An ionic liquid extraction that preserves the molecular structure of cutin shown by nuclear magnetic resonance

An ionic liquid extraction that preserves the molecular structure of cutin 4 shown by nuclear magnetic resonance 5 6 Carlos J.S. Moreira, a,± Artur Bento, a,± Joana Pais, a Johann Petit, b Rita Escórcio, a Vanessa G. 7 Correia, a Ângela Pinheiro, a , Łukasz P. Haliński, c Oleksandr O. Mykhaylyk, d Christophe 8 Rothan, b Cristina Silva Pereira a, * 9 10 a Instituto de Tecnologia Química e Biológica António Xavier, Universidade Nova de Lisboa (ITQB NOVA), Av. 11 da República, 2780-157, Oeiras, Portugal 12 b UMR 1332 BFP, INRAE, Univ. Bordeaux, F-33140 Villenave d’Ornon, France 13 c Department of Environmental Analysis, Faculty of Chemistry, University of Gdańsk, Wita Stwosza 63, 80-308 14 Gdańsk, Poland 15 d Soft Matter Analytical Laboratory, Dainton Building, Department of Chemistry, The University of Sheffield, 16 Sheffield, S3 7HF, UK 17 ± equally contributing authors; * corresponding author: Cristina Silva Pereira (spereira@itqb.unl.pt) 18 19


55
Plant polyesters, namely cutin and suberin, are the third most abundant plant polymers after 56 cellulose/hemicellulose and lignin. Naturally, due to their high abundance in nature, plant 57 polyesters are considered as promising substitutes to petroleum-based plastics (Heredia-Guerrero 58 et al., 2017). In particular, cutin makes up the polymeric matrix of the cuticle that builds the 59 protective layer of the aerial parts of land plants; an evolutionary feature acquired during the 60 colonization of terrestrial environments (Fich et al., 2016). The cuticle constituents (cutin and 61 efficiency of the reaction was higher when cholinium hexanoate was used (Fig. 1). Cholinium 123 hexanoate did not catalyse the cleavage of octyl octanoate (Fig. 1A), contrary to the BMIM 124 acetate that catalysed this reaction though inefficiently (Fig. 1B). As previously reported, 125 cholinium hexanoate catalyses specifically the hydrolysis of acylglycerol esters (Ferreira et al., 126 2014), regardless that in the present study the absence of agitation and the higher water content 127 of the ionic liquid reduced the reaction efficiency. 128 We then tested the potential of these ionic liquids for the isolation of cutin from tomato 129 peels after 2, 15 and 170 hours compared to a conventional method (i.e. enzymatic removal of 130 polysaccharides followed by organic solvent mediated dewaxing). In the process of suberin 131 extraction from cork using cholinium hexanoate, suberin in the filtrate is recovered by 132 precipitation in an excess of water (Ferreira et al., 2012). We preliminarily tested a 2 hours 133 reaction of cutin peels in cholinium hexanoate, and verified using Attenuated total reflection-134 Fourier transform infrared spectroscopy (ATR-FTIR) that the archetypal bands assigned to cutin, 135 i.e. long chain aliphatics (CH 2 and C=O), were detected in the insoluble fraction (not in the 136 filtrate as observed for cork suberin) whereas the filtrate shows enrichment in bands usually 137 assigned to polysaccharides (C-O-C) (Supplemental Fig. S1). Accordingly, the produced 138 insoluble fractions were characterised using SEM (Fig. 2) and 13 C MAS NMR (Fig. 3A). SEM 139 imaging of the cutins extracted with either ionic liquid are virtually identical: a clean thick cutin-140 continuum showing the epidermal cell grooves ( Fig. 2A-F). In the reference cutin, i.e. obtained 141 through the conventional enzymatic-based process, the cutin-continuum apparently overlaps with 142 other cellular components, and many intracellular spaces are not hollow (Fig. 2G). 