Spectral mixture model Fresh leaf surfaces are composed of many organic compounds and produce complex spectra showing many absorption bands. At the present level of understanding it is not possible to identify all of the specific compounds responsible for every spectral feature. However, it is possible to identify major classes of compounds and to recognize whether a particular class is especially abundant in a particular leaf sample. Figure 1 is a mixture simulation of Olea europea, a well-studied species (Frega & Lercker, 1985; Bianchi et al., 1992, 1993). As detailed in the figure caption, the simulation was made by digitally summing ATR spectra of cellulose, water, nonacosane (C29H60), carnauba wax and oleanolic acid (C30H48O3). Although carnauba wax is not found in O. europea, the wax has high ester content (Vandenburg & Wilder, 1970; Tulloch, 1973) and provides a convenient ester standard. The simulation illustrates many of the common ATR spectral features of leaves and provides a framework for understanding details of leaf spectra. Note, however, that the mixture proportions used in the simulation do not necessarily represent the mass proportions for reasons discussed later.
Figure 1. Simulation of Olea europea leaf. (1a) Water: (a) 3288 cm−1; (b) 2125 cm−1; (c) 1633 cm−1. (1b) Cellulose: (a) 1055 cm−1; (b) 1032 cm−1. (1c) Nonacosane (C29H60): (a) 2914 cm−1; (b) 2846 cm−1; (c) 1472 cm−1; (d) 1462 cm−1; (e) 729 cm−1; (f) 719 cm−1. (1d) Oleanolic acid (C30H48O3) (University of Sao Paulo, Phytochemistry laboratory): (a) 1688 cm−1; (b) 1029 cm−1; (c) 996 cm−1. (1e) Carnauba wax (see composition in Vandenburg & Wilder, 1970): (a) 1735 cm−1; (b) 1166 cm−1. (1f) Simulation of the composition of Olea europea adaxial surface. Sum: 0.2 water + 10 cellulose +1 nonacosane + 0.5 oleanolic acid + 7 carnauba wax. (1g) Adaxial surface of fresh leaf of Olea europea. Dotted line indicates x-axis scale change at 2000 cm−1. Unless otherwise noted, all reagents were obtained from Aldrich Corp. (Sigma Aldrich).
Download figure to PowerPoint
The water spectrum in Fig. 1a, shows an intense and broad OH stretching band near 3288 cm−1 (arrow ‘a’), a broad shallow band at 2125 cm−1 (arrow ‘b’), and an HOH bending feature at 1633 cm−1(arrow ‘c’). This last feature is not evident in the Olea europea spectrum, but is common in many of the other leaves studied. The steep downward slope on the right side of the water spectrum is the shoulder of a strong absorption band at 686 cm−1 that is outside of the measurement range. This shoulder is seen in most of the spectra studied. Cellulose (Fig. 1b) is mainly characterized by two strong bands at 1055 cm−1 and 1032 cm−1 (arrows ‘a’ and ‘b’), which were assigned by Maréchal and Chanzy (2000) to the C–O stretching of primary and secondary alcohols, respectively. Nonacosane, a compound found in many plant waxes, including O. europea (Fig. 1c), shows the main features of long chain aliphatic compounds: the methylene (CH2) stretching features at 2914 cm−1 and 2846 cm−1 (arrows ‘a’ and ‘b’), methylene bending features at 1472 cm−1 and 1462 cm−1 (arrows ‘c’ and ‘d’) and the rocking doublet of methylene at 729 cm−1 and 719 cm−1 (arrows ‘e’ and ‘f’) (Silverstein & Webster, 1998). Oleanolic acid (Fig. 1d), which is abundant in O. europea, displays a strong band at 1688 cm−1 assigned to the carbonyl (C=O) stretching vibration mode in acids (arrow ‘a’) (Silverstein & Webster, 1998), and two characteristic unassigned bands at 1029 cm−1 (arrow ‘b’) and 997 cm−1 (arrow ‘c’). Carnauba wax (Fig. 1e) illustrates the main spectral features of esters: carbonyl (C=O) stretching at 1735 cm−1 (arrow ‘a’), and C-C(= O)-O stretching at 1166 cm−1 (arrow ‘b’) (Silverstein & Webster, 1998), which from now on will be called simply the carboxyl or C–O stretching feature. Both of these ester features are strongly expressed in the leaf spectrum.
