Evaluation of the functional role of the maize Glossy2 and Glossy2-like genes in cuticular lipid deposition

Plant epidermal cells express unique molecular machinery that juxtapose the assembly of intracellular lipid components and the unique extracellular cuticular lipids that are unidirectionally secreted to plant surfaces. In maize (Zea mays L.), mutations at the glossy2 (gl2) locus affect the deposition of extracellular cuticular lipids. Sequence-based genome scanning identified a novel gl2 homolog in the maize genome, Gl2-like. Sequence homology identifies that both the Gl2-like and Gl2 genes are members of the BAHD superfamily of acyltransferases, with close sequence homology to the Arabidopsis CER2 gene. Transgenic experiments demonstrate that Gl2-like and Gl2 functionally complement the Arabidopsis cer2 mutation, with differential impacts on the cuticular lipids and the lipidome of the plant, particularly affecting the longer alkyl chain acyl lipids, particularly at the 32-carbon chain length. Site-directed mutagenesis of the putative BAHD catalytic HXXXDX-motif indicates that Gl2-like requires this catalytic capability to fully complement the cer2 function, but Gl2 can accomplish this without the need for this catalytic motif. These findings demonstrate that both Gl2 and Gl2-like overlap in their cuticular lipid function, however the two genes have evolutionary diverged to acquire non-overlapping functions. One-sentence summary Transgenesis dissection of the functional roles of the maize Glossy2 and Glossy2-Like genes in cuticular lipid deposition.

The functionality of the ZmGl2 and ZmGl2-like genes was evaluated by their transgenic 153 ability to complement the cer2 mutant of Arabidopsis. These effects are interpreted in the 154 context of the effect of removing such functionality in the homologous host, i.e., the maize gl2 155 mutant (note a mutation at the gl2-like locus is not currently available). The gl2 mutation 156 primarily affects the accumulation of the major components of the cuticular lipids on maize 157 seedling leaves, reducing the levels of C32 primary alcohol and C32 aldehyde, which are 158 associated with only a partial compensatory increase in C28 primary alcohol and aldehyde 159 (Supplemental Figure 1, Supplemental Table 1). 160 The ORFs coding for the ZmGl2 and ZmGl2-like proteins were expressed in Arabidopsis 161 under the transcriptional control of the constitutive 35S promoter in homozygous lines that 162 carried either wild-type or mutant cer2-5 alleles. Expression of the ZmGl2 and ZmGl2-like 163 transgenes was confirmed at the protein and mRNA levels by western blot analysis (Figure 2A) 164 of extracts using a GLOSSY2 antibody, and by RT-PCR analysis of RNA isolated from these 165 plants ( Figure 2B), respectively. As expected, the GL2 antibody detects a 46-kDa-polypeptide 166 band in extracts from maize silk tissues, and this protein band is also detected in extracts 167 prepared from Arabidopsis transgenic lines that are expressing the Gl2 transgene (i.e., genotype: 168 ZmGl2 in WT and ZmGl2 in cer2-5); this protein band is absent from the control, non-transgenic 169 Arabidopsis plants. ZmGl2-like expression was detected via RT-PCR analysis with RNA-170 template preparations made from Arabidopsis plants carrying the ZmGl2-like transgene (in the 171 WT and cer2-5 mutant lines), and this transcript was undetectable in the non-transgenic wild-172 type and cer2-5 mutant control plants ( Figure 2B). 173 Typical of cer mutants, the stems of the cer2-5 mutant show the eceriferum phenotype, 174 presenting bright green stems (Koornneef et al., 1989), rather than the dull green appearance of 175 the wild-type plants ( Figure 2C and 2D). This phenotype is associated with epicuticle-deficiency 176 and is indicative of changes in the cuticular surface lipid composition (Koornneef et al., 1989). 177 As with the transgenic expression of the CER2 protein (Xia et al., 1996), the transgenic 178 expression of the ZmGL2-LIKE protein in the cer2-5 mutant background fully restores the stem 179 eceriferum phenotype to the dull green, wild-type appearance ( Figure 2D). In contrast, the 180 transgenic expression of the ZmGL2 protein in the cer2-5 mutant only partially restored this 181 phenotype to the wild-type appearance ( Figure 2C). As control experiments, we transgenically 182 expressed the ZmGl2 or ZmGl2-like transgenes in wild-type Arabidopsis plants, and this did not 183 2C and 2D). 185 As is expected from prior studies (Xia et al., 1996)  transgenic expression of ZmGl2-like in the wild-type background also increased the total 205 cuticular lipid load by ~15% compared to wild-type (statistically significant by t-test, p-value 206 <0.05), whereas the expression of the ZmGl2 transgene had no such effect on the total cuticular 207 lipid load of the wild-type stems. 