FERONIA and microtubules independently contribute to mechanical integrity in the Arabidopsis shoot

To survive, cells must constantly resist mechanical stress. In plants, this involves the reinforcement of cell walls, notably through microtubule-dependent cellulose deposition. How wall sensing might contribute to this response is unknown. Here, we tested whether the microtubule response to stress acts downstream of known wall sensors. Using a multistep screen with 11 mutant lines, we identify FERONIA (FER) as the primary candidate for the cell’s response to stress in the shoot. However, this does not imply that FER acts upstream of the microtubule response to stress. In fact, when performing mechanical perturbations, we instead show that the expected microtubule response to stress does not require FER. We reveal that the feronia phenotype can be partially rescued by reducing tensile stress levels. Conversely, in the absence of both microtubules and FER, cells appear to swell and burst. Altogether, this shows that the microtubule response to stress acts as an independent pathway to resist stress, in parallel to FER. We propose that both pathways are required to maintain the mechanical integrity of plant cells.


Introduction
All living organisms use mechanical forces as instructive cues during their development [1,2]. They also share a common mechanical property: Cells are pressurized by osmotic pressure and thus experience cortical tension. Osmotic pressure in plants is several orders of magnitude higher than that of animal cells, and it is counterbalanced by stiff cell walls [3]. Regulating the mechanical properties of cell walls, through the perception of wall tension and integrity, is thus crucial for plant growth and development [4,5].
Plant cell walls are composed of load-bearing cellulose microfibrils, tethered by a matrix made of polysaccharides and structural proteins [6,7]. The deposition of cellulose microfibrils is generally guided by cortical microtubules [8]. Beyond  of cellulose microfibrils controls the mechanical anisotropy of the wall. There is now ample evidence showing that cortical microtubules align with maximal tensile stress directions in the wall [9,[10][11][12][13]. This provides a feedback loop in which shape and growth, whether at the individual cell or whole organ scale, prescribes a pattern of stress, to which cells resist by reinforcing their walls along maximal tensile stress directions [12,14]. Although they are significantly less stiff, other wall components contribute to wall properties. In particular, pectins can partially rescue defects in cellulose synthesis in young cell walls. For instance, isoxaben treatment, which inhibits cellulose deposition through the internalization of CESA complexes, leads to thicker walls that are enriched in pectin [15]. Similarly, in young hypocotyls, pectin polarities precede the formation of mechanically anisotropic walls [16]. In contrast to cellulose deposition, pectin, as well as all other matrix components, are secreted to the cell wall [6,7]. Therefore, in principle, this provides an alternative mechanism for the cell to resist wall tension or damage. As for microtubules, the related mechanotransduction pathway is largely unknown. Yet, over the past decade, Catharanthus roseus Receptor-Like Kinases (CrRLKs) have emerged as key players. Although the link with mechanical stress remains to be formally established, THESEUS1 (THE1) has been implicated in the wall integrity pathway [17][18][19]. Based on defective root growth behavior on stiff interface, calcium signaling, pH response, and TOUCH gene expression, FERONIA (FER) has emerged as a candidate mechanosensor [20]. FER can sense the status of the cell wall, notably when salinity rises, through pectin binding [21]. It was recently proposed that FER also monitors microtubule behavior through a cascade involving Rho GTPases (ROP6) and the microtubule severing protein katanin [22]. Here, through a reverse genetic screen on wall sensors and using a suite of mechanical tests, we show that our best wall sensor candidate FER is not required for the microtubule response to stress, further suggesting that the microtubule response to stress can be more autonomous than anticipated. We also reveal that FER-dependent wall integrity pathway depends on wall tension and that both FER and the microtubule response to stress contribute to wall integrity.

Altered pavement cell shape as a proxy for defective response to mechanical stress
We first used a morphometric proxy to test the involvement of wall sensors in the microtubule response to stress. The jigsaw puzzle shape of Arabidopsis pavement cells has been proposed to be actively maintained and amplified by the microtubule response to mechanical stress. Indeed, necks in such cells prescribe highly anisotropic tensile stresses locally, to which microtubule arrays and thus cellulose deposition align [14,23,24]. We reasoned that the shape of pavement cells could be used in a mutant screen as a proxy for defects in that mechanical feedback loop. In past research, such screens have targeted the intracellular biochemical cues behind cell-cell coordination [25] and the contribution of cell wall properties in cell shape [26]. Whether wall sensors are involved in pavement cell shape remains ill described. Here, we focused on mutants impaired in receptor-like kinases that are highly expressed in the epidermis and aerial parts of the plant during early development and that exhibit an established link with the cell wall (and their closest homologs), namely feronia (fer), theseus1 (the1), theseus1/ feronia-related1 (tfr1; at5g24010), curvy1 (cvy1), hercules receptor kinase 1 (herk1), herk2, mdi-s1-interacting receptor-like kinase2 (mik2-1), wall-associated kinase 1 (wak1), wak2, wak3, and wak4 (see S1 Table). We imaged and quantified the pavement cell shapes in receptor-like kinase candidate mutants, with the aim to select the ones with the strongest cell shape defects (Fig 1A).
To do so, we extracted the epidermal signal and used PaCeQuant to segment pavement cells and measure 27 shape descriptors (see Materials and methods [27] (S1A Fig)). To select the most discriminating PaCeQuant shape descriptor, we first performed a principal component analysis on our data. We compared the contribution of each parameter to the 2 axes with the most variation ( Fig 1B). The cell perimeter was first, followed by the convex hull perimeter, the convex hull area, the cell area, the nonlobe area, and the longest path length, then by circularity (for relevant definition, see S1B Fig).
Note that the same trend was obtained when considering solidity, another parameter often used to characterize lobe formation [38] (S1D Fig). Note also that while circularity and solidity can be impacted when stress response levels change, they do not necessarily directly scale with stress response level. Different shapes can lead to similar circularity and solidity values (e.g., see last figure); nevertheless, this cell shape-based screening allowed us to identify the RLK mutants with the most affected pavement cell shape, possibly through a defective microtubule response to stress. Because fer-4 stands out, this first screening suggests that FER could play a role in the microtubule response to mechanical stress.

Differential response of receptor-like kinase mutants to isoxaben
To challenge the results from this initial screen, we next used a well-established protocol to mechanically perturb cell walls. Isoxaben inhibits the biosynthesis of cellulose [39], and thus weakens the wall. In past work, such treatment were shown to induce a hyperalignment of cortical microtubules at the shoot apical meristem and in cotyledon pavement cells [14,40], consistent with a response to increased tensile stress levels in the cell wall. Note that isoxaben can also trigger other responses, including reactive oxygen species (RAU : PleasenotethatROShas OS) production, lignification, and changes in gene expression [17]. Thus, depending on time and dose, isoxaben may also ultimately reduce stress level [41]. Here, we use this drug as a screening tool, complementary to the pavement cell shape screen, to identify mutants insensitive or hypersensitive to mechanical perturbation and which are thus likely to be defective in mechanosensing.
Altogether, our primary screening approach using both pavement cell shape analysis and isoxaben sensitivity test as a combined proxy for the microtubule response to mechanical stress highlights FER as a primary candidate. So far, our data are also consistent with the proposed scenario in which FER and katanin belong to the same pathway [22], but these remain primary screening approaches that do not directly test the involvement of our main candidate in the microtubule response to mechanical stress. Based on this screening, in the following we decided to focus on FER to investigate its involvement in the mechanical integrity of the shoot and to directly test its contribution to the microtubule response to mechanical stress.

