DEFECTIVE KERNEL1 regulates cellulose synthesis and affects primary cell wall mechanics

The cell wall is one of the defining features of plants, controlling cell shape, regulating growth dynamics and hydraulic conductivity, as well as mediating plants interactions with both the external and internal environments. Here we report that a putative mechanosensitive Cys-protease DEFECTIVE KERNEL1 (DEK1) influences the mechanical properties of primary cell walls and regulation of cellulose synthesis. Our results indicate that DEK1 is an important regulator of cellulose synthesis in epidermal tissue of Arabidopsis thaliana cotyledons during early post-embryonic development. DEK1 is involved in regulation of cellulose synthase complexes (CSCs) by modifying their biosynthetic properties, possibly through interactions with various cellulose synthase regulatory proteins. Mechanical properties of the primary cell wall are altered in DEK1 modulated lines with DEK1 affecting both cell wall stiffness and the thickness of the cellulose microfibril bundles in epidermal cell walls of cotyledons.


Introduction
The polysaccharide rich extracellular matrix of plant cells has crucial roles in providing mechanical support, enabling water transport, cell-to-cell communication, and is a determining factor for turgor driven morphogenesis (Cosgrove, 2005). Cell walls vary in chemical composition and mechanical properties to support the different functional requirements of the cell. In primary walls of Arabidopsis the major components are cellulose, xyloglucan and pectins. Changes in cell wall composition and interaction between its component polymers greatly affect its mechanical properties and in turn impact growth and development of the plant (Zhang et al., 2019;Wang et al., 2020). Cellulose is a significant component of most types of cell walls, organized into cellulose microfibrils (CMF) that impart mechanical strength to the cell wall (Johnson et al., 2018;Zhang et al., 2019). Cellulose is synthesized by membrane bound CELLULOSE SYNTHASE (CESA) complexes (CSCs) organized in heterotrimers (Persson et al., 2007;Nixon et al., 2016;Polko and Kieber, 2019). Cellulose chains synthesized by CESAs form CMFs in the apoplast and become entangled in the cell wall matrix (Johnson et al., 2018). Mutants for either different CESA genes, such as prc1-1, or for regulatory proteins, such as csi1/pom2 and korrigan (kor), have decreased cellulose content and growth defects (Fagard et al., 2000;Lane et al., 2001;Bringmann et al., 2012;Vain et al., 2014). Mutants for pectin and hemicellulose synthesis , such as either quasimodo2 or xyloglucan xylosyltransferase 1, 2 (xxt1xxt2) have less stiff cell walls than wild type (WT) plants. Lack of either pectin or xyloglucans (XGs) in these mutants decreases bundling of CMFs, changing the mechanical properties of cell walls and initiating complex signaling cascades that result in developmental defects (Xiao and Anderson, 2016;Du et al., 2020). Decreased synthesis of pectins and XGs also results in slower CSCs and less cellulose being produced, demonstrating that all components of the cell wall contribute to maintenance of its biochemical composition and integrity, in a continuous feedback loop-like manner (Xiao and Anderson, 2016;Du et al., 2020).
Proteins capable of sensing changes in biomechanical properties/mechanical signals at the level of the apoplast (PM/cell wall) are important for maintenance of cell wall integrity (CWI) (Engelsdorf and Hamann, 2014;Vaahtera et al., 2019). Previous studies have identified several protein families involved in CWI sensing and monitoring the mechanical status of the cell wall. These proteins belong to a family of Catharanthus roseus receptor like kinases (CrRLK), wall associated kinases (WAKs) and mechanosensitive (MS) ion channels (Vaahtera et al., 2019). Some of these proteins (eg. WAK2 and FERONIA (FER)) interact directly with pectins via their extracellular domains, while others (eg. MS ion channels), sense mechanical changes in the PM caused by mechanical perturbations of the cell wall. Changes in cell wall mechanics caused by either different developmental processes or environmental stress factors are sensed by these CWI sensors, eliciting different intra-and extra-cellular responses (Kohorn, 2009;Feng et al., 2018;Gjetting et al., 2020). Rapid mechanically activated (RMA) calcium currents are known to be activated by mechanical signaling. A functioning DEFECTIVE KERNEL1 (DEK1) protein has been shown to be crucial for RMA activity in Arabidopsis (Tran et al., 2017). DEK1 has been identified as a regulator of cell wall biochemical composition and structure (Johnson et al., 2008;Demko et al., 2014;Galletti et al., 2015;Amanda et al., 2016;Amanda et al., 2017).
DEK1 is a major regulator of plant development and growth (Johnson et al., 2005;Johnson et al., 2008). Loss-of-function mutants of dek1 result in embryo lethal phenotypes in both Arabidopsis and maize (Becraft et al., 2002;Johnson et al., 2005). Plants with reduced levels of DEK1 such as dek1 RNA interference and artificial microRNA (amiRNA) lines show a range of severe phenotypes, such as an almost complete absence and misspecification of the epidermal layer (Johnson et al., 2005), an irregular epidermal layer made of large cells (Johnson et al., 2008) or a loss of cell-to-cell adhesion and occasional gaps in the epidermal layer (Galletti et al., 2015). In addition, DEK1 indirectly controls HD-ZIP IV transcription factors which are involved in the regulation of epidermal identity (Galletti et al., 2015). It was proposed that DEK1 could act as a mediator of cellular adhesion by receiving and integrating signals from the apoplast and facilitating cell-to-cell contact through maintenance of epidermal identity (Galletti et al., 2015). DEK1 belongs to the calpain superfamily of regulatory proteases. Calpains function in a range of cellular signaling pathways by modulating their targets through proteolytic processing resulting in altered protein activity, localization, substrate specificity or stability (Ono and Sorimachi, 2012). DEK1 is the only identified member of the calpain family in land plants (Lid et al., 2002;Demko et al., 2014) and its overall structure is distinct from the mostly cytosolic animal calpains (Lid et al., 2002;Wang et al., 2003). DEK1 is an approx. 240 kDa transmembrane protein with a complex structure including 21-23 transmembrane domains (TM), a loop domain, a cytoplasmic regulatory juxtamembrane domain and a proteolytically active CALPAIN domain close to its C-terminus (Lid et al., 2002;Johnson et al., 2008). DEK1 has been shown to undergo autocatalytic cleavage to release the CALPAIN domain into the cytoplasm where it is proposed to proteolytically cleave target proteins to alter their activity (Johnson et al., 2008;Amanda et al., 2016). The catalytic activity of DEK1 CALPAIN domain is dependent on calcium (Wang et al., 2003), as shown for animal calpains (Garcıá Dıáz et al., 2006).Complementing dek1 mutant lines of either Arabidopsis or Physcomitrella patens with only the CALPAIN domain is sufficient to rescue mutants, confirming the CALPAIN region as the catalytically active domain of the protein (Johnson et al., 2008;Demko et al., 2014). Over-expression of CALPAIN (OE CALPAIN) results in altered cell division and expansion, unlike over-expression of the full length DEK1, suggesting the TM and juxtamembrane regions of DEK1 act as regulatory domains (Johnson et al., 2008).
Studies performed on an over-expressor of CALPAIN (OE CALPAIN) and on amiRNA DEK1 lines have shown that DEK1 is a major regulator of cell wall composition and structure in the epidermis of Arabidopsis leaves (Amanda et al., 2016). Immunohistochemical studies on OE CALPAIN lines showed a significant increase in antibody density for both cellulose, high-and low-methyl esterified homogalacturonan (HG) and arabinan while the same cell wall components were less abundant in epidermal cell walls of amiRNA DEK1 compared with the WT (Amanda et al., 2016).
Increased labelling of epitopes for crystalline cellulose were specifically observed in epidermal cell walls of leaves and stems with no change in overall cellulose content in leaves of OE CALPAIN-GFP in a dek1-3 mutant background and an amiDEK1 line (Amanda et al., 2016). In this study, we investigated the role of DEK1 in the regulation of cellulose synthesis at a molecular level. An overexpressor of CALPAIN domain of DEK1 in WT background (pRPS5A:CALPAIN-6XHIS, OE CALPAIN, Galletti et al., 2015) and an EMS single point mutant, dek1-4, which has a C to T nucleotide substitution in the CALPAIN domain and has mild phenotypes (Roeder et al., 2012) were used in this study to enable more direct comparisons to WT plants. Results of this study demonstrate that DEK1 is involved in cellulose synthesis in Arabidopsis cotyledon pavement cells by regulating the dynamics of CSCs. Our data also indicate that DEK1 influences cell wall mechanics.

