Patterns of shoot architecture in locally adapted populations are linked to intraspecific differences in gene regulation


  • Robert L. Baker,

    1. Department of Ecology and Evolutionary Biology, Campus Box 334, University of Colorado at Boulder, Boulder, CO 80309, USA
    2. Current address: Department of Botany, University of Wyoming, Laramie WY 80271, USA
    Search for more papers by this author
  • Lena C. Hileman,

    1. Department of Ecology and Evolutionary Biology, University of Kansas, 1200 Sunnyside Ave, Lawrence, KS 66045, USA
    Search for more papers by this author
  • Pamela K. Diggle

    1. Department of Ecology and Evolutionary Biology, Campus Box 334, University of Colorado at Boulder, Boulder, CO 80309, USA
    Search for more papers by this author

Author for correspondence:
Robert L. Baker


  • Shoot architecture, including the number and location of branches, is a crucial aspect of plant function, morphological diversification, life history evolution and crop domestication.
  • Genes controlling shoot architecture are well characterized in, and largely conserved across, model flowering plant species. The role of these genes in the evolution of morphological diversity in natural populations, however, has not been explored.
  • We identify axillary meristem outgrowth as a primary driver of divergent branch number and life histories in two locally adapted populations of the monkeyflower, Mimulus guttatus.
  • Furthermore, we show that MORE AXILLARY GROWTH (MAX) gene expression strongly correlates with natural variation in branch outgrowth in this species, linking modification of the MAX-dependent pathway to the evolutionary diversification of shoot architecture.


Branching is a fundamental process contributing to shoot architecture (Sussex & Kerk, 2001) and the evolution of morphological diversity among plants (Barthelemy & Caraglio, 2007; Bell, 2008) from the first branched sporophytes of early land plants (Remy & Hass, 1996) to intricate patterns of axillary branching among flowering plants. Shoot architecture is also variable within species, where it influences leaf placement and light interception (King, 1998; Bell, 2008), contributes to performance (Niinemets et al., 2004) and affects fitness (Lortie & Aarssen, 2000). Because both vegetative branches and flowers develop from axillary meristems, vegetative branching may preclude flowering, and trade-offs between flowering and branching can play a critical role in the evolution of life-history strategies (Geber, 1990; Bonser & Aarssen, 2006). Variation in the development of branches is subject to natural selection, is an important component of adaptation to novel environments (Bonser & Aarssen, 1996) and was central to the artificial selection that led to crop domestication (Doebley et al., 1997; Gepts & Papa, 2001; Weeden, 2007). Understanding the developmental genetic basis of intraspecific variation in branching is critical for understanding plant evolution and may provide important insights for further crop improvement. Despite its importance, however, shoot branching has not been studied from a molecular genetic perspective in natural populations, which harbor developmental variation and are where evolutionary processes such as selection act.

An excellent opportunity for studying natural variation in the developmental genetics of shoot architecture is provided by two locally adapted populations of the monkeyflower, Mimulus guttatus, which have well-characterized and contrasting life histories and patterns of branching. In their native environment, individuals from the coastal dunes of Oregon (DUN population) are perennials that branch frequently. In contrast, individuals from Iron Mountain in the Oregon Cascades (IM population) are alpine annuals that rarely branch (Hall & Willis, 2005; Hall et al., 2006). These contrasting branching patterns persist when plants are grown in a common environment and maternal effects are minimized (Baker & Diggle, 2011). Differences in branch number between populations are specific to the two basal-most nodes of the main axis and are expressed early during ontogeny; they are therefore not simply due to differences in lifespan (Baker & Diggle, 2011).

Here, we first demonstrate that population-level differences in DUN and IM branching patterns are caused primarily by differences in the frequency of axillary meristem outgrowth, as opposed to differences in axillary meristem initiation. Second, to further understand the developmental basis of divergent branching patterns, we examine expression of candidate genes that regulate branch outgrowth in these plants. In model species such as Arabidopsis thaliana MORE AXILLARY GROWTH (MAX) genes inhibit branch outgrowth and constitute one of the best-characterized and least pleiotropic pathways regulating shoot architecture (Stirnberg et al., 2002). In A. thaliana, there are four single copy MAX genes: AtMAX1 is a cytochrome p450 (Booker et al., 2005), AtMAX2 is an F-box gene (Stirnberg et al., 2007), and AtMAX3 and AtMAX4 are carotenoid cleavage dioxygenases (Sorefan et al., 2003; Booker et al., 2004). The protein products of AtMAX1, 3 and 4 function in roots to convert a β-Carotene precursor into strigolactone, an upwardly mobile hormone (Schwartz et al., 2004). Wild-type roots grafted to mutant shoots are sufficient to restore the wild-type branching pattern. However, reciprocal grafts demonstrate that wild-type AtMAX1, 3 and 4 in shoots are also sufficient to restore a wild-type phenotype (reviewed in Beveridge & Kyozuka, 2010). Mutant phenotypes are observed only when AtMAX1, 3 or 4 are knocked out in both shoots and roots (Turnbull et al., 2002; Sorefan et al., 2003; Booker et al., 2005). In contrast, AtMAX2 acts specifically in shoots at individual nodes to perceive the AtMAX1/3/4 signal and to negatively regulate shoot branching (Stirnberg et al., 2007) via modulating expression of AtPIN1 (PIN-FORMED1), an auxin efflux transporter (Bennett et al., 2006).