Cutin vs cutan Lycopersicon esculentum and B. vulgaris are well-studied species that illustrate ATR spectral differences between cutin and cutan (Fig. 2). Whereas cutin is composed mainly of esterified monomers of hydroxyl- and epoxy-fatty acids (Holloway, 1982b; Kolattukudy, 1996), cutan is an unsaponifiable polymer made of polymethylene chains linked by ether bonds (Heredia, 2003). The polyester material cutin comprises more than 80% (by weight) of the tomato cuticle (Baker, 1982a) and the ATR spectrum shows strong features at 1727 cm−1 (Fig. 2, arrow ‘a’), 1165 cm−1 (Fig. 2, arrow ‘b’) and 1103 cm−1 (Fig. 2, arrow ‘c’), which were assigned to the ester carbonyl stretching, the carboxyl asymmetrical stretching and the carboxyl symmetrical stretching vibrations, respectively, by Ramirez et al. (1992). According to Baker (1982a), the epicuticular waxes of the tomato fruit comprise only 3% of the cuticle's total weight and are mainly composed of fatty acids, flavonoids, triterpenoids, and hydrocarbon homologues. Consistent with this small wax amount, the bands observed in the tomato fruit spectrum (Fig. 2) do not appear to be characteristic of these wax-forming compounds.
Figure 2. Cutin vs cutan spectra. Comparison between cutin-rich Lycopersicon esculentum (tomato) fruit and cutan-rich Beta vulgaris (sugar beet) leaf. Lycopersicon esculentum: (a) 1727 cm−1 (νC=O) (b) 1165 cm−1 (νasC–O–C) and (c) 1103 cm−1 (νsC-O-C). Beta vulgaris: (d) 1735 cm−1 (ν–COOH or ν–C=O) (e) 1606 cm−1 (νas–COO−). Dotted line indicates x-axis scale change at 2000 cm−1.
Download figure to PowerPoint
Beet leaves lack cutin (Jeffree, 1996) and therefore the ATR spectrum of a beet leaf (Fig. 2) should not show ester bands. The major band at 1606 cm−1 (Fig. 2, arrow ‘e’) is similar to a band assigned to the stretching vibration of carboxylate anion (–COO−) by Villena et al. (2000) in a study of a nondegradable fraction of Clivia miniata cuticle, presumably an analog for cutan. These same authors assigned a band at 1730 cm−1 to the carbonyl stretching vibration of fatty acids, rather than to esters, inasmuch as esters had already been extracted from their sample. A similar feature occurs near 1735 cm−1 in beet (Fig. 2, arrow ‘d’).
Figure 3 shows spectra of a fresh leaf of Quercus rubra (red oak) (Fig. 3b) compared with its own epicuticular wax (Fig. 3a) and a treated leaf that was immersed in chloroform to remove the surface waxes (Fig. 3c). The epicuticular wax of the red oak displays a double carbonyl band at 1732 cm−1 and 1720 cm−1 (Fig. 3a, arrows ‘a’ and ‘b’), and the fresh leaf similarly displays bands at 1730 cm−1 and 1714 cm−1 (Fig. 3a, arrows ‘d’ and ‘e’). By contrast, the leaf washed with chloroform has only a single carbonyl feature at 1729 cm−1 (Fig. 3c, arrow ‘h’) that is likely analogous to the band seen in tomato at 1727 cm−1 (Fig. 3d, arrow ‘k’). Therefore the band at 1729 cm−1 is inferred to be caused by cutin and not by wax. The wax spectrum displays a relatively weak C–O band at 1171 cm−1 (Fig. 3a, arrow ‘c’) compared to its C=O feature (Fig. 3a, arrows ‘a’ and ‘b’), whereas the leaf washed with chloroform has a C–O band at 1169 cm−1 (Fig. 3c, arrow ‘i’) that is stronger than the corresponding C=O band (Fig. 3c, arrow ‘h’). The fresh leaf shows the C–O band at 1168 cm−1 (Fig. 3b, arrow ‘f’) as having an intermediate strength. Therefore, the 1169 cm−1 band is inferred to be produced by a constituent situated beneath the surface wax layer, and by analogy with the tomato spectrum (Fig. 3d, arrow ‘l’), is probably caused by cutin. The band near 831 cm−1 (Fig. 3, arrows ‘g’, ‘j’ and ‘m’) has not been assigned, but is seen in all the spectra except for the wax. The band may be related to phenolic compounds present in cutin (Holloway, 1982b).
Figure 3. Wax vs cutin features. Comparison between (a) the epicuticular wax of Quercus rubra, (b) the fresh leaf of Q. rubra (adaxial surface, baseline corrected), (c) the leaf washed with chloroform (adaxial surface, baseline corrected) and (d) Lycopersicon esculentum as a standard for cutin. Arrows within parts (a–d): (a) 1732 cm−1; (b) 1720 cm−1; (c) 1171 cm−1; (d) 1730 cm−1; (e) 1714 cm−1; (f) 1168 cm−1; (g) 829 cm−1; (h) 1729 cm−1; (i) 1169 cm−1; (j) 831 cm−1; (k) 1727 cm−1; (l) 1165 cm−1; (m) 832 cm−1. Spectra are offset for clarity.