143 In the 13 C MAS NMR spectrum of the reference cutin, the major structural classes 144 assigned to cutin include the long methylene chains -(CH 2 ) n with major peaks at 26, 29 and 145 34 ppm, the oxygenated aliphatics -CH 2 O (63 ppm) and CHO (73 ppm), and the carboxyl 146 groups at 172 ppm, comprising the contribution of both esters and acids (Chatterjee et al., 2016) 147 (Fig. 3A). Only minor signals can be assigned to the aromatic region (105 and 130 ppm). The 148 spectral signatures of the remaining cutins are very similar regardless of the ionic liquid used and 149 extraction time and also similar to the reference cutin spectrum ( Fig. 3 and Table 1). The relative 150 contributions of the signals assigned to aromatics for the cutins purified with either ionic liquid 151 increased along the reaction time, possibly an artefact derived from phase corrections. The 152 relative contributions of the oxygenated aliphatics region (57-92 ppm) are higher in the ionic 7 originated from the 2 hour ionic liquid reactions are less broad compared to the reference cutin, 185 suggestive of increased homogeneity. This feature was lost when extensive reaction times were 186 used, consistent with the estimated reduction in the biopolymer reticulation and esterification 187 ( Fig. 3B-D). The WAXS patterns of all cutin samples (Fig. 4B) are mainly represented by a 188 broad diffuse peak with the maximum intensity at q ~ 1.41 Å -1 which most likely corresponds to 189 an amorphous structure commonly formed by organic polymeric materials with an inter-chain 190 distance of around 4.5 Å. Considering cutin's composition, this amorphous structure should be 191 related to randomly-packed acyl chains. In contrast to the reference cutin, the scattering patterns can be assigned to an orthorhombic crystal structure (space group Pnma, Miller indexes 110 and 196 200, respectively) commonly formed by compounds comprised of alkane-like chains such as 197 triacylglycerols (b' phase) (Mykhaylyk et al., 2007) or polyethylene Southern et 198 al., 1972). This suggests that the extraction of cutin by the ionic liquids enriches this material 199 with a crystalline component where some acyl chains tend to form crystals. A low level of 200 branching of acyl chains in cutin possibly is favourable for the formation of an orthorhombic unit 201 cell, which is thermodynamically more stable than the rotator phase formed by distorted alkane 202 chains packed in a hexagonal array (Small, 1984). This observation is consistent with the DSC 203 measurements (Fig. 4A) indicating that the cutin extracted by cholinium hexanoate for 170 h, 204 containing the highest fraction of the acyl crystalline component, has the highest peak melting 205 point. The third diffraction peak at q = 2.46 Å -1 , observed for cutin extracted by cholinium 206 hexanoate (Fig. 4B, purple and blue curves), cannot be related to the acyl chain crystalline 207 structure. Its position is significantly shifted from a possible 020 peak at q = 2.55 Å -1 generated 208 by the orthorhombic structure (Fig. 4B). The third peak is likely to be associated with a 209 crystalline cellulose and can be assigned to 004 reflection of monoclinic cellulose I β (space group 210 P12 1 1) (Rongpipi et al., 2019). It has to be noted that the most intense 200 diffraction peak of the 211 cellulose I β expected at q = 1.63 Å -1 is not visible because of an overlap with the intense broad 212 peak corresponding to the amorphous structure. Crystalline cellulose usually coexists with 213 amorphous cellulose (Rongpipi et al., 2019). However, it would be difficult, if possible at all, to 214 identify the cellulose amorphous component with its expected peak maximum intensity at q = also to the starting material) ( Table 2). Both the abundance and the diversity of fatty acids 230 decreased as the reaction time in the ionic liquid increased (Table 2). This was more pronounced 231 when BMIM acetate was used for 170 hours, which rendered a cutin that is almost devoid of 232 fatty acids and also containing nearly two times less dicarboxylic acids. The fatty acids carry a 233 methyl end-group that is esterified to the biopolymer through a single bond. 234 A snap-shot of the molecular structure of cutin purified by cholinium hexanoate reveals 235 extant free hydroxyls and free acids 236 Our data made evident the potential of using short time reactions with either ionic liquid to 237 recover from tomato peels a cutin continuum displaying esterification/reticulation levels and 238 composition near to that found in planta. In addition, cholinium hexanoate presents several 239 advantages compared to the BMIM acetate. It cleaves fewer ester bonds, rendering a more 240 esterified biopolymer (Fig. 3d), and contrary to the BMIM acetate, it is also biocompatible and 241 biodegradable (Petkovic et al., 2010). 