In A. thaliana, inactivation of any one of the four MAX genes results in a substantial increase in branch outgrowth compared to wild-type plants (Stirnberg et al., 2002; Booker et al., 2004; Bainbridge et al., 2005). MAX orthologs have been characterized in other model taxa including Pisum, Petunia, Oryza and Solanum, where they all function to negatively regulate outgrowth of axillary meristems (Sorefan et al., 2003; Snowden et al., 2005; Johnson et al., 2006; Zou et al., 2006; Ledger et al., 2010; Vogel et al., 2010; Waldie et al., 2010; Drummond et al., 2012), as well as in nonmodel organisms such as Actinidia chinensis (kiwifruit) where they are hypothesized to perform the same function (Ledger et al., 2010).

The broad conservation of MAX gene function across all flowering plants studied to date indicates a crucial role for the MAX pathway in the development of shoot architecture. However, these data are primarily from laboratory mutants and transgenic analyses of model organisms, and do not address the potential role of MAX genes, or a MAX-dependent pathway, in generating natural variation. Quantitative Trait Loci (QTL) and linkage disequilibrium studies indicate that AtMAX2 and 3 are associated with differences in branching patterns among A. thaliana accessions (Ehrenreich et al., 2007). Thus, although the role of MAX gene expression in M. guttatus or any natural population has never been assessed, the MAX pathway may be able to produce the spectrum of branching phenotypes observed within and among populations. Here we show that MAX expression correlates with contrasting patterns of branch outgrowth between two locally adapted populations of M. guttatus.

Materials and Methods

Description of species

Mimulus guttatus, DC (Phrymaceae) is widespread across much of western North America ( Mimulus guttatus is characterized by an acropetal, basitonic branching pattern with most branches occurring at the two basal-most nodes (Baker & Diggle, 2011). Previous studies report multiple serial axillary meristems per leaf axil (Moody et al., 1999). The distal, primary axillary meristem may develop into a branch or a single flower while the proximal, secondary axillary meristem often remains quiescent unless products of the primary axillary meristems are damaged (Moody et al., 1999).

We focus on two well-studied, locally adapted populations (Hall & Willis, 2006) from similar latitudes that differ in number and frequency of branches and in life history (Willis, 1993a,b; Sweigart et al., 1999; Willis, 1999; Fishman et al., 2002; Hall & Willis, 2005; Hall et al., 2006; Lowry et al., 2008, 2009; Lowry & Willis, 2010; Baker & Diggle, 2011). Plants from the DUN population are located at sea level in the Oregon Dunes National Recreation Area, and are perennials with numerous branches and flower relatively late (Hall & Willis, 2006; Baker & Diggle, 2011). Plants from the Iron Mountain (IM) population are short-lived, alpine (occurring at 1463 m) annuals that rarely branch vegetatively and produce one or occasionally two flowers (Hall & Willis, 2006).

Plant growth

Seed was collected, grown, transplanted and staged as previously described to facilitate comparisons to the morphological descriptions in Baker & Diggle (2011). We defined germination – which corresponds to week (wk) –1 in Fig. 1– as occurring when cotyledons were first visible above the soil surface. All tissues were collected at the same developmental stages as in Baker & Diggle (2011)– at germination, transplantation (wk 0), and weekly thereafter through wk 4.

Figure 1.

