Specificity in auxin responses is not explained by the promoter preferences of activator ARFs

Auxin is essential for almost every developmental process within plants. How a single small molecule can lead to a plethora of downstream responses has puzzled researchers for decades. It has been hypothesized that one source for such diversity is distinct promoter-binding and activation preferences for different members of the AUXIN RESPONSE FACTOR (ARF) family of transcription factors. We systematically tested this hypothesis by engineering varied promoter sequences in a heterologous yeast system and quantifying transcriptional activation by ARFs from two species, Arabidopsis thaliana and Zea mays. By harnessing the user-defined and scalable nature of our synthetic system, we elucidated promoter design rules for optimal ARF function, discovered novel ARF-responsive promoters, and characterized the impact of ARF dimerization on their activation potential. We found no evidence for specificity in ARF-promoter interactions, suggesting that the diverse auxin responses observed in plants may be driven by factors outside the core auxin response machinery.

binding and activation preferences for different members of the AUXIN RESPONSE FACTOR 23 (ARF) family of transcription factors. We systematically tested this hypothesis by engineering 24 varied promoter sequences in a heterologous yeast system and quantifying transcriptional 25 activation by ARFs from two species, Arabidopsis thaliana and Zea mays. By harnessing the 26 user-defined and scalable nature of our synthetic system, we elucidated promoter design rules 27 for optimal ARF function, discovered novel ARF-responsive promoters, and characterized the 28 impact of ARF dimerization on their activation potential. We found no evidence for specificity in 29 7 al., 2019). We tested how AuxRE sequence impacts activation by AtARF5 and AtARF19 on two 170 AuxREs facing towards each other by comparing activation on the AuxREs TGTCTC/GAGACA 171 and on the AuxREs TGTCGG/CCGACA. We found that all tested ARFs activate more strongly 172 on the TGTCGG/CCGACA AuxREs (Figures 2A and 2B). The difference in AtARF5 activation 173 on the canonical AuxRE sequence and the novel sequence, nearly a nine-fold increase, was 174 striking. In combination with previous protein binding microarray data (Boer et al., 2014), this 175 may suggest AtARF5 has a strong preference for activation on TGTCGG/CCGACA, at least 176 with this promoter orientation and spacer. Similarly, while the maize ARF5-like protein ZmARF4 177 does not activate well on TGTCTC/GAGACA, it does show transcriptional activity on the 178 TGTCGG/CCGACA AuxREs at levels similar to ZmARF27. These results again do not show 179 divergent promoter preferences among ARFs-while the relative degree of preference may 180 differ between ARFs, they all activate more strongly on the same promoter variant. 181 182

AtARF19 can activate on a single AuxRE in yeast 183
Our results suggested that the AuxRE sequence TGTCGG and its reverse complement may be 184 more optimal than the canonical AuxRE for ARF activation on the promoter. While common 185 synthetic auxin responsive reporters have high copy numbers of AuxREs within a short 186 sequence, in native auxin responsive promoters it is rare for two AuxREs to occur close 187 together (Grigolon et al., 2018 (Korasick et al., 2014, Nanao et al., 2014. Structural studies indicate that ARFs require 199 dimerization at the DD to bind to DNA (Boer et al., 2014). In addition, mutations in either DD or 200 PB1 of AtARF19 reduce ARF activity (Pierre-Jerome, 2016), though these studies only 201 addressed ARF behavior on promoters with multiple AuxREs. We tested the activity of AtARF19 202 mutations that disrupt ARF dimerization in either the DD (G247I and A50N) or the PB1 domain 203 (termed ARF19 KO-a triple mutation K962A; D1012A; D1016A) (Pierre-Jerome, et al. 2016) 204 and compared these to the activity of a DNA-binding mutant AtARF19 H138A ( Figure 3A, B). 205 The dimerization mutations caused a loss of activation on the single AuxRE (TGTCGG) 206 promoter to nearly the same extent as the DNA-binding mutation ( Figure 3C), suggesting that 207 dimerization is necessary for ARF activation on the promoter despite the presence of only a 208 single optimal binding site. Interestingly, when we tested the activity of these dimerization 209 mutants on the two TGTCGG AuxREs facing towards each other, they caused a loss of 210 activation but not to the same extent as on the single AuxRE, suggesting that multiple AuxRE 211 sites may compensate for a loss of dimerization of the ARFs themselves. As ARFs were 212 crystallized as a dimer pair with each monomer bound to a separate AuxRE (Boer et al., 2014), 213 how an ARF dimer contacts the DNA when there is a single AuxRE present is unknown. It is 214 possible that only a single ARF-AuxRE interaction is required to bring the dimer to the DNA, and 215 the other ARF forms transient interactions with multiple DNA sequences, which may serve as 216 cryptic, low-affinity binding sites. Or the proximity of ARFs within a dimer pair may allow one to 217 bind a single AuxRE promoter as soon as the other falls off, increasing the on rate of ARF 218 binding to the promoter. 219 220 AtARF19 has a unique residue in the dimerization domain required for activity on a single 221