242 Recently, we resolved the molecular structure of in situ suberin using solution state 243 NMR, upon its solubilisation in heated DMSO directly from cork after four hours of cryogenic 244 milling (Correia et al., 2020). This inspired us to apply cryogenic milling for the solubilisation of 245 a cutin extracted with cholinium hexanoate after two hours. Solving cutin's molecular structure 246 9 would create conditions to look "inside" its backbone, specifically to its esterification 247 arrangement. The GC-MS analyses disclosed only the composing hydrolysable constituents 248 (Table 2) and the solid-state analyses -13 C MAS NMR, DSC and WAXS ( Fig. 3 and 4) -249 revealed only the bulky chemical functionalities and properties of the purified cutin materials. 250 Only after 10 hours of cryogenic milling the cutin was solubilised in DMSO, reflecting cutin's 251 much lower solubility compared to suberin. We analysed the impact of the cryogenic milling 252 process, especially the occurrence of oxidation reactions inside the grinding jar due to possible 253 condensation of oxygen at low temperatures. Elemental analysis of cutin before and after the 254 cryogenic milling process revealed that the relative percentage of the tested elements, including 255 oxygen (Supplemental Table S2), were unaltered after the treatment. Therefore, despite this 256 solubility drawback a solution state 1 H NMR could be acquired with good resolution showing the 257 presence of many overlapping signals (Fig. 5A); an archetypal feature observed in other complex 258 multifunctional polymers (Lyerla, 1980). The relative abundances of aliphatics, CH/CH 2 -X 259 oxygenated aliphatics and aromatics were estimated through the integration of the 1 H-spectrum 260 as 70%, 27% and 3%, respectively. The assignment of 1 H chemical shifts for the constituent 261 monomers was then achieved through a combination of 1 H-1 H (correlation spectroscopy -262 COSY) and 1 H-13 C (heteronuclear single-quantum correlation spectroscopy -HSQC, 263 heteronuclear multiple-bond correlation spectroscopy -HMBC) correlation experiments 264 (Supplemental Table S3, Supplemental Fig. S4 to S6). Previous NMR-based data of tomato cutin 265 were attained through solution state NMR analyses of oligomeric structures obtained by 266 methanolysis of tomato peels (Graça and Lamosa, 2010) and through high resolution-magic 267 angle spinning (HR-MAS) NMR analyses of the tomato cutin swelled in DMSO (Deshmukh et 268 al., 2003). These studies provided important baseline information for the assignment of the 269 spectrum of cutin extracted with cholinium hexanoate for two hours (Supplemental Table S3). 270 The full range of the HSQC spectrum of cutin is depicted in Fig. 5B, highlighting the 271 regions corresponding to aliphatics and CH/CH 2 -X aliphatics as well as aromatics. A detailed 272 analysis of the HSQC spectrum of the two aliphatic regions with the assignment of CH 2 and CH 3 273 groups from the aliphatic chains, the ester bonds and the free mid-chain hydroxyl groups is 274 shown in Fig. 5C not cleave primary esters bonds (Fig. 1). Here we assigned the β-(C=O) esters to a 1 H shift of 278 1.49 ppm and 13 C shift of 24 ppm but we could not detect the signal of β-(C=O) acids, regardless 279 that they have been assigned before in tomato cutin using HR-MAS NMR (Deshmukh et al., 280 2003).  