Building Blocks of the Protective Suberin Plant Polymer Self-Assemble into Lamellar Structures with Antibacterial Potential.

The protection of terrestrial plants from desiccation, mechanical injury, and pathogenic invasion is achieved by waxes and cutin polyesters on leaf and fruit surfaces as well as suberin polymers that are embedded in the cell walls of roots, but the physicochemical principles governing the organization of these biological composites remain incompletely understood. Despite the well-established enzymatic mediation of suberin formation in the skins of potato tubers, cork oak trees, and internal plant tissues, the additional possibility of self-assembly in this system was suggested by our serendipitous finding that solvent extracts from potato phellem tissues form suspended fibers and needles in the absence of such catalysts over a period of several weeks. In the current study, we investigated self-assembly for three-component model chemical mixtures comprised of a hydroxyfatty acid, glycerol, and either of two hydroxycinnamic acids that together typify the building blocks of potato suberin biopolymers. We demonstrate that these mixtures spontaneously form lamellar structures that are reminiscent of suberized plant tissues, incorporate all constituents into self-assemblies, can form covalently bound ester structures, and display antibacterial activity. These findings provide new perspectives on the self-association and reactivity of these classes of organic compounds, insights into agriculturally important suberin formation in food crops, and a starting point for engineering sustainable materials with antimicrobial capabilities.

Introduction

Self-assembly of nanoscale molecular units into ordered structures is a widely observed phenomenon in living organisms. From familiar cellular phospholipid multibilayers to self-aggregating peptides, polyamines, and porphyrins, this natural capability has also inspired the creation of novel engineered materials for biotechnology and biomedicine. For instance, collagen has been found to self-assemble into helical structures;[1] polyamines such as spermidine and putrescine can self-aggregate under simulated physiological conditions,[2] similarly to their in vivo arrangement around DNA molecules. Many types of self-assembly phenomena, based on hydrogen bonding or other weak interactions and which need not entail covalent bonding, are known in biological systems, e.g., the tetrameric heme units of hemoglobin and the base pairs of the DNA double helix. Among the diverse self-assembled designed materials under development are urea-based supramolecular hydrogels for gel electrophoresis,[3] oligosaccharide-based block copolymers for charge storage in memory devices,[4] and DNA-inspired soft materials for the regulation of enzymatic catalysis.[5]

For the long-chain amphiphilic ω-hydroxyfatty acids (ω-hydroxyFA) of interest in studies of protective leaf and fruit cuticles in plants, self-assembly to form micelles or lamellar vesicles has been reported: for mixtures of structurally related compounds, in conjunction with agents that prevent crystallization, or on mica templates.[6−8] Such association phenomena have also gained credence from molecular dynamics simulations.[8,9] In addition, self-aggregation of free fatty acids, amylose, and a whey protein has been reported to produce a nanocomplex that could carry lipophilic molecules for functional food or drug delivery applications, where the fatty acid acts as a bridge between the other two incompatible constituents,[10] and a glycerol-fatty acid ester was found to self-assemble into bilayers in water.[11] As these thermodynamically controlled assemblies slowly grow in size, they build up more intermolecular interactions (typically van der Waals, hydrogen bond, and hydrophobic effects); they can then gain enthalpic stability and reach equilibrium in a fully assembled state.[12]

The ω-hydroxyFA are also major constituents of plant suberin polymers, which are embedded in the cell walls of, e.g., cork oak trees and potato tubers during either normal plant development or as a wound-healing response that mitigates desiccation and pathogenic invasion.[13−16] Indeed, suberin isolated from cork with an ionic liquid, which retains a portion of its linear and acylglycerol ester structure, displays bactericidal properties.[17,18] Analyses of extracts from both native and wound potato phellem tissues implicate these natural materials as potentially useful antioxidants and antimicrobials.[19−21] Although distinctive potato suberin lamellar structures observed by transmission electron microscopy (TEM) have been attributed to polyphenolic (phenylpropanoid) and polyaliphatic (ω-hydroxyFA) domains[13,22,23] linked by glycerol bridges,[16] the macromolecular architecture of these assemblies remains incompletely determined.[15] The intriguing possibility of self-assembly in this system was suggested by our serendipitous observation that soluble potato periderm extracts formed visible fibers and needles reproducibly without any added catalysts during a span of several weeks.

In the current work, we designed three-component model chemical mixtures with several purposes in mind: to extend our fundamental understanding of self-assembly by hydroxyfatty acids, phenylpropanoids, and glycerol; to gain insight into suberin formation for terrestrial plant protection on the outer surfaces of potato tubers; and to develop engineering strategies for sustainable materials that could range from packaging to floor covering.[24,25] We mixed ferulic or sinapic acids to represent the aromatic domain, 16-hydroxyhexadecanoic (juniperic) acid for the aliphatic domain, and glycerol for a probable bridge between them as proposed by previous researchers.[16,26] The suspended solid material formed after 3–21 day incubation periods was analyzed using both imaging and spectroscopic techniques to measure their surface and bulk properties, respectively, probing the macromolecular structural organization from micron to nanometer length scales. Parallel assays of antibacterial activity were used to evaluate the functional potential of these self-assembly systems.

