Leaflet anatomy verifies relationships within Syagrus (Arecaceae) and aids in identification

Abstract The current investigation was carried out to examine how palm anatomy may coincide with the current molecular analysis including the three recognized clades of Syagrus Mart. and to justify the splitting of acaulescent Syagrus species (e.g. Syagrus petraea (Mart.) Becc.) into several species. Free-hand cross-sections of leaflets were made and the comparison of these verifies the relationships suggested by the molecular data. Free-hand leaflet sections were also found to be useful in the identification of otherwise difficult-to-identify acaulescent Syagrus species. The result and conclusion is that anatomical data is valuable in helping to verify molecular data and that splitting the acaulescent species of Syagrus is justified by the differences discovered in their field habit and anatomy. These differences were used to produce an identification key that is based on the anatomy.


Introduction
Syagrus is part of the largest subfamily of palms, Arecoideae (Dransfi eld et al. 2005(Dransfi eld et al. , 2008Gunn 2004). Recent construction of a supertree using a supermatrix consisting of DNA sequence data, plastid restriction fragment length polymorphism data and morphological data further supports that Syagrus belongs to the Cocoseae (Baker et al. 2009). Within the Arecoideae there is strong support via the phylogenetic analyses of the low copy genes PRK and RPB2 that it belongs to the RRC (Roystoneeae, Re-inhardtieae, Cocoseae) clade and specifi cally to the Cocoseae and Attaleinae (Baker et al. 2011). Attaleinae include genera like Allagoptera, Attalea, Beccariophoenix, Butia, Cocos, Jubaea, Jubaeopsis, Lytocaryum, Parjubaea, Syagrus, and Voanioala or in other words all of the non-spiny palms with small hard coconut-like fruits.
Palm leafl et anatomy has been useful in identifi cation and has been used to suggest systematic relationships. Tomlinson (1961) examined and described the leafl et anatomy of some 250 species of palms in 137 genera and suggested some systematic relationships among genera. Horn et al. (2009) took it a step farther and mapped out the lamina anatomy using the data matrix for the palm family resulting from a phylogenetic analysis by Asmussen et al. (2006) in order to understand the lamina anatomy evolution. Tomlinson's "brief survey" (Tomlinson 1961) inspired Glassman (1972Glassman ( , 1987 to examine Syagrus and its closely related genera in greater detail. Glassman (1972) emphasized that his survey of the genus was "based on mostly one collection for each taxon." However by the time he completed his revision (Glassman 1987), slides of two or more specimens were made for most taxa. His key was written as a convenient tool for identifi cation and was not intended to show close relationships. Meerow et al. (2009) showed how leafl et anatomy further supported the molecular relationships between Allagoptera, Parajubaea and Polyandrococos (now synonym of Allagoptera). Tomlinson et al. (2011) expanded his original work and presented information on 183 palm genera (out of 185 now recognized) and suggested relationships based on anatomy and the use of modern phylogenetic approaches. Glassman (1972) emphasized the following anatomical characters: (1) nature of adaxial and abaxial surfaces (straight, wavy or furrowed), (2) number of cell layers comprising both adaxial and abaxial hypodermis, (3) relative frequency and location of larger veins and whether they bulge at the laminar surfaces, (4) frequency and location of intermediate and minor veins (whether they are abaxial, in the middle or both) and whether they are attached to nonvascular fi bers (5) relative abundance, shape and location (adaxial and abaxial, or adaxial only) of clusters of nonvascular fi bers (N.V.F.), called fi ber bundles by Tomlinson et al. (2011), (6) relative frequency and location of stomata (abaxial surface only or both leaf surfaces) and whether they are sunken or not, (7) size of midrib and its shape in transection (truncate, rounded or pointed), (8) size and shape of expansion cell tissue (E.C.T.), (9) relative frequency of dark staining bodies (probably groups of cells containing tannins) in mesophyll and midrib, (10) number and relative size of fi ber clusters and veins at extremities of pinna. Tomlinson et al. (2011) suggested that the following anatomical features vary in Syagrus: (1) abundance and distribution of trichomes; (2) adaxial epidermal cells varying from rectangular, with distinctly sinuous anticlinal walls to rhombohedral, obviously extended cells with straight or at most undulate walls; (3) the thickness of the outer epidermal wall, varying from very thick with the cell lumen scarcely one tenth its depth, to thinner walls only 2-3 times thicker than the remaining walls, the cell lumen large; (4) the abundance and location of adaxial non-vascular fi bers, varying from an almost continuous layer within the hypodermis to few fi bers; (5) the extent and location of abaxial non-vascular fi bers; (6) the extent to which the minor abaxial veins are in contact with the abaxial hypodermis; (7) the degree to which the inner sheath of major veins develop fi brous extensions to the upper surface layers.