Axillary meristem (AXM) development at node 1. Only branches develop at node 1, and DUN plants produce significantly more branches than IM plants starting at wk 4 (a; error bars are 95% confidence intervals and arrow indicates time of transplant; adapted from Baker & Diggle, 2011, with permission from The Botanical Society of America). At germination (wk –1), there are no AXMs at node 1 (b–e) or node 2 (d and e) of DUN (b and d) or IM (c and e) plants. At wk 0, DUN plants have initiated a single AXM (AXM1; f and j), as have IM plants (g and k). At wk 1, the AXMs from DUN (h) and IM (i) plants are the same size and have both initiated prophylls (l (DUN) and m (IM)). By wk 2, DUN (n and r) and IM (o and s) have both initiated a secondary serial AXM that is subjacent to the primary AXM. The primary AXM in DUN plants has initiated multiple sets of leaves and begun to grow out as a branch (n and r) whereas the primary AXMs in IM plants have only prophylls (o and s). At wk 3, the DUN primary AXMs have grown out into branches (p and t) whereas IM primary axillary meristems often remained at the prophyll stage (q) or occasionally grew out into branches (u), albeit smaller than comparable DUN branches (t). At wk 3, secondary AXMs for DUN (p and t) and IM (q and u) did not grow out beyond the initiation of prophylls. SAM, shoot apical meristem; L1 and L2, leaves at node 1 and 2 (respectively) of the main axis; MA, main axis; AXM1, primary axillary meristem; AXM2, secondary axillary meristem; P, prophyll; Br, branch. Bars, 100 μM except inset in (t), 50 μM.


Tissue was collected from the first two nodes of a minimum of ten plants per population per time interval. ‘Node 1’ is the cotyledonary node. All tissue was prepared, stained and visualized as in Moody et al. (1999). Axillary meristems were recorded as present when a multicellular dome-like structure (Scanning Electron Micrographs; SEM) that was densely staining (Light Micrographs; LM) was observed in a leaf axil. Axillary meristems were considered vegetative buds once the first leaf pair (prophylls) was initiated or floral when the meristem began broadening and directly initiated 5 sepals without prophylls or other vegetative appendages. Vegetative axillary buds were considered vegetative branches (hereafter, ‘branches’) once they had initiated additional leaf pairs and internode expansion had begun.

Gene expression

Arabidopsis thaliana MORE AXILLARY GROWTH (MAX)1, 2, 3 and 4 were used as query sequences in tBLASTx searches of the M. guttatus genome (DoE Joint Genome Institute, annotation v1.1, assembly v1.0; Altschul et al., 1997). Matches with E values < 1 × 10−50 were considered putative M. guttatus MAX orthologs. In the case of AtMAX2, AtMAX3 and AtMAX4, there was only one putative M. guttatus ortholog for each A. thaliana gene (hereafter, MgMAX2, MgMAX3 and MgMAX4). There were two putative M. guttatus MAX1 orthologs. Based on identical scores and E-values, nucleotide sequence similarity (> 99%, including > 99% similarity in putative introns), and the location of one putative ortholog on a small, potentially orphan scaffold (scaffold 853), these sequences likely represent the same gene and do not reflect a duplication of MAX1 in M. guttatus (hereafter, MgMAX1; see Supporting Information Fig. S1). The cDNA sequences of MgMAX1, 2, 3 and 4 were confirmed in plants derived from both the DUN and IM populations of M. guttatus using the DUN10 and IM62 (the line used for the genome sequence) lines provided by the Mimulus seed stock center at Duke University. For candidate gene sequence confirmation, RNA was extracted from the shoot apicess of one DUN10 and one IM62 plant using RNeasy mini kits and treated with DNAseI following the manufacturer’s protocols (Qiagen). DNAse treated RNA was used to generate cDNA using SuperScript III (Invitrogen) reverse transcriptase and random hexamer primers according to the manufacturer’s protocols. Gene-specific primers designed to amplify 500 bp segments of each of the candidate genes were based on the draft M. guttatus genome sequence. PCR was performed using the Expand High Fidelity PCR system (Roche) following the manufacturer’s instructions using a gradient of 42–65°C and 40 cycles on an Eppendorf Mastercycler Gradient thermocycler. PCR products of expected size were gel extracted from three independent reactions for both DUN and IM cDNA templates (Qiagen Gel Extraction Kit) and cloned into the pCR 4-TOPO TA cloning vector, which was transformed into alpha-DH5 chemically competent cells (Invitrogen) following the manufacturer’s protocols and screened using carbenicillin. Plasmids were isolated from ten positive clones per PCR reaction using the Plasmid Mini Kit following the manufacturer’s instructions (Qiagen). Sequencing reactions were performed using M13F and M13R primers by Functional Biosciences. Vector and primer sequences were trimmed and base calls confirmed by visual inspection of chromatograms in Sequencher v4.6-4.10 (Gene Codes).

Tissue collection  MAX expression in Arabidopsis, Pisum and Petunia occurs in both shoots and roots. The highest levels of AtMAX1, 3 and 4 (and their respective orthologs in Pisum, Petunia, Orzya and Actinidia) are in roots, where these genes function (Sorefan et al., 2003; Booker et al., 2004; Snowden et al., 2005; Johnson et al., 2006; Zou et al., 2006). AtMAX1, 3 and 4 (and their orthologs in other species) are part of a pathway that synthesizes an upwardly mobile strigolactone hormone signal (Beveridge et al., 1996; Schwartz et al., 2004; Johnson et al., 2006; Gomez-Roldan et al., 2008; Umehara et al., 2008). AtMAX2 (and its ortholog in Pisum) is expressed in the shoot at nodes where it perceives the hormone generated by AtMAX3, AtMAX4 and AtMAX1, and negatively regulates branch outgrowth (Stirnberg et al., 2007).

