TRANSPARENT TESTA GLABRA 1-Dependent Regulation of Flavonoid Biosynthesis

The flavonoid composition of various tissues throughout plant development is of biological relevance and particular interest for breeding. Arabidopsis thaliana TRANSPARENT TESTA GLABRA 1 (AtTTG1) is an essential regulator of late structural genes in flavonoid biosynthesis. Here, we provide a review of the regulation of the pathway’s core enzymes through AtTTG1-containing R2R3-MYELOBLASTOSIS-basic HELIX-LOOP-HELIX-WD40 repeat (MBW(AtTTG1)) complexes embedded in an evolutionary context. We present a comprehensive collection of A. thaliana ttg1 mutants and AtTTG1 orthologs. A plethora of MBW(AtTTG1) mechanisms in regulating the five major TTG1-dependent traits is highlighted.


Motivation
Flavonoids are secondary plant metabolites with a broad spectrum of functions for plants, including UV protection, pollination and feeding attraction, rhizosphere signaling and pathogen defense [1][2][3][4][5]. Early on, health-promoting effects of flavonoids were under investigation, and a mixture of flavonoids from citrus was initially named "vitamin P" by Szent-Györgyi in 1936 [6].
Generations of researchers have been attracted by flavonoids, their function and the regulation of their biosynthesis pathways. Potential health-promoting benefits for human nutrition, as well as medical topics like binding and inhibition of proteins, absorption, conversion and localization within the human body were in the focus of more than 70% of flavonoid reviews. Apart from these, two major types of topics dominated other reviews since 1964: (1) the substances themselves (e.g., lists of substances, identification of individual substances, chemical properties, synthesis and conversion, analytical methodology, composition in specific species, tissues and developmental stages) and (2) their function (e.g., functions tested in vitro, like radical scavenging, functions in plants like UV protection and functions related to the plant s environment like exudate signaling within the rhizosphere, pathogen defense, pollination or feeding attraction).
Different types of flavonoids are known to have different functions (e.g., [3,[7][8][9][10]). Therefore, when it comes to breeding efforts, understanding the varying flavonoid composition of specific plant organs throughout development is of utmost relevance. So far, only a few reviews have centered on specific aspects of the flavonoid biosynthesis pathway's regulation through plant development.
Where We Are and Where to Go?
Screening for plants with deviating coloration of plant organs facilitated the identification of the core pathway's enzymes and essential regulators. These mutant sets, combined with sequencing

Flavonoid Biosynthesis Pathway
At the end of the last century, A. thaliana, an annual Brassicaceae with a small genome, emerged as the model organism for molecular geneticists. Its genome was the first sequenced genome of a flowering plant and was released in the year 2000 [17]. Extensive detailed knowledge about the flavonoid biosynthesis pathway and its regulator stems from this workhorse of molecular genetics. Studies on this pathway in A. thaliana benefit from the fact that the core enzymes are only encoded by a single gene except for six copies for FLAVONOL SYNTHASE (FLS), out of which two were shown to be expressed [18,19].
Flavonoids are widely distributed within the plant kingdom and can be found in bryophytes, vascular plants and spermatophytes. It is discussed that flavonoids initially evolved as regulators or messengers and then as UV protectants [23]. The early steps of the pathway are shared by mosses and, e.g., angiosperms and therefore thought to be of ancient origin. As both parts of the pathway have evolved separately, but have evolved for a long time, it might not be surprising to find both early and late biosynthesis of enzymes: specific and overlapping sets of regulators. AtTTG1-dependent regulation is predominantly specific for the late biosynthesis of enzymes.  [22] with an emphasis on A. thaliana seeds. Black arrows highlight the core pathway and substances in A. thaliana discussed in this review. The respective enzymatic steps from the different levels of the pathway are extracted on the right. Examples for different products of OMTs (O-METHYLTRANSFERASEs) are given (isorhamnetin, peonidin, petunidin, malvidin). Note that not all steps occur in A. thaliana and other substances can also occur in A. thaliana: e.g., pelargonidin. See [22] for details. The dashed arrow indicates that several proteins and enzymes are required for proanthocyanidin synthesis, deposition and/or further conversion to condensed tannins including transport processes (for a review on flavonoid transport and a few recent findings, see [24,25]). R2R3-MYELOBLASTOSIS-basic HELIX-LOOP-HELIX-WD40 repeat (MBW(AtTTG1)) complexes predominantly regulate late biosynthetic genes encoding for the core pathway enzymes. Dashed line: the assumed "border" between early and late biosynthetic genes and products in A. thaliana [20,26,27]. Boxed enzymes are under differential developmental regulation of MBW(AtTTG1) complexes. See Section 6 for details. Reduced levels of isorhamnetin in seeds of ttg1-1, but not transparent testa (tt)2 and tt8 mutants, indicate that the border might need some adjustment or that additional AtTTG1-dependent regulation occurs [22]