Th e fi rst part of this paper investigates leafl et anatomy to see how it coincides and possibly even verifi es the relationships supported by the molecular analysis of Meerow et al. (2009). Leafl et anatomy was useful in producing valuable anatomical characters for a phylogenetic analysis of the Attalinae palms based purely on morphological and anatomical characters (Noblick et al. 2013). Some of these anatomical characters were found to coincide with the molecular analysis of the Attaleinae (Meerow et al. 2009) as previously mentioned above. Using seven low copy nuclear WRKY genes, Meerow et al. (2009) inferred that the coconut (Cocos nucifera) was sister to new world Syagrus species and that the genus Syagrus was composed of at least three distinct clades: Rain Forest, Eastern Brazilian and Cluster-stem (Fig. 1). Th e three clades were examined anatomically to see if their anatomy supported these relationships.
Th e second part of this paper deals with the problem of identifying the "acaulescent" species of Syagrus. Most of these palms grow in Brazilian savannas ( cerrados) and high altitude rocky fi elds (campo rupestre). Many species of Syagrus are described as acaulescent and Tomlinson et al. (2011) admitted that based on the low number of collections, especially of Syagrus, there was a great need for more detailed studies within this group. In fact, about 25 out the 54 currently recognized species of Syagrus are acaulescent or short stemmed (Table 1) and their identifi cation remains challenging with many having formerly been dismissed as Syagrus petraea . Palms that do not have visible above ground stems are often referred to as acaulescent meaning "without a stem." In fact all palms have a stem, whether it is suberect, short-stemmed with the crown remaining at ground level, persistent juvenile which remains at ground level or rhizomatous stems that remain horizontal at ground level (Tomlinson et al. 2011 and similar looking fl owers and fruits. Many herbarium specimens of these acaulescent Syagrus look alike, but in the fi eld they display characters that are not well preserved on herbarium sheets or that are not reported on the labels (Table 1), like defl exed pinnae (leafl ets that are bent or turned abruptly or sharply downwards) or petioles that bend strongly downwards causing the leaf to lie parallel or fl at on the ground (prostrated) as opposed to having leaves that are strongly ascending with straight petioles. Th e second part of this paper will therefore confi rm, add to and hopefully correct some of the work already started by Glassman (1972Glassman ( , 1987. Not counting the two acaulescent Butia that Glassman (1987) included in the Syagrus portion of the leafl et anatomical key for his revision, Glassman only includes about half or 12 of the 26 currently recognized acaulescent Syagrus taxa. In addition, he misidentifi ed some of the specimens used in that key. For his anatomical studies, he sometimes neglected to use specimens collected in the vicinity of their holotypes. Syagrus  6M) whose Bolivian holotype is now believed to be a local endemic is just such an example. His anatomical leafl et drawing of Glaziou 22254 (Glassman 1987) from central Brazil in the state of Goiás matches S. glazioviana (Dammer) Becc. anatomy (Fig. 7E) more than it does the Bolivian S. petraea (Fig. 6M). Also instead of Glassman's simple drawings of the middle portion of the lamina, I believe the images of leafl et marginal cross-sections to be more useful and more informative. Th erefore, the second intent of this paper is to facilitate the identifi cation of these diffi cult acaulescent Syagrus species by (1) making use of poorly recorded fi eld characters ( Table 1) that are not easily interpreted from a fl attened dried specimen and (2) their leafl et anatomy ( Table 2) making use of digital images of the cross-sections of leafl et margins. Major veins attached to the adaxial hypodermis but separated from the abaxial hypodermis by an additional cell layer or layers X X X X X X X X X X X X X X X X X X X X X X Major veins mostly attached to the adaxial surface often by a short or long fi brous sheath extension (girder) X X X X X X X Minor veins on both surfaces often nearly pairing up acoss from each other X X X X S Minor veins near the middle of the mesophyll or slightly below it (closer to the middle than to the abaxial) X X X S S X X X X X Many minor veins adjacent to the abaxial surface or in the lower third of the lamina (closer to the abaxial surface than to the middle) X X X X X X X X X X X S X X X X X X

Veins with exaggerated fi berous sheath
Vein with large exaggerated fi brous sheath running along the margin S S X X X X X X X One minor vein with exaggerated fi brous sheath running along the margin Two or more minor veins with exaggerated fi brous sheath running along the margin X X X S X X X X

Fiber bundles