Because a subset of MAX genes function in roots, we collected tissue from both roots and shoots. We assayed MgMAX1, 2, 3 and 4 expression at node 1 in DUN and IM plants and MgMAX1, 3 and 4 in roots of DUN and IM plants. We did not assay MgMAX expression at node 2 because at node 2 axillary meristem fate (floral vs vegetative in IM plants) contributes to differences in branch number (see Results section). It is unclear whether MAX genes are involved in floral meristem outgrowth and we did not want to confound our results concerning branch outgrowth with the additional variable of meristem fate. Therefore, node 1 (where all meristems in both populations were vegetative) is a much more promising location for examining MgMAX expression.

All tissue collected was staged to correspond to previously published data on the morphological development of branches in DUN and IM populations of (Baker & Diggle, 2011; Fig. 1a) and some samples from this previous study were also used for histology and SEM in the current study. Tissue was collected from two maternal families from each population that represented the most extreme branching phenotypes typical of each population (Baker & Diggle, 2011). Tissue was collected from 15 plants per population per time point. Tissue from 2 to 5 plants per population per time point was pooled into a single biological sample. Biological samples were replicated three times for each population, time point and category of tissue collected. Shoot tissue was collected by quickly dissecting out node 1 of main axis of DUN and IM plants. Subtending leaves were trimmed away but all axillary products were left intact. Roots were quickly cleaned of all substrate under cold water. Roots from each plant were divided into two samples: (1) the lower half (R2), from the distal end of the root mass to half the length of the roots, and (2) upper half (R1), from the base of the hypocotyl to half way down the length of the root mass because in Petunia, DECREASED APICAL DOMINANCE1 (dad1) – the Petunia MAX4 ortholog – expression decreases from distal to proximal root tissue (Snowden et al., 2005). At wk 1, all root tissue was collected in a single sample (Ra).

Quantitative reverse transcription PCR (qRT PCR)  RNA for qPCR was extracted from 10 to 100 mg of tissue from each sample using RNeasy mini RNA extraction kits according to the manufacturer’s protocol (Qiagen). All RNA samples were treated with TURBO DNA-free DNAse (Ambion) and RNA concentration was verified with a Qubit v2.0 fluorometer (Invitrogen). A subset of extractions was concentrated for 20 min at room temperature in a vacufuge (Eppendorf, Hamburg, Germany). RNA quality was assessed using an Agilent 2100 BioAnalyzer; samples with strong 26S and 18S ribosomal RNA peaks that also did not exhibit degradation were reverse transcribed into cDNA (+ RT) using 1.0 μg RNA and an iScript reverse transcriptase kit (including random primers) according to the manufacturer’s protocol (BioRad). Ten randomly selected DNAse-treated RNA samples were used for additional cDNA synthesis reactions without reverse transcriptase (− RT). Actin primers (that do not span an intron) were used in quantitative RT PCR (protocol described below) with + RT and − RT templates in tandem (Scoville et al., 2011). Amplification from + RT cDNA but not − RT samples indicated successful cDNA synthesis and absence of contaminating genomic DNA.

Quantitative reverse transcription PCR (qRT PCR) was performed according to standard methods (Vandesompele et al., 2002; Scoville et al., 2011) on shoot (node 1) and root tissue. All qRT-PCR reactions were performed on an Agilent Technologies Mx3005P thermocycler (Stratagene, La Jolla, CA, USA) using 1.0 μl of cDNA template, diluted 1 : 10 in water, Brilliant III Ultra Fast SYBR Green QPCR Master Mix (Agilent Technologies, Santa Clara, CA, USA) with a final concentration of 1×, and a final concentration of 500 nM for all primers. All reactions were performed using a Master Mix from a single lot.

Gene-specific primer pairs for each candidate gene were designed using Primer3 (v0.4.0) software to amplify short (100–300 bp) DNA segments. Primers were designed based on inclusive DUN and IM consensus sequences from IM and DUN plants (primer sequences may be found in Table S1) to bind without mismatch to coding regions that are conserved between the populations but otherwise represent unique sequences within the genome, thereby minimizing potential population biases in the reactions. PCR amplicons spanned putative introns when possible (MgMAX1, 3 and 4; MgMAX2 lacks introns). Six sets of primer sequences for five housekeeping reference genes (one primer pair for ubiquitin (Ubq5), actin (Act), 26S ribosomal RNA (26S), and ribosomal binding protein L2 (L2), and two primer pairs for elongation factor 1α (EF1α, EF1α2)) previously identified from the M. guttatus draft genome (Scoville et al., 2011) were also evaluated.