Trait Complexity and the Evolution of Transcription Factor Families
With increased trait complexity in plant organisms, transcription factor gene families have expanded [28]. For bHLH (basic HELIX-LOOP-HELIX), as well as for many other transcription factors, a strong expansion accompanied the period of land invasion [29]. Transcription-associated proteins (TAPs) can be separated into transcription factors (like R2R3-MYB(MYELOBLASTOSIS (homolog)) or bHLH), binding cis-regulatory elements in a sequence-specific manner and transcription regulators (like TTG1) that are of relevance within gene regulatory networks, but do not necessarily bind DNA themselves [28]. The combinatorial control in transcription factor complexes with components from different transcription factor classes allowed the integration of different developmental and environmental signals for further trait diversification [30]. In agreement with this, transcription factor family expansion correlated with morphological complexity in land plant evolution. This was not the case for transcriptional regulators [28].
Within the AtTTG1-dependent gene regulatory network, the highest trait specificity is found for the R2R3-MYB factors (see the next section), indicating that these might have evolved the most towards specific adaptive traits. A reduced complexity is seen for bHLH factors with partially overlapping function. Differential complex composition of MBW(AtTTG1) complexes can explain the differential AtTTG1-dependent regulation of the flavonoid biosynthesis pathway. In particular, differing DNA binding motif preferences of the respective R2R3-MYB components have been reported [31][32][33].

AtTTG1-Dependent Gene Regulatory Network
AtTTG1 is the head of an evolutionarily-conserved gene regulatory network, regulating five AtTTG1-dependent traits with adaptive value for the plant: (1) trichome and (2) root hair patterning, (3) accumulation of anthocyanidin (and its derivatives) (in seedlings), (4) seed coat pigmentation and (5) seed coat differentiation including mucilage production ( Figure 2). Differential MBW(AtTTG1) complex composition underlies the regulation of the five major traits [15,20,[34][35][36][37][38]. Within the AtTTG1-dependent gene regulatory network, the highest trait specificity is found for the R2R3-MYB factors (see the next section), indicating that these might have evolved the most towards specific adaptive traits. A reduced complexity is seen for bHLH factors with partially overlapping function. Differential complex composition of MBW(AtTTG1) complexes can explain the differential AtTTG1-dependent regulation of the flavonoid biosynthesis pathway. In particular, differing DNA binding motif preferences of the respective R2R3-MYB components have been reported [31][32][33].

AtTTG1-Dependent Gene Regulatory Network
AtTTG1 is the head of an evolutionarily-conserved gene regulatory network, regulating five AtTTG1-dependent traits with adaptive value for the plant: (1) trichome and (2) root hair patterning, (3) accumulation of anthocyanidin (and its derivatives) (in seedlings), (4) seed coat pigmentation and (5) seed coat differentiation including mucilage production ( Figure 2). Differential MBW(AtTTG1) complex composition underlies the regulation of the five major traits [15,20,[34][35][36][37][38]. AtTTG1-dependent gene regulatory network. Proteins encoded by the AtTTG1-dependent gene regulatory network and possible differential MBW(AtTTG1) complex formation, regulating the five major AtTTG1-dependent traits. Shown are all bHLH and R2R3-MYB proteins that can form a ternary complex with AtTTG1 and regulate one of the five major AtTTG1-dependent traits indicated on the right. Lines represent interactions as documented in the BioGrid database (www.thebiogrid.org) [39]. Inhibitors/competitors compete with respective R2R3-MYBs for binding to bHLH proteins. We also extracted from the BioGrid database non-MYB and non-bHLH factors that interact with TTG1 and added the recently found interactors SPL4 and SPL5. Interactions are documented in a variety of papers [40][41][42][43][44][45][46][47][48][49][50][51][52][53]. For the inhibitors, see e.g., [54][55][56]. The seed coat picture was acquired by Hans-Peter Bollhagen. Photography of the seedling was done by Siegfried Werth.  AtTTG1-dependent gene regulatory network. Proteins encoded by the AtTTG1-dependent gene regulatory network and possible differential MBW(AtTTG1) complex formation, regulating the five major AtTTG1-dependent traits. Shown are all bHLH and R2R3-MYB proteins that can form a ternary complex with AtTTG1 and regulate one of the five major AtTTG1-dependent traits indicated on the right. Lines represent interactions as documented in the BioGrid database (www.thebiogrid.org) [39]. Inhibitors/competitors compete with respective R2R3-MYBs for binding to bHLH proteins. We also extracted from the BioGrid database non-MYB and non-bHLH factors that interact with TTG1 and added the recently found interactors SPL4 and SPL5. Interactions are documented in a variety of papers [40][41][42][43][44][45][46][47][48][49][50][51][52][53]. For the inhibitors, see e.g., [54][55][56]. The seed coat picture was acquired by Hans-Peter Bollhagen. Photography of the seedling was done by Siegfried of bHLH proteins [40,42,58]. The highest specificity of the trait regulation by MBW(AtTTG1) complexes exists on the level of the R2R3-MYB factors, while the bHLH factors act redundantly towards the different traits, and AtTTG1 is indispensable for all five traits. While we focus on the flavonoid biosynthesis in this review, we would like to point to previous reviews covering the role of AtTTG1 in trichome and root hair patterning (e.g., [59,60]). AtTTG1 has also been positioned in the seed coat mucilage pathway [61].
The bHLH motif was first discovered and described in murine muscle development transcription factors and found to mediate dimerization and DNA binding [84]. It consists of a basic region at the N-terminus that binds specific DNA motives and an HLH region that mainly forms homo-and hetero-dimers with bHLH proteins [85]. Several reviews have focused on plant bHLH transcription factors in general and in specific pathways (see, e.g., [81,86,87]).