Results and Discussion

TEM Reveals Lamellar Microstructures Formed by Suberin-Inspired Assemblies

Lamellar structural arrangements are considered to be a hallmark of suberized plant systems,[27] particularly in potato periderm tissues.[28,29] These structural features were replicated in self-assemblies made from an ω-hydroxyfatty acid, glycerol, and either of the two phenolic (hydroxycinnamic) acids during a 21 day incubation period (Figure ). Both the self-assembled solid suspensions that formed spontaneously (Figures and S1) and the potato tuber-derived surface tissues that develop via enzyme-catalyzed biosynthesis (Figure S1) display repeating lamellar organizational patterns. Whereas potato periderms exhibit lamellar spacings of 55 ± 4.5 and 50 ± 3.7 nm for native and wound-healing tissues, respectively, the FerGlyJun and SinGlyJun self-assemblies display spacings an order of magnitude smaller: 4.4 ± 0.3 and 2.6 ± 0.1 nm, respectively. These disparities in spacing are not unexpected given the simplicity of our three-component model system, but our findings nonetheless offer clear evidence for the propensity of major suberin constituents to form organized structural arrangements analogous to the protective coverings of potato tubers,[14,16,29]Zea mays roots,[27] and Arabidopsis roots.[30] Given that fatty acids such as juniperic acid are known to self-assemble into bilayer structures,[6−8] it is notable that the FerGlyJun and SinGlyJun spacings correspond numerically to extended bilayers or monolayers, respectively. The 5–8% spread of observed lamellar spacings is roughly comparable for the native potato periderms and self-assembly materials (Figures S1 and S2), suggesting a similar degree of architectural regulation in the two types of systems. Comparing the morphology of our two model assemblies, SinGlyJun exhibits extended striated lamellar regions that appear to stack in layers or sheets, whereas FerGlyJun has “patches” of separately ordered domains containing “snakelike” structures, each with narrowly spaced lamellar arrangements (Figure ). This comparison suggests that a greater proportion of the original Sin, Gly, and Jun constituents may be tied up in a well-organized self-assembled state. Independent experiments with shorter incubation times (3, 7, and 14 days; Figure S2) also show lamellar regions that grow progressively in size with time but have spacings of <1 nm.

Figure 1

Typical preparation scheme for suberin-inspired self-assemblies from glycerol and a phenolic acid (ferulic or sinapic acid in 60% v/v methanol:water) and a hydroxyfatty acid (juniperic acid in acetonitrile).

Figure 2

TEM images of FerGlyJun and SinGlyJun assemblies that were incubated for 21 days. FerGlyJun images illustrate lamellar domains (top left), individual lamellar structures (bottom left), and illustrative spacing measurements of 2.41, 2.65, and 2.54 nm (top right); the SinGlyJun image illustrates extended layered sheets (bottom right).

Solid-State NMR Supports Self-Assemblies with Retained Molecular Structures of the Three Constituents

To assess possible structural alterations of ω-hydroxyfatty acid, glycerol, and phenolic acid compounds in our model mixture upon incubation to form ordered threadlike suspended solids, we evaluated the chemical composition of dried materials using solid-state 13C NMR. In principle, the observed spectral features could be attributed to self-assemblies and/or covalently bound reaction products. Quantitatively reliable DPMAS 13C spectra (Figure ) displayed resonances from each of the three constituents that were combined in the model mixture: long-chain aliphatics (15–45 ppm, from juniperic acid), CH3O (45–60 ppm, from sinapic or ferulic acid), CH2O (60–70 ppm, from glycerol and juniperic acid), CHO (70–92 ppm, from glycerol), arenes (92–160 ppm, from sinapic or ferulic acid), and COX (160–185 ppm, from sinapic, ferulic, or juniperic acids). These are the same chemical moieties evidenced in the 13C NMR spectra of lamellae-forming suberized cell walls of potato periderm tissues.[20,29,31] The resonances are broader than in typical crystalline solids but are generally well defined, as expected in light of the locally ordered structural arrangements found by TEM methods. The chemical shifts observed for the assemblies, which were within ∼1 ppm of reported values for the starting materials (not shown), argue against significant conversion to esters or other covalently bound structures that would produce new chemical bonding patterns. That said, the breadth of the spectral features could make it challenging to discern the modest fractions of carboxylic acid esters with slightly shifted resonances of their COX groups.

Figure 3

Comparative 150 MHz natural abundance solid-state direct polarization magic-angle spinning (DPMAS) 13C NMR spectra of three-component self-assemblies incubated for 21 days, showing color-coded resonance assignments for the major functional groups and hypothetical structures shown to illustrate possible covalent connections between the constituents that can be verified by liquid chromatography–mass spectrometry (LC–MS) and gas chromatography–mass spectrometry (GC–MS). A: SinGlyJun and relative numbers of each carbon type (bottom); B: FerGlyJun spectrum (top) and relative numbers of each carbon type (bottom).

