Multi-level analysis of the interactions between REVOLUTA and MORE AXILLARY BRANCHES 2 in controlling plant development reveals parallel, independent and antagonistic functions

Plant development is highly plastic, and both leaf form and plant stature strongly depend on the environment a plant is exposed to. In suboptimal light conditions, e.g. in the canopy shade of other plants, Arabidopsis plants develop flattened leaves with extended petioles, show less branching and flower early. This suggests a tight connection between the pathways that control environmental responses and the gene networks that operate to control pattern formation in developing organs such as the leaves. In this context, the shoot apical meristem plays a decisive role as it maintains a pool of stem cells at the shoot tip. Cells that leave the stem cell niche at the lower side form stem tissue and cells at the flanks of the upper side produce new leaves. Factors controlling the establishment and maintenance of the shoot apical meristem have been identified decades ago. Members of the class III homeodomain leucine zipper (HD-ZIPIII) family of transcription factors play instrumental roles, starting with determining apical cell fate in the developing embryo (Smith and Long, 2010). Plants carrying loss-of-function mutations in three out of the five HD-ZIPIII genes, REVOLUTA (REV), PHAVOLUTA (PHV) and PHABULOSA (PHB), fail to establish a shoot meristem and develop pin-formed seedlings that arrest in development soon after germination (Emery et al., 2003; Prigge et al., 2005). HD-ZIPIII mRNA stability is controlled by the microRNA miR165/6 (Emery et al., 2003). Gain-of-function mutations in HD-ZIPIII genes that disrupt microRNA regulation cause respective mRNAs to accumulate to high levels (Emery et al., 2003; McConnell et al., 2001) and mutant plants display strong developmental phenotypes. In the case of PHB and PHV, respective gain-of-function mutants often develop radialized leaves and have a stunted appearance (McConnell et al., 2001). The rev10D gain-of-function mutant displays radialized vasculature and leaf-to-stem fusions that cause bending of the main stem (Emery et al., 2003). Another miR-disrupting gain-of-function mutation in the REV gene is amphivasal vascular bundle 1 (avb1), which causes similar growth defects as rev10D but additionally promotes the formation of ectopic axillary meristems (Zhong and Ye, 2004). Thus, avb1 mutant plants have a very bushy appearance due to the outgrowth of side shoots from the axillary meristems. Interestingly, both avb1 and rev10D carry the same mutation (P190 L) even though they have been isolated in different genetic screens. Thus, it appears that the differences in phenotype are either caused by variation in the growth conditions or by modifier mutations that segregate in the different genetic backgrounds. Loss-of-function mutations affecting REV (such as rev5) fail to initiate axillary meristems and have a barren appearance (Otsuga et al., 2001; Talbert et al., 1995), which agrees with REV being a positive regulator of shoot branching. Furthermore, REV directly activates the expression of SHOOTMERISTEMLESS to enable axillary meristem formation (Shi et al., 2016). Beside the negative regulation by microRNAs, HD-ZIPIII proteins are negatively controlled by LITTLE ZIPPER (ZPR)-type microProteins (Kim et al., 2008; Wenkel et al., 2007). MicroProteins are small single-domain proteins that are related in sequence to larger, multi-domain proteins and often act by establishing negative-feedback loops (Eguen et al., 2015; Staudt and Wenkel, 2011). ZPR1, ZPR3 and ZPR4 are direct and positive REV targets (Brandt et al., 2013). Thus, REV regulates its own protein activity by upregulating different ZPR microProteins.

Shoot branching is controlled by auxin in the course of establishing apical dominance of the main shoot. In this process, auxin is transported from the shoot to the root and represses the formation of side shoots (Booker et al., 2003). Strigolactones, plant hormones that are derivatives of the carotenoid pathway, have been identified as prominent branching suppressors by inhibiting outgrowth of lateral buds (Gomez-Roldan et al., 2008; Umehara et al., 2008). Recent work revealed that strigolactones and cytokinin function by affecting auxin transport (Shinohara et al., 2013; Waldie and Leyser, 2018) to control shoot branching. In this context, the MORE AXILLARY BRANCHES (MAX) genes encode components required for the production and perception of the strigolactone signal. Here, MAX1, MAX3 and MAX4 are biosynthetic enzymes that lead to the production of strigolactones (Booker et al., 2005; Stirnberg et al., 2002), while MAX2 is an F-Box protein that acts to initiate degradation of target substrates (Stirnberg et al., 2007). In Arabidopsis, it has been shown that MAX2 interacts with the strigolactone receptor D14 and upon hormone perception initiates the degradation of a SMXL6/7/8-type repressor complex (Wang et al., 2015).

MAX2 was identified as a potential direct REV target gene (Brandt et al., 2012). Loss of MAX2 function results in a pleiotropic phenotype and mutant plants have elongated hypocotyls, produce more side shoots and are delayed in leaf senescence. Here, we have validated that MAX2 can be upregulated by REV and that both REV and MAX2 have largely overlapping patterns of expression. To investigate the regulatory relationship between REV and MAX2 in the biological processes that they both seem to affect, we crossed available mutants to generate higher-order mutants. We assessed the phenotypic changes of respective single- and double-mutant plants that were grown with wild-type plants in different conditions. Our analysis revealed that shoot branching and hypocotyl elongation were subject to oppositional control by REV and MAX2. While REV promoted shoot branching and hypocotyl elongation, MAX2 acted as a suppressor in both processes. Hence, respective double mutants showed intermediate phenotypes. In the regulation of leaf senescence, however, both REV and MAX2 act as positive regulators and respective double mutants show partially additive phenotypes. During vascular development, REV acts as a polarizing factor that promotes the formation of xylem tissue (Zhong and Ye, 1999). Histological analysis revealed that the rev10D-induced radialization of vascular elements was suppressed in rev10D max2 double mutant plants. This indicated that proteins that antagonize REV function might have accumulated in max2 mutant plants. Taken together, this study uncovers context-dependent molecular mechanisms underlying REV functions in shoot and vascular architecture, photobiology, and senescence.

