Increased accumulation of specific carotenoids in plastids through plant breeding or genetic engineering requires an understanding of the limitations that storage sites for these compounds may impose on that accumulation. Here, using Capsicum annuum L. fruit, we demonstrate directly the unique sub-organellar accumulation sites of specific carotenoids using live cell hyperspectral confocal Raman microscopy. Further, we show that chromoplasts from specific cultivars vary in shape and size, and these structural variations are associated with carotenoid compositional differences. Live-cell imaging utilizing laser scanning confocal (LSCM) and confocal Raman microscopy, as well as fixed tissue imaging by scanning and transmission electron microscopy (SEM and TEM), all demonstrated morphological differences with high concordance for the measurements across the multiple imaging modalities. These results reveal additional opportunities for genetic controls on fruit color and carotenoid-based phenotypes.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
The carotenoid biosynthetic pathway, together with the specific carotenoid compositions of many plants, have been well described and reviewed extensively (Fraser and Bramley, 2004; von Lintig, 2010). Plants are excellent dietary sources of pro-vitamin A, macular pigments and carotenoid-based anti-oxidants, and there have been many efforts to improve the content of these compounds in crop plants using both traditional breeding and genetic engineering efforts (Schmidt-Dannert et al., 2000; Sandmann et al., 2006; Bai et al., 2011; Farre et al., 2011; Shumskaya and Wurtzel, 2013). In general, the results of these efforts produce the expected outcome, i.e. increased carotenoid content; however, there are also unexplained outcomes. In particular, the relationship between the structure of the organelle (chromoplast) and the increased carotenoid content has been highlighted as an area for further investigation (Egea et al., 2010; Shumskaya and Wurtzel, 2013).
Chile peppers (Capsicum sp.) accumulate very high concentrations and complex mixtures of secondary metabolites, particularly carotenoids, in pericarp (fruit wall) tissue. This accumulation of carotenoids results in brightly colored red, yellow and orange fruits, which are characteristic of this genus (Davies et al., 1970; Thorup et al., 2000; Ha et al., 2007; Guzman et al., 2010; Wahyuni et al., 2011; Rodriguez-Uribe et al., 2012). These carotenoids are synthesized and accumulate within chromoplasts, which are differentiated plastid organelles (Kuntz et al., 1992; Bouvier et al., 1994; Ytterberg et al., 2006; Shumskaya and Wurtzel, 2013). The chromoplasts that reside in the chile pepper fruit wall develop from the chloroplasts present in these cells when the fruit was immature green (Spurr and Harris, 1968). Green chloroplasts also accumulate carotenoids as essential components of their photosynthetic structures, although the concentration and diversity of carotenoids increases dramatically during the chromoplast transition (Egea et al., 2010). We have characterized the differences in carotenoid biosynthetic gene expression in chile varieties that accumulate distinct carotenoid profiles, resulting in red- and orange-colored fruit (Guzman et al., 2010, 2011; Rodriguez-Uribe et al., 2012). The differences in gene sequence and transcript accumulation for the carotenoid biosynthetic enzymes are not sufficient to explain all of the fruit phenotypes observed. The presence of additional control measures, potentially in the chromoplast structures in these fruit, may explain the accumulation of different carotenoid compositions and amounts.
The concept that there is an association between chromoplast morphology and carotenoid composition has been proposed since the 1880s, as discussed in an early review of these organelles (Straus, 1953). More recent investigations have focused on comparative studies between varieties with differing pigment profiles, such as Daucus carota lines that accumulate various levels of β–carotene (Kim et al., 2010), red and white loquat lines (Fu et al., 2012), red and yellow papaya (Schweiggert et al., 2011), and Capsicum annuum lines with distinctly colored fruit (Simpson et al., 1977). Differences in the abundance of chromoplast sub-organellar structures were observed in these comparative studies. In some studies, carotenoid compositions were determined, but direct subcellular detection of specific carotenoids was not performed. Inferences were drawn based on these pairwise comparisons: lines with abundant specific carotenoids show specific appearance of crystalloid structures in the chromoplasts, suggesting that the crystalloid structures are composed of the abundant carotenoid (Simpson et al., 1977; Schweiggert et al., 2011).
