Regulation of xylem fiber differentiation by gibberellins through DELLA-KNAT1 interaction

How the environment regulates the development of multicellular organisms is a fundamental issue in biology about which little is known. Gibberellins (GAs) are plant hormones that integrate environmental information and translate it into developmental outputs (i.e. by promoting developmental transitions or initiating specific differentiation programs). For example, in many plant species, GA levels increase in seeds when exposed to appropriate light, temperature and humidity conditions, and such an increase is a trigger for mobilization of nutrient resources and growth of the previously dormant embryo (Shu et al., 2016). Similarly, GAs are necessary to establish the morphogenesis of trichomes (specialized epidermal cells with attributed functions in pathogen resistance) (Pattanaik et al., 2014).

From a mechanistic perspective, GA signaling is initiated by binding of GAs to the GA receptor, which then recognizes and promotes the degradation of DELLA proteins assisted by F-box proteins (Daviere and Achard, 2013; Hirano et al., 2008; Schwechheimer, 2011). In Arabidopsis, it has been shown that DELLA proteins act as transcriptional regulators by directly interacting with specific transcription factors that, owing to such interaction, alter their function (Locascio et al., 2013b; Marin-de la Rosa et al., 2014). Based on the identification of DELLA interactors (over 60 are known so far) and the characterization of the impact of particular interactions in development, we are beginning to understand the molecular mechanisms by which GAs regulate specific developmental processes. Clear examples of these are the control of the meristematic activity in the root via the interaction of DELLAs with B-type ARRs (Marin-de la Rosa et al., 2015); the transition to the reproductive phase through the DELLA-SPL transcription factors interaction (Yu et al., 2012); or the control of germination through the interaction of DELLA proteins with ABI3, ABI5, AtML1, and TCP14 and TCP15 (Lim et al., 2013; Resentini et al., 2015; Rombolá-Caldentey et al., 2014). However, GAs are also central to many other developmental processes in which the molecular mechanisms underlying their activity remain unknown. A remarkable example is the regulation of vascular development.

Plant vasculature originates during embryogenesis, but its development is not restricted to that stage. Indeed, depending on the environmental conditions or on the specific necessities that plants may encounter, vascular development can stop and resume multiple times during the plant life cycle. In adult plants, most of the new vascular cells are derived from the cambium, a specialized pool of undifferentiated meristematic cells that is programmed to develop exclusively the vascular tissues, namely xylem and phloem, that conduct water and solutes, and the assimilates, respectively. GAs have been shown to promote at least two aspects of vascular development: xylem expansion and the differentiation of a specific cell type within the xylem – the xylem fibers (Aloni, 2013; Eriksson et al., 2000; Mauriat and Moritz, 2009; Ragni et al., 2011). Here, we focus on the molecular mechanism underlying the activity of the GAs in xylem fiber differentiation. Recent discoveries indicate that KNOX-I genes [KNAT1/BREVIPEDICELLUS and SHOOTMERISTEMLESS (STM)] promote xylem fiber differentiation during vascular development (Liebsch et al., 2014) and that GA-dependent promotion of fiber differentiation, indeed, depends on the presence of active KNAT1 (Ikematsu et al., 2017). Importantly, we have identified KNAT1 as an interactor of the Arabidopsis DELLA protein GAI. In such a conceptual framework, we have tested the hypothesis that the function of GAs in the regulation of fiber development is regulated by the DELLA-KNAT1 physical interaction.

Using a previously described yeast two-hybrid (Y2H) screen with the GRAS domain of the DELLA protein GAI (‘M5-GAI’) as bait, we identified a number of putative DELLA interactors (Locascio et al., 2013a), among which KNAT1/BREVIPEDICELLUS(BP) was present. KNAT1/BP belongs to the KNOX family of transcription factors (which are general regulators of plant development) and has been shown to regulate the activity of the shoot apical meristem and the cambium (Byrne et al., 2002; Liebsch et al., 2014; Lincoln et al., 1994). To corroborate the observed DELLA-KNAT1 interaction and, at the same time, to map the interacting domains of GAI with KNAT1, we expanded the Y2H assay by including several truncated versions of the GAI clone as bait (Fig. 1A) and another DELLA protein, RGA (Fig. S1). Results showed that only the full-length and the M5 versions (but not the other truncated versions) of GAI, as well as the M5-like version of RGA, were able to interact with KNAT1 (Fig. 1A and Fig. S1). This result resembles the interactions with other transcription factors, i.e. with BZR1, PIF4 or JAZ1 (de Lucas et al., 2008; Gallego-Bartolomé et al., 2012; Hou et al., 2010), and indicates that the LR1 domain of the protein is necessary, but not sufficient, for the GAI-KNAT1 interaction. We also verified the interaction between GAI and KNAT1 in planta by co-immunoprecipitation studies in Nicotiana benthamiana leaves (Fig. 1B; Fig. S2).