Expression of a dominant version of REV (REVd) fused to the rat glucocorticoid receptor under control of the 35S promoter resulted in strong developmental defects when transgenic plants were additionally treated with dexamethasone (DEX) (Brandt et al., 2012). Here, DEX induces the translocation of the chimeric GR-REVd protein into the nucleus where it can interact with chromatin to control gene expression. This system was used to perform ChIP-seq studies and identify direct REV target genes (Brandt et al., 2012). The further study of REV targets revealed novel roles for REV in the shade avoidance response (Bou-Torrent et al., 2012; Brandt et al., 2014, 2012; Merelo et al., 2017, 2016) and the control of leaf senescence (Xie et al., 2014). Among the potential direct target genes, we also identified MAX2. The max2 loss-of-function mutants are known to have an altered leaf morphology and increased branching (Stirnberg et al., 2002), delayed senescence (Woo et al., 2001) and a constitutive photomorphogenic response (Shen et al., 2007). Because several of these and related traits are also regulated by REV and the other HD-ZIPIII transcription factors in a redundant manner, we studied the relationship between REV and MAX2. With ChIP-seq we identified two potential REV binding sites in the MAX2 promoter (Fig. 1A) located −1.1 and −1.9 kb upstream of the transcription start site. To validate binding of REV to either one or both of these sites, we performed independent ChIP-qPCR experiments. For this, we treated one set of our transgenic plants expressing FLAG-GR-REVd with DEX and the other set with a mock solution. No binding to the MAX2 promoter was observed in the mock-treated samples. The analysis of the different binding sites revealed an interaction of REV with the more proximal region that was identified by ChIP-seq (Fig. 1B). Expression of MAX2 was then analyzed in different rev mutant plants (rev5, rev10D), of which rev5 plants are loss-of-function mutants and rev10D mutants are insensitive to miRNA regulation and resemble REV-overexpressing plants. Moreover, MIR165a-OX plants with reduced HD-ZIPIII mRNA (MIR165a-OX) and ZPR3-OX plants showing a reduced activity of HD-ZIPIII proteins, through an interaction with the overexpressed LITTLE ZIPPER3 protein, were also included (Fig. 1C). Analysis of young leaves revealed no changes in MAX2 expression levels in genetic backgrounds with lost or reduced REV or HD-ZIPIII function (Fig. 1C). In the rev10D gain-of-function background, we detected a strong increase of MAX2 mRNA, suggesting that HD-ZIPIIIs are not required to maintain the basal levels of MAX2 expression at early stages of development but can act as positive regulators of MAX2 expression. Analysis of older leaves revealed significantly reduced levels of MAX2 expression in genetic backgrounds with lost or reduced REV or HD-ZIPIII function and slightly increased levels in rev10D mutant plants. Furthermore, the analysis of MAX2 expression in different tissues showed a reduced expression of MAX2 in flowers of plants overexpressing ZPR3 and MIR165a (Fig. S1A,B). These findings support a role for REV as a positive regulator of MAX2 expression, especially at later stages of development.

We also assessed the spatial pattern of MAX2 expression in rev5 and rev10D mutant plants by in situ hybridization. We observed that MAX2 expression was slightly lower in rev5 mutant plants compared with the Col-0 wild type. In rev10d mutant plants, we observed signal intensities that were comparable with the Col-0 wild type (Fig. 1D). Analysis of the pMAX2::GUS reporter in wild-type and rev10D mutant plants yielded stronger GUS expression in the latter mutant background (Fig. 1E). Moreover, the GUS signal was, in both cases, restricted to the vasculature, indicating that REV is not active outside its expression domain. In summary, these findings support a role for REV as an upstream regulator of MAX2 in adult tissue that can be modulated in response to age or physiological status.

REV can upregulate MAX2 expression in a direct manner. We next compared the spatial patterns of expression of REV, LITTLE ZIPPER3 (ZPR3; a known direct REV target gene) and MAX2. To this end, we analyzed respective promoters fused to the β-glucuronidase (GUS) gene in stably transformed plants. We reasoned that if REV acts on MAX2, we should observe largely overlapping patterns of expression of all three genes. In flowers, strong GUS staining indicative of REV expression was detected at the base and top of the pistil and the stamens (Fig. 2A), whereas MAX2 was expressed only in stamens (Fig. 2I) and ZPR3 expression was not detectable (Fig. 2Q). In cotyledons, and young and old rosette leaves, REV and MAX2 expression was detectable in the vasculature, while ZPR3 expression was also vascular localized but very weak in the vasculature of cotyledons (Fig. 2B-D,J-L,R-T). At the shoot apex, strong GUS signal indicative of REV and MAX2 expression was observed in the shoot meristem region (Fig. 2E,M), while ZPR3 expression was restricted to the vascular strands that frame the meristem (Fig. 2U). In hypocotyls and roots, we observed largely overlapping patterns of expression of all three genes and here the GUS signals were mostly present in the vasculature (Fig. 2F-H,N-P,V-X). Taken together, we found that all genes are expressed at different levels and they exhibited partially overlapping patterns of expression, indicative of a regulatory relationship. This suggests that REV has the potential to act as a modulator of MAX2 expression.