Download figure to PowerPoint
Sources of spectral variability (1) Leaves from different times of the growing season: Young leaves collected in May, and mature leaves collected from the same individuals during July through September, show complex spectral differences involving changes in the intensities of features at 1008, 1032 and 1050 cm−1 associated with cellulose, and possibly other polysaccharide-rich substances such as hemicelluloses and pectins (Fig. 4a–c). The band at 1008 cm−1 showed little change in A. hippocastanum, and A. octandra, but became more developed in mature leaves of C. ovata, C. florida, L. tulipifera, and M. pomifera (Fig. 4a). Wilson et al. (2000) assigned the 1008 cm−1 band to polygalacturonic acid, which is a variety of pectin (Fry, 2004). The band at 1032 cm−1 became more intense in mature leaves of A. rubrum, Q. alba, Q. rubra and T. cordata (Fig. 4b), whereas the band at 1050 cm−1 became stronger in F. grandifolia and M. grandiflora (Fig. 4c). The 1050 cm−1 band in F. grandifolia and M. grandiflora may be related to the abundance of amorphous silica on the leaf surfaces (Postek, 1981). As shown in Fig. 4c, the mature leaves display a broadening and/or displacement of the spectral feature near 1050 cm−1 compared with the immature leaves that could be consistent with increased amounts of amorphous silica. Several other species exhibited only minor variations in all of these bands between their young and the mature leaf spectra, including C. caroliniana, L. styraciflua and P. serotina (Fig. 4d). However, a band near 840 cm−1 in P. serotina, which is probably caused by aromatic compounds, shows a marked decrease in intensity in the mature leaf. An opposite change in the 840 cm−1 feature intensity can be seen in G. biloba (Fig. 4e). Unlike the other species examined, G. biloba did not display bands at 1008, 1032, and 1050 cm−1, but did show band variations between immature and mature leaves at 957, 981 and 1020 cm−1 (Fig. 4e). Causes of these band variations in G. biloba are not presently known.
Figure 4. Spectral comparisons between young leaves (continuous lines) and mature leaves (dotted lines). The data are organized into three main groups reflecting changes in particular spectral bands. (a) Species showing band changes at 1008 cm−1 (from top to bottom) Aesculus hippocastanum, Aesculus octandra, Carya ovata, Cornus florida, Liriodendron tulipifera and Maclura pomifera; (b) species showing band changes at 1032 cm−1 (from top to bottom) Acer rubrum, Quercus alba, Quercus rubra and Tilia cordata; (c) species showing band changes at 1050 cm−1 (from top to bottom) silica gel (SiO2), Fagus grandifolia and Magnolia grandiflora; (d) species showing little variation (from top to bottom) Carpinus caroliniana, Liquidambar styraciflua and Prunus serotina (with arrow marking possible phenolic feature); (e) Ginkgo biloba shows stronger bands (marked with arrows) in the mature leaves at 1104, 1020, 981, 957 and 840 cm−1. Both spectra in each pair are shown on the same y-axis.
Download figure to PowerPoint
Spectral differences in the C=O bands near 1727 cm−1 and the C–O features near 1165 cm−1 (Fig. 4) were also seen in young and mature leaves. These differences did not follow a simple pattern of behavior. Both bands were stronger in young leaves of A. rubrum (4.0% and 4.5% greater intensity for the C=O and C–O features, respectively), C. ovata (7.2 and 4.5%), C. florida (4.4 and 3.5%), M. pomifera (7.6% and 1.9%), and T. cordata (8.0 and 4.0%). Conversely, the same features were stronger in the mature leaves of L. tulipifera (6.2% and 5.2% greater intensity for the C=O and C–O features, respectively), M. grandiflora (4.0% and 4.2%) and Q. alba (3.9% and 6.8%). A number of other species displayed almost no variation in these band intensities for young and mature leaves, including A. hippocastanum (0.7% and 0.9%), A. octandra (1.0% and 3.2%), C. caroliniana (1.2% and 0.1%), F. grandifolia (2.1% and 0.6%), G. biloba (0.3% and 0.0%), L. stiraciflua (1.6% and 0.9%), P. serotina (1.0% and 2.2%) and Q. rubra (2.6% and 0.6%). The C=O absorption features showed additional variations – in some species a single feature occurring near 1730 cm−1 or 1727 cm−1 in young leaves became a doublet feature in the mature leaves with band centers near 1730 cm−1 or 1727 cm−1 and 1716 cm−1. This was observed for A. rubrum, A. hippocastanum, C. ovata, F. grandifolia, M. pomifera, P. serotina, Q. alba, and Q. rubra. In other cases the opposite spectral behavior was observed, for example, in A. octandra and L. tulipifera two carbonyl features merged into one as the leaves matured. Finally, there were cases where no changes in the C=O band(s) occurred in the course of leaf development, including C. caroliniana, C. florida, L. styraciflua and M. grandiflora. Possible factors involved in these band changes are discussed later.