The relative numbers of carbons derived from integrated peak area ratios deviate from expectations based on the original 1:1:1 molar ratios of hydroxyfatty acid, glycerol, and sinapic or ferulic acid starting materials, which were chosen to allow ample opportunity for intercomponent interactions. Those ratios would correspond to 12 long-chain methylene carbons per three alkoxy (CHO and CH2O) groups, roughly akin to native suberized potato cell walls.[29] The long-chain methylene groups are under-represented for SinGlyJun (Figure A); our observation of films stuck to the sides of incubation vials suggests that a portion of the juniperic acid constituent may not be completely recovered. For FerGlyJun (Figure B), the long-chain methylene groups are better represented: (CH2): (CH2O + CHO): arene ratios are 1.5:1:1.2, indicating that this latter lyophilized solid incorporates the three components in more comparable proportions. Together, these observations underscore how modest structural differences between ferulic and sinapic acids can lead to divergent self-assembly preferences that produce the contrasting lamellar spacings and morphologies displayed in TEM micrographs.

Evidence for the incorporation of glycerol into a macromolecular assembly was also obtained from the 13C NMR spin relaxation behavior of this constituent. Figure A compares the spectra of the alkoxy region for neat natural abundance glycerol (red), 10% [U-13C3]-enriched glycerol and 10% [1,2,3-13C3]-ferulic acid within the FerGlyJun self-assembly (blue), and natural abundance glycerol within the FerGlyJun self-assembly (green). After corrections for line broadening introduced during signal conditioning, disparities in the values of full width at half-height for the glycerol CH2O and CHO groups (G1 and G2) are striking: 15 and 12 Hz for neat glycerol, 170 and 140 Hz for unlabeled assemblies, and 500 and 470 Hz for partially labeled assemblies. These latter values reflect 10- to 40-fold shorter values of the apparent spin–spin relaxation times (T2*), supporting slower overall molecular tumbling of glycerol associated with assembly formation in the three-component complex.

Figure 4

Comparative 150 MHz 13C NMR spectra (A) and spin–spin (T2) relaxation times (B–D) for glycerol in three samples: natural abundance glycerol starting material (red); glycerol (13C-labeled) after incubation with 10% [1,2,3-13C3]-ferulic acid) and juniperic acid to form FerGlyJun self-assemblies (blue); and natural abundance FerGlyJun self-assemblies (green). Panel (A) was processed with VNMRJ software, and panels (B–D) were analyzed with EXCEL. Data shown in panels (B–D) follow the color scheme above, with CH and CH2 carbons denoted by squares and circles, respectively. Acquisition parameters are detailed in the Experimental Section.

To make a more rigorous comparison of molecular motions in the starting materials and the self-assemblies, we first measured the spin-lattice relaxation times (T1’s) in natural abundance glycerol as a neat liquid and in the self-assembly (Figure ), finding roughly 4-fold longer values for the solid self-assemblies that indicated diminished local segmental motions. We then determined the respective T2’s (Figure B–D), which were 50-fold shorter for both glycerol carbons in the partially 13C-enriched self-assembly samples (Figure C) as compared with neat glycerol (Figure B) and thus confirmatory of slower overall motions upon formation of the glycerol-containing macromolecular aggregate. Whereas 13C–13C and 13C–1H couplings could contribute to the spin–spin relaxation of Figure C, a 20-fold shortening is retained for the natural abundance mixture (Figure D), confirming the impact of assembly formation on molecular tumbling.

For self-assemblies in which ferulic acid and glycerol constituents were enriched with 13C stable isotopes, it was possible to verify the preservation of the original structural architectures more definitively in the three-component FerGlyJun assembly by examining pairwise 13C–13C proximal interactions in 2D dipolar-assisted rotational resonance (DARR) spectra (Figure ). We note first that the 10% 13C-labeled carbons in ferulic acid display NMR signals at 112, 145, and 173 ppm corresponding to multiply bonded and carboxylic acid moieties, respectively, but because each aromatic carbon is present at 1% natural abundance, their possible proximal 13C–13C pairs will not be observable in these experiments. For glycerol, the 13C enrichment includes both CH2O (62 ppm) and CHO (72 ppm) groups, so 13C–13C pairs can be detected. Finally, 13C-enriched juniperic acid was not available commercially; 30 ppm fatty acid chain signals are diminished in intensity and, like the ferulic acid aromatics, exhibit no carbon–carbon correlations.

Figure 5

Two-dimensional (2D) contour plot of a 13C–13C dipolar-assisted rotational resonance (DARR) solid-state NMR experiment conducted at 150 MHz and using a 300 ms mixing time for a FerGlyJun self-assembly with the hypothetical structure shown. 13C-enriched sites and through-space correlations on ferulic acid and glycerol constituents are marked with asterisks and curved arrows corresponding to the observed through-space correlations. In addition to the structural fragments verified directly from these 2D experiments, for concreteness, we show a hypothetical molecular structure that joins the three constituents via ester linkages.

