Overexpression of AHL proteins enhances root hair production by altering the transcription of RHD6‐downstream genes

Abstract AT‐HOOK MOTIF NUCLEAR LOCALIZED (AHL) proteins occur in all sequenced plant species. They bind to the AT‐rich DNA sequences in chromosomes and regulate gene transcription related to diverse biological processes. However, the molecular mechanism underlying how AHL proteins regulate gene transcription is poorly understood. In this research, we used root hair production as a readout to study the function of two Arabidopsis AHL proteins, AHL17, and its closest homolog AHL28. Overexpression of AHL17 or AHL28 greatly enhanced root hair production by increasing the transcription of an array of genes downstream of RHD6. RHD6 is a key transcription factor that regulates root hair development. Mutation of RHD6 completely suppressed the overproduction of root hairs by blocking the transcription of AHL17‐activated genes. The overexpression of AHL17 or AHL28, however, neither affected the transcription of RHD6 nor the accumulation of RHD6 protein. These two AHL proteins also did not directly interact with RHD6. Furthermore, we found that three members of the Heat Shock Protein70 family, which have been annotated as the subunits of the plant Mediator complex, could form a complex with both AHL17 and RHD6. Our research might reveal a previously unrecognized mechanism of how AHL proteins regulate gene transcription.


| INTRODUCTION
AT-HOOK MOTIF NUCLEAR LOCALIZED (AHL) is a group of proteins that exist in eukaryotes, including all sequenced plant species (Zhao et al., 2014). AHL proteins contain two conserved structural units: an AT-hook motif located at their N-terminus and a plant-and prokaryote-conserved (PPC) domain located at their C-terminus (Fujimoto et al., 2004). The AT-hook motif has a conserved palindromic core sequence of Arg-Gly-Arg, which binds to the AT-rich region in the minor groove of DNA (Aravind & Landsman, 1998;Reeves & Nissen, 1990;Strick & Laemmli, 1995). The PPC domain is involved in protein-protein interaction and nucleus targeting (Zhao et al., 2013). In higher plants, AHL proteins have been shown to regulate a variety of processes involved in growth and development, such as flower initiation (Yun et al., 2012), flower organ patterning (Ng et al., 2009), vascular tissue patterning (Zhou et al., 2013), leave Qike Zeng and Li Song contributed equally to this work. senescence (Lim et al., 2007), hypocotyl growth (Zhao et al., 2013), pollen wall formation (Lou et al., 2014), gibberellin biosynthesis (Matsushita et al., 2007), axillary meristem maturation (Karami et al., 2020), and embryo development (Karami et al., 2021). The model plant Arabidopsis thaliana (Arabidopsis) has 29 AHL family members, which can be divided into two phylogenic clades ( Figure S1A). Several Arabidopsis AHL proteins have been shown to regulate gene transcription by binding to a specific AT-rich DNA sequence and by interacting with some transcription factors (Lee & Seo, 2017;Ng et al., 2009;Zhao et al., 2013). To date, however, it is still unknown whether nuclear proteins other than the general transcription machinery and transcription factors are involved in AHL-mediated gene transcription in higher plants.
In this research, we used root hair production in Arabidopsis as a readout to study the working mechanism of two Arabidopsis AHL proteins, AHL17 and AHL28, in regulating gene transcription. Root hairs are tubular outgrowths of root epidermal cells (Dolan et al., 1994;Gilroy & Jones, 2000). Their formation is controlled by a complex regulatory network of gene transcription. In Arabidopsis, the fate of root hair cells is determined by a homeodomain-type transcription factor GLABRA2 (GL2) (Di Cristina et al., 1996;Masucci & Schiefelbein, 1996;Rerie et al., 1994). After root hair cells are specified, the signaling components downstream of GL2 promote root hair initiation and elongation, which involves cell expansion. These signaling components include the following bHLH-type transcription factors: ROOT HAIR DEFECTIVE6 (RHD6), its homologous proteins RHD6-LIKE1 to 5 (RSL1-5) (Menand et al., 2007), and orthologs of Lotus japonica-ROOTHAIRLESS1-LIKE1 to 3 (LRL1-3) (Karas et al., 2009). Among them, RHD6 and RSL1 act immediately downstream of GL2 to control the transcription of RSL2-5 and LRL1-3, in which RHD6 plays a key role (Bruex et al., 2012;Lin et al., 2015;Pires et al., 2013;Yi et al., 2010). Downstream of these transcription factors are the proteins that directly participate in cell expansion, which involves several cellular processes, such as Ca 2+ and H + transport, cytoskeleton reorganization, vesicular trafficking, ROS production, synthesis of cell wall materials, and cell wall modifications (Grierson et al., 2014;Ishida et al., 2008).
Heat Shock Protein 70s (HSP70s) are the most abundant proteins induced by high temperature and are highly conserved in both prokaryotes and eukaryotes (Nover & Scharf, 1997). They function as molecular chaperons that help proteins fold properly and that prevent proteins from forming aggregates under stress conditions (Usman et al., 2017). HSP70s are also involved in protein translation and translocation and in stabilization of the cell membrane. In Arabidopsis, the HSP70 family has 18 members, which are located in various subcellular compartments (Usman et al., 2017). The Arabidopsis mutants with knockout of multiple HSP70 genes exhibited altered developmental phenotypes and decreased tolerance to several abiotic stresses (Leng et al., 2017).
Using combined genetic and molecular approaches, we found in the current research that overexpression of AHL17 and AHL28 promoted root hair production through enhancing the transcription of an array of the genes downstream of RHD6. Further investigation showed that three members of the HSP70 family might serve as the molecular links between AHL17/AHL28 and RHD6 in regulating gene transcription. Our research therefore revealed a previously unrecognized mechanism of how AHL proteins regulate gene transcription. This work also established an ideal experimental system for further studying the function of AHL proteins and may have some applications in engineering transgenic crops with high water and nutrition efficiency.
2 | RESULTS 2.1 | Overexpression of AHL17 and AHL28 enhances root hair production We previously identified an Arabidopsis mutant, hps5, that has enhanced root hair production . hps5 carries a gainof-function mutation in the ethylene receptor ERS1 and its roots display constitutive ethylene responses. In roots of hps5, 61 genes are upregulated relative to the wild type (WT, Columbia-0 background).
When one of the upregulated genes, that is, AHL17 (AT5G49700) was overexpressed in WT plants under a strong constitutive CaMV 35S promoter ( Figure S1B), the transgenic plants (35S::AHL17) exhibited enhanced root hair production, in terms of both root hair length and root hair density (Figure 1a-b). The increased root hair density in the overexpressing lines suggested that some of the non-hair cells (atrichoblasts) were transformed into root hairs. This inference was confirmed by histological analyses (Figure 1d-e) and quantification of ectopically produced root hairs (Table S1). AHL28 (AT1G14490) is the closest homolog of AHL17 in the Arabidopsis AHL protein family ( Figure S1A). The overexpression of AHL28 ( Figure S1B) also caused ectopic root hair formation and increased root hair length and root hair density (Figure 1a,c-e, Table S1). The length of primary roots and root epidermal cells of 35S::AHL17 and 35S::AHL28 plants did not differ from those of the WT ( Figure S2), excluding the possibility that the increased root hair density in the transgenic plants was due to the change of the length of epidermal cells.
We next examined the root hair phenotypes of the single and double mutants of AHL17 and AHL28, which were generated by the CRISPAR/Cas9 technique ( Figure S3A-B). Root hair development of these single and double mutants did not significantly differ from that of the WT ( Figure S3C), suggesting that there might be genetic redundancy among AHL17, AHL28, and other AHL proteins.

