... „ —J ... -J rhe peculiar leaves of Wilkesia, which he excludes from Madinae omparison of them with and places in Galinsoginae; and he suggests form the subject of an that Dt/baiitia (sens// lato), Argyroxiphium, and ist. 1957^). The present Wilkesia form a related endemic group. The egetative anatomy of the fact that he relates Di/buutia to elements of n respects. For the pur- both Heliantheae and Senecioneae is not in- the writer interprets Du- congruent, in his opinion, because Dubautia 196 PACIFIC SCIENCE, Vol. XIII, April, 1959

1 part glacial acetic acid) provided good ma­ Gaud. var. platyphylla Hillebr.. Carlquist terial for anatomical study. The others of the H20a (UC); D. platyphylla (Gray) Keck, species, however, were studied from frag­ Forbes 1101M (UC); D. railliardioidesHillebr., ments of herbarium specimens treated accord­ Carlquist Hl6 (UC); D. reticulata (Sherff) ing to techniques described earlier (Carlquist, Keck, Rock 8573 (UC); D. Rockii (Sherff) 1957a: 207). The liquid-preserved material Keck, Rock 8601 (BISH); D. scabra (DC) was prepared by means of techniques de­ Keck, Carlquist H20 (UC); D. Sherffiana Fosb., scribed in that paper. Preparations made from St.John 23924 (BISH); D.struthioloides (Gray) herbarium material, both as whole mounts Keck, Wilkes Exped. s. n. (GH, type); D. and as sections, gave quite satisfactory infor­ ternifolia (Sherff) Keck, Forbes 1175 (BISH); mation concerning vegetative anatomy. Every D. thyrsiflora (Sherff) Keck, Forbes 1203M effort was made to secure mature leaves, to (GH : /). waialealae Rock, Rock Oct. L$>11 analyze structure of several portions of a leaf, (GH, cotype). Argyroxiphinm Caliginii Forbes, and to secure stems both in primary condi­ Carlquist H28 (UC); A. Grayanum (Hillebr.) tion, for a study of stem structure, and with Degener, Carlquist H27 (UC); A. sand­ some secondary growth, for study ot mature wichense DC, Carlquist H19 (UC); Wilkesia pith types. Pith of the rosette species of gymnoxiphium Gray, Carlquist H10 (UC). Ariyroriphium {A. sandwichense) was taken Gratitude is expressed to the curators of from basal (epicotyl) portions ot the rosette. these herbaria for use of their materials. Data Leat anatomy was studied in all species except concerning die stem of D. platyphylla were D. coriacea (Sherff) Keck, D. demissifolia obtained from a slide (prepared from a liquid- (Sherff) Keck, and D. kohalae (Skottsb.) St. preserved collection by Degener, no. 19188) John. These species would probably add little kindly given to the writer by Dr. John W. to the gamut of variation presented here. Hall. Thanks are due Dr. Harold St. John for Stem anatomy was studied for the majority of his assistance during the author"s field work the species of Duhautia and Argyroxiphium, as in the Hawaiian Islands. well as tor Wilkesia. Subspecific variation was ANATOMY disregarded in this study, and analysis of the hybrids recognized by Sherff (1935.) did not sections seem feasible at present. Characters of leaf anatomy as seen in tran­ Duhautia arborea (Gray) Keck, Rock 8344 sections constitute the best indications of (UC); D. ciliolata (DC.) Keck, var. laxiflora specific diversity in Duhautia. These charac­ (DC) Keck, Rock 10326 (UC); D. Hille- ters include bifacial or isolateral organization, brandit (H. Mann) Keck, Hillebrand s.n. relative size of cells in upper and lower epi­ (GH, syntype); D. Knudsenii Hillebr., Carl­ dermis, thickness of cell walls in epidermis, quist H15 (UC); D. Gray, Heller width of palisade, frequency ot bundle- 2616 (UC); D. latifolia (Gray) Keck, Heller sheath extensions, presence of fibers within 2887 (UC); D. laxa H. & A., Carlquist H12 the bundle sheath, presence of secretory ca­ (UC); D. linearis (Gaud.) Keck, Rock 8123 nals in the bundle sheath, and occurrence of (UC); D. lonchophylla (Sherff) Keck, St. John uniseriate nonglandular or biseriate glandular 10303 (UC); D. magnifolia Sherff, Rock 9012 hairs. Cuticle is not included in description ot (BISH); D. Menziesii (Gray) Keck, Carlquist epidermis cell wall thickness, because it was invariably a very thin layer. Numerous prep­ H17 (UC); D. microcephala Skottsb., Carl­ arations showed separation ot the cuticle from quist H14 (UC); D. molokaknsis (Hillebr.: the epidermis (e.g., Fig. 3). The species of Keck, Forbes 86Mo (UC); D. montana (H. Duhautia are compared below by reference to Mann) Keck, Rock 8594 (UC); D. paleata "types" based on the species illustrated. No Gray, Forbes 914K (BISH); D. plantaginea Vegetative Anatomy — CARLQL'IST 197 taxonomic significance is necessarily implied D. Knudsenii (Fig. 1). Leaf bifacial; cells of by this arrangement. The groups below, how­ upper epidermis larger than those of lower ever, do appear to be natural groups in some epidermis, thin walled; 1 layer of palisade instances. For each group, the illustrated spe­ present; bundle-sheath extensions present cies is used as a basis for description, and the on major veins only; fibrous bundle cap species most closely resembling it are listed present only on major veins; secretory- beneath, with the features by which they canals 4 near larger veins, 1 or 2 near smal­ differ mentioned in parentheses. ler veins; trichomes absent at maturity.