DEK1 modulated lines show altered responses to isoxaben-induced cellulose perturbations
To investigate the possible involvement of DEK1 in regulation of cellulose biosynthesis, WT and DEK1 modulated lines were treated with isoxaben. Isoxaben is a plant-specific herbicide which inhibits the synthesis of cellulose through interacting with CESAs and causing them to be removed from the PM and internalized into the cell (Heim et al., 1990;Shim et al., 2018). Isoxaben-induced disruption of cellulose synthesis causes reduced root growth and cell swelling. Cellulose deficient mutants have been shown to display altered responses to isoxaben treatment (Engelsdorf and Hamann, 2014;Vaahtera et al., 2019). We measured root lengths of 6-day old WT and DEK1 modulated plants grown on media supplemented with either 2 nM isoxaben or DMSO as a control ( Figure 1A). In response to isoxaben treatment WT plants showed a significant reduction of root length ( Figure 1B). For DMSO control plants no differences in root lengths between WT and OE CALPAIN seedlings were observed. In contrast, we observed significant differences between the mean root length of WT and OE CALPAIN when grown on isoxaben, shown by the 95% CIs of the mean differences of these two groups not overlapping ( Figure 1B). Estimation statistics analysis showed high mean differences in mean root length between OE CALPAIN plants grown on 2 nM isoxaben and the DMSO controls ( Figure 1B). In dek1-4 mutants grown on DMSO media the mean difference in root length compared to WT indicated significantly shorter root growth ( Figure 1A). In response to isoxaben-induced mechanical perturbation of the cell wall dek1-4 mutants resulted in no change in root growth compared to controls ( Figures 1A, B). In addition, no significant differences in the mean difference were observed between dek1-4 and WT ( Figure 1B, 95% CI-s are overlapping). Changes in cellulose levels are known to influence plant tolerance to abiotic stresses such as salt (Endler et al., 2015). The response of DEK1 modulated lines to salt treatment was investigated. WT showed significantly reduced root growth after transfer to salt treatment compared to plants transferred to control media (Supplementary Figure 1). The response of OE CALPAIN plants to salt was more moderate with a smaller reduction in root growth than WT (Supplementary Figure 1). A similar insensitive response to that of isoxaben treatment was observed for dek1-4, with a small mean difference between root length in control and salt-grown plants (Supplementary Figure 1).