| Gene expression patterns and subcellular localization of AHL17
Before investigating the molecular functions of AHL17, we first determined the expression patterns of the AHL17 gene and the subcellular localization of AHL17 protein. A 1.2-kb promoter sequence of AHL17 was fused to the GUS reporter gene and introduced into WT Arabidopsis plants. Eleven independent AHL17::GUS lines were generated, and the results from a representative line are presented here. AHL17:: GUS was mainly expressed in the vascular tissues of primary and lateral roots and young leaves (Figure 2a-e). It was also expressed in the stigma and at the junctions between flower/silique and pedicles . No GUS activity was detected in root apical meristem and root hair cells. A similar vascular tissue-preferred expression pattern was observed in the roots of the transgenic gene when the GUS gene was replaced with a GFP gene (Figure 2k). In the AHL17::GFP lines, however, GFP signal could also be detected in root apical meristem and throughout a whole root hair (Figure 2l-m). This was probably because the transcription of AHL17 was very low and the GFP protein was more stable than the GUS protein. We then generated the transgenic lines expressing the AHL17-GFP fusion gene under AHL17's own promoter. In the roots of these lines, the AHL17-GFP F I G U R E 1 Root hair phenotypes of AHL17and AHL28-ovrexpressing lines. (a) Root hair phenotypes of 7-day-old seedlings of the WT, three independent 35S::AHL17 lines and 35S:: AHL28 lines. (b) and (c) Root hair length and density of the WT, 35S::AHL17 lines and 35S:: AHL28 lines shown in (a). Values are means ± SD of at least 15 roots for each line. Asterisks indicate a significant difference from the WT (ttest, **P < .01). (d) Roots of 7-day-old seedlings of the WT, 35S::AHL17, and 35S::AHL28 lines were examined under a stereomicroscope. Red arrows indicate the root hairs produced from the non-hair position. (e) Cross-section of 7-day-old seedlings of the WT, 35S::AHL17, and 35S::AHL28 lines. Red arrows indicate root hairs produced at non-hair positions.

| AHL17 and AHL28 enhance root hair production via RHD6
RHD6 is the key regulator of the signaling cascade below GL2 and plays a critical role in root hair initiation and elongation (Franciosini et al., 2017;Huang & Zheng, 2014). Downstream of RHD6 are the proteins that directly participate in cell expansion, resulting in root hair initiation and elongation (Grierson et al., 2014;Ishida et al., 2008). To determine whether the function of AHL17 in root hair production depended on RHD6, we crossed the 35S::AHL17 plant (hereafter, referred as AHL17 OX plant) with rhd6 and generated progeny that were homozygotes for both the 35S::AHL17 transgene and the rhd6 allele (AHL17 OX rhd6 line). These plants showed strong defects in root hair production, which were similar to the defects of rhd6 in terms of both root hair length and root hair density (Figure 3a 2.4 | Overexpression of AHL17 and AHL28 enhances the transcription of root hair-related genes downstream of RHD6 To understand why the function of AHL17 in root hair production was dependent on RHD6, we compared the gene transcription profiles among the WT, 35S::AHL17, rhd6, and AHL17 OX rhd6. The cDNA libraries were constructed using the total RNA isolated from the roots of 7-day-old seedlings. For each genotype, three biological replicates were used. Log2 ≥ 1 or ≤ À1 (2-fold change in expression levels) and FDR (false discovery rate) ≤ .01 were used as the cut-off for the selection of the genes whose expression levels significantly differed from those of the WT. In the AHL17 OX line, there were 544 genes whose expression was upregulated and 361 genes whose expression was downregulated compared with the WT (Table S2). GO term analysis indicated that the differentially expressed genes are enriched in several biological processes, including trichoblast differentiation, NADPH regeneration, photosynthesis, and response to vitamin (Table S3).
Among the genes related to trichoblast differentiation were those involved in the formation of Ca 2+ gradient in the root tip, synthesis and modifications of cell wall components, small G protein-mediated signal transduction, cytoskeleton rearrangement, and ROS production.
In rhd6, there were 34 genes whose expression was upregulated and 313 genes whose expression was downregulated compared with the WT (Figure 3e and Table S2). The gene expression heatmap clearly showed that almost all of the downregulated genes in rhd6 were upregulated in the AHL17 OX line (Figure 3e). We then compared the RNA expression profiles among AHL17 OX, rhd6, and AHL17 OX rhd6 lines. We found that 81 genes whose expression was upregulated in AHL17 OX but downregulated in rhd6 were also downregulated in the AHL17 OX rhd6 line ( Figure 3e and Table S4). Among these 81 genes, some genes, including COW1 (Grierson et al., 1997), PRP3 (Bernhardt & Tierney, 2000), LRX1 (Baumberger et al., 2003), six RHS (ROOT HAIR-SPECIFIC) (Won et al., 2009), and five EXTENSIN (Velasquez et al., 2011), have been experimentally demonstrated to be involved in root hair development. There were also other cell wall proteins, cell wall modification enzymes, peroxidases, and so forth.
Hereafter, collectively, these genes are referred as F I G U R E 3 Effects of AHL17 overexpression are suppressed by RHD6 mutation. (a)-(d) Root hair phenotypes of 7-day-old seedlings of the WT, rhd6, AHL17 OX, and AHL17 OX rhd6, respectively. In (c) and (d), a one-way ANOVA analysis was carried out for the whole dataset and post hoc comparisons were conducted using the SPSS Tukey HSD test at P < .05 level. (e) Hierarchical clusters displaying the differentially expressed genes in the 7-day-old seedlings of AHL17 OX, rhd6, and AHL17 OX rhd6 lines compared with the WT. (f) Relative expression of six selected genes in the 7-day-old WT, rhd6, AHL17 OX, and AHL17 OX rhd6 seedlings as determined by qPCR. Values are means ± SD of three technical replications in one experiment. The experiment was repeated three times with similar results. The expression of each of these genes in the WT was set to 1.0. Asterisks indicate a significant difference from the WT (t-test, **P < .01). (g) ChIP-qPCR assays of the binding of AHL17 to the promoters of the COW1 and PRP3 genes. Chromatins from 35S::GFP-AHL17 transgenic plants were isolated and immuno-precipitated with GFP antibodies. The levels of enrichment of the precipitated DNA fragments were quantified by qPCR assays. The ChIP signal was normalized to TUB8 gene. Values are means ± SD of three technical replicates. Experiments were repeated two times with similar results. Asterisks indicate a significant difference from that of TUB8 gene (t-test, *P < .05, **P < .01). Schematic diagrams showing the putative AT-hook binding elements in the promoters of these two genes are placed on the top. Black vertical lines indicate the putative AT-hook binding elements. Letters P represent the PCR fragments. (h) EMSAs show the binding of AHL17 to the putative AT-hook binding elements in the promoters of the COW1 and PRP3 genes. The relative positions of the probes used in EMSAs in each promoter are indicated with red triangles in (g). The unlabeled probe (cold probe) with different concentrations was used as a competitor.
RHD6-downstream root hair-related genes. To confirm the results of the RNA-seq experiment, we performed qRT-PCR to analyze the expression of six root hair-related genes, that is, LRX1, COW1, PRP3, RSH15, PER8, and EXT17, in different genotypes. The expression patterns of these genes shown by qRT-PCR experiments ( Figure 3f) and RNA-seq (Table S4) were consistent, indicating that the RNA-seq experiments were set up properly.
The above results suggested that AHL17 enhanced root hair production through increasing the transcription of root hair-related genes downstream of RHD6.

| AHL17 and AHL28 bind to the promoters of RHD6-downstream genes
Next, we determined whether AHL17 and AHL28 could bind to the promoters of RHD6-downstream genes. We first analyzed DNA sequences of the promoters of COW1 and PRP3 by PlantPan 3.0 (http://plantpan.itps.ncku.edu.tw/index.html). The potential AT-hook binding motifs were identified in the promoters of these two genes ( Figure 3g and Table S5). We then conducted chromatin immunoprecipitation assays followed by quantitative PCR (ChIP-qPCR) in the roots of 7-day-old 35S::GFP-AHL17 seedlings. Chromatin was isolated, cross-linked, and precipitated with anti-GFP antibodies. The DNA fragments that precipitated with GFP-AHL17 proteins were analyzed by qPCR. The primers were designed to amplify the regions that contain one or more putative AT-hook binding motifs. The results showed that some of the motifs in the promoters of these two genes amplified by qPCR were significantly enriched compared with that of the TUB8 gene (Figure 3g), suggesting that AHL17 could bind to the promoters of RHD6-downstream genes in vivo.
To further confirm the physical association of the AHL17 and AHL28 proteins with the promoter of COW1 and PRP3, we carried out electrophoretic mobility shift assays (EMSA) using the full-length recombinant AHL17 and AHL28 proteins produced in E. coli cells. The recombinant AHL17 protein contained an MBP (maltose-binding protein) tag at its N-terminus and a 6Â His tag at its C-terminus (MBP-AHL17-His), whereas the recombinant AHL28 protein only contained an MBP tag at its N-terminus (MBP-AHL28). The DNA probes used in EMSA were sequences that displayed high binding affinity to AHL17 in the ChIP assays (the sequences of the DNA probes are listed in  Figure S7A). We then performed a Western blot experiment using the proteins extracted from the roots of plants carrying the RHD6::RHD6-GFP construct in the WT, 35S:: AHL17, or 35S::AHL28 background using anti-GFP antibodies. The results showed that the levels of RHD6-GFP were not affected by the overexpression of AHL17 or AHL28 ( Figure S7B).

| AHL17 and AHL28 do not directly interact with RHD6
Next, we used the luciferase complementation imaging (LCI) technique (Chen et al., 2008) to determine whether AHL17 and AHL28 enhanced the expression of RHD6-downstream genes via direct protein-protein interactions with RHD6. We first investigated whether AHL17 and AHL28 could interact with themselves and with each other. The coding sequences (CDS) of AHL17 and AHL28 were fused to the N-terminal (nLUC) or C-terminal half (cLUC) of the LUC gene and were co-transformed into the leaves of N. benthamiana with proper controls.
The co-expression of cLUC-AHL17 and AHL17-nLUC, cLUC-AHL28 and AHL17-nLUC, cLUC-AHL17 and AHL28-nLUC, or cLUC-AHL28 and AHL28-nLUC all resulted in a strong fluorescence signal from reconstituted LUC activity ( Figure S8A-B). In contrast, no reconstituted LUC signals were seen for the co-expression of cLUC-AHL17 and nLUC, cLUC and AHL17-nLUC, cLUC-AHL28 and nLUC, or cLUC and AHL28-n-LUC. These results indicated that AHL17 and AHL28 could interact with each other as well as with themselves. We then assessed the interaction between AHL17 and AHL28 with RHD6, RSL1, RSL2, RSL4, and GL2. The CDS of these transcription factors were fused to the N-terminal or C-terminal half of the LUC gene and co-transformed with corresponding AHL17-and AHL28-fused LUC constructs into the leaves of N. benthamiana. No signals of reconstituted LUC activity were detected in any of these combinations, indicating that AHL17 and AHL28 did not directly interact with RHD6 and other transcription factors ( Figure S8A-B). The absence of protein-protein interactions between RHD6 with AHL17 or between AHL28 with RHD6 was confirmed by bimolecular fluorescence complementation (BiFC) assays in the leaves of N. benthamiana ( Figure S8C).