FlGS. 1-6. Leaves of Dubatitia. (1-5) Leaf transections. (1) D. Knudsenii. X 75. (2) D. railliardioides. X 67. (3) D. latifolia, X 95. (4) D. uaialealae, X 87. (5) D. Menziesii, X 67. (6) D. lomhophylla, portion of cleared leaf to show marginal trichomes. below, and secretory canals, which appear as darker lines (on account of resins) beside two larger veins, X 58. 198 PACIFIC SCIENCE, Vol. XIII, April, 1959

Like D. Knudsenii: D. latifolia (Fig. 3). Leaf bifacial, thin; up­ D. laxa (a few uniseriate hairs on lower per and lower epidermis cells of approxi­ surface). mately the same size, very thick walled: a D. molokaiensis (thicker outer walls on single layer of cells weakly defined as upper epidermis). palisade; bundle-sheath extensions fre­ D. reticulata (thick-walled upper epi­ quent, consisting of isodiametric to fiber­ dermis; uniseriate hairs frequent on like sclereids, although no "bundle cap" both surfaces, especially in the libers are present on the vascular bundles: grooves overlying veins on the 1 or 2 secretory canals present within the upper surface). bundle sheath or sheath extension of many- 2. D. railliardioides (Fig. 2). Leaf bifacial; veins; hairs absent at maturity. cells ot upper epidermis larger than those Like D. latifolia: of lower epidermis; prominent grooves on D. laevigata (sclerenchyma of bundle upper surface above major veins; 2 pali­ sheaths and sheath extensions con­ sade layers present; sheath extensions sisting of thick-walled fibers; secre­ present on many larger veins; fibers com­ tory canals absent'. mon at both poles of larger veins; 4 secre­ i. D. uaialealae (Fig. 4). Leaf bifacial; cells tory canals in outer portion of bundle of upper and lower epidermis of about the sheath of larger veins, 1 or 2 near smaller same size, very thick walled; 1 or 2 layers veins; scattered uniseriate trichomes on somewhat differentiated as palisade; sheath lower surface and margins (upwardly extensions present only on midvein; prom- appressed). nent strands of thick-walled fibers on Like D. railliardioides: phloem face of veins, also on xylem pole D. magnifolia, D. microcephala, D. of larger veins; secretory canals absent; plantaginea (virtually identical with large uniseriate trichomes abundant on D. railliardioides). upper surface and on margins. D. paleata (sheath extensions on ma­ D. Menziesii (Fig. 5). Leaf somewhat iso- jor veins only: fibers more common lateral (stomata frequent on upper surface; at phloem pole of veins; secretory- adaxial palisade cells short, nearly isodia­ canals absent; numerous uniseriate metric; cells on abaxial face like adaxial hairs on both surfaces). palisade cells in shape, with more nu­ D. ternifolia (fibers more common on merous chloroplasts than cells in the cen­ phloem face of bundles). tral portion of the leaf); leaf very thick; D. scabra, D. ciliolata, D. lonchophylla, bundle-sheath extension only on midvein; D. linearis, D. Sherjfiana, D. thyrsi- large groups of fibers on phloem face of flora (upper epidermis very large veins (a few libers on xylem pole of celled; bundle-sheath extension on larger veins); secretory canals infrequent, midvein only; fibers mostly on 1 or 2 in bundle sheath ot larger veins phloem face of bundle; usually 1 near leaf base; thick-walled uniseriate secretory canal above xylem pole of trichomes scattered over both surfaces and larger veins). margin of leaf. D. platyphylla (fibers at phloem face of Like D. Menziesii: bundles only; secretory canals ab­ D. Hillebrandii (nearly identical with sent; biseriate glandular hairs with D. Menziesii). large multicellular capitate heads D. arborea (fibers lacking near veins; and long stalks abundant on both biseriate glandular hairs present on surfaces; uniseriate hairs absent). both leaf surfaces). Vegetative Anatomy — CARLQUIST 199

D. Rockii (a single secretory canal impracticable because the acid released such adaxial to most veins). compounds from the wall and caused swelling D. montana (fiber strands near veins and distortion of the leaves. The phenomenon small :. described by the writer as "abaxial meristem" D. struthioloides (more nearly isolateral in Argyroxiphium is totally lacking in Du­ than D. Menziesii). bautia, as it is in Wilkesia. The different types of leaf structure de­ Leaf Venation tailed above reflect ecological conditions for Dubautia to a certain extent, as well as being Figures 7-9 show the range of venation indicative of taxonomic groupings. The thin, types which may be seen in the genus Du­ glabrous leaves of D. Knudsenii, with thin- bautia. Dubautia latifolia (Fig. 7) is excep­ walled epidermal cells, correspond to the tional for the genus in its reticulate vein pat­ moist forest areas on Kauai where that spe­ tern with polygonal areoles containing nu­ cies grows. On the contrary, the tendency merous freely-terminating veins. Except for the toward thick, isolateral leaves with thick- major veins (two of which are shown, left and walled epidermal cells in