MAX2 has a known function as a branching inhibitor in Arabidopsis; in max2 loss-of-function mutants axillary shoots grow out and plants have a bushy appearance. In contrast, REV plays a known role as activator of shoot branching and rev loss-of-function mutants develop fewer side shoots. To investigate the genetic relationship of REV and MAX2 with regard to their role in shoot branching, we generated sets of double mutants with max2 in combination with rev gain- and loss-of-function mutants. As observed before, we detected a higher number of side shoots in max2 mutant plants compared with the corresponding Col-0 wild type (Fig. 3A,B, Table 1). In line with the known functions of REV, we observed reduced shoot numbers in rev5 mutant plants, as well as in transgenic plants overexpressing either miR165a (MIR165a-OX) or the ZPR3 microProtein (ZPR3-OX), which leads to a reduction in HD-ZIPIII RNA levels or in HD-ZIPIII activity, respectively. In contrast to our expectation, we also observed reduced numbers of side shoots in the rev10D gain-of-function mutant (Fig. 3A,B). However, a closer inspection of this mutant revealed that the main stem appears fasciated and is likely a product of a fasciated meristem producing an unknown number of shoot fusions precluding detailed analysis. The combination of max2 with the different REV/HD-ZIPIII loss-of-function mutants revealed a strong suppression of the max2 phenotype and double mutant plants resembled the REV/HD-ZIPIII loss-of-function mutants (Fig. 3A,B, Table 1). The finding that the max2 phenotype was suppressed by rev loss-of-function mutations implies that the max2 mutant phenotype is dependent on the activity of REV. However, as the double mutants were not completely identical to the REV/HD-ZIPIII loss-of-function mutants, we conclude that REV and MAX2 play at least partially independent roles in the regulation of shoot branching.

Plants carrying loss-of-function mutations in MAX2 not only show branching defects but have also altered rosette and leaf morphologies. Compared with wild type, max2 mutants develop a more compact rosette with rounder leaves, which strongly resembles that in the rev10D mutants (Fig. S2A). Measuring the lengths of the petioles of wild type and the different single and double mutant plants revealed that all single mutants tend to have shorter petioles compared with the wild type. The rev5 and ZPR3-OX plants have petioles comparable with Col-0 wild-type plants, while MIR165a-OX plants have strongly reduced petiole lengths. The latter indicates that ectopic microRNA overexpression produces more-complex phenotypes compared with the loss of REV function or the ZPR3 overexpression. It is well established that miR165/6 interacts with both ARGONAUTE1 (AGO1) and the PINHEAD/ZWILLE/AGO10: here, AGO1 initiates miR-dependent HD-ZIPIII degradation (Kidner and Martienssen, 2004), while AGO10 sequesters miR165/6, leading to stabilization of HD-ZIPIII mRNAs (Zhu et al., 2011). The rev10D and max2 single mutants both have significantly reduced but similar petiole lengths (Fig. S2B). This contrasts with max2 rev10D double mutants, which have petiole lengths comparable with wild type, suggesting a complex interaction. In mutant plants with rounder leaves (such as rev10D and max2), quantification was easier compared with mutant plants that had narrower, downward curling leaf blades (such as rev5, ZPR3-OX and MIR165a-OX). In the latter mutants it was sometimes difficult to pinpoint the position at which the blade ended and the petiole started. The analysis of respective double-mutant plants revealed petiole length similar to the Col-0 wild type, except max2 rev5 that had slightly shorter petioles, which is again indicative of a complex genetic interaction. In addition to measuring petiole length, we also assessed the overall leaf morphologies by measuring both leaf lengths and leaf widths. As pointed out before, max2 mutant plants had rounder leaves with a length-to-width ratio of around 1.3, compared with Col-0 wild-type leaves that were more elongated with a length-to-width ratio of around 1.7 and rev10D with a ratio around 1.5 (Fig. S2C). Transgenic ZPR3-OX plants and rev5 mutant plants had downward curled narrow leaves with length-to-width ratios of around 2.1 and 2.4, respectively. Transgenic MIR165a-OX plants had a low length-to-width ratio (around 1.5) but were severely stunted. The analysis of respective double-mutant plants revealed wild-type-like leaf morphologies for max2 rev5 and max2 MIR165A-OX double mutants, with a length-to-width ration of around 1.6 (Fig. S2A,C), while the leaf morphologies of transgenic max2 ZPR3-OX plants resembled the respective transgenic lines with wild-type MAX2 alleles. Taken together, we can conclude that REV and MAX2 play opposing roles in regulating leaf morphology: REV acts as a positive regulator, promoting leaf width; MAX2 acts as a negative regulator, restricting leaf width. The combination of both mutants generates an intermediate phenotype and respective double mutants resemble the Col-0 wild type. This indicates that, under these circumstances, the direct regulation of MAX2 through REV appears to play a subordinate role and other factors appear to be more important.

Arabidopsis is a shade-sensitive plant that reacts to both qualitative and quantitative changes of light. When shaded by other plants, the ratio of the red to far-red light is reduced and plants show a set of growth responses that include hypocotyl elongation, petiole elongation, premature flowering and early senescence, commonly referred to as shade-avoidance syndrome (Roig-Villanova and Martinez-Garcia, 2016). To determine how wild-type and mutant plants respond to shade, we grew seedlings in white light growth chambers equipped with additional red and far-red LEDs. This setup allowed us to keep the photosynthetic active radiation (PAR) at a constant level but specifically alter the red to far-red light ratio. When grown in white light (WL), Col-0 wild-type plants showed very short hypocotyls compared with seedlings grown in shade (WL+FR), which developed very long hypocotyls (Fig. 4). In both WL and WL+FR, max2 mutant plants had very long hypocotyls and elongation in shade was strongly reduced, resulting in a growth ratio (hypocotyl length in WL+FR/hypocotyl length in WL) close to 1. Plants carrying mutations in REV (here rev5) or transgenic plants with reduced HD-ZIPIII mRNA or HD-ZIPIII activity, as in MIR165a-OX and ZPR3-OX, respectively, showed normal hypocotyl length when grown in WL but displayed much shorter hypocotyls when grown in WL+FR. This is consistent with previous findings (Brandt et al., 2012). In addition, rev5 mutants, and MIR165a-OX and ZPR3-OX transgenic plants had growth ratios smaller than wild-type plants, which is indicative of a disturbed shade response, supporting a positive role for HD-ZIPIIIs in the shade-avoidance response. In agreement, rev10D mutant plants appeared mildly hypersensitive to shade and developed slightly longer hypocotyls in both WL and WL+FR conditions, with an increased growth ratio (Fig. 4). The max2 rev5 double-mutant plants had slightly shorter hypocotyls in WL compared with max2 single mutants, indicating the constitutive growth in WL requires REV. When assessing the shade-avoidance response, it seems that respective double mutant regained the ability to respond to shade with a growth ratio comparable to that of rev5. The combination of max2 with either miR165a-OX or ZPR3-OX resulted in a complete shade insensitivity, with shorter hypocotyls in both WL and WL+FR conditions (relative to the max2 single mutant); max2 rev10D double mutants developed long hypocotyls in WL but were able to significantly elongate in WL+FR conditions, indicating that enhanced REV expression can partially rescue the shade-insensitive phenotype of max2 (Fig. 4). In summary, these results confirm a positive role for REV in promoting elongation growth in shade and a negative role for MAX2 as a suppressor of elongation growth.

