We previously described that members of the S21 MYB transcription factor subfamily are targets of a CUL3-based E3 ligase that uses BPM proteins as substrate adaptors (Chen et al., 2015). In anticipation that these E3 ligases can potentially target additional MYB transcription factors outside the S21 subfamily, we analyzed the S23 subfamily, which includes three members: MYB1 (At3g09230; 393 AA; 42.81 kDa), MYB25 (At2g39880; 367 AA; 40.91 kDa), and MYB109 (At3g55730; 399 AA; 43.53 kDa) (Stracke et al., 2001). Homology and similarity comparisons by us (Supplementary Figure 1), and phylogenetic analysis done by Stracke et al. (2001) has shown that MYB25 and MYB109 group most closely together, while MYB1 forms a separate clade. It needs to be mentioned here that SALT-RELATED MYB1 (SRM1/MYBS2/At5g08520) (Wang et al., 2015), despite its listing on TAIR as MYB1¹, is not a member of the S23 subfamily, and has not been assigned to the R2R3 family (Dubos et al., 2010).

The R2R3 S23 MYB transcription factor subfamily contains several known motifs. Using the ELM server tool², R2R3 DNA binding motifs were identified in the N-terminal regions (MYB1: 54–103 and 106–154; MYB25: 49–98 and 101–146; MYB109: 55–104 and 107–155). In addition, PEST motifs were predicted by the epestfind tool³ for MYB25 (AA 186–200) and MYB109 (AA 177–191), but not for MYB1 (Figure 1). PEST motifs are enriched in the amino acids proline (P), glutamate (E), serine (S), and threonine (T), and are often connected with protein instability (Salama et al., 1994; Tsurumi et al., 1995; Berset et al., 2002). They have been demonstrated in some BPM substrates to be critical for BPM-substrate interaction (Mooney et al., 2019).

In humans and plants, a conserved SPOP-binding Consensus (SBC) motif was identified that facilitates interaction between substrate and BPM proteins (Zhuang et al., 2009; Morimoto et al., 2017). In humans, the cassette is often enriched in serine and threonine residues, and generally follows the sequence motif ϕ-π-S-S/T-S/T (ϕ, non-polar; π, polar; S, serine; T, threonine), while in plants a slightly modified version is also recognized by BPMs (ϕ-π-S-X-S/T, with X being any amino acid). Queries of the S23 MYB sequences revealed that MYB1 contains an SBC (LESSS, AA 350–355) and an SBC-like (LDSPT, AA 323–328) motif, while MYB25 only contains an SBC (LSSSS; AA 282–287) and MYB109 an SBC-like (ANSVT; AA 167–172) motif (Figure 1). The presence of SBC and/or PEST motifs indicates that all three MYB proteins may interact with BPM proteins and be targets of the ubiquitin-proteasome pathway, respectively. In agreement with this, in all three cases, the UbPred tool⁴ predicted several lysine residues that have a high confidence level to be ubiquitylation sites (Figure 1). MYB56–a previously confirmed CRL3^BPM substrate (Chen et al., 2015)–also contains PEST and SBC-like motifs as well as ubiquitylation sites (Figure 1). Because of these similarities, we decided to further verify whether the S23 MYB proteins are BPM substrates.

To test interaction between the S23 subfamily members and BPMs, we followed two approaches. Yeast 2-hybrid (Y2H) assays were performed between the three S23 MYBs and all Arabidopsis BPMs. As shown in Figure 2A, all of the BPMs are able to interact with MYB1, MYB25, and MYB109. To further corroborate these findings, pulldown assays were performed with three His-tagged BPM proteins that represent members from two of the three BPM clades present in Arabidopsis (BPM1, 3, and 5) (Weber et al., 2005), and GST-tagged MYB proteins. While GST:MYB1, :MYB25, and :MYB109 were all able to co-precipitate with the different His:BPMs, GST alone was not (Figure 2B). These results provide evidence that MYB1, MYB25, and MYB109 can be substrates of a CUL3^BPM E3 ligase.

