Elucidating connections between the strigolactone biosynthesis pathway, flavonoid production and root system architecture in Arabidopsis thaliana

Abstract Strigolactones (SLs) are the most recently discovered phytohormones, and their roles in root architecture and metabolism are not fully understood. Here, we investigated four MORE AXILLARY GROWTH (MAX) SL mutants in Arabidopsis thaliana, max3‐9, max4‐1, max1‐1 and max2‐1, as well as the SL receptor mutant d14‐1 and karrikin receptor mutant kai2‐2. By characterising max2‐1 and max4‐1, we found that variation in SL biosynthesis modified multiple metabolic pathways in root tissue, including that of xyloglucan, triterpenoids, fatty acids and flavonoids. The transcription of key flavonoid biosynthetic genes, including TRANSPARENT TESTA4 (TT4) and TRANSPARENT TESTA5 (TT5) was downregulated in max2 roots and seedlings, indicating that the proposed MAX2 regulation of flavonoid biosynthesis has a widespread effect. We found an enrichment of BRI1‐EMS‐SUPPRESSOR 1 (BES1) targets amongst genes specifically altered in the max2 mutant, reflecting that the regulation of flavonoid biosynthesis likely occurs through the MAX2 degradation of BES1, a key brassinosteroid‐related transcription factor. Finally, flavonoid accumulation decreased in max2‐1 roots, supporting a role for MAX2 in regulating both SL and flavonoid biosynthesis.


| INTRODUCTION
As sessile organisms, plants require roots to anchor themselves, absorb water and nutrients and interact with microbiota in the rhizosphere. Soils are inherently heterogenous, with varying water and nutrient levels. The architecture of the root system exhibits high developmental plasticity to adapt to these environmental stressors.
The primary and lateral roots (LRs) of plants are able to modify their directionality and growth for more effective nutrient mining, known as 'root foraging' (Motte & Beeckman, 2019;Waidmann et al., 2020).
On a finer scale, root hairs (RHs) extending from root epidermal cells also finetune the uptake of nutrients and water through increasing root surface area (Gahoonia & Nielsen, 2003). RH growth and development is the result of dynamic interplay within a gene regulatory network, which integrates both developmental and environmental signals (Shibata & Sugimoto, 2019).
As with aboveground growth, phytohormones-and the crosstalk between them-are heavily involved in coordinating root system architecture in response to endogenous and exogenous signals. For example, auxin is a positive regulator of both LR and RH growth, and strigolactones (SLs) functionally interact with the transport of auxin by affecting the abundance and polarisation of PIN-FORMED (PIN) auxin efflux carriers (Bennett et al., 2016;Shinohara et al., 2013;Weijers & Wagner, 2016).
Despite being the most recently discovered class of phytohormones, SLs are already known to have major functions in plant development and plant-microbial communications (Bhoi et al., 2021). One of the first of many identified roles of SLs in plant development was the finetuning of shoot architecture, hence the name "MORE AXILLARY GROWTH" (MAX) given to some of the key Arabidopsis thaliana SL biosynthesis genes. It has been suggested that SLs inhibit shoot branching, acting antagonistically against the cytokinin promotion of bud growth (Gomez-Roldan et al., 2008;Umehara et al., 2008).
More recently, the SL control of auxin transport has also been implicated in shoot branching regulation (Van Rongen et al., 2019). It has previously been suggested that SL addition has an effect on root system architecture in A. thaliana, such as by promoting RH elongation, and decreasing LR growth (Jiang et al., 2016;Kapulnik et al., 2011;Ruyter-Spira et al., 2011). However, much of previous research regarding phenotypic and transcriptomic responses to SL addition has been performed using the SL analogue rac-GR24, a racemic mixture of two GR24 enantiomers GR24 5DS and GR24 ent-5DS . It was later revealed that whilst GR24 5DS preferentially binds and activates the SL receptor DWARF14 (D14), GR24 ent-5DS can activate the karrikin receptor KARRIKIN INSENSITIVE 2 (KAI2) as a karrikin analogue (Scaffidi et al., 2014). Thus, the use of rac-GR24 means that the exact signalling pathway responsible for certain gene regulatory networks and phenotypes remains elusive.