In the 2D 13C–13C DARR spectra of Figure , strong dipole–dipole interactions are evident between CHO and CH2O carbons resonating at 72 and 62 ppm (blue), respectively, indicating that the glycerol moiety is preserved in the self-assembled mixture. Additionally, a portion of the ferulic acid moiety is verified by several cross-peaks that are indicative of through-space connections: 173 × 112 (red shadowed, COO x C=C–C); 173 ×145 (red, COO x ϕ-C=C); and 145 ×112 (green, ϕ-C=C x C=C–C). The DARR results are consistent with either associated monomers that retain their molecular identities or covalently linked structures such as the hypothetical diester shown in Figure ; these solid-state NMR data could be consistent with either architectural arrangement. Supporting evidence for the presence of covalently bound structures is drawn from the MS results presented below.

LC–MS Reveals Chemical Bonds among the Three Self-Assembled Constituents

The LC–MS data presented in Table provide clear evidence for compounds with ester linkages between both pairs and triples of ferulic acid, glycerol, and juniperic acids in the FerGlyJun solid sample. In addition to ferulic and juniperic acids with protonated adducts at m/z 195.0657 (retention time (RT) 9.2 min) and m/z 273.2430 (RT 17.8 min), respectively, the MS data show an ester of ferulic acid with juniperic acid ([M + H]+, m/z 449.2898 for FerJun) and esters of glycerol with each of the hydroxyfatty- and phenolic acids ([M+], m/z 268.0941 for FerGly; [M+], m/z 444.1415 and [M + H]+, m/z 445.1493 for GlyFer2; [M + H]+, m/z 347.2790 and [M + Na]+, m/z 369.2610 for GlyJun). Both a molecular ion ([M+], m/z 522.3245) and a protonated adduct ([M + H]+, m/z 523.3266) (Figure S3) are consistent with the presence of the three-component diester FerJunGly such as that shown in Figure . The observed diester molecular ion could also correspond to GlyJunFer or FerGlyJun, that is, each possible bonded arrangement is consistent with the esters described above. Based on these results, we infer that all three components can form ester linkages, in pairs and/or among all three constituents.

Table 1

Liquid Chromatography–Mass Spectrometry of a FerGlyJun Self-Assembly

retention time (RT, min)	m/z, observed molecular ion [M + H]+a	m/z, predicted	difference, ppm	identified compound	m/z, observed adducts and fragmentsb
9.2	195.0657	195.0652	2.6	Ferulic acidc
12.3	445.14931	445.1495	–0.42682	GlyFer2	444.1415 [M+]
16.4	347.2792	347.2793	–20.099	GlyJun	369.2612 [M + Na]+
17.8	273.2430	273.2424	2.2	Juniperic acidc
19.1	449.2898	449.2898	1.71382	FerJun
20.7		268.0828	42.1512	FerGly	268.0941 [M+]
21.3	523.3266	523.3278	–2.3886	FerJunGly or GlyJunFer or FerGlyJun	522.3245 [M+]

Extracted ion chromatogram (EIC).

Fragments and adducts observed at the same RT.

Checked by LC–MS runs of authentic standard compounds.

Largely similar trends of covalent bond formation are seen in the LC–MS data of SinGlyJun (Table S1). The spectroscopic results again implicate the esterification of sinapic acid and glycerol ([M + H]+, m/z 299.1125 for SinGly); esters of glycerol with juniperic acid confirming the results shown in Table ([M + H]+, m/z 347.2792 and [M + Na]+, m/z 369.2619 for GlyJun; [M + H]+, m/z 601.5038 and [M+] m/z 600.4965 for GlyJun2); as well as an ester of sinapic and juniperic acids ([M + H]+, m/z 479.3003 for SinJun).

Both sets of LC–MS results support the presence of several types of ester linkages, illustrating the versatility with which the three components of this model system can esterify spontaneously as well as self-assemble. Thus, starting from suberinlike monomeric building blocks including glycerol, ω-hydroxyfatty acids, and a phenolic acid, we observe several esters to glycerol (FerGly and JunGly; SinJun and SinGly) as well as diesters that incorporate three building blocks (FerGlyJun, GlyFer2, and GlyJun2).

GC–MS Reveals the Percentage of Chemically Bonded Species in Three-Component Self-Assemblies

Because all three original constituents of the self-assemblies could be detected by GC–MS and compared with authentic materials, this spectroscopic technique allowed us to make quantitative estimates of the prevalence of ester-linked compounds involving pairs of ferulic acid, glycerol, and juniperic acids in the FerGlyJun and SinGlyJun self-assembly samples. In contrast to LC–MS, the chemical functionalization used for GC–MS enables ionization of compounds containing hydroxyl groups: glycerol ([M + 3TMS]+, m/z 308), ferulic acid ([M + 2 TMS]+, m/z 338), and juniperic acid (m/z 385 and m/z 401, matched to an authentic compound). Along with established MS libraries, the retention times and MS fragments observed for authentic standards were used to confirm the identifications of the individual Fer, Sin, Gly, and Jun constituents in each of the FerGlyJun and SinGlyJun self-assembly solutions (Table S2).