| AHL17 and AHL28 directly interact with HSP70 proteins
Because AHL17 did not directly interact with RHD6, we wanted to identify the missing component that linked AHL17 and RHD6. An immunoprecipitation (IP) assay was performed using total proteins extracted from 7-day-old seedlings of 35S::GFP-AHL17. The proteins in the extracts were precipitated with anti-GFP antibodies and were subjected to mass spectrometer analysis. Among the precipitated proteins, AHL17 and AHL28 had the first and third broadest peptide spectrum matches of the peptides that resulted from trypsin digestion (Table S6). These results further suggested that AHL17 and AHL28 function as a complex. Five other AHL family proteins were also coprecipitated with AHL17, including AHL15, 19, 22, 24, and 27 (Table S6). At least one peptide, which was specific to each AHL protein, was detected in the IP assay. All of these AHL proteins belong to the same clade in the AHL family of Arabidopsis ( Figure S1A). Interestingly, the precipitated proteins with the second, fourth, and fifth highest coverage scores were HEAT SHOCK PROTEIN 70-1 (HSP70-1, AT5G02500), HSP70-2 (AT5G02490), and HSP70-4 (AT3G12580), respectively (Table S6). These proteins have also been annotated as the subunit 37e, 37d, and 37c of the plant Mediator complex (Mathur et al., 2011).
To verify the physical interactions between AHL17 and these HSP70s in planta, we performed LIC assays in the leaves of N. benthamiana. The CDS of AHL17 was fused to the N-terminal half of LUC (AHL17-nLUC), and the CDS of HSP70-1 was fused to the C-terminal half of LUC (cLUC-HSP70-1). The leaves that were cotransformed with AHL17-nLUC and cLUC-HSP70-1 displayed strong fluorescence, whereas those co-transformed with AHL17-nLUC and cLUC-HSP70-5 or AHL19-nLUC and cLUC-HSP70-1 displayed no fluorescence, demonstrating that the interaction between AHL17 and HSP70-1 in vivo was sequence-specific ( Figure 4a). Next, we performed co-immunoprecipitation (Co-IP) assays in leaves of N. benthamiana that co-expressed myc-tagged HSP70-1 (HSP70-1-myc) and GFP-tagged AHL17 (AHL17-GFP). The results showed that the HSP70-myc protein could be co-precipitated with AHL17-GFP by anti-GFP antibodies (Figure 4b). The interaction between HSP70s and AHL17 was further confirmed in the nucleus of the leaves of N. benthamiana by BiFC assays (Figure 4c).
To determine which part of AHL17 interacts with HSP70, we divided AHL17 into three parts: (1)  we found that the PPC domain but not the AT-hook domain of AHL17 was required and sufficient for AHL17 to interact with HSP70-1 and HSP20-2 ( Figure S10).

F I G U R E 4
Interactions between AHL17 and HSP70-1. (a) LCI assays. Agrobacterium carrying the construct AHL17-nLUC and cLUC-HSP70-1 or cLUC-HSP70-5 were co-infiltrated into the leaves of N. benthamiana. The constructs of AHL19-nLUC and cLUC-HSP70-1, as well as the nLUC and cLUC, were also co-infiltrated into the same leaf as the control. LUC activity was detected 2 days after infiltration. (b) Co-IP assay. Total proteins were extracted from N. benthamiana leaves co-transformed with 35S::AHL17-GFP and 35S::HSP70-1-myc and immunoprecipitated by anti-GFP beads. The precipitated proteins were analyzed by western blot with anti-GFP and anti-myc antibodies. IP: immunoprecipitated. (c) BiFC assays. The construct cYFP-AHL17 was co-infiltrated with HSP70-1-nYFP or HSP70-5-nYFP into the leaves of N. benthamiana. The fluorescence signals of YFP were examined 2 days after infiltration. The co-infiltration of cYFP-AHL19 and HSP70-1-nYFP was used as the control.

| HSP70s form a complex with AHL17/28 and RHD6 in planta
Because the mutations of RHD6 could suppress the effects of overexpression of AHL17 on root hair production and gene transcription and because HSP70 proteins interact with AHL17, we wondered whether HSP70 proteins serve as the molecular links that enable AHL17 and RHD6 to form a complex in planta. To test this hypothesis, we first demonstrated the interactions between HSP70-1 and RHD6 using LIC assays in the leaves of N. benthamiana (Figure 5a). The interaction was further confirmed by BiFC assay (Figure 5b) and Co-IP assays ( Figure 5c) performed in the leaves of N. benthamiana.
expressed RHD6-GFP, HSP70-1-myc, and AHL17-FLAG in the leaves of N. benthamiana. First, anti-myc beads were used to immunoprecipitate HSP70-1-myc protein. Both RHD6-GFP and AHL17-FLAG protein could be co-precipitated as detected by anti-GFP and anti-FLAG antibodies, respectively (Figure 5d). This result reconfirmed the interactions between HSP70-1 and AHL17, as well as the interactions between HSP70-1 and RHD6. Second, we used anti-GFP beads to immune-precipitate RHD6-GFP protein. We found that the AHL17-FLAG proteins could be co-precipitated in the presence of HSP70-1-myc but that AHL17-FLAG could not be co-precipitated by RHD6-GFP in the absence of HSP70-1-myc (Figure 5e). These results demonstrated that HSP70-1 could form a complex with both AHL17 and RHD6 in planta.  (Figure 3a-d). However, what our work differed from previous studies is that AHL17 and AHL28 do not directly interact with RHD6 ( Figure S8). And, the overexpression of AHL17/ AHL28 neither affects the transcription of RHD6 nor the accumulation of RHD6 protein ( Figure S7). Therefore, there must be something missing between AHL27/AHL28 and RHD6 for their interactions to regulate gene transcription. Using the IP-MS technique and yeast two-hybrid experiments, we identified three HSP70 proteins as AHL17-interacting proteins (Table S6). We further demonstrated that they could directly interact with both AHL17/AHL28 and RHD6 to form a protein complex in planta (Figures 4 and 5). These HSP70 proteins seem to serve as molecular glues to bring AHL17 and RHD6 together to enhance the transcription of RHD6-downstream genes.