Major fi ber bundle running along or very near the leafl et margin X S X S First or second fi ber bundle on adaxial surface larger than rest X X X X X S X X X X X S Most large adaxial fi ber bundles reach ca. 1/3 to 1/2 across the mesophyll X X X X X X X X X X X X X X X X X X Most small adaxial fi ber bundles reach 1/5 to 1/4 across the mesophyll X X X Adaxial fi ber bundles mostly long and thick X X X X X X X X X Adaxial fi ber bundles mostly short and thick X X X X X X S Minor fi ber bundles present among or between larger adaxial fi ber bundles and veins S X X X X X X X X X S X X X X X X X X Minor fi bers or fi ber bundles scattered in the mesophyll X X X Minor fi ber bundles along the abaxial surface absent X X X X X X X X X S X Few minor fi ber bundles along the abaxial surface X S X X X S Many minor fi ber bundles adjacent to the abaxial surface X S X X X S X X X X X X Abaxial minor fi ber bundles occasionally alternating with abaxial minor veins X S X X X S Sometimes minor fi ber bundle adjacent to the margin X X X X X X X X X X X X X X X X X X Th ick-walled fi ber-like hypodermal cells protecting the margin X

Plants examined
Both live material and preserved herbarium material (Table 3) were used in this project. Th e living material used in this study came from the collections at Montgomery Botanical Center (MBC, Miami, FL). Th e dried material was from collections made while doing fi eldwork in Brazil, from the garden and herbarium at Jardim Instituto Plantarum (HPL, Novo Odessa, São Paulo, Brazil) and from dried collections at the Fairchild Tropical Botanic Garden herbarium (FTG, Miami, FL) and a few specimens from the following herbaria: G, IBGE, IPA, K, MO, NY and US.

Anatomical preparation
Two methods were employed for expedient identifi cation. First, one side of the middle section of a middle leafl et was folded back and forth on itself in accordion fashion; the folded leafl et was then held down on a cutting board, while using a double-edged razor blade to cut thin cross-sections. Th e sections were rinsed into a watch glass with water and a thin brush was used to select the thinnest sections under a dissecting scope and then placed on a microscope slide in a droplet of 1:1 glycerin/water solution. A cover glass was placed over the specimen and the slide was placed under a compound light microscope and photographed under the 10× objective (100× magnifi cation). Most of the sections were unstained, but in rare cases toluline blue (0.01%) was tested to see if it made it easier to view certain characters, which it did not (Fig. 3C, 6D).
In the second method, better suited when material is limited, a small square of carrot of the appropriate size is cut to fi t in an inexpensive hand-held student microtome. I purchase my hometrainingtools hand-held microtome online. A vertical slit is cut in the carrot and a small piece of leafl et is inserted in the appropriate orientation. Th e carrot is clamped into the hand-held microtome. Th e microtome is screwed to the appropriate level and an ordinary folding straight edge razor, the kind used for shaving, is utilized to cut the cross-sections and honed occasionally to keep it sharp. Sections are handled the same way as above. Scale was later added using a stage micrometer. Dried material can also be sectioned and photographed after rehydrating in a 5% solution of Contrad 70® (Decon Labs, King of Prussia, Pennsylvania) for a period of 24 hours (Tomlinson et al. 2011).

Characters utilized
Th is paper is focused mainly on characters of the more easily sectioned marginal and laminal portions of the leafl et and not so much on the harder to section midrib. Trichomes, epidermis and dark staining bodies were also not looked at.   Pink arrows = minor veins; Black arrows = vein with an exaggerated fi brous sheath; Yellow arrows = major fi ber bundles; Red arrows = minor fi ber bundles; Green arrows = Cuticle; Orange arrows = epidermis; Purple arrows = hypodermis; White star = mesophyll A S. allagopteroides illustrates a large marginal vein with an exaggerated fi brous sheath (black), an unattached major vein (white), the presence of minor veins on both the adaxial and abaxial surfaces (pink) and an occasional minor adaxial fi ber bundle (red) B S. caerulescens illustrates a large marginal fi ber bundle (yellow), a major vein attached to the adaxial surface by a fi brous extension (white), a small vein with an exaggerated fi brous sheath (black), minor veins (pink) sometimes alternating with minor fi ber bundles (red) along the abaxial surface, and adaxial fi ber bundles reaching nearly ½ the distance across the mesopyll (white star) C S. vagans illustrates the fi rst or second fi ber bundle as being the largest along the adaxil surface (yellow) and minor fi ber bundles (red) scattered throughout the mesophyll (white star), minor veins located near the middle or just slightly below D S. gouveiana illustrates the cuticle (green), epidermis (orange), hypodermis (purple) E S. harleyi illustrates a protective layer of thick-walled hypodermal cells (purple) on the margin, which is characteristic of this species. A, B, C Scale = 0.1 mm; D, E scale = 0.2 mm.