WDRs-TTG1
WD40 repeat (WDR) proteins are conserved in eukaryotes [88]. They have evolved in plants in various protein families with diverse functions: e.g., signal transduction, cytoskeletal dynamics, chromatin modification or transcriptional regulation [88]. This is in part due to diversification of regulators and targets up-and downstream the WDR proteins that act as interaction platform, constituents for protein complexes and sites of transient protein contacts [88]. WDR proteins are characterized by different numbers of WD40 repeats (usually 4-10 in plants) [88]. WDR proteins share a stretch of about 40 amino acids that usually end with Trp-Asp (WD) in each WD40 repeat [89]. Four and more WD40 repeats in one protein can form so-called β-propellers, a cylindrical formed series of four-stranded antiparallel beta sheets [88,90]. In the mammalian G-protein subunit Gβ, it is shown that the first and last WD 40 repeat contribute to the same beta-blade [90]. There are 237 WDR proteins with more than four repeats in A. thaliana [88], including TTG1, CONSTITUTIVELY PHOTOMORPHOGENIC 1 (COP1) and the four SUPPRESSOR OF PHYA-105 (SPA) proteins [34,[91][92][93].
TTG1 sequences were known to be present in angiosperms, but not in gymnosperms and older plant lines [31]. This requires an update, as recently, an ortholog has been identified from Norway spruce (Picea abies) [94]. For an overview of (putative) AtTTG1 orthologs, see Section 9 of this review.