| DISCUSSION
The HSP70 family proteins are highly conserved throughout prokaryotes and eukaryotes (Boorstein et al., 1994;Gupta & Golding, 1993). They are well known for functioning as molecular chaperons that facilitate protein folding (Boston et al., 1996;Brkljacic et al., 2009;Clément et al., 2011;Noël et al., 2007;Rosenzweig et al., 2019). The direct participation of HSP70s in the regulation of gene transcription, however, has seldom been reported in literatures.
In one example with mammalian cells, HSP70 was found to regulate retinoid acid-induced gene transcription of the retinoid acid receptor β2 by working with the Mediator complex (Gao et al., 2015). In another example, HSP70 was shown to affect the conformation of tumor suppressor p53 to regulate the transcription of the genes downstream of p53 (Dahiya et al., 2019). Before the current study, there was only one report with plants showing that HSP70 could mediate the activity of transcription factors (Tiwari et al., 2020).
Mediator is a protein complex that works with RNA polymerase II, general transcription factors (TFIIs), and certain gene-specific transcription factors to form the pre-initiation complex that initiates transcription (Chadick & Asturias, 2005;Dotson et al., 2000). Arabidopsis Mediator complex consists of 34 subunits, and subunits 34, 35, 36, and 37 are plant-specific (Backstrom et al., 2007;Mathur et al., 2011;Yang et al., 2016). The different subunits of the Mediator complex are associated with specific biological processes (Yang et al., 2016). In Arabidopsis, researchers have used bioinformatics approaches to predict that HSP70-1, -2, and -4 are the subunits MED37c, 37d, and 37e, respectively, of the Mediator complex (Mathur et al., 2011). If these three HSP70 proteins are indeed the subunits of the Mediator complex, then it is not surprising that they can participate in regulating gene transcription, although their exact working mechanism is not yet clear. We speculate that under stress conditions, HSP70s help maintain the conformation of AHL17/28 and RHD6 so that they can function properly in regulating gene transcription. To further confirm the role of HSP70s in AHL-mediated gene transcription, it is necessary to test whether the mutation of these three HSP70 genes can block the effects of AHL17/28 overexpression in altering gene transcription and in promoting root hair production.
These works are underway.
AHL proteins have been shown to regulate a variety of processes in plant growth and development. The functions of several AHL proteins were revealed by a gene overexpression strategy, including AHL27 and AHL29 in hypocotyl growth (Street et al., 2008), AHL27 in leaf senescence (Lim et al., 2007), and AHL22 in flowering regulation (Xiao et al., 2009). That the overexpression of AHL17 or AHL28 enhanced root hair production might suggest a new function of AHL proteins in plant development (Figure 1). The single and double mutants of AHL17 and AHL28, however, did not exhibit obvious defects in root hair development ( Figure S3). One possibility for not seeing defects in root hair development in the ahl17/ahl28 mutants is that the basal level expression of these two genes may be very low; therefore, they might not play a significant role in root hair development under normal growth conditions. However, when plants are exposed to stress conditions, such as nutrition deficiency (González-Fontes et al., 2016;Schmidt & Schikora, 2001), mechanical stress (Okamoto et al., 2008), or bacterial infection (Galland et al., 2012), the level of ethylene is increased, which may induce the expression of AHL17 and AHL28, therefore enhancing root hair production. The molecular mechanism of how ethylene regulates root hair development has also been extensively studied (Qiu et al., 2021;Song et al., 2016). In fact, we found that both transcription and protein accumulation of AHL17 are induced by ethylene ( Figure S5). Like that under normal growth condition, the ahl17/ahl28 mutants also did not display obvious defects in root hair production in response to ethylene treatment ( Figure S3). Another explanation for this is the genetic redundancy that exists among the members of the AHL family. Genetic redundancy among the AHL proteins has been reported for AHL-mediated hypocotyl growth (Zhao et al., 2013), flower initiation (Xiao et al., 2009), axillary meristem maturation (Karami et al., 2020), and embryo development (Karami et al., 2021). Consistent with this possibility, our IP-MS analysis indicated that six other AHL proteins (all in the same clade of AHL protein family), including AHL28, could be co-immunoprecipitated with AHL17 (Table S6), suggesting that some of these AHL proteins may co-exist in the same protein complex and may perform similar functions. We used LCI assays to test whether AHL17 and AHL28 could interact with AHL19, which was co-precipitated with AHL17 in IP assay. Indeed, the results showed that AHL17/AHL28 could interact with AHL19 ( Figure S11).
Currently, we are generating the knockout mutants with multiple AHL genes to test whether the genetic redundancy indeed exists among the members of the AHL family in regulating root hair formation.
Because of the lack of root hair phenotype of ahl17ahl28 double mutants, we are still not certain whether the AHL17 and AHL28 overexpression-enhanced root hair production is an artifact or not.
However, in this work, it is more important that we revealed a previously unrecognized mechanism of AHL proteins-mediated gene transcription rather than determining whether AHL17 and AHL28 are genuine regulators of root hair development or not. It would not be possible to reveal such a mechanism if we did not use the overexpression strategy. The overexpression strategy has been demonstrated to be very useful in revealing a new biological process or identifying new components involved in a regulatory pathway. A great example is that overexpression of a chalcone synthase gene in petunia flowers resulted in a chimera color pattern of petals, instead of a uniformly enhanced coloration of petals as usually expected (Metzlaff et al., 1997). This phenomenon led to the discovery of RNA interference in higher plants. In our previous research, we found that overexpression of a sucrose transporter gene SUC2 in Arabidopsis using the CaMV 35S promoter inhibited primary root growth and enhanced plant responses to phosphate starvation, supporting that sucrose signaling is involved in these processes (Lei et al., 2011). To search for signaling components downstream of sucrose, Lei et al. (2014) screened for the genetic suppressor of 35S:SUC plants with restored primary root growth. Unexpectedly, they found that the restored primary root growth in all suppressors was due to the silencing of the 35S promoter activity rather than the suppression of sucrose signaling. Therefore, by serendipity, the use of these overexpressing materials resulted in the discovery of several novel components involved in gene silencing pathway (Duan et al., 2017;Lang et al., 2015;Lei et al., 2014;Nie et al., 2019;Zhou et al., 2022). Similarly, with these examples, we would like to emphasize that although the overexpression strategy used in this work was not sufficient to determine whether AHL17 and AHL28 are intrinsic regulator of root hair development, it indeed increased our understanding of the function of AHL proteins in regulating gene transcription.
Another significance of this work is that it established an ideal experimental system for further studying the functions of AHL proteins. This is because as a readout for AHL17/28's function in vivo, root hair development is very easy to score. In our future plan, we will use this system to investigate how HSP70 proteins mediate the interactions between AHL17/28 and RHD6 to regulate gene transcription.
We will also test how different variants of AHL17/28 and RHD6 interact with HSP70 proteins and target DNA sequences to exert their functions, which will provide further insights into the mode of