D. Menziesii seems right of center, respectively), there is no related to the dry, sunny alpine habitat of this tendency for orientation of veins longitudi­ plant. Most of the species indicated here as nally in the leaf. D. latifolia is apparently the having similarity to D. Menziesii are also only species which has this venation pattern. alpine types. This distinctive feature is recognized by the The highly specialized leaf anatomy of segregation of D. latifolia as a separate section Argyroxiphium (Carlquist, 1957^) bears little of the genus, " Venoso-reticulatae" (Sherff, superficial resemblance to that of even the 1935: 111), whereas the remainder of the isolateral-leaved species of Dubautia. The Railliardia species are included by Sherff in presence of fiber strands along veins, secre­ rhe section "Nervosae." The most typical ve­ tory canals (in A. Caliginii), and the presence nation condition in Dubautia is illustrated by of uniseriate hairs (much more abundant in D. linearis (Fig. 9). In this species, the five Argyroxiphium) are similar features. The capi­ main veins run the length of the leaf, as in tate biseriate glandular hairs on inflorescence D. latifolia. Marginal veins and those between bracts of Argproxiphium and Wilkesia are very the primary veins also have a distinctly longi­ similar to those on leaves of Dubautia platy- tudinal course. Areoles are more nearly rec­ phylla. Dubautia shows little differentiation of tangular, elongated with the long axis of the leaves near inflorescences into inflorescence leaf. Vein endings are frequent. Dubautia bracts, whereas in Argproxiphium a marked microcephala (Fig. 8), D. plantaginea, and D. difference, reflected in the anatomical struc­ railliardioides show a modification of this ture, is seen between vegetative leaves and pattern. In these species all of the major vena­ bracts of the inflorescence. The leaf of tion and much of the minor venation show a Wilkesia, which differs markedly from that of longitudinal orienration. Areoles are very Argyroxiphium (Carlquist, 1957t/), appears elongate, and vein endings are more infre­ much like the leaf of Dubautia laevigata in quent than in D. linearis. The predominant transection. Although the genera Dubautia pattern of venation in the genus Dubautia, and Wilkesia lack the intercellular deposition then, is like that of D. linearis, and is therefore of pectic compounds which is characteristic a deparrure from typical reticulate dicoty­ of Argyroxiphium, cell walls in leaves of ledonous venation. One may question whether Dubautia and Wilkesia are rich in pectic com­ the reticulate venation of D. latifolia is a pounds. Treatment ofleaves of the latter with primitive condition or a "reversion." The ad­ hydrofluoric acid prior to sectioning proved vanced position of D. latifolia in characters 200 PACIFIC SCIENCE, Vol. XIII, April, 1959 Vegetative Anatomy — CARLQUIST 201 other than venation may suggest the latter the genus. The types indicated include all interpretation. Probably the type of venation that have been reported previously for the shown by D. linearis is basic for the genus be­ family, except for mukilacunar nodes with an cause it is found in conjunction with other alternate-leaved condition. This latter type unspecialized characters in some of the puta- does not occur in any of the alternate-leaved tively primitive species. The type of venation species, which are uniformly trilacunar. As illustrated by D. microcephala seems clearly a some of the examples below show, transitions specialized condition. between the recognized types may occur in a Comparison of leat venation in Dubautia single stem. with that of Argyroxipbium and Wilkesia is 1. Trilacunar, alternate leaves. Examples: D. pertinent in this regard. Paradental sections Sberffiana, D. lenchophylla. of Arpyroxiphi/im leaves reveal that the basic 2. Trilacunar, opposite (decussate) leaves. venation is similar to that of D. linearis if the Examples: D. Menziaii (Fig. 12), D. mon- additional vein series denoted by the writer as tana, D. latifolia, D. platyphylla, D. a "upper" and "lower" sets or veins (Carlquist, The drawing of D. Menziesii shows, below, 1957;/1 are disregarded. Wilkesia I Fig. 10), on a trilacunar node with only the traces of the contrary, is markedly different from the facing leaf indicated. Above is a node Argyruxiphium in its venation. This remarkably in which both leaves of the pair are sup­ monocotyledonous venation pattern is char­ plied by 4 traces. This node, then, is transi­ acterized by longitudinal orientation of vir­ tional between trilacunar and mukilacu­ tually all of the major and minor veins. nar. The additional pair of traces, on the Cross-connections between the longitudinally far side of the stem, fuse to a single trace oriented veins are infrequent. Although this a short distance below the node. Such fu­ pattern appears to be anomalous in compari­ sion of adjacent laterals is not character­ son with the types described above, it may istic of trilacunar nodes in Dubautia. well have been derived from a type like that of Dubautia microcephala by means of a more 3. Trilacunar, verticillate leaves. Examples: nearly exclusive production of longitudinally D. ternifiilia, D. wai . i he 3 traces of oriented veins. Venation similar to that of each leaf at a node continue downward Wilkesia has been described for another mem­ into the vascular cylinder without fusion ber of Compositae in Schlectenclalia luzulae- of adjacent laterals. folia, a member of the tribe Mutisieae (Urban 4. Mukilacunar. opposite (or decussate) and Mobius, IXH4). leaves. Examples: D. laxa (Fig. 14), D. laevigata (Fig. 13), O. magnifolia, D. Knud- senii. As comparison of Figures 13 and 14 Nodal Anatomy shows, mukilacunar nodes can differ in The <;enus Dubautia is curious because of number of traces and in the fusion of the variation in phyllotaxy and the presence traces below a node. In D. laxa (Fig. 14), of both petiolate and sessile clasping leaves each leaf is provided with 5 (or 6) traces, within the genus. Both of these characters of which continue downward into the vascu­ leaf attachment are related to the vascular lar cylinder independently. The most lat­ pattern of nodes. Although the author's ma­ eral traces in each leaf base increase the terial did not prove sufficient for placing each number of veins in the leaf margin by species in a category, the data below probably branching toward the margin. In D. laevi­ indicate the full range of types which occur in gata (Fig. 13), more numerous traces con-

FIGS. 7-10. Portions of cleared leaves to show venation. (7) D. latifolia; margin at left; X 10.5. (8) D. micro­ cephala, X 13-5. (9) D. linearis; margin at right; X 95. (10) Wilkesia ®mnoxiphium; margin below; X 9-5. 202 PACIFIC SCIENCE, Vol. XIII. April, 1959

cinue downward into the vascular cylinder. cance because, according to Dr. I. W. In addition, the most lateral pair shown Bailey (personal communication), this on the near side of the stem unite. This condition has never been reported in di­ union of adjacent laterals, mentioned cotyledons. The author (1957^/) suggested above for D. Menziesii, is characteristic of that it may occur in Argyroxiphit/m Gray- some Heliantheae with opposite leaves, aniim, a species which was demonstrated and has been reported by Carlquist (1957f) to have alternate leaves with multilacunar for Fitchia and Oparanthus. nodes, but in which rare individuals have 5. Multilacunar, verticillate leaves. Example: verticillate leaves. D. railliardioides (Fig. 11). Five traces, re­ The great diversity of nodal types in Du- lated to 5 gaps, enter each leaf at a node, bautia is of considerable interest. Despite this which usually consists of 3 leaves. As in diversity, however, a marked contrast with D. laxa and D. laevigata, increase in num­ Wilkesia occurs in regard to branching of ber of veins at the leaf base is accom­ veins in the leaf base. Wilkesia has verticils of plished by branching of the laterals. The numerous leaves, each of which is related to a occurrence of a multilacunar condition at trilacunar condition (Carlquist, 1957;/. Iig. a verticillate node is of particular signili- 18). At the leaf base, the laterals are at the

FIGS. 11-14. Reconstructions of nodes of Dubai, (based on serial sections). (11) D. raiUiardioieUs, X 6. (12) D. Menziesii, X 11.5. (13) D. laevigata, X 5.6. ( D. laxa, X 6.5. Darker lines represent veins; lighter lines represent the outlines of a nodal segment including r bases. Vegetative Anatomy — CARLQUIST 203 margins of the leaf and branches are given some cells in center of pith scleri- off toward the margin, whereas in Argyroxi- fied; carbonized resin deposits in phium the branches are given off toward the intercellular spaces). midveins. Dubautia resembles Wilkesia in this D. laevigata (no secretory canals; en­ respect. In Argyroxiphium (Carlquist, I957d, tire pith sclerined, more markedly fig. 17), however, the alternate leaves are re­ so at periphery). lated to multilacunar nodes, and adjacent 2. D. linearis (Fig. 16). Entire pith sclerified, laterals of successive leaves in the spiral are not more so at periphery than at center; united below the leaf base. intercellular spaces small; no secretory Dubautia, Argyroxiphium, and Wilkesia, then, canals; carbonized resin deposits not are amply distinct from each other in nodal observed. anatomy. The taxonomic interest of the dif­ Like D. linear:.;: ferent types within Dubautia seems evident; D. Sherfjiana, D. reticulata (carbonized the phylogenetic relationship of the various resins in intercellular spaces). types in the three genera, however, will be 3. D. latifolia (Fig. 17). Pith thin walled ex­ better understood when greater knowledge of cept at periphery, where it is prominently nodal anatomy in Compositae at large is sclerified, and in the center, where nests available. of sclereids occur; the central sclereids are much narrower in diameter