Optimal annealing temperatures and reaction efficiencies were established for primers. Each housekeeping gene primer pair was amplified using a gradient of annealing temperatures from 52 to 62°C in triplicate. Primers that did not amplify a single target (Ubq5 and EF1α2) as assessed by disassociation curves without a single, clean peak were not considered for further evaluation. All remaining housekeeping primers amplified a single product consistently at 60°C, and all future amplifications were conducted with a 60°C annealing temperature. All gene-of-interest primers also consistently produced a single amplicon with a 60°C annealing temperature. Reaction efficiencies were established for the remaining primers for housekeeping genes and all primers for genes of interest by conducting qRT PCR using four separate serial dilutions of (1 : 10, 1 : 100, 1 : 1000, 1 : 10 000) of the same cDNA template in triplicate. Reaction efficiencies were calculated by fitting a line to the number of cycles (c(T)) vs log(relative quantity) using MxPro QPCR software (Stratagene); reactions that failed or amplified inconsistently were not included in the calculations. Efficiencies (E) for each qPCR were calculated using the equation inline image (Rasmussen, 2000). Reaction efficiencies between 94% and 106% were treated as ≤ 100% for calculations of relative gene expression.

The housekeeping genes with the most stable expression patterns were determined by amplifying DNA from a subset of DUN and IM cDNA samples representing roots, shoots and the two most extreme developmental time points. qRT PCR was conducted in triplicate and the c(T) values were converted to relative quantities for outlier analysis using the geNORM visual basic application (Vandesompele et al., 2002).

We assessed relative expression levels of MgMAX1, 2, 3 and 4 in tissue from node 1 and relative expression levels of MgMAX1, 3 and 4 in root tissue samples. For each time point and tissue type, we performed qRT PCR for genes (See Table S1 for gene-of-interest primer sequence) of interest and the two most stably expressed reference genes (EF1α and Actin) on three biological samples from the DUN and IM populations in triplicate on a single plate along with no template negative controls for each primer pair, in triplicate. The critical threshold levels were set automatically for each of the plates.

We calculated relative gene of interest expression levels normalized with the geometric mean of the two most stable reference genes (EF1α and Actin) following the delta c(T) method described in (Scoville et al., 2011). Samples that had evidence of spurious amplification (indicated by multiple peaks in the disassociation curves or incorrect melting temperatures) were excluded form the study. Samples that failed to amplify were also excluded from the study, unless all three technical replicates from the same sample failed to amplify, which indicated a relative quantity of zero. For each gene of interest, time point and tissue type, we calculated relative expression levels by subtracting the minimum critical threshold (c(T)) value of each gene from the c(T) value of each reaction to generate delta c(T) values. Relative expression quantities were calculated by raising the specific efficiency of each primer pair (E) to the –delta c(T) of each reaction. The relative quantity of the triplicates was averaged to produce a mean relative quantity for each gene, tissue type, time point and biological replicate. The relative quantity of each biological replicate was normalized by dividing by the geometric mean of the two reference genes for that replicate (Vandesompele et al., 2002). For each gene of interest, tissue type and time point, normalized expression levels were divided by the sample with the highest normalized expression level, thereby setting the expression level in the sample with the highest expression to 1 for ease of interpretation (Scoville et al., 2011). Comparisons within a tissue type and time point do not require adjustment for inter-plate variability because they were run on a single plate. Nonoverlapping standard errors were considered evidence of significant differences at the < 0.05 level (Cumming et al., 2007).


Meristem outgrowth explains differences in branch number at node 1

In all DUN and IM plants examined, two serial axillary meristems (a distal, primary and a proximal, secondary axillary meristem) were observed at each leaf axil at the first node of the main axis. For plants of both populations, the primary axillary meristem was always vegetative. The secondary axillary meristem was also vegetative, initiated after and proximal to the primary axillary meristem, and never developed > 1 set of leaves (prophylls).

At germination (wk –1) neither DUN nor IM plants bore axillary meristems at node 1 (Fig. 1b–e). Plants from both populations had initiated a single primary axillary meristem in each leaf axil by wk 0. These axillary meristems were approximately the same size, and leaf buttresses were visible, indicating that the axillary meristems were initiated at the same time (Fig. 1f–g,j–k). By wk 1, the primary axillary meristem of plants from DUN and IM populations were clearly vegetative, both having a pair of prophylls (Fig. 1lm).