Once each “unbound” constituent was identified, its integrated MS peak area was compared with the total peak areas from ions derived from both bound and unbound forms. This procedure yielded estimates of the relative amounts of covalently bound vs total constituents in each self-assembly sample, assuming comparable ionization efficiencies. Dimers (or dimer fragments) of each pair were identified by their molecular ions: e.g., m/z 489 for GlyJun + 2 TMS and 503 for FerJun + 2 TMS, respectively. Analogous identifications of m/z 489 (GlyJun + 2 TMS) and m/z 501 (SinJun + Na+) were made for the SinGlyJun self-assembly. Thus, in addition to using LC–MS to confirm that all three components can form ester linkages with each other, we were able to estimate that 2.1 ± 1.2 and 11.0 ± 0.1% of the FerGlyJun and SinGlyJun assemblies, respectively, are covalently bound species. The apparently more robust ability of sinapic acid-containing assemblies to form esters should be viewed with caution in light of limited GC–MS mass accuracy and the absence of juniperic acid in the NIST library. Nonetheless, the ability of these three compounds to self-assemble into well-defined lamellae even with a limited prevalence of covalent linkages is notable.

Antibacterial Assays Illustrate the Potential of Self-Assemblies against Pectobacterium carotovorum Infection

At concentrations of 10 μg/mL, each of the constituents—ferulic acid, juniperic acid, glycerol, and sinapic acid—as well as their FerGlyJun and SinGlyJun self-assemblies—displayed inhibitory activity against P. carotovorum (Figure ). Among the individual constituents, Sin displayed the most modest inhibition (19.8 ± 3.3%); the other three constituents, namely, Fer, Jun, and Gly, displayed significantly higher (p < 0.05) values in the range of 27–30% that were not statistically different from one another. For the two FerGlyJun and SinGlyJun self-assemblies, antibacterial activities were statistically indistinguishable, though SinGlyJun displayed a higher inhibition value of 27.6 ± 4.0%.

Figure 6

Percentage growth inhibition for 10 μg/mL of the chemical constituents, the assembly solutions, and an ampicillin positive control against P. carotovorum bacteria after 4 h of incubation. Error bars indicate standard error values; asterisks indicate statistical significance with respect to the ampicillin positive control, determined by Tukey pairwise analysis based on five replicates of each experiment. Tukey analysis was also used to compare each pair of materials shown above.

If we compare inhibitory activities against P. carotovorum bacteria for the individual glycerol, juniperic, and ferulic acid constituents with their corresponding FerGlyJun assembly, again at a common concentration of 10 μg/mL, we observe no statistically significant differences (p < 0.05). However, the activity trend is different for the SinGlyJun assembly, which exhibits 27.6 ± 4.0% inhibition. Sinapic acid has a statistically significant lower percentage inhibition (19.9 ± 3.3%) than SinGlyJun and the other compounds comprising the assembly, namely, glycerol and juniperic acid. Together, these results indicate that, whereas assembly formation in FerGlyJun and SinGlyJun does not reveal potentiation or addition interactions among the constituents, neither is there antagonism observed among them.[32] Importantly, the measured growth inhibition of the chemical constituents and their self-assembly solutions greatly exceed that observed for ampicillin, which was used as our positive antibacterial control. Thus, these findings suggest that related SinGlyJun and FerGlyJun assemblies could be potential candidates for the prevention of P. carotovorum infections in potatoes or other plant species.

A rather different picture emerges from tests against Escherichia coli, which was chosen for testing as a reference Gram-negative bacterium including some toxic strains that are transmitted to humans via contaminated vegetables or meat products. This bacterium is also implicated in increasing the risk of colon cancer.[33] First, there are statistically significant differences in antibacterial activity among the chemical constituents (Figure ). The phenolic ferulic and sinapic acids, with percentage inhibition values of 24.4 ± 4.5 and 23.5 ± 3.1%, respectively, displayed significantly higher activities than glycerol or juniperic acid. The two assemblies showed no statistically significant differences in antibacterial activity (p < 0.05). Nonetheless, FerGlyJun had an activity (23.5 ± 2.2%), similar to ferulic acid but significantly higher than its other two constituents, glycerol and juniperic acid. In contrast, the activity of SinGlyJun did not differ significantly from any of its chemical constituents. Just as for P. carotovorum, then, the formation of FerGlyJun and SinGlyJun did not potentiate the E. coli inhibitory activities of the constituent compounds, but neither was there any antagonism demonstrated for their antibacterial capabilities.[32] In the case of FerGlyJun, the assembly displayed a statistically significant activity that was higher than either glycerol or juniperic acid. In contrast to the P. carotovorum case, the activity of the ampicillin positive control (36.6 ± 6.7%) against E. coli was significantly higher than that of the assemblies or their corresponding constituents.

Figure 7

Percentage inhibition for 10 μg/mL of the chemical constituents, the assembly solutions and an ampicillin positive control against E. coli bacteria after 6 h of incubation. Error bars indicate standard error values; Tukey pairwise analysis based on five replicates of each experiment revealed that the differences with respect to the ampicillin positive control were not statistically significant. Tukey analysis was also used to compare each pair of materials shown above.

