EDTA-free mini protease inhibitor cocktail and lysozyme, followed by sonication for 2 min in 5-s blocks and 10-s intervals. Purification was performed using GE Healthcare HisTrap FF columns according to the manufacturer’s instructions. Oligonucleotide probes (40 bp; table S12) were annealed and phosphorylated using NEB T4 polynucleotide kinase with gamma ³²P-ATP (PerkinElmer) and purified using G-25 Sepharose Quick spin columns (Roche). Binding reactions were performed as previously described (71), separated on 6% acrylamide 0.5× tris-borate EDTA gels, dried, and imaged using a phosphorimager.

Arabidopsis seedlings expressing CAMTA3-YFP were grown vertically on MS plates for 5 days, then subjected to control (untouched) or touch treatment, and stained with 4′,6-diamidino-2-phenylindole (DAPI). Root cells of the stained plants were used to detect the fluorescence under a Leica SP8 TCS confocal microscope. Samples were imaged with laser excitation at 514 nm and emission at 500 to 570 nm for YFP, while DAPI nuclear fluorescence was observed with excitation at 405 nm and emission at 415 to 476 nm.

RNA-seq data files are available from ArrayExpress with accession number E-MTAB-10920. All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials.