Development and Fine Structure of the Glandular Trichomes of Artemisia annua L.

Development of capitate glands on the leaves of annual wormwood (Artemisia annua L.) was monitored with scanning and transmission electron microscopy. Differentiation of foliar cells into gland cells began in the youngest leaf primordia. After differentiation into a 10-celled biseriate structure of two stalk cells, two basal cells, and three pairs of secretory cells, the cuticle of the six secretory cells separated from the cell walls to form a bilobed sac that eventually splits to release its contents. At every developmental stage, the cells of the the gland contained relatively little vacuolar volume. The secretory cells contained extensive endoplasmic reticulum. The plastids of each cell pair were different. At maturity, the apical cells contained proplastids or leucoplasts with only occasional thylakoids. The cell pair below the apical cell pair contained amorphous chloroplasts without starch grains. The basal cell pair contained proplastids or leucoplasts and the stalk cells contained chloroplasts. The stroma to thylakoid ratio in the secretory cell chloroplasts was high. Initially, osmiophilic product was observed most frequently associated with stacked thylakoids, plastid envelopes, and smooth endoplasmic reticulum, although it was associated with all cell membranes. Near the plasma membrane adjacent to cell walls bordering the subcuticular space, the cytoplasm was enriched in smooth endoplasmic reticulum containing osmiophilic material. The apical cell wall of the apical secretory cell pair was reticulated on the inner cytoplasmic side and contained osmiophilic staining on the cuticular side. During early senescence, osmiophilic product was commonly associated with outer mitochondrial membranes.


Introduction
The glandular trichomes of the genus Artemisia are the site of accumulation of a variety of secondary metabolic products (Slone and Kelsey 1985). Artemisia species are the source of many sesquiterpenoid lactones (Duke et al. 1988; Marco and Barbera 1990), and some of these have been histochemically located in the subcuticular space of the head cells of the glandular trichomes of Artemisia umbelliformis Lam. (Cappelletti et al. 1986). The antimalarial drug artemisinin, a sesquiterpenoid lactone, is produced by Artemisia annua (annual wormwood; Klayman 1985). The leaves and stems of this species are covered with glandular trichomes (Duke et al. 1988).
There is considerable worldwide interest in the production of this valuable compound (Klayman 1985; World Health Organization 1988; Bryson and Croom 1991), yet little is known of the development and structure of the glands in which it is assumed to accumulate. There is also interest in this plant, as a dried herb for the floral and craft trades, where it is used in aromatic wreaths (Simon et al. 1990), and as a potential source of naturally occurring pesticides (Duke et al. 1987(Duke et al. , 1988Chen and Leather 1990;Chen et al. 1991;DiTomaso and Duke 1991;Duke 1991). Many of the compounds found in A. annua (Charles et al. 1991) are bioactive and phytotoxic (Duke et al. 1988;Duke 1991). Artemisinin is highly phytotoxic to A. annua itself (Duke et al. 1987 ing sequestration within or secretion from the producing tissues a necessity. Autoallelopathic compounds of velvetleaf (Abutilon theophrasti), are found only in glandular trichomes (Sterling et al. 1987). The subcuticular spaces of the glandular trichomes are likely sites of sequestration of artemisinin and other phytotoxic secondary compounds produced by A. annua. Nothing has been published on the structure and development of the glands of this species.
In this article we examine the developmental structure of the specialized cells of A. annua glandular trichomes at the scanning and transmission electron microscope levels. Special emphasis has been placed on the plastids of these cells because the plastids have distinctive ultrastructural and morphological features. The plastids of the glandular trichome secretory cells were found to be specialized and structurally dissimilar to plastids in other parts of the plant. Others have previously noted that plastids of resin-secreting cells sometimes have plastids with unusual morphology (Dell and McComb 1978b).

PLANT MATERIAL
Seeds of Artemisia annua were planted shallowly (1-2 mm) in 1.2-L pots filled with Jiffy-Mix (Ball Jiffy) potting media. The pots were watered with tap water once every 4 d until seedling establishment and then watered every other day with a dilute solution of Peters 20-20-20 general purpose fertilizer (0.25 g/L). Leaves from mature plants 30-60 d old were sampled for microscopy studies. JOURNAL OF PLANT SCIENCES SCANNING ELECTRON MICROSCOPY (SEM) Meristematic and mature leaf tissue was fixed overnight in 4% glutaraldehyde in 0.1 M cacodylate buffer. After washing in distilled water and dehydrating in ethanol, the tissues were dried in a critical-point drier (Balzers 020). The dried specimens were then mounted on aluminum specimen stubs with silver conductive paint, coated with 20 nm of gold-palladium in a sputter coater (Anatech Hummer X), and observed and photographed in a scanning electron microscope (JEOL JSM-840) at 15 kV. In addition, SEM was performed on the etched faces of resin-embedded material prepared as described below. Etching was accomplished by immersing the faced specimen block for 5 min in a saturated solution of sodium hydroxide in absolute ethanol (sodium ethoxide).

