Pl. Syst. Evol. 260: 199–221 (2006) DOI 10.1007/s00606-006-0443-8 Floral structure and systematics in four orders of rosids, including a broad survey of floral mucilage cells M. L. Matthews and P. K. Endress Institute of Systematic Botany, University of Zurich, Switzerland Received November 11, 2005; accepted February 5, 2006 Published online: July 20, 2006 Ó Springer-Verlag 2006 Abstract. Phylogenetic studies have greatly ened mucilaginous inner cell wall and a distinct, impacted upon the circumscription of taxa within remaining cytoplasm is surveyed in 88 families the rosid clade, resulting in novel relationships at and 321 genera (349 species) of basal angiosperms all systematic levels. In many cases the floral and eudicots. These cells were found to be most structure of these taxa has never been compared, common in rosids, particulary fabids (Malpighi- and in some families, even studies of their floral ales, Oxalidales, Fabales, Rosales, Fagales, Cuc- structure are lacking. Over the past five years we urbitales), but were also found in some malvids have compared floral structure in both new and (Malvales). They are notably absent or rare in novel orders of rosids. Four orders have been asterids (present in campanulids: Aquifoliales, investigated including Celastrales, Oxalidales, Stemonuraceae) and do not appear to occur in Cucurbitales and Crossosomatales, and in this other eudicot clades or in basal angiosperms. paper we attempt to summarize the salient results Within the flower they are primarily found in the from these studies. The clades best supported by abaxial epidermis of sepals. floral structure are: in Celastrales, the enlarged Celastraceae and the sister relationship between Celastraceae and Parnassiaceae; in Oxalidales, the Key words: androecium, Celastrales, Crossoso- sister relationship between Oxalidaceae and Con- matales, Cucurbitales, gynoecium, Oxalidales. naraceae, and Tremandraceae embedded in Elae- ocarpaceae; in Cucurbitales, the sister relationship between Corynocarpaceae plus Coriariaceae, and ‘Very large’, ‘poorly understood’ but ‘well the grouping of the core Cucurbitales (Cucurbit- supported’ are three good descriptors for the aceae, Begoniaceae, Tetramelaceae, Datiscaceae); rosid clade (Fig. 1). It comprises perhaps a in Crossosomatales, the sister relationship between third of the species of angiosperms and includes Ixerbaceae plus Strasburgeriaceae, and between 14 orders (Savolainen et al. 2000; Soltis et al. this clade and Geissolomataceae. The core Crosso- somatales (Crossosomataceae, Stachyuraceae, 2000, 2005; APG 2003), of which many have Staphyleaceae) and Celastrales as an order are had complex taxonomic histories. Following not strongly supported by floral structure. In molecular analyses, new orders have appeared addition, a new floral feature of potential in some cases, such as Crossosomatales (APG systematic interest is assessed. Specifically the 2003), while in others the traditional composi- presence of special cells in flowers with a thick- tion of the orders has been altered, such as in 200 M. L. Matthews and P. K. Endress: Floral structure and systematics in four orders of rosids Saxifragales 98 Matthews and Endress 2002), Cucurbitales Vitaceae 100 (Matthews et al. 2001, Matthews and Endress 100 Myrtales 80 Geraniales 2004), Crossosomatales (Matthews and En- 100 Crossosomatales Zygophyllales dress 2005a) and Celastrales (Matthews and 100 62 Celastrales Endress 2005b). Our reason for selecting these Oxalidales COM 99 51 100 clade Malpighiales orders of rosids was based upon an initial 99 Fabids 100 Rosales (=EurosidsI) study comparing what appeared to be the 77 Fabales N-fixing 98 very similar flowers of Cunoniaceae (Oxali- 68 100 Cucurbitales clade 60 100 Fagales dales) and Anisophylleaceae (Cucurbitales) 100 Brassicales Malvids (Matthews et al. 2001, Scho¨nenberger et al. 95 100 Malvales (=EurosidsII) 51 100 Sapindales 2001). The ensuing results prompted us to compare the remaining families from these Fig. 1. Cladogram of rosids and potential sister two orders (Matthews and Endress 2002, groups (Saxifragales and Vitaceae) (modified after 2004). When Crossosomatales appeared as a APG 2003 and Soltis et al. 2005, with jackknife newly circumscribed order (Sosa and Chase values after Soltis et al. 2005). Orders whose floral 2003, APG 2003) they became our next focus structure was comparatively studied are highlighted of study (Matthews and Endress 2005a). in bold (Matthews and Endress 2002, 2004, 2005a, b) Finally Celastrales were compared as we had already studied Oxalidales and planned to study Malpighiales in the future, thus com- Cucurbitales (e.g. Zhang et al. 2006). Addi- pleting the COM (Celastrales-Oxalidales-Mal- tionally, floral structure for many members of pighiales) clade in which these three orders these orders is either unknown or poorly currently form a polytomy. In addition we represented (see also Stevens 2001 onwards, surveyed one feature, elaborate petals, Judd and Olmstead 2004). throughout all orders of the eudicots (Endress In view of this progress in our phyloge- and Matthews 2006a). netic understanding of rosids, we commenced In this paper we attempt to summarize the a project five years ago focussing upon the results from these four large ordinal studies, comparative floral structure of rosid orders specifically mentioning only those special with either new or novel circumscription. Our (uncommon) features which are of interest overall aim was to determine whether floral as potential synapomorphies for various structure (morphology, anatomy and histol- clades. Additionally, a new feature of poten- ogy) either supports or questions the rela- tial systematic interest, special cells in flowers tionships proposed by molecular phylogenetic with a thickened mucilaginous inner cell wall studies, and additionally to tease out poten- is assessed, which has emerged from our tial floral synapomorphies for these orders. intitial studies (Matthews et al. 2001, Mat- We aimed to describe the same features in thews and Endress 2002) and was corrobo- the same way and to ensure that the same rated by the later studies (Matthews and features were always described for each Endress 2004, 2005a,b). We have surveyed the study, but adding new features as they presence of this feature in flowers for 321 emerged. Our motivation based on a com- genera (349 species) from 88 families of basal mon problem faced when surveying literature, angiosperms and eudicots and also assessed that different studies use different terms to the distribution of these cells in leaves based describe the same feature or certain features on published records (West 1969; Metcalfe are just not mentioned, all of which leads to and Chalk 1950, 1979, 1983; Napp-Zinn 1973; confusion. Metcalfe 1987; Gregory and Baas We have studied four orders to date 1989; Bakker 1992; Bakker and Gerritsen including Oxalidales (Matthews et al. 2001, 1992b; Mariani et al. 1988; Bakker and Baas M. L. Matthews and P. K. Endress: Floral structure and systematics in four orders of rosids 201 1993; Bredenkamp and Van Wyk 1999). surveyed per family was not proportional. In Finally some general comments and conclu- total, flowers from 88 families (321 genera sions are given with respect to our experience and 349 species) of basal angiosperms and from these broad studies. eudicots were surveyed for the presence or absence of these cells and type of mucilage cell noted. Approximately 74 families of the Materials and methods 86 families listed by Metcalfe and Chalk Survey of mucilage cells. We have used the (1950) and Napp-Zinn (1973) were surveyed. publications by Metcalfe and Chalk (1950) The additional families were from our own and Napp-Zinn (1973) to compile a com- ordinal studies (Matthews et al. 2001; Mat- bined list of taxa for which either cells with a thews and Endress 2002, 2004, 2005a, b). mucilaginous inner cell wall and distinct Special mucilage cells are clearly recognisable cytoplasm (here called ‘special mucilage in plastic-embedded material, less so in the cells’) or entirely mucilaginous cells with an paraplast sections, although best attempts indistinct cytoplasm (here called ‘unspecified were made. In cases where it was not mucilage cells’) are mentioned for vegetative possible to clearly distinguish between the parts. In cases where the names (family or two cell types (in paraplast sections), the genus) have changed since these publications species was noted as having unspecified the new name is used (Appendix I). The mucilage cells. Species were assigned to ordinal classification is based on APG having special mucilage cells only when (2003), and Ka˚ rehed (2001) is used for clearly visible cytoplasm was present (posi- Icacinaceae. tioned almost always to the outside of the Metcalfe and Chalk (1950) mention 69 mucilaginous cell wall) including cases where families whose leaves have a mucilaginous the cytoplasm was reduced but still clearly epidermis, and of these, 13 are described as visible (e.g. Fig. 9 Strasburgeria robusta). In having cells with a mucilaginous inner cell other cases, both special and (what appeared wall. Napp-Zinn (1973) mentions 14 families to be) unspecified mucilage cells were present for which the inner wall of all or most cells of in the same epidermis (e.g. Fig. 3 Connarus one or both of the leaf epidermises are conchocarpus), however, always in these cases mucilaginous, and nine families for which he the presence of special mucilage cells was supposes the cells are ‘‘mimicking’’ an epider- clearly evident. Advanced