Tree architecture: A strigolactone-deficient mutant reveals a connection between branching order and auxin gradient along the tree stem

Significance

What makes a tree a tree instead of a bush? Through a candidate gene approach, we identified a natural bush-like (short and highly branching) SL-deficient birch mutant, kanttarelli, with an early STOP codon in an essential SL biosynthesis gene, BpMAX1. The number of higher-order branches was increased in the mutant and in phenocopying transgenic RNAi -lines. Intriguingly, the auxin concentration formed a gradient along the main stem in the WT, with more auxin in the uppermost internodes and less toward the base, whereas in the transgenic line this gradient was absent. Mathematical modeling showed that this difference in auxin distribution may result from the differing architectures. Our results could be applied in the breeding of trees with an optimized architecture.

Abstract

Due to their long lifespan, trees and bushes develop higher order of branches in a perennial manner. In contrast to a tall tree, with a clearly defined main stem and branching order, a bush is shorter and has a less apparent main stem and branching pattern. To address the developmental basis of these two forms, we studied several naturally occurring architectural variants in silver birch (Betula pendula). Using a candidate gene approach, we identified a bushy kanttarelli variant with a loss-of-function mutation in the BpMAX1 gene required for strigolactone (SL) biosynthesis. While kanttarelli is shorter than the wild type (WT), it has the same number of primary branches, whereas the number of secondary branches is increased, contributing to its bush-like phenotype. To confirm that the identified mutation was responsible for the phenotype, we phenocopied kanttarelli in transgenic BpMAX1::RNAi birch lines. SL profiling confirmed that both kanttarelli and the transgenic lines produced very limited amounts of SL. Interestingly, the auxin (IAA) distribution along the main stem differed between WT and BpMAX1::RNAi. In the WT, the auxin concentration formed a gradient, being higher in the uppermost internodes and decreasing toward the basal part of the stem, whereas in the transgenic line, this gradient was not observed. Through modeling, we showed that the different IAA distribution patterns may result from the difference in the number of higher-order branches and plant height. Future studies will determine whether the IAA gradient itself regulates aspects of plant architecture.

Woody plants display two major forms in nature, growing as either a tree or a bush. A tree is tall, with a clearly defined main stem, while bushes are short without an evident main stem, and they produce an abundance of branches (1). Tree architecture is a three-dimensional concept describing all of the aerial elements of the plant, including its stem shape, branching pattern, leaves, and reproductive organs, all composing a functional general structure. The long lifespan of perennial woody species allows trees and bushes to grow taller and add higher-order branches annually (2–5). Vegetative shoot development occurs in repeating modules called phytomers, each composed of a leaf, an axillary meristem, and an internode (6). Axillary buds produced by axillary meristems can either stay dormant or branch out (Fig. 1A). As such, they provide a high degree of phenotypic plasticity for tree growth. The number, size, and angle of branches determine the overall form and architecture of a tree.

Fig. 1.

Identification of the natural bpmax1 mutant. (A) A schematic view of higher-order branches and other major components of a tree crown. (B) Phenotypes of 3-mo-old bushy birches grown in greenhouse conditions (Scale bar, 30 cm.) (C) Gene structure of BpMAX1; F and R are the primer positions designed for the RNAi construct; the kanttarelli mutation is highlighted in the fourth exon, showing a single nucleotide polymorphism (G -> T) causing an early stop codon (Scale bar, 500 bases.) (D) Histochemical GUS staining of pBpMAX1::GUS plants. Panel i) a representative shoot of in vitro pBpMAX1::GUS (Scale bar, 1 mm.) The upper right corner of the image features a magnified view, highlighting the axillary bud. Panel ii) a representative stem of in vitro pBpMAX1::GUS (Scale bar, 200 μm.) Panel iii) internode5 of a branch from a 2-mo-old greenhouse-grown pBpMAX1::GUS tree (Scale bar, 100 μm.) Tissues and xylem cell types in the vasculature are marked in the figure. Ray parenchyma cells are highlighted with red arrows. (E) qRT-PCR analysis of BpMAX1 expression, with the reference genes PP2A and Atub, n = 3, error bars show SEM. A t-test was performed to compare BpMAX1 expression. P = 0.078 (WT vs. RNAi2), P = 0.087 (WT vs. RNAi60). (F) Birch SL quantification throughout 2 mo, n = 9. Structures of CLA and 18-OH-CLA are shown in the WT column. (G) Germination response of Orobanche minor seeds to birch SLs, n = 3, error bars show SD. There are three germination peaks from a previously uncharacterized potential SL, 18-OH-CLA, and CLA at t_R 6.9 min, 9.8 min, and 11.9 min, respectively.

