Structure-function analysis of the maize bulliform cell cuticle and its role in dehydration and leaf rolling

The cuticle is a hydrophobic layer on the outer surface plant shoots, which serves as an important interaction interface with the environment. It consists of the lipid polymer cutin, embedded with and covered by waxes, and provides protection against stresses including desiccation, UV radiation, and pathogen attack. Bulliform cells form in longitudinal strips on the adaxial leaf surface, and have been implicated in the leaf rolling response observed in drought stressed grass leaves. In this study, we show that bulliform cells of the adult maize leaf epidermis have a specialized cuticle, and we investigate its function along with that of bulliform cells themselves. Analysis of natural variation was used to relate bulliform strip pattering to leaf rolling rate, providing evidence of a role for bulliform cells in leaf rolling. Bulliform cells displayed increased shrinkage compared to other epidermal cell types during dehydration of the leaf, providing a potential mechanism to facilitate leaf rolling. Comparisons of cuticular conductance between adaxial and abaxial leaf surfaces, and between bulliform-enriched mutants vs. wild type siblings, provided evidence that bulliform cells lose water across the cuticle more rapidly than other epidermal cell types. Bulliform cell cuticles have a distinct ultrastructure, and differences in cutin monomer content and composition, compared to other leaf epidermal cells. We hypothesize that this cell type-specific cuticle is more water permeable than the epidermal pavement cell cuticle, facilitating the function of bulliform cells in stress-induced leaf rolling observed in grasses. One sentence summary Bulliform cells in maize have a specialized cuticle, lose more water than other epidermal cell types as the leaf dehydrates, and facilitate leaf rolling upon dehydration.

Plants display a variety of responses to environmental stresses. An important drought stress 51 response in grasses is reversible leaf rolling along the longitudinal leaf axis upon water limitation 52 Disproportionate shrinkage of bulliform cells, located only on the adaxial side of the leaf, during 203 leaf dehydration is thought to create a hinge-like effect promoting leaf rolling. However, we 204 could not find published experimental evidence confirming this hypothesis. Indeed, it is 205 technically challenging to investigate this, since conventional methods permitting visualization of 206 plant tissues at the cellular level have the potential to cause cell shrinkage in their own right 207 (e.g. fixation and dehydration prior to embedding in a sectioning medium), or reverse cell 208 shrinkage (e.g. if freshly cut hand sections of dehydrated tissue are mounted in aqueous 209 medium under a cover slip). To overcome these problems, a newly-established cryo-confocal 210 imaging method was employed. After 4 hours of dehydration of intact (detached) adult leaves, 211 or no dehydration, tissue fragments were shock-frozen in optimal cutting temperature compound 212 (OCT) and cross-sectioned in a cryo-microtome. Autofluorescence of tissue cross sections 213 exposed at the block face was then imaged in a custom-built liquid N 2 -cooled chamber mounted 214 on a confocal microscope ( Figure 2). In comparison with cross sections of fully hydrated control 215 leaves (Figure 2A), pavement cells as well as bulliform cells (red circles) were smaller in 216 dehydrated (rolled) leaves ( Figure 2B). Bulliform cell shrank more than pavement cells ( Figure  217 2C,D), but no difference in shrinkage was observed comparing adaxial and abaxial pavement 218

cells. 219