Conclusions

Inspired by the protective capabilities of suberized cell walls on the surface of plant-based crops such as potatoes and the serendipitous observation of self-assembled solids formed spontaneously from their organic solvent extracts, we engineered three-component systems comprised of a long-chain hydroxyfatty acid, a phenolic carboxylic acid, and glycerol to represent the three major domains of the suberin biopolymer found in potato tubers. The self-assemblies formed after 3–21 days of incubation incorporated all three constituents into their macromolecular structures, which exhibited lamellar ordering analogous to potato phellem tissues and antibacterial properties akin to cork-derived aggregates and films obtained by partial suberin hydrolysis in ionic liquids[18,34] and potato tuber-derived phellem tissue extracts.[19,21] These observations support a phenomenon of inherently favorable self-association and nanoscale ordering for these diverse fatty acid, phenolic acid, and glycerol molecular structures.

The current study was initiated with equimolar ratios of each component in a mutually miscible solvent system that would dissolve all three chemicals, and the solid-state NMR results showed that all three constituents were present in the final resulting solid for both FerGlyJun and SinGlyJun. Moreover, the spin relaxation data offered strong evidence that glycerol was present in a large molecular assembly after incubation with ferulic acid and juniperic acid. Whereas the designed materials consisted primarily of noncovalently bound self-assemblies that preserved the original chemical structures of their constituents, up to 11% of the macromolecular structures contained covalent ester linkages between pairs of the starting materials. By contrast, suberin formation is thought to occur by enzyme-catalyzed chemical reactions within a narrow apoplastic region outside the plasma membrane[35] and involves deposition of end-product polyesters within a complex polysaccharide cell wall. Nonetheless, this model chemical system opens the possibility of exploiting intrinsically favored interactions among aliphatic hydroxyl fatty acids, phenolic acids, and glycerol to promote self-assembly that could be useful for new protective industrial materials derived from natural sources.

Experimental Section

Materials

Ferulic acid (Sigma-Aldrich), sinapic and juniperic acids (Fluka), and glycerol (Thermo-Fisher, Waltham, MA) were used as received. Methanol (ACS grade, Honeywell, Charlotte, NC), deionized water, and acetonitrile (J.T. Baker, Philipsburg, NJ) were also used as solvents. Additional materials included formic acid (Sigma-Aldrich, St. Louis, MO), pyridine (EMD Millipore Corporation, Billerica, MA), N-methyl-N-(trimethylsilyl) trifluroacetamide, and 1% trimethylchlorosilane (Thermo Scientific, Belfonte, PA). [U-13C3]-glycerol was obtained from Cambridge Isotopes (Tewksbury, MA) and [1,2,3-13C3]-ferulic acid was obtained from Sigma-Aldrich.

Preparation of Self-Assemblies

Ferulic acid (Fer) or sinapic acid (Sin), glycerol (Gly), and juniperic acid (Jun) starting materials were each dissolved in either 60% (v/v) methanol (Fer, Sin, Gly) or acetonitrile (Jun) to make a mutually soluble 3 mM solution. To form an equimolar mixture that would maximize the likelihood of intercomponent interactions, 4.0 mL of each stock solution was combined in a glass vial. The mixtures were incubated at room temperature without agitation for 3–21 days, in separate experiments, in a cool, dark, dry place. Threadlike solids were visible to the naked eye from 3 days onward, remaining suspended in the solution rather than amenable to pelleting or creaming. The solvents were evaporated using a nitrogen gas manifold and then removed by lyophilization to obtain a solid material for analysis. Parallel self-assembly experiments with 13C-enriched starting materials used 10% (w/w) [1,2,3-13C3]-ferulic acid and 10% [U-13C3]-glycerol.

Transmission Electron Microscopy (TEM)

A small portion (ca. 1 mg) of the lyophilized solid was suspended in a 50% (v/v) ethanol/water mixture. Negative staining was done by adapting standard procedures[36] as follows. A 5 μL drop of the sample suspension was placed on a 300-mesh nickel-covered carbon film TEM grid (FCF300-Ni Formvar Carbon Film) from Electron Microscopy Sciences (Hatfield, PA), allowed to evaporate for 30 s, and then blotted with filter paper. A 5 μL drop of 1% uranyl acetate dye was then placed on the grid and blotted immediately. A JEOL 2100 TEM instrument (JEOL Ltd., Tokyo, Japan) was used with a LaB6 beam source, a beam strength of 200 kV, and a current density of 30–60 pA/cm3. Images were captured with an Ultrascan 1000 camera (Gatan Inc., Pleasanton, CA) and processed to derive lamellar spacings with Digital Micrograph software (ver. 2.11.14.04.0, Gatan Inc.). At least eight spacing measurements were made at different locations for each of the self-assemblies and plant phellem tissues, allowing us to determine mean values and standard errors for each value.