than other Pith cells of the pith; intercellular spaces small; The wide range of pith types in Dubautia secretory canals absent; carbonized resin is suggested by comparison of Figures 15-18 deposits in intercellular spaces of thin- (see also Figs. 21, 22, right). As with patterns walled area of pith. This pith type appears of leaf construction, these can be referred to to be a marked specialization over the a number of types; these should not be re­ preceding types. garded as categories, however, although some 4. D. Menziesii (Fig. 18; Fig. 21, right). Pith reflect natural assemblages of species: highly lacunose, owing to large intercel­ 1. D. plantaginea (Fig. 15). Pith cells thin lular spaces; periphery consisting of small walled with small intercellular spaces; fiberlike sclereids adjacent to bundles; periphery (inner margins of bundles) scle- remainder of pith composed of large cells, rified; secretory canals present (opposite both sclerified and unsclerified; no secre­ inner faces of larger bundles); no car­ tory canals; carbonized resins abundant in bonized resin deposits present. intercellular spaces. Figure 21 shows an Like D. plantaginea: early stage in the development of this D. microcephala, D. railliardioides, D. pith, whereas pith from an old stem is hnchophylla (virtually identical with shown in Figure 18. The progressive scle- D. plantaginea). rification is evident, although some thin- D. paleata (peripheral sclerenchyma walled cells remain in the pith shown in very thick walled). Figure 18. D. scabra (secretory canals infrequent). Like D. Menziesii: D. magnifolia (periphery unsclerified). D. montana, D. strutbioloides, D. Hille- D. ternifolia (patches of sclereids pres­ brandii (virtually identical with D. ent throughout pith). Menziesii \. D. Knudsenii, D. platyphylla, D. laxa, D. arborea (fewer sclereids). D. ciliolata (secretory canals ab­ This type of pith would seem to represent sent) . a specialization over the condition indi­ D. tbyrsiflora (no secretory canals; cated for D. plantaginea. An intermediate 204 PACIFIC SCIENCE, Vol. XIII, April, 1959

FIGS. 15-20. Transections of pith. (15) D. plantaginea, X 75. (16) D. linearis, X 80. (17) D. latifolia, X 84. (18) D. Menziesii, X 78. (19) Argyroxipbium Caliginii. X 72. (20) A. sanduichense; packets of pith tracheids at right; groups of procambial cells below and left; X 90; inset shows bordered pits between two pith tracheids, X 870. Vegetative Anatomy — CARLQUIST 205

stage in specialization may be represented in the experiments of Wetmore and Sorokin by D. uaialealae. (1955). The formation of nests of sclereids 5. D. waialeaiae (Fig. 22, right). Pith entirely and pith tracheids in A. sandwichense pith, thin walled, with large intercellular spaces; however, is apparently not related to pre­ no secretory canals present; carbonized existing vascular tissue, with which the pith resin deposits in intercellular spaces. tracheids are not in continuity. The formation The pith of Argyroxiphi/nn Caiiginii (Fig. of these nests is centripetal: the central por­ L9) is highly lacunose, and carbonized resin tion of the pith was observed to contain only deposits are present. The inner margins of thin-walled cells, whereas such packets have vascular bundles are clothed with a con­ matured into sclereids and tracheids near the spicuous zone of very thick-walled hbers. The periphery of the pith. A given nest may con­ remainder of the pith, however, remains thin tain tracheids only, sclereids only, or a mix­ walled and unlignified. An identical condition ture of the two cell types. As in the experi­ occurs in A. Grayannm. Wilkesia differs only ments of Wetmore and Sorokin, no phloem in lacking carbonized resin deposits. could be detected in association with the The pith of Argyroxiphium sand wichense pith tracheids. (Fig. 20), however, shows a remarkable anom­ Although the diversity of pith types in alous feature. Basically, pith in this species is Dubautia is considerable, the pith of Argyro- identical with that of A. Caiiginii in that it is xiphium and Wilkesia does not match that of lacunose and has fibrous caps on inner faces any given species of Dubautia. The presence of the bundles. Into the large intercellular of certain features, such as carbonized resins, spaces of the cells of mature pith, however, is very similar, and pith of *)hium and new cells are proliferated by pre-existing pith Wilkesia can be considered to be within the cells. These new cells subdivide in various range of the variation pattern of Dubautia pith. planes, often forming packets oriented in a particular direction. Most of these packets Stem mature into sclerihed cells (Fig. 20, right) al­ The structure of the seem in respects other though some of them (Fig. 20, left) appar­ than nodal anatomy and pith anatomy is ently remain thin walled. Some of these scle­ relatively uniform within Dubautia, Argyro- rihed cells, although somewhat elongate and xiphiunu and Wilkesia, so that a small