208 Figure 4 shows the compositional changes in the cuticular lipid profiles of these stems; 209 these data are focused on alkyl chain-lengths of 26-carbons and longer, whereas the effect on the 210 accumulation of the shorter length alkyl chains was negligible (these latter data are included in 211 Supplemental Figure 2 and 3). The cer2-5 mutation primarily affects the accumulation of the 212 major components of the cuticular lipids, which are nearly eliminated in the mutant; these 213 constituents are C30 fatty acid alkyl derivatives, specifically C29 alkane, C29 secondary alcohol, 214 and C29 ketone. These decreases in accumulation are associated with a partial compensatory 215 increase in shorter chain fatty acid derivatives, predominantly the C26 primary alcohol. The 216 transgenic expression of either Gl2, Gl2-like or the CER2 transgenes in the cer2-5 mutant 217 background results in a cuticular lipid compositional profile that is near identical to the wild-type 218 profile in terms of the major cuticular lipid components. Thus, as compared to the cer2-5 219 mutant, these transgenic lines show increased accumulation of the C29 alkane, C29 secondary 220 alcohol, C29 ketone. However, the effect on the less abundant cuticular lipids is different among 221 the three transgenes. Thus, like the CER2 transgene, the Gl2 transgene decreased the 222 accumulation of the C26 components (i.e., primary alcohols) to wild-type levels, but the Gl2-like 223 transgene was not capable of this effect, and the C26 primary alcohol remained at elevated levels 224 as in the cer2 mutant ( Figure 4A). 225 In addition to the transgenic compensatory effects on the cuticular lipid profiles, the 226 expression of the two maize transgenes in the cer2-5 mutant induced novel changes to the 227 profiles that are not normally present in wild-type controls. In particular, ZmGl2 expression in 228 cer2-5 mutant induced the formation of longer alkyl chains; namely derivatives of the C32 fatty 229 metabolites are also observed when the ZmGl2 transgene is expressed in the wild-type 231 background ( Figure 4). These latter novel components are not manifest by the transgenic 232 expression of either the CER2 or the ZmGl2-like genes in either the wild-type or the cer2-5 233 mutant ( Figure 4). 234 Collectively therefore, these biochemical changes establish that both the ZmGL2 and 235 ZmGL2-LIKE proteins are functional homologs of CER2 and fully complement the biochemical 236 deficiency associated with the cer2 mutation. However, the two maize genes are not equivalent 237 in how they affect cuticular lipid profiles; namely the ZmGl2 transgene induces additional 238 capabilities by enabling the production of even longer chain constituents than normal (up to 32-239 carbon fatty acids, and their alkyl derivatives), and the ZmGl2-like is incapable of reversing the 240 effect on the accumulation of 26-carbon atom constituents.  , 2008). This is the HXXXDX-motif, which contains the catalytic His-residue that is 249 responsible for deprotonating the alcohol or amine acyl acceptor-substrate, and is thus crucial to 250 the acyltransferase catalytic mechanism. As with other biochemically characterized BAHD 251 Interaction between potential substrates and this BAHD catalytic domain was 254 computationally explored with structural models of ZmGL2 and ZmGL2-LIKE proteins, 255 generated by Phyre2 (Kelley et al., 2015). Using these structural models, the 3DLigandSite 256 algorithm (Kelley and Sternberg, 2009) identified the Sorghum hydroxycinnamoyl transferase 257 (HCT) as the best BAHD structural template for ZmGL2 and ZmGL2-LIKE proteins. Using the 258 experimentally determined structure of the substrate-enzyme complex of HCT, we identified that 259 in addition to the potential catalytic His residue, the last "X" residue of the HXXXDX-motif of 260 ZmGL2 and ZmGL2-LIKE (i.e., Ile-174 and Ile-190, respectively) may be sufficiently close to a 261 potential substrate to directly interact. Furthermore, the Ile residue at this position is rare among 262 BAHD homologs (occurs <1% of 1085 sequences that we analyzed), and residues with 263 considerably smaller side chains are the most prevalent residues at this position (~70% are Gly 264 and ~20% are Ala residues). As a contrast, the middle 3 "X" residues are more conserved as 265 hydrophobic amino acids. 266 Therefore, site-directed mutagenesis was used to experimentally explore the functional 267 role of the H, D and final "X" residues in the HXXXDX-domain of the ZmGL2 and ZmGL2-268  Table 2). 274 complement the cer2-5 chemotype; namely either the expression of the wild-type ZmGl2 276 transgene or any of the three ZmGl2 point mutants, which should have destroyed the BAHD 277 catalytic capability, are still capable of reversing the reduction in epicuticular lipid accumulation 278 that is characteristic of cer2-5 ( Figure 5A). In contrast, the H185A, and I190A point mutants of 279 ZmGl2-like transgene cannot fully complement the cer2-5 chemotype, and reverse the cuticular 280 lipid load on these stems ( Figure 5B). These results demonstrate that the HXXXDX BAHD 281 catalytic motif is not required for the ability of the ZmGL2 protein to functionally replace the 282 CER2 function, however this motif is required for the ability of the ZmGL2-LIKE protein to 283 fully replace the CER2 function. 284 The ZmGL2 and ZmGL2-LIKE transgenes alter the VLCFA and hydroxy-VLCFA 285