To verify whether the S23 subfamily members are substrates of the 26S proteasome, we utilized a cell-free degradation assay that contains all the necessary components to degrade a protein through the ubiquitin proteasome pathway (Wang et al., 2009; Mooney et al., 2019). Purified, recombinant S23 proteins tagged with a GST incubated in native plant extracts and tracked over a 60 min time course to determine if their amount decreased over time (Figure 3). As controls we included GST:MYB56 and GST alone. In addition, the proteasomal inhibitor MG132 was used to test whether degradation depends on proteasomal activity. GST:MYB1, :MYB25, :MYB109, and :MYB56 were entirely degraded after 30 min, and in all cases degradation was inhibited in the presence of the proteasomal inhibitor MG132. In comparison, GST alone, which served as a negative control, was fully stable over the tested time range showing that degradation of the recombinant proteins was MYB-dependent. Overall, these results indicate that all members of the S23 subfamily are substrates of the 26S proteasome.

Since no detailed functional descriptions of S23 family members are available, we analyzed MYB1, MYB25, and MYB109 to determine expression patterns in tissue and intracellular locations.

Tissue specific expression patterns for MYB1, MYB25, and MYB109 were investigated by generating transgenic plants that expressed a GUS reporter under the control of a genomic construct that comprised the promoter (MYB1 and MYB25, 1600 bp upstream of the ATG; MYB109, 1592 bp upstream of the ATG) and the entire gene without the 3′ untranslated region (further referred to as proMYB1:MYB1:GUS, proMYB25:MYB25:GUS, and proMYB109:MYB109:GUS, respectively) (Figure 4A). Several independent transgenic lines were obtained for each construct and analyzed for GUS activity. Plants expressing the proMYB1:MYB1:GUS construct mainly showed GUS expression in developing leaves (cotyledons and young rosette leaves) and in the root tips (Figure 4B). Staining was not observable in stem, flowers or seeds, while there was strong staining at the pedicels connecting the flowers and siliques with the main stem (Figure 4B). Generally, expression of the reporter among the proMYB25:MYB25:GUS and proMYB109:MYB109:GUS constructs was very comparable (Figures 4C,D). The corresponding plants showed GUS expression in roots, young leaves, anthers, and the extremities of siliques, and expression patterns for the MYB25 and MYB109 expression constructs were highly comparable (Figures 4C,D).

While GUS protein was detectable in proMYB1:MYB1:GUS expressing plants, it was not detectable in the proMYB25:MYB25:GUS and proMYB109:MYB109:GUS lines. Figure 4E indicating that MYB25:GUS and MYB109:GUS are highly unstable, possibly due to degradation by the proteasome. Supporting this notion, GUS protein was detected (single band of around 100 kDa) in proMYB25:MYB25:GUS and proMYB109:MYB109:GUS plants treated with the proteasomal inhibitor bortezomib for 16 h, while no protein was detectable for bortezomib-treated wild type plants on Western blot (Figure 4D). We did observe increased protein amounts MYB1:GUS after plants were treated with bortezomib (Figure 4E). These data further corroborated findings from the cell-free degradation assays and provide strong evidence that all three S23 members are targets of the 26S proteasome in planta.

Subcellular localization studies were done with either a YFP-reporter (MYB1) or a GFP reporter (MYB25 and MYB109) fused C-terminally to the respective MYB protein. The constructs were expressed under the control of either a UBQ10 promoter (MYB1) or a 35S promoter (MYB25 and MYB109) and tested in transgenic Arabidopsis plants (Figures 5A–C). While MYB1:YFP was located in the cytosol and the nucleus (Figure 5A), the GFP-tagged MYB25 and MYB109 proteins predominantly accumulated in the nucleus where nuclear speckles were consistently observed, suggesting that the fusion proteins are localized at sites of active transcription (Spector and Lamond, 2011).

Based on sequence similarity and expression patterns, it seems likely that the S23 MYB family members are functionally redundant. To further evaluate in planta activity, we chose to characterize in greater detail (phenotypically and by RNA-seq) 35S:MYB25:GFP expressing plants. We worked with two independent transgenic lines from MYB25:GFP expressing plants, for which expression was confirmed through confocal microscopy and RT-qPCR analysis (Figure 5B and Supplementary Figure 2).