(2) Adaxial and abaxial leaf surfaces: The adaxial and abaxial variations were of two main types reflecting differences in wax composition and structure (Fig. 5), or in the abundance of trichomes on the abaxial surface (Fig. 6). The spectra of wax extracts from the two surfaces of A. rubrum show absorption bands in the same positions, but with some bands having different relative intensities (Fig. 5a). The wax spectra point to an overall similarity between the adaxial and abaxial wax compositions, with some differences in the concentrations of the wax constituents.
Figure 5. Acer rubrum. (a) Spectra of the adaxial and abaxial waxes; (b) spectra of the adaxial (top) and abaxial surfaces (bottom). The wax of the adaxial surface (scanning electron microscopy image, top right) is much thinner than the abaxial wax and permits the infrared to penetrate more deeply into the cuticle and cell wall, resulting in relatively stronger cutin (C=O and C–O) and cellulose/hemicellulose features with a maximum at 1032 cm−1. Bars, 14 µm. Thicker wax on the abaxial surface results in relatively stronger aliphatic CH2 features.
Download figure to PowerPoint
Figure 6. Comparison of attenuated total reflectance (ATR) spectra and scanning electron microscopy images of adaxial and abaxial leaf surfaces of Fagus grandifolia. Abundant trichomes on the abaxial surface may be responsible for the pronounced polysaccharide feature at 1031 cm−1.
Download figure to PowerPoint
The whole leaf spectra of A. rubrum display significant spectral differences (Fig. 5b). The abaxial surface has stronger aliphatic features (CH2 bands at 2914 cm−1 and 2847 cm−1, 1472 cm−1 and 1462 cm−1, and 729 cm−1 and 719 cm−1) likely owing to greater wax thickness. The adaxial surface shows stronger contrast in the principal ester bands (C=O at 1725 cm−1 and C–O at 1165 cm−1) and in the major cellulose/polysaccharides feature at 1032 cm−1. This may be explained as a result of a thinner wax layer, possibly coupled with differences in cell structure that allow more ATR interaction with cutin and cellulose on and near the adaxial surface.
Scanning electron microscopy images of the Acer rubrum abaxial and adaxial leaf surfaces (Fig. 5b) show the structure of the epicuticular waxes. The abaxial surface has a porous meshwork of thin wax plates, while the adaxial surface has a striated wax, forming subparallel rodlets. The chemical origins of these marked structural differences are not presently known.
Figure 6 compares the adaxial and abaxial spectra of F. grandifolia. The main difference is that the adaxial surface shows a broad band at 1050 cm−1 while the abaxial feature displays a narrower band at 1031 cm−1. The 1050 cm−1 feature resembles the feature attributed to silica in Magnolia grandiflora (Fig. 4c) and the 1031 cm−1 feature is similar to features seen in a variety of species and attributed to cellulose or other polysaccharides (Fig. 4b). A possible interpretation is that the 1031 cm−1 band is related to a chemical constituent present in trichomes, as many abaxial leaf surfaces with abundant trichomes display this absorption band (Fig. 6, bottom right).
(3) Sun and shade leaves: Figure 7 shows a spectrum of extracted cuticular wax of M. grandiflora (Fig. 7a) compared with the spectra of the adaxial surfaces of sun and shade leaves (Fig. 7b). The main spectral differences between the sun and shade leaves are the presence of stronger C=O and C–O stretching bands in the sun leaf spectrum. Both bands are absent in the wax spectrum (Fig. 7a), and probably originate from cutin, indicating a thicker cuticle that is typically characteristic of sun leaves (Osborn & Taylor, 1990). The sun leaf also has a more intense CH2 stretching feature (c. 11% stronger than the shade leaf), indicating a thicker wax layer. Conversely, the shade leaf has a stronger feature at 1046 cm−1 compared with the sun leaf, probably reflecting greater ATR interaction with cell wall materials, including cellulose, which are partly obscured by the thicker wax and cutin layers of the sun leaf surface.
Figure 7. Spectral comparison between (a) Magnolia grandiflora wax and (b) M. grandiflora sun (continuous line) and shade (dotted line) leaves (adaxial surfaces). The CH2, C=O and C–O features produced by waxes and cutin are stronger in the sun leaf, whereas the 1046 cm−1 feature related to cell wall materials is stronger in the shade leaf.
Download figure to PowerPoint
Some other species for which sun and shade leaves were compared, including A. hippocastanum, F. grandifolia and L. tulipifera also showed stronger ester (cutin) bands in the sun leaves. Conversely, G. biloba, Q. robur and Q.rubra showed only minor variations in the spectra for both types of leaves.