Regulation of leaf senescence is another trait influenced by REV. Overlapping patterns of expression of MAX2 and REV were observed in older leaves and therefore we examined the involvement of MAX2 and HD-ZIPIII mutants during the onset and progression of senescence. We analyzed different senescence-related parameters to gain a comprehensive understanding. Leaves of one rosette were sorted according to their age and photographed, and leaves of the same position in the rosette were compared. One typical example of 8-week-old plants is presented in Fig. 5A, already indicating differences in senescence strength. In order to quantify these differences, leaves of the same position of at least six plants were harvested in a weekly rhythm from week 5 to week 9. Visible phenotypical changes in leaf color due to chlorophyll degradation were categorized using an automated colorimetric assay (ACA; Bresson et al., 2018) in which the leaf color is analyzed pixel-wise and the percentage of a specific color is calculated relative to the whole leaf area (Fig. 5B). Compared with wild type (Col-0), rev5 mutant plants showed a delay in senescence progression, while overall development was not retarded. Likewise, max2 mutant plants were delayed in senescence progression; however, the delay appears to be more pronounced at the beginning of senescence in 5- and 6-week-old plants, whereas at later stages rev5 plants appear to be significantly delayed, indicating that REV regulates additional senescence-associated genes. The rev5 max2 double mutant displays an additive effect: while it resembles more the max2 phenotype in 5- and 6-week-old plants, it is very similar to the rev5 phenotype in 8-week-old plants and more senescent than both single mutants in 9-week-old plants. These findings support a role for REV and MAX2 in the same physiological pathway in which they are both main players, but additional independent effects also exist. Plants with reduced HD-ZIPIII mRNA levels by microRNA165a overexpression (MIR165a-OX) and plants with reduced activity of HD-ZIPIII proteins by interaction with LITTLE ZIPPER3 proteins (ZPR3-OX) revealed that not only REV but also other HD-ZIPIII proteins are involved in senescence regulation, as their phenotype is even more severely delayed in senescence progression than rev5. This is especially visible in later stages of development. When the max2 mutation was introgressed into the genetic background of reduced HD-ZIPIII mRNA (MIR165a-OX), plants displayed a very similar phenotype to MIR165a-OX plants, suggesting that HD-ZIPIII proteins also have target genes besides MAX2. Interestingly, an additional effect was observed when HD ZIPIII protein activity, but not mRNA levels, was inhibited by the overexpression of ZPR3. When the max2 mutant was combined with reduced activity of HD-ZIPIII proteins (ZPR3-OX), this line showed an enhanced senescence phenotype in 5- and 6-week-old plants compared with max2 or ZPR3-overexpressing plants, but progression was almost completely arrested in the two following weeks so that in 8-week-old plants senescence was similarly delayed compared with max2 MIR165a-OX. As the MAX2 protein is absent due to mutation of the gene in the max2 ZPR3-OX plants, this could indicate that a potential MAX2 protein interaction partner might be liberated in early stages of development (5 and 6 weeks) and could inhibit ZPR3 action. As the max2 ZPR3-OX line showed first accelerated and then delayed senescence compared with rev5 max2 double mutants, the interplay between MAX2, HD ZIPIII and LITTLE ZIPPER proteins appears to be more complex and differs in early and late stages. In addition, the miRNA-insensitive rev10D mutant shows an antagonistic phenotype to the loss-of-function allele, but only at the beginning of senescence in 5- and 6-week-old plants. At later time points, rev10D mutant and max2 rev10D double mutant plants also exhibit delayed senescence, again pointing at a more complex regulatory network between MAX2, HD-ZIPIIIs, LITTLE ZIPPER proteins and miRNAs during senescence. Besides the phenotypic observations, the analyses of further physiological senescence-related parameters confirmed these results. The pulse amplitude modulation (PAM) method images the chlorophyll fluorescence and describes the activity of photosystem II. This activity declined earlier in the Col-0 leaves compared with all mutants. This decline was more pronounced in max2 than in rev5 plants, and the max2 rev5 double mutants showed an intermediate course (Fig. S3A), which is consistent with the phenotypic analyses. In addition, in all max2 mutants with other backgrounds, photosystem II activity also remained higher compared with the max2 in Col-0 background, which also matched the phenotypic analyses. In later stages of senescence, plasma membranes become more and more fragile and permeable, finally leading to the collapse of the cells. The membrane integrity was measured by ion leakage (Fig. S3B) and further supports the results of the colorimetric analysis. Col-0 displayed higher membrane deterioration than all other mutants and max2 plants were more similar to Col-0 than to rev5 plants and respective double mutants. Consistent with the accelerated senescence phenotype, max2 ZPR3-OX had a slightly higher membrane deterioration than wild-type plants in 6-week-old leaves. However, rev10D did not show the same effect, even though it also showed a very similar accelerated senescence phenotype to max2 ZPR3-OX, again indicating that the interplay appears not to be simple. The same held true for intracellular hydrogen peroxide contents. Hydrogen peroxide serves as a signal molecule for the initiation and progression of leaf senescence. Therefore, we measured the intracellular ROS contents by H₂DCFDA fluorescence (Fig. S3C). All mutant lines had a lower H₂O₂ content compared with wild type in later stages of senescence, which is consistent with the delayed progression of senescence of all tested plant lines. Here, max2 rev10D levels appear to be very similar to the wild type in 8-week-old plants, even though senescence was delayed in this line, suggesting that different parts of the senescence program are altered in the different mutants.

