We reconstructed phylogenies of Acereae at the section-level using sequences of nuclear Internal Transcribed Spacer (ITS) and the chloroplast spacer trnD-psbM (hereafter, psbM). We selected these markers because of their utility as DNA barcodes in plants (Dong et al. 2012; Li et al. 2011; Shaw et al. 2005; Zuo et al. 2011, 2017), their demonstrated utility in Acer for the sections in question (Grimm et al. 2006; Renner et al. 2008), and our preliminary observations about the information content of psbM for Acereae. We reconstructed phylogenies for this study even though prior studies have generated phylogenies of Acereae using chloroplast DNA, ITS, and nuclear genes (Suh et al. 2000; Tian et al. 2002; Grimm et al. 2006; Buerki et al. 2010; Renner et al. 2008; Harris et al. 2017), because doing so gave us more control over sampling of representative species, the ability to curate data and make decisions about data quality, and firsthand knowledge of all analysis parameters.

We obtained sequences of psbM and ITS from GenBank for representative samples of sections of AcersensuWolfe and Tanai (1987), A. laurinum, and all species of sect. Rubrasensuvan Gelderen et al. (1994). We used Wolfe and Tanai (1987) to guide our taxonomic sampling, because their treatment splits Acer into smaller sections, which are more consistent with large, published molecular phylogenies (e.g., Grimm et al. 2006, Renner et al. 2008) than the most recent treatment by van Gelderen (1994). Wolfe and Tanai recognized 21 sections of extant Acer, and maintained A. saccharinum in a separate, monotypic section from A. rubrum and A. pycnanthum. Our section-level sampling according to Wolfe and Tanai (1987) may underrepresent diversity in Acer, especially within sect. Acer, which has been the least taxonomically stable section and probably includes species that are phylogenetically distant (Ogata 1967; Wolfe and Tanai 1987; van Gelderen et al. 1994; Grimm et al. 2006, Renner et al. 2008). Nevertheless, resolving relationships in sect. Acer is beyond the scope of our study and, based on outcomes from prior molecular phylogenetic studies (Grimm et al. 2006, Renner et al. 2008), species variously treated in sect. Acer are distant from sects. Rubra and Hyptiocarpa. Of the 21 sections recognized by Wolfe and Tanai, we sampled 20, but the missing section, sect. Integrifolia, may be represented by Acer pentaphyllum Diels. Wolfe and Tanai (1987) treated A. pentaphyllum in sect. Acer, but the species is included in sect. Pentaphylla in van Gelderen et. al (1994) with other species of Wolfe and Tanai’s (1987)sect. Integrifolia and is resolved with species of sect. Integrifolia in molecular phylogenies (Suh et al. 2000; Grimm et al. 2006; Renner et al. 2008). For all sections of Acer, we sampled the type species when possible. In addition to species of Acer, we included both species of Dipteronia in our analyses, and we utilized one sequence each of Sapindus Linnaeus, Koelreuteria Laxmann, and Xanthoceras sorbifolium Bunge as outgroups. Xanthoceras may be sister to all Sapindaceae and Sapindus and Keolreuteria represent the core Sapindaceae (sensuBuerki et al. 2010), which is sister to Hippocastanoideae (Buerki et al. 2010). We did not include Hippocastaneae among the outgroup or ingroup, because it has ITS sequences that are very difficult to align with Acer according to a prior report (Grimm et al. 2006) and our personal experience. Nevertheless, prior molecular phylogenetic studies of Acer have used cpDNA and have included Aesculus of Hippocastaneae (Renner et al. 2008; Tian et al. 2002). Therefore, we compare results of those studies with our own. The ITS and psbM datasets comprised 27 sequences each. The details of our sampling, including additional explanation of taxonomic representativeness and GenBank accession numbers, are presented in Table 1.