AtTTG1 Mutants
A. thaliana mutants of AtTTG1 were fundamental in elucidating its gene and proteins functions. One of the often used and stronger mutants is ttg1-1, which carries a mutation leading to a premature stop codon and thereby to a truncation of the AtTTG1 protein at its far C-terminus. A closer view of the AtTTG1 gene and protein structure will help to understand the relevance of the C-terminus for AtTTG1.
In AtTTG1 loss-of-function mutants, leaves are glabrous as no trichomes are formed, ectopic root hairs are additionally produced, seeds are yellow instead of brownish, seedlings do not accumulate anthocyanidin and the columella, as well as the mucilage of seed coat epidermal cells is missing.
Both flavonoid-related AtTTG1-dependent traits, anthocyanidin accumulation and seed coat pigmentation melt down to a differential regulation of the flavonoid biosynthesis pathway. The TTG1 syndrome of ttg1 mutants varies in dependence of the position and type of mutation. Table 1 lists 22 A. thaliana ttg1 mutants and the observed flavonoid-related pigmentation phenotype of seeds.  103] * Note that ttg1-21 has been assigned twice and "-CI" was added here for clarification, CI = carbon ion-induced chromosomal rearrangement (breakage in exon 1 of TTG1 on chromosome 5 and rejoining with chromosome 3), tt: transparent testa phenotype, g: glabra phenotype, severity of glabrous phenotype: g > (g) > ((g)); 1 origin: Kranz collect. = Kranz collection; 2 GABI ("Genomanalyse im biologischen System Pflanze") -Kat ("Kölner Arabidopsis T-DNA") lines [98]; 3 SK (Saskatoon) population [103]: activation tagging lines (insertion of pSKI015 [104]); # enlarged trichome cells usually with no aerial extension, but sometimes stubby un-branched trichomes. Please note that not in all cases has the function been linked to the mutation. Not all studies conducted a rescue experiment or tested for allelism with a previously established mutant. Please refer to the references for more details. Col: Columbia, Ler: Landsberg erecta, Est: Estland, EMS: Ethyl methanesulfonate (used for mutagenesis), T-DNA: transfer-DNA, UTR: untranslated region, NA: no information available, urm: unarmed, gl1: glabra1, ND: position of insertion was not determined in the respective study.
The position of the mutations within these mutants indicates that proper AtTTG1 function requires an intact C-terminus. In Gβ, the first and last WD40 repeat contribute to the same β-blade; a similar scenario might occur in AtTTG1. Together with the modeled 3D-structure of AtTTG1, this points to a high relevance of the C-terminus for the protein's proper folding and domain structure (see Figure 3).  [105]. Exons, introns, 5′ and 3′UTR, mutants and protein domains were visualized using CLC DNA workbench 5.3.1 (CLC bio, Aarhus, Denmark). Please note that the intron is not part of the 3′UTR, which comprises the last nucleotide of exon 1 and the full exon 2. When searching databases for domains and repeats of AtTTG1, the resulting number, positioning and confidence for the detected WD40 repeats are diverse (e.g., SMART, PROSITE) [106][107][108][109]. Shown here is the annotation provided by UniProt. The colored bar below the protein sequence indicates the confidence of the Phyr2 model shown in (b) along the protein [110]. (b,c) AtTTG1 tertiary structure. The AtTTG1 protein tertiary structure was modeled using (b) Phyre2 or (c) loaded into SwissProt PdbViewer v.4.1.0 for marking the amino acids that are mutated or changed to a stop codon in the mutants [110,111]. Both (b,c) underline a part of the β-sheet structure, in particular the relevance of the C-terminus of AtTTG1 for the tertiary structure (highlighted by an ellipse in (b)). For the list of mutants and respective references, please refer to Table 1. CDS: coding DNA sequence, UTR: untranslated region.
Interestingly, according to the reported insertion loci (Table 1; [105]), two of the activation tagging T-DNA insertions were shown to be located in intron 1. This might change the 3′UTR due to differential splicing and possible loss of the 3′UTR portion originating from exon 2 in the transcript (exon 1 includes the stop codon and the first nucleotide of the 3′UTR; exon 2 comprises the remaining 3′UTR; see Figure 3). Moreover, similar as recently reported for GLABRA3 (GL3) [112], the intron might have a regulatory function, which could be impaired by the T-DNA insertion (size of inserted T-DNA > 6 kb). A T-DNA insertion in the AtTTG1 intron might change the splicing, the 3'UTR and its function, transcript stability and movement properties. Furthermore, an effect of the activation Figure 3. (a) AtTTG1 gene structure, mutants and protein domains. The AtTTG1 gene structure was retrieved from TAIR [105]. Exons, introns, 5 and 3 UTR, mutants and protein domains were visualized using CLC DNA workbench 5.3.1 (CLC bio, Aarhus, Denmark). Please note that the intron is not part of the 3 UTR, which comprises the last nucleotide of exon 1 and the full exon 2. When searching databases for domains and repeats of AtTTG1, the resulting number, positioning and confidence for the detected WD40 repeats are diverse (e.g., SMART, PROSITE) [106][107][108][109]. Shown here is the annotation provided by UniProt. The colored bar below the protein sequence indicates the confidence of the Phyr2 model shown in (b) along the protein [110]. (b,c) AtTTG1 tertiary structure. The AtTTG1 protein tertiary structure was modeled using (b) Phyre2 or (c) loaded into SwissProt PdbViewer v.4.1.0 for marking the amino acids that are mutated or changed to a stop codon in the mutants [110,111]. Both (b,c) underline a part of the β-sheet structure, in particular the relevance of the C-terminus of AtTTG1 for the tertiary structure (highlighted by an ellipse in (b)). For the list of mutants and respective references, please refer to Table 1. CDS: coding DNA sequence, UTR: untranslated region.
Interestingly, according to the reported insertion loci (Table 1; [105]), two of the activation tagging T-DNA insertions were shown to be located in intron 1. This might change the 3 UTR due to differential splicing and possible loss of the 3 UTR portion originating from exon 2 in the transcript (exon 1 includes the stop codon and the first nucleotide of the 3 UTR; exon 2 comprises the remaining 3 UTR; see Figure 3). Moreover, similar as recently reported for GLABRA3 (GL3) [112], the intron might have a regulatory function, which could be impaired by the T-DNA insertion (size of inserted T-DNA > 6 kb). A T-DNA insertion in the AtTTG1 intron might change the splicing, the 3'UTR and its function, transcript stability and movement properties. Furthermore, an effect of the activation tagging insert including 35S promoters and a possible additional transcript is possible. Helen A. Stafford described the differential accumulation of proanthocyanidin between plants with different developmental strategies. While abundant in gymnosperms and some ferns, proanthocyanidins are still found as defense compounds in leaves of long-lived woody plants, and they are rare in short-lived, herbaceous plants in which they are found in the seed coat of some of these [23]. In A. thaliana, seeds off ttg1 null mutants appear yellow due to the missing pigments [15]. They have a transparent testa. These pigments require the activity of DIHYDROFLAVONOL 4-REDUCTASE (DFR) and other downstream enzymatic steps [14]. These include LEUCOANTHOCYANIDIN DIOXYGENASE (LDOX)/ANTHOCYANIDIN SYNTHASE (ANS), ANTHOCYANIDIN REDUCTASE (ANR)/BANYLUS (BAN) and, in other species, also LEUCOANTHOCYANIDIN REDUCTASE (LAR) activity [20].