The process of architectural development differs fundamentally between tree species. Some species tend to have a relatively long period of unbranched growth characterized by very limited primary branching of the main stem. For instance, under greenhouse conditions, Populus spp. generally do not develop primary branches until they are over ca. 1.5 m in length (ca. 3 mo old). In contrast, some tree species already develop primary branches as seedlings; e.g., birch (Betula spp.) seedlings generally branch out when they are only ca. 25 cm (less than 1 mo old) under greenhouse conditions, which makes it possible to study complex architectural traits in young birch trees. Furthermore, birches have highly variable naturally occurring architectural forms; they display a wide architectural spectrum from small bushes to tall trees. In Betula spp., almost all of the axillary buds on the main stem grow out at a young age. Besides its architectural variance, birch is also attractive to researchers because of the multiple genomic (7) and transcriptomic resources (8) that are available, together with well-established genetic tools (9, 10). We therefore focused our study on the genetic basis of architectural development in trees on silver birch (Betula pendula). Through a candidate gene approach of several naturally occurring architectural birch variants, we identified one, kanttarelli, having an early STOP codon in an essential strigolactone (SL) biosynthesis gene, BpMAX1 (MORE AXILLARY GROWTH 1). While the role of SL phytohormones in plant architecture, especially their inhibition of branching, is well-established in annual herbaceous plants, it is less well understood in long-living tree species, which have more complex branching patterns (11, 12).

SLs are a structurally diverse class of carotenoid-derived plant hormones; to date, more than 30 different biologically active compounds have been identified (13, 14). They are also exuded by plants into the rhizosphere, where they promote symbiotic interactions with arbuscular mycorrhizal fungi (15, 16) and induce the germination of root parasitic plants of the Orobanchace and Striga genera (17, 18). Despite their structural variety, all of the elucidated structures of biologically active SLs are characterized by the presence of a butenolide lactone ring (D-ring). In canonical SLs, it is connected by an enol ether bridge to a second moiety that consists of a tricyclic lactone (ABC-ring); in noncanonical SLs, it is connected to other secondary moieties that are not as well conserved (19, 20). The enol ether is unstable and easily cleaved or hydrolyzed, so SLs exhibit a major common fragment ion at m/z 97 in liquid chromatography with tandem mass spectrometry (LC–MS/MS) reflecting the D-ring moiety.

In the SL synthesis pathway, MAX1, MAX3, and MAX4 are the three best-characterized SL biosynthesis genes in Arabidopsis. Previous studies have suggested that MAX3 and MAX4 are carotenoid dioxygenases, while MAX1 belongs to the cytochrome P450 family (21–23). Other growth regulators that have been implicated in the regulation of branching include auxin and sucrose. Auxin has been proposed to inhibit lateral branching through two parallel (and not mutually exclusive) mechanisms, a canalization-based and a secondary messenger model (24). Sucrose has also been indicated to trigger bud outgrowth and antagonize the effect of auxin and SLs on bud outgrowth (25, 26).

Our study explores the links between tree architectural development and SL content, auxin and sucrose distribution in WT, and SL-deficient birch lines. We combine physiological studies with a modeling approach to form a detailed analysis of the branching pattern in a tree species.

Results

kanttarelli Is a Naturally Occurring SL-deficient Birch Mutant.