DELLA interaction with transcription factors has been shown to either impair their ability to bind their target cis elements (de Lucas et al., 2008; Gallego-Bartolomé et al., 2012) or to promote target transactivation (Fukazawa et al., 2014; Marin-de la Rosa et al., 2015). To establish the possible molecular effect of DELLA on KNAT1, we examined the ability of KNAT1 to bind a sequence containing a previously identified KNAT1-bindng cis element (Fig. 2A; Mele et al., 2003), using electrophoretic mobility shift assays (EMSAs). As expected, bacterially produced KNAT1 was able to specifically bind this cis sequence (Fig. 2B) and, more importantly, the addition of increasing amounts of RGA competitively impaired the binding of KNAT1 to the corresponding probe (Fig. 2C). This result is in agreement with a model in which DELLAs impair the recognition by KNAT1 of its target promoters.

Given that KNAT1 plays a central role in the regulation of meristematic activity, understanding the biological meaning of the DELLA-KNAT1 interaction appears to be of general relevance for plant development. We hypothesized that KNAT1 would control different gene sets depending on the presence or absence of DELLA proteins. To test such a hypothesis and to gain more insights into the general biological significance of the DELLA-KNAT1 interaction, we performed comparative transcriptomic analyses. We first treated seedlings of the KNAT1 overexpressor transgenic line 35S::KNAT1 (Lincoln et al., 1994) (Fig. S3) and its wild type (No-0) with paclobutrazol (PAC), an inhibitor of GA biosynthesis that induces the accumulation of DELLAs (Silverstone et al., 1998), for 18 h. At that point, seedlings of each genotype were separated into two blocks: one that was treated with GA for 5 h (to induce DELLA degradation) and another one that remained in PAC (to maintain the accumulation of DELLA). In this way, we generated four different classes of samples: (1) wild type treated with PAC; (2) 35S::KNAT1 treated with PAC; (3) wild type treated with PAC plus GA; and (4) 35S::KNAT1 treated with PAC and GA. Through immunodetection of the DELLA protein RGA, we confirmed that PAC-treated samples of each genotype contained higher levels of DELLA proteins than GA-treated samples (Fig. S4). We then performed RNA-seq transcriptomic analyses with samples of all four classes and established comparisons between the transcriptomic profiles. We compared the transcriptome of wild type with that of 35S::KNAT1 treated with PAC and, in parallel, the transcriptomes of wild type with that of 35S::KNAT1 treated with GA (Fig. 3A; Table S1). The comparison between PAC-treated samples yielded genes that are KNAT1 targets in the presence of DELLAs, whereas the comparison between PAC+GA samples yielded genes that are KNAT1 targets in the absence of DELLAs (Fig. 3A). Using a statistical level of P<0.01, we identified 985 genes misregulated by KNAT1 only in the presence of DELLA and 776 genes misregulated only in the absence of DELLA. Out of those, 262 and 183 did it with a fold change higher than 2 (Fig. 3B). Our results show that KNAT1 has different targets depending on the presence of DELLAs, reinforcing the hypothesis that KNAT1 plays different biological roles depending on whether it interacts with DELLA proteins or not, and suggest, therefore, that the DELLA-KNAT1 interaction is relevant.

To investigate the particular KNAT1 functions that would be modulated by DELLA-KNAT1 interaction, we focused on the genes that were differentially affected by KNAT1 overexpression only in the absence or only in the presence of DELLAs. In both sets of genes, our Gene Ontology analysis showed a statistically significant enrichment of categories involved in cell wall metabolism, phenylpropanoid and lignin biosynthesis, and the response to hormones (Fig. S5). In fact, 63 and 78 of the genes induced and repressed by KNAT1, respectively, had been previously identified as ‘cell wall-associated genes’ by gene co-expression studies (Wang et al., 2012). To calibrate the involvement of KNAT1 and DELLA in this process, we then generated a co-expression network using those 141 genes as seed in the ATTED tool (Obayashi et al., 2007) and found that the KNAT1-DELLA interaction could potentially affect a total of almost 200 genes involved in secondary cell wall production, with 34 of them being differentially regulated by DELLA-dependent KNAT1 activity (Fig. 3C-E; Table S2). Thus, although KNAT1 was already known as a regulator of lignin biosynthesis (Mele et al., 2003), our results point out that some aspects of such regulation are DELLA dependent. Moreover, when we examined the expression of randomly selected SCW-related genes (Wang et al., 2012; Table S2), which had been tagged as ‘KNAT1 targets’ according to our RNA-seq experiment (Table S1), we found that six out of eight genes tested reproduced DELLA-dependence in mature hypocotyls undergoing secondary growth (Fig. S6), supporting the relevance of the interaction between DELLAs and KNAT1 for vascular development. In addition, this relevance was further supported by the colocalization of KNAT1-CFP and GFP-RGA in the nuclei of vascular cells of hypocotyls undergoing secondary growth (Fig. 1C). The signal was maximized when the plants were grown in the presence of 10 µM PAC.