Rescue test: Decreasing the matrix potential of the growth medium rescues the burst cell phenotype in feronia
We first decided to further investigate the isoxaben hypersensitivity phenotype of fer-4. It is commonly believed that WT seedlings display a shorter hypocotyl on isoxaben because cell wall defects are perceived and compensated through wall reinforcement, ultimately leading to reduced growth. Both wall reinforcement and reduced growth would prevent the cells from bursting. This notably builds on the analysis of the1 mutant, which exhibits a longer hypocotyl than the WT on isoxaben because of the lack of wall sensing [17]. In fer, the hypocotyl is in contrast even shorter than the WT. Thus, either wall sensing is enhanced in fer thus strongly repressing growth, or, by contrast, wall sensing is impaired in fer, even more than in the1, and cells burst before walls can even be reinforced. In the rest of the article, we focus the analysis of the fer-4 allele, but similar results could be obtained in fer-2 (S3 Fig). When looking closely at fer-4 hypocotyls grown on isoxaben and stained with propidium iodide (PI), we observed many dead cells, as revealed by PI staining (Fig 3A). We calculated a bursting index, i.e., the percentage of burst cell area over total area of a field of epidermal cells in a given image. We found the bursting index to be more than one order of magnitude higher in fer-4 isoxaben-grown hypocotyls (20 ± 18%) than in the WT (0 ± 1%, p < 10 −3 ; n Col-0,iso,0.7%agar = 16; n fer-4,iso,0.7%agar = 16, Fig 3B). This observation thus seems consistent with the latter scenario. To test this hypothesis further, we reasoned that lowering the water potential in the medium should reduce water intake for seedlings, reduce tensile stress level [13], and in the end, reduce the number of burst cells. We thus tested different agar concentration in the medium and analyzed the impact on fer-4 phenotype. When increasing the agar concentration to 2.5%, the bursting index in hypocotyls was reduced by 70% in isoxaben-grown seedlings, supporting our hypothesis (p = 0.002; n fer-4,isoxaben,0.7%agar = 16; n fer-4,isoxaben,2.5%agar = 16, Fig 3A and 3B).
We then took a closer look at the pavement cells of fer-4 cotyledons grown in a medium containing a standard concentration of agar (0.7%) and over time. When looking at orthogonal sections, the outer cell wall appeared very deformed, consistent with cell swelling in fer-4, when compared to WT (S4 Fig). This geometry echoes previously observed phenotypes and is consistent with increased wall stretch in fer [20,21,42].
In these growth conditions, the mechanical integrity of the cell was not maintained in fer: Many burst cells were present, as marked by a strong PI coloration (Fig 3C, S4 Fig). The bursting index increased over time, from 16 ± 16% at 4 DAG (days after germination) to 60 ± 30% at 12 DAG. At both time points, it was significantly higher to that of the WT (p 4DAG,0.7%agar < 10 −3 ; p 12DAG,0.7%agar < 10 −3 ; n Col-0,4DAG,0.7%agar = 14, n Col-0,12DAG,0.7%agar = 9, n fer-4,4DAG,0.7%agar = 14, n fer-4,12DAG,0.7%agar 8; Fig 3D). To further confirm that burst cells appear in response to mechanical stress, we exposed cotyledon to an osmotic shock and monitored cell behavior. We transferred seedlings grown on 2.5% agar for 4 days to 0.35% agar, reasoning that this sudden change should challenge the fer-4 cell walls. As predicted, many fer-4 cells swelled and burst in the hours following the transfer (S5 Fig, S1 Movie).
Note that in all these experiments, the aerial phenotype is rescued, even though only the roots are in contact with a medium with different agar concentrations. This is consistent with the idea that the differences in matrix potential is the cause of the rescue (and not the contact with soft or stiff agar). This is also consistent with previous results obtained in other mutants with cell wall defects, such as the quasimodo mutants, where it was also confirmed, using atomic force microscopy, that growth on 2.5% agar medium significantly decreased the apparent tensions in the cotyledon's outer epidermal cell walls [13].
Thus, we find a correlation between the mechanical status of the fer-4 cotyledons (as monitored by the bursting index) and growth. Taken together, these results suggest that the fer-4 mutant is unable to sense the mechanical status of the tissue and that the fer-4 phenotype reflects a passive turgor-dependent wall defect.

Cortical microtubules orient with predicted maximal tensile stress after ablation in fer-4
So far, all the tests suggest that FER is a major player in the mechanical integrity the shoot, consistent with signaling and growth data obtained in the root [20] and with growth data in the hypocotyl [43]. However, it is not clear whether this involves the microtubule response to mechanical stress. To formally check this, we analyzed the behavior of cortical microtubules in response to local ablation in fer-4. From a mechanical standpoint, the sudden drop of turgor pressure in the ablated cells, together with epidermal tension [44], generates a circumferential tensile stress around the dead cells [12]. Such a perturbation causes the cortical microtubules to reorient in the new maximal tensile stress direction (circumferential) around the ablation [12,14]. We used the microtubule reporter pPDF1::mCit-MBD to monitor the microtubule response in Col-0 and fer-4 cotyledons. As a 0.7% agar medium triggers widespread cell death in fer-4 ( Fig 3C, S6A-S6D Fig), all tests were performed on 2.5% agar. We measured both the anisotropy and the average orientation of cortical microtubule arrays (relative to the ablation site) in cells surrounding the ablation (Fig 4, S7 and S8 Figs).
Similar trends could be observed when using the p35S::GFP-TUB microtubule marker line: cortical microtubule orientation appeared circumferential around ablations in both WT and fer-4 (S9 Fig). However, the diffuse fluorescent signal hindered quantitative analysis with FibrilTool. Because the microtubule response to ablation is comparable in fer-4 and in the WT, this formally shows that the microtubule response to stress can be independent from FER.