DEK1 is involved in fine tuning of epidermal cell wall composition
Studies of DEK1 modulated lines grown on isoxaben and salt suggest cellulose content and/or assembly is altered. Cell wall linkage analysis was undertaken on alcohol insoluble residue (AIR) wall preparations to investigate potential wall compositional differences in 10-day old seedlings of DEK1 modulated lines compared to WT. No statistically significant differences were found (Supplementary Figure 2; Supplementary Table 1). A trend towards reduced levels of cellulose in OE CALPAIN lines was observed and this was investigated further using the Updegraff (1969) method which determines crystalline cellulose content. Of the AIR, crystalline cellulose constituted 51% (w/w) in WT, 42% in OE CALPAIN and 48% in dek1-4 ( Figure 1C). Previous studies have shown DEK1-related changes in cellulose and pectin content predominantly occurs in the outer epidermal cell walls of both young and mature leaves and inflorescence stems (Amanda et al., 2016;Amanda et al., 2017).
Immunofluorescence and immuno-gold labelling of cell wall epitopes was used to determine if wall composition differs at the tissue/cellular scale that may be beyond the resolution of the whole seedling chemical analyses. CBM3a labelling of cellulose epitopes showed slightly lower fluorescence intensity in the outer periclinal epidermal cell wall of both DEK1 modulated lines compared to WT (Supplementary Figures 3A-C). TEM imaging of the cotyledon epidermal cell walls showed significantly decreased immuno-gold labelling of cellulose epitopes with CBM3a in both DEK1 modulated lines compared with WT ( Figures 1D-G). Cotyledon sections labelled with Calcofluor white to detect cellulose, JIM5 and JIM7 to detect partially and highly methyl-esterified HG pectin, respectively, and LM15 to detect XG revealed no obvious differences in fluorescence signal between WT and DEK1 modulated lines (Supplementary Figure 4). We performed immuno-gold labelling with JIM5 and JIM7 antibodies that detect HG pectins in cotyledon epidermal cell walls (Supplementary Figure 5). Our results show that dek1-4 had significantly higher average density of JIM5 immuno-gold particles compared with WT, while OE CALPAIN has a significantly lower count of JIM5 immuno-gold particles ( Supplementary Figures 5A-C, G). Levels of partially-methyl esterified HG detected by the JIM7 antibody were significantly higher in OE CALPAIN and significantly lower in dek1-4 compared to WT (Supplementary Figures 5D-F, H).

CESA transcript levels show no differences between WT and DEK1 modulated lines
Isoxaben growth assays, immuno-labelling and cell wall compositional analysis suggest DEK1 influences levels of cellulose, particularly in epidermal walls. To determine how changes in cellulose may have come about we investigated CSCs at both the transcriptional and post-transcriptional levels. Previous studies have shown that modulation of DEK1 affects expression of numerous cell wall-related genes, however CESA levels were unaffected (Johnson et al., 2008;Amanda et al., 2016). CESA transcript levels were investigated in 6-day old seedlings of WT, OE CALPAIN and dek1-4 lines using qRT-PCR. No significant differences in transcript levels of CESA1, CESA3 and CESA6, encoding primary wall CESAs, were observed between WT and DEK1 modulated lines (p>0.05) using fold change analysis (Supplementary Figure 6). These data suggest DEK1 influences cellulose biosynthesis at the post-transcriptional level. density between WT and DEK1 modulated lines ( Figure 2B). We then proceeded to investigate if DEK1 could affect the velocity of the CSC. Investigation of CSC velocity is commonly used as a proxy for determining cellulose synthesis rates at the cellular level. In brief, the more cellulose CESAs synthesize, the faster CSCs will move along within the plane of the PM (Paredez et al., 2006;Diotallevi and Mulder, 2007;Fujita et al., 2011). CSCs at the PM of the periclinal epidermal cells of 3-day old cotyledons were tracked during 10-min time lapses ( Figure 2C) and kymographs of GFPlabelled CESA3 particles ( Figure 2D) were used to obtain velocities of CSCs ( Figure 2E). CSC velocities in both OE CALPAIN and dek1-4 mutant lines were significantly reduced compared to WT ( Figure 2E). A reduction of CSC velocities in the periclinal cell walls of cotyledon pavement cells is consistent with the findings of  immuno-histochemical assays that indicate significantly reduced cellulose levels in OE CALPAIN compared to WT ( Figure 1D).

CSC membrane trafficking is significantly altered in DEK1 modulated lines
To investigate if DEK1 might be regulating other functional properties of CSC complexes, delivery rate of CSCs and their lifetime at the PM was examined. Fluorescence recovery after photobleaching (FRAP) experiments were performed on OE CALPAIN and dek1-4 plants crossed into a GFP : CESA3 reporter line to determine if the rate of exocytosis of CSCs into the PM was altered. Imaging of GFP : CESA3 particles during the 10 min following photobleaching ( Figure 3A) showed CESAs in OE CALPAIN had significantly faster recovery after FRAP than WT and dek1-4 had significantly slower recovery ( Figures 3B, C). Studies of the lifetime of CSC particles at the PM showed the opposite trends ( Figure 3D). OE CALPAIN CSCs spent significantly shorter time in the PM than WT CSCs, while dek1-4 CSCs stayed s i g n i fi c a n t l y l o n g e r i n t h e P M b e f o r e u n d e r g o i n g endocytosis ( Figure 3D).