components of the cutin and cellular lipidome profiles 286
Because the genetic lesion that determines extracellular lipid traits occurs in the context 287 of intracellular lipid metabolic processes, we profiled and compared the intracellular lipidomic 288 pools in the stems of the different genotypes generated in this study. Furthermore, as a potential 289 BAHD enzymes, ZmGL2 and/or ZmGL2-like maybe involved in the assembly of the ester 290 bonds, which are prevalent in the assembly of the cutin polymer. For these reasons therefore, we 291 also evaluated the effect of these genetic manipulations on the cutin monomers of the isolated 292 cutin prepared from stems. 293 As with the cuticular lipid analyses described above, these comparisons are interpreted in 294 the context of the effect of removing the gl2 functionality in maize. Thus, lipidomics analyses 295 indicate that mutation at the gl2 locus in maize affects the VLCFA pool associated with the 296 intracellular lipidome. These alterations are primarily associated with elongation of C26 and C28 297 ZmGl2-like (Fig 6B) on the accumulation of the three lipid classes, the cellular lipidome, cutin 302 monomers and extractable cuticular lipids. These data indicate that quantitatively, the major 303 effect of the cer2 mutation is in halving the total accumulation of the extractable cuticular lipids, 304 with minimal or no effect on the total accumulation of the cellular lipidome or cutin monomers. 305 Moreover, the transgenic expression of ZmGl2 or ZmGl2-like reversed the effect of the cer2 306 mutation, and returned cuticular lipid content to wild-type levels, and had no quantitative effect 307 on the total accumulation of the other lipid components that were evaluated ( Figure 6A and B). 308 However, these transgenic manipulations had compositional effects on these lipids, and 309 we focus the following text on the effect on VLCFA and derivative components. Although the 310 cutin polymer is thought not to have VLCFA components, cutin preparations often contain 311 VLCFA derivatives (i.e., 2-hydroxy-VLCFAs), which are probably associated with sphingolipids 312 that co-purify with cutin (Molina et al., 2006). Thus, regardless of their complex-lipid origins, 313 because these VLCFAs are FAE-generated products, we evaluated the effect of ZmGl2 and 314 ZmGl2-like transgenic expression on their abundance. 315 Specifically, in the cer2-5 mutant there are reductions in accumulation of the C22-and 316 C24-fatty acids and their 2-hydroxy-derivatives, and an increase in the C26 fatty acid and the 317 corresponding 2-hydroxy-derivative that was associated with the cutin preparations (i.e., 318 metabolite #163, #165, #167, #176, #178, and #181, Supplemental Figure 5A, Supplemental 319 Table 3). Transgenic expression of Gl2 in the cer2-5 mutant reversed these alterations in VLCFA 320 alterations in the total levels of cutin monomers (Fig. 6B) and of the VLCFA components 324 associated with the cutin preparations, and this was irrespective of whether Gl2-like is expressed 325 in a wild-type or cer2-5 mutant background (Supplemental Figure 5E and 5F, Supplemental 326 Table 3).  Table 3). Among these chemically defined lipids, 333 the most striking alterations are the changes in the accumulation of the free fatty acids (i.e., 334 metabolites #148 to #153, Supplemental Figure 5A-F, Supplemental Table 3). The cer2-5 335 mutation reduced the accumulation of C30 fatty acid (metabolite #150), resulting in the higher 336 accumulation of the precursor fatty acids of 26 and 28 carbon chain lengths (i.e., metabolites 337 #148 and #149, Supplemental Figure 5A, Supplemental Table 3). The transgenic expression of 338 maize Gl2 and Gl2-like in this mutant reversed these cer2-5-induced effects (Supplemental 339 Figure 5C and 5E, Supplemental Table 3). Moreover, ZmGl2 has additional capabilities, 340 inducing the increased accumulation of C32 fatty acid (i.e., metabolite #151, Supplemental 341 Figure 5C and 5D, Supplemental Table 3). This latter effect correlates with the increased levels 342 of C31 alkyl derivatives that were detected in the epicuticular lipid profiles of wild-type and 343 Another alteration in the lipidome that is consistent with the shared functionality between 347 CER2, ZmGl2 and ZmGl2-like in fatty acid elongation is the alteration in the accumulation 348 pattern of the glycosylceramides that utilize a 2-hydroxy C26 fatty acid building block (i.e., 349 metabolites #58 and #59, Supplemental Figure 5A, Supplemental Table 3). The accumulation of 350 these metabolites is doubled in the cer2-5 mutant, and this effect is reversed by the transgenic 351 expression of either ZmGl2 or ZmGl2-like (Supplemental Figure 5C and 5E, Supplemental Table  352 3). An additional similar genetic modulation, but not associated with fatty acid elongation, is the 353 accumulation of the minor sterol esters; i.e., -sitosterol esterified with linoleic or linolenic acid 354 (i.e., metabolites #106 and #107, Supplemental Figure 5A, 5C and 5E, Supplemental Table 3). 