To understand the genetic basis of tree architecture, we identified four natural bushy variants of B. pendula in Finland, namely kanttarelli, “Luutakoivu,” “Pöytäkoivu,” and “Luuta E8032” (Fig. 1B), and we studied their branching patterns under greenhouse conditions. During 3 mo of growth, all of the bushy variants were shorter than the designated WT V5834 (a B. pendula clone from the Finnish breeding population) in the greenhouse. Meanwhile, all the birches in this experiment formed first-order branches in 3 wk. Luutakoivu had more first-order branches than the WT. In contrast, Pöytäkoivu and Luuta E8032 had fewer first-order branches than the WT. Intriguingly, kanttarelli birch (from Kuopio, Finland, 60°90′N, 27°62′E) had a similar number of first-order branches and internodes in its main stem as the WT (SI Appendix, Fig. S1) even though the main stem of kanttarelli was shorter. By systematically searching in kanttarelli, Luutakoivu and Pöytäkoivu for disruptive mutations in candidate genes that share homology with previously identified branching-linked genes in other plant species (7), we identified a loss-of-function mutation due to an early stop codon in the fourth exon of BpMAX1, an ortholog of the Arabidopsis MORE AXILLARY GROWTH 1 (AtMAX1) gene (Fig. 1C and SI Appendix, Fig. S2A and Table S2) in the kanttarelli variant. In contrast to kanttarelli, we did not find disruptive mutations in the SL-related genes in the other bushy variants.

BpMAX1 is the only ortholog of AtMAX1 in the silver birch genome (SI Appendix, Fig. S2B); its protein sequence is 74.2% identical with the AtMAX1 protein (SI Appendix, Fig. S2C). To investigate the conservation of BpMAX1 and its promotor pBpMAX1, a complementation assay was performed in Arabidopsis. Both branch number (SI Appendix, Fig. S2 D and E) and plant height (SI Appendix, Fig. S2 F and G) phenotypes were rescued in the Arabidopsis max1 mutant by the expression of a pBpMAX1::BpMAX1 construct.

To assess the pattern of BpMAX1 expression, we generated a promoter-GUS reporter construct, pBpMAX1::GUS, using 2,215 bp of the genomic sequence upstream of the BpMAX1 transcription start site. In the shoot, pBpMAX1::GUS was expressed in both the vasculature and the axillary buds. In the vasculature, it was observed to be prominently expressed in the ray parenchyma cells (Fig. 1D). This result is consistent with published birch RNAseq data (8), where the BpMAX1 transcript also peaks in the xylem tissue (SI Appendix, Fig. S3 A and D). Furthermore, the other two orthologs of SL biosynthetic genes in birch, BpMAX3 and BpMAX4, are also expressed specifically within the xylem (SI Appendix, Fig. S3 B–D). Previous studies in Arabidopsis have shown that genes involved in SL production and response are expressed in both shoot and root vascular tissues (23, 27–29). This indicates that birch may be able to produce SL and/or its biosynthetic intermediates in the context of xylem tissues.

To test whether the kanttarelli phenotype results from the naturally occurring mutation in the BpMAX1 gene, we used RNA interference (RNAi) to generate transgenic birch lines with downregulated BpMAX1 expression in the designated WT. Suppression of BpMAX1 expression was achieved by generating an RNAi construct targeting a 336-bp fragment from the first exon to the beginning of the second exon (Fig. 1C). Transcription levels of BpMAX1 in the representative downregulation lines (RNAi60 and RNAi2) were reduced compared with WT (Fig. 1E).

To examine whether SL content was downregulated in the RNAi60 and RNAi2 lines, we profiled SL in root exudates of these lines as well as in kanttarelli. We observed that RNAi60, RNAi2, and kanttarelli produced very limited SL compared to the WT during 2 mo of growth in hydroponic solutions (Fig. 1F and SI Appendix, Fig. S4 A–D). Two SL compounds, carlactonoic acid (CLA) and 18-hydroxycarlactonoic acid (18-OH-CLA) (SI Appendix, Fig. S4A), were identified in the birch root exudates by comparing their retention times (t_R) in LC–MS/MS with the corresponding synthetic standards. We then confirmed their identity and function by testing their ability to promote the germination of O. minor seeds, a standard bioassay for SL activity (13, 18, 30). The germination assay also identified another active germination stimulant with a t_R of 6.9 min (Fig. 1G). The stimulant had major fragment ions at m/z 97 (SI Appendix, Fig. S4E), characteristic of a SL D-ring moiety, indicating that this is a previously uncharacterized SL. The biological assay not only verified the retention times of the birch SLs but also validated reduced SL levels in RNAi60, RNAi2, and kanttarelli, since their root extracts had much lower O. minor germination promotion activity (Fig. 1G). Collectively, these results indicate that kanttarelli is indeed a natural SL mutant and that we successfully generated SL downregulation lines by silencing BpMAX1.