To investigate whether differential shrinkage of bulliform cells is due to water loss to the 220 atmosphere (i.e. via evaporation across the bulliform cuticle) or to movement of water into 221 neighboring cells, pavement cell volumes were examined as a function of proximity to bulliform 222 supporting the hypothesis that the bulliform cuticle is more water permeable. To further 297 investigate this hypothesis, a similar dehydration experiment compared g c of adaxial leaf 298 surfaces of adult wild-type leaves (containing bulliform cells), to that of bulliform-free abaxial 299 surfaces. This was achieved by covering one or the other surface with petroleum jelly to prevent 300 water loss from the covered side of the leaf (Fig 6B). Increased dehydration of the bulliform-301 containing, adaxial side of the leaf was observed compared to the abaxial, bulliform cell-free 302 side, while the control without petroleum jelly constituted the full cuticular conductance. In 303 conclusion, complementary experiments investigating the relationship between bulliform 304 differential shrinkage, assisting the rolling of grass leaves upon dehydration. 307 308 309 310 Bulliform cell cuticle nanoridges are not the main driver of increased dehydration 311 Bulliform cells show a reticulate pattern of cuticle nanoridges on their surfaces (Becraft et al., 312 2002), increasing cuticular surface area relative to overall cell surface area, and providing a 313 possible mechanism to increase the rate of dehydration of bulliform cells relative to other 314 epidermal cell types. To address the question of whether the observed increase in water loss 315 from bulliform cells and bulliform-enriched tissues could be explained by this mechanism, high 316 resolution surface imaging of leaf glue impressions was performed with a Keyence VHX-6000 317 digital microscope system (Figure 7). Bulliform cells in wild-type leaves (white arrowheads) were 318 organized in strips of 3-5 cells, and their cuticles displayed nanoridges aligned with the 319 proximodistal axis of the leaf, often appearing to span cell-to-cell boundaries (white arrow) 320 ( Figure 7A-C). Adjacent pavement cells (black arrow) lacked these nanoridges. wty2 mutants 321 displayed normal bulliform strips (white arrowhead) as well as abnormal, epidermal cell bumps 322 which, upon closer inspection, revealed an even denser than normal pattern of cuticular 323 nanoridges ( Figure 7D-F). dek1-D mutants usually had wider than normal bulliform strips, but 324 cells in these strips varied with respect to cuticular nanoridges: cells in the center had 325 nanoridges, whereas those towards the outer edges had few or no nanoridges ( Figure 7G-I). 326 Interestingly, in the last BC mutant, Xcl1, abnormal bulliform-like cells displayed no cuticular 327 nanoridges whatsoever ( Figure 7J-L). These differences are not due to variations in leaf water 328 content, as all leaves were fully turgid at the time glue impressions were made. Imaging of 20 abaxial leaf impressions of some of the mutants (Supplemental Figure S3) revealed bulliform-330 observed in only some areas, while others displayed bulliform-like cells with a smooth surface 332 (Supplemental Figure S3A,B). As for the adaxial surface, no nanoridges could be detected on 333 the abaxial sides of Xcl1 mutant leaves (Supplemental Figure S3C,D). In conclusion, increased 334 water loss in the three analyzed bulliform-enriched mutants cannot be caused solely by surface 335 area increase due to the presence of cuticle nanoridges on their excess bulliform-like cells, 336 since not all of the mutants displayed these cuticle features despite a higher cuticular 337 conductance rate. 338 339 Bulliform-enriched cuticles have a unique biochemical composition with major 340

differences in cutin 341
In an effort to identify unique and possibly functionally significant components of bulliform cell 342 cuticles, we sought to biochemically characterize them. Since no method was available to 343 physically separate bulliform from pavement cells on the scale needed to biochemically analyze 344 their cuticles directly, two complementary approaches were taken to compare these cuticle 345 types indirectly with respect to both wax and cutin monomer composition: (1) adaxial (bulliform-346 containing) and abaxial (bulliform-free) cuticles were compared (see methods for information on 347 how this was achieved), and (2) leaf cuticles of the bulliform-overproducing mutants (wty2, 348 dek1-D, and Xcl1) were compared to their respective wild-type siblings. We reasoned that 349 compositional changes seen in both comparisons should provide insight into the specific 350 composition of the bulliform cell cuticle. 