DEK1 affects cell wall mechanics of cotyledon pavement cells
The defects in CESA dynamics observed in cotyledon pavement cells prompted us to examine the possible role of DEK1 in influencing structure and physical properties of the CMFs and consequently the mechanical properties of the cell wall. We used atomic force microscopy (AFM) to image and mechanically characterize native cell walls.
We performed AFM imaging on dry cell wall monolayers of isolated abaxial epidermal cell walls from cotyledons of 10-day old Arabidopsis seedlings (Figures 4A-C; Supplementary Figures 7, 8). Imaging was done on the inner surface of the abaxial epidermis outer periclinal walls, which is the newest deposited layer of the cell  Figure 9), as reported for onion epidermal cell wall monolayers imaged by AFM (Kafle et al., 2014;Zhang et al., 2016). Differences in the thickness of CMFs were observed in OE CALPAIN and dek1-4 (Figures 4A-C) compared to WT. In OE CALPAIN, thicker bundles of CMFs were observed as well as amorphous deposits on the cell wall surface ( Figure 4B). Diameters of the CMF bundles were quantified from height channel images (Supplementary Figure 9). Results showed that the CMF bundles of both OE CALPAIN and dek1-4 had significantly higher mean diameters than WT CMFs ( Figure 4D). Analysis of the distribution of CMF dimeters shows that both DEK1 modulated lines have a higher frequency of bundled CMFs in the range of 15-40 nm whereas WT CMFs are mostly represented in the range of 1-15 nm, implying that WT has either an abundance of single CMFs (~3.5 nm is the diameter of a typical CMF (Zhang et al., 2016)) or smaller bundles of CMFs ( Figure 4E).
Previous studies have shown that changes of cell wall composition in cell wall mutants of Arabidopsis can affect mechanical behavior of the cell wall and CMF properties (Xiao and Anderson, 2016). To examine cell wall mechanical properties in DEK1 modulated lines, we performed AFM nanoindentation on the isolated abaxial epidermal cell walls of 10-day old Arabidopsis cotyledons. Nanoindentation generated force-displacement curves, were used to calculate cell wall elastic moduli (Supplementary Figures 10A, C, E). It is worth noting that within all analyzed genotypes, a large range of elastic moduli was observed ( Figure 4F; Supplementary Figures 10B, D, F). In WT, values ranged from 0.1 Mpa to 5 Mpa, which is consistent with previous data showing highly heterogenous distribution of cell wall mechanical properties (Yakubov et al., 2016). Analysis of the apparent elastic modulus in OE CALPAIN and dek1-4 suggests that both lines had significantly stiffer cell walls compared to WT ( Figure 4F). These results suggest DEK1 is involved in regulating and maintaining mechanical properties of the cell wall.

Discussion
In this study we show that DEK1 affects the synthesis of both cellulose and pectins in early developmental stages of Arabidopsis. Both DEK1 modulated lines show decreased crystalline cellulose in outer epidermal cell wall of cotyledons. Velocity of CSCs in the epidermal pavement cell PMs, which is used as a proxy for the biosynthetic activity, is significantly reduced in both DEK1 modulated lines, while the density of CSC complexes is similar to that of the WT. PM dynamics of CSCs is significantly altered in both lines, with OE CALPAIN CSCs having significantly slower CSC exocytosis and faster endocytosis, while these features of CSC dynamics were reversed in dek1-4 plants. Epidermal cell wall stiffness was significantly increased in both DEK1 modulated lines compared to WT.

