The Linear Angiosperm Phylogeny Group (LAPG) III: a linear sequence of the families in APG III

Abstract

Linear sequences of families are required for herbarium curators who wish to arrange collections systematically rather than alphabetically, and there are currently a wide range of systems in use. Over time, in most herbaria, as the understanding of relationships changed, these classifications have become outdated and no longer represent the best estimate of relationships. Curators have updated these systems, usually by adding suffixes or making composite numbers such as ‘145a’ or ‘145.01’ to incorporate changes in an existing sequence. In many cases, however, these modifications have not been published and have often been herbarium-specific, resulting in herbaria that originally used the same system diverging over time.

The linear sequence of families based on the Angiosperm Phylogengy Group (APG) II classification system (LAPG II; Haston et al., 2007) was founded on the best estimate of family relationships at that time. In revising the linear sequence of families to the new APG III classification (APG III, 2009), we provide a clear and explicit update according to the current best estimate of relationships. In LAPG II, we decided to allocate a family number to each of the families listed in the APG II classification, including the ‘bracketed’ families that were presented as ‘acceptable monophyletic alternatives to the broader circumscription favoured here’ (APG II, 2003). In doing this, we felt that we would be giving herbaria the greatest flexibility in accepting or rejecting the ‘bracketed’ families. The APG III classification has tended to accept the broader circumscription of families and does not allow the option of ‘bracketed’ families. The number of accepted families has therefore significantly declined. In addition, our understanding of relationships in several parts of the phylogenetic tree has improved and these changes have been incorporated here.

In considering linear sequences derived from phylogenetic trees, it is important to remember that the sequence is dependent on the methodology used and that loss of phylogenetic information in the conversion is unavoidable. The choice of methodology is controversial (Hawthorne et al., 2008), and the linear sequence can be misleading if not understood in the context of a tree. Linear sequences systematically ordered are, however, used to arrange most of the large herbaria worldwide and therefore merit some attention. This paper provides, for herbarium curators, a simple linear numbered sequence to the families recognised in APG III (Table 1).

The methodology used to translate the phylogenetic tree into a linear sequence was described in detail in Haston et al. (2007). This methodology was questioned by Hawthorne et al. (2008), but in the absence of any obviously better way of generating a linear sequence from a phylogenetic tree we have retained the same method. We agree that there is potential to modify this method in the future, especially to ensure stability.

The main changes in the order of families in the sequence are within the rosid clade. Additional resolution has clarified the circumscription of the fabid (rosid I) and malvid (rosid II) clades. Of these changes, the following have had the greatest impact on the linear sequence: Geraniales are now sister to Myrtales in the malvid clade; Crossosomatales are sister to the remaining malvid families; Malvales are now sister to Brassicales (Wang et al., 2009); and Huerteales are now sister to the Malvales/Brassicales clade (Worberg et al., 2009).

Elsewhere in the sequence there have been more minor changes, the more significant of which are listed here. Ceratophyllaceae have moved from within magnoliids and are now sister to the eudicots (Jansen et al., 2007; Moore et al., 2007). Aextoxicaceae and Berberidopsidaceae are now sister to the Santalales/Caryophyllales/asterid clade (Moore et al., 2009). Dilleniaceae are probably now sister to the Saxifragales/rosids clade (Moore et al., 2009), but this is not well supported at present. In addition, there were changes to the sequence when clades were rotated as a result of changes in the number of terminals caused by the sinking of families or, more rarely, the recognition of new families.

The individual trees on which this sequence is based are mostly published (Stevens, 2001; Jansen et al., 2007; Moore et al., 2007 onwards; and papers referenced in Haston et al., 2007), but see also Wang et al. (2009), Tank & Donoghue (in press), Moore et al. (2008, 2009); Refulio-Rodriguez & Olmstead (2008); Wurdack & Davis (2009) and papers in Stevens (2001 onwards). For justification of family limits adopted, see APG III (2009).

In LAPG III, we suggested that composite numbers be used for minor changes such as new families (Haston et al., 2007). We would continue to recommend the use of composite numbers for families that are added or moved and would add that they may be useful to enable herbaria to change systems in a more gradual process. We would also reiterate that amendments that involve major changes in relationships will require a revised linear sequence. However, we anticipate that changes in the future will be infrequent and less disruptive to the linear sequence.

REFERENCES