Finally, we analyzed the expression of leaf senescence marker genes by quantitative real-time PCR. Therefore, we chose three senescence-associated genes (SAGs) SAG12 and SAG13, encoding a cysteine protease and a short-chain alcohol dehydrogenase, respectively, as well as a NAC-domain transcription factor (ANAC092/ORE1). SAG12 is often used as a marker gene in late senescence and therefore is not detectable in week 6. In week 7, Col-0 plants reached the highest expression level, whereas low levels could be detected in max2, rev5 and ZPR3-OX. For the other mutant lines, SAG12 expression starts even later, suggesting a more delayed senescence process (Fig. S3D). The expression levels of SAG13 (Fig. S3E) were also highest in wild-type plants, but the differences in expression to the mutants were not very pronounced, which is consistent with the role of SAG13 in early senescence and the delay of the senescence progression in the mutants at later stages. For ANAC092 (Fig. S3F) a similar pattern to that of SAG13 was observed, but in this case the expression in all 8-week-old mutant plants was clearly diminished compared with wild type. Interestingly, ANAC092 expression was slightly higher in 6-week-old ZPR3-OX and rev10D plants, which corresponded to the accelerated phenotype of these two lines in this early stage.

Taken together, we can conclude that MAX2 and REV, as well as some other HD-ZIPIIIs, act in the same pathway to control the progression of senescence; however, MAX2 and REV also have independent effects. Overall, a complex regulatory network between HD-ZIPIII, miRNAs, LITTLE ZIPPER proteins and the different target genes appears to be at work.

To further explore the independent functions of REV in the regulation of shoot branching, we investigated potential direct REV target genes that operate at the nexus of growth regulation. We performed comparative analysis of available ChIP-seq, RNA-seq and microarray data (Brandt et al., 2012; Reinhart et al., 2013), and paid particular attention to genes with known roles in hormonal signaling pathways. Brassinosteroid hormones act as global growth regulators. Mutations in genes encoding enzymes that synthesize the brassinosteroid hormone or receptors that perceive the hormone show a severe dwarf phenotype. BAS1 encodes an enzyme that acts in the brassinosteroid catabolism and bas1-D gain-of-function mutants suppress hypocotyl growth (Neff et al., 1999). In organ boundaries, LATERAL ORGAN BOUNDARIES (LOB) suppresses brassinosteroid actions by inducing BAS1 (Bell et al., 2012). Previous ChIP-seq and microarray data show that REV can upregulate BAS1 expression. We performed gene expression analysis of BAS1 with transgenic 35S::FLAG-GR-REVd plants treated with either a mock solution or a dexamethasone-solution (DEX) that induces the translocation of the FLAG-GR-REVd protein from the cytoplasm to the nucleus. In addition, plants were pre-treated with cycloheximide, a protein biosynthesis inhibitor, allowing us to identify only direct REVd target genes. In support of the previous findings, we found that REVd induced BAS1 expression (Fig. S4).

The hormone abscisic acid (ABA) acts as a stress hormone and controls seed germination but has recently been shown to inhibit axillary bud outgrowth. ABA is induced by a transcriptional cascade in developing shoot buds to suppress branching under suboptimal light conditions (González-Grand; et al., 2017). Several of the potential REV target genes play a role in ABA-related signaling. The GTL1 transcription factors control drought tolerance (Yoo et al., 2010), HSL1 and PIF1 control seed germination (Oh et al., 2004; Zhou et al., 2013), and SnRK3.9 is a protein kinase active in polarized pollen tube growth (Steinhorst et al., 2015); however, several related kinases act in salt signaling (Barajas-Lopez et al., 2018). We found that GTL1, HSL1, PIF1 and SnRK3.9 were transcriptionally upregulated in response to REVd induction, supporting the idea that they are all bona fide target genes. We have previously shown that REV controls the expression of the class II HD-ZIP (HD-ZIPII) transcription factor genes HAT2, HAT3, HAT4 and ATHB4 (Brandt et al., 2012). Loss-of-function mutants in these HD-ZIPII genes result in strong leaf defects (Bou-Torrent et al., 2012) that are due to unregulated microRNA MIR165/6 expression (Merelo et al., 2016). Here, we tested whether HAT22, which encodes another homeodomain leucine zipper protein, is also a direct REV target gene and we found that, as with the other candidates, HAT22 expression was induced by REVd induction (Fig. S4). The latter finding is particularly important, as overexpression of HAT22 has been shown to accelerate senescence (Kollmer et al., 2011). These results confirm previous findings and support a role for REV as a transcriptional activator that can induce expression of key regulators of diverse growth-regulating pathways.