AuxRE 222
Alignments among Arabidopsis and maize ARFs ( Figure 3A) showed a difference in sequence 223 within the DD of AtARF19 when compared to its maize homologues ZmARF27 and ZmARF35 224 ( Figure 3B). We hypothesized that this single residue difference, so close to the monomer-to-225 monomer contact residues within the DD, could explain AtARF19's unique ability to activate 226 transcription on promoters with only a single AuxRE. To test this, we generated a mutated form 227 of AtARF19 that replaced the asparagine residue with an alanine, the same amino acid found in 228 ZmARF27 (N256A). This single residue change abolished AtARF19 activity on a single AuxRE, 229 while leaving its activity on a two-AuxRE promoter essentially unchanged ( Figure 3D). The 230 polarity of the asparagine may help stabilize the dimeric form of AtARF19, leading to higher 231 transcriptional activation overall and greatly increasing the number of potential promoters it can 232 act on. While N256 is necessary for AtARF19's ability to activate on promoter with a single 233 AuxRE, it is not sufficient. AtARF7, which shares the same asparagine residue in its DD, cannot 234 activate on a single AuxRE (Supplemental Figure S2). This difference, in combination with the 235 critical role of the PB1 domain in ARF transcriptional activation ( Figure 3C), implicates the still 236 poorly understood inter-domain interactions in determining overall protein function. 237

238
Discussion 239   240 It has been widely speculated that specificity within ARF-promoter interactions is responsible for 241 the observed diversity in transcriptional and developmental responses triggered by auxin. Our 242 results suggest that this model is unlikely to be true, at least among Class A ARFs. All the ARFs 243 tested showed similar promoter preferences, and all required dimerization for full activity. We 244 were able to elucidate a set of promoter design rules for maximizing response across the A 245 clade, and found that these design rules were conserved across Arabidopsis and maize. Simply 246 stated, these rules are as follows: (1) ARFs most strongly activate on promoters with at least 247 four AuxREs arranged facing towards one another ( Figure 1); (2) the non-canonical TGTCGG 248 sequence can further boost expression, especially by ARFs in the AtARF5 clade ( Figure 2). This 249 second rule has relevance for the design and interpretation of auxin reporters. For example, 250 DR5v2, which uses TGTCGG (Liao et al., 2015), may over-report responses driven by AtARF5 251 and its homologues relative to other Class A ARFs. Our study also highlights the complexity of 252 inter-domain interactions within the ARFs, as dimerization at both N-and C-terminal 253 dimerization domains was found to be critical for maximal transcriptional activation. 254 255 The differences between the architecture of auxin reporters and native auxin responsive 256 promoters are striking. The rules derived from the systematic analysis presented here are 257 generally consistent with the construction of auxin reporters, where there is a trend towards high 258 copy numbers of canonical AuxREs in a short sequence space (Ulmasov et al., 1997a;Ulmasov 259 et al., 1997b). Closely spaced AuxREs are found only rarely in the Arabidopsis genome 260 (Grigolon et al., 2018), and frequently are neither the ideal sequence nor in the ideal orientation 261 relative to the TSS. One possible explanation for the rarity of "ideal" auxin promoters is that it 262 allows for integration of signals from multiple pathways, a hypothesis supported by the 263 enrichment for transcription factor binding sites for other proteins in auxin-responsive promoters. 264

265
Our results showed that heterodimerization between ARFs is essential for ARF function, but 266 importantly heterodimerization between ARFs and other transcription factors could support ARF 267 activity on non-ideal native promoters and potentially act as a locus for specificity within auxin 268 response. Bioinformatics analyses of auxin-induced genes show that many promoters of these 269 As we continue to elucidate the rules of ARF-activated transcription, synthetic tools should 283 make it possible to examine each of these aspects in turn. Future efforts that combine synthetic 284 and native approaches will ultimately be needed to pinpoint the combination of factors that 285 make up the "auxin code", as well as to make it possible to retrace the evolutionary path that 286 connected novel auxin response modules to diversity in plant form and function. 287 288

290
Yeast integrating plasmid construction 291 Oligonucleotides were obtained from Integrated DNA Technologies with standard desalting 292 purification. All cloning was done by Gibson assembly unless otherwise specified, using 293  Figure 1 Arabidopsis and maize ARFs share promoter preferences. A) Schematic of yeast engineered to constitutively express ARF proteins and promoter variants. All promoter variants were inserted into the A1 site of a pIAA19 promoter with mutated AuxREs. The transcription start site (TSS) is to the right and arrowheads indicate the orientation of the AuxRE, starting with 5'-TGTC-3'. Fluorescence was measured by flow cytometry with the results depicted as median values and 95% confidence intervals. B) AtARF19 and AtARF5 show strong activation on promoters with four AuxREs (five base pair spacer). C) AtARF19 and AtARF5 show stronger activity on promoters with two AuxREs facing towards each other rather than away from each other (seven base pair spacer). D) ZmARF4, ZmARF27, and ZmARF29 show stronger activity on promoters with two AuxREs facing towards each other rather than away from each other (seven base pair spacer). E) AtARF19 and AtARF5 show stronger activity on promoters where the two AuxREs face towards rather than away from the TSS (five base pair spacer). F) ZmARF4, ZmARF27, and ZmARF29 show stronger activity on promoters where the two AuxREs face towards rather than away from the TSS (five base pair spacer).