Differences in axillary bud development were first apparent at wk 2 (Fig. 1n–o,r–s). The primary bud of DUN plants was larger than that of IM plants, and bore multiple leaf pairs, indicating that it had begun growing out as a branch (Fig. 1n,r). In contrast, the prophylls of axillary buds of IM plants had not expanded and no new leaf primordia were initiated (Fig. 1n,s). At wk 2 plants from both populations had initiated a secondary axillary meristem in each leaf axil, proximal to the primary vegetative axillary bud.

By wk 3, internodes of the primary DUN axillary bud had expanded as it continued to grow out as a branch (Fig. 1p,t). In contrast, at wk 3, the primary vegetative bud on the majority of IM plants had initiated only prophylls (Fig. 1q), or occasionally multiple leaf pairs (Fig. 1u) but showed no internode expansion. At wk 3, the secondary axillary meristems of both DUN and IM plants had initiated only prophylls (Fig. 1t–u) and the prophylls were roughly the same size.

Meristem fate and outgrowth explain differences in branching at node 2

At node 2, DUN and IM plants initiated primary and secondary axillary meristems at similar times. Differences in branch number at node 2 were caused by both differences in axillary meristem fate (always vegetative in DUN plants vs occasionally floral in IM plants) and the earlier and more frequent outgrowth of DUN vegetative axillary meristems as branches (Fig. 2).

Figure 2.

Axillary meristem (AXM) development at node 2. IM plants produce branches and occasionally flowers; DUN plants produce only branches and significantly more of them than IM plants (a; error bars are 95% confidence intervals and arrow indicates time of transplant; adapted from Baker & Diggle, 2011, with permission from The Botanical Society of America). (b and c) SEM of node 2 at wk 2. (b) DUN. (c) IM. (d and e) longitudinal sections through node 2 at wk 3. (d) DUN. (e) IM. (f and g) longitudinal sections through node 2 at wk 4. (f) DUN. (g) IM. (h and i) SEMs of node 2 at wk 4. (h) DUN. (i) IM. L2, leaves at node 2 of the main axis; MA, main axis; AXM1, primary axillary meristem; AXM2, secondary axillary meristem; P, prophyll; Br, branch. Bars, 100 μM.

At germination (wk –1), plants from both populations had already initiated a second set of leaf primordia, but lacked meristems in the axils of these primordia (Fig. 1d,e). By wk 0, plants from both populations had initiated a single axillary meristem at node 2. At wk 1, plants from both populations had only a single primary axillary meristem per leaf axil, and it was not until wk 2 that the axillary meristems of plants from both populations had initiated prophylls and become axillary buds (Fig. 2b,c). Plants from both populations had also initiated a secondary, serial axillary meristem proximal to the primary axillary bud by wk 2 (Fig. 2b,c, see insets). At wk 3, the primary axillary bud on DUN plants had initiated three leaf pairs whereas the primary vegetative axillary bud on IM plants bore only unexpanded prophylls (Fig. 2d,e). By wk 4, internodes of the primary vegetative axillary bud of DUN plants had expanded, resulting in an axillary branch (Fig. 2f,h). On some IM plants, primary vegetative axillary buds had grown out into branches (Fig. 2a,g); however, many remained arrested as vegetative axillary buds with only prophylls visible (Fig. 2i). Primary axillary meristems of some IM plants developed into a floral bud at node 2 and occasionally this floral bud became necrotic and failed to develop. By wk 4, the secondary vegetative axillary buds on DUN plants had initiated a second leaf pair (Fig. 2h) whereas secondary vegetative axillary buds on IM plants had not and, in some cases, had not yet initiated prophylls (Fig. 2i, inset). Secondary axillary meristems at node 2 from both populations were always vegetative.

Gene expression analyses

Because differences in branch number at node 2 are caused by axillary meristem outgrowth and fate (Fig. 2), we examined candidate gene expression only at node 1 where axillary meristem outgrowth alone determines branch number (Fig. 1). MgMAX expression for DUN plants was very low at some stages (e.g. DUN MgMAX3,Fig. 3), but was never zero. MgMAX1 expression did not correlate with branching phenotype (Fig. S2).

Figure 3.

Relative expression levels of Mimulus guttatus MORE AXILLARY GROWTH2, 3 and 4 genes during development in plants from the Dunes (DUN; solid bars) and Iron Mountain (IM; striped bars) populations. (a) In shoot tissue (node 1). (b) MgMAX3 and 4 expression in roots at wk1 (b), and the top half (c) and bottom half (d) of roots at wks 2–4. Error bars are standard errors and asterisks denote significant (< 0.05) differences in expression.