The process of SL biosynthesis in A. thaliana involves the successive action of MAX proteins. First, the carotenoid isomerase DWARF27 (D27) converts all-trans-β-carotene into 9-cis-β-carotene (Abuauf et al., 2018;Booker et al., 2004). MAX3, also termed carotenoid cleavage dioxygenase-7 (CCD8), then cleaves 9-cis-β-carotene into 9-cis-β-10 0 -carotenal, which is then further cleaved by MAX4 (aka carotenoid cleavage dioxygenase-8 [CCD8)]), into carlactone (Sorefan et al., 2003). MAX1 is a Class III cytochrome P450 monooxygenase that catalyses the oxidation and hydroxylation of carlactone in the cytosol, producing SLs (Booker et al., 2005). SLs then move to other cells through PLEIOTROPIC DRUG RESISTANT1 (PDR1) transporters to induce downstream SL responses (Borghi et al., 2016). When SL enters a cell, it is first perceived by the downstream SL receptor and the α/β-fold hydrolase DWARF14 (D14). MAX2 is an F-box protein involved in targetting proteins for degradation as a subunit of an S phase kinase-associated protein 1-cullin-F-box (SCF) E3 ubiquitin ligase (Stirnberg et al., 2007). When SL is bound to D14, D14 can either signal for downstream SL responses or be degraded by SCF MAX2 , thereby limiting the intensity and duration of SL signalling. SCF MAX2 also recruits and degrades other proteins than D14, including SUPPRESSOR OF MAX2 1 (SMAX1) and SMAX1-LIKE (SMXL) family proteins, which regulate downstream SL responses (Soundappan et al., 2015). It was found that the transcription factor BRI1-EMS-SUPPRESSOR 1 (BES1) is both able to interact with and be degraded by MAX2 (Wang et al., 2013). In the presence of SL, evidence suggests that D14 activates BES1 degradation via the SCF MAX2 complex, thereby regulating downstream signalling. BES1 is most well-known as a positive regulator in the brassinosteroid (BR) signalling pathway, directly regulating BR-responsive gene expression, but it has also recently been found to inhibit flavonoid biosynthesis via targeting key enzymes such as TRANSPARENT TESTA4 (TT4) and TRANSPARENT TESTA5 (TT5) (Liang et al., 2020).
In a very similar mode of action to D14, MAX2 is also able to recruit the karrikin receptor KARRIKIN INSENSITIVE 2 (KAI2) in the presence of karrikin or KAI2-ligands (Waters & Smith, 2013). The large similarities and overlaps between the SL and karrikin signalling pathways have made research about their specific functions challenging.
For example, much of previous work regarding phenotypic and transcriptomic response to SL addition has been performed using rac-GR24. This has created confusion regarding which signalling pathway is responsible for control over which gene regulatory networks and phenotypes, or if an interplay between the two mediates the effects.
Karrikin and SL pathways are also both involved in the regulation of flavonoid and anthocyanin biosynthesis, being downregulated in both max2 and kai2 plants during drought conditions (Li et al., 2017). In addition, when d14 plants are treated with rac-GR24, they have enhanced flavonol accumulation.
Within this work, we implicate MAX2-BES1 degradation in the control of flavonoid biosynthesis, pinpointing an intersection with flavonoid biosynthesis. We found that the max2 mutant has a different flavonoid profile to the three SL knockdown biosynthesis mutants, suggesting that MAX2 acts as a key regulator of flavonoid biosynthesis.

| Root system architecture analysis
Twelve days after the transfer of seeds to the growth cabinet, root systems were imaged via a flatbed scanner and root samples were flash-frozen in liquid nitrogen. The images were analysed using Fiji software (Schindelin et al., 2012). Primary root and LR measurements were taken for each seedling: primary root length, average LR length, LR number and LR density (the LR number divided by primary root length). For each genotypes, 5-20 seedlings per biological repeat were measured, and a total of three to four independent biological repeats were conducted. For RH analysis, the images of the first section of fully elongated RHs from the root tip were taken using a Leica MZ FLIII fluorescence stereomicroscope; then, RH length and number counts were measured in a 3 mm area using ImageJ (5-20 seedlings per biological repeat, three to four biological repeats). Measurements were considered statistically significantly different if they had a p < 0.05 (analysis of variance and post hoc Tukey or pairwise Wilcoxon rank-sum test with FDR adjustment).