TRANSMISSION ELECTRON MICROSCOPY (TEM)
Meristematic and mature leaf tissue was fixed for 4 h in 0.2% glutaraldehyde in 0.1 M cacodylate buffer. After rinsing for 30 min in the buffer, the tissue was postfixed for 2 h in 1% OS04 in 0.5 M cacodylate buffer. The tissue was then rinsed in distilled water, dehydrated in an acetone series, and embedded in Spurr's resin. Thin sections obtained with a diamond knife on an ultramicrotome (Reichert-Jung Ultracut E) were stained with 2% uranyl acetate for 5 min and poststained with Reynolds's (1963)  The internal volume densities of the components of chloroplast types were compared by the method of Mayhew and Reith (1988). Micro-graphs were printed with a superimposed grid, and the intersections of the lines were used for point counts. Each point overlaying a chloroplast was recorded according to the chloroplast constituent on which it fell. The number of hits per constituent was divided by the total chloroplast hits to obtain the volume density for the chloroplast component.  1B). At this stage, no filamentous trichomes were seen on the leaf surfaces. Later during the early leaf development the capitate glands were sometimes obscured by filamentous trichomes ( fig. 1 C). On the adaxial surfaces of young leaves, filamentous trichomes and capitate glands were arranged in two rows in troughs along either side of the leaf midrib ( fig. 1 C). The pattern of glands on the abaxial leaf surface and on stems was random ( fig. 1D).

Results
By these stages, the capitate glands had expanded, and the glands had taken on a characteristic bilobed shape of a mature structure resulting from the expansion of the cuticle covering the apical cells of the gland ( fig. 2A, B). The cuticular sac was oroken in many of the more mature glands ( . 4). There were three apical cell pairs that we term secreting cells because, at later developmental stages, the subcuticular space was bordered by these cells. The subcuticular space is filled with secretory products that cause the cuticle to expand to its bilobed external appearance. Below these cells were two basal cells and a stalk cell pair, attaching the gland to the leaf epidermis. The gland cells ( fig. 4) are more vacuolated than those of the average gland at this stage. The outer portion of the cuticle stained more intensely than the leaf cuticle at this stage. After all 10 cells had been formed, the cuticular surface of the gland began to separate from the cell wall near the tip of the gland (fig. 4, inset). The onset of cuticular detachment may be as-sociated with the onset of secretory activity. Little or no osmiophilic material was found in the cells at this point; however, osmiophilic material could be seen below the cuticle, in the outer portion of the cell wall.
The plastids of the two apical cells at the 10cell stage were amoeboid and contained no thylakoids, whereas those of the three cell pairs distal to the apical cell pair were amoeboid with thylakoids. Those (fig. 6D). The thylakoids were intimately associated with this osmiophilic material, appearing sometimes to abut or envelop the material ( fig. 6A-C).
As the glands matured further, the amount of osmiophilic material increased to the point that good ultrastructural definition was difficult to obtain, and normal plastid structure began to deteriorate ( fig. 7). The osmiophilic material at this stage could in part result from membrane deterioration caused by senescence. At this stage, the plasmalemma appeared to be separated from the cell wall, perhaps because of deposition of material between the cell wall and the plasmalemma. The cell wall of the apical cells of the gland that bordered the subcuticular space had an osmiophilic layer of granular appearance that at the gland apex was thicker than the nonosmiophilic portion of the cell wall ( fig. 7B). In this micrograph, the reticulated inner side of the cell wall can also be seen. Osmiophilic globules were more commonly associated with the outer membranes of mitochondria or contained within other intracellular membranes. No intercellular spaces developed between gland cells at any point during gland development. After this stage, cellular  . 4-6). The stroma to thylakoid volume density ratio within these gland chloroplasts was only slightly higher than in mature mesophyll cells, if the volume density of starch grains of the mesophyll chloroplasts was subtracted ( fig. 9).

Discussion
Biseriate, capitate glandular trichomes are common in certain genera of the Compositae (Fahn 1988) such as Artemisia (Kelsey and Shafi-zadeh 1980; Ascensao and Pais 1982, 1985Slone and Kelsey 1985;Cappelletti et al. 1986) and Chrysanthemum Peterson 1979a, 1979b;Peterson and Vermeer 1984). The glandular trichomes of Artemisia annua that we have described are quite similar to those of Artemisia umbelliformis (Cappelletti et al. 1986). However, in A. umbelliformis the biseriate glandular trichomes were obscured on leaves by other types of trichomes. Histochemical stains for sesquiterpenoid hydroperoxides resulted in staining of the subcuticular space and head cells of the glandular trichomes. Artemisia campestris glandular trichomes (Ascensao andPais 1985, 1987) were virtually identical to those of A. annua except that the amoeboid chloroplasts that we have described in the subapical two cell pairs were not noted in studies with A. campestris. Ascensao and Pais (1987) found chloroplasts only in the second and third cell pairs (from the base) in the glands of A. campestris. (Franceschi and Giaquinta 1983). In chloroplasts, starch accumulation indicates that production exceeds export. In the tissues that we sampled, mesophyll chloroplasts contained high levels of starch ( fig.  6D). Thus, if photosynthetic rates of gland and mesophyll chloroplasts were similar, it follows that the glandular chloroplasts are exporting photosynthate at a dramatically higher rate than mesophyll chloroplasts, or starch metabolism differs between the two chloroplast types. This is consistent with the view that the chloroplasts of the gland cells are highly involved in converting photosynthate into terpenoids and exporting them to the cytoplasm. The more amoeboid form of gland cell chloroplasts as compared with the typically ellipsoid chloroplasts of mesophyll tissues may in part result from the absence of a large central vacuole in the gland cells. Compaction between the plasma membrane and tonoplast may contribute to the ellipsoid shape of chloroplasts of mature mesophyll cells.