assigned the signals of aliphatic esters, primary and secondary 281 alcohols, free acids and α-branched carboxylic acids, yet the last two assignments could not be 282 confirmed by HMBC. In the present study, the signals of the β-(C=O) acids possibly overlap 283 with that of the esters and their differentiation from the small chemical shift differences observed 284 in the acquired HSQC is virtually impossible. The α-(C=O) signal displays two 1 H signals with a 285 13 C shift of 33 ppm, namely at 2.25 ppm and 2.19 ppm, which can be assigned to esters and 286 acids, respectively. The α-(C=O) signal with a 1 H shift of 2.17 ppm has been previously assigned 287 to xylan esters (Zhang et al., 2016). Based on the detection of vestigial amounts of 288 microcrystalline cellulose in the cutin extracted with cholinium hexanoate for 2 hours (Fig. 4B), 289 this signal may be associated to the presence of cellulose esters. Analysis of the cutin extracted 290 with BMIM acetate (also cryogenically milled), which is apparently devoid of microcrystalline 291 cellulose (Fig. 4B), showed that the α-(C=O) signal displays a 13 C shift of 33 ppm and only a 1 H 292 shift of 2.26 ppm (Supplemental Fig. S7). Finally, to precisely assign the free acids in the cutin 293 extracted with cholinium hexanoate for 2 hours, we acquired the HMBC spectrum that confirmed 294 their signal at a 13 C shift of 35 ppm and a 1 H shift of 2.02 ppm (Supplemental Fig. S6). This 295 observation is consistent with that previously assigned in cork suberin where the signal of the 296 acid is at a 13 C shift of 36 ppm and a 1 H shift of 2.03 ppm, and that of the esters displays a 13 C 297 shift of 34 ppm and a 1 H broad shift from 2.33-2.27 ppm (Correia et al., 2020). 298 Based on the assignments defined above, we calculated through integration of the signals 299 in the 1 H NMR the relative abundance of free acids, of total esters (comprising primary and 300 secondary aliphatic esters yet excluding sugar esters) and of linear esters as 38%, 52% and 10%, 301 respectively ( Fig. 5A, see text-insert). No acylglycerol bonds were detected in the HSQC 302 analyses of cutin (Fig. 5B), consistent with the very low abundance of glycerol in tomato cutin 303 (Fich et al., 2016). We hypothesise that the free acids detected in the cutin spectra might mostly 304 account for their natural occurrence, though one cannot exclude, at this stage, that some aliphatic 305 esters might have undergone cleavage in the presence of cholinium hexanoate. 306 in the greenhouse (season, light, temperature and hygrometry). In addition, the cutins which 328 originated from the mutants show an increase in the relative abundance of non-hydrolysable 329 constituents compared to the wild type (ca. 10% increase), and their identification yields 330 decreased nearly 20% due to higher diversity of unidentified monomers (Table 3). Cutin from 331 both mutants display higher relative abundance of fatty acids and dicarboxylic acids (nearly 332 tenfold and twofold, respectively) and lower relative abundance of ɷ-hydroxyacids (five to three 333 times) compared to the wild-type cutin. 334 To confirm that free acids naturally occur in cutin (hence differentiating these from free 335 acid groups formed during the ionic liquid extraction), we compared the spectra of cutin from the 336 wild-type cultivar obtained through the ionic liquid process followed by cryogenic milling with 337 that of the cuticle solubilised solely via cryogenic milling (Fig. 6, Supplemental Fig. S8 to S12). 338 The obtained 1 H NMR ( Fig. 6A-B) and HSQC spectra ( that the presence of non-cutin constituents in the cuticle contributes to the appearance of many 340 signals that have yet to be assigned, e.g. in the CH 3 region (Fig. 6D). Importantly, the signals 341 previously assigned to free acids α-(C=O) acids  are visible in both samples ( Fig. 6C-D), 342 which were confirmed in the corresponding HMBC spectra (Supplemental Fig. S10). 