Since BPMs and other MYB transcription factors have been previously brought into context with abiotic stress responses (Jung et al., 2008; Lechner et al., 2011; Cui et al., 2013; Kim et al., 2013; Wang et al., 2015; Julian et al., 2019; Chico et al., 2020; Skiljaica et al., 2020), we analyzed MYB25:GFP overexpressor lines under abiotic stress conditions. We focused on the sensitivity of the plants at the germination stage since this is a straightforward and frequently used stage to investigate aberrant stress sensitivities in Arabidopsis. In particular, we tested osmotic (300 mM sorbitol), and salt stress (150 mM NaCl) conditions, as well as ABA treatments (0.5 and 0.75 μM). Monitoring the germination rate over time showed that MYB25:GFP seeds consistently germinated faster compared to wild type seeds on sorbitol, NaCl, and ABA (Figures 6A–E). Because ABA is connected with cold stress responses (Xiong et al., 2001; Zhu et al., 2005; Chen et al., 2009; Lu et al., 2019), we also tested our transgenic plants for germination sensitivity at 4°C, but did not observe any significant differences compared to wild type (Figure 6F).

Additionally, we generated a polycistronic artificial amiRNA construct that downregulates all three S23 MYBs (Supplementary Figure 3). Transgenic plants constitutively expressing the construct showed hypersensitivity toward salt and sorbitol as well as ABA at the germination stage (Supplementary Figure 4). These findings are genetically in agreement with the observed higher tolerance of the MYB25:GFP overexpression lines.

Sensitivities of MYB25:GFP overexpressor lines in root elongation assays was tested on salt (100 and 150 mM NaCl) and ABA (20 μM), but we only observed a significant difference to wild type on ABA (Figures 7A,B). While ABA was less effective in inhibiting elongation growth, it needs to be emphasized that MYB25:GFP seedlings already had significantly shorter roots than the wild type (Figure 7A). In addition to shorter roots, we also observed that MYB25:GFP overexpressor lines had slightly smaller seeds and an early flowering phenotype compared to wild type plants (Figures 7C–E), but were otherwise indistinguishable from wild type.

Overall, these findings indicated that MYB25 regulates osmotic and salt stress as well as ABA responses, and impacts Arabidopsis development in specific aspects, such as flowering time, seed development, and primary root length.

Because it is currently unclear what genes and processes are regulated by MYB25 in Arabidopsis, we decided to perform a transcriptional profiling on the 35S:MYB25:GFP overexpressing seedlings. We looked separately at root and shoot tissues due to the observed differences in GUS staining (Figure 4C).

As shown in the Venn diagram in Figure 8A, expression of a broad range of genes was significantly changed (p < 0.001, q < 0.05) in 35S:MYB25:GFP plants when compared to Col-0 wild type (for a complete list of genes see Supplementary Table 1). Interestingly, only a set of 151 genes was overlapping between shoot and root data, whereas the majority of expression changes were specific for either root or shoot. A significantly higher number of genes were changed in the shoots of 35S:MYB25:GFP plants (3261 genes), while in 35S:MYB25:GFP roots nearly seven-fold less genes (477 genes) showed aberrant expression compared to Col-0. We also verified RNA-seq data by RT-qPCR on four selected genes in independently grown 14-day old whole seedlings, and confirmed changes in gene expression (Supplementary Figure 5 and Supplementary Table 1) (log2 fold-change by RNA-seq: AT5G46830, +3.8 in shoots; AT2G30770, +2.5 in shoots; AT5G04120, −3.3 in roots; AT3G21352, 2.9 in roots). These genes were selected as they were among the most differentially regulated genes identified and demonstrate the role of MYB25 in influencing biotic and abiotic stress.