DEK1 in cellulose assembly
DEK1 is proposed to act as a mechano-sensor protein at the PM (Amanda et al., 2016;Tran et al., 2017). Upon perception of mechanical stimuli, originating either from other growing cells in the tissue or from a cell's own turgor pressure, or an external environmental signal, the CALPAIN domain of DEK1 is proposed to be auto-catalytically released into the cytoplasm (Amanda et al., 2016;Tran et al., 2017). DEK1 CALPAIN likely acts as a regulatory protease (Wang et al., 2003;Johnson et al., 2008) and downstream signaling initiates changes in cell wall biosynthesis/remodeling and epidermal specification (Demko et al., 2014;Galletti et al., 2015;Amanda et al., 2016;Amanda et al., 2017). In this study we show DEK1 influences CESA dynamics, cellulose content, pectin restructuring and mechanical properties of epidermal walls.
It is currently understood that slower movement of CSCs across the PM, a proxy for a lower rate of cellulose biosynthesis, results in less cellulose synthesized (Diotallevi and Mulder, 2007). The significant reduction of CSC velocity in OE CALPAIN and dek1-4 ( Figures 2D, E) support a role for DEK1 in regulation of CESA dynamics. Mutants for CESA6/prc1-1, and CESA1/any1, have significantly reduced CESA velocities in dark grown hypocotyls (Bischoff et al., 2011) that is proposed to be the major cause of a 30% reduction in growth (Fagard et al., 2000). There is no data in the literature that CESA proteins undergo proteolytic processing after insertion (by exocytosis) into the PM. RT-qPCR analysis of CESA1,3 and 6 expression levels, and measurement of CESA density at the cotyledon pavement epidermal cell PMs showed no difference between WT and DEK1 modulated lines. These data support DEK1 regulation of CSCs at the post-translational level. An interesting possibility is that DEK1 regulates CSC activity indirectly, potentially through interactions with either CESA regulatory proteins or the cytoskeleton components that guide CSCs. Candidates would include CSI1/POM2 that facilitate binding between CSCs and CMTs (Gu et al., 2010;Bringmann et al., 2012), the endo-1,4-beta-glucanase KORRIGAN (KOR) involved in CESA regulation (His et al., 2001;Vain et al., 2014) and PATROL1 (PTL1) that interacts with exocyst complex proteins and CS1/POM2 to deliver CSCs to the PM (Zhu et al., 2018). Mutants kor1-1 and kor1-3 display a similar reduction in CSC mobility to DEK1 modulated lines when compared to WT plants (Vain et al., 2014). In addition kor mutants have been shown to have decreased cellulose content, but also increased pectin abundance, which was interpreted as a compensatory mechanism and part of feedback loops connecting cellulose and pectin synthesis to changes in the mechanical properties of the cell wall (His et al., 2001).
It could be conceivable that DEK1 is involved in CSC guidance along CMTs, either by modifying the activity and properties of CSI1/POM2 or some other CESA-CMT interface proteins. Reduced CSCs velocities and cellulose content is observed in pom2-1 mutants compared to WT plants as well as spiral twisting of the entire rosette and leaves. Previous studies of lines with reduced levels/activity of DEK1, including dek1-4, showed epinastic curling of cotyledons and leaves (Roeder et al., 2012;Galletti et al., 2015). CMTs are responsive to mechanical stress and assumed to align in the direction of maximal stress (Hamant et al., 2008). During salt stress, CMTs are rapidly disassembled, and CSCs removed from the PM (Endler et al., 2015). COMPANION OF CELLULOSE SYNTHASE (CC) proteins bind to CMTs and promote salt tolerance by stabilizing CMTs and enabling cellulose synthesis. CMTs were shown to re-organize normally in dek1-4 mutants in response to ablation in cotyledons, however, were more highly orientated than in WT pavement cells due to changes in cell shape (Galletti et al., 2015). A study in animals has shown that a nonproteolytic variant of CALPAIN, CALPAIN6 (CAPN6), stabilizes microtubules by binding to them and preventing their depolymerization (Tonami et al., 2011). Unlike CAPN6, DEK1 CALPAIN has proteolytic activity, and it would be interesting to investigate if DEK1 CALPAIN can interact with CMT and potentially disrupt binding of CESA complexes (Tonami et al., 2011).
Upregulation of the exocytotic module of the CSC trafficking machinery in OE CALPAIN (Figures 3B, C) could be one explanation for the faster exocytosis of the CESAs, as observed by faster post-FRAP recovery of CSC particles (Figures 3A, B). Exocytosis of CSCs is regulated by a large protein complex (Polko and Kieber, 2019), which interacts with CSI1 and PTL1 to deliver CSCs to the PM (Zhu et al., 2018). The frequency of CSC insertions and alignment with the CMT was highly decreased and the delivery rate of CSC severely diminished in csi1-3/ptl1-2 double mutants. In addition, ptl mutants also have lower CSC velocities, and both ptl1-2 and csi1-3/ptl1-2 lines have significant decreases in cellulose content (Zhu et al., 2018). The contrasting effect which OE CALPAIN and dek1-4 have on CSC PM trafficking could mean that DEK1 is interacting with several components involved in CSC trafficking to (and in) the PM, and that its role in signaling involves multiple target proteins.