num­ fiberlike, have simple pits and should prob­ ber of examples may be used, and variations ably be termed sclereids. Others, as the inset may be described where they occur. In Du­ in Figure 20 shows, have bordered pits and bautia Menziesii (Fig. 21), the vascular bun­ must therefore be termed tracheids. However, dles develop prominent bundle caps of proto- because Compositae do not have tracheids in phloem fibers. The layer of cells immediately the proper sense in their xylem, these trache­ exterior to the fibers is a well-marked endo­ ids are perhaps more nearly comparable to dermis (Fig. 21, inset) in which a Casparian the "vascular tracheids'* (i.e., degenerate ves­ strip may easily be demonstrated. The endo­ sels which lack perforation plates I which do dermis accompanies the bundle cap of a leaf occut in the secondary xylem of A. sand- trace in its departute from the vascular cylin­ wichense (Carlquist, 1958). For the phenom­ der, although no endodermis occurs in leaf enon described above, the term "pith tra- base or leaf. Although stem endodermis is cheid" might be invented to accomodate uncommon in dicotyledons at large, it is their distinctive origin and mode of occur­ frequent in Compositae, as the data of Met­ rence. The formation of these pith tracheids calfe and Chalk (1950: 788) and Carlquist might be likened to the formation of tra­ (1957c 47) show. Numerous secretory canals cheids in callus, such as those which occurred are arranged adjacent to the vascular bundles. 206 PACIFIC SCIENCE, Vol. XIII, April, 1959 T Vegetative Anatomy — CARLQUIST 207

Although the secretory canals appear at first cell layers below the epidermis. As in D. glance to be arranged at random, examination Menziesii, therefore, cork formation is not of serial sections shows that they occur in limited to a single layer. The hairs which pairs beside larger bundles which, farther up thickly clothe the stem of D. waialealae are the stem, depart from the vascular cylinder as nonglandular and uniseriate in their upper leaf traces. Cortical secretory canals, then, are portions. The multicellular nature of the initiated continuously with growth in length trichome base results from elongation of epi­ of the stem, because departure of secretory dermal cells subtending the hair. The tri­ canals in association with leaf traces would chome proper, therefore, may be termed result in their total absence in upper portions uniseriate. of a stem if new ones were not initiated. In The differences between D. Menziesii and D. Menziesii, cortical secretory canals con­ D. waialealae mentioned above are not great, tinue into the leaf base for a short distance, and represent the extremes of variation ob­ although distal portions of the leaf lack them. served in the genus. The majority of the spe­ In species which characteristically have such cies have moderately lacunate cortex, inter­ canals in the lamina, such as D. Knudsenii mediate between these two types. Although (Fig. 11, these canals continue from the stem secretory canals were found to be scarce or into the leal and either ramify or (particularly lacking in the stems of D. waialealae ami /). in die case of canals adaxial to veins) originate pLitypbylla examined, they are present to some de novo. As the portion of a cleared leaf of D. extent in the remaining species. lonchophylla in Figure 6 shows, these canals mainly parallel the larger longitudinally ori­ In the aspects of stem structure mentioned ented veins. The cortex of D. Menziesii ex­ above, Argyroxiphini// and Wilkesia are refera­ terior to the secretory canals is highly lacu- ble to exactly the same pattern. In Argyro- nate. In the stem shown in Figure 21, cork xiphium, carbonized resin deposits, like those formation is beginning. Periclinal divisions of the pith, occur in intercellular spaces in the may be seen both in the epidermis and in the cortex, as well as between fibers in bundle layer immediately beneath the epidermis. Cork caps. Argyroxiphi/im Caliginii is the only spe­ formation, then is not restricted to a single cies of its genus in which secretory canals are layer of the stem. At extreme left in Figure 21, present in the leaves (Carlquist, I957d): secre­ two degenerating biseriate glandular hairs tory canals accompanying leaf traces continue may be seen. into the leaves. In other species, they termi­ nate below the level at which leaf traces enter The stem of D. waialealae (Fig. 22) illus­ leaf bases. In Wilkesia, which also lacks foliar trates several points of contrast. Only small secretory canals, many of the leaf traces are lacunae are present in the cortex (gaps in that not accompanied by secretory canals, al­ region in Fig. 22 probably resulted from the though canals are abundant around the vas­ fact that a dried specimen was used for this cular cylinder proper. Secretory canals in preparation). Secretory canals are infrequent, Wilkesia, for the most part, terminate below although a small canal may be seen below the separation of leaf traces from the cylinder. large bundle cap, lower right. Cork formation Abundant cork formation from subepider­ is not initiated in the epidermis, which is al­ mal layers is characteristic of Argyroxipbium most entirely involved in the formarion of Caliginii and A. Grayanum, whereas little trichomes, but may be seen in the two or three cork occurs in the rosette species A. sand-