Both REV and MAX2 control shoot patterning and loss-of-function mutations show the opposite phenotypes. REV is a known vascular regulator and it has been reported that rev gain-of-function mutants (rev10D) develop radial amphivasal vascular bundles (Emery et al., 2003). This prompted the analysis of whether max2 mutant plants have altered vascular structures and whether there is a genetic relationship with regard to the vascular defects observed in hd-zipIII mutants. We sectioned the bases of the main stems of Col-0 wild type, and max2 and hd-zipIII mutants, and respective max2 hd-zipIII double mutants, and examined morphologies of the vascular bundles. In comparison with Col-0 wild-type plants, max2 mutant plants showed very similar vascular morphologies, except that the xylem parts contained an increased number of enlarged tracheary elements (Fig. 6A,B). Plants with loss-of-function mutations in REV (rev5) or transgenic plants that overexpress the LITTLE ZIPPER3 gene or MIR165a all showed flattened vascular bundles (Fig. 6C-E), whereas bundles of rev10D plants were almost entirely radialized with xylem surrounding the phloem (Fig. 6F). The combinations of max2 with rev5 (max2 rev5), or transgenic plants that overexpress the LITTLE ZIPPER3 (max2 ZPR3-OX) gene or MIR165a (max2 MIR165a-OX) resembled the respective hd-zipIII single mutants (Fig. 6G-I). The radial vascular bundles observed in rev10D mutant plants were, to a large degree, suppressed in max2 rev10D double mutant plants (Fig. 6J), indicating that either the REV10D-induced radialization requires a functional MAX2 protein or protein targets that are normally degraded by MAX2 accumulate and antagonize REV function. In agreement with this, we found that the known REV target genes ZPR3, HOMEOBOX-LEUCINE ZIPPER PROTEIN 3 (HAT3) and ARABIDOPSIS THALIANA HOMEOBOX-LEUCINE ZIPPER PROTEIN 4 (ATHB4) were slightly upregulated in the max2 mutant background, indicative of MAX2 or MAX2 substrates influencing REV/HD-ZIPIII activity (Fig. S5A). We also assessed the expression of REV and several REV target genes in wild-type and transgenic MAX2-OX plants. While REV expression was slightly lower in plants overexpressing MAX2, we found increased levels of ZPR3 and slightly lower levels of HAT3 and TAA1 (Fig. S5B). These findings indicate post-translational alterations in REV.

To successfully grow and reproduce in a dynamic environment, plants evolved efficient signal perception and transduction systems. Downstream of the signal perception, transcriptional regulators dissipate the signal and convert it into transcriptional output. In that regard, several transcription factors can act as regulatory hubs and integrate signals from multiple pathways. Here, we report the identification of two signaling hubs regulating each other and thereby influencing diverse developmental processes. We found that the REV transcription factor physically interacted with the promoter of MAX2, and rev10D gain-of-function mutants showed elevated levels of MAX2 expression. In agreement, the depletion of REV function or protein activity caused MAX2 expression to decline in older leaves (Fig. 1C). This indicates that the regulation of MAX2 is either stage specific or occurs in response to a specific signal. We have previously shown that the ability of REV to bind DNA in vitro is redox sensitive (Xie et al., 2014). It is therefore conceivable that the cellular redox-state of older leaves renders REV more active to regulate ‘late’ targets such as MAX2. Circumstantial support for this hypothesis is provided by a recent study on the role of the ZPR2 microProtein in regulating stem cell maintenance under hypoxia (Weits et al., 2019). The authors showed downregulation of many of the REV targets (including MAX2) upon ZPR2 induction specifically under hypoxic conditions. Hypoxia and leaf senescence both affect the cellular redox state but likely through different pathways. In summary, our results support the notion that MAX2 transcriptional regulation requires additional environmental inputs. Changes in redox state seem likely and it is imaginable that the PAS domain of REV acts as a redox sensor. Gain-of-function mutants in rev (here rev10D) had elevated levels of MAX2 mRNA, which indicates that REV acts as an inducer of MAX2 expression. A comparative study of the expression patterns of REV, MAX2 and the known REV-target ZPR3 revealed overlapping patterns of expression (Fig. 2), supporting the idea that MAX2 is a direct target gene.

REV is a positive regulator of axillary meristem initiation. Loss-of-function rev mutants have a reduced number of axillary buds compared with rev gain-of-function mutants (avb1) that develop an increased number of axillary buds. On the contrary, MAX2 is a negative regulator of axillary bud outgrowth and loss-of-function mutants have a bushy appearance (Fig. 3). The finding that the combination of rev and max2 mutants produced an intermediate phenotype implies that both factors partially operate in separate pathways. Side-shoot formation requires two steps: axillary bud formation and bud outgrowth. Here, REV promotes axillary bud formation (Otsuga et al., 2001) by promoting STM expression (Shi et al., 2016). In this study, we showed that REV also affected the subsequent bud outgrowth by regulating MAX2. Given that REV acts upstream of MAX2 but MAX2 antagonizes REV function, there might be a negative-feedback mechanism involved in side shoot production: REV induces side shoots but the concomitant upregulation of a negative factor (MAX2) prevents excessive side shoot production. The fact that loss of REV function does not result in overall reduced expression levels of MAX2 (Fig. 1) explains the barren appearance of rev5 mutant plants. The finding that hd-zipIII max2 double mutants (this includes rev5 max2, ZPR3-OX max2 and MIR165a-OX max2) showed an intermediate phenotype suggests that HD-ZIPIII activity is required for the max2 mutant phenotype and both factors control shoot branching in an interdependent manner.

Arabidopsis seedlings respond to shade by inducing elongation growth of the hypocotyl. Previous research has established a positive role for HD-ZIPIII transcription factors in promoting shade-induced growth (Baima et al., 2014; Brandt et al., 2012) and a negative role for MAX2 in repressing hypocotyl elongation (Shen et al., 2007). Similar to the control of shoot branching, we also found the combination of the hd-zipIII and max2 mutants resulted in double-mutant plants that exhibited intermediate growth responses (Fig. 4). Some combinations, such as max2 rev5, showed elongated hypocotyls in white light and shade, and a partially restored shade-avoidance response, while max2 ZPR3-OX and max2 MIR165a-OX plants produced generally shorter hypocotyls; however, these plants were almost shade insensitive. In summary, these results show that the long hypocotyl phenotype of max2 mutants is partially dependent on HD-ZIPIII activity, supporting a similar relationship to that previously observed in the control of shoot branching. Thus, HD-ZIPIIIs and MAX2 control shade-induced hypocotyl growth in an interdependent manner. MAX2 is a E3-ubiquitin ligase that potentially regulates many target substrates. It is possible that some of these target substrates directly or indirectly influence REV/HD-ZIPIII activity. The analysis of ChIP-seq and microarray data revealed that REV acts upstream of diverse growth-regulating pathways (Brandt et al., 2012; Merelo et al., 2013; Reinhart et al., 2013). Some of these regulations are positive and could promote context-dependent growth, whereas others would result in a context-dependent suppression of growth. A factor that, upon induction, would suppress growth is the brassinosteroid catabolism enzyme BAS1. The finding that REV acts as a direct and positive regulator of BAS1 expression (Fig. S4) supports a rheostat-like function of REV: promoting growth but at the same time limiting excessive growth.