Others have noted head or stalk cells of capitate glands containing chloroplasts without starch grains (Vermeer and Peterson 1979b; Ascensao and Pais 1982). The plastids of soybean (Glycine max) glandular trichome cells contain starch grains but have no thylakoids
The stalk cells of resin-secreting glandular hairs of Beyeria viscosa have chloroplasts, whereas the cells at the hair tip have amoeboid leucoplasts (Dell and McComb 1978b). The amoeboid leucoplasts are ensheathed in an extensive network of smooth ER. These cells are ultrastructurally similar to the apical cells of the A. annua capitate gland (figs. 4, 5A). The apical or head cells of the glandular hairs of Newcastelia viscida (Dell and McComb 1978b) and Eremophila fraseri (Dell and McComb 1977) also contain amoeboid leucoplasts in an ER network.
The two-and four-cell stages of development of the A. annua gland are similar to those of Cannabis sativa L.; however, after the next periclinal cell division in C. sativa, the apical cells undergo anticlinal proliferation to form a secretory disc (Kim and Mahlberg 1991). The cuticle separates only from the one-cell-thick discoid tier of cells. In Chrysanthemum (Vermeer and Peterson 1 979a), the stalks of the capitate glands are also composed of 10 cells in a biseriate structure, as in Artemisia species. In both Chrysanthemum and Artemisia, the secretory sac is composed of the greatly expanded cuticle of the last three pairs of cells. As in C. sativa (Kim and Mahlberg 1991), the cuticle of the secretory sac is significantly thicker than the cuticle of other leaf structures. We did not find this in A. annua.
The majority of evidence indicates that the terpenoids generated by capitate glands are produced by the cells of the gland rather than by other cells of the leaf (Mahlberg et al. 1984;Gershenzon et al. 1991;Wagner 1991). While terpenoids can be synthesized outside of plastids (Kleinig 1989), plastids appear to be highly involved in terpenoid production in epidermal glands of plants. Nielsen et al. (1991) found that a nonsecreting line of tobacco lacked chloroplasts in gland cells. Spring and Bienert (1987) found production of terpenoids in sunflower (Helianthus annuus L.) capitate glands to be highly dependent on light level, indicating photosynthetic participation in synthesis. Mahlberg et al. (1984) reported that material interpreted as secretory product of C. sativa L. gland cells appeared at the plastid surface and migrated through the cytoplasm to accumulate along the cell surface adjoining the secretory cavity between the cuticle bonded to a portion of the primary cell wall and the remainder of the primary cell wall. This is very similar to our interpretation of the apparent movement of osmiophilic material in A. annua glands.
An ultrastructural study of terpenoid-producing epithelial cells of A. campestris resin ducts showed a slightly different pattern (Ascensao and Pais 1988). During early stages of secretion, somewhat isodiametric plastids with poorly developed thylakoids contained osmiophilic material in the stroma and thylakoids. These plastids were surrounded by smooth ER. Later, the plastid envelope and ER contained osmiophilic material and the plastid had lost internal osmiophilic material. The inner sides of cell walls bordering the resin-duct lumen had ingrowths similar to the reticulations of the inner cell wall of the apical gland cells of A. annua (figs. 5A, B, 7D). Furthermore, the smooth ER-enriched cytoplasm adjacent to the reticulated cell wall was similar to that observed in apical stalk cells ( fig. 5B). Dell and McComb (1977) found resin formation and secretion in the cells of glandular hairs of E. fraseri to be similar in most respects to what we have described. The only significant differences are that no reticulation of the inner cell walls of secreting cells was noted and that plastids did not appear to be as actively involved in the process.
Differences in plastid morphology between cell pairs of the gland stalk were striking. Of the five cell pairs, only three pairs appeared to have chloroplasts with thylakoids capable of photosynthesis ( fig. 5). Several questions are posed that cannot be answered by ultrastructural studies such as presented here. Do all or only certain of the gland cells participate directly in terpenoid production? Three of the five cell pairs of the gland border the secretory cavity, implying that three pairs contribute to filling the cavity with terpenoids. Does each cell type specialize in synthesis of different terpenoids? The morphological differences imply a division of biosynthetic function. Do those cells with photosynthetically active plastids provide substrate for terpenoid production in adjacent cells? These and other questions remain to be answered. Although intact glands can be isolated and studied biochemically (Dell and Mc-Comb 1978a; Keene and Wagner 1985;Gershenzon et al. 1991;Yerger et al. 1992), no methods are currently available for separating the different cell types of the gland. Determination of the identity of the material in the subcuticular space will require cytochemical methods and/or selective extraction and chemical analysis.