SLs Affect Tree Architecture By Regulating Higher-order Branching.

To better characterize the bushy phenotypes of the various genetic lines with reduced BpMAX1 activity, we documented the growth of WT, RNAi60, RNAi2, and kanttarelli under greenhouse conditions for 1 y (Fig. 2A). Similarly to previous studies in other species (23, 30–32), the BpMAX1::RNAi lines and kanttarelli were shorter than the WT (SI Appendix, Fig. S5A). Unlike herbaceous SL mutants (in Arabidopsis, rice, and pea), which had more first-order branches (23, 30–32), the BpMAX1::RNAi lines had a very similar number of first-order branches as the WT (Fig. 2F). It should be noted that in WT birch seedlings, most of the axillary buds on the main stem usually form primary branches, so there is little opportunity for an increase in first-order branching. However, whereas the WT formed very few second-order branches during the first year, both RNAi60 and RNAi2 developed several second-order branches (Fig. 2G), indicating that SL deficiency promotes secondary axillary bud growth.

Fig. 2.

Increased higher-order branching in BpMAX1::RNAi lines and kanttarelli vs. WT. (A) Upper: Phenotypes of WT, RNAi60, RNAi2, and kanttarelli at week 14 in the greenhouse. Lower: Phenotypes of WT, RNAi60, RNAi2, and kanttarelli at week 32 in the greenhouse. The WT was held in place by a vertical stick. (B) Phenotype of 10-y-old WT-looking bred population of birch trees in the field. (C) A close-up view of the highest-order branching of a 10-y-old birch tree from the field. The branch order is indicated with red numbers. (D) Phenotype of a 13-y-old kanttarelli in the field. (E) A close-up view of the highest-order branching of a 13-y-old kanttarelli. The branch order is indicated with red numbers. (F) Quantification of the number of first-order branches of WT, RNAi60, RNAi2, and kanttarelli throughout 31 wk. The t-test was performed on the data from week 11 and week 31; 7 ≤ n ≤ 8; error bars show SEM. (G) Quantification of the number of second-order branches of WT, RNAi60, RNAi2, and kanttarelli throughout 31 wk. The t-test was performed on the data from week 11 and week 31; 7 ≤ n ≤ 8; error bars show SEM. (H) Quantification of the number of branches per order of a WT-looking bred population of birch trees (abbreviated WT in the figure) in the field, n = 20. (I) Quantification of the number of branches per order of kanttarelli, 3 ≤ n ≤ 4. (J) A schematic model of tree architecture: trees vs. bushes. The x-axis represents averaged branch orders; the y-axis represents plant height. Vertical thick blue lines are stems; the horizontal thin green (WT-looking bred population of birch trees, abbreviated WT in the figure) and purple (kanttarelli) lines are the first-order branches carrying higher-order branches. The further the horizontal thin lines reach, the higher order branches they have.

Another characteristic feature of SL deficiency in herbaceous species is reduced branch angle (33, 34). To evaluate this phenotype in birch, we performed a 3D terrestrial laser-scanning (3D-TLS) experiment (35–37) on greenhouse-grown 14-wk-old plants and again after the onset of dormancy in the first winter. The results showed that BpMAX1::RNAi lines had smaller branch angles than the WT (SI Appendix, Fig. S5 B and C). The shorter stems, higher branching order, and more erect branches in BpMAX1::RNAi resulted in a rounder and more compact form than the WT (SI Appendix, Fig. S5 D–G). These parameters were analyzed by the NaPPI facility (the National Plant Phenotyping Infrastructure) (38).

To better understand how SL deficiency influences the architecture of birch, we followed the development of tree architecture over time in a WT-looking bred population of birch trees and in the kanttarelli mutant. The bred birch population was grown in three stands in Southern Finland (Hollola): a 3-y-old, 6-y-old, and 10-y-old stand. We conducted 3D-TLS scans of these stands. The scanning data revealed that 3-y-old trees of the bred population (avg. height 1.3 m) rarely had any second-order branches, while the 6-y-old (avg. height 3.9 m) and 10-y-old trees (avg. height 9.2 m) had fourth-order branches (Fig. 2 C and H and SI Appendix, Fig. S5H).

