Reply to Doyle: Emphasizing independent gene duplications and convergent molecular function

The origin(s) and evolution of root nodule symbiosis (RNS) have long attracted biologists (1–6). Using gene history as a proxy for trait (RNS) history, we (7) reported independent duplications in a gene involved in recognition between host plants and their bacterial symbionts and inferred that these duplications promoted convergent acquisition of gene function in legumes and Cucurbitales/Rosales, providing a possible mechanism for multiple gains of RNS (6). Doyle (8) argues that our reference to independent gene duplications in NFP as “nonhomologous” and our failure to refer to these collective gene products as co-orthologs negate our findings of molecular convergence in NFP function. Doyle (8) then claims that NFP function can only be convergent if we assume recruitment of NFP for nodulation was independent in these two groups and argues that a single origin of RNS is “simpler” than independent origins. Here, we reply to these critiques.

First, our gene tree clearly demonstrates a single copy of NFP in the common ancestor (CA) of the nitrogen-fixing clade (NFC), with evidence for i) multiple independent duplications in rosids outside the NFC and ii) three relevant independent duplications in the NFC (7). Doyle (8) emphasizes that the duplicate genes are co-orthologs. We agree with Doyle’s terminology for gene products (8: figure 1) but focus instead on the independent (i.e., nonhomologous) duplication events themselves. Doyle (8) apparently misinterpreted our statements of nonhomology of duplication events as nonorthology of duplicate genes.

Second, the designation of products of independent duplications as co-orthologs is not informative regarding acquisition of signaling function because gene genealogy alone does not directly equate to co-option of function (Figs. 1 and 2). When and where co-option from the ancestral mycorrhizal function of NFP occurred phylogenetically, and the relationship of co-option to gene duplication, are addressed in part (7). The current signaling function was likely absent in the CA of the NFC, given the phylogenetic distribution of functional NFP copies. We make no a priori assumptions about whether RNS was present in the CA but instead falsify the previous co-option hypothesis (4), which Doyle fails to acknowledge, and stress: “Future studies will be necessary to identify the convergent functional changes that ancestral NFP underwent to function in … symbiosis” (7, p. 6). The relevant NFP function cannot be mapped to the NFC phylogeny until after the clearly independent duplications in Cucurbitales/Rosales and legumes, arguing against a single origin of RNS. The NFP topology directly contradicts massive loss of genes and/or function in alternative gene copies as required by the single-origin hypothesis.

Fig. 1.

Simple schematic figure showing single or multiple gains of function associated with gene duplication. (A) Single co-option event, as in ref. 4. (B) Multiple co-option events, as in ref. 7. Species trees (species labeled A–I, duplications indicated by black diamonds above the branches) are depicted with included gene trees, with colors indicating function (ancestral in blue vs. derived in red). The gray bar indicates the clade with the derived functional trait of interest, carrying no implication as to origin.

Fig. 2.

Gene trees as in Fig. 1, expanded for easier reading. Tip labels indicate co-orthology relationships: the first number indicates the deep duplication, and subsequent numbers represent subsequent duplications and thus subsequent co-orthology (e.g., E1.1 = deep duplication and one subsequent duplication). In (A), the entire gene copy set with the derived function is co-orthologous to a clade of genes that retained the ancestral function. In (B), species C does not have a co-option event for the specified gene and may have undergone co-option through another mechanism (7). Terminology is relative to the frame of comparison. In (B), the gene set [D1, E1.1, F1.1, D1.2, F1.2] is co-orthologous to D2 or to the set [D2, E2, F2] as [E2, F2] is to D2. However, [D1, E1.1, F1.1, D1.2, F1.2, G1, I1] is not co-orthologous to [D2, E2, F2, G2] because the underlying duplications yielding the set are not homologous (i.e., the sets induced by the comparison and the sets induced by the duplications conflict); this is not true of the single duplication in (A).

The duplication history of NFP (7) is thus inconsistent with a single origin of RNS (4). With no new data, Doyle has challenged our interpretation of molecular convergence of NFP by focusing on co-orthology of duplicate genes rather than the independent events that generated them. As stressed previously (7), and in agreement with Doyle (8), new data on history, function, and evolution of NFP and additional RNS genes are needed to clarify the complex history of nodulation (6) and how evolution builds complexity.