DEK1 and cell wall mechanics
The epidermis is a key regulator of plant growth and this is influenced by the mechanical properties of the cell wall Moulia et al., 2021). In DEK1 modulated lines, pectins and cellulose are altered yet the dynamics of these changes are unclear and if direct or indirect. Previous studies (Amanda et al., 2016;Amanda et al., 2017), have shown that DEK1 affects cell wall composition almost exclusively on the level of leaf outer epidermal cell walls. In leaf 4 of 17 day old Arabidopsis plants, increased labelling of CBM3a recognizing cellulose and JIM5 detecting lowmethyl esterified HG was found in dek1-3 OE CALPAIN GFP compared to WT and an amiRNA DEK1 line. In this study, decreased levels of CBM3a and JIM5 labelling was found in epidermal walls of cotyledons of OE CALPAIN. DEK1 potentially acts in different regulatory networks and/or on tissue specific targets during early and late developmental stages of Arabidopsis. Alternatively, changes in cell walls at later stages could be indirect and result from compensatory changes initiated by cell wall integrity pathways (Vaahtera et al., 2019). Further analysis is required to uncover the dynamics of DEK1-regulated wall changes in future.
Compensatory mechanisms to counteract changes in cell wall composition and biomechanics are known to occur, for example, pectins have been shown to increase in response to reduced levels in cellulose and lead to increased stiffness of the cell wall (His et al., 2001). DEK1 has previously been shown to regulate expression of p e c t i n -r e l a t e d g e n e s , i n c l u d i n g G A U T s a n d p e c t i n methyltransferases as well as altered deposition of pectic polysaccharides in leaf epidermal cell walls of OE CALPAIN GFP (Amanda et al., 2016;Amanda et al., 2017). Mutants in pectin methyltransferase, quasimodo2 and tumorous shoot development2, have decreased levels of both HGs and cellulose, decreased CESA velocities, thinner and more fragmented CMFs (Mouille et al., 2007;Du et al., 2020), while mutants in putative galacturonic acid transferase, quasimodo1, experience severe defects in cell-to-cell adhesion due to low HG-content (Bouton et al., 2002;Verger et al., 2018). Increased partially methyl-esterified HG in dek1-4 is consistent with a role for pectin-related increased bundling of CMFs in this line. Stiffer cell walls in OE CALPAIN, with reduced levels of partially methyl-esterified HG and increased levels of highly methyl-esterified HG requires further investigation. The relationship between the degree of HG methyl-esterification and cell wall stiffness is unresolved. It could depend upon local concentrations of Ca 2+ in the apoplast , orientation of the cell wall, such as anticlinal vs periclinal cell walls (Haas et al., 2020), activity of polygalacturonases (Yang et al., 2018) and many as yet unknown factors. Changes in interactions between pectins and CMF could also influence mechanical properties of the cell wall in DEK1 modulated lines ( Figure 4E). NMR studies of 'never dried' onion epidermal cell walls have shown that HGs make the majority of contacts with CMFs, which could mean that they act as primary mechanical tethers . Pectins affect cell wall mechanical properties through poroelastic effects, without impacting the structural properties of cellulose (Lopez-Sanchez et al., 2015).
In future it will be essential to identify the proteolytic targets of DEK1 CALPAIN and the dynamics of cell wall changes in response to altered DEK1 to determine direct vs indirect effects. In this study, pectin composition and root inhibition by isoxaben have opposite patterns in OE CALPAIN and dek1-4 lines, whereas cellulose composition and cell wall stiffness show similar trends. Further studies are required to determine if, for example, the CALPAIN domain is required for the former effects, whereas the latter effects are a response to the induced cell wall modifications. Different stress conditions are known to lead to similar responses of plants in terms of growth and morphogenesis (Tenhaken, 2014). However, the underlying molecular mechanism of these changes and their associated signaling cascades can differ greatly. CWI sensors of the CrRLK family, including FER and THE1 are involved in responses to stress, mediated by RALF peptide binding (Feng et al., 2018;Blackburn et al., 2020;Zhang et al., 2020). RALF peptides are known to be processed by substilin-like proteases (Stegmann et al., 2017), however, other proteases such as DEK1 may be involved in generating mature RALF ligands. DEK1 signaling cascades are still completely unknown and further work is required to determine if cross-talk between CrRLK signaling and DEK1 pathways occur. Identification of CALPAIN targets through the use of targeted proteomics approaches is essential for understanding the pathways through which DEK1 acts, and its precise role in plant growth and development.
Plants were grown vertically in petri dishes, on MS 1% (w/w) sucrose media, 0.8% agar, pH 5.8 (Murashige and Skoog, 1962), without vitamins. Plants were grown in Percival growth cabinets, model CU36L/LT (Percival Scientific, Perry, Iowa, USA), at a temperature of 21°C and light intensity of 95-111 mol m -2 s -1 . Growth conditions used in different experiments are detailed in sections below.
F2 DEK1 modulated lines crossed with GFP : CESA3 were used for CESA dynamics experiments. Non-genotyped segregating F2 plants were grown on MS 1% sucrose media for 3 days, imaged and then genotyped for identification of DEK1 modulated lines containing pCESA3:GFP : CESA3 constructs. After confocal imaging, OE CALPAIN pCESA3:GFP : CESA3 F2 plants were transferred onto MS 1% sucrose media supplemented with 10 µg ml -1 hygromycin. Plants were grown horizontally for 10-14 days together with Col-0 and GFP CESA3 plants as positive controls. Only imaging data for OE CALPAIN pCESA3:GFP : CESA3 plants which survived the hygromycin selection were used for CESA dynamics analysis.
After CESA imaging, dek1-4 pCESA3:GFP : CESA3 plants were transferred to MS 1% sucrose media supplemented with Plant Preservative Mixture (Plant Cell Technology, Washington DC, USA) diluted in ratio 1:1000. Plants were grown until they developed the first true leaves (7-10 days post imaging). A leaf was used to isolate DNA using Edwards buffer method (Edwards et al., 1991) and the DNA analyzed by dCAPS (Neff et al., 1998) (Supplementary Table 2) or KASP genotyping (LGC Genomics, Berlin, Germany) to identify dek1-4 homozygous plants. Only imaging data for dek1-4 homozygous plants was used in further CESA dynamics analysis.  Supplementary Table 2. Lightcycler ABI 7900 HT (Applied Biosystems, Foster City, California, United States) lightcycler was used for RT-qPCR using amplification, conditions: 50°C -2 min; 95°C -10 min; 40 cycles of 95°C for 15 s, 60°C for 1 min; 95°C for 15 s, 60°C for 15 s, 95°for 15 s. qPCR data was analyzed using SDS v 2.4.1 analysis software (Thermo Fisher Scientific, Waltham, MA, USA). CESA expression levels were normalized against GLYCERALDEHYDE-3-PHOSPHATE DEHYDROGENASE (GADPH) which was used as a reference gene.

Analysis of CESA transcript levels
Relative gene expression values expressed as fold changes were calculated using the method of (Livak and Schmittgen, 2001). CESA expression levels were treated as 1 in Col-0 WT and values in DEK1 modulated lines were calculated compared to Col-0. Gene expression fold changes lower than 0.5 and higher than 2.5, compared to Col-0, were treated as significant.

Cell wall composition analysis
Cell wall composition analysis was performed on pooled 10-day old whole seedlings. Analysis of the cell wall polysaccharide composition of alcohol insoluble cell wall residues (AIR) from Col-0 and DEK1 modulated lines was performed according to (Pettolino et al., 2012). Analysis of the total crystalline cellulose in seedlings was performed using Updegraff assay (Updegraff, 1969) of AIR (10-50 mg).

Spinning disc confocal microscopy of CSC
Three-day old seedlings were mounted on microscope object slides and imaged using instruments and settings as described in (Sampathkumar et al., 2013). Photobleaching used to access CSC trafficking dynamics was performed using a FRAP/PA system (Roper Scientific, Acton, MA, USA) integrated into the spinning disc confocal imaging system, as described in (Sampathkumar et al., 2013).