To understand the role of BpMAX1 in defining tree architecture, we analyzed the branching pattern in young and old kanttarelli trees. We compared 6-y-old trees of the bred population with 4-y-old kanttarelli growing in an open-air hangar in Haapastensyrjä’s research infrastructure in Loppi, and 10-y-old trees of the bred population with 13-y-old kanttarelli growing in the field in Kuopio. The branching pattern of kanttarelli contrasted with the trees of the bred population. The scanning data revealed a maximum of sixth-order branches in 4-y-old mutant plants (avg. height 1.4 m) and a maximum of seventh-order branches in 13-y-old mutant plants (avg. height 3.4 m) (Fig. 2 E and I and SI Appendix, Fig. S5H), while the WT plants never exceeded fourth-order branching in the period of this study. In conclusion, we observed consistently higher-order branching in the kanttarelli mutant than in the bred population in both young and old trees. Additionally, kanttarelli exhibited smaller branch angles at young stages than trees in the bred population. In older trees, the branch angles of the two lines were similar (SI Appendix, Fig. S5 I and J). To quantify the higher-order branching phenotype over time, we computed the branch distribution model (Fig. 2J). Our model highlights that shorter stems and advanced higher-order branching are two elements that distinguish the architecture of a tree versus a bush.

SL Levels Affect the Auxin Distribution Along the Main Stem.

Classical decapitation experiments have shown that the removal of the shoot apex (the auxin source) can activate branching (39–41). To mimic the higher-order branching phenotype in the WT at an earlier stage, we performed a decapitation assay and explored the most efficient way to induce branching in birch. This strategy was chosen due to the complex architecture of birch. The assay included three treatments: i) stem tip removal; ii) first-order branch tip removal from the seven basal branches; iii) a combination of both treatments (SI Appendix, Fig. S6A). One-month-old WT and RNAi60 lines underwent all three treatments, and 1 mo after the treatments, we then measured how many second-order branches had grown on first-order branches that had been decapitated. Intact WT did not have second-order branches; some WT that underwent treatments i and ii had second-order branches; all the WT undergoing treatment iii had second-order branches at an early stage. In contrast, RNAi60 always had second-order branches, with or without the decapitation treatments (SI Appendix, Fig. S6 B and C).

Previous studies have indicated that auxin may act through SLs to facilitate apical dominance and that mutations impairing SL activity significantly increase polar auxin transport (PAT) from the apical to the basal end of excised stem segments (42–45). We therefore analyzed PAT in internodes from the upper half of WT, RNAi60, and kanttarelli stems. These three lines exhibited similar patterns of PAT, with decreasing PAT from the tip to the middle of the stems (SI Appendix, Fig. S7A). Quantification of internode3, internode7, and internode11 from WT and RNAi60 showed no statistical difference between stem cross-sectional area in the two lines (SI Appendix, Fig. S7 B and C).

To further understand the interaction between auxin and SL, we investigated the auxin content in the primary stem. Auxin profiling was conducted in two sets of experiments. In the first set (in 2021), IAA levels were quantified in internode3, internode7, and internode11 from the top to the middle of 2-mo-old trees (Fig. 3D). We observed a reduction in IAA concentration from the top to the middle in the WT. By contrast, we observed little, if any, reduction in RNAi60 and an increase of IAA concentration in the kanttarelli mutant (Fig. 3A). In the second set (in 2022), the IAA profile was investigated in both intact and decapitated trees to understand how auxin changes after the initiation of second-order branching. Treatment iii [with both stem and branch tips removed (SI Appendix, Fig. S6A)] was applied to 1-mo-old WT and RNAi60 trees. One month later, we sampled the decapitated trees and intact controls. Samples were taken from the tip to the base at four internode intervals (Fig. 3E). We observed an alteration of the auxin distribution pattern in the WT after decapitation. Decapitated WT plants with higher-order branches showed an increase in IAA concentration from the top to base (Fig. 3B), similar to the intact RNAi60 (in 2022) and kanttarelli previously tested in 2021.

Fig. 3.