The onset of senescence initiates the last phase in the life of an annual plant. During this phase, carbon, mineral and nitrogen resources are remobilized from the senescing leaves to the reproductive organs and developing seeds; thus, senescence is important for the reproductive success, but can also be used as an exit strategy under harsh stress conditions. Senescence is mainly driven by transcriptional changes affecting more than 8000 genes during onset and progression (Breeze et al., 2011). Therefore, transcription factors play a pivotal role in this process. REV was already discovered as one of the factors regulating leaf senescence in a redox-dependent manner and thereby established a relationship between early and late leaf development (Xie et al., 2014). To date, the regulatory function of REV in senescence has been explained by its positive regulation of WRKY53 expression, one of the central regulators of leaf senescence (Miao et al., 2004; Zentgraf et al., 2010). However, as the rev5 senescence phenotype was more severe than the wrky53 phenotype in this former study, it is now clear that REV targets additional SAGs (Xie et al., 2014). Here, we could show that one of these genes is MAX2. REV acts also upstream of MAX2 and positively affects MAX2 expression in some conditions. Accordingly, max2 mutant plants show delayed senescence. The rev5 max2 double mutant resembled max2 mutants during early senescence and rev5 mutants during late senescence, indicating that additive and independent effects exist and that they act in parallel. Surprisingly, rev10D, which contains a miRNA-insensitive version of REV and would therefore resemble a REV-overexpressing plant, shows only an accelerated senescence phenotype at early stages of senescence. This effect was negated in the absence of MAX2 (max2 rev10D), indicating that MAX2 is one of the most important downstream genes of REV in this phase. Overall, it was remarkable that the phenotype developed in two different phases: an early (5- and 6-week-old plants) and a late (8- and 9-week-old plants) phase. No significant difference was observed in 7-week-old plants in all lines compared with wild type, even though there is already a tendency towards a delay of senescence in almost all lines. Whereas in the late phase all lines showed significantly delayed senescence, in the early phase max2 ZPR3-OX and rev10D, in particular, showed significantly accelerated senescence, whereas max2, max2 rev5 and max2 rev10D exhibit already delayed senescence, indicating an important but diverse function of MAX2 in early senescence. Taken together, our results reveal that there is no simple interplay between HD-ZIPIIIs, miRNA165a, LITTLE ZIPPER3 and MAX2, as a biphasic (early and late) development of senescence exists and different parts of the senescence program are touched in the different mutants. Most likely more players and more feedback loops exist in this complex regulatory expression network. Moreover, MAX2 might also regulate senescence through strigolactone signaling. MAX2 encodes a F-Box protein, which can interact with the strigolactone receptor D14 and, upon hormone perception, leads to the degradation of a SMXL6/7/8-type repressor complex (Wang et al., 2015). Consistently, strigolactones can regulate leaf senescence in concert with ethylene in Arabidopsis (Ueda and Kusaba, 2015). Therefore, we can conclude that HD-ZIPIII proteins regulate senescence via WRKY53 and MAX2, and most likely other target genes, and that MAX2 is part of the complex multilevel network regulation of senescence influencing strigolacone signaling.

Growth responses, such as shade-induced elongation growth and shoot branching are controlled by REV and MAX2 in an interdependent fashion. In the senescence process, REV and MAX2 act synergistically, and with regard to vascular patterning it seems that MAX2 controls factors or processes that control REV activity (Fig. 6). This is supported by the finding that rev10D plants have strongly radialized vascular bundles compared with the bundles of max2 rev10D double mutant plants that resemble the wild type. To validate whether MAX2 controls unknown factors that influence REV protein activity, we tested the expression of three REV direct target genes in the max2 mutant background. This analysis revealed that the expression levels of ZPR3, HAT3 and ATHB4, three direct REV target genes (Brandt et al., 2012, 2013), were slightly upregulated in max2 compared with wild-type plants (Fig. S5). This finding suggests feedback regulation of REV by MAX2, which could be direct or indirect. In the regulation of senescence, where REV and MAX2 have parallel functions, the loss of MAX2 might result in the upregulation of REV target genes and thereby enhance REV function. In summary, our findings show that different growth processes are controlled by REV and MAX2, but their regulatory hierarchy is dependent on the respective process that is likely a product of altered cellular environments (Table 1). Our results demonstrate that in order to achieve a holistic understanding of genetic regulation, a multilevel analysis is valuable, even though the results can turn out to be complex and difficult to understand.

All experiments were performed in the Arabidopsis thaliana Col-0 ecotype. The following mutant lines were used in this study: max2-1 (Stirnberg et al., 2002), rev5 (A260V), rev10D (gain-of-function allele), ZPR3-OX (Wenkel et al., 2007) and MIR165a-OX (Kim et al., 2010).

For all experiments, seeds were stratified for 3 days at 4°C in the dark before being plated on growth medium (GM) containing 4.3 g/l Murashige and Skoog medium (Duchefa), 10 g/l saccharose, 0.5 g/l MES and 5.5 g/l plant agar (Duchefa) (pH 5.8). The MS medium-grown and the soil-grown Arabidopsis plants were incubated under 100 µmol m⁻² s⁻¹ in a 16 h light/8 h dark cycle.

