Hypothesis of parallel genetic evolution at the Tb1 locus for the adaptation of vegetative branching during maize and pearl millet domestication. A. The phylogenetic tree shows that Zea mays and Pennisetum glaucum are two wild grasses from the Panicoid subfamily that separated 30 million years ago (dotted lines, scale not respected), wild Z.mays (teosinte) growing in America and wild P.glaucum in Africa. About 9,000–4,000 years ago, they were independently domesticated into maize and pearl millet, respectively. Pictures below the tree illustrate the parallel morphological evolution of both wild progenitors during their domestication, in particular the reduction of tillering and branching. Z.mays 

Context in source publication

Context 1

... independent emergence and selection of new adaptive mutations at homologous genes (reviewed in [8,9,10,11]). This initial view has been refined as the genetic control of developmental traits targeted by domestication is gradually revealed in model systems like rice and maize. Prior to domestication, wild progenitor grass species have diverged over 65 million years during which they have strongly diversified morphologically through the evolution of gene networks. It is now clear that many of these networks control the same developmental traits as those later targeted by domestication [7]. For some traits, genes have conserved their role in different grass species and sometimes even across monocots and dicots. In contrast, for other traits, developmental gene networks have evolved specifically in such a way that key determinants differ in related species (e;g. the ramosa gene in Panicoideae, which cannot be found in rice despite extensive efforts to clone it) [7]. Positional cloning of domestication genes is still tedious, slowing the advances to identify these determinants and compare them across species. Therefore, the hypothesis of parallel genetic evolution during domestication is not trivial and needs to be tested directly by a candidate-gene approach for a given domestication gene. Like in maize, vegetative branching has been considerably reduced during pearl millet domestication ( Pennisetum glaucum ) [12] ( Figure 1). Even though branch number still segregates in domesticated pearl millet populations, cultivated varieties produce much less branches than wild P.glaucum (Figure 1A–B). In some areas, segregation of wild features in the domesticated gene pool may be due to the occurrence of weedy plants, which display intermediate branching phenotypes [13] (Figure 1B). Previously, we reported a domestication QTL for reduced vegetative branching in this species which covered a region predicted to harbor the Teosinte-branched1 ( Tb1 ) gene according to comparative mapping [12,14]. Tb1 is a plant-specific transcription factor [15] and a major domestication gene in maize [16]. While the barley Tb1 orthologue has recently been shown to contribute to spike architecture differences between two-rowed and six-rowed varieties [17], Tb1 has mainly been associated to the development of vegetative branches. Its specific targets and mode of action are yet unknown but transgenic and mutant studies of Tb1 homologs in rice, sorghum and A.thaliana showed that it contributes to repress the activity of vegetative axillary meristems where it is expressed, and their expansion into branches [18,19,20,21]. Vegetative branching is a very complex and highly multigenic trait requiring the coordination of meristem growth by multiple pathways, including local meristematic and long-distance hormonal signals from roots and shoots, as well as environmental cues (reviewed in [22,23]). Surprisingly, Tb1 was singled out as the only major gene involved in the adaptation of vegetative branching during the domestication of maize, accounting for 35% of the trait variance [16,24,25], even though stem number is controlled by at least 8 other loci in the wild progenitor teosinte [26]. Further studies revealed that human selection targeted adaptive sites located upstream of the gene, possibly in regulatory sequences related to a hypothetic dosage effect of Tb1 on development or to the strong pleiotropy of the gene over inflorescence structure [27,28]. Tb1 is an obvious a priori candidate gene for the adaptation of vegetative branching in other domesticated species due to its conserved function in the control of this trait in dicots [19] and monocots (grasses) [18,20]. However, it has never been formally proven to be involved in the evolution of branching during domestication other than in maize. In fact, patterns of evolution in the coding sequence of the gene suggest that changes in the TB1 protein did not contribute to the morphological diversification of grasses [29]. This does not preclude the eventuality of positive selection on other Tb1 regions, such as its regulatory sequences. In the single study published to date examining Tb1 roles in the evolution of tillering during domestication other than in maize [30], a cDNA clone of the maize Tb1 gene was shown to coincide with a domestication QTL in some foxtail millet crosses ( Setaria italica ). This QTL was minor and its effects considerably smaller than those of Tb1 in maize (9% vs 35% on average). Therefore, Tb1 effects seem to vary greatly between species, making it difficult to predict if the gene may be a ‘‘key’’ locus recurrently recruited for the evolution of branching during domestication. The ontogeny of axillary stems from different types of vegetative meristems (see first section of results and ref.