Auxin distribution along the birch main stem. (A) Auxin concentration from the 3rd (uppermost part of the tree) to 11th internode (middle of the tree) of 2-mo-old WT, RNAi60, and kanttarelli trees in 2021, n = 5. (B) Auxin concentration from the 3rd (uppermost part of the tree) to 27th internode (base of the stem) of 2-mo-old WT and RNAi60 in 2022, n = 3. Decapitation was performed on 1-mo-old WT and RNAi60 trees. Sampling took place 1 mo after decapitation to give time for second-order branch development. As a control, 2-mo-old WT and RNAi60 were sampled together with the decapitated trees. (C) Sucrose distribution from the 4th (uppermost part of the tree) to 28th internode (base of the tree) of 2-mo-old WT and RNAi60 in 2022, n = 3. Decapitation and sampling took place at the same time as in Fig. 3B. The Pearson correlation coefficient was calculated for the relation between IAA/sucrose levels and the internode position. r indicates how far the data points are from the line of the best fit. p measures the likelihood of observing a correlation of this strength under the null hypothesis. (D) The sample scheme in 2021—auxin samples are in pink and unsampled internodes in gray. (E) The sample scheme in 2022—auxin samples are pink, sucrose samples blue, and unsampled internodes gray. More internodes are presented as a lighter gray dashed line at the bottom to represent the complete tree architecture.

Sucrose has been reported to be another major regulator of branching (25, 46). Therefore, we explored the sucrose distribution in trees and whether it was affected by decapitation. We found that sucrose is evenly distributed from the tip to the base in both intact WT and intact RNAi60 lines. Moreover, decapitation (treatment iii) did not significantly alter the sucrose distribution pattern in stems (Fig. 3C).

Based on our findings, we hypothesized that the structural differences between WT and RNAi60 may suffice to cause differences in the global auxin concentration pattern, since young expanding leaves at active shoot apices are a major source of auxin in shoots. We used computational modeling to test the plausibility of this hypothesis (Fig. 4). We formally described the following assumptions, which we call the long-distance signaling model (LSM): i) Auxin is synthesized at apices; ii) auxin is synthesized at positions on the branch where leaves are located; iii) auxin is transported basipetally; iv) the magnitude of transport is proportional to the auxin flux; and v) auxin decays at a fixed global rate. This basic LSM model conceptually follows previously published formalisms that simulate long-distance auxin-mediated signaling (47). Structurally, the two lines were designed to have the same number of first-order branches (Fig. 2F) and internodes (SI Appendix, Fig. S5K), following the quantitative phenotyping data of 2-mo-old plants under greenhouse conditions. The main differences were that the RNAi60 model had shorter internodes than the WT and that the RNAi60 model had second-order branches while the WT had none (Fig. 4 A and B).

Fig. 4.

Computational model simulating auxin transport for WT and RNAi60. (A) Model structure of intact WT and RNAi60 trees, and decapitated WT and RNAi60. RNAi60 has shorter internodes than the WT and more second-order branches than WT; decapitation induced second-order branch development in both lines. The schematic models are also examples of 3D simulations of auxin concentration values in the WT and RNAi60 reached steady state. The blue color denotes the auxin concentration. (B) Simulated trees represented as a connected graph of vertices representing branch segments, apices, and/or leaves. Six segments were used to represent an internode in the WT and four segments for an internode in RNAi60. (C) Each cell in the grid represents an individual simulation run under the specified parameter value settings until homeostasis. Red color denotes a positive and blue color a negative gradient of auxin concentration. (D) Stimulated auxin concentration values in model stems. Data are presented also in (A).

To model this, we ran a set of simulations with varying configurations of auxin dynamics parameter values, covering a range of plausible values. Each simulation was run until the auxin concentration values between simulation steps reached a steady state. Subsequently, we documented all simulation runs that matched the diverging global auxin concentration patterns that we measured in the WT and RNAi60 under greenhouse conditions. In Fig. 4C, we show a slice of the parameter space of our computational model. Specifically, in the WT, we expect auxin concentration to decrease toward the base of the stem (blue color), while in RNAi60, the auxin concentration increased along the stem (red color). A wide range of our in silico experiments generated global auxin concentration patterns along the stem that matched the auxin concentration values observed in vivo. We show an example rendering of our dynamic 3D simulation at homeostasis in Fig. 4A, where we used identical parameter value settings for both trees (p = 0.7, d = 0.155 in Fig. 4C). In this case, the blue color indicates the auxin concentration for a given segment, which declines from the tip to the base in the WT but increases in RNAi60. The auxin concentration values (Fig. 4D) along the model stems (Fig. 4A) were consistent with measurements of the trees under greenhouse conditions (Fig. 3B). This implies that under the same auxin synthesis and decay process, global auxin concentrations are more significantly reduced in the WT along the primary axis compared to RNAi60, which has shorter internodes and additional auxin sources provided by second-order branches.