Target Enzymes of MBW(AtTTG1) Complexes in the Core Flavonoid Biosynthesis Pathway
Progress made in the last decade in differentiating target-specific activity of MBW(TTG1) complexes is diverse and uses different experimental systems (see previous reviews for a developmental and hormonal embedding [20,113,114]). In Figure 4, we combine exemplary results from at set of studies. These analyze the regulation of the flavonoid biosynthesis' core enzymes through differential sets of MBW(AtTTG1) components. The compilation highlights the need for systematic tissue-specific studies in this area of research. We apologize for all other examples not chosen here.
Many studies have focused on the regulation of DFR through MBW(AtTTG1) complexes. This is not surprising as no DFR is expressed in absence of AtTTG1 as shown in ttg1 mutants [14]. This leads to a complete blockage of the downstream pathway in these mutants [22]. On the level of bHLH factors, redundancy has been observed for AtTTG1-dependent traits. In this line, Zhang and co-workers observed a significantly reduced DFR expression as compared to the wildtype when analyzing seedlings. They also found a decrease of the DFR transcript level gl3egl3 double mutants and an even lower level in gl3egl3tt8 triple mutants (Figure 4, Study 1 [42]).
Further downstream, the pathway, BAN expression was modulated in young siliques that carried constructs overexpressing AtTTG1, TT8 and TT2, which were fused to the glucocorticoid receptor (GR). These move into the nucleus upon dexamethasone (DEX) treatment. The increased BAN expression in these AtTTG1-GR and TT2-GR lines could be even more elevated when using cycloheximide (CHX) to block protein translation and thereby block secondary transcriptional effects possibly evoked by transcriptional activity of direct targets of AtTTG1 and TT2. This result was in contrast to TT8-GR induced BAN expression upon DEX treatment counteracted by CHX treatment suggesting a more complex regulation through TT8 (Figure 4, Study 2 [47]). In the same study, a strong induction by adding AtTTG1 to TT2-GL3/EGL3/TT8 dimer combinations was found. Here, A. thaliana protoplasts were used as the test system. Interestingly, TT2-EGL3 and TT2-TT8 did not require AtTTG1 for a significant induction of BAN transcription. MBW complexes with MYC1 could not activate the BAN promoter:GUS construct.
Xu and co-workers also acquired the expression of flavonoid biosynthetic genes in mutant background and compared it to the respective wildtype [115]. Not only the LBGs DFR, LDOX and BAN are downregulated in tt2, tt8 and ttg1 mutants, but also the EBGs F3H and F3'H in tt2 and ttg1 mutants. As MBW(AtTTG1) complexes affect transcriptional cascades, it is relevant to point out that direct activation of DFR, LDOX and again BAN is shown for AtTTG1-GR in siliques using the same DEX/CHX approach as described above. The regulation of F3'H is indirect for AtTTG1, but direct for GL3 (Figure 4, Study 4 [27]).
An important step towards understanding MBW(AtTTG1)-regulated flavonoid compositions will be tissue-specific studies using, e.g., promoter:GUS constructs for the pathways' core enzymes as done for seeds [115]. Other important lines of experiments will be to bypass redundancies through systematic double and higher order mutants (e.g., [27,42]) and to analyze direct promoter binding. An important step towards understanding MBW(AtTTG1)-regulated flavonoid compositions will be tissue-specific studies using, e.g., promoter:GUS constructs for the pathways' core enzymes as done for seeds [115]. Other important lines of experiments will be to bypass redundancies through systematic double and higher order mutants (e.g., [27,42]) and to analyze direct promoter binding. (3) induced upon DEX treatment combined with or without CHX when fusions with GR were used. White boxes: the respective promoter/transcript level was not significantly affected in the experiment, and this was documented in the respective study. Grey boxes: not tested/no results presented in the respective study. Note the differential regulation based on complex composition, e.g., for AtTTG1-GL3/EGL3/TT8-TT2/MYB5 complexes in [115]. AtTTG1-GL3/EGL3/TT8-PAP are known to be relevant in vegetative tissue (e.g., [115,118]).
The following studies were used for compiling this figure and are indicated in the figure as numbers below the MBW factors: 1: [ overexpressor (ttg1, tt2, myb5, gl3egl3, gl3egl3tt8, gl3egl3tt8myc1, MYBRNAi, (3) induced upon DEX treatment combined with or without CHX when fusions with GR were used. White boxes: the respective promoter/transcript level was not significantly affected in the experiment, and this was documented in the respective study. Grey boxes: not tested/no results presented in the respective study. Note the differential regulation based on complex composition, e.g., for AtTTG1-GL3/EGL3/TT8-TT2/MYB5 complexes in [115]. AtTTG1-GL3/EGL3/TT8-PAP are known to be relevant in vegetative tissue (e.g., [115,118]).
The following studies were used for compiling this figure and are indicated in the figure as numbers below the MBW factors: 1: [42] [117]. *: wt vs. mutant or overexpressor (ttg1, tt2, myb5, gl3egl3, gl3egl3tt8, gl3egl3tt8myc1, MYBRNAi, pap1/pap1-D, Pro35S:MYB113). Black boxes mean in this case: The respective promoters are significantly downregulated in the respective mutants. #: upregulation effect of transcription upon overexpression of TTG1-GR or GL3-GR encoding constructs in ttg1 or in the gl3egl3 double mutant. ##: effect is also seen for GL3-GR when EGL3 is present in gl3 single mutants (height sucrose concentration (3%) in this experiment); F3H and CHX were not tested in this experiment. x: fusion to GR in the presence of DEX and CHX leads to upregulation of the respective promoter tested (all in respective mutant background). Pp: Physcomitrella patens protoplasts At: A. thaliana (A7) protoplasts, siliques: young siliques, endoth.: endothelium, seedling +: seedlings treated with cycloheximide (CHX) when testing for the effect of overexpressing, e.g., AtTTG1-GR (glucocorticoid receptor) or GL3-GR, respectively. For chalaza and endothelium, the results are from the visual impression from GUS-stained seeds. Thick boxes around enzymes: LBGs. See, e.g., Figure 1 for the other abbreviations.