351 The total lipid polyester (cutin) monomer load of wild-type adaxial cuticles was 352 significantly higher than on the abaxial side ( Figure 8A). The relative abundance of individual 353 monomer classes in both adaxial and abaxial cuticles was determined via normalization to the 354 total monomer load on the respective side. Some monomers, including 18:0 FA, several hydroxy 355 9-epoxy-18-OH, showed a major accumulation in the adaxial surface containing BCs. All three 357 bulliform-enriched mutant cuticles ( Figure 8B) also had a higher total cutin monomer load 358 compared to wild-type, thus we compared their relative lipid polyester monomer composition. 359 Most monomers found to be different in the adaxial/abaxial comparison did not overlap or 360 overlapped only partially with the differential abundances detected in the bulliform mutant 361 analysis. However, the polar cutin component that is highly enriched in the BC-containing 362 adaxial wild-type cuticles, 18:0 9-epoxy-18-OH, was also significantly increased in all three 363 bulliform-enriched mutant cuticles. These findings point to 18:0 9-epoxy-18-OH as a cutin 364 monomer that is enriched in bulliform cell cuticles. 365 Cuticular wax analysis of bulliform-enriched tissues created a more complex picture 366 ( Figure 9). Total wax load on the adaxial (bulliform-containing) side of wild-type leaves was 367 increased ( Figure 9A), but decreased in two out of three of the bulliform-enriched mutants 368  showed relative abundance changes that were in agreement between adaxial/abaxial 377 comparisons and bulliform mutants vs. wild-type comparisons. While all three mutants showed 378 increased free fatty acid content ( Figure 9D), this difference was not seen in the adaxial/abaxial 379 comparison ( Figure 9B). Fatty alcohols and wax esters were decreased in two of the three 380 24 mutants, but this difference was also not observed in the adaxial/abaxial analysis. Aldehydes, 381 a difference in the bulliform mutant comparisons. Analysis of single compounds within each wax 383 class (Supplemental Figure S4) also did not clearly point to specific wax molecules as specific 384 to or enriched in BC cuticles. Thus, apart from evidence for hydrocarbon enrichment, we found 385 no clear indication from these analyses of enrichment or depletion in individual wax classes or 386 molecules in bulliform cuticles. 387 388

Transcriptome analysis suggests a role of ferulate in BC cuticle maturation 389
In order to identify genes whose functions underlie unique features of bulliform cell cuticles, 390 gene expression analysis of the three bulliform mutants was conducted. The data were 391 analyzed to search for genes differentially regulated in the same direction during cuticle 392 maturation in all three mutants compared to their wild-type siblings. To this end, the previously 393 characterized zone of cuticle maturation in developing adult leaves (from 10-30% of the length 394 of a partially expanded leaf #8; Bourgault et al., 2020) was harvested from mutants and 395 corresponding wild-types, and analyzed via RNAseq (Supplemental Figure S5). The three 396 mutants displayed a varying degree of differential gene regulation, with 4833 differentially 397 expressed genes (DEG) for wty2 in the analyzed zone (Supplemental Figure S5A), comparably 398 less differential gene expression for dek1-D with only 132 DEG (Supplemental Figure S5B), and 399 527 DEG detected for Xcl1 (Supplemental Figure S5C). However, the overlap between the 400 DEGs in all three mutants was minimal, with only 4 genes showing a differential regulation in all 401 three datasets (Supplemental Figure S5D, Supplemental Table S2), and only two of these 402 genes deviating from wild-type values in the same direction in all three mutants (Supplemental 403 Figure S5E). One of these genes, Zm00001d008957, which showed increased expression in all 404 three BC mutants, encodes a putative indole-3-acetic acid-amido synthetase and is annotated 405 as aas10 (auxin amido synthetase10). BLAST analysis identified AtJAR1, an enzyme 406 26 responsible for the last step of the biosynthesis of bioactive jasmonic acid (JA), JA-Isoleucine 407 proteins with high amino-acid sequence identity to AAS10 mostly belong to the auxin-409 responsive GH3 protein family. The other gene differentially regulated in the same direction in 410 all three BC mutants (reduced expression) is Zm00001d050455, which is not functionally 411 annotated in the maize genome but its closest relative in Arabidopsis encodes a hydroxy-412 cinnamoyl-CoA shikimate/quinate hydroxy-cinnamoyl transferase in Arabidopsis (HCT 413 (Hoffmann et al., 2004), 37% identical at the protein level) (Supplemental Table S4). 