| Diphenylboric acid-2-aminoethyl ester staining and confocal microscopy
Ten-day-old seedlings grown as above were immersed in a solution containing 0.01% Triton X-100 and 2.52 mg/ml diphenylboric acid-2-aminoethyl ester for 7 min, then rinsed with distilled water for a further 7 min (Sanz et al., 2014). Roots from individual plants were mounted in distilled water and visualised on a Zeiss 880 Laser-Scanning Confocal Microscope (Carl Zeiss Microimaging) (excitation at 458 nm, emission band pass 500-630 nm, Zeiss EC Plan-Neofluar Â10 objective lens). Micrographs were taken of optical sections of the epidermal layer. Integrated density measurements were taken using Fiji software (Schindelin et al., 2012) from the beginning of the peak fluorescence in the elongation zone, with this location chosen in order to account for differences in the growth of different genotypes.

| RNA extraction, sequencing and data processing
Frozen root tissue (stored at À80 C) from 12-day-old seedlings was homogenised by placing sterile beads in sampling tubes and using a Tissuelyzer (Qiagen). RNA was extracted from homogenised material using the Monarch RNA MiniPrep Kit (New England Biolabs) with the recommended on-column DNase treatment. For RNA sequencing, the quantity and quality of RNA were determined with a Bioanalyzer 2100 RNA 6000 Pico Total RNA Kit (Agilent Technologies). Extracted root RNA from three replicates of samples were sequenced with reactions for paired-end, unstranded, 150 bp read-length sequences using an Illumina NovoSeq 6000 by Novogene. The depth of sequencing was determined to be at least 20 million reads per sample. Initial quality control of raw reads was performed using FastQC. Raw reads were then subject to trimming and adaptor removal using Trimmomatic v0.33 (Bolger et al., 2014) with the following parameters: LEADING = 20, TRAILING = 20, SLIDINGWINDOW = 4:20, MINLENGTH = 75 and using PHRED33 quality scores; FastQC was used for quality control of the clean reads. Alignment of trimmed reads was performed using STAR (Dobin et al., 2013) using only paired reads. The index files for STAR were generated with the TAIR10 genome and genome annotation with the sjdbOverhang parameter set to 149 (read length À1). Raw counts data for 32,833 genes in the A. thaliana genome were obtained from the alignment files using STAR; raw RNA-seq data for this manuscript was deposited in the NCBI SRA database (PRJNA784414). Differential expression analysis was carried out using DESeq2 (Love et al., 2014). Genes were considered differentially expressed (DE) if they had a log2 fold change >1 and or <À1, and had a Bonferroni-corrected p < 0.05. Lists of DE genes generated from DESeq2 were queried for overrepresented gene ontology (GO) terms using the 'enrichGO' function of the R package clusterProfiler 4.0 (Wu et al., 2021). GO terms were considered overrepresented if they had a p < 0.05.

| RESULTS
3.1 | SL analogue rac-GR24 reduces LR length and number in Col-0 and all max mutants, with the exception of max2-1 In previous work, the max mutants had been found to have an aboveground branching phenotype (Gomez-Roldan et al., 2008;Umehara et al., 2008), but given the links between MAX function and hormone pathways related to root development, such as the auxin response pathway, we were interested in analysing belowground phenotypes (Bennett et al., 2016;Shinohara et al., 2013;Weijers & Wagner, 2016). In order to assess this, SL biosynthesis mutants max3-9, max4-1 and max1-1, and the SL-and karrikin insensitive mutant max2-1 root system architecture was analysed in the absence and presence of rac-GR24 (Table S1).
3.2 | SL mutant max4-1 has a significantly shorter primary root length compared to Col-0 under a 12 light/12 dark photoperiod, but not under a 16 light/8 dark photoperiod Phenotyping of max mutants with or without rac-GR24 has been performed previously and some of our results agree with those studies, including the finding that max4-1 but not max2-1 has a lower density of LRs with GR24 treatment (Kapulnik et al., 2011) (Table S1).
