TCP and MADS-box transcription factor networks regulate heteromorphic flower type 4 identity in Gerbera hybrida 5

One sentence summary: The development of marginal ray flowers in Asteraceae flower heads is 13 regulated by a network of CINCINNATA- and SEPALLATA-like transcription factors upstream of 14 a CYCLOIDEA-like gene. ABSTRACT 26 The large sunflower family, Asteraceae, is characterized by compressed, flower-like inflorescences 27 that may bear phenotypically distinct flower types. The CYCLOIDEA/TEOSINTE BRANCHED1 28 (CYC/TB1)-like transcription factors (TFs) belonging to the TEOSINTE BRANCHED 29 1/CYCLOIDEA/PROLIFERATING CELL FACTOR (TCP) protein family are known to regulate 30 bilateral symmetry in single flowers. In Asteraceae, they function at the inflorescence level, and 31 were recruited to define differential flower type identities. Here, we identified upstream regulators 32 of GhCYC3 , a gene that specifies ray flower identity at the flower head margin in the model plant 33 Gerbera hybrida . We discovered a previously unidentified expression domain and functional role 34 for the paralogous CINCINNATA-like (CIN) TCP proteins. They function upstream of GhCYC3 35 and affect the developmental delay of marginal ray primordia during their early ontogeny. At the 36 level of single flowers, the Asteraceae CYC genes show a unique function in regulating the 37 elongation of showy ventral ligules that play a major role in pollinator attraction. We discovered 38 that during ligule development, the E class MADS-box TF GRCD5 activates GhCYC3 expression. 39 We propose that the C class MADS-box TF GAGA1 contributes to stamen development upstream 40 of GhCYC3 . Our data demonstrate how interactions among and between the conserved floral 41 regulators, TCP and MADS-box TFs, contribute to the evolution of the elaborate inflorescence 42 architecture of Asteraceae. 43


INTRODUCTION
The Asteraceae family is the largest family of flowering plants. Phylogenetically that ray identity itself evolved multiple times independently in the family (Panero and Funk, 2008). 71 In Gerbera hybrida (gerbera), overexpression of CYC2 clade genes GhCYC2, GhCYC3,or GhCYC4 72 in disc flowers converted them into ray-like with elongated petals and disrupted stamen  Among the six CYC2 clade genes in gerbera, GhCYC3 is the strongest candidate gene to be 87 responsible for the regulation of ray flower identity as well as the growth of the ventral ligule.

88
GhCYC3 is exclusively expressed in marginal ray flower primordia (Tähtiharju et al., 2012). It also 89 shows the greatest expression in elongating ligules in contrast to the other five CYC2 clade genes

97
Identification of putative upstream transcriptional regulators of GhCYC3 98 We performed a yeast one-hybrid (Y1H) screen to identify putative upstream TFs interacting with 99 the gerbera GhCYC3 cis-regulatory region. We first performed a sequence analysis on 21 promoter 100 regions of selected Asteraceae CYC2 clade genes (Supplemental Table S1). MEME analysis  Homeodomain-leucine zipper (HD-ZIP I), and TCP TF gene families, and ten gerbera homologs 110 corresponding to these TFs were identified in BLAST searches (Supplemental Table S2). 111 We also applied in silico analysis to identify putative transcription factor binding sites (TFBSs) the 27 bp consensus sequence strongly resembles the GC-rich core motif (Fig. 1A). We also 118 focused on MADS-box TF binding sites, called as CArG boxes, CC(A/T) 6 GG, as MADS-box 119 proteins have been suggested to operate in the same regulatory cascades as TCP factors (Kaufmann 120 et al., 2009). We identified two candidate CArG boxes within the GhCYC3 promoter, 121 CCTAAAAGAG at -155 to -164 bp, and CCAATTCTGA at -192 to -201 bp (Fig. 1C).

122
GhCIN1/2, GRCD5, and GAGA1 TFs activate the PGhCYC3:LUC reporter 123 We tested the ability of the ten candidate TFs of gerbera (Supplemental Table S2 benthamiana. We did not observe reporter activation with any of the candidate proteins of the 126 AP2/ERF, NAC, DOF, or HD-ZIP I families (Supplemental Fig. S3A). However, two out of the 127 three CIN-like TCP domain TFs, GhCIN1 and GhCIN2, activated the reporter construct (Fig. 1D).

128
Both GhCIN1 and GhCIN2 showed activation only when fused to the VP16 domain, suggesting 129 that they may bind the target DNA, but require (an)other co-factor(s) for transcriptional activation.

144
All candidate proteins activated the reporter constructs including either the 1900 bp, 878 bp, or 367 145 bp 3' fragments of the promoter (constructs 2, 4, and 5, respectively), while the lack of the 3' region 146 (construct 3) abolished activation (Fig. 2B, C). This result corresponds with the presence of the TCP 147 TFBS and the two CArG boxes within the 367 bp sequence upstream of the TSS (Fig. 1).