343 Accordingly, the free acids detected in the ionic liquid purified cutins ( contrary to that observed for the cutin derived from the peels of processing tomatoes (Fig. 5D). 348 One possibility is that the cleavage of primary esters is greatly influenced by the native 349 arrangement of the polymer. 350 The impact of the mutations is seen by the relative abundances of aliphatics, CH/CH 2 -X 351 oxygenated aliphatics, and aromatics, in the 1 H-spectra which were estimated as 71%, 29% and 352 0% for the wild-type (Fig. 6A), as 46%, 50% and 4% for the gpat6 mutant (Fig. 7A), and as 353 39%, 59% and 2% for the cus1 mutant ( Fig. 7B), respectively. Contrary to the wild type, in both 354 mutants the signal assigned to free acids could not be detected ( Fig. 7) (Fig. 6A, see text-inserts). 355 To confirm this observation, we compared the spectrum of the cutin from the cus1 mutant 356 purified by the ionic liquid followed by cryogenic milling with that of the cus1 cuticle 357 solubilised solely via cryogenic milling (Supplemental Fig. S13). The obtained HSQC spectra 358 confirmed the absence of free acids in this mutant, furthering that the observed absence of this 359 chemical group in the cus1 and gpat6 cutins is a consequence of the mutations and not of the 360 sample processing (further details in Supplemental Fig. S8 to S19). 361 Based on the 1 H-spectral information, we also estimated the relative abundance of 362 aliphatic esters (total) and primary aliphatic esters in the cutin of the mutants (Fig. 7, see text-363 inserts). Accordingly, the ratio of total esters versus linear esters is comparable in the wild type 364 and the gpat6 mutant but substantially lower in the cus1 mutant. In other words, compared to the 365 wild type, the gpat6 mutant shows similar amounts of linear esters and of secondary esters, 366 contrary to the cus1 mutant that shows more than a twofold increase in linear esters but also the 367 lowest esterification level (i.e. amount of secondary esters) ( Fig. 7A-B, see text-insert). 368 The magnification of the HSQC regions corresponding to aliphatics and CH/CH2-X 369 aliphatics for the cutins of the mutants is also shown ( 13 both mutants, the signals assigned to terminal hydroxyls are visible ( Fig. 7E-F), similar to that 371 observed in the wild-type cutin (Fig. 6E). The detected CH 3 groups are apparently enriched in 372 both mutants compared to the wild type, consistent with the observed increase in the relative 373 abundance of hydrolysable fatty acids for the mutants (Table 3) and with that reported before 374 (Petit et al., 2016;Philippe et al., 2016). No acylglycerol was visible in the HSQC spectra of the 375 cutin from the mutants, possibly as their abundance are below the detection limits of the 376 analytical technique. The mutants show more non-assigned signals compared to the wild-type 377 cutin, consistent with the observed lower identification yields in the GC-MS data (Table 3) Mazurek et al., 2017), resistance 397 to pathogens  and fruit quality (Petit et al., 2017). 398

Advantages of ionic liquid extraction with respect to conventional cutin extraction methods 399
Our ionic liquid cutin extraction method performed on tomato peels demonstrates that 400 subcuticular polysaccharides (which were found in the filtrate) are removed similar to that of the 401 https://plantphysiol.org Downloaded on May 2, 2021. -Published by Copyright (c) 2020 American Society of Plant Biologists. All rights reserved. enzyme treatment but in a considerably shorter period of time (i.e. 2 h instead of days, and even 402 weeks) . Also, the ionic liquid extraction does not require any specific 403 dewaxing step. When extracted with either ionic liquid, a cutin-continuum is isolated, 404 strengthening that the ionic liquid does not substantially impact the cutin polyester. This is 405 contrasting to that previously reported by us for suberin extraction from cork using cholinium 406 hexanoate where nanoparticles of the biopolymer are isolated (Correia et al., 2020)