Two sequences were new to this study: psbM of Acer sterculiaceum K. Koch subsp. franchettii (Pax) A.E. Murray and A. cissifolium (Siebold & Zucc.) K. Koch. We obtained the new sequences using fresh material, which we collected from the United States National Arboretum. Our collections consisted of leaves for DNA extractions, which we preserved in silica at the time of sampling, and voucher specimens, which we deposited at the United States National Herbarium (US; http://n2t.net/ark:/65665/396759747-a431-4859-b4a7-8c57db1cc2a2 and http://n2t.net/ark:/65665/36583930c-3354-4039-9e29-f9e0f9699ecb). We performed DNA extractions using a Qiagen Plant Mini Kit (Venlo, Netherlands) according to manufacturer recommendations, and we amplified psbM using forward and reverse PCR primers from Lee and Wen (2004). We performed PCR, sequencing, and purification steps using the reactions, thermocycling scheme, and protocols reported in Harris et al. (2017), except that the thermocycling included 35, rather than 40, cycles. Our primers for sequencing were the same as those that we used for PCR amplification. We reported the new sequences to GenBank (Table 1).

We performed sequence alignment using the MAFFT algorithm (Katoh et al. 2002; Katoh et al. 2005) on the GUIDANCE 2 (Sela et al. 2015) webserver (http://guidance.tau.ac.il/ver2/; Penn et al. 2010). GUIDANCE 2 helps to identify uncertain regions of an alignment by comparing alignments derived from bootstrap guide trees. The GUIDANCE 2 webserver also facilitates removing uncertain portions of an alignment and realigning through an iterative, interactive process. We performed initial alignments on our ITS and psbM data matrices with up to five MAFFT iterations for refinement and 100 bootstrap replicates. We used a conservative confidence score of 0.853 (GUIDANCE 2 Overview, http://guidance.tau.ac.il/ver2/overview.php), and we removed all sites with lower confidence scores. Following this step, we performed a new alignment in GUIDANCE 2 with the uncertain sites excluded, and we checked that the new alignment had a confidence score of at least 0.95 (out of 1.0 possible) averaged across all sites. We also checked the final alignment visually with sites color-coded according to their GUIDANCE 2 score using JALVIEW (Waterhouse et al. 2009) on the GUIDANCE 2 webserver. We concatenated aligned matrices using SEQUENCEMATRIX (Vaidya et al. 2011), and our concatenated matrix comprised composite taxonomic entities of the same section in Acer and usually of the same species, except in the case of sects. Rubra and Hyptiocarpa, for which composite entities were always of the same species (see Table 1 for Dipteronia and outgroups). We provide all final alignments in Dryad: http://dx.doi.org/10.5061/dryad.n26nd

Prior to phylogenetic analyses, we assessed the data matrices for base compositional heterogeneity and to determine the best nucleotide substitution model. We sought to detect base compositional heterogeneity, because it can lead to errors in phylogenetic inferences especially in the placement of outgroups and other long branches (Tarrı́o et al. 2000; Jermiin et al. 2004; Sheffield 2013). We performed the analysis for base compositional heterogeneity using a chi square test in PAUP* (Swofford 2002). We estimated the best model of nucleotide substitution from among 1-, 2-, and 6- parameter models with and without gamma rate variation (see Yang 1996 regarding invariance) in JMODELTEST (Posada 2008) under the Bayesian information criterion (BIC), and determined that the 6-parameter SMY+G (BIC=6544.4) and 2-parameter K80+G (BIC=5387.7) models were the best fit for ITS and psbM, respectively.

We performed phylogenetic analyses using neighbor-joining (NJ), maximum likelihood (ML), and Bayesian inference (BI) methods independently for ITS and psbM as well as for the concatenated data matrix. We performed the NJ analyses in GENEIOUS TREE BUILDER using Jukes Cantor distance and 1000 NJ bootstrap (BS) replicates to assess support. We reconstructed the ML trees in MEGA 6.06 (Tamura et al. 2013). In MEGA, we set models according to the results from JMODELTEST except that we used GTR+G for ITS, because it is the only 6-parameter model available in MEGA. We performed the analyses with five gamma rate categories and the subtree pruning and recrafting method of branch swapping. We also performed 500 BS replicates under the same parameters to determine support for clades. For BI, we utilized the GTR+G model of nucleotide substitution a priori (see Huelsenbeck and Rannala 2004; Ronquist et al. 2011) and unlinked models for the two markers in the analysis of concatenated data. The BI analysis comprised two simultaneous runs of 20 million generations with 12 incrementally heated MCMC chains each in MRBAYES 3.2.6 (Ronquist and Huelsenbeck 2003; Ronquist et al. 2011; Ronquist et al. 2012). We sampled the MCMC every 5000 generations and used Tracer 1.6 (Rambaut and Drummond 2007) to confirm stationarity and that a 10% burnin per independent analysis was appropriate. We combined results for simultaneous analyses using LOG COMBINER of the BEAST 1.8.0 software package (Drummond and Rambaut 2007; Drummond et al. 2012). We summarized the combined trees for each gene by selecting a maximum clade credibility tree with TREE ANNOTATOR, also of the BEAST 1.8.0 software package, and we obtained branch lengths for the selected tree using the median lengths from among the posterior distribution of trees. We also generated alternative summaries of the combined BI trees in GENEIOUS using 50% majority rule consensus with compatible groups with less than 50% support allowed. We visualized and rooted the final NJ, ML, and summarized BI trees in FIGTREE 1.4.0 (Rambaut and Drummond 2009). All final trees with clade support values are available in Dryad: http://dx.doi.org/10.5061/dryad.n26nd.