FER and microtubules independently contribute to the mechanical integrity of the shoot in response to mechanical stress
Our ablation results may seem at odds with the fact that fer-4 and bot1-7 belong to the same cluster when analyzing pavement cell circularity (see Fig 1D). We thus revisited our quantification of pavement cell shape to identify other shape descriptors amenable to discriminate bot1-7 and fer-4. We focused on lobe size in pavement cells. In the katanin mutant bot1-7, which displayed a higher circularity than the WT (see Fig 1), the average basal lobe width was 3% smaller than that of the WT (p < 10 −3 ; n WS-4 = 294; n bot1-7 = 252; Fig 5A). Although this is a rather small difference, this is consistent with a reduced microtubule response to stress in the katanin mutant. By contrast, the average basal lobe width in fer-4 was 12% larger than that of the WT (p < 10 −3 ; n Col-0 = 428; n fer-4 = 321; Fig 5A). Similar trends in basal lobe width were observed when seedlings were grown on 2.5% agar, albeit with lower values, also consistent with a reduced microtubule response to mechanical stress in such conditions (S10 Fig). Thus, both fer-4 and bot1-7 pavement cells exhibit higher circularity than the WT through different geometries. Pavement cell shape may thus reflect different responses to stress: Reduced microtubule dynamics in bot1-7 would generate smaller lobes, whereas increased stress levels in fer-4 would lead to hyperaligned cortical microtubules and larger lobes. Such hyperaligned cortical microtubules can be observed in fer-4 pavement cells when seedlings are grown on 0.7% agar ( S6C Fig). Alternatively, and nonexclusively, the presence of dead cells in fer-4 may affect the stress pattern, and thus the cortical microtubule response, further increasing the circularity of pavement cells.

Discussion
From a reverse genetic screen, we show that FER plays a primary role in the mechanical integrity of the shoot (using both hypocotyl and cotyledons as model systems). We then find that FER prevents turgor-dependent cell bursting in the shoot. We also find that the microtubule response to stress does not require FER. This does not exclude the possibility that FER may modulate the microtubule response to stress, indirectly. Finally, we show that shutting down both FER and microtubule pathways has an additive effect on turgor-dependent cell bursting. This provides a scenario in which the mechanical feedback in the shoot that is required to maintain mechanical integrity, involves 2 largely independent modules, cortical microtubules and FER (Fig 5D).

FERONIA-and MT-dependent mechanical integrity
This result is consistent with recent findings showing how pectin and cellulose deposition are controlled with largely disconnected networks: Pectins are important for the initiation of pavement cell formation, and the deposition of cellulose microfibril rather amplifies a preexisting mechanical pattern [24]. Similarly, mechanical polarities in growing hypocotyl cells precedes the cellulose-derived mechanical anisotropy in hypocotyl cells [16].
We also provide evidence that the fer mutant can be largely rescued through changes in the mechanical environment of the plant despite its central and somewhat pleiotropic role in plant development. This echoes recent findings where essential regulators are found optional, when challenged with different mechanical environments. This is the case for instance for key regulators of pectin synthesis, where decreasing tensile stress can rescue cell-cell adhesion defects in the quasimodo mutants [13]. This was also shown for the katanin mutant, where increasing tensile stress levels with isoxaben treatment can generate WT-like cortical microtubule arrays [29]. More recently, this was even extended to signaling in the context of apical hook formation: arf7 arf19 auxin transduction mutant seedlings exhibit a WT phenotype when mechanically constrained [45].
If the microtubule response to stress does not depend on FER, what could be the most relevant mechanotransduction pathway? As shown with optical tweezers, pulling on microtubules promote their polymerization in vitro [46,47]. Cortical microtubules also align with predicted maximal tension when protoplasts are confined in rectangular microwells and pressurized by hypoosmotic conditions [48]. These data, together with the FER-independent microtubule response to stress, further support the hypothesis that microtubules may be their own mechanosensor [49].
Conversely, microtubule (and cellulose microfibril) alignment likely results from a combination of several cues, beyond tensile stress. Cell geometry can affect microtubule behavior, independent of cortical tension. In particular, due to their high persistence length, microtubules tend to become longitudinal in vitro [50] or in depressurized protoplasts [51]. Furthermore, cell edge factors can affect cortical microtubule behavior, leading to cell-scale aligned arrays [52][53][54]. Although microtubule, FER, stress, and cell geometry can be uncoupled in experiments, their interplay may provide synergies in vivo. In particular, the deposition of matrix material in the wall depends on exocytosis, which is also promoted by membrane tension. Conversely, affecting matrix deposition may weaken the wall and increase the tensile stress levels. Consistently, microtubule arrays exhibit an enhanced anisotropy when the cell edge GTPase Rab-A5c dependent trafficking is affected in roots [54].
All living organisms constantly sense and respond to mechanical stress. Interestingly, so far, all known mechanosensors at the extracellular matrix (e.g., integrins) also display ligand binding activity. FER may belong to this category as well, as FER also acts as a RALF peptide receptor. Whether this dual role provides synergistic activity for signaling remains to be investigated.
By impairing FER and microtubules, we reveal how plant cells behave when they become unable to respond to stress: They switch to a more passive mode, like a balloon or a soap