414 Interestingly, a slightly less similar gene (28% identity at the protein level, Supplemental Table  415 S4 Zm00001d050455 in bulliform-enriched tissue indeed has an influence on HCA content 423 specifically in the BC cuticle by using our previously described dataset. In the comparison of 424 wild-type adaxial vs. abaxial cutin, isolated from epidermal peels, decrease in HCAs could be 425 observed on the adaxial (bulliform-containing) side ( Figure 10A), which was solely due to 426 reduced ferulate content ( Figure 10B). Analysis of relative HCA content revealed an increase of 427 coumarate and decrease of ferulate in the adaxial cuticle ( Figure 10C). This suggests that BC 428 cuticles are indeed reduced in ferulate, supporting the hypothesis that the putative HCA 429 biosynthetic gene Zm00001d050455 promotes ferulate incorporation into the polyester and is

Maize leaf rolling is impacted by variation in bulliform strip patterning 460
The role of BCs in leaf rolling has been a matter of ongoing debate for decades, and no clear 461 conclusion has been drawn about their functional contribution to this important drought 462 response (Ellis, 1976;Moulia, 2000). Loss of turgor in bulliform cells on the adaxial leaf surface 463 has long been thought to induce rolling, with additional contribution by shrinkage of 464 subepidermal sclerenchyma and mesophyll tissue due to water loss (Redmann, 1985). But 465 rolling can also occur in leaves that lack bulliform cells (Shields, 1951), questioning the 466 establish a functional role for BCs in leaf rolling in maize by examining the relationship between 468 bulliform patterning and leaf rolling across a large collection of genetically diverse maize lines. However, these mutations usually lead to a constantly rolled leaf status rather than alterations in 486 the inducibility or speed of rolling upon drought or heat stress. In general, our data support the 487 conclusion that bulliform strip architecture and distribution across the leaf play a role in 488 regulation of leaf rolling. Although a microscopic phenotype, bulliform strip patterning could 489 represent an important agronomic trait with consequences on macroscopic phenotypes such as 490 plant architecture and drought resistance. 491

BC cuticles of the adult maize leaf are structurally and compositionally unique 493
In the adult maize leaf, pavement cells, BCs, stomatal guard, and subsidiary cells all show 494 different cuticle ultrastructure (Bourgault et al., 2020). The present study examined the 495 ultrastructure of BC cuticles in detail. BCs exhibit a roughly 4-fold thicker cuticle with a 496 prominent cuticular layer, which was not evident in pavement cell cuticles. This cuticular layer is 497 ultrastructurally defined by the presence of osmiophilic fibrils oriented perpendicular to the plane 498 of the cuticle, likely to be polysaccharides (Jeffree, 2006;Mazurek et al., 2017). This likens the 499 BC cuticle more to the classic three-layered cuticle model than the pavement cell cuticle of 500 maize leaves, which lack a well-defined cuticular layer but have two distinct layers of the cuticle 501 proper (Bourgault et al., 2020). 