For the morphological study of leaves, we examined individuals representing all four species comprising Acer sects. Rubra and Hyptiocarpa. We sampled leaves from all available specimens of A. laurinum and A. pycnanthum and four specimens each of Acer rubrum and A. saccharinum (Table 2). Our sampling of A. rubrum and A. saccharinumincluded mid- and late-season specimens from two or more geographically distant parts of the ranges of the species and was designed to facilitate detection of population-level and seasonal variation in cuticular wax features (Sargent 1922; de Jong 1976; Delendick 1981). We obtained leaf samples near the center of leaves from sites adjacent to the midvein. The samples were dry when we obtained them from herbarium sheets. Air-dried samples, such as from herbarium sheets, are suitable for examination of cuticles without additional preparations and do not typically develop structural artifacts from drying or during examination with SEM (Pathan et al. 2010). We used specimens deposited at the United States National Herbarium (US) to obtain all leaf materials (Table 2).

We used a Hitachi TM300 scanning electron microscope (SEM) to examine the ultrastructure of the abaxial and adaxial surfaces of the leaves following standard protocols. We used a standard working depth of 10mm and took SEM micrographs under 15kv after determining that this intensity of the electron beam would not melt the cuticular wax. All of our scanning electron micrographs of the leaf surfaces are available from in Dryad: http://dx.doi.org/10.5061/dryad.n26nd.

Throughout, we apply the term ‘cuticle’ to all parts of the wax layer(s) above the cellulose wall of the epidermal cells. We acknowledge that the cuticle is a complex structure comprised of many well-delimited and/or intergrading components (reviewed in Fernández et al. 2016). However, our imaging is from a birdseye view, such that we are not able to distinguish among cuticular layers. We use terminology for cuticular wax forms following Barthlott et al. (1998). For discussion of leaf characters, especially veins, we follow The Manual of Leaf Architecture of the Leaf Architecture Working Group (1999).

We examined numerous herbarium specimens to complete this study. In particular, we examined specimens in person at US, South China Botanical Garden (IBSC), and the United States National Arboretum (NA). We also examined high resolution images of specimens online using JSTOR Global Plants (http://plants.jstor.org/) and SEINet (http://swbiodiversity.org/seinet).

Our phylogenetic results are congruent with previous molecular studies, which have found well-supported close relationships between Acer sects. Rubra and Hyptiocarpa. For example, Renner et al. (2008) reconstructed a phylogeny of Acereae from six chloroplast genes, including psbM, and using all four species comprising sects. Rubra and Hyptiocarpa. They found 99%BS support for a Rubra-Hyptiocarpa clade based on an ML analysis and showed the same relationships within the clade as in our analyses (Fig. 2; data in Dryad) (Renner et al. 2008). Similarly, Li et al. (2006) performed an NJ analysis of Acereae and found 100%BS support for a clade of sects. Rubra and Hyptiocarpa according to two chloroplast genes, including psbM, and Grimm et al. (2006) obtained the same result using MP and BI analyses of ITS. Tian et al. (2002) also recovered the Rubra-Hyptiocarpa clade from concatenated ITS and one chloroplast gene, trnL-F, except that they did not include Acer pycnanthum in their study. In addition to phylogenetic reconstruction, network analyses have also shown strong support for the grouping of the Rubra-Hyptiocarpa (Grimm et al. 2006; Renner et al. 2008). By comparison to other studies, we found relatively low molecular phylogenetic support for the Rubra-Hyptiocarpa clade, and this is probably due to stringent removal of uncertain portions of our alignments and because our chloroplast dataset is small, comprising only one gene.