MgMAX2, 3 and 4 expression in shoots correlates with branching phenotype  At node one, DUN plants expressed one or more of the MgMAX genes at significantly lower levels than IM plants throughout development. IM plants never expressed any of the four MgMAX genes at significantly lower levels than DUN plants (Figs 3, S2). DUN plants expressed MgMAX2 at significantly lower levels than IM plants at wk 2 and expressed MgMAX3 and 4 at significantly lower levels than IM plants at wks 1–4 (Fig. 3a).

MgMAX3 and 4 expression in roots correlates with branching phenotype  At wk 1, there were no significant differences between DUN and IM MgMAX expression in root tissue (Fig. 3b). In the top (proximal) half of roots, DUN plants expressed both MgMAX3, and 4 at significantly lower levels than IM plants at wk 2 (Fig. 3c). In the bottom half of roots, DUN plants expressed MgMAX3 at a significantly lower level than IM plants at wk 4 (Fig. 3d).


Mimulus guttatus plants from the DUN population are perennials that flower relatively late in ontogeny and produce many vegetative branches. In contrast, M. guttatus from the IM population are annuals that flower early and have few vegetative branches (Hall & Willis, 2006). Differences in branch number between DUN and IM plants occur early in ontogeny and primarily at the two basal-most nodes on the main axis (Baker & Diggle, 2011). In plants from both populations, primary meristems at the first two nodes are initiated at the same time; differences in branch number are not due to differences in the timing or frequency of axillary meristem initiation.

At node 2, branch number differences between populations are caused by both meristem fate (solely vegetative in DUN plants, vegetative and floral in IM plants) and vegetative axillary meristem outgrowth (Fig. 2). At node 1 on the main axis, however, all primary axillary meristems on both DUN and IM plants, regardless of whether they grow out or not, initiate prophylls and are clearly vegetative (Fig. 1a–u). Therefore, differences in branch number at node 1 cannot be attributed to contrasting meristem fate. Rather, population level differences in branch number at node 1 are entirely due to differences in axillary meristem outgrowth vs quiescence, including onset (earlier in DUN plants) and frequency (greater in DUN plants; Fig. 1a,n–u). The differential outgrowth of meristems into vegetative branches is first detected at wk 2, well before differences in branch number are visible at the macromorphological level (wk 4).

Levels of MgMAX2, 3 and 4 expression are negatively correlated with branch outgrowth in DUN and IM populations of M. guttatus, which is consistent with the function of all known MAX orthologs in inhibiting branch outgrowth (Napoli & Ruehle, 1996; Stirnberg et al., 2002; Sorefan et al., 2003; Booker et al., 2004, 2005; Snowden et al., 2005; Auldridge et al., 2006; Johnson et al., 2006; Arite et al., 2007; Vogel et al., 2010). Specifically, at wk 2 we observe a sharp decrease in MgMAX2 expression in DUN shoots (relative to IM shoots). At the same time, there is a similar decrease in DUN MgMAX3 and 4 in the top half of roots (relative to IM roots). The decrease of MgMAX2, 3 and 4 expression in DUN relative to IM plants coincides temporally with axillary meristem outgrowth in DUN plants. The pattern of MgMAX2, 3 and 4 expression is strong evidence suggesting that the MgMAX pathway represses branch outgrowth and is involved in regulating variation in branch phenotype between these two naturally occurring populations.

In model taxa, MAX gene function has been studied in genetically homogeneous lines. In contrast, our study populations likely harbor considerable background genetic variation that likely contributes to substantial variation observed in branch number (Figs 1a, 2a) and much of the variation in MgMAX1, 2, 3 and 4 expression (Figs 3, S2). Despite the high levels of variation in expression we observed, all instances of significant differences in MgMAX2, 3 and 4 expression were in the predicted direction: lower levels of MAX expression occurred in the relatively highly branched DUN plants.

Despite, or perhaps because of, the high level of functional conservation, the role of MAX genes (unlike MADS-box and many other developmental genes; reviewed in Becker & Theißen, 2003) in the evolution of morphological diversity has not been well explored (Drummond et al., 2012). Our results, and those of other studies, indicate that these genes are particularly promising candidates for functional studies of natural variation in branch outgrowth leading to evolutionary transitions in plant architecture. For instance, in A. thaliana, MAX2, 3 and 4 do not function solely as binary switches regulating branch outgrowth. Instead, QTL studies (MAX2 and 3) and heterozygous mutants (MAX3/max3 and MAX4/max4) with intermediate phenotypes demonstrate the ability of MAX2, 3 and 4 to regulate branching in a quantitative, dosage-dependent manner (Booker et al., 2004; Auldridge et al., 2006; Ehrenreich et al., 2007). Branch number varies both within and between DUN and IM populations, indicating that the mechanism(s) that control natural variation in M. guttatus branch outgrowth also may act in a dosage-dependent manner similar to that of the MAX pathway in A. thaliana. Although the appropriate QTL studies of branch outgrowth remain to be done in M. guttatus, the association of MAX expression with contrasting patterns of branch number in natural populations implies a role for MAX expression in generating natural variation in branch number and therefore a role for MAX genes in the evolution of morphological diversity in shoot architecture (summarized in Fig. 4). Of course, MAX genes do not act in isolation. Regulation of branch outgrowth is a complex and multigenic phenomenon involving many non-MAX genes such as BRANCHED1 and BRANCHED2 and various phytochromes (Aguilar-Martinez et al., 2007; Finlayson et al., 2010). These genes may also be involved in variation in branching within and among lineages.