FIGS. 21-23. Transections of stems of Dubautia. (21) D. Menziesii: epidermis at left; X 98; i nset shows endo- dermis with Casparian strip (line running through cells above fibers), X 350. (22) D. waialealae, X 115. (23) D. platypbyUa; stem showing secondary growth; cork above, secondary phloem below; X 107; inset shows an area of lacunate cortex, enlarged, in which pectic warts are present on cell walls, X 218. 208 PACIFIC SCIENCE, Vol. XIII. April, 1959

wichense and A. virescens. In Wilkesia, the xiphiitm sand wichense (Carlquist, 1957;/). layer beneath the epidermis is differentiated Root into a discrete hypodermis consisting of thick Roots of the genus Dubautia were not walled sclereids, and cork formation is initi­ studied. In Argyroxiphiitm and Wilkesia, which ated in a single layer immediately interior to the writer was able to grow from seed he col­ the hypodermis. lected in the Hawaiian Islands, anatomical The stems of both Argyroxiphiitm and studies were made. Adventitious roots of A. Wilkesia lack endodermis demarcated by Cas- Caliginii were preserved in the field. Roots in parian strips, although the formation of tan­ these genera show typical dicot root struc­ gential sheets of cytoplasm in the layer of ture. In A. sandwichense, roots varied from cells immediately exterior to bundle caps diarch to pentarch, depending on the diam­ indicates that this layer of cells may be some­ eter of the root. The primary root of the seed­ what differentiated as an endodermis. ling, and all of the finer roots of a mature Biseriate glandular hairs with capitate mul­ plant were found to be diarch. In the ad­ ticellular heads are characteristically present ventitious roots of A. Caliginii, which have a on the stems of both Argyroxiphiitm and relatively larger diameter, five to seven xylem Wilkesia. These trichomes are of the same poles were observed. A variation of two to type as those illustrated by Carlquist (1957

Fitchia. The distribution of secretory canals in in the application of anatomical characters to vegetative parts or Dubautia, Argyroxiphium, taxonomy of the genera discussed, descrip­ and Wilkesia follows very closely the diagram tion of relationships among the three genera, given by Carlquist (1957c 58) for F. speciosa, and in determination of the relationships of differing only in those species which do not these seemingly isolated genera. Within the characteristically have secretory canals in pith, genus Dubautia, characters of vegetative anat­ cortex, or leaves. The occurrence of secretory omy seem to offer good criteria for taxonomic canals in inflorescence structures of Dubautia, decisions, although inflorescence anatomy Argyroxiphium, and Wilkesia will be consid­ cannot be overlooked. ered in subsequent studies. The results of the present study, and one reported separately (Carlquist, 195HJ indicate Trichomes: Summary that Keck (1936: 24) is justified in uniting Although occurrence and types of tri­ the species of Railliardia with Dubautia. No chomes have been mentioned above, the basic anatomical character, or combination of such pattern underlying trichome occurrence needs characters, seems to support the division of to be stated. In the genera Dubautia, Argfro­ the species into the two genera as they have xiphium, and Wilkesia, two types of trichomes been traditionally delineated. For example, D. are present: uniseriate nonglandular and bi- waialealae (invariably treated as a Dubautia) seriate glandular. Examination of shoot apices seems closer in node and leaf anatomy to D. indicates that these two types may be formed Menziesii (included under Railliardia by au­ initially, but are variously matured or de­ thors recognizing that genus) than to Du­ generate upon maturation ot organs of the bautia plantaginea (type species of the genus) plant. In Dubautia railliardioides, for example, or its close relatives (D. microcephala, D. both uniseriate hairs and biseriate glandular railliardioides). trichomes with subdivided multicellular tips Rather, the writer believes that a picture of may be seen on young leaf primordia. On species groups emerges, although these can­ mature portions of the plant, both types of not yet be defined with sufficient precision to hairs are absent except for retention of uni­ permit division ot the genus into subgenera seriate hairs in leaf axils. Thus, differences in on the basis of the characters described here. indument ot the mature plant are the result of Attention may be called, however, to certain elimination or retention of these two basic alliances which seem