502 This study also investigated the composition of the BC cuticle. The thickness and 503 ultrastructure of pavement cell cuticles in adult maize leaves are indistinguishable on adaxial 504 and abaxial surfaces (Bourgault et al., 2020). Thus, differences in cuticle composition between 505 the two surfaces are likely due primarily to the presence of BCs and other cell types (e.g. hairs) 506 present only on the adaxial side. To exclude contributions of other cell type-specific cuticles, a 507 complementary analysis of bulliform-enriched tissue was undertaken by comparing three 508 different bulliform-enriched mutants (wty2, dek1-D and Xcl1) to their respective wild-type 509 siblings. Differences observed in both comparisons (bulliform mutants vs. wild-type, and adaxial 510 vs. abaxial) very likely depict a true compositional difference between BC and pavement cell 511 cuticles, and might indicate important functional components of this special cuticle type. 512 An overall increase of cutin (but not wax) load was found in bulliform-containing 513 or -enriched tissue in all the comparisons. This increase is consistent with the dramatically 514 increased thickness of the BC cuticle, an increase that is mostly due to the presence of a 515 comparisons agreed in identifying the cutin monomer 9,10-epoxy-18-hydroxyoctadecanoic acid 517 as being highly enriched in bulliform cuticles. Together, our findings identify cutin load and 518 monomer composition as the main differences between BC and pavement cell cuticles, 519 potentially changing the physical properties of the cuticle due to different degrees of cross-520 linking in the polymer scaffold (Fich et al., 2016). Analysis of the petal cuticle in Arabidopsis 521 revealed that cutin biosynthesis is required for the formation of cuticle nanoridges (Li-Beisson et 522 al., 2009;Mazurek et al., 2017), supporting the idea that nanoridges found on the BC cuticle 523 might be present due to different cutin load and/or composition, likely resulting a different 524 polyester structure, compared to pavement cells, which lack nanoridges. The functional 525 significance of 9,10-epoxy-18-hydroxyoctadecanoic acid in the BC cuticle is unclear, but it has 526 previously been implicated in freeze-resistance of cold-hardened rye (Griffith et al., 1985). application of these and related methods seem to be able to deliver cell-to-cell resolution of 546 cuticle composition. 547 Gene expression analysis of BC-enriched mutants yielded fewer candidates regulators 548 of BC differentiation than anticipated, but identified two genes that were differentially expressed 549 in the same direction the cuticle maturation zone of all three mutants compared to wild-type. 550 One of these, Zm00001d050455, is a homolog of Arabidopsis HCT, which is involved in 551

Bulliform-enriched tissue shows increased water loss upon dehydration 576
Several lines of evidence gathered in our study point to the conclusion that the bulliform cuticle 577 could be more water permeable, leading to higher water loss of bulliform cells upon dehydration: 578

1) BCs show increased volume loss upon dehydration compared to adjacent pavement cells, as 579
shown in the cryo-confocal analysis of dehydrated tissue in situ. Water lost from BCs does not 580 seem to be redistributed into the neighboring epidermal cells, as no decreasing gradient in 581 volume loss of adjacent cells respective to their position to BCs was detected, which would be 582 expected upon additional water entry to these cells from the BCs. Partial direction of water flow 583 from BCs to mesophyll tissue cannot be excluded but could not be quantified with the images at 584 hand. Indeed, Haberlandt and Drummond (1928) state that in rapidly transpiring organs the 585 epidermis loses water to photosynthetic tissue with its higher osmotic pressure. Nevertheless, at 586 least some of the water lost by BCs in this process has to cross the water barrier of the cuticle 587 and exit the tissue, since the overall weight of the leaf is decreasing over time, as seen in our 588 and many other studies using this or similar methods to evaluate cuticular conductance 589 (Kerstiens, 1996;Ristic and Jenks, 2002;Lin et al., 2019). Importantly, all our dehydration 590 stomatal transpiration, so increased volume loss of BCs upon dehydration suggests increased 592 water loss over the cuticle. 