Solid-State 13C NMR

Solid-state NMR experiments were carried out on a four-channel Agilent (Varian) DirectDrive2 spectrometer (Agilent Technologies, Santa Clara, CA) operating at a 1H frequency of 600 MHz (150 MHz 13C) with either a 1.6 mm FastMAS probe (5–8 mg of the sample) or a 3.2 mm T3HXY probe (15–20 mg). Typical 1H and 13C hard 90° pulse durations for the 1.6 mm probe were 1.5 and 1.7 μs; for the 3.2 mm probe, the values were 2.3 and 2.7 μs; 1.5 μs corresponds to a radiofrequency (rf) field of 167 kHz). The small-phase incremental alternation (SPINAL) method[37] was used to produce 1H decoupling fields of 167 kHz (1.6 mm probe) or 109 kHz (3.2 mm probe) during signal acquisition unless otherwise specified. Typical 13C spectral widths were 30 kHz (200 ppm).

Direct polarization magic-angle spinning (DPMAS) 13C NMR spectra were collected with a recycle delay of 50 s to permit nuclear spins to return to equilibrium between successive acquisitions, a spin rate of 10.00 ± 0.02 kHz, and a nominal temperature of 25 °C following the procedures described for suberized potato periderms.[29] Spectral processing was done using VNMRJ software (ver. 4.2, Agilent) with 150 Hz line broadening and external referencing of 13C chemical shifts to the −CH2– group of adamantane (Sigma-Aldrich) at 38.48 ppm.[38] The resulting peak areas were analyzed by pixel counting using Adobe Photoshop software. This method has been validated against the traditional cut-and-weigh procedure in previous studies.[20,29] The chemical shift regions were defined as follows: aliphatics ((CH2), 15–45 ppm), alkoxy groups (CH2O and CHO, 45–92 ppm), arenes (C=C, 92–160 ppm), and carbonyls (COX, acids or amides, 160–185 ppm).

Cross polarization (CP) conditions were set up using a [U-13C,15N]-glutamine reference sample. 13C power was typically set at full strength during CP, i.e., rf fields of 147 and 93 kHz for 1.6 and 3.2 mm probes, respectively, and the 1H power level was optimized at the Hartmann–Hahn match condition for the −1 sideband with respect to the MAS spinning speed. A 10–15% linearly ramped 1H spin-lock field and 1 or 2 ms 1H–13C contact time were used. 2D dipolar-assisted rotational resonance (DARR) experiments[39,40] were conducted with CP for partially 13C-enriched FerGlyJun assemblies (described above) to determine particular through-space 13C–13C correlations that define the macromolecular structure. This experiment used a spin diffusion mixing time of 300 ms, a recycle delay of 3 s between each of 400 scans, and a MAS rate of 10.00 ± 0.02 kHz.

Values of the 13C spin-lattice and spin–spin relaxation times (T1 and T2) were determined for glycerol and for the FerGlyJun assemblies. (a) Neat glycerol liquid was placed into a 3.2 mm rotor and spun at 5 kHz MAS, with a 10 kHz continuous-wave 1H decoupling field applied during acquisition and 5 Hz of line broadening for data processing. T1’s were measured with DP acquisition using the saturation recovery method.[41] Based on the T1 results, a 4 s recycle delay was deemed more than adequate for determinations of transverse relaxation times (T2); these latter values were measured with a DP Hahn echo pulse sequence. Substantial flattening of the relaxation curve at time points shorter than 1 ms suggested spin-lock artifacts due to short echo times, so only time points greater than 2 ms were included in the analysis. (b) To determine spin relaxation times for the glycerol constituents in FerGlyJun assemblies, two samples were spun at 20 kHz: the aforementioned partially 13C-enriched assembly (in a 1.6 mm rotor) and an unlabeled sample (in a 3.2 mm rotor) that removes any effects of 13C labels on values of T2. The former sample had good CP sensitivity due to enrichment; T2 was measured with CP (recycle delay 3 s, contact time 1 ms) followed by a Hahn echo and using 167 kHz SPINAL decoupling. The latter sample had poor CP sensitivity; T2 was measured with DP and a Hahn echo. To establish a suitable recycle delay for DP, the glycerol T1 was estimated from the saturation recovery experiment conducted on the labeled self-assembly sample. A 5 s recycle delay was used for the T2 measurement; only the signals from glycerol and the methoxy of ferulic acid were observed. Continuous-wave proton decoupling of 5.5 kHz was found to suffice for these peaks. The acquisition times for T2 measurements of neat glycerol, partially labeled FerGlyJun, and unlabeled FerGlyJun were 12 min, 1 h, and 5.7 days, respectively.

Liquid Chromatography–Mass Spectrometry (LC–MS)

Absolute ethanol was used to dissolve the self-assembled lyophilized solids and the authentic starting materials at a concentration of 1.0 mg/mL and analyze them by LC–MS. A Bruker ESI-LC-MS instrument (Bruker, Billerica, MA) with a 2.1 mm × 150 mm Acclaim RSLC 120 2.2-μm 120 Å reverse phase C18 column (Agilent, Santa Clara, CA) was used for these experiments. The mobile phase consisted of 0.1% formic acid in H2O (A) and 0.1% formic acid in acetonitrile (B). Gradient elution proceeded as follows: 0–2 min: 10% B; 2–20 min: 10–100% B; 20–25 min: 100% B; and 25–30 min: 100–10% B. The column temperature was 30 °C, and the declustering potential was 0 V. The ion source temperature was 200 °C. A flow rate of 200 μL/min and an injection volume of 5 μL were used, with each run lasting 30 min. Bruker data analysis software (ver. 4.1) was used to process and analyze the results. The individual Fer, Sin, Jun, and Gly chemical constituents were identified by comparisons of their retention times (RT) and m/z fragments with authentic standards. Heterodimers of these compounds were identified based on a comparison of the m/z data with libraries available from SciFinder, PubChem, and Chemspider.