149
To verify the DNA binding activity of the CIN-like factors with the core TCP binding motif GGtCC 150 at position -272 bp upstream of the 27 bp consensus, we mutated the site into GGtTG within the 367 151 bp promoter sequence (mTCP2, Fig. 1 and 2). Using the mutated reporter ( Fig. 2A, construct 6),  Table S4). The reconstructed phylogeny indicates that these genes are grouped into 167 the previously identified subclades (Fig. 3)   For the MADS-box genes, we focused on GRCD4 and GRCD5 as we have previously defined their 189 functions in ray flower ligule development (Zhang et al., 2017). At this stage, we omitted GRCD8 190 from further analyses. Compared to its paralog GRCD5 that is predominantly expressed in ligules,

191
GRCD8 shows more ubiquitous expression in all floral organs, and we still lack transgenic lines to 192 verify its function (Zhang et al., 2017). Regarding GAGA1, our previous data indicates that it 193 represents a classical C class gene being expressed only in stamens and carpels (Yu et al., 1999).

194
Silencing of GAGA1 in transgenic gerbera led to homeotic conversion of stamens into petals and 195 carpels into sepal-like structures (Yu et al., 1999;Kotilainen et al., 2000). 196 Here, we analyzed the expression of GRCD5 and GRCD4 during ray flower ligule development

228
In transgenic lines, the phenotypes of mature inflorescences especially regarding ligule growth were 229 minor and variable, however, phenotypic changes during early primordia initiation were obvious 230 and were observed in all four GhCIN1 RNAi lines and in one GhCIN2 RNAi line (TR15) (Fig. 6). 231 We have previously shown that development of ray primordia in wild-type (WT) gerbera is   GhCIN1 is expressed in the incipient ray primordia at the axils of the involucral bracts (Fig. 5, Fig.   305 7A). Our previous data showed that the flower meristem identity gene GhLFY is also expressed at reporter. However, GhCIN2 is expressed in involucral bract primordia that lack GhCYC3 expression 312 (Fig. 4B, C, Fig. 5G, I, Fig. 7A), and thus GhCIN2 may act as a repressor, most likely through 313 interaction with yet unknown factors, to exclude GhCYC3 from bracts. As the transgenic gerbera Firstly, we showed that the SEP3-like proteins GRCD5 and GRCD8, could effectively activate the 334 PGhCYC3:LUC reporter while GRCD4 did not (Fig. 1E). The expression of GRCD5 closely 335 overlaps with GhCYC3 specifically during ray flower ligule development (Fig. 4E, F, Fig. 7D, E). 336 We also showed that GhCYC3 expression was significantly reduced in GRCD5 RNAi and in   In silico analysis of the Asteraceae CYC2 clade promoter sequences 398 We performed in silico analyses using MEME (http://meme-suite.org) (Bailey and Elkan, 1994) and  Table S1). For the analyses, we used 1000 bp   Table S6.

422
To generate the yeast bait strain, the bait plasmid was linearized with BstBI, transformed into the 423 Y1H Gold strain (Clontech, Palo Alto, CA, USA), and plated on SD-Ura growth medium.

424
Integration into the yeast genome was confirmed by colony PCR combining a vector-specific 425 forward primer and an insert-specific reverse primer (Supplemental Table S6). The bait strain was 426 tested for the minimal inhibitory concentration of AbA. We also integrated the pAbAi vector 427 without any insert into Y1HGold, and used it as a negative control bait strain. All yeast 428 transformations were done following either small-or library-scale LiAc-transformation protocols described in Yeastmaker TM Yeast transformation system 2 manual (PT1172-1, Clontech, Palo Alto, 430 CA, USA).

433
Approximately two million colonies were screened and selected on SD-Leu-Ura/900 ng/ml AbA 434 selection plates. The candidate Arabidopsis transcription factor gene sequences obtained from the 435 library screen were used in BLAST searches against the gerbera RNASeq databases (T. Teeri and P. 436 Elomaa, unpublished data) to identify the gerbera homologs (Supplemental Table S2). In addition, 437 we defined the expression patterns for the gerbera homologs based on the read counts in our 438 RNASeq data and identified candidate genes that are co-expressed with GhCYC3 (Supplemental 439 Fig. S2). Their ability to activate the GhCYC3 reporter construct was examined in planta using Based on the Y1H result, the sequences of corresponding gerbera homologs were identified from 444 the gerbera RNAseq database using BLAST searches (Supplemental Table S2). The full-length  Table S3).   Table S6. GhACTIN was used as an internal control. Statistical differences in expression levels 505 between the control and the transgenic samples were analyzed using the independent samples t-test.

506
In situ hybridization 507 The preparation of the plant samples, sectioning, and hybridization steps were performed as   Table S4). An alignment of the TCP domain was 519 generated using Clustal Omega and was converted into a corresponding nucleotide alignment. The 520 resulting codon alignment was then subjected to phylogenetic analysis. The best-fit substitution