Understanding the structural requirements for packaging carotenoids in organellar spaces will improve our ability to increase the levels of these important dietary sources of pro-vitamins (e.g. β–carotene), macular pigments (lutein and zeaxanthin) and antioxidants (von Lintig, 2010; Farre et al., 2011). Further, the sub-organellar location of these carotenoids has been proposed to explain differences in dietary bioavailability among a number of crop systems (Kopsell and Kopsell, 2006; Jeffery et al., 2012).
Differentiation of individual carotenoid compounds and their relative abundances at subcellular and sub-organellar levels is possible using hyperspectral confocal Raman microscopy and multivariate imaging analysis (Collins et al., 2011). Raman spectroscopy has been shown to be a powerful technique for plant science due to its ability to provide quantitative chemical information based on the interaction of excitation radiation with specific molecular vibrations (Baranska et al., 2013). Carotenoids contain conjugated polyene chains that result in electronic absorption bands in or very near the visible excitation wavelengths typically used for Raman spectroscopy. When this phenomenon occurs, vibrations associated with that chromophore are strongly enhanced. Hyperspectral confocal Raman microscopy combines the chemical information available from Raman spectroscopy with the three-dimensional spatial resolution of confocal microscopy. The result is a Raman spectrum from each voxel within a sample. Multivariate spectral analysis tools (Schoonover et al., 2003) are able to deconvolute the small changes in the resonance Raman spectrum that result from differences in the conjugation length and substitution of the carotenoid backbone, making localization and quantification of specific carotenoid molecules possible, even in complex biological tissues.
We have successfully used hyperspectral confocal Raman microscopy to determine carotenoid distribution in unicellular photosynthetic organisms at the subcellular level in cyanobacterial cells (Vermaas et al., 2008; Collins et al., 2012) and the sub-organellar level in green microalga (Collins et al., 2011). Pudney et al. (2011) demonstrated use of Raman microscopy to assess the location of the dietary carotenoids β–carotene, lycopene and lutein in fruit exudate prepared from three tomato varieties (Solanum lycopersicum). They observed carotenoids in ‘inclusion bodies’ which probably represented chromoplasts; however, the work was not performed in intact fruit tissue. These methods may be extended to the study of chromoplast composition and structure in intact fruit samples based on the diversity of fruit color found in C. annuum cultivars.
There are numerous examples of chile cultivars with specific carotenoid compositions comprising differing abundances of carotenes versus specific xanthophylls (Guzman et al., 2010; Wahyuni et al., 2011), providing excellent germplasm resources to explore relationships between cellular ultrastructure and carotenoid chemical diversity. In this work, we directly tested the hypothesis that unique carotenoids accumulate in specific subcellular and sub-organellar sites, resulting in varietal specific chromoplast structures. This approach used four independent imaging modalities with companion analysis of the carotenoid composition performed using standard chromatographic methods. Tissue, cell and subcellular structures were resolved by scanning electron microscopy (SEM) and transmission electron microscopy (TEM), with detail at the single-cell level being obtained utilizing laser scanning confocal microscopy (LSCM) and hyperspectral confocal Raman microscopy to achieve spatial resolution of five specific carotenoids (β–carotene, zeaxanthin, lutein, violaxanthin and capsorubin) in live samples of fruit walls from five distinct chile (C. annuum) cultivars.