| Root hair observation and measurement
Images of root hairs on the agar plates were captured with a stereomicroscope (Olympus SZ61) equipped with a digital camera (Olympus DP72) with 1.5Â or 4.5Â magnification. The lengths of root hairs located 5 to 6 mm from the root tip were measured at 1.5Â magnification with the assistance of Digimizer software. For root hair numbers, both mature and immature root hairs including bulges in the region 5 to 6 mm from the root tip were counted. At least 15 seedlings for each genotype were used for root hair measurements.
To quantify root hair cells and non-root hair cells in the root epidermis, two areas along the root of a 7-day-old seedling were selected randomly and were examined at 4.5Â magnification. For each region, five contiguous cells from an H file and five contiguous cells from a neighboring N file were examined, and the numbers of hair cells and non-hair cells were recorded.

| Histological analysis of root hair formation
The histological analysis of root hair formation was performed according to Hou et al. (2021). Basically, the roots of 5-day-old seedlings were sectioned with a Leica microtome (Leica EM UC7). The sections were stained with periodic acid-Schiff's reagent and observed using a differential interference contrast microscope (Nikon 80i).

| Vector construction and plant transformation
To generate plant overexpressing vectors of 35S::AHL17 and 35S:: AHL28, the genomic sequences of the corresponding genes were amplified from genomic DNA of WT Arabidopsis plants. The genomic sequence of each gene was cloned into the BamHI and SacI restriction sites between the CaMV 35S promoter and the NOS terminator of the plant expression vector pZH01.
The mutated alleles of AHL17 and AHL28 were also generated in the WT using a CRISPR/Cas9-based genome editing system developed by Xing et al. (2014). The targeted editing sites in the corresponding genes were determined using the online service To generate vectors RHD6::RHD6-GFP, the CDS of RHD6 was first inserted into pJG186 using SacI and ApaI restriction enzymes, which resulted in the 35S::RHD6-GFP construct. The CaMV 35S promoter on this construct was then replaced by the promoter regions of RHD6. All of the primers used for the construction of the vectors are listed in Table S7.
All constructs were mobilized into the Agrobacterium tumefaciens strain GV3101 using the freeze-thaw method and were transformed into Arabidopsis plants via the Agrobacterium-mediated flower dip method (Clough & Bent, 1998). The stable transgenic lines were selected using antibiotic-containing media.