Characters examined during this study follow some of Glassman's 4, 5 and 10 characters listed above and Tomlinson's characters 4, 5, 6, and 7 listed above. Figure  2 will clarify much of the terminology and characters used in this paper. In each leaf cross-section the upper or superior side of the lamina is called the adaxial surface, meaning "towards the axis", since this side of the leaf faces towards the axis or center of the plant as it grows out. Th e lower or inferior side is called the abaxial, meaning "away from the axis", since this side faces away from the center of the plant (Dransfi eld et al. 2008, Esau 1977. Th e outer most layer of the leaf is the cuticle (Fig. 2D, 2E green arrow), a non cellular waxy layer produced by epidermis (Dransfi eld et al. 2008). Th e cuticle is followed by the epidermis, "outer skin" (Fig. 2D, 2E orange arrows), followed by the hypodermis, "under skin" (Fig. 2D, 2E purple arrows), which is fi nally followed by the mesophyll, "middle leaf", region (Fig. 2 white stars). Within the mesophyll are vascular bundles, or fi brovascular bundles or veins of various sizes (Tomlinson et al. 2011) that will be referred to as major veins (Fig. 2 white arrows), intermediate veins ( Fig. 2 blue arrows), and minor veins ( Fig. 2 pink arrows). Some major and intermediate veins are often attached to the adaxial hypodermis and sometimes to both the surfaces by fi brous sheath extensions. If the attachment extends to both surfaces via a fi brous sheath extension, the vein becomes girder-like and is indeed referred to as a girder (Tomlinson et al. 2011) (Fig. 5A). In some veins the fi brous sheath becomes so enlarged with fi bers that such veins are referred to as veins with exaggerated fi ber sheaths (Tomlinson et al. 2011) (Fig. 2 black arrows). In addition to the veins, the laminal tissues are supported by nonvascular fi bers or fi ber bundles of various sizes. Some have major fi ber bundles adjacent to or near their margins (Fig. 2 yellow arrows). Many fi ber bundles are adaxial and may reach close to ½ the distance across the mesophyll (Fig. 2B, 2D). Minor, intermediate and major fi ber bundles can be found adaxially ( Fig. 2A red arrows; 2B, 2C yellow arrows). Most minor fi ber bundles are mainly abundant abaxially (Fig. 2B, 2C red arrows) and occasionally scattered throughout the mesophyll (Fig. 2C red arrows).
To keep things simple for fi eld examination, the following qualitative characters were examined: (1) location, attachment or lack of attachment of the major veins to one or both surfaces and method of attachment (fi ber sheath extension or not); (2) location, attachment or unattachment of intermediate veins to one or both surfaces and method of attachment (e.g. sheath extensions, formation of girders); (3) location of the minor veins (e.g. adaxial, abaxial, abaxial and adaxial, middle, marginal); (4) presence, size and location of veins with an exaggerated fi brous sheath (large ones often located on the leafl et margin); (5) presence, location, size and sometimes cross-sectional shape of fi ber bundles and the extent they reach across the mesophyll. Th ese characters can also be further summarized as follows: (1) Major vein location {adjacent to the margin; near the margin but not adjacent to it (this means that along a horizontal plane there is a maximum of one minor vein or one fi ber bundle separating it from the actual margin); not adjacent to nor near the margin} (2 Th e key was designed for fi eld use, which means minimal equipment, no staining, and low magnifi cation and the use of simple characters. Refer to the characters in the methods for clarifi cation of terminology. By using the methods listed above and following many of the simple techniques mentioned by Tomlinson et al. (2011), rapid results can be achieved in a laboratory provided with only the simplest equipment. Th is simple approach was also successfully used in a signifi cant study of palm leaf development by Nowak et al. (2007).