Heritable variation in carotenoid composition in fruit pericarp of chile cultivars
Chile cultivars were selected for analysis based on a preliminary LSCM screen performed to identify chromoplast shape classes as well as on β–carotene and total carotenoid accumulations based on chromatographic data (Guzman et al., 2010). This resulted in selection of four varieties with terminal red-colored fruit and one variety with terminal orange-colored fruit (Figure 1). Additional carotenoid compositional data on these varieties were obtained using HPLC (Table 1). The five cultivars grouped into three classes: NuMex Garnet and LB–25 had similar high abundance of β–carotene and capsanthin, NuMex Heritage 6–4 and NuMex Nematador had lower total carotenoid and low β–carotene levels, and Costeño Amarillo had high total carotenoids and a uniquely high violaxanthin content.
Table 1. Capsicum annuum fruit carotenoid levels and chromoplast dimensions
NuMex Heritage 6–4
ND, not detected. Values are means ± standard deviation; means within a row that are followed by different letter are significantly different (α = 0.05; Duncan's multiple range test).
Carotenoids (μg g−1 fresh weight pericarp)
795.73 ± 110.03 A
791.74 ± 66.24 A
714.98 ± 142.99 A
468.52 ± 97.09 B
455.11 ± 109.10 B
106.66 ± 15.94 A
93.05 ± 10.30 B
27.67 ± 9.53 E
61.10 ± 13.78 C
43.81 ± 10.68 D
28.76 ± 4.10 B
28.26 ± 2.66 B
193.36 ± 58.13 A
18.77 ± 6.29 B
16.19 ± 3.34 B
39.29 ± 10.91 A
41.86 ± 3.2 A
14.38 ± 4.87 C
23.59 ± 5.49 C
29.75 ± 12.39 C
199.93 ± 23.51 A
203.69 ± 14.52 A
128.47 ± 25.79 B
141.95 ± 46.30 B
39.84 ± 9.92
Chromoplast volume (μm3)
17.04 ± 5.72 C
23.80 ± 4.42 BC
27.52 ± 9.65 B
21.01 ± 2.99 BC
57.67 ± 8.67 A
17.05 ± 4.53 C
24.73 ± 5.99 B
24.72 ± 4.34 B
18.29 ± 3.27 C
60.65 ± 13.68 A
19.98 ± 5.96 C
25.17 ± 8.90 BC
27.52 ± 5.66 B
20.33 ± 3.22 C
62.20 ± 16.74 A
Chromoplast aspect ratio (length/width)
1.25 ± 0.30 B
1.55 ± 0.14 B
3.14 ± 0.53 A
1.19 ± 0.11 B
1.31 ± 0.26 B
1.77 ± 0.61 B
1.74 ± 0.37 B
3.29 ± 0.41 A
1.00 ± 0.00 C
1.24 ± 0.14 C
1.28 ± 0.21 C
1.64 ± 0.31 B
3.18 ± 0.38 A
1.00 ± 0.00 D
1.30 ± 0.21 C
Dimensions of fruit pericarp chromoplasts vary across chile cultivars
The overall shape and volume of the chromoplasts for each cultivar were determined using multiple methods of microscopic analysis (Figure 1 and Table 1). The chromoplasts in two varieties, NuMex Garnet and LB–25, had the classic ovoid chromoplast shape (aspect ratios ranged between 1.2 and 1.7 depending on microscopic method of analysis). The chromoplasts in Costeño Amarillo had an unusual elongated, sickle-shaped morphology (aspect ratio > 3), while the chromoplasts in NuMex Heritage 6–4 and NuMex Nematador were spherical (aspect ratios ranged between 1.0 and 1.3 depending on microscopic method of analysis). These distinct chromoplast shapes were detected by all four microscopy modalities (Figure 1 and Table 1). We observed these distinct chromoplast shapes for these varieties using LSCM in fruit collected from field-grown or greenhouse-grown plants and in fruit harvested over a 2-year period. These distinct shapes and sizes are therefore specific to the genetic background of the cultivars and are not due to environmental effects of the plant cultivation or technical artifacts introduced by the microscopic processing steps.