CSC timelapse processing
Image processing was performed using ImageJ software (Rasband, W., NIH, Bethesda, MD, USA). Brightness and contrast of the raw time-lapses were modified. Subtract background function (rolling ball size 30-40 pixels) and Walking average plugin (three frames averaged) were used to remove background noise and to obtain clearer images.
Density of CSC particles was determined from confocal timelapse images using IMARIS 7.4 image analysis software (Bitplane, Oxford Instruments, Oxford, United Kingdom). Time-lapse image stacks were pre-processed in ImageJ as described above and imported into IMARIS. Spot detection and tracking function of the IMARIS was used to detect and label CSC particles. CSC particle numbers used to calculate mean CSC density were taken from three random timeframes from each analyzed pavement cell form every imaged cotyledon for each of the analyzed genotypes. Density was calculated as mean of CSC particles density per one mm 2 for each cell, plant and genotype.
Velocity analysis of CESA particles imaged using confocal microscopy was performed using Fluorescent Image Evaluation Software for Tracking and Analysis (FIESTA) software (Ruhnow et al., 2011), which is available under open license from FIESTA, University of Dresden Fusion Forge Wiki webpage (https:// fusionforge.zih.tu-dresden.de/plugins/mediawiki/wiki/fiesta/ index.php/FIESTA). Raw confocal time-lapse images were imported into FIESTA. Maximum signal intensity projections were used to generate kymographs. Kymographs were generated from the CSC trajectories manually drawn using segmented line tool in FIESTA.
Analysis of the data obtained by FRAP was performed using a modified version of the CSC density measurement method described above. CSC particles detection and quantification was performed on 9 x 9 mm cell patches rather than 10 x 10 mm, which were the dimensions of bleached surface. After performing detection of fluorescently labelled CSC particles using IMARIS software, CSC post-FRAP density, recovery rates and average plasma membrane life-time of the complexes was calculated as described in (Sampathkumar et al., 2013).

Cell wall isolation for AFM topographical imaging and nanoindentation
Cotyledon pavement epidermal cell walls were isolated for analysis with the atomic force microscope (imaging and nanoindentation). Rectangular microscope coverslips (Roth, Karlsruhe, Germany) were covered with 200 ml of 0.01% poly-Dlysine-hydrobromide (PDL; P6407, Sigma Aldrich, St. Luis, MO, USA) in deionized water. Coverslips were left for 15-20 min to allow the PDL to polymerize, after which coverslips were washed thoroughly with deionized water and dried in a stream of nitrogen.
We used 10-day old seedlings of Arabidopsis to obtain cotyledons for cell wall isolation. The larger of the two cotyledons was cut off using surgical tweezers, placed on a PDL coated coverslip (with the abaxial side facing the coverslip) and the following steps were performed under a stereomicroscope (Zeiss Stemi 508, Zeiss, Jena, Germany). The cotyledon was pressed gently on one side using surgical tweezers and then a vertical cut was made using a micro-surgery scalpel (Fine Scientific Tools, Vancouver, Canada), across the middle of the cotyledon in a direction perpendicular to its longer axis. After the vertical cut, a second cut was performed in parallel to the longer cotyledon axis to remove all leaf tissues and expose the abaxial epidermis with patches of the cell wall monolayers on its edge (Supplementary Figure 6). Regions with cell wall monolayers were observable under 50x magnification as small opaque patches of material on the edge of the abaxial epidermis (Supplementary Figure 6).
The remaining parts of the dissected cotyledon were carefully attached to the surface with nitrocellulose based red stained adhesive (nail polish) to immobilize the sample during imaging and mechanical probing. Adhesive was applied with an injection syringe needle, carefully avoiding regions with cell wall monolayers and left to cure for 5-10 seconds. Samples were then incubated with 500 ml of 1% SDS for 20-30 s to remove organelles, proteins and other cellular components released during the cotyledon dissection. SDS was removed by washing with deionized water, ensuring that the dissected cotyledon does not curl up on the coverslip. For highresolution AFM topological imaging the samples were left overnight to completely dry while, samples used for mechanical measurements were stored in water to prevent loss of native mechanical properties.

AFM topographical imaging
For measuring the CMF diameter, a Dimension 3100 AFM (Veeco, Plainview, NY, USA) was used. Imaging was performed in tapping mode in air, using silicon ARROW-NCR cantilevers (tip radius <10 nm; Nanoworld, Neuchatel, Switzerland).The optical camera integrated into the AFM setup was used to identify regions of interest for imaging. Regions with a size of 2 x 2 mm 2 were imaged with a resolution of 512 x 512 pixels. We recorded height channel images (Supplementary Figure 8) and amplitude channel images ( Figures 4A-C). Imaging of all genotypes was performed in biological duplicate; that is, the same experiments were repeated twice with two sets of plants grown independently at different times.

Analysis of the cellulose microfibril images
High-resolution, high-quality cell wall monolayers imaged in dry conditions were analyzed using custom made MATLAB script to determine the diameter of the imaged CMFs. For the analysis of CMF diameter, we used the height channel AFM images (Supplementary Figure 8) as obtained from the AFM software.
The extracted values of the CMF diameter were pooled for each genotype and CMF diameter values.