In these conditions, all plants grown under a 16-h light/8-h dark photoperiod had significantly shorter PRs than those in 12-h light/12-h dark. However, we found that primary root length in max4-1 was significantly shorter than in Col-0 in the 12-h light/12-h dark photoperiod (p = 9.40EÀ15), a phenotype that was not observed in the 16-h light/8-h dark conditions (Figure 2A,E). This suggests that the shorter primary root in max4-1 is at least partly daylength-regulated.
3.3 | Karrikin receptor mutant kai2-2 is not altered in LR length, number or density when treated with rac-GR24 We then investigated the max2-1 rac-GR24 insensitive LR phenotype in more detail in the 12-h light/12-h dark photoperiod. As MAX2 interacts with both the SL and karrikin pathways, we analysed the root system architecture of d14-1 and kai2-2 mutants in the absence and presence of rac-GR24 (Table S1). As with Col-0, primary root length was not significantly affected by rac-GR24 addition in either d14-1 or kai2-2 ( Figure 3A,E). However, d14-1 had a significantly longer primary root length than Col-0 under mock-treated conditions (p = 0.037). A significant decrease in LR length with rac-GR24 addition was seen in Col-0 (p = 4.1EÀ09) and d14-1 (p = 0.024), but not kai2-2 ( Figure 3B,E). There was also a significant decrease in LR number with rac-GR24 in Col-0 compared to mock-treated conditions (p = 0.024) and d14-1 (p = 0.001), but not kai2-2. The lack of LR number and length response to rac-GR24 in kai2-2 but not d14-1 suggests that this LR developmental regulation in response to rac-GR24 is partly driven by karrikin signalling.
Only Col-0 had a significant decrease in LR density in response to rac-GR24 compared to mock-treated conditions (p = 0.001). As this effect on LR density is not seen in d14-1 or kai2-2, our data agrees with the hypothesis presented by Villaécija-Aguilar et al. (2019) that KAI2 and D14 coregulate LR density. Under mock conditions, both d14-1 and kai2-2 had significantly shorter average LR lengths than Col-0 (p = 4.9EÀ05 and 0.01, respectively), and kai2-2 also had significantly lower number and density of LRs per plant (p = 0.001 and 0.032, respectively). Thus, both D14 and KAI2 seem to be involved in the LR responses to rac-GR24 and SL signalling.
In line with the phenotypes described by Kapulnik et al. (2011), RH length was found to increase with rac-GR24 addition for Col-0 (p = 0.03) and max4-1 (p = 0.049) ( Figure 4B,C). However, there was a significant decrease in RH length in rac-GR24-treated max1-1 compared to mock-treated (p = 4.07EÀ02). Also, max4-1 and max1-1 had longer RHs than Col-0 without rac-GR24 treatment (p = 1.40EÀ08 and 8.70EÀ08, respectively) ( Figure 4B,C). As max2-1 does not exhibit an increase in RH number or length with rac-GR24, these findings support the conclusion that rac-GR24 response in RH number and length is MAX2-dependent (Kapulnik et al., 2011). It also suggests a role for MAX3, MAX4 and MAX1 in regulating RH number and length. 3.5 | SL receptor mutant d14-1 does not show RH number or length changes with rac-GR24 As the RH length and number increase in response to rac-GR24 appears to be MAX2-dependent, we next sought to decipher whether it is also a D14-or KAI2-dependent process. We found that there is an increase in RH number with rac-GR24 addition in Col-0 (p = 0.01) and kai2-2 (p = 0.001). In addition, there is an increase in average RH length in Col-0 (p = 3.11EÀ02) and kai2-2 (p = 2.48EÀ03). In contrast, d14-1 mutants exhibited no change in RH number or length ( Figure 5A-C). This implies that the rac-GR24 3.6 | Triterpenoid biosynthetic genes are commonly downregulated in max2-1 and max4-1 To investigate the mechanisms behind the different root phenotypes in max mutants, we performed RNA sequencing on root tissue from 12-day-old Col-0, max2-1 and max4-1 seedlings without rac-GR24 treatment (Table S2). We selected max4-1 and max2-1 for their different phenotypes; max4-1 has a short primary root with longer and more numerous RHs, whilst max2-1 has short RHs. We found a total (À1 < FC > 1 and p < 0.05) between Col-0 and either max2-1 and max4-1 roots ( Figure 6A). Using silhouette plotting, DEGs could be grouped into seven clusters, and we then analysed the functions and processes (via GO terms) that were overrepresented within each cluster ( Figure 6D) (FDR-adjusted p < 0.05, Table S3).