Results
Anatomical characters observed on marginal palm leafl et cross-sections have been found to verify the Rain Forest and Eastern Brazilian clades and to some extent the Cluster-stem clade found within Syagrus (Fig. 1). In the Rain Forest clade, there is a continuous hypodermal layer of fi brous to thick-walled cells, one to two layers thick just below the adaxial epidermis (Fig. 3A, 3B, 3C, 3D). In the Eastern Brazilian clade, there are many thick, closely-spaced, multicellular fi ber bundles running along the adaxial surface of the leafl et (Fig. 4A, 4B, 4C, 4D). Finally the Clusterstem clade is usually characterized by minor sparsely spaced fi ber bundles on the adaxial side and minor veins adjacent to the abaxial surface (Fig. 5B, 5C) or with minor veins on both surfaces (Fig. 5D) that make the anatomy of S. macrocarpa Barb. Rodr. S. fl exuosa (Mart.) Becc. and S. cerqueirana look interestingly similar to one another.
After examining many leafl et hand sections of various acaulescent palm specimens, it was discovered that many had very diff erent leafl et anatomy. Th e presence and absence of the anatomical characters in all 25 species is recorded in Table 2. Useful anatomical characters were found to separate the 25 known species of acaulescent and short-stemmed Syagrus and an identifi cation key was developed. Several acaulescent Syagrus specimens frequently identifi ed as Syagrus petraea were found to have distinctive fi eld habits and leafl et anatomies (Table 1 and 2).

Distinguishing Major Clades
Species of the Rain Forest clade (Fig. 1), which includes many Amazonian species, are distinguished anatomically by an almost continuous adaxial fi brous layer, one or a few cells thick just under the epidermis (the hypodermal layer) (Fig. 3A, 3B, 3C, 3D). I speculate that perhaps this nearly continuous fi brous layer strengthens the lamina while maintaining its fl exibility (Vincent 1982), helps the leaf shed water and discourages fungus infection. Species of the Eastern Brazilian clade (Fig. 1) are distinguished by thicker and stiff er leafl ets reinforced by many adaxial, thick, multicellular fi ber bundles along the adaxial side of the leaf and these fi ber bundles may extend as far as ½ the distance across the mesophyll (Fig. 4A, 4B, 4C, 4D). Fibers assume much of the load-bearing capacity of the lamina (Horn et al. 2009, Vincent 1982. Th ese fi bers and fi ber bundles help the leafl et to retain its shape, fl exibility and form when leaf turgidity wanes during the periodic dry spells that frequent the seasonally dry forests, cerrados and caatingas of Eastern Brazil. Th e Cluster-stem clade (perhaps a misnomer, since not all are cluster-stemmed) is not as clear. Syagrus campylospatha (Figs. 5A, 6A) appears anatomically diff erent from the rest of the group having girder type intermediate veins that are attached to both sides of the leaf by fi brous sheath extensions. Syagrus fl exuosa and S. macrocarpa have nearly identical anatomies with minor, sparsely spaced fi ber bundles running along the adaxial surface (Fig. 5B, 5C) and S. cerqueirana (formerly identifi ed as S. petraea) has a similar aspect but replaces these adaxial fi ber bundles with minor veins, which are found on both sides of the leaf (Fig. 5D).
Most acaulescent Syagrus exhibit the Eastern Brazilian pattern (e.g. S. gouveiana; Fig. 7A) with the large, multicellular fi ber bundles running along the adaxial side of the leafl et and the Cluster-stem pattern, similar to that of S. cerqueirana (Fig. 5D, 6D, 6H), with minor veins on both surfaces (e.g. S. lilliputiana, Fig. 6G), each attached to either the adaxial or abaxial surface by short, fi brous extensions. Since most acaulescent palms grow in seasonally dry areas (cerrados) that require stiff er leafl ets, it is perhaps understandable why the Rain Forest pattern is not seen among them.