PLOS BIOLOGY
bubble, and because of turgor pressure, swell and ultimately die by bursting. Revisiting living systems with the lens of active matter physics is thus particularly suited to understand how cells manage to resist to stress, notably in their response to a fluctuating environment.
Seeds were surface sterilized and individually sown on Murashige and Skoog medium (MS medium, Duchefa, Haarlem, the Netherlands) or Arabidopsis medium (custom-made Duchefa "Arabidopsis" medium (DU0742.0025, DAU : PleaseprovidethemanufacturerlocationforDuc uchefa Biochemie, The Netherlands) with different agar concentrations (see S2 Table for detailed description of the different media used). For drug treatments,AU : PleaseprovidethemanufacturerlocationforSigmaatfirstmentioninthes media were supplemented with isoxaben (Sigma, Germany) or oryzalin (CAU : PleasenotethatasperPLOSdoesnotallowtermslikeInc:; Ltd:; etc:; inthemanuscriptexce hem Service, USA) from stock solutions in dimethyl AU : Pleaseprovidethemanufacturerl sulfoxide (DMSO, Sigma). All plants were placed 2 days in the dark at 4˚C then transferred in a 20˚C long days growth chamber. When seedlings were maintained in the dark, petri dishes were covered with aluminum foil in the 20˚C long days growth chamber.
For every experiment, 3 biological replicates or more were obtained. Col-0 or WS-4 seedlings were included as controls in all experiments and replicates. This also explains why the sample size of Col-0 images used in Figs 1 and 2 is bigger than for the other genotypes. Note that some of the Col-0 cotyledon data were previously used in our methodological article [55], as templates to introduce the SurfCut ImageJ tool (see below in Image analysis) for cell contour extraction.

Statistical analysis
Statistical analyses were performed with R software (https://www.R-project.org). The sample size is indicated in the main text. For pavement cell shapes (Figs 1 and 5A)AU : AcitationforFig6Awasde , we used PaCe-QuantAna, the R script that accompanies the PaCeQuant analysis [27], and the FactoMineR and factoextra R libraries for the principal component analysis [58]. Violin plots were shown with the corresponding p-value of Kruskal-Wallis tests. For hypocotyl length (Fig 2), the control distribution of hypocotyl length was standardized and the same parameters (μ ctrl , o' ctrl ) used to shift the isoxaben distribution similarly (S2 Fig). A Wilcoxon-Mann-Whitney test was then performed on the shifted distributions of the mutant and of the WT for each genotype. For the orientation of cortical microtubules, a Kolmogorov-Smirnov test was performed to compare the angle distributions to a uniform distribution between 0˚and 90˚. All other quantitative measures were compared using Wilcoxon-Mann-Whitney tests. As a Wilcoxon-Mann-Whitney test can be directed or not, the p-value shown in all experiments with a Wilcoxon-Mann-Whitney test was that of a nondirected test for nonsignificant p-value (to ensure that neither distribution was higher than the other) and that of a directed test in the significant direction for a significant p-value. Differences with p-values that were under 1% were considered significant, and those between the commonly used thresholds of 5% and 1% were considered as tendencies. All ± values referred to the standard deviation of the distribution.
SAU : AbbreviationlistshavebeencompiledforthoseusedthroughoutSupportinginformat upporting information S1  Fig 2B). To further ease the comparison, the differences between WT and mutants are then plotted as the hypocotyl length index deviation from the WT, which are the values obtained when subtracting the mutant treated normalized mean to the WT treated normalized mean. This generates a negative value in the hypocotyl length index for mutants being more sensitive to the treatment and a positive value for mutants less sensitive to the treatment ( Fig  2C). WT, wild type. Representative confocal images of fer-4 pavement cells, from seedlings grown on a medium containing 2.5% agar for 4 days, transferred on a medium containing 0.35% agar, stained with PI and imaged in water immediately after the transfer to a new medium (t0) and 1 to 16 hours later. Cell contours were extracted with SurfCut and projected in 2D. Orthogonal projections were also extracted with Fiji, focusing on the cell indicated with red arrows. See also S1 Movie. Scale = 100 μm.