Detection of sub-organellar sites of specific carotenoid accumulation
In a previous study, we demonstrated optical separation of the carotenoid and chlorophyll auto-fluorescence in our LSCM system (Kilcrease et al., 2012). Similar LSCM methods have been used in tomato fruit studies (Egea et al., 2011). Representative images of the optical separation of this auto-fluorescence in chile fruit are shown in Figure S1, and Movie S1 shows a 3D display of the ‘turning’ region of the fruit wall containing sectors that are both green and red in color.
In mature chile fruit, there was no detectable chlorophyll signal in the LSCM images; only carotenoids are detected (Figure 1). We complemented this approach by confocal Raman microscopy and multivariate curve resolution (MCR) analysis to speciate carotenoids based on their resonance-enhanced Raman vibrations in the fingerprint region (900–1600 cm−1) and weakly coevolving fluorescence emission, thus enabling their relative abundances at the subcellular level to be assessed (Tauler, 1995; Collins et al., 2011; Baranska et al., 2013). Spectra from individual carotenoid standards as well as the MCR models developed for the pericarp samples are shown in Figure 2. The carotenoid chemical standards had unique spectra that are characterized by strongly enhanced ν1 (approximately 1520 cm−1) and ν2 (1157 cm−1) frequencies arising from the C=C and C–C stretches of the polyene backbone, respectively, as well as vibrational overtones (2000–4000 cm−1) superimposed on a fluorescence background (Figure 2a).
Multivariate curve resolution is a linear unmixing method that uses iterative numerical analysis to deconstruct overlapping spectral features into linear contributions of the underlying spectral components and their corresponding relative abundance (Haaland et al., 2003; Jones et al., 2012). Two independent MCR models were developed to describe the carotenoid composition in the complex live pericarp samples (Figure 2b,c), and comparison with the reference spectra permitted assignment of the MCR-derived spectra to specific carotenoids. Examples of images generated from the MCR-derived spectra using each of these two models and the resulting false-color merged image are shown in Figure S2A,B.
Among the four cultivars that shared the same MCR model (LB–25, NuMex Garnet, NuMex Heritage 6–4 and NuMex Nematador), it was possible to compare specific carotenoid abundances detected by the Raman approach at the per-chromoplast level (Figure S3). The rank order for abundance of specific carotenoids was the same for detection by Raman microscopy and HPLC extraction (Table 1) for β–carotene and zeaxanthin. This concordance further supports the reliability of the spatial distribution obtained by the Raman microscopic method.
The relative abundance of chromoplast internal structures visible in TEM was compared across all five cultivars using various chromoplasts from multiple fruit collected from at least three individual plants per line; representative fields are shown in Figure 1. Chromoplast structures were identified based on published descriptions (Spurr and Harris, 1968; Harris and Spurr, 1969; Simpson et al., 1977). Plastoglobules were detected in chromoplasts for all five varieties, with NuMex Garnet having a higher number than the other varieties. Other organellar structures were found in some but not all varieties, for example fibrillar carotenoid crystals were observed in four varieties, but not in LB–25. A scheme of the structures observed is shown in Figure S4. A unique structure was observed in only one cultivar; in Costeño Amarillo, the chromoplasts contained numerous electron-dense bodies, which were named ‘violaxanthin-associated’ bodies (Figure 3). The confocal Raman images of this cultivar demonstrated localized sites of violaxanthin accumulation within the chromoplast, consistent with the structures detected by TEM.
Subcellular distribution of carotenoids
Extra-chromoplast lipid bodies were detected that had significant carotenoid content (Figures 1 and 4). In the case of Costeño Amarillo, many of these lipid bodies contained lutein. These structures were immediately adjacent to chromoplasts (Figure 4). Further, colorless lipid bodies were also present that appeared to lack carotenoid content as determined by Raman detection. In contrast, NuMex Garnet, LB–25 and NuMex Heritage 6–4 accumulated β–carotene in extra-chromoplast lipid bodies (Figure 1). Although these extra-chromoplast lipid bodies were observed in all chile varieties, they were less frequent in NuMex Nematador and LB–25 (Figure 1).