The distribution pattern of auxin differs in the WT and RNAi60, likely because of architectural differences. We also observed that decapitation altered the distribution of IAA. Therefore, we modeled global auxin concentration patterns in both decapitated WT and RNAi60 using the same LSM method. The auxin concentration values (Fig. 4D) along the decapitated WT stems in the model (Fig. 4A) were consistent with the trend of IAA distribution in the experiments conducted under greenhouse conditions (Fig. 3B). The auxin concentration gradients in the models mainly result from structural differences, and the greater similarity of the auxin concentration in the decapitation simulations highlights the similar morphology of these plants (SI Appendix, Fig. S6C). In conclusion, the architectural differences between WT and RNAi60—the shorter internodes and increased second-order branching in RNAi60—are sufficient to cause different auxin distribution patterns.

Discussion

To address a major question in tree biology: “What separates a tree from a bush?,” we took advantage of architectural variation present in birch species. Through a candidate gene approach, we identified a natural mutant, kanttarelli, with deficient SL biosynthesis because of a mutation in BpMAX1. We found that pBpMAX1::GUS was expressed specifically in xylem ray parenchyma cells (Fig. 1D), indicating that ray parenchyma cells may contribute to SL synthesis. Previously, grafting experiments had suggested that SL synthetic genes control long-distance and graft-transmissible signals that move upward (21, 48–52). It is a reasonable hypothesis that SL intermediates travel in the xylem. Moreover, orobanchol, a SL compound, was detected in xylem sap in Arabidopsis and tomato (53) evidence that SLs are transported through the xylem, at least orobanchol. In this study, we have shown that BpMAX1 is strongly expressed in the xylem-associated ray parenchyma cells, an essential tissue for the storage and transport of water, nutrients, and nonstructural carbohydrates (NSC) (54). The spatial expression of BpMAX1 is consistent with the idea of acropetal movement of SLs in plants to regulate shoot branching. Furthermore, highlighting the diversity of naturally occurring SL forms in different plant species, we found that birch most probably contains previously unidentified noncanonical SLs (exhibiting mass fragments of m/z 97, m/z 139, and m/z 157) (SI Appendix, Fig. S4E), which have not been structurally studied in detail.

Due to the long lifespan of woody species, understanding tree architecture requires long-term observation. In this study, we conducted a detailed phenotyping of young greenhouse-grown SL-deficient birch saplings (Fig. 2 A, F, and G and SI Appendix, Fig. S5A). Subsequently, we tracked the development of tree architecture over time by 3D laser scanning of field-grown WT and kanttarelli trees (Fig. 2 B–E and H–J and SI Appendix, Fig. S5H). Taken together, our data show that the appearance of bushiness in the kanttarelli mutant is due to the combined effect of its shorter stature and accelerated higher-order branching under both greenhouse and field conditions.

Previous studies in various species have shown that SL mutants have significantly increased PAT, with 1.5-cm stem segments typically transporting on the order of 20% more auxin than the wild type in six-hour assays (42–45). However, our study revealed similar PAT in the upper half of stems from WT, RNAi60, and kanttarelli. This could reflect little or no effect of SLs on auxin transport in birch or insufficient sensitivity in the assay. The lack of a discernible difference in auxin transport is not due to gross differences in stem anatomy, since the upper half of the stem appears to have similar anatomy in both WT and RNAi60 (SI Appendix, Fig. S7 B and C). IAA profiling along the main stem demonstrated that IAA levels and internode position correlate negatively in the WT, consistent with the pattern expected if auxin is synthesized in the apex, transported basipetally, and decays at some rate. The IAA distribution pattern can be altered by decapitation treatment in the WT (Fig. 3 A and B), with an increase in the number of active apices correlating with increased auxin in more basal internodes, eliminating or even reversing the gradient seen in intact plants. Consistent with this idea, the distribution of auxin in SL-deficient birch is similar to that in decapitated WT trees. Indeed, our computational model suggests that shorter stature and increased second-order branching are sufficient to account for the different auxin distributions in the WT and the SL-deficient lines, with no requirement for differences in auxin synthesis, transport, or decay.