Figure 4.

Proposed mechanism for divergent development of branches in DUN and IM populations of M. guttatus. (a) Plants from both populations initiate primary (distal) axillary meristems at the same location, frequency and ontogenetic stage. Subsequently, plants from both populations also initiate secondary (proximal) axillary meristems subjacent to the primary axillary meristem at the same location, frequency and ontogenetic stage. (b) Low MAX expression in DUN plants (relative to IM plants) releases repression of primary axillary meristem outgrowth, allowing these meristems to produce branches. High MAX expression in IM plants (relative to DUN plants) represses primary axillary meristem outgrowth, preventing these meristems from growing out as branches. (c) The cumulative result of differential axillary meristem outgrowth generates highly branched plants typical of the DUN population and unbranched plants typical of the IM population.

DUN and IM plants both initiated secondary meristems in each leaf axil. Regardless of MgMAX1, 2, 3 and 4 expression, these secondary meristems remained quiescent in both populations. The MAX pathway may not be involved in secondary meristem development. Alternatively, the differential activities of these meristems may be related to the development of vascular tissue. The strigolactone signal generated by AtMAX1, 3 and 4 appears to be conducted through the plant vasculature (Kohlen et al., 2011). The vasculature associated with the primary axillary meristems in M. guttatus is well differentiated and likely capable of transporting signal molecules when meristems begin to develop into branches. In contrast, associated secondary axillary meristems lack differentiated conducting tissues (Fig. 1n) and may not receive strigolactone (or other) signals. Finally, more precise spatial regulation of MgMAX2 specific to the individual meristem may control primary vs secondary axillary meristem outgrowth.

In order to fully understand the evolution of plant development, it is necessary to examine developmental genetic pathways that control well-characterized and evolutionarily relevant morphological variation within and among populations, where genetic divergence, adaptation and speciation occur (Cresko et al., 2007; Johnson, 2007). We demonstrate that the MAX pathway, which is conserved across at least 150 million years of evolutionary history and is integral to plant body plan development, also is strongly associated with contemporary, natural variation in branch outgrowth in M. guttatus. We provide some of the first quantitative molecular genetic data from plants that associates variation in gene expression with well-characterized developmental events to explain differences in a functionally and evolutionarily central aspect of morphology. Trade-offs between branching and flowering can influence flowering time and life history evolution (Geber, 1990; Zopfi, 1995; Bonser & Aarssen, 1996, 2006; Baker & Diggle, 2011). In M. guttatus, flowering time, a critical component of local adaptation in DUN and IM populations (Hall & Willis, 2006), is negatively correlated with branch number (Hall et al., 2006; Baker & Diggle, 2011). Because of the tight relationship between flowering, branching and life history (Zopfi, 1995; Bonser & Aarssen, 1996; Prati & Schmid, 2000; Bonser & Aarssen, 2006), branching, in addition to flowering time, likely contributes to local adaptation. Our data imply a role for the MAX pathway in regulating developmental variation in shoot architecture among populations and therefore in the evolution of divergent, and ecologically important, life history traits.


The authors thank W Friedman and the Arnold Arboretum of Harvard University for generous access to facilities, equipment, and supplies; F Rosin for invaluable technical insight; and W Friedman, F Rosin, J Bachelier, R Povilus, W Adams, D Stock, R D Baker and S S Baker for helpful comments and discussion. J Willis, J Kelley, B Blackman, D Lowry, J Mojica and J. Preston shared important insights, provided seed for preliminary work, and helped with field logistics. Three anonymous reviewers contributed helpful comments. Funding was provided by the Dept. of Ecology & Evolutionary Biology at CU Boulder, Beverly Sears Graduate Student Grant, The Botanical Society of America, The CU Museum of Natural History Walker van Riper Fund, Molecular and Organismal Research in Plant History (MORPH) Research Coordination Network, and a Sigma Xi Grant in Aid of Research.