MBW(AtTTG1) Complexes, Multilayered Regulatory Mechanisms at Work: A Possible Scenario
Atttg1 mutant analysis suggests that AtTTG1 is indispensable to regulate all AtTTG1-dependent traits in concert. It is assumed that AtTTG1, MYB and bHLH factors can form a ternary complex that acts as a transcriptional activator [31,35,38,119,120]. Interestingly, overexpression of the maize R gene in Atttg1 mutant background could rescue all AtTTG1-dependent traits [121]. It has also been shown that bHLH-R2R3-MYB-dimers can activate tested promoters [122]. This might also have been the ancient status, as, e.g., mosses possess type IIIf bHLH factors, but no functional TTG1 ortholog has been described to our knowledge [29,31]. GLABRA1 (GL1) and GL3 were shown to be able to activate CPC expression in Arabidopsis suspension cultures [122], and TT2 and TT8 overexpression is sufficient for BAN and LDOX activation, as well as PAP1 with EGL3 or TT8 for LDOX activation in A7 protoplasts, which express low levels of TTG1 according to the authors [47,117,118]. Therefore, the test system might not have been devoid of TTG1. Is this amount of putatively translated TTG1 sufficient to mediate a minimal required TTG1 function when the other factors are overexpressed? Here, studies in ttg1 mutant background are required. Since bHLH and an R2R3-MYB factor can activate transcription without TTG1, what is the function of TTG1? This has not been solved yet, and many aspects require further investigation. Other possible scenarios are described below (see Section 8).
One possible function of AtTTG1 might be to shield the bHLH and R2R3-MYB factor from (negative) regulators. AtTTG1 was shown to be localized at promoters (e.g., TRANSPARENT TESTA GLABRA2 (TTG2) and CAPRICE (CPC)) by semi-quantitative PCR of ChIP experiments [123]. It might modulate the chromatin at the target promoter to set the stage for the other factors (early in development). It might also be directed to target promoters by non-MYB, non-bHLH interaction partners (e.g., TTG2 [51]) or the following scenario might ( Figure 5 The possible scenarios are still diverse and several might occur in different conditions, tissues and developmental stages. Many regulatory aspects are still to be uncovered for these well-investigated MBW(AtTTG1)-complexes. Careful hypothesis testing is required in the future and testing which mechanisms also apply for the other AtTTG1-dependent traits. Therefore, it is worth summarizing the regulatory mechanism stemming not only from work on the flavonoid-related AtTTG1-dependent traits. Which of these also act in flavonoid biosynthesis? Which evolved independently?

AtTTG1 and MBW(AtTTG1) Complexes: Modes of Regulation
Accumulating findings suggest that the AtTTG1-dependent regulation is more diverse than initially expected and point to a multilayered regulation of the flavonoid biosynthesis core enzymes. The relevance of TTG1 as a regulatory hub is pointed out by a plethora of regulatory pathways and mechanisms regulating upstream, downstream TTG1 and even within MBW(AtTTG1) complexes. This diverse AtTTG1-dependent trait regulation seems to require up-and down-stream partners either within MBW(AtTTG1) complexes or additional partners. As transcriptional regulators like AtTTG1 are not expected to expand with organism complexity to the extent as transcription factor families like R2R3-MYB or bHLH factors do [28], regulatory mechanisms centered on TTG1 might be conserved between TTG1-dependent traits.
In the following, we are listing various modes of MBW(AtTTG1) regulation and provide example publications. On the level of bHLH/MYB factors, the relevance of post-translational modifications has been reviewed recently [125]. Many findings have extended the picture presented more than a decade ago, e.g., by Broun in 2005 [38], already suggesting some scenarios for which supportive results were found. Here, we do not differentiate between traits, as some of the mechanisms can be expected to be utilized for more than one trait, but this remains to be investigated in the future: Instead of ternary complex formation and binding to the promoter as a complex, a R2R3-MYB factor binds to a specific promoter. It cannot activate the promoter on its own, at least this was shown for the AtDFR promoter and R2R2-MYB proteins of MBW(AtTTG1) complexes [46]. bHLH factors could dock to the WD40-platform TTG1 (Figure 5a). When the appropriate bHLH-TTG1 pair is available, it binds the R2R3-MYB factor, facilitated maybe by a bHLH binding motif at an appropriate distance, and the ternary complex is (at least transiently) formed (Figure 5b). Prior DNA binding of the R2R3-MYB factor might not be required. TTG1 possibly even dissociates again (Figure 5c), but was required for joining the bHLH-R2R3-MYB pair. This scenario also fits with the recently described mechanism of competition of TTG1 and GL1 for binding to GL3 [122].
The requirement of two transcription factors for target gene expression allows for integration of two or more signaling pathways through the regulation of protein abundance for both, the respective bHLH and R2R3-MYB factor. A third layer is added through the required presence of TTG1, which is needed in the scenario described above for initial complex formation. TTG1 might also be necessary to protect the bHLH factor from degradation or inactivation to achieve a threshold of molecules which jointly activate (with the respective MYB factor) the target gene's transcription. Therefore, the rescue of ttg1 mutants overexpressing the R gene from maize could be explained if R is not selective towards its A. thaliana R2R3-MYB interaction partner. Moreover, its protein levels, due to the overexpression, are high enough to compete with the negative regulation of the non-endogenous R protein.
In the above-described scenario, the competitors of the R2R3-MYB proteins for bHLH binding, R3-MYB factors (e.g., [56]) that do not act as transcriptional activators, could act very efficiently: they might bind to the TTG1-bHLH dimer also prior to DNA-binding (Figure 5d), evoke TTG1 release and thereby subject the bHLH protein to its negative regulators, e.g., the ubiquitin ligase UBIQUITIN-PROTEIN LIGASE 3 (UPL3)/KAKTUS [124]. The transcription of the target gene through an MBW(AtTTG1) complex is thereby counteracted.
The possible scenarios are still diverse and several might occur in different conditions, tissues and developmental stages. Many regulatory aspects are still to be uncovered for these well-investigated MBW(AtTTG1)-complexes. Careful hypothesis testing is required in the future and testing which mechanisms also apply for the other AtTTG1-dependent traits. Therefore, it is worth summarizing the regulatory mechanism stemming not only from work on the flavonoid-related AtTTG1-dependent traits. Which of these also act in flavonoid biosynthesis? Which evolved independently?