In diverse C. annuum fruit, distinct chromoplast architecture and composition were detected, characterized and confirmed using four independent microscopic methods, with the sub-organellar localization of specific carotenoids correlating with carotenoid composition as determined by chromatographic methods. The shapes of the chromoplasts detected by confocal Raman microscopy were similar to the shapes detected by the other microscopic methods (Figure 1 and Table 1). The distribution and abundance of specific carotenoids were readily observed in these images and correlated with the HPLC data in Table 1. Further, there were specific sub-organellar regions that had increased concentrations of specific carotenoids, e.g. violaxanthin in Costeño Amarillo or β–carotene in NuMex Heritage 6–4. There was distinct cultivar-specific spatial organization of the carotenoid signal in the chromoplasts.
Several key findings result from this study. First, the confocal Raman method was able to assign a unique chemical composition to subcellular structures detected by TEM and LSCM. Second, the numerous electron-dense bodies observed in the TEM images of Costeño Amarillo were determined to be localized sites of violaxanthin accumulation within the chromoplast and thus were named ‘violaxanthin-associated’ bodies (Figure 3). Third, the per-chromoplast quantification of carotenoid detected by the Raman signal (Figure S3) is in agreement with the cultivar-specific pattern of carotenoids determined by HPLC analysis (Table 1). Finally, carotenoids were detected in lipid bodies external to the chromoplast. These bodies were observed in all of the varieties tested, although they differed in abundance and apparent chemical composition (Figures 1 and 4). For example, in Costeño Amarillo, these bodies had Raman signatures consistent for lutein, while in NuMex Heritage 6–4, the lipid bodies were found to be composed almost exclusively of β–carotene. Importantly, there were numerous instances of colorless lipid globules lacking any Raman signal bands. LSCM did not detect any carotenoid signal outside the chromoplasts. This is due to the lower sensitivity of the LSCM compared to the confocal Raman microscopy, and the inability of LSCM to speciate carotenoids.
Although the Raman/MCR method is more sensitive than LSCM, it is not equally sensitive to all carotenoids, because the resonance enhancement of the Raman bands is expected to scale as the square of the absorption probability at the excitation wavelength (Johnson and Peticolas, 1976). Thus, we cannot detect all the carotenoids measurable by chemical extraction and HPLC separation using Raman/MCR analysis of living tissue. The two MCR models (Figure 2 and Figures S2 and S3) demonstrate these detection limits. Although present at low levels in Costeño Amarillo fruit (Table 1), β–carotene was not identified by the Raman/MCR model, and, similarly, violaxanthin, which is present but at low levels in the other four fruit types was also not identified by the relevant Raman MCR model.
The diversity of fruit color in well-characterized C. annuum cultivars makes this system especially powerful for studies relating carotenoid chemical composition to plastid morphology and ultrastructure. Carotenoid accumulations in chile fruit are among the highest in plant organs, with some varieties exceeding 800 μg g−1 fresh weight (10 mg g−1 dry weight) (Guzman et al., 2010). In contrast, the high pigment–1 mutants of tomato accumulate carotenoids to 300 μg g−1 fresh weight (Cookson et al., 2003), and transgenic tomatoes engineered for increased carotenoid content via multiple gene constructs may achieve only 1 mg g−1 dry weight carotenoid in the fruit (Apel and Bock, 2009). However, transgenic versions of various cereal grains have higher levels of total carotenoids than wild-type Capsicum fruit (Bai et al., 2011).