Given that both MAX2 and MAX4 are involved in the SL signalling pathway at different stages, we first investigated common changes between max2-1 and max4-1 compared to Col-0 in Cluster 1, which has reduced expression in the max mutants, and Cluster 6, which has increased expression in the max mutants. Cluster 1, the largest cluster with 117 genes, had an overrepresentation of genes with the associated biological process (BP) GO terms 'triterpenoid biosynthetic process' and 'triterpenoid metabolic process', as well as related molecular function (MF) GO terms (Table S3). DEGs involved in this cluster include MARNERAL SYNTHASE (MRN1) and putative PENTACYCLIC TRITERPENE SYNTHASE 3 (PEN3), suggesting that SL biosynthesis and signalling also regulates synthesis of other terpenoid metabolites. Cluster 6 includes 31 genes whose expression is higher in both max mutants, with enriched GO terms involved in amino acid transport; these include the USUALLY MULTIPLE ACIDS MOVE IN AND OUT TRANSPORTERS (UMAMIT28) and AMINO ACID PERMEASE 6 (AAP6).

| Fatty acid metabolism is downregulated and xyloglucan metabolism is upregulated specifically in max4-1
Whilst max2-1 and max4-1 have some similar SL-related transcriptome profiles (clusters 1 and 6), there are also significant differences between them (DEGs in Clusters 2, 3, 5, 6 and 7). Cluster Arabidopis leaves (Ueda & Kusaba, 2015) and it is likely to affect root metabolism as well.
In Cluster 4, there are 12 genes with lower expression in max4-1 that show no changes in max2-1. This cluster is enriched for GO terms involved in triterpenoid metabolism, including lanosterol synthase, oxidosqualene cyclase and beta-amyrin synthase activity. One such DEG in this cluster is the terpenoid cyclase gene LUPEOL SYNTHASE 5 (LUP5), which is downregulated in max4-1 but not max2-1. This suggests that some SL-related terpene metabolism is reliant on MAX4  Table S3), including 2-oxoglutarate (2OG) and Fe(II)-dependent oxygenase superfamily protein and GIBBERELLIN 20 OXIDASE 2 (GA20OX2). Particularly striking in this cluster was the significant upregulation of MAX4 in max2-1 (p = 2.73EÀ11). This is consistent with the previous finding that SL biosynthesis is upregulated in max2, but not max4-1 compared to wild-type .
Overall the gene expression analysis both helped to understand which molecular processes were commonly regulated by MAX2 and MAX4 but also distinguished their roles related to root development.
3.8 | Flavonoid metabolism-related genes are differentially expressed between max2-1 and Col-0, but not max4-1 It has been previously shown that MAX2 is capable of degrading the master transcription factor BES1 (Wang et al., 2013). Also, it has been shown that BES1 inhibits CHI and TT4 transcription by repressing the transcription of MYB11, MYB12 and MYB111. These transcription factors activate CHS and CHI transcription, as well as downstream flavonoid biosynthesis genes FLAVANONE 3-HYDROXYLASE (F3H) and FLAVONOL SYNTHASE1 (FLS1) (Liang et al., 2020). We, therefore, hypothesised that in the max2-1 mutant, flavonoid biosynthesis might be downregulated due to lack of BES1 degradation.