quite certain: types. The leaves of D. platyphylla furnish an (1) D. Menziesii, D. arborea, D. Ilillebrandii, example of marked development of the bi­ D. montana, and D. struthioloides agree in seriate hairs with subdivided pits, whereas leaf, pith, and wood anatomy. leaves and stem of D. waialealae demonstrate conspicuously the uniseriate nonglandular (2) The species with narrow strap-shaped hairs with no evidence, in the mature plant, leaves and wide leaf bases, D. plantaginea, of the biseriate type. D. microcephala, D. railliardioides, D. pale- ata, D. magnifolia, and D. ternifolia, seem The precise nature and distribution of tri­ close in leaf anatomy, presence of secre­ chomes in Compositae is inadequately known. tory canals in pith, and, to a certain extent. The occurrence of the capitate glandular hairs in nodal anatomy. D. Knudsenii appears mentioned above, however, appears signifi­ close to this assemblage on the basis of its cant in connection with relations of the three foliar secretory canals and its pith structure. genera to certain non-Hawaiian groups. (3.) A grouping of species including D. cilia- lata, D. lonchophylla, D. scabra, D. Sherff- DISCUSSION ana, and D. thyrsiflora seems justifiable on The importance of the data developed lies the basis of near identity in leaf structure, 210 PACIFIC SCIENCE, Vol. XIII, April, 1959

particularly in regard to the single secre­ peromannia. Pacific Sci. 11: 207-215. tory canal on the xylem lace of veins. 1957^. Wood anatomy of Mutisieae Dubautia waialealae and D. Litifolia show pe­ (Compositae). Trip. Woods 106: 29 o. culiarities of anatomy not closely matched by — 1957f. The genus Fitchia (Composi­ other species, and are probably rather iso­ tae). Univ. Calif. Publ. Bot. 29: 1-144. lated from whatever species each of them 1957;/. Leaf anatomy and ontogeny in may most closely resemble. Each of these two Argyroxipbium and Wilkesia (Compositae). species should probably form a "group" by Amer. J. Bot. 44: 696-705. itselt. Inclusion or species other than those 1958. Wood anatomy of Heliantheae mentioned does not seem to be justified on (Compositae). Trop. Woods 108: 1-30. the basis of present anatomical information. HOFFMANN, O. 1890. Compositae. IN Engler The material for Argyroxiphiuw supports and Prantl, Die natiirlicben Pflanzenfamilten the interspecific relationships in that genus 4(5): 87-402. Wilhelm Engelmann. Leipzig. suggested earlier by the writer (1957*/). KECK, DAVID D. 1936. The Hawaiian Silver­ The suggestion of Keck (1936: 10) that 's. Occ. Pap. Bishop Mus. Vol. 11, Dubautia, Argyroxipbinm, and Wilkesia must No. 19. 38 pp. be considered a group within themselves METCALFE, C. R., and L. CHALK. 1950. Anat- could be defended on the basis ot vegetative tomy of the Dicotyledons. Vol. 2. Clarendon anatomy. The presence and distribution of Press, Oxford. trichome types, the patterns and nature of ST. JOHN, HAROLD. 1950. The subgenera of secretory canals, the lacunate character of the Dubautia (Compositae): Hawaiian plant cortex, and the presence ot carbonized resins studies 18. Pacific Sci. 4(4): 339-345. in intercellular spaces are notable in this re­ SHERFF, EARL E. 1935. Revision of Tetra- gard. In fact, the three genera could not be molopium, Lipochaeta, Dubautia, and Railli- separated on the basis of some anatomical ardia. Bull. Bishop Mus. 135: 1-136. characters. Other structural features offer ex­ SKOTTSBERG, CARL. 1931. Remarks on the cellent criteria, however, and Dubautia, At- flora of the high Hawaiian volcanoes. gyroxiphium, and Wilkesia are amply distinct Medd. Gbteborgs Bot. Tradg. 6: 47-65. in respect to leaf and node anatomy. 1956. Derivation of the flora and Because there is little published information fauna of Juan Fernandez and Easter Island. on anatomy of genera which are putatively Nat. Hist. Juan Fernandez and Easter Island related to these three genera, no comment can 1: 193-439- be made in this regard at present. The writer TETLEY, URSULA. 1925. The secretory system hopes to demonstrate, however, that charac­ of the roots of the Compositae. New ters described here are worthy of considera­ Pbytol. 24: 138-162. tion in comparing these genera with such taxa URBAN, I., and M. MOBIUS. 1884. Ueber as the American genera of Madinae, Railfi- Scblectendalia luzulaefolia Less., eine Mono- ardella, and certain Senecioneae. kotylen-ahnliche Composite und Eryngium eriopborum Cham., eine grass-blattrige Um- REFERENCES bellifere. Bet: Dtscb. Bot. Ges. 2: 100-107. BENTHAM, GEORGE, and JOSEPH D. HOOKER. WETMORE, RALPH H., and SERGEI SOROKIN. 1873- Genera Plantarum. Vol. 2. Lovell 1955. On rhe differentiation of xylem. J. Reeve and Co., London. Arnold Arbor. 36: 305-317. CARLQUIST, SHERWIN. 1956. On the occur­ WILLIAMS, BERT C. 1954. Observations on rence of pectic warts in Compositae. Amer. intercellular canals in root tips, with special J. Bot. 43: 425-429. reference to Compositae. Amer. J. Bot. — 1957^. Systematic anatomy of Hes- 41: 104-106.