2) Bulliform-enriched tissue shows higher water loss rates in 593 detached leaf drying assays than comparable control tissue. Three different mutants with 594 elevated bulliform cell surface areas showed increased cuticular conductance compared to their 595 wild-type siblings, and wild type leaf adaxial leaf surfaces containing BCs displayed higher 596 cuticular conductance than the abaxial surfaces lacking BCs. While studies of cuticular 597 conductance have investigated this trait in a multitude of plant species (e.g. Table 1 in 598 Kerstiens, 1996), only a few compare the g c of adaxial and abaxial tissues. For example, no 599 difference in cuticular conductance between adaxial and abaxial leaf surfaces could be 600 measured for holm oak (Fernández et al., 2014) or beech (Hoad et al., 1996). While these are 601 dicot species lacking BCs on their adaxial surface, a study in rice also showed increased 602 adaxial cuticular conductance in leaves (Agarie et al., 1998), in agreement with our results of 603 higher water loss of bulliform-containing tissue. Again, also under these circumstances water 604 must be crossing the cuticle surface, since there is overall loss of water from the tissue shown 605 by weight decrease of the leaf over time. These findings lead us to speculate that, indeed, the 606 presence of bulliform cells could drive increased cuticular conductance, possibly facilitating leaf 607 rolling upon dehydration. 608 Is it likely that the much thicker BC cuticle, with major changes in cutin and less in 609 waxes, is more water permeable than other epidermal cuticle types? While some data connect a 610 thicker cuticle to a lower cuticular water loss rate in maize (Ristic and Jenks, 2002), there is a 611 long line of evidence that cuticle thickness in general cannot be taken as measure for water 612 permeance (Riederer and Schreiber, 2001; Jetter and Riederer, 2016). Cuticle composition 613 rather than thickness seems to be the determining factor for cuticular permeability, which 614 generally is attributed to wax components rather than cutin monomers (Kerstiens, 1996;615 wax or cutin composition showed higher cuticle permeability despite increased cuticle thickness 617 (Xiao et al., 2004;Kurdyukov et al., 2006;Sadler et al., 2016). The accumulation of epoxy-618 monomers in the cutin fraction of BCs might be an indication of a less-cross-linked cutin scaffold 619 in these cells, since cuticles high in these monomers with unused functional groups for cross-620 In conclusion, we demonstrate that the maize BCs show higher water loss upon 629 dehydration compared to other epidermal cells. The exact role of BCs and their specialized 630 cuticle in the leaf rolling response of maize has yet to be elucidated, but leaf rolling appears to 631 be facilitated by this thicker, yet likely more water permeable cuticle type unique to BCs. 632 Integration of biochemical, transcriptomic, ultrastructural, and functional data suggest an 633 important role for the cutin matrix in this cuticle type, including the compound 9,10-epoxy-18-634 hydroxyoctadecanoic acid. Together, our findings advance knowledge of cuticle 635 composition/structure/function relationships, and how cuticle specialization can contribute to cell 636 and organ functions.  The raw RNAseq data will be deposited at NCBI SRT with SRA accession number xxx.   Cuticular conductance, representing rates of water loss across the cuticle in the dark when 877 stomata are closed, was measured during dehydration of detached leaves (20-22 °C, 55-65 % 878 humidity). A) Cuticular conductance for three BC-enriched mutants wty2, dek1-D, and Xcl1 and 879 corresponding wild-types). CC rate is calculated as water loss (g) per hour per g dry weight 880 (DW). Values are given as means ± SD (n = 3-5 biological replicates per genotype). Statistical 881 analysis used two-tailed unpaired Student's t-test, with ***P < 0.001. B) Cuticular conductance 882 rate of adaxial and abaxial leaf surfaces (one-sided dehydration was achieved by covering one 883 side of the leaf with petroleum jelly) compared to full CC rate, calculated as water loss (g)  normalized to overall cutin monomer load (inset). Values are given as means ± SD (n = 4 906 biological replicates per surface/genotype). Statistical analysis used two-tailed unpaired 907 Student's t-test, with *P < 0.05, **P < 0.01, and ***P < 0.001. 908 Values are given as means ± SD (n = 4 biological replicates per surface). Statistical analysis 923 used two-tailed unpaired Student's t-test, with **P < 0.01 and ***P < 0.001. 924 925 926