Mechanical characterization of isolated cell walls using atomic force microscopy
To determine the apparent elastic modulus of the isolated cell wall monolayers, we performed AFM-based nanoindentation (manual force mapping mode), applying a method previously described (Yakubov et al., 2016) and modified for our experimental conditions. Nanoindentation experiments were performed using a JPK Nanowizard 3 AFM (Bruker Nano GmbH, Berlin, Germany) mounted on an Olympus IX71 inverted phase-contrast microscope (Olympus Corporation, Shinjuku City, Tokyo, Japan). Coverslips with prepared samples were mounted onto the AFM sample holder as described in (Yakubov et al., 2016).
An optical CCD camera integrated into the AFM setup was used during the entire measurement; first to align the cantilever and then to perform constant checks of the sample and cantilever properties. We used DNP-10 cantilevers (cantilever A) with a nominal spring constant of 0.35 N m -1 (Bruker Corporation, Billerica, MA, USA). The optical sensitivity of the cantilevers was determined from contactless oscillations after which the thermal noise method was used to obtain the spring constant. The used cantilevers had spring constants ranging from 0.30-0.38 N m -1 .
Before starting the mechanical characterization, topological imaging of the cell wall monolayer was performed to identify the thinnest regions of the sample. First, using intermittent contact mode, we imaged an area of 5 x 5 mm 2 with a resolution of 512 x 512 pixels. Imaging of a 5 x 5 mm 2 region was always performed near the edge of the dissected cotyledon, so that part of imaged area represented the bare glass slide to which the cotyledon was attached. The flat glass slide surface, which contained some tissue debris after dissection, was used as a reference (position 0 nm), to determine the thickness of the imaged sample (Supplementary Figure 7). If sample thickness was in the 0.5-2.0 mm range, we selected a smaller 1 x 1 mm 2 region within the original 5 x 5 mm area and repeated the topography imaging to re-confirm cell wall thickness (Supplementary Figure 7). If the thickness of the topological features inside this region remained in the 0.5-2 mm range, we performed nanoindentation on that region. This range of sample thicknesses values was empirically chosen. During method development, we determined that regions thicker than 2.0 mm were either wrinkled cell wall patches or cell wall bilayers (i.e. from both adaxial and abaxial sides; thickness 2.5-4 mm). Samples thinner than 0.5 mm were not used to avoid non-specific surface interactions and to minimize the contribution of the underlying glass surface to the elastic modulus.
For nanoindentation, the 1 x 1 mm 2 region of interest was further divided into a 16 x 16 array. Each position was indented once to obtain a force map consisting of 256 force-displacement curves. For each approach-retract cycle, the contact force was set to 2.5 nN and the z-piezo speed was 1 mm s -1 . The average indentation depth was 180 nm for Col-0, 164 nm for OE CALPAIN and 147 nm for dek1-4, while indentation depth values where mostly in the range of 100-350 nm (Supplementary Figure 9G). While the indentation depths partially exceeded 10% of the sample thickness, the values lie in a similar range. Contributions from the underlying glass surface can thus be expected to be similar for all three samples.
One force map was obtained from each dissected cotyledon. Imaging of all genotypes was performed in biological duplicate; the same experiments were repeated two times, with two sets of plants grown independently at different times. From each replicate, cotyledons from 4-6 different plants were imaged, one cotyledon per plant. Force maps were analyzed using JPK SPM processing software, version 6.1.9 (Bruker Nano GmbH). As adhesion was observed in the retract segments of the force-displacement curves, the approach segments were fit with the Hertz model to obtain the apparent elastic modulus. The Poisson ratio was set to 0.5 and the baseline value was unpinned and set to 0. The extracted apparent elastic moduli from different cotyledons for the same genotype were pooled and dek1-4 and OE CALPAIN values were compared with Col-0.

Isoxaben and salt assays
Plants used for the isoxaben treatments were sterilized and grown for 6 days on MS media containing 1% sucrose, with the addition of 2 nM isoxaben (Sigma-Aldrich, St. Louis, MO, USA) dissolved in DMSO (Honeywell, Charlotte, NC, USA), and on MS 1% sucrose containing media supplemented with 2nM DMSO as a control. Experiments were performed in biological duplicates and technical triplicates. Salt stress experiments were performed as described in (Feng et al., 2018).
Plants were imaged using a Keyence VHX-6000 digital microscope (Keyence Corporation of America, Itasca, IL, USA). Whole plates were placed on a motorized microscope stage and series of images were acquired in sequence covering the entire surface of the plate, and then automatically stitched by the software to generate a single whole-plate image. We used objective ZS20, 20X magnification, on black field, with epi-illumination, for imaging 6day isoxaben grown and salt-treated plants. Root lengths were quantified using Measure length function of ImageJ software (Rasband, W., NIH, Bethesda, MD, USA).

Statistical analysis
Unless stated otherwise, all statistical analysis in this study were performed using unpaired Student t-test, results were treated statistically significant if p<0.05. We used online Student t-test calculator at the webpage https://www.graphpad.com/ quickcalcs/ttest1.cfm.
Analysis of data from isoxaben growth and salt assays was performed using estimation statistics. After obtaining root length data, statistical significance of the results was assessed using the online, open-access Estimation Stats software (https:// www.estimationstats.com/#/) and data are plotted using Cumming estimation plot as described in (Ho et al., 2019).

Data availability statement
The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding authors.

Author contributions
LN performed CSC imaging, AFM measurements and imaging and CWI sensing experiments, and analyzed all the data from these experiments. KJ, TB, and AS conceptualized and designed the research. LN, GY, KJ, and KB conceptualized, developed, and validated cell wall isolation method and AFM imaging and force spectroscopy protocols. KJ and AS made DEK1 crosses with CSC reporter line. YM performed immunofluorescences and immunohistochemical imaging experiments and analyzed the data. GY wrote the MATLAB script for analysis of AFM cellulose images. LN wrote the original draft of the article manuscript. KJ, TB, KB, and AS edited, curated and corrected the original draft of the manuscript. KB, GY, and YM gave valuable input and comments on the article manuscript.