Acaulescent species
Some of the problems of identifying acaulescent Syagrus species were covered previously in the introduction concerning the lack of good label information in relation to how leaves and leafl ets are displayed or arranged on the plant before pressing and drying. Having observed most of these variations personally in the fi eld has led me to the challenging process of trying to straighten out this much neglected complex of species. For me, it started in Bahia, Brazil with the misidentifi cation of the acaulescent cerrado palm, S. glazioviana. Many palm taxonomists, including Glassman and myself (Noblick 1991), have erroneously identifi ed it as S. petraea. Initially, Glassman (1965) placed it in synonomy with S. petraea, based on the shape and size of their female fl owers. A few years later, he considered them distinct species (Glassman 1968) after he had seen the lectotype for S. glazioviana due to diff erences in the width of the leafl ets and shape of the leafl et tips (symmetrical verses asymmetrical). Finally in his revision (Glassman 1987), he synonomized it once again with S. petraea, concluding that the diff erences seen must have been due to favorable versus unfavorable growing conditions. In western Bahia it is often used to make brooms. It has meter-long leaves with long, regular to loosely clustered pinnae and a spike infl orescence. In the same cerrados, one will encounter another acaulescent, spicate palm with shorter leaves and tightly clustered pinnae that are unsuitable for broom making, which I had previously identifi ed as S. petraea as well. I currently believe that the smaller western Bahian one is S. loefgrenii, which has also been proposed as a synonym of S. petraea (Henderson et al. 1995), but the leafl et anatomy of S. petraea (Fig. 6M) is very diff erent from the anatomy of both S. loefgrenii (Fig. 6I) and S. glazioviana (Fig. 7E). It has been an unfortunate fact that most acaulescent, spicate Syagrus have gotten automatically classifi ed as S. petraea. In truth, acaulescent palms with a sparsely branched infl orescence have also been classifi ed as S. petraea (i.e. S. cerqueirana from Paraguay). However Syagrus petraea is not the only problematic acaulescent Syagrus. In his anatomy paper (Glassman 1972), Glassman misidentifi ed S. cerqueirana (Swallen 9520 from Mato Grosso) as S. graminifolia. He then proceeded to use the anatomy of that misidentifi ed specimen to represent S. graminifolia in his publications (Glassman 1972(Glassman , 1987. When I examined the anatomy of collections of S. graminifolia from Goiás I discovered that they had a very diff erent anatomy from what Glassman had published, but I resolved the issue by comparing the anatomy of my specimens with that of a leafl et fragment borrowed from the original 1827 holotype (Burchell 5956) and found them to be a match. In addition, Glassman mistakenly reported Burchell's collection from the state of Piauí, but Burchell's fi eld notes and itinerary (Smith and Smith 1967) clearly place him in southern Goiás at the time.
Many of the S. petraea-types have very diff erent leafl et anatomies. Th eir visible fi eld characters (Table 1) and their distinctive anatomy has justifi ed splitting up the complex Lorenzi et al. 2010) by resurrecting formerly synonymized names (e.g. Syagrus glazioviana, S. loefgrenii) and by describing several new species (e.g. S. allagopteroides, S, angustifolia, S. caerulescens, S. cerqueirana, S. evansiana, S. gouveiana, S. itacambirana, S. minor, S. pleiocladoides, S. procumbens, S. rupicola). Currently, there are about 26 taxa of Syagrus without visible above ground stems or with very short stems and it is strongly suspected that there will be several more based on the anatomy that has so far been observed. As a disclaimer, I feel that this key is still not the fi nal word and will need further revision as new species are discovered and others get reworked. Th e same species grown under diff erent growing conditions or adult and juvenile forms may look slightly diff erent, but the anatomy maybe an important tool in helping us to resolve these issues. I suspect that some species which appear to be morphologically diff erent but anatomically similar may turn out to be the same species (e.g. S. allagopteroides and S. minor are suspicious).
In conclusion, leafl et anatomy has been found to be useful in helping to confi rm or verify relationships discovered through the molecular analysis and in identifying some of the diffi cult acaulescent Syagrus species.
Anatomical key to acaulescent Syagrus Margin with a huge fully functional major vein with a somewhat exaggerated fi brous sheath at or near the margin (Fig. 6C)  Minor veins adjacent to both the adaxial and abaxial surface (Fig. 6D-H A minor to intermediate vein with an exaggerated fi brous sheath adjacent to the margin and occupying less than half of the margin (Fig. 6D) ... S. cerqueirana -A major vein with an exaggerated fi brous sheath adjacent to the margin and occupying more than half to nearly the entire margin ( Fig. 2A)  Marginal vein with exaggerated fi brous sheath occupies over half of the margin but not the entire margin (Fig. 6E)  Major vein usually unattached separated from the hypodermis by another cell layer or two ( Fig. 2A, 6F)  Minor fi ber bundles along the adaxial nearly round in shape (Fig. 6G)  Most large adaxial fi ber bundles reaching less than 1/4 to 1/5 across the mesophyll (Fig. 6I)  Large major vein near the margin but not adjacent to it (Fig. 6N)  Adaxial fi ber bundles long and skinny and reaching less than 1/5 to ¼ across the mesophyll (Fig. 6M)  Major veins are usually attached adaxially (Fig. 2D, 7A)