355 Compared to the wild-type, the accumulation of these two sterol esters is increased by between 356 10 and 25 fold in the cer2-5 mutant, and their levels are decreased to normal when ZmGl2 is 357 transgenically expressed in the mutant; in contrast, the transgenic expression of ZmGl2-like does 358 not have this latter effect (Supplemental Figure 5C and 5E, Supplemental Table 3). 359 These lipidomics analyses also quantified changes in the accumulation of 1,219 analytes 360 that are chemically undefined (Supplemental Table 3). On a mole basis these chemically 361 undefined lipids account for ~15% of all the lipids that were collectively profiled in the wild-362 type and in the cer2-5 mutant tissue. These data would be more informative of ZmGl2 or ZmGl2-363 like functions once their chemical identities are determined, however, we evaluated their relative 364 accumulation patterns as molecular markers to gain insights into the relative functionalities of 365 the two maize genes. In the tabular data presented in Supplemental Table 3, these analytes are 366 positive ion modes. 368 The accumulation of approximately half of these chemically undefined lipid analytes are 369 unaffected by the cer2-associated genetic manipulations. The Venn-diagram shown in Figure 6C  both transgenic lines ( Fig 6D). Therefore, although the two maize homologs can functionally 384 complement the cuticular lipid phenotype of the cer2 mutation, the two maize genes also express 385 unique attributes that are not equivalent to each other or to the CER2 gene. 386

DISCUSSION 387
The extracellular cuticular lipids that coat the outer surfaces of the aerial organs of 388 terrestrial plants (Fernandes et al., 1964;Kolattukudy 1980) are specifically produced by the 389 cellular population of these aerial organs (Jellings and Leech, 1982), elucidating the molecular and 391 biochemical mechanisms that regulate their biogenesis is confounded by the other 90% of the cell 392 population that is not involved in these processes. This technical barrier has been partially 393 overcome by utilizing forward genetic approaches to identify and characterize genes involved in 394 This strategy has been successful in the isolation of causative genes that generate the cuticular 396 lipid phenotypes (glossy in maize and eceriferum in Arabidopsis). Yet in many cases, the exact 397 mechanisms by which these gene products affect cuticular lipid deposition is still unclear. 398 Exemplary of such molecular characterization of cuticular lipid genes is the gl2 gene of 399 maize, and its Arabidopsis homolog, cer2 (Tacke et al., 1995;Xia et al., 1996). In this study we 400 specifically implemented a transgenic strategy to characterize the functional inter-relationship 401 between the maize Gl2 gene, and the homologous Gl2-like gene, and the Arabidopsis homolog, 402 CER2. We identified the maize Gl2-like gene by the shared 63% sequence similarity with Gl2. 403 Although ZmGl2 is known to be involved in cuticular lipid biosynthesis (Hayes and Brewbaker, 404 1928;Bianchi, 1975;Tacke et al., 1995), ZmGl2-like may have overlapping but distinct roles in 405 this pathway. The transgenic experiments conducted in this study utilized Arabidopsis as the 406 vehicle for these evaluations, testing for the ability of each maize gene to compensate for the 407 missing function associated with the homologous Arabidopsis CER2 gene; in parallel we also 408 evaluated the effect of each transgenic event in the wild-type background. 409 The ability of Gl2 and Gl2-like to contribute to extracellular cuticular lipid deposition 410 The functionality of ZmGl2 in cuticular lipid biosynthesis was initially identified by the 411 genetic observations that mutations at this locus eliminate the deposition of these constituents on 412 identification of ZmGl2-like is solely based on sequence homology with the ZmGL2 and CER2 414 proteins, without any functional data to support its role in cuticular lipid biosynthesis. Mutations 415 at the cer2 locus block the normal accumulation of cuticular lipids, by apparently blocking the 416 conversion of C26 and/or C28 fatty acid to C30 fatty acid (Mcnevin et al., 1991;Jenks et al., 1995).