AtTTG1 and MBW(AtTTG1) Complexes: Modes of Regulation
Accumulating findings suggest that the AtTTG1-dependent regulation is more diverse than initially expected and point to a multilayered regulation of the flavonoid biosynthesis core enzymes. The relevance of TTG1 as a regulatory hub is pointed out by a plethora of regulatory pathways and mechanisms regulating upstream, downstream TTG1 and even within MBW(AtTTG1) complexes. This diverse AtTTG1-dependent trait regulation seems to require up-and down-stream partners either within MBW(AtTTG1) complexes or additional partners. As transcriptional regulators like AtTTG1 are not expected to expand with organism complexity to the extent as transcription factor families like R2R3-MYB or bHLH factors do [28], regulatory mechanisms centered on TTG1 might be conserved between TTG1-dependent traits.
In the following, we are listing various modes of MBW(AtTTG1) regulation and provide example publications. On the level of bHLH/MYB factors, the relevance of post-translational modifications has been reviewed recently [125]. Many findings have extended the picture presented more than a decade ago, e.g., by Broun in 2005 [38], already suggesting some scenarios for which supportive results were found. Here, we do not differentiate between traits, as some of the mechanisms can be expected to be utilized for more than one trait, but this remains to be investigated in the future: • Differential complex formation (for an example of differential target regulation in seeds: [115]). • Insights from trichome and root hair patterning: activator/inhibitor and lateral inhibition, feedback loops' de novo pattern formation (including cell-to-cell movement) [36,60,126,127]. • Trapping and deletion model [48,128]. • Competitors (R3-MYBs) [54][55][56]129,130].

•
Competition of TTG1 and an R2R3-MYB factor for binding to a bHLH factor [122].
To transfer this wealth of knowledge not only from trait to trait, but also between species, an evolutionary embedding of MBW(AtTTG1) components (here, emphasis on AtTTG1) is required.