We observed very similar cuticular wax configurations on the abaxial leaf surfaces of species of sect. Rubra and in A. laurinum. In general, these configurations comprised membranous crystals that coalesce in formations appearing as wax splatters on the surface. We unexpectedly showed evidence that cuticles comprised of membranous plates are the source of the classic silvery appearance in sects. Rubra and Hyptiocarpa by showing that when cuticular wax formation is absent in A. pycnanthum (Fig. 7A), so is the silvery color. We expect that cuticular waxes are probably responsible for the silvery color in all species of sects. Rubra and Hyptiocarpa, and the relationship between the silver color and cuticular waxes has been previously noted and explored (e.g., Baker 1974; Caddah et al. 2012).

Some authors have speculated that cuticular wax configurations may be of limited taxonomic value, because they could vary with environment (Baker 1974; Mayeux et al. 1981). However, cuticular waxes have been informative in other groups (e.g., Jatropa Linnaeus, Dehgan 1980; Rosa Linnaeus sect. Caninae, Wissemann 2000; and Aralia Linnaeus, Wen 2011) and often have clear evolutionary significance (Eglinton and Hamilton 1967). Moreover, we did not find notable differences in the waxes among specimens collected in different parts of their geographic ranges or during different seasons (compare images at http://dx.doi.org/10.5061/dryad.n26nd from specimens of Acer rubrum and A. saccharinum). The striking cuticular waxes on the abaxial surfaces of all four Rubra-Hyptiocarpa species probably reflects descent from a common ancestor and could function in insect interactions (e.g., limiting insect walking on the abaxial surfaces; Baker 1974; Eigenbrode and Espelie 1995; Federle et al. 1997; Gorb et al. 2008; Müller 2008) or reducing water loss (Sutter and Langhans 1982; Clarke and Richards 1988).

The cuticle layer on the adaxial surface of Acer laurinum appears less similar to the species of section Rubra. While both sects. Rubra and A. laurinum have striations, these differ in the size of the striae, or ridges, which are wider and taller in A. laurinum (compare Fig. 6B with Figs 3B, 4B, 5B). Additionally, the size of the striae in A. laurinum makes the cuticle appear thicker than in the other species. A thick cuticle in A. laurinum would be consistent with its distribution in subtropical and tropical regions (Bloembergen 1948; van Gelderen et al. 1994; Xu et al. 2008), as tropical species often exhibit thick cuticles to reduce leaching via regular rainfall (Martin and Juniper 1970; Boeger et al. 2004). Striations of different widths and heights between A. laurinum and sect. Rubra may represent specialized adaptations to local conditions but the presence of striations may arise from a common genetic architecture. Recent studies on the genetic basis for cuticular wax phenotypes in model organisms such as Sorghum L. (Punnuri et al. 2017) and Arabidopsis Heynh. (Lee and Suh 2015) provide a foundation for future investigations of the evolutionary origins of cuticular wax forms in sects. Rubra and Hyptiocarpa and other maples.

Acer sects. Rubra and Hyptiocarpa cannot be united strictly based on the appearance of the abaxial surfaces of their leaves. Although this feature may have taxonomic value (Merrill 1941; Krause 1978; Delendick 1981) and it appears monomorphic in sects. Rubra and Hyptiocarpa, it also occurs elsewhere in the genus (van Gelderen et al. 1994). In particular, silvery or glaucous surfaces occur in most species of sect. Pentaphyllum and in some species of sect. Acer. Nevertheless, the taxonomic informativeness of cuticular waxes in Acer may warrant further investigation to compare both the fine features of ultrastructure and wax chemical composition especially within and among glaucous and non-glaucous sections and species.