For senescence-phenotyping, Arabidopsis thaliana plant lines were grown in a climate chamber under long day conditions (16 h light, 8 h dark) with moderate light intensity (80-100 µE/m2/s) at an ambient temperature of 20°C. Individual leaves were labeled with threads of different color according to their age to identify the leaf age while harvesting within the first rosette even in late stages of senescence (Bresson et al., 2018). Leaves of the same position were used for the same specific analysis. Plants were harvested in a weekly rhythm and at the same time of the day to avoid circadian effects. Bolting and flowering occurred within 5 and 6 weeks, respectively.

For the GUS construct, a DNA fragment containing a 3001 bp upstream promoter region of MAX2 was PCR amplified using DNAs from Col-0 as templates. The resulting PCR product was cloned into pENTR/SD-TOPO vector and subsequently cloned into the pBGWFS7 vector to generate the proMAX2-GUS construct. The final constructs were introduced into Col-0 wild-type plants by an Agrobacterium tumefaciens-mediated floral dip. Transformed seeds were initially screened for resistance to the herbicide Basta, followed by confirmation by RT-PCR. Homozygous lines were selected and used for phenotypic analysis.

In situ hybridization was carried out with 12-day-old seedlings. MAX2 probes were generated through amplifying of nucleotides 1 to 618 of the MAX2-coding sequence by PCR with the forward primer 5′-ATGGCTTCCACTACTCTCTC-3′ and reverse primer 5′-GCTTGATTTGTATCCCTCGG-3′.

For ChIP experiments and subsequent qPCR analysis, 35S::FLAG-GR-REVd transgenic plants were grown in liquid MS medium supplemented with vitamins for 15 days in continuous light and induced with 25 μM dexamethasone (DEX) for 60 min. Samples were harvested and ChIP was carried out as described previously (Brandt et al., 2012).

For hypocotyl measurements, Arabidopsis thaliana plants were grown for 2 days in constant white light conditions to induce germination and then kept for another 4 to 5 days either in the same conditions or transferred to simulated canopy shade conditions.

To investigate a difference in the progression of senescence between the plant lines, we performed a variety of methods that are described in detail by Bresson et al. (2018). Therefore, we harvested six plants (n=6) per plant line weekly, starting in the state before bolting at week 4 up to week 9, when leaves are heavily senescent. For the evaluation of leaf senescence phenotypes, rosette leaves were aligned according to their age with the help of color threads. Leaves of the first rosettes were aligned next to each other in order of their age. The automated colorimetric assay (ACA) defines leaf color pixel-wise and calculates the percentage of each category within the leaves, thereby providing information about chlorophyll degradation during senescence visible by a color change from green over yellow to brown. The activity of photosystem II (PSII) was assessed using a PAM chlorophyll fluorometer (Maxi version; version 2-46i, Heinz Walz). Therefore, Fv/Fm values were determined from leaf 5 and leaf 10. In addition, we measured the intracellular hydrogen peroxide level in leaf 8 using the fluorescent dye H₂DCFDA. Electrolyte leakage was determined using a conductivity meter (CM100-2, Reid & Associates).

QRT-PCR analysis was performed using senescence marker genes such as SAG12 (AT5G45890), which encodes a cysteine protease, SAG13 (At2g29350), which encodes a short-chain alcohol dehydrogenase, or ORE1/ANAC092 (AT5G39610), a member of the NAC transcription factor family that plays a role in senescence. Expression was normalized to ACTIN2, relative quantification to either GAPDH (AT1G13440) or ACTIN2 (At3g18780) was calculated using the ΔΔCT-method according to Pfaffl (Pfaffl, 2001). Total RNA was extracted using either the Spectrum Plant Total RNA Kit (Sigma) or the RNeasy Plant Mini Kit (Qiagen), followed by cDNA synthesis with RevertAid Reverse Transcriptase (Thermo Scientific). For the qPCR, KAPA SYBR Fast BioRad iCycler (KAPA Biosystems) master mix was used following the manufacturer's instruction.

The position of plants within flats was randomized to account for environmental variation. Total rosette branches at least 1 cm in length were counted for each plant at 6 weeks.

The 7th leaf of each plant was marked with indelible marker at ∼3 weeks post-germination. The maximum length and width of the leaf blade were measured, as well as the length of the petiole.

Tissue of mutants and wild type were vacuumed for 15 min, fixed for 4 h in Karnovsky's Fixative, and embedded in resin, according to Spurr's procedure (Spurr, 1969). Sections (2 μm) were made on a SuperNova Reichert-Jung microtome, then stained with Toluidine Blue-O (0.05%, pH 4.4) and visualized in bright field using a Nikon Eclipse 80i Fluorescence microscope.

The REV::GUS (rev-9), MAX2::GUS and ZPR3::GUS plants were incubated overnight in GUS staining solution [100 mM NaPO₄ (pH 7.0), 10 mM EDTA, 0.5 mM K₄Fe(CN)₆, 0.5 mM K₃Fe₆, 0.01% Triton-X and 1 mM X-Gluc]. Samples were infiltrated under vacuum for 10 min and then incubated at 37°C overnight. The staining buffer was removed and the samples were cleared in 70% ethanol. All observations by light microscopy were made with the Nikon Eclipse 80i Fluorescence microscope.

Sequence data used in this article can be found in the Arabidopsis Genome Initiative or ENA databases under the following accession numbers: MAX2 (AT2G42620), REV (AT5G60690), ZPR3 (AT3G52770), MIR165a (AT1G01183), SAG12 (AT5G45890), SAG13 (AT2G29350), ORE1/ANAC092 (AT5G39610), BAS1 (AT2G26710), GTL1 (AT1G33240), HAT22 (AT4G37790), HSL1 (AT1G28440), PIF1 (AT2G20180), SnRK3.9 (AT4G18700), HAT3 (AT3G60390) and ATHB4 (AT2G44910).

The peer review history is available online at https://dev.biologists.org/lookup/doi/10.1242/dev.183681.reviewer-comments.pdf