3.9 | BES1 targets are over-enriched amongst max2-1 DEGs F I G U R E 6 Root transcriptome analysis of 12-day-old max2-1 and max4-1 seedlings implicates metabolic differences in mediating the max phenotypes. (A) Heatmap of differentially expressed genes between Col-0 and either max2-1 or max4-1 grouped into clusters of similar increased/decreased expression patterns. Genes were considered differentially expressed (DE) if they had a log2 fold change >1 and or <À1, and had a Bonferroni-corrected p < 0.05. (B) Plots of expression values (log 2 (counts)) for selected DE genes per cluster. Asterisks indicate statistically significant difference compared to Col-0 as indicated by brackets (***p < 0.001; n = 20 pooled root samples per biological repeat, three biological repeats). (C) Venn diagram of 292 DE genes in max2-1 and max4-1 root tissue, showing their shared DE genes and number of upregulated and downregulated DE genes per genotype tissue, we identified eight as DEGs in our max2-1 dataset, which represented a 3.7-fold enrichment (hypergeometric test, p = 0.001). Of these, three were upregulated: AT3G49620, AT1G56600 (GAL-ACTINOL SYNTHASE 2) and AT1G58340, and five downregulated, AT2G01422, AT1G25240, AT1G54970, AT3G55120 (CHI), AT1G06923. A previous work identified putative BES1 target transcription factors through ChIP-chip and gene expression studies, including AT1G56600 and AT1G06923 (Yu et al., 2011). Using 14-day-old gain-of-function bes1-D seedlings-in which BES1 accumulates to high levels- Yu et al. (2011) also identified upregulated and downregulated genes compared to wild-type. Of these, 33 upregulated genes and 38 downregulated genes matched with the 378 direct BES1 targets in roots. Three of the 5 downregulated BES1 direct interactors identified in max2-1 seedlings (AT2G01422, AT1G25240, AT1G54970) were also found to be downregulated in bes1-D seedlings (Yu et al., 2011). In comparison, in the max4-1 dataset, there was no significant enrichment of root BES1 target genes, further suggesting a direct and specific MAX2-BES1 link in integrating SL biosynthesis and flavonoid biosynthesis. Furthermore, analysis of microarray data from bes1-D plants found that the most significantly enriched downregulated genes by BES1 were flavonoid biosynthesisrelated (Liang et al., 2020).
Using RT-qPCR on RNA extracted from whole seedlings, we then asked if the BES1-targeted flavonoid-related genes identified in our max2-1 dataset (TT4 and TT5) were also downregulated in max2-1 whole seedlings. Consistent with the downregulation of flavonoid F I G U R E 7 Flavonoid biosynthesis and root accumulation is downregulated in max2-1, but not max4-1. (A) RT-qPCR was performed using RNA extracted from 12-day-old max2-1 and max4-1 seedlings with primers for TT5 and TT5 (n = 20 seedlings per genotype per biological repeat, three biological repeats). (B) Ten-day-old seedling root tips were stained with 2.52 mg/ml diphenylboric acid-2-aminoethyl ester (DPBA), which results in flavonoid fluorescence when excited at 458 nm. Roots were imaged and integrated density measurements were taken from the beginning of the peak fluorescence in the elongation zone as fluorescence arbitrary units (AU) per micron 2 . (C) Representative images of DPBA root tips for Col-0, max2-1 and max4-1; bars represent 100 microns. Asterisks indicate statistically significant difference as indicated by brackets (Student's t-test, *p < 0.05, **p < 0.01) biosynthesis genes in max2-1 root tissue, transcripts of both TT4 and TT5 were significantly reduced in max2-1 whole seedlings, but not max4-1 ( Figure 7A). This suggests a global effect of lack of MAX2 on flavonoid biosynthesis gene transcription.

| Flavonoid production in max2-1 roots is decreased
Given the evidence for downregulated flavonoid-related transcripts in max2-1 in both root tissue and whole plants, we next asked if this leads to decreased flavonoid accumulation in roots of the max2-1 mutant. To see if there are earlier changes in max2-1 flavonoid accumulation, leading to phenotypic changes, we imaged 10-day-old root tips as this is a known area of peak flavonoid accumulation in A. thaliana seedlings (Buer et al., 2013). We found that fluorescence in the peak zone in max2-1 was significantly reduced (Student's t-test, p < 0.002) ( Figure 7B,C), confirming that flavonoid accumulation happens less in max2-1 root tips compared to wild-type and max4-1.

| DISCUSSION
SLs are important phytohormones involved in plant development, coordinating developmental cues and stress responses to regulate growth of aerial plant organs (Omoarelojie et al., 2019;Saeed et al., 2017).