The Gene Ontology (GO) term enrichment analysis showed that a majority of the changes observed in the MYB25:GFP seedlings were related to cellular stress-response pathways, either to abiotic stress (e.g., salt, drought, osmotic, or reactive oxygen stress) or to biotic stress (e.g., jasmonic acid signaling, glucosinolate metabolism, and pathogen responses) (Figure 8B and Supplementary Table 2). For example, some of the stress responsive genes with aberrant expression in MYB25:GFP plants include slow type anion channel-associated homologue 1 (SLAH1) (Qiu et al., 2016), the dehydration-responsive element-binding protein 2C (DREB2C) (Je et al., 2014; Song et al., 2014), and responsive to dessication29a (RD29a) (Yamaguchi-Shinozaki and Shinozaki, 1994) (Supplementary Table 2). However, GO term enrichment analysis also identified changes in other processes related to, for example, general development or light response, and developmental pathways. Gene examples here include late elongated hypocotyl (LHY) (Park et al., 2016), somatic embryogenesis receptor-like kinase 4 (SERK4) (Albrecht et al., 2008), related to ABI3/VP1 1 (RAV1) (Fu et al., 2014), or circadian clock associated 1 (CCA1) (Lau et al., 2011; Park et al., 2016). GO enrichment analysis of biological pathways revealed that the roots of MYB25:GFP seedlings are altered in glucosinolate and glycosyl compound biosynthesis, while the shoots displayed aberrant biotic stress responses (Supplementary Figure 6). This is reflected among the most differentially regulated genes, as, for example, UDP-Glycosyltransferase superfamily protein members, UGT78D4 and UGT91A1 are strongly downregulated in the root, while jasmonate-zim-domain protein 10 (JAZ10) and tyrosine aminotransferase 3 (TAT3) are highly downregulated in the shoots (Figures 8C,D).

Overall, the data indicate that MYB25 fulfills distinct functions in root and shoot, and that it affects a broader range of processes in the shoot than in the root. They also show that many of the processes affected are related to stress response, while general developmental pathways are proportionally less affected, which corroborates with our phenotype observations.

As plant material, Arabidopsis thaliana (variety Col-0) were used. Plants were grown at 20°C, with a day:night cycle of 16 h light to 8 h dark. Arabidopsis plants cultured under sterile conditions grew on minimal medium without supplemental sucrose according to Estelle and Somerville (1987). Stable Arabidopsis transformations were accomplished following the floral-dip method (Clough and Bent, 1998), while transient expression in tobacco leaves was done as described by Sparkes et al. (2006). For all phenotypical and stress related experiments, stable transgenic plants of the T3 generation were used. Salt (NaCl) and abscisic acid (ABA) (Sigma-Aldrich, St. Louis, MO, United States) dependent germination assays were performed by plating seeds on minimal medium supplemented with different NaCl or ABA concentrations, respectively. Germination was defined when radicles first emerged from the seed coat. For salt-dependent root length assays, 3-day old Arabidopsis seedlings were transferred individually to plates supplemented with or without different NaCl concentrations. Growth of the primary root was measured daily for up to 2 weeks. For gene expression analysis in conjunction with salt stress, we adopted a protocol described in Wang et al. (2015). In brief, 9-day-old Col-0 seedlings, grown on vertical minimal media plates containing 1% agarose, were transferred to 5 mL minimal media solution supplemented with or without 200 mM NaCl for 4 h. Quantification of seed size was done using ImageJ software (Schneider et al., 2012). For flowering time points, seeds were directly brought out into soil, and transferred to individual pots 10 days later. Beginning of flowering was defined when first inflorescences became visible (time of bolting). All experiments were repeated at least three times independently as biological replicates.