Different taxonomic treatments of Hyptiocarpa do not all agree on species delimitation within the section. The large number of synonyms in Hyptiocarpa and confusion over the boundaries of species may reflect high variability and the need for additional field work to elucidate species limits or intergradation (Bloembergen 1948; van Gelderen et al. 1994). The most recent treatment of sect. Hyptiocarpa in Flora of China (Xu et al. 2008) recognizes two species: Acer laurinum and A. pinnatinervium. Acer pinnatinervium is considered a synonym of A. laurinum by van Gelderen (1994) and in the Plant List (http://www.theplantlist.org), but its status within Hyptiocarpa merits discussion here.

According to Xu et al. (2008), Acer laurinum and A. pinnatinervium differ in fruit size, the number of primary veins per leaf, and their geographic distributions. Acer laurinum has fruits 4-7 cm and leaves with three primary veins, while Acer pinnatinervium has fruits 2-4 cm and only one primary vein, i.e., it is truly pinnately veined. The pinnate venation in Acer pinnatinervium may be particularly noteworthy, because most species of Acer have leaves with three main veins (Merrill 1941). Therefore, pinnate venation in A. pinnatinervium is considered the primary character for distinguishing it from A. laurinum (Merrill 1941). With respect to geographic distributions, Xu et al. (2008) report that Acer laurinum has a broader range, being found from southwestern China to India, Vietnam, Indonesia, and the Philippines, while A. pinnatinervium occurs in southwest China, Thailand, and India.

Closer examination of Acer laurinum and A. pinnatinervium shows that they intergrade on the number of primary veins. Some collections of A. laurinum (e.g., Blume 466, L; Blume s.n., L) show strong basal acrodromous veins, while isotypes of Acer pinnatinervium (F. Kingdon-Ward 9102, A, BM) show pinnate venation with brochidodromous secondary veins near the leaf base. However, the holotype and isotype of A. laurinum (F.W. Junghuhn s.n., L, U, respectively) each show variability in venation such that some leaves have acrodromous veins and others are pinnately veined with weak brochidodromous secondaries. We also observed this variability within a specimen of A. laurinum utilized in the SEM component of this study, Cult., in Hort. Bog. III,K,37 (see Table 2), and in many specimens that are ascribed to A. pinnatinervium and digitized in the Chinese Virtual Herbarium (http://www.cvh.ac.cn/). In the latter case, intra-individual variability of leaf veins may account for recent disagreements in the identities of specimens as either A. laurinum or A. pinnatinervium evidenced by the annotation labels. Based on these observations, we suspect that the number of primary veins is not be sufficient to distinguish Acer pinnatinervium from A. laurinum, and combining the two species may be needed pending an additional study of more strategically samples individuals.

Leaves in sects. Rubra and Hyptiocarpa, hereafter sect. Rubrasensu latu, exhibit shapes that vary within and among species from elongate to orbicular (Fig. 1). Acer laurinium and A. pinnatinervium have highly elongate leaves, while, in A. rubrum, leaves vary from being orbicular (Fig. 1A) to having slight elongation (Fig. 8A). Similarly, leaves in mature A. pycnanthum may also possess roughly orbicular leaves (Fig. 1B) to leaves that are highly elongated and nearly lacking lobes (Fig. 8C–D). In A. saccharinum, most individuals have leaves that are more-or-less orbicular (Fig. 1C), but some have elongated leaves (e.g., Chaney 290, LSU). Many species of Acer exhibit elongation of juvenile leaves, including in A. saccharinum (e.g., Longbottom 8925, DOV), A. pycnanthum (e.g., Meyer 12513, NA), and A. rubrum (Fig. 8B). Additionally, leaves and leaflets in other Sapindanceae are also often elongate (Acevedo-Rodríguez et al. 2011; Harris et al. 2017). Leaf elongation in seedlings of Acer may indicate an underlying genetic architecture in the genus and, consequently, ontogenic recapitulation (Haeckel 1866; e.g., Mishler 1998). Thus, while variable leaf shape in sect. Rubra does not unite its species, the tendency towards elongation is likely a noteworthy plesiomorphy in Acer.