The differences in carotenoid accumulation in plant tissues have been attributed to a variety of factors, from environmental effects through biochemical regulation to developmental signals (Bramley, 2002; Liu et al., 2004; Egea et al., 2010). In addition to these factors influencing carotenoid accumulation, two genes with predicted structural roles in plastid development have been described: fibrillin or plastid-associated protein (Deruere et al., 1994) and Orange (Or) (Lu et al., 2006). Transgenic tomato plants expressing the fibrillin gene from C. annuum have twofold increased levels of carotenoids compared to non-transgenic controls, but the transgenic plants also have a delayed persistence of thylakoid structures in the ripening fruit (Simkin et al., 2007). In ripening chile fruit, fibrillin is proposed to cause self-assembly of carotenoid and polar lipid fibrils surrounded by the fibrillin protein (Deruere et al., 1994). In potato (Solanum tuberosum), fibrillin transcripts were detected in screens of drought-responsive transcripts, and expression in leaves increased in response to oxidative stress, suggesting a role in protection of the thylakoid membranes (Langenkamper et al., 2001). No comparative expression profiles for fibrillin in chile varieties with different carotenoid accumulation profiles are available.
This current study clearly demonstrates there is diversity in fruit pericarp chromoplast architecture and associated differences in carotenoid content. Crosses between chile lines with high carotenoid content (e.g. LB–25) and lines with large chromoplast volumes (e.g. Nematador) would produce hybrid lines that may be used to test hypotheses about the role of organellar structure in the regulation of carotenoid accumulation.
Confocal Raman microscopy on living fruit tissue is an independent method that allows detection of compositional differences without the limitations of extraction processes and saponification steps. Other technologies to develop a spatial map of important metabolites in plants, e.g. mass spectrometry approaches, are able to detect a wide range of compounds, but the spatial resolution is still at the tissue level and microscopic processing and embedding may be required (Li et al., 2013; Ye et al., 2013). Our success in the carotenoid-rich structures of Capsicum fruit suggests that the confocal Raman approach may be useful for cellular and subcellular spatial analysis of carotenoids in other multicellular plant organs. Further, use of this technical advance allows us to propose that there are probably structural aspects to the controls on the accumulation and storage of these essential dietary metabolites. Given the extensive germplasm base with well-described fruit color phenotypes, further investigations into the genes responsible for these structures or patterns of storage are possible in Capsicum fruit.
Seeds of five Capsicum annuum cultivars (NuMex Heritage 6–4, NuMex Nematador, NuMex Garnet, LB–25 and Costeño Amarillo) were germinated in a controlled greenhouse on the New Mexico State University Las Cruces campus according to standard chile propagation practices (Bosland et al., 2012). Seedlings were transplanted to the field (Leyendecker Science Center, 13 km south of Las Cruces), where plants were cultivated using standard cultural practices for growing chiles in southern New Mexico (Bosland and Walker, 2004). Fully mature, ripe fruits were collected and stored under protection from light at 4°C. Fruit were processed for carotenoid extraction or prepared as specimens for specific microscopic analysis within 24–48 h of collection.
Carotenoid extractions and analysis
Pericarp tissue was hand-dissected and frozen at −80°C for further analysis. Frozen samples were chopped into small pieces measuring approximately 4 mm2 using a manual food chopper (Farberware Soft Grips food chopper model #83427–93), and homogenized in CHCl3 using a Polytron generator (model #PT 10–35, Polytron, http://www.kinematica.ch/en.html) and then phase-separated by centrifugation (1000 g, 5 min) at 4°C. The CHCl3 phase was collected and dried down under N2 in a warm water bath. Freshly dried samples were saponified, and carotenoids were quantified after HPLC essentially as previously described (Rodriguez-Uribe et al., 2012). Calibration curves were generated at 450 nm using the reference standards β–carotene, lutein and lycopene (Sigma, http://www.sigmaaldrich.com/united-states.html) and capsanthin, capsorubin, zeaxanthin, antheraxanthin, violaxanthin and β–cryptoxanthin (CaroteNature GmbH, http://www.carotenature.com/). Total carotenoid content was calculated using the total chromatogram peak area at 450 nm and the β–carotene calibration curve. The mean concentration for each specific carotenoid was calculated using the GLM procedure (SAS 9.2; SAS Institute Inc., http://www.electronmicroscopysciences.com/), based on independent HPLC analyses of multiple individuals for each cultivar (n =7–11). Duncan's multiple range test was run using the GLM procedure to determine whether there were significant differences between the means for the total carotenoid content and specific carotenoid contents among the cultivars.