Any genetic material for MYB1, MYB25, and MYB109 used in this work was amplified either from Col-0 genomic DNA or cDNA generated from total RNA via a High Capacity cDNA Reverse Transcription kit according to the kit’s manual (Applied Biosystems, Foster City, CA, United States). PCR-based amplification of DNA was accomplished with Phusion High-Fidelity DNA Polymerase under standard conditions (Thermo Fisher Scientific, Waltham, MA, United States). For the MYB1 and MYB25 promoters a 1600-bp long region upstream of the ATG was amplified, while for the MYB109 promoter this comprised 1592 bp. PCR products were first sub-cloned into pCR8 (LifeScience, Darmstadt, Germany) and sequenced before being further processed. For expression of GST-fusion proteins in Escherichia coli, MYB1, MYB25, MYB56, and MYB109 cDNAs were shuffled via LR-GATEWAY-reactions to pDEST15 (LifeScience, Darmstadt, Germany). A cDNA construct received from the Arabidopsis Stock Center Resource for MYB1 (DKLAT3G09230) was cloned into pHB2-GST (ABRC, Columbus, OH, United States) using NdeI/EcoRI sites. For expression of His-tagged BPM1, 3, 5 proteins, the corresponding gene was cloned into pET21b (Novagen Inc., Madison, WI, United States) using the NdeI/XhoI restriction sites. For in planta expression of C-terminally tagged MYB1:YFP fusion proteins under the control of a UBQ10 promoter was cloned into the XmaI/BamHI sites of the modified pGREEN vector pG20-YFP (Pratt et al., 2020). C-terminally tagged MYB25: and MYB109:GFP fusion proteins under the control of a 35S promoter, and C-terminal GFP constructs using the genomic DNA regions (introns and exons but no promoter) of MYB25 and MYB109 were generated in the Gateway-compatible vectors pMDC43 and pMDC83, respectively (Curtis and Grossniklaus, 2003). For GUS reporter assays, a promoter genomic MYB1 DNA region (introns and exons plus promoter) was cloned in the modified pGREEN vector pG20-GUS (Pratt et al., 2020) via Gibson technology using the HindIII/XmaI sites. Promoter:genomic constructs for MYB25 and MYB109 were cloned via Gateway technology using the destination vector pMDC163 (Curtis and Grossniklaus, 2003). A polycistronic artificial microRNA (amiRNA) construct with two independent amiRNAs was generated based on predictions from the WMD3 server⁵. One amiRNA targets MYB25 and MYB109 (CAGCGCTCTGGTTAATCTTT), while the second one targets MYB1 and MYB25 (TAAGATTTACCAGAGCGCCGT). No off targets were predicted. The two amiRNAs are co-expressed, and separated by RNASe P and Z cutting site according to Xie et al. (2015), that allow in planta processing to gain individual amiRNAs. The two amiRNAs are followed by an active ribosomal complimentary sequence (ARC) which functions as internal ribosome entry site to allow uncapped translation of any protein behind the ARC sequence (Chappell et al., 2000; Akbergenov et al., 2004; Pfeiffer et al., 2012). The construct was synthesized by Bio Basic Inc., Amherst, NY, United States, and cloned into the modified pGREEN vector gG20-GUS (Pratt et al., 2020) via Gibson technology using the SmaI/BamHI. The vector allows constitutive expression based on a UBQ10 promoter, and easy tracking of expression using its GUS reporter (Supplementary Figure 3). All primers used in this work are listed in Supplementary Table 3.

Recombinant proteins were expressed and purified using standard methods as described earlier (Chen et al., 2013). In brief, protein cultures were grown overnight at 37°C, diluted to an OD₆₀₀ of 0.1–0.3, and grown to an OD₆₀₀ of 0.6–0.8. The cultures were then induced with 0.1 M IPTG (Isopropyl β-D-1-thiogalactopyranoside), grown for an additional 2 h, pelleted, and stored at −20°C before protein extraction. Expressed proteins were affinity purified from E. coli via glutathione agarose beads (Sigma-Aldrich, St. Louis, MO, United States), or via Ni-NTA agarose beads (Sigma-Aldrich, St. Louis, MO, United States). For elution of GST-tagged recombinant proteins from beads, proteins were incubated for 1h at 4°C in GST-elution buffer (25 mM reduced glutathione; 50 mM Tris pH 8.8; 200 mM NaCl).

Tissue specific expression patterns of GUS reporter constructs introduced to Arabidopsis were assessed according to Jefferson (1989). In brief, plant tissues were harvested at various ages and incubated if not otherwise stated in GUS staining solution for 12 h at 37°C. Tissue was cleared using 70% ethanol and incubated in 50% ethanol prior to microscopic imaging. Plant tissues were imaged using a Leitz fluorescence stereomicroscope and a Leica MZ8 Camera. For detection of GUS protein, an α-β-Glucuronidase (N-Terminal) antibody (Sigma-Aldrich, St. Louis, MO, United States) was used. For bortezomib (ApexBio, Houston, TX, United States) treatment plants were incubated in liquid minimal medium supplemented with the inhibitor as indicated in the figure and legend for the duration of the experiment.

The constructs were initially screened for localization using transient expression in Nicotiana benthamiana epidermal cells as a preliminary assay as described by Chen et al. (2015). Subsequently, transgenic A. thaliana seedling roots were screened for GFP expression. Arabidopsis tissues were imaged using a Leica SP8 confocal compound microscope.