Section Rubra s.l. has variable inflorescence architecture (Fig. 9). Acer rubrum (Fig. 9A–B), A. pycnanthum (Fig. 9C), and A. saccharinum (Fig. 9D) have inflorescences that are umbels (de Jong 1976; van Gelderen et al. 1994), while A. laurinum and allied taxa have inflorescences that may be racemes (F.W. Junghuhn s.n., L) or paniculate thyrses (Lindley, 418, K) (de Jong 1976; van Gelderen et al. 1994). The umbels, which are unique within Acer, probably represent evolutionarily reduced racemes, while the racemes, which are more common in Acer, may represent reduced paniculate thyrses (de Jong 1976, Singer 2008). Inflorescences throughout sect. Rubra s.l. are almost exclusively lateral (Ohwi 1965, de Jong 1976, van Gelderen et al. 1994), although some authors report occasional terminal inflorescences in A. pycnanthum (van Gelderen et al. 1994; but contrast with A. pycnanthum in Ohwi 1965, de Jong 1976). While lateral inflorescences are common to other sections of Acer, exclusively lateral ones (or nearly so) occur only in sects. Rubra, Lithocarpa, and Glabra.

Species of sect. Rubra s.l. except A. pycnanthum may be monoecious or dioecious and exhibit labile sex expression among individuals (de Jong 1976; Primack and McCall 1986; Santamour 1993), and within-individual and within-clade labile sex expression occurs in some other groups of Acer and other Sapindaceae (Acevedo-Rodríguez et al. 2011; Renner et al. 2007). Acer pycnanthum is thought to be exclusively dioecious (de Jong 1976, Saeki 2008). Among monecious individuals of Acer rubrum, A. saccharinum, and A. laurinum, individual inflorescences are usually exclusively comprised of staminate or pistillate flowers. One prior study inferred that dioecy was ancestral in sect. Rubra s.l., but that inference was based on scoring A. laurinum as dioecious (Renner et al. 2007), which is not accurate (Bloembergen 1948; de Jong 1976; Xu et al. 2008). All flowers in sect. Rubra s.l. emerge from leafless buds, and this is a taxonomically informative trait that delimits some sections of Acer from others (de Jong 1976; van Gelderen et al. 1994).

Fruits in sect. Rubra s.l. also share many features (Fig. 10), especially from among those identified as taxonomically informative in a comprehensive study by Wolfe and Tanai (1987). We have observed that the fruits of all species in sect. Rubra have slightly inflated seed locules without keels, wings that are straight at the base, and mericarps forming an acute angle with respect to one another. Each of these traits tends to be monomorphic within sections. Each trait occurs in about half of all sections, but this suite of traits may be unique to sect. Rubra s.l. Additionally, species in sect. Rubra s.l. are capable of producing partially developed seedless mericarps (Fig. 10), compared to complete or extremely minimal (e.g., roughly pinhead-sized) development in other species and sections (de Jong 1976). The degree of development of seedless mericarps in Acer is well-characterized by de Jong (1976) and is taxonomically informative. Partially developed, seedless mericarps occur in about half of sections of Acer, and most sections are monomorphic for this trait. Fruits of sect. Rubra s.l. are highly variable in size within species with the largest fruits occurring in A. saccharinum and A. laurinum (Townsend 1972; van Gelderen et al. 1994; Xu et al. 2008).

Prior studies have proposed other plausible relationships for sect. Hyptiocarpa based on morphology. In particular, leaf morphology has often been used to link sect. Hyptiocarpa with Acer oblongum Wallich ex de Candolle and its close relatives in sect. Pentaphylla or Integrifolia (Pax 1885; Momotani 1962; Fang 1966). Acer oblongum has entire, unlobed elongate leaves and silvery abaxial surfaces (van Gelderen et al. 1994) that are similar to leaves in A. laurinum. Nevertheless, any association between Acer oblongum and sect. Hyptiocarpa has not been supported by molecular phylogenies, which show that Acer oblongum is associated with sect. Pentaphyllum and distant from sect. Rubra (Suh et al. 2000; Renner et al. 2008). Morphologically, A. oblongum differs from A. laurinum by having mostly terminal inflorescences and by flowers and leaves arising from the same buds (van Gelderen et al. 1994). Additionally, the waxes of A. oblongum may differ from those in sect. Rubra by extending partially onto the midrib. While we made this observation on many specimens at IBSC, we used a low magnification hand lens, and a more detailed study using higher magnification may be warranted. Another possible association for sect. Hyptiocarpa was with sect. Lithocarpa, which has a relatively large number of bud scales, axillary inflorescences from leafless buds, and insertion of stamens on a staminal disk (Ogata 1967); features that are also shared with sect. Rubra s.s., except for stamen insertion (Pax 1885; Ogata 1967; de Jong 1976; van Gelderen et al. 1994). In sect. Rubra stamens are inserted outside of the disk or the disk is absent in some individuals of each species (van Gelderen et al. 1994). Thus, the disk may be relatively labile within sect. Rubra s.s. and in sect. Rubra s.l. Section Hyptiocarpa differs from sect. Lithocarpa (except A. macrophyllum Pursh.) by having wood rays 3-4 cells wide rather than cells wide. Overall, in prior taxonomic work, recognition of a distinct sect. Hyptiocarpa, seems more motivated by uncertainties about its affinities (Ogata 1967; de Jong 1976; Delendick 1981; Wolfe and Tanai 1987; van Gelderen et al. 1994) than affirmation of its significant uniqueness within Acer (e.g., contrasted with A. carpinifolium and A. negundo Linnaeus).