AtTTG1: Evolutionary Embedding and Functional Orthologs
The flavonoid biosynthetic pathway is found in a wide range of land plants, even in the bryophytes (mosses), and it has been suggested that synthesis of flavones, flavanones and flavonols may have evolved, first, to provide chemical messengers and then UV sunscreens [10,23]. Maize (Zea mays), petunia (Petunia hybrida) and snapdragon (Antirrhinum majus) emerged as the major models for the study of flavonoid biosynthesis, and genes encoding R2R3-MYB and bHLH proteins were identified as regulators of flavonoid structural genes, demonstrating broad conservation of this regulatory mechanism in these plants [121,[138][139][140][141][142][143][144][145][146][147][148]. By then, the relationship between WDR proteins and the R2R3-MYB/bHLH transcriptional regulators had not been revealed. The first anthocyanin regulatory locus that was cloned from petunia, AN11, encodes a protein containing five WD40 repeats [149]. Further support for the formation of an MBW complex in plants came from interaction studies conducted in A. thaliana [40]. Similarly, the maize WDR PAC1 is specifically involved in the flavonoid pathway, but can complement all ttg1 phenotypes in A. thaliana, and it became the first identified TTG1-like protein in monocots [138,150]. Later, WDR proteins were found to regulate the flavonoid biosynthetic pathway in other plants such as Arabis alpina, Perilla frutescens, Ipomoea nil, Medicago truncatula, Malus domestica and many others (see Figure 6 and Table 2) [138,[151][152][153][154][155].
Regularly, new orthologs are identified, which might be of relevance for breeding purposes. Not only are allelism tests and rescue experiments within the respective species conducted (if applicable) to explore the ortholog's function, when mutants or TTG1 variants are identified. More often, the function of the orthologs is estimated by using the model species A. thaliana. In cross-species rescue experiments, TTG1 orthologs from other species are expected to take over AtTTG1 function at least in part within respective MBW(TTG1) complexes. This year, for example, ectopic expression of SiTTG1 (a newly identified AtTTG1 ortholog in Setaria italica) in the A. thaliana ttg1-13 background was shown to fully rescue the glabrous trichome and the flavonoid phenotype. This suggests that SiTTG1 is a member of flavonoid biosynthesis regulators in monocots [156]. Another example is BrTTG1, which was isolated from a brown-seeded hairy Brassica rapa and found to functionally complement an A. thaliana ttg1 mutant, while another ortholog, isolated from Brassica rapa yellow-seeded glabrous germplasm, was not functional [157].
It needs to be mentioned that promoter sequences and MBW components might have differentially evolved in the respective other species and thereby led to shifts in MBW(TTG1) function. Nevertheless, newly-identified orthologs also provide novel insights into the evolutionary aspects of MBW(TTG1)s.
MBW complexes have been identified as common and conserved flavonoid biosynthesis regulators, similarly reported for various land plants; although in several cases, not all components have been identified or characterized ( Figure 6 and Table 2). Recently, MBW complexes (PaWD40-1-PabHLH1/2-PaMYB29/32/33/35) were characterized in Norway spruce (Picea abies) and shown to be involved in the regulation of the flavonoid biosynthesis pathway [94]. This reveals a full MBW regulator in gymnosperms, which were previously thought to be devoid of TTG1 orthologs [31].
MBW complexes determine the spatiotemporal expression of flavonoid biosynthesis target genes that account for tissue-specific accumulation of flavonoids. Some MBW complexes from monocots can control the expression of enzymes of the entire pathway, while others specifically control late flavonoid biosynthesis genes in eudicots [158]. Nevertheless, the bHLH interaction motif ([DE]Lx 2 [RK]x 3 Lx 6 Lx 3 R) found in R2R3-MYB members of MBW complexes is highly conserved among higher plant species [125], suggesting that at least MYB and bHLH interactions arose early during the event of land plant evolution. genes that account for tissue-specific accumulation of flavonoids. Some MBW complexes from monocots can control the expression of enzymes of the entire pathway, while others specifically control late flavonoid biosynthesis genes in eudicots [158]. Nevertheless, the bHLH interaction motif ([DE]Lx2[RK]x3Lx6Lx3R) found in R2R3-MYB members of MBW complexes is highly conserved among higher plant species [125], suggesting that at least MYB and bHLH interactions arose early during the event of land plant evolution.  Interestingly, in the rosid clade, besides regulation of the flavonoid biosynthesis pathway, combinatorial MBW complexes evolved several extra functions: for instance, trichome patterning, root hair patterning and seed coat mucilage production in A. thaliana [41][42][43]45,128,[160][161][162][163][164][165] or Arabis alpina [155] and seed hair formation in cotton (Gossypium hirsutum) [166][167][168][169]. These observations imply that a common regulatory MBW module has been adapted for controlling specific epidermal cell fates in rosids.
However, such pleiotropic functions of MBW complexes have neither been observed in the asterid clade nor in monocots. Based on this, a speculation is raised: new roles of MBW complexes in controlling epidermal cell fate may have diverged since the evolutionary separation of these major plant groups, although the details of this are still not clear [170,171]. This is supported by the findings that multicellular trichome and conical cell formation in asterids, like Antirrhinum and Solanaceae species, are regulated by MIXTA-like R2R3-MYB-related proteins in which the bHLH interaction motif ([DE]Lx 2 [RK]x 3 Lx 6 Lx 3 R) is devoid. MIXTA gene overexpression in rosids does not affect trichome formation [171][172][173]. AtGL1, a trichome patterning specific-R2R3-MYB protein, was grouped phylogenetically together with AtPAP and AtTT2 that act in the regulation of the flavonoid biosynthesis pathway. This clade is distinct from the MIXTA-like regulators branch [72,174,175]. It is assumed that duplication and subsequent divergence, as known for other protein families [28], has been the driving force to evolve new roles of MBW complexes and other epidermal cell fates in rosids as compared to asterids [174].

Perspective
"Who does what, where, when and why?" There is constant progress in understanding the spatiotemporal regulation mediated by MBW(TTG1) complexes in the model species A. thaliana, but also in fruits and crops. Understanding changing flavonoid compositions, e.g., for the benefit of pre-and post-harvest fruit quality in dependence of developmental and environmental cues (see, e.g., [225,226]) includes understanding the activity of major regulators of the flavonoid biosynthesis pathway like the MBW(TTG1) complexes. However, also biotechnological engineering of the flavonoid biosynthesis might benefit from recent findings on the pathways' regulation [116].
Why plants accumulate specific flavonoid compositions, especially a wide range of specific derivatives in different tissues throughout development, is still not fully understood, and the function of individual substances, as well as mixtures could be a central question driving future research.

Methods
Seven hundred and sixty reviews containing "flavonoid(s)" in the title were extracted with EndNote X8 (Clarivate Analytics, Philadelphia, PA, USA) from the PubMed and Web of Science databases in October 2017 and sorted by topic. The same databases were searched with EndNote X8 for an initial overview of literature related to TTG1/URM23 (search for full gene name and abbreviation in title/abstract/keywords). Additional references were manually selected. We apologize to any researcher who s work on the topic this review could not cover.
Author Contributions: Andrea Schrader wrote the manuscript. Bipei Zhang provided Figure 6, Table 2 and most of Section 9. Both authors read, corrected and approved the manuscript.

Conflicts of Interest:
The authors declare no conflict of interest.