| Quantitative real-time PCR analysis
qPCR analyses for gene expression were performed as described by Song et al. (2016). The primers used for qPCR analyses are listed in Table S7.

| RNA-Seq analyses
Total RNAs were extracted from the roots of 7-day-old seedlings using the RNeasy Plant Mini kit (Qiagen). The RNA-seq analyses were performed at Bionova Company. The RNA-seq libraries were constructed through adaptor ligation and were subjected to single-ended sequencing with a 50-nucleotide reading length. FastQC software was used to assess the quality of raw sequencing reads. The adaptor and the low-quality reads were removed before data analysis. The remaining reads were aligned to the Arabidopsis TAIR 10.0 reference genome using TopHat2. After the sequences of rRNA or tRNA were removed, the TopHat read alignments were assembled by Cufflinks software to produce a transcriptome annotation of the genome. The expression levels for all transcripts were normalized to per million mapped reads. The differentially expressed genes were identified using DESeq2 (Love et al., 2014). The cutoff value for differentially expressed transcripts was a ≥2-fold change in expression with an FDR ≤ 0.05.

| Analysis of GUS activity
The histochemical analyses of GUS activity were performed as described by Jefferson (1989).

| Confocal microscopy
The fluorescence signals produced from the roots of 7-day-old seedlings of the transgenic plants carrying a GFP gene or from BiFC assays (signals from YFP) were observed with a confocal microscope (Zeiss LSM710META or Leica STED). The wavelengths of excitation/ emission were 488 nm/491-535 nm for GFP and 514 nm/520-550 nm for YFP proteins. The captured fluorescence images were processed and analyzed with Zen Black, Zen Blue software, or LAS X.

| EMSAs
The full-length CDSs of AHL28 were cloned into vector pMAL-c5x (NEB) using NotI and SalI restriction sites at the N terminus of an MBP tag to generate MBP-AHL28 recombinant protein. For better purification, the full-length CDSs of AHL17 were cloned into vector pMAL-c5x (NEB) with a 6x His tag sequence in its C terminus, thus generating MBP-AHL17-His. These constructs were separately transformed into E. coli strain Rosseta. The MBP-AHL28 was purified using Amylose Resin gravity flow columns (NEB), whereas the MBP-AHL17-His was purified using Ni-NTA agarose columns (QIAGEN).
The FAM-labeled hot probes containing the putative AT-hook binding motifs and its cold probes were generated by annealing the FAM-labeled complementary oligonucleotides. The sequences of the oligonucleotides used for generating various probes are listed in Table S7. EMSAs were performed as described in Sun et al. (2016).

| ChIP-qPCR assay
ChIP-qPCR assays were performed essentially as described by Saleh et al. (2008). Briefly, the chromatins were isolated from the 14-dayold transgenic plants overexpressing GFP-AHL17. The isolated chromatins were cross-linked with 1% formaldehyde, sonicated, and precipitated by Anti-GFP Magarose Beads (Smart Lifesciences). The precipitated DNA fragments were released by 200-mM NaCl and subjected to qPCR analysis.

| Co-immunoprecipitation assays
For testing the interaction between AHL17 and HSP70-1/2 as well as the interaction between RHD6 and HSP70-1/2, the full-length CDSs of HSP70-1 and HSP70-2 were cloned into the vector pMYC2 (modified form pROK2) through Xbal and BamHI restriction sites to generate 35S::HSP70-1-myc and 35S::HSP70-2-myc constructs. Agrobacterium strains carrying 35S::AHL17-GFP and 35S:: HSP70-1-myc or 35S::HSP70-2-myc, as well as 35S::RHD6-GFP and 35S::HSP70-1-myc or 35S::HSP70-2-myc, were co-infiltrated into the leaves of N. benthamiana. The infiltrated leaves with empty GFP construct were used as a negative control. Two days after infiltration, the infiltrated leaves were harvested, ground in liquid nitrogen, and suspended in an ice-cold protein extraction buffer to extract the total proteins. Anti-GFP Magarose Beads (Smart Lifesciences) were then added to the extracts, which were incubated at 4 C for 1 h with gentle shaking. The precipitated proteins were washed at least six times using ice-cold protein extraction buffer, and bound proteins were eluted by heating the beads in SDS protein loading buffer at 95 C for 10 min. The precipitated AHL17 and RHD6 proteins were detected by Western blotting using anti-GFP antibody (Abmart), whereas the HSP70-1 and HSP70-2 proteins were detected by Western blotting using anti-myc antibody (Abmart).