N-terminal GST-tagged MYB proteins were purified and eluted before incubated with His:BPM1, :BPM3, and :BPM5 that remained on Ni-NTA agarose beads in a Tris Buffer (0.1 M Tris pH 7.5, 150 mM NaCl, 0.5% NP-40 (Igepal CA-630) containing 100 mM PMSF protease inhibitor for 2 h at 4°C followed by three ten minute washes. Samples were boiled in a 4× SDS loading buffer (0.25 M Tris pH 6.8, 8% SDS, 20% β-mercaptoethanol), run on a 10% SDS-PAGE gel, and transferred to a 0.45 μm PVDF membrane (TISCH Scientific, North Bend, OH, United States) for immunodetection via monoclonal GST and His antibodies (LifeTein, Somerset, NJ, United States), and a horse radish-coupled secondary donkey anti-mouse (Santa Cruz Biotechnology, Dallas, TX, United States). Samples were initially detected using an α-GST antibody and stripped and re-probed using an α-His antibody.

Yeast 2-Hybrid studies were done as described earlier (Weber et al., 2005). Selection medium II (SDII) was supplemented with histidine and uracil and represented a transformation control. SDIV minimal medium lacked uracil and histidine and was used for interaction studies. Photos were taken from individual spots 1 week after plating.

Cell-free degradation assays generally followed a protocol described by Wang et al. (2009). In brief, a master mix that allowed 20 ul samples to be removed at each time point of eluted recombinant GST-tagged proteins (100 ng) were incubated in native plant extract (200 μg) from 12-day-old A. thaliana seedlings (Chen et al., 2013). GST-tagged proteins were incubated in a plant buffer (4 mM PMSF, 10 mM ATP, and 5 mM DTT) rocking at room temperature, with samples taken at 0, 30, and 60 min. An additional sample was treated with the proteasome inhibitor MG132 (20 μM; Selleckchem, Houston, TX, United States), and incubated for the duration of the experiment.

Approximately 50 mg of tissue was frozen in liquid nitrogen and ground using a mortar and pestle. RNA was extracted from Arabidopsis tissues using a RNeasy Plant Mini RNA Kit (Qiagen, Germantown, MD, United States) with an RLT lysis buffer as described by the manufacturer. Reverse transcription was performed using an Applied Biosystems high-capacity kit for reverse transcription into cDNA. The cDNA was either used for basic PCR (GFP, DDB1a) or for RT-qPCR (MYB1, MYB25, MYB109, ACTIN2, BHLH28, CYP71A13, dPGM, AT3G21352). For expression analysis by RT-qPCR an ABI 7500 Fast Real Time PCR machine was used. All primers used are listed in Supplementary Table 3.

Overexpression lines for MYB25:GFP, along with Col-0, were grown vertically in sterile culture on minimal medium and harvested at 14-days after plating. Seedling roots and shoots were separated for each line and used for RNA extraction with a RNeasy Plant Mini RNA (Qiagen, Germantown, MD, United States) extraction kit. RNA extractions were collected in triplicates. RNA sequencing and quality control were done at the Oregon State University’s Center for Genome Research and Biocomputing with each sample barcoded, pooled, and split between two Illumina HiSeq 3000 lanes for sequencing (150 bp single end). The barcodes from each sample were trimmed, and the samples were filtered for quality. Sequencing data were processed using the Tuxedo Suite and analyzed using the protocol from Trapnell et al. (2012). In brief, reads were aligned with Top Hat against the TAIR10 A. thaliana genome, and transcripts were assembled using Cufflinks. The reads were analyzed for differential gene expression using Cuffdiff. Visualization of data was performed using Cummerbund on significant results (p < 0.01, q < 0.05 for MYB25 root and shoot). GO analysis was performed using GOTermFinder⁶. The most differentially regulated GO processes were selected for the creation of pie charts and heatmaps. Further heatmaps were generated with HeatMapper⁷ expression plots. Additional GO enrichment analysis of the top 30 differentially regulated biological process pathways was performed using ShinyGO v0.66⁸ to demonstrate the interplay of these GO categories (p < 0.05).