Cutin Extractions 497
Enzymatic Process. Cutin was isolated from tomato as previously described (Chatterjee et al., 498 2012). In brief, tomato peels were immersed in an enzymatic cocktail containing 4 ml of 499 pectinase, 0.2 g of cellulase, 13 mg NaN 3 and 196 ml of 50 mM sodium acetate buffer, and 500 incubated at 31 °C for 24 hours with constant shaking. The isolated cuticles were successively 501 dewaxed for 36 hours by Soxhlet extraction with methanol, chloroform and hexane (1:1:1), 502 finally freeze dried and stored at room temperature. This cutin enriched material was used as a 503 reference material for the optimisation phase of the present study, i.e. selection of a suitable ionic 504 liquid process. 505 Ionic liquid Process. Cutin was extracted from tomato peels as previously described for the 506 extraction of suberin from cork (Ferreira et al., 2014), with slight modifications. In brief, 2 g of 507 tomato peel powder were mixed in 20 g cholinium hexanoate or BMIM acetate and incubated for 508 a defined period of time (100 °C, without stirring). The reaction was stopped by the addition of 509 160 mL of DMSO. The polymer was recovered by filtration using a nylon membrane filter (0.45 510 µm); then washed with an excess of deionized water with the aid of centrifugation (Eppendorf 511 5804 R centrifuge, 5000 rpm at 4 °C for 30 minutes). 512 were selected to systematically process cutin samples prior to their 2D NMR analysis. 529

Treatment of cutin extracted using cholinium hexanoate (2h) with Trifluoroacetic Acid
Nuclear Magnetic Resonance (NMR) analyses 530 13 C magic angle spinning nuclear magnetic resonance ( 13 C MAS NMR) spectra were acquired on 531 the cutin enriched materials (± 250 mg) which were packed into 7 mm o.d. zirconia rotors (after 532 grinding if needed), equipped with Kel-F caps. 13 C MAS with High-Power CW Decoupling 533 spectra were obtained at 75.49 MHz, on a Tecmag Redstone/Bruker 300WB, with spinning rates 534 of 3.1-3.3 kHz. In these experiences 90˚ radio frequency (RF) pulses of around 4.5 µs and 535 relaxation delays of 3 s were used. 13 C chemical shifts were referenced with respect to external 536 glycine ( 13 CO observed at 176.03 ppm). 537 Solution state NMR spectra of the cutin samples solubilised with the aid of cryogenic milling 538 (see above) were recorded using an Avance II + 800 MHz (Bruker Biospin, Rheinstetten, 539 Germany) spectrometer, with the exception of 1 H-13 C HMBC spectra that were acquired using an 540 Avance III 800 CRYO (Bruker Biospin, Rheinstetten, Germany). All NMR spectra ( 1 H, 1 H-1 H 541 COSY, 1 H-13 C HSQC) were acquired in DMSO-d 6 using 5 mm diameter NMR tubes, at 60 °C as 542 follows: 3 mg of cryomilled cutin in 400 μL of DMSO-d 6 . 543 MestReNova, Version 11.04-18998 (Mestrelab Research, S.L.) was used to process the raw data 544 acquired in the Bruker spectrometers. 545

Differential scanning calorimetry (DSC) 546
Calorimetric analyses of the cutin enriched materials were carried out in a TA Instruments Q200 547 calorimeter connected to a cooling system and calibrated with different standards (indium, empty 548 cap). The sample weights ranged from 9 to 11 mg. A temperature interval from -80 °C to 220 °C 549 has been studied and the heating/cooling rate was 10 °C ·min -1 . 550

dimensional (2D) pixel WAXS detector (Dectris, Switzerland). Loose cutin powder samples 556
were enclosed between two flat kapton films and mounted on the beamline sample stage (the 557 total sample thickness is about 1 mm). 2D WAXS patterns were recorded in a transmission mode 558 over a q range of 1.3 Å -1 to 3.5 Å -1 [where q = (4πsinθ)/λ = 2π/d is the length of the scattering 559 vector, θ is one-half of the scattering angle and d is spacing in real space] using an exposure time 560 of 1200 seconds. The WAXS data were reduced (calibrated, integrated and background-561 subtracted) using the Foxtrot software package supplied with the instrument. 562