[7]), as well as the pleiotropy of Tb1 on inflorescence architecture [16,17,24,25] are further a priori arguments against the possibility of parallel genetic evolution at this locus. In this study, we asked if the Tb1 locus played a role in the evolution of tillering in pearl millet, using a candidate-gene approach to investigate the parallel evolution observed between maize and pearl millet during their domestication (Figure 1A). To test this hypothesis in the absence of routine transgenic technology in non-model species, we first checked whether polymorphism in the gene segregates with branching variation in pearl millet genetic crosses. We also extended this survey to rice, sorghum and foxtail millet. Secondly, we verified that Tb1 ’s expression pattern is conserved in pearl millet. Thirdly, we tested whether sequence polymorphism at the Tb1 locus in domesticated and wild populations is consistent with a recent event of human selection. For these purposes, we cloned PgTb1 , the orthologue of Tb1 in pearl millet and used a combination of QTL mapping, expression and molecular evolution analyses. Grasses produce two types of axillary stems from their main primary shoot. Tillers are issued at a basal position, from nodes that are put in place early during seedling development, and they often develop their own adventitious roots independent from the main stem. Branches grow from nodes located higher up on the stem, after this latter starts elongating (after flowering induction) [7]. Both types of branching have been reduced in domesticated sorghum [7], foxtail millet [30], maize [25] and pearl millet [12] (Figure 1A). Absence of branching at a node can arise from various developmental defects related to different genetic networks [23]. The vegetative axillary meristem can either fail to initiate at the axil of the leaf, as observed in some foxtail millet varieties [7], or it can be arrested in its organogenic activity, like it is the case in maize [31], sorghum [20] and foxtail millet [7]. To investigate whether it is so in pearl millet as well, we dissected domesticated and wild plants at different stages of development. As illustrated in Figure 1C, tillers and branches fail to develop in domesticated plants due to the arrested activity of their vegetative axillary meristems which remain dormant either as meristems or as small buds with one or two leaf primordia. Therefore, branching adaptation during domestication has comparable developmental origins in maize, sorghum, foxtail and pearl millet, and could be caused by orthologues of the same genes involved in axillary meristem activity. By assembling a comprehensive comparative genetic map of QTLs for axillary branching in these four species (Figure 2), we observed that QTLs for branching reduction are consistently detected in the predicted region for Tb1 in sorghum and pearl millet, in addition to the previously described cases of association with the gene in maize [16] and foxtail millet [30]. These QTLs reflect adaptation of branching during both domestication (in ‘‘wild progenitor x cultivated landrace’’ crosses) and secondary crop diversification (in crosses between varieties). Interestingly, the Tb1 region of perennial species of sorghum also harbors QTLs for the production of rhizomes (Figure 2), which are structurally equivalent to underground tillers [7]. On the other hand, the Tb1 region is not associated to domestication QTLs in wheat (not shown) or rice (Figure 2), although transgenic experiments have shown that Tb1 orthologues of these species have conserved a role in tiller development [18]. This is consistent with the fact that the vegetative architecture of pooids (wheat) and ehrhartoid (rice) cereals was not altered by domestication. Instead, they produce a profuse number of tillers (and no upper branches), like their wild progenitors [7]. However, QTLs for tiller number map close to OsTb1 in crosses involving rice varieties that have been specifically selected for a low-tillering phenotype during secondary crop diversification (Figure 2). This comparative map also revealed that the genetic basis of branching adaptation during the domestication of sorghum and millets is in sharp contrast with maize. It involves multiple genes in addition to Tb1 , some of which have much stronger effects on the trait than Tb1 (Figure 2). As opposed to observations in maize, Tb1 effects in those species are usually moderate to low and sometimes depend on environmental conditions (e.g. in foxtail millet [32]). Altogether, these results suggest a consistent pattern of parallel evolution of vegetative branching in cereals based in part on the repeated selection of Tb1 , despite strong differences from a ...