Figure 8. 

Elongate leaf shape in Acer rubrum and A. pycnanthum. A–B A. rubrum C–D A. pycnanthum. Unfortunately, there is no scale for the images of A. pycnanthum, but the leaf size is similar to that illustrated in Figure 1B. Herbarium specimens in A and B deposited at US, and accession information visible in images. Detailed specimen records are available via the US online catalog (http://collections.nmnh.si.edu/search/botany/).

Figure 9. 

Inflorescences of Acer sects. Rubra and Hyptiocarpa. A Acer rubrum with umbels of pistilate flowers B Acer rubrum with umbels of staminate flowers C A. pycnanthum with umbels of pistilate flowers C A. saccharinum with umbels of pistilate flowers. Note flowers with two, divided persistent styles D A. pinnatinervium with racemose thyrse. Specimens in A–D deposited at US, and specimen in D deposited at the British National Museum (BM). Accession information visible in images, and detailed specimen records are available via the US online catalog (http://collections.nmnh.si.edu/search/botany/) and at the data portal of BM (http://data.nhm.ac.uk/).

Figure 10. 

Fruits of species of Acer section Rubra. A A. rubrum. Specimen on left deposited at US National herbarium (US) with collection name and number: Lilian 62. Specimen on right deposited at Kew (K) as Acer drummondii Nutt. (= A. rubrum) with collection name and number: Drummond 53. Image of fruits obtained from image of specimen deposited in JSTOR Plants (http://plants.jstor.org/) B A. pycnanthum, used with attribution to Chinese Virtual Herbarium(http://www.cvh.ac.cn/); Miyoshi Furuse 54050, PE C A. saccharinum showing fruit with two fertilized ovules (upper) compared with one fertilized ovules and one partially developed, unfertilized ovule (lower). Specimens deposited at US with collection name and number: Wolf s.n. and Pringle s.n., respectively D A. laurinum Specimen deposited at K with collection name and number: Lindley, 418. Image of fruits obtained from image of specimen deposited in JSTOR Plants. Scale bar of 1cm applies to all images.

Section Rubra s.l. may have radiated out of the tropics and into temperate areas of Japan and North America based on our phylogenetic results (Fig. 2) and results presented in other molecular phylogenetic studies (Grimm et al. 2006; Renner et al. 2008). In particular, results suggest that A. laurinum is the earliest diverging species within sect. Rubra s.l. An out- of-the-tropics radiation in sect. Rubra s.l. may have been accompanied by, or even spurred by, polyploidization. According to the literature, Acer laurinum is diploid (2n=26), A. saccharinum is tetraploid, A. pycnanthum is hexaploid, and A. rubrum includes hexaploid, heptaploid (rarely), and octaploid individuals (Fig. 2), and these counts reflect attempts to avoid hybrid individuals and use materials originating from wild populations (Löve 1971; Santamour 1965; van Gelderen et al. 1994; Chromosome Count Database, http://ccdb.tau.ac.il/home/). Polyploidization is assumed to have played a role in adaptation to less equitable environments in some plant groups, and polyploidy sometimes shows clear positive correlation with latitude (Beaton and Hebert 1988). Future studies may examine the timing of evolutionary radiation, such as by using fossils and divergence time dating, to better understand possible correlations between ploidy level and past environments.