Here, we add evidence that SLs also play a role in shaping root system architecture. In all max mutants, except SL receptor max2-1, we demonstrate that rac-GR24 treatment results in plants with fewer and shorter LRs, concurring with previous studies (Jiang et al., 2016;Kapulnik et al., 2011;Ruyter-Spira et al., 2011). As rac-GR24 is known to stimulate both SL and karrikin signalling pathways via MAX2, this does not confirm a solely SL-induced response. Indeed, the formation of junction roots has been attributed to both SL and karrikin signalling through MAX2 (Swarbreck et al., 2020), and so it is likely that LR development is similarly multiregulated. We demonstrate here that, as with max2-1, rac-GR24 addition does not lead to inhibition of kai2-2 mutant LR elongation or formation, which is seen for Col-0, implying that this effect on LR development is a KAI2-dependent response. However, rac-GR24 did also not affect d14-1 nor kai2-2 LR density, so it is plausible that both karrikin and SL pathways are responsible for LR rac-GR24 response as discussed by Villaécija-Aguilar et al. (2019).
As with previous work, we also found that rac-GR24 increases RH length in the max mutants max3-9 and max4-1 and RH number in all SL biosynthesis mutants (Kapulnik et al., 2011). Previous authors have suggested rac-GR24 RH elongation to be karrikin-driven (Villaécija-Aguilar et al., 2019). However, we did see a decrease in RH length in max1-1 with rac-GR24. This indicates that the accumulation of bioactive SL precursors, such as carlactone, alongside rac-GR24 may influence RH length to some extent. Similarly, if SLs had a positive effect on RH length and density, we would have expected SL biosynthesis mutants to have shorter and less numerous RHs than wild-type without rac-GR24 treatment. However, we found all max biosynthesis mutants to have longer and more numerous RHs than wild-type. Thus, it is possible that karrikins and SLs both have roles in RH elongation and number. Indeed, whilst max2-1 had fewer RHs than wild-type, we found that neither kai2-2 nor d14-1 had this phenotype, suggesting a cumulative effect of disrupting both karrikin and SL pathways in max2-1.
Whilst Col-0 and kai2-2 both had an increase in RH number and length with rac-GR24 addition, we confirmed that d14-1 did not, suggesting that rac-GR24 response in RHs is dependent on D14 signalling, but not KAI2. Both kai2-2 and d14-1 had longer average RH lengths than wild-type, reminiscent of the max biosynthesis mutants.
Thus, as with LR growth, it is likely that both karrikin and SL signalling contribute to the regulation of RH length.
We also found that max4-1 had a short primary root phenotype which was not rescuable by rac-GR24 treatment. Through growing max4-1 under two different photoperiods, we found that the max4-1 short primary root phenotype is present when plants are grown in a 12-h light/12-h dark photoperiod, but not in a 16-h light/8-h dark photoperiod. This suggests that the mechanism behind the shorter primary root in max4-1 is daylength-regulated. In support of this hypothesis, it was recently found that D14, which is essential for SL signal transduction and regulation in the SL biosynthesis pathway, is regulated by the circadian clock in rice since overexpression of rice CIRCADIAN CLOCK ASSOCIATED1 (OsCCA1) positively regulated D14 (Wang et al., 2020a). Given the direct role of D14 in regulating SL biosynthesis, if AtD14 is also entrained to the circadian clock, it is likely that this is affecting overall SL or precursor levels, and therefore this could explain the phenotypic differences between the plants.
Through RNA sequencing, we identified a number of processes that are commonly altered in max2-1 and max4-1, including amino acid transport. It is possible that this alteration in amino acid transport is a stress response to disturbed SL regulation. Indeed, an increased expression of UMAMIT transporters has been found to induce stress phenotypes, such as stunted growth and increased expression of stress markers in A. thaliana (Besnard et al., 2021). The exact mechanism by which increased amino acid transport results in this constitutive stress response is unknown. Given the short primary root phenotype in max4-1 but not max2-1, the downregulation of VLCFA-related genes in max4-1 seems of significance. VLCFAs have been implicated in stimulating primary root growth by activating ethylene biosynthesis, which in turn promotes the division of quiescent centre cells (Ortega-Martínez et al., 2007;Qin et al., 2007). Kcs mutants with dysfunctional VLCFA production have significant primary root growth retardation in both A. thaliana and barley (Lee et al., 2009;Weidenbach et al., 2015), further supporting that the stunted primary root growth of max4-1 but not max2-1 could be potentially explained by the lower levels of VLCFA-related transcripts. As this shorter primary root growth phenotype is present only in max4-1, it could be that this reduction in VLCFA metabolism is a response to low levels of MAX4 and the SL precursor carlactone, as it is a signal molecule able to bind D14 (Abe et al., 2014).