Scanning electron microscopy
Mature pepper pericarp tissue was hand-dissected into sections of approximately 6 × 18 mm from areas between inner placental veins. Excised tissue was then immediately fixed in a solution of 4% glutaraldehyde, 0.1 m imidazole/HCl buffer (pH 7.2) in filtered deionized water (Milli-Q, http://www.millipore.com) for 24 h at room temperature, and then dehydrated through ethanol using a Spurr's kit (Electron Microscopy Sciences, http://www.electronmicroscopysciences.com/) with extended step times. Dehydrated pericarp tissue was plunge-frozen in liquid nitrogen. Following full temperature equilibrium (within seconds, specimens stop bubbling), sections were freeze-fractured on a brass block in a liquid nitrogen bath. Fractured sections were collected and thawed into fresh absolute ethanol.
Fully thawed and fractured sections were layered into a drying basket with Whatman No. 1 (http://www.whatman.com/) trimmed filter paper discs as dividers. The loaded basket was critical point-dried using a LADD Model 28000 (http://www.laddresearch.com/about-us) critical-point dryer with five or six liquid CO2 exchanges of 2 min duration until no ethanol remained. Dry sections were placed at a density of 3–5 sections per plug, using carbon adhesive sticky tabs. The sides of each section were coated with colloidal silver paint and thereafter sputter-coated completely with nano gold using a Denton Vacuum Desk IV cold sputter coater (http://www.dentonvacuum.com/). Images were acquired using a Hitachi S–3400 II scanning electron microscope (http://www.hitachi-hta.com/).
Laser scanning confocal microscopy
Mature pepper pericarp tissue was hand-dissected into sections of 3 × 12 mm from areas between inner placental veins, and placed cut side down in a 40 μm thick glass-bottomed Petri dish. Sections were imaged using a Leica SP5–II laser scanning confocal fluorescence microscope (http://www.leica-microsystems.com/home/) with a 20× water immersion lens and correction collar (HCX PL APO20x/0.70). Autofluorescence was captured using modified settings from (Egea et al., 2010) with excitation at 488 nm and emission windows for carotenoids at 515–590 nm and for chlorophyll at 650–700 nm. A 3.55× electronic zoom was utilized, bringing the total magnification to approximately 70×. A 5 μm stack of images was collected for five areas per section.
Transmission electron microscopy
Mature pepper pericarp tissue was hand-dissected into sections of 2 × 10 mm from areas between inner placental veins. Samples were fixed and dehydrated through ethanol as for the SEM processing. Dehydrated samples were infiltrated with Spurr's resin using protocols modified from Spurr and Harris (1968) in 12 h steps of 25, 50, 75% and absolute resin on a rocking platform shaker, with an additional exchange of absolute resin, then specimens in 25 ml screw cap vials were placed on a rotator for 24 h. Infiltrated sections were placed into molds and polymerized at 40°C for 24 h.
Resin-embedded specimens were trimmed and faced using a Leica UC6 electronically assisted microtome with a glass knife. Thin sectioning to a thickness of 70 nm was performed using a Diatome histo diamond knife (https://www.diatomeknives.com/). Sections were examined for thickness using color of the section, fumed with chloroform to flatten the sections and placed on copper–Formvar 100 mesh grids (Electron Microscopy Sciences). The sections were next allowed to thoroughly dry before staining on a droplet (50 μl) of uranyl acetate for 5 min. Grids were briefly washed with approximately 2 ml of de-ionized H2O before being placed, section side down, on a droplet of lead citrate for 5 min. Grids were once again washed with de-ionized H2O and allowed to fully dry. Dried, stained grids were visualized using a Hitachi H–7650 transmission electron microscope.