Gas Chromatography–Mass Spectrometry (GC–MS)

The lyophilized solids and the starting materials were dissolved in 100 μL of pyridine at a concentration of 1.0 mg/mL, to which were added 100 μL of N-methyl-N-(trimethylsilyl)trifluoroacetamide and 1% trimethylchlorosilane. The mixture was heated to 50 °C for 1 h in an incubator-shaker set to 250 rpm. Samples were then analyzed using a QP2010 GC–MS (Shimadzu, Canby, RI) equipped with a Durabond-5 column (30 m × 0.25 mm id, film thickness 0.25 μm; Agilent Technologies, Santa Clara, CA). Injections of 1.0 mL were made using the splitless mode and an oven temperature of 250 °C. The temperature program was adapted from the work of Yang et al.[42] on extracts from wound-healing potato periderm tissues: 0–5 min after reaching 70 °C, to clear the solvent from the system; 5–53 min, 70–310 °C at 5 °C/min; 53–64 min, 310 °C; cooling to 70 °C. The instrument was calibrated with perfluorotributylamine, and two injections were made for each sample. GCMSolution software (Shimadzu) was used for all postrun processing. Authentic samples of the individual Fer, Sin, Jun, and Gly chemical constituents were injected to establish retention times that could be used to identify MS signals from these compounds in the self-assembly mixtures. To estimate the percentage of, e.g., Gly in free vs covalently bound forms in the self-assembly mixture, the integrated peak area of Gly in the MS spectrum was compared with the peak areas from all ions that contained Gly fragments. Where possible, the spectra of the compounds were also compared with mass spectral libraries from the National Institute of Standards and Technology (NIST)[43] and Wiley[44] to aid in structural identification.

Antibacterial Assays

The individual analyte compounds (ferulic acid, sinapic acid, juniperic acid, glycerol, and ampicillin positive control) and self-assemblies (FerGlyJun, SinGlyJun) were dissolved in either ethanol (for organic compounds) or autoclaved deionized H2O (for ampicillin) at concentrations of 10 μg/mL. Mueller–Hinton (M–H) liquid medium was prepared following the manufacturer’s instructions (VWR International, Radnor, PA) and autoclaved for 68 min. Nonpathogenic Escherichia coli (Strain MG1655) and the potato pathogen P. carotovorum (formerly designated as Erwinia carotovorum) (Strain ECC15) bacteria were each cultured on agar plates and incubated at 30 °C for a minimum of 24 h. A colony forming unit (CFU) was taken from the solid medium and placed in a Falcon tube containing 2.0 mL of M–H broth and then placed in an incubator-shaker set to 30 °C and 250 rpm for 24 h; another Falcon tube containing 2.0 mL of M–H media was prepared and stored under the same conditions to serve as a control. The following day, three UV–vis cuvettes were prepared: one with 800 μL of M–H broth (a blank), one with 800 μL of the incubated M–H broth (to check for contamination while in the incubator), and one containing 720 μL of M–H broth and 80 μL of bacterial culture mixture (10% v/v). Absorbance values at 600 nm were obtained using a VWR UV1600PC UV–vis spectrophotometer (VWR International, Radnor, PA) and used as indicators of bacterial cell viability or subsequent lysis. Both M–H blank samples gave readings between 0.03 and 0.06; for E. coli, the absorbance values were 0.400–0.600, and for P. carotovorum, they were 0.250–0.400. Following the Beer–Lambert equation, an amount of the bacteria/M–H mixture was added to 20 mL of M–H media to yield an initial absorbance reading of ca. 0.05, thus ensuring that the subsequent spectrometer readings would not exceed a reliable range of values. Then, a 96-well plate was filled with 2.0 μL of each analyte dissolved in 198 μL of either M–H broth (blank) or the bacterial mixture. A mixture containing 2.0 μL of ethanol dissolved in 198 μL of both M–H and the bacterial mixture was used as a negative control. In the case of ampicillin (positive control[21]), 2.0 μL of deionized H2O was dissolved in M–H broth or the bacterial mixture as the control/blank for the antibiotic. A Spectramax 190 plate reader (Molecular Devices, San Jose, CA) was set to 30 °C; readings were taken every 15 min for 4 h (P. carotovorum) or 6 h (E. coli),[21,45] corresponding to the log phase for each bacterium. For data analysis, we used the readings at 4 and 6 h for P. carotovorum and E. coli, respectively. Tukey pairwise analysis was used to assess statistically significant differences among each of the constituents, assemblies, and positive controls in antibacterial assays against both P. carotovorum (five replicates) and E. coli (six replicates).