Both ethylene and VLCFAs are also implicated in senescence and plant lifespan, as are SLs (Iqbal et al., 2017;Ueda & Kusaba, 2015). Indeed, MAX4 is upregulated in senescing leaves, and so a similar role for SL in root senescence seems likely (Ueda & Kusaba, 2015).
It has been shown previously that xyloglucan is required for normal RH tip growth and expansion and that interrupting xyloglucan metabolism results in a severe short RH phenotype (Cavalier et al., 2008). In contrast to our finding that max4-1 has longer and denser RHs, it is possible that this upregulation in xyloglucan metabolism results in the 'hairy' phenotype. Some exemplar xyloglucanrelated DEGs which were significantly upregulated in max4-1 but not max2-1 include XYLOGLUCAN ENDOTRANSGLUCOSYLASE/ HYDROLASE PROTEIN 15 and 20 (XTH15 and XTH20) as well as GLU- Given that none of these DEGs are differentially expressed in max2-1, which has no significant difference in its RH average length compared to Col-0, this data supports the idea that regulation of xyloglucan metabolism could be a SL or carlactone biosynthesis-related process and is MAX2-independent. As MAX2 is part of a SL regulatory feedback loop and is able to degrade D14 to limit SL signal transduction, it is plausible that another feedback loop exists whereby overaccumulation of D14 without bound SL results in upregulation of SL biosynthesis genes to maintain basal SL levels. Though, as the other MAX SL biosynthetic genes are also expressed in roots, it is interesting that MAX4 specifically is upregulated in max2-1.
It is striking that the majority of DE gene clusters were involved in metabolic pathways. SLs, in the roots at least, appear to be part of a metabolite regulatory hub for triterpenoids, flavonoids, fatty acids and xyloglucans. The key to connecting these metabolites seems to be the response to stress since SL production is promoted during nutrient starvation, such as lack of phosphate. Thus the regulation of xyloglucans and fatty acids to shape the root system architecture for F I G U R E 8 Model showing the hypothetical interaction between strigolactone and flavonoid biosynthesis pathways via MAX2-BES1, related to the work presented here. BES1 inhibits transcription of MYB11, MYB12 and MYB111 which, in turn, promotes transcription of key flavonoid biosynthesis genes (TT4, TT5, F3H and FLS). BES1 is also able to be degraded by MAX2. With our new phenotypic and transcriptomic data, we hypothesise that the downregulation of flavonoid biosynthesis in max2-1 is due to lack of BES1 degradation by MAX2. We propose that plants without a functional MAX2 protein accumulate BES1, leading to over-inhibition of flavonoid biosynthesis optimal root foraging makes sense. The self-regulation of triterpenoid metabolism could act as a SL feedback loop as well as promote the release of specific triterpenoids into the surrounding soil. As flavonoids are also important components of root-microbe interactions, co-regulation of flavonoid biosynthesis under low nutrient stress would be of great benefit to the plant. Equally, flavonoids are important antioxidants that scavenge reactive oxygen species, alleviating general cellular stress. Here, we present a model of co-regulation of SL and flavonoid biosynthesis via MAX2-mediated BES1 degradation ( Figure 8). The regulation of flavonoid biosynthesis by BES1 has been previously documented; microarray data from bes1-D plants presented the most significantly enriched downregulated genes by BES1 as flavonoid biosynthesis-related (Liang et al., 2020). Links between SL and flavonoid biosynthesis and MAX2 also have recently suggested, with D14-specific GR24-4DO-treated wild-type plants showing upregulated flavonoid biosynthesis (Wang et al., 2020b).
Our finding that flavonoids accumulate to a lower level in max2-1 supports this, elucidating an exciting link between two key plant metabolic pathways that coordinate root developmental growth and responses to environmental stress.

SUPPORTING INFORMATION
Additional supporting information may be found in the online version of the article at the publisher's website.