Statistical analysis of chromoplast dimensions
For determination of chromoplast volume and aspect ratio by LSCM or SEM (n =15 each), independent lengths and widths were determined for each cultivar using image fields from at least five individuals; for TEM, 30 independent lengths and widths were determined for each cultivar using image fields from at least five individuals. Chromoplast volumes and aspect ratios were calculated based on these measurements. Duncan's multiple range test was performed using the GLM procedure of SAS 9.2 (SAS Institute Inc.) to calculate the means of the volumes and aspect ratios and to determine whether there were significant differences between these means among the cultivars.
Hyperspectral confocal Raman microscopy
Mature pepper pericarp tissue was hand-dissected into sections of 1 × 10 mm from areas between inner placental veins. Samples were placed onto a well-slide with a drop of water and covered with a #1.5 coverslip. Confocal Raman images were acquired using a custom hyperspectral microscope (Sinclair et al., 2006) with 488 nm excitation and a 60× oil objective (Nikon Plan Apo, NA = 1.4, http://objectives.nikoninstruments.com/). This microscope has demonstrated the ability to detect the Raman vibration of xanthophyll carotenoids due to the well-known resonance enhancement effect, with spatial resolution of 250 nm in the x and y dimensions and 600 nm in the z dimension (Vermaas et al., 2008). Spectra from 500 to 800 nm were acquired at a rate of 2100 spectra/sec using an xIon, an electron-multiplying charge-coupled device (Andor, http://www.andor.com/).
Multivariate image analysis
Spectral images were pre-processed as described previously (Jones et al., 2012) and combined into composite image datasets for analysis based on carotenoid composition. One dataset included spectral images for Costeño Amarillo, while a second dataset combined spectral images for LB–25, NuMex Garnet, NuMex Heritage 6–4 and NuMex Nematador. MCR analysis (Tauler, 1995; Schoonover et al., 2003; Van Benthem and Keenan, 2004) was performed independently on the composite datasets as described previously (Haaland et al., 2003; Collins et al., 2011, 2012), and two MCR models were developed to model the signal variance to >95% in both instances.
Statistics on a per-chromoplast basis were generated by masking off spectral images to exclude regions without chromoplasts and colored lipid bodies. The mean per-chromoplast intensities were calculated by summing the per-pixel MCR component intensities across the mask and dividing the total by the number of pixels that comprised the mask. Appressed chromoplasts were manually segmented using a custom-written program (Collins et al., 2012).
In total, approximately 180 chromoplasts (and associated colored lipid bodies) (n =20, 42, 47 and 72 for NuMex Nematador, NuMex Garnet, LB–25 and NM Heritage 6–4, respectively) were analyzed across all chile types except for Costeño Amarillo, which was analyzed using a separate MCR model (to account for its unique carotenoid composition), and therefore direct comparison with the other chile types is not valid.
The authors thank P. Bosland for Capsicum seeds and for field space, P. Cook, Core University Research Resources Laboratory, NMSU for LSCM, TEM and SEM methods, M. Sinclair, Electronic, Optical, and Nano Materials, SNL, for maintenance and use of the hyperspectral confocal microscope, and H. Jones for development and support of the MCR software. The plant cultivation, carotenoid composition, SEM, LSCM and TEM work were performed at New Mexico State University and supported in part by the New Mexico Agricultural Experiment Station and grants from the US National Science Foundation award numbers MRI-DBI-095817 and MRI-DBI-0520956, from the US Department of Agriculture (National Institute of Food and Agriculture 2010-34604-20886), Raman microscopy and image analysis were performed at Sandia National Laboratories and were supported as part of the Photosynthetic Antenna Research Center, an Energy Frontier Research Center funded by the US Department of Energy, Office of Science, Office of Basic Energy Sciences (award number DE–SC 0001035). Sandia National Laboratories is a multi-program laboratory that is managed and operated by Sandia Corporation, a wholly owned subsidiary of Lockheed Martin Corporation, for the US Department of Energy's National Nuclear Security Administration (contract number DE–AC04-94AL85000).