Western blot analysis 592
Protein-extracts were prepared by homogenizing plant leaf tissue with a buffer consisting 593 of 62.5 mM Tris-HCl, pH 6.8, 30% glycerol, 10% SDS and 10% 2-mercaptoethanol. Samples were 594 vortexed for 5-min, boiled for 10-min and centrifuged at 13000g for 2-min. The clarified 595 supernatant protein extracts were subjected to SDS-PAGE and proteins were transferred to a 596 nitrocellulose membrane according to manufacturer's instructions (Bio-RAD, Hercules, CA). 597 Protein blots were first probed with GL2-specific antibody, recovered from ascites fluid recovered 598 (Thermo Scientific, Rockford, IL) and visualized on the ChemiDoc XRS+ gel documentation 602 system (Bio-Rad). 603

Scanning Electron Microscopy 608
The 1-cm-long stem segments were mounted on aluminum stubs with double-sided 609 carbon tape, dried in a desiccator and sputter-coated (www.tedpella.com) with a Cressington 610 HR208 sputter coater with platinum for 90s at 40 milliamps, depositing a 10 nm-thick coating. 611 The segments were examined at 10 kV with a Hitachi SU4800 field emission 612 SEM (www.hitachi-hightech.com), and images were digitally captured in TIFF format. held at 120°C, then ramped at 10°C/min to 260°C and held at this temperature for 10 min, and 634 then ramped at 5°C/min to 320°C and held there for 4 min. EI-MS ionization energy was set to 635 70 eV and the interface temperature was 280°C. Resulting chromatograms and mass-spectra 636 were deconvoluted and queried against an in-house Mass-Spectral library and the NIST 14 Mass 637 Spectral Library using the NIST AMDIS software (Stein, 1999). 638 GC-FID analysis was conducted with an Agilent 6890 GC, equipped with a DB-1 MS 639 capillary column (15 m x 0.25 mm x 0.25 m, Agilent 122-0112). Chromatography was 640 conducted with helium gas, at a flow-rate of 1.2 mL/min, and an inlet temperature at 280 o C. The 641 column oven temperature was initially held at 80°C, then ramped at 15°C/min to 220°C, then 642 ramped at 7.5°C/min to 310°C, and finally ramped at 20°C/min to 340°C and held for 5 min. The 643 used to chemically identify eluting peaks. 645 The relative abundance of extracellular lipids/mg dry weight of plant material was 646 calculated based on the ion-strength of the internal standard. Statistical significance was 647 determined using Student's t-test. 648     wild-type background (D). Each subset identifies the number of lipid analytes whose abundance 784 is either increased (blue digits) or decreased (red digits) in the indicated genetic background. The 785 digital data can be found in Supplemental     Transgenic expression of ZmGl2 and ZmGl2-like in the wild-type background. All numeric and statistical data, including the C20-C24 carbon chain-length minor components can be found in Supplemental Table 1. The data represent mean + standard error of 10-15 replicates  Table  2. The data represent mean + standard error of 10-40 replicates. . Each subset identifies the number of lipid analytes whose abundance is either increased (blue digits) or decreased (red digits) in the indicated genetic background. The digital data can be found in Supplemental  Table 3. The altered accumulation levels are statistically significant based on evaluation by Student's T-test (pvalue < 0.05).