These results are consistent with data from Arabidopsis, though there are differences. In birch, the primary shoot is highly branched, with buds released from apical dominance as the primary shoot grows, increasing the distance between it and the axillary buds below it, resulting in a classical acropetal branching habit. In contrast, in Arabidopsis, apical dominance during vegetative growth is much stronger, which is likely in part due to the very short internodes of the rosette, resulting in the close proximity of all axillary buds to the active primary apex. At floral transition, rapid internode elongation, coupled with the suppression of leaf development, correlates with a dramatic reduction in apical dominance, with axillary buds released in a basipetal pattern (47). This branching is therefore conceptually similar to decapitation, and produces an auxin distribution pattern similar to those seen in decapitated birch (Fig. 3B).

The effect of SL deficiency in Arabidopsis on these auxin distribution patterns is striking. The gradient of accumulation of auxin down the main stem is much steeper in SL mutants than in the WT, with relatively little difference in auxin levels in apical segments, but approaching an order of magnitude more auxin in basal stem segments of SL-deficient mutants compared to the WT (47). This is qualitatively similar to the situation in birch, with a shift toward greater basal auxin accumulation along the stem in both cases (Fig. 3 A and B). However, in Arabidopsis, the magnitude of the shift cannot be explained solely by differences in stem length and active shoot number since these do not differ greatly between SL mutants and WT at the nodes where auxin has been measured, though the differences in auxin levels are large. In Arabidopsis, the auxin differences can be explained by changes in auxin transport, which is significantly increased in SL-deficient mutants. Interestingly, no differences in auxin transport were detected between WT and SL-deficient lines in birch. SL has at least two modes of action, and it is possible that in birch, its effects on gene expression are more important than any effects on auxin transport.

In conclusion, our study comparing the WT and genetic lines that have reduced BpMAX1 function highlights a different auxin distribution between the tree-like and bushy plants. It is possible that such a difference in auxin distribution could further contribute to various morphological adaptions in the two architectural forms. From an economical point of view, architecture strongly affects the value of a tree: A tall, straight primary stem with few, thin primary branches provides the best timber material for sawmill products, whereas fruit production and ornamental purposes benefit from the shorter forms with abundant branches. Consequently, we hope that our interdisciplinary study will facilitate genome-based breeding of SL-activity-optimized tree architecture for both timber and horticultural purposes.

Materials and Methods

Silver birch WT clone V5834 was used for transgenic plant generation. Plants for GUS-staining, qRT-PCR quantification, IAA profiling, NaPPI phenotypical analysis, auxin transport capacity analysis, and histological analysis were 2 mo old under greenhouse conditions. Plants for decapitation were 1-mo-old, while plants for second-order branch quantification and sampling were 2-mo-old, and plants for both assays were grown under greenhouse conditions. Plants for SL profiling were hydroponically. Details for the following are available in SI Appendix, SI Materials and Methods: genetic materials and growth conditions, histochemical staining for GUS activity, qRT-PCR, O. minor germination assay, SL profiling, laser scanning and scanning data analysis, branching distribution modeling, IAA profiling, sucrose profiling, computational model implementation, phylogenetic analysis, NaPPI phenotyping, decapitation assays, polar auxin transport capacity measurement, and stem area measurement. The primers used in this study are listed in SI Appendix, Table S1. The BpMAX1 gene sequence and kanttarelli mutation are given in SI Appendix, Table S2. The pBpMAX1 promotor sequence is given in SI Appendix, Table S3.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Dataset S04 (XLSX)

Dataset S05 (XLSX)

Dataset S06 (XLSX)

Dataset S07 (XLSX)

Dataset S08 (XLSX)

Dataset S09 (XLSX)

Dataset S10 (XLSX)

Dataset S11 (XLSX)

Dataset S12 (XLSX)

Dataset S13 (XLSX)

Dataset S14 (XLSX)

Data, Materials, and Software Availability

Code and script data have been deposited in ChangBio/kanttarelli (https://zenodo.org/badge/633505761.svg) (55). The modeling algorithms, computer code, and R code are available at https://github.com/ChangBio/kanttarelli.git (56). All other data are included in the manuscript and/or SI Appendix.