As opposed to its role in the shoot apical meristem, where KNAT1 prevents early cell differentiation (Byrne et al., 2002), previous reports have shown that, during secondary growth, KNAT1 promotes xylem fiber differentiation (Liebsch et al., 2014; Mele et al., 2003). A recent report suggested a genetic link between KNAT1 and GA during xylem fiber differentiation, by which the developing xylem would gain the capacity to respond to GA in a KNAT1-dependent manner through a currently unknown molecular mechanism (Ikematsu et al., 2017). Having confirmed the DELLA-KNAT1 physical and functional interactions (Fig. 1; Fig. 3), we decided to test the relevance of this particular interaction in the control of xylem differentiation. We therefore treated a knat1 loss-of-function mutant (bp-11) and the 35S::KNAT1-overexpressor line (together with their respective controls, Col-0 and No-0) with GA₃ or PAC. In order to analyze the differential development of fibers across samples, hypocotyls were collected, sectioned and stained with phloroglucinol to detect lignin deposition. Similar to previous observations (Ikematsu et al., 2017), our GA treatments did not induce fiber formation in the bp-11 mutant, but we observed that they promoted fiber differentiation in all the other genotypes (Fig. 4). KNAT1 overexpression also promoted the formation of xylem fibers (Fig. 4), which was especially evident as the No-0 accession typically produces less fiber development than other accessions such as Col-0. More importantly, DELLA hyperaccumulation achieved with the PAC treatment completely abolished fiber production even in the 35S::KNAT1 line (Fig. 4D), and also reduced cambial activity (Fig. 4B). This result, together with the fact that KNAT1 expression levels are not affected by GA or PAC treatments (Fig. 5), support the proposed model of post-transcriptional modulation of KNAT1 activity through physical interaction with DELLAs.

Finally, to confirm that the observed xylem differentiation phenotype implied the alteration of fiber production, we carried out expression analyses of NST1, NST3 and SND2, master regulators of secondary cell wall production during xylem fiber development (Mitsuda et al., 2007, 2005; Zhong et al., 2006). As expected, the expression of the three genes was reduced in bp-11, and relatively increased in 35S::KNAT1 (Fig. 5), in agreement with the observed effects on actual fiber production (Fig. 4). Similarly, PAC-dependent DELLA accumulation prevented the induction of NST1 by KNAT1, and even caused an 85% decrease in NST3 expression both in wild-type and 35S::KNAT1 plants (Fig. 5). Consistent with KNAT1 acting downstream of DELLAs, altering GA levels did not significantly alter the expression of these genes in the bp-11 mutant. As NST1 and NST3 have been shown to upregulate a number of MYB transcription factors that regulate secondary cell wall developmental aspects during fiber development (including lignin biosynthesis) (Ohashi-Ito et al., 2010; Zhong et al., 2008), the effect of altering DELLA levels, i.e. by PAC or GA treatments, on NST1 and NST3 expression is in agreement with the role of KNAT1 in lignin biosynthesis and the enrichment of these genes among the DELLA-dependent KNAT1 targets set (Fig. 3; Fig. S5).

In conclusion, we propose that KNAT1-mediated xylem fiber development is negatively regulated by physical interaction with DELLA proteins. Given that DELLA protein levels have been shown to vary under different environmental conditions (Achard et al., 2006, 2007, 2008; Arana et al., 2011; Djakovic-Petrovic et al., 2007), it will be interesting to ascertain whether the mechanism proposed here mediates the regulation of specific aspects of cambial activity by the environment, and also whether this module regulates the development of other organs where KNAT1 and DELLAs are co-expressed, e.g. the shoot apical meristem (Hay and Tsiantis, 2010).

Arabidopsis thaliana accessions Col-0 and No-0 were used as wild type. The bp-11 mutant and the KNAT1ox (35S::KNAT1) line have been previously described (Lincoln et al., 1994; Venglat et al., 2002). The reporter lines RGA::GFP-RGA and KNAT1::KNAT1-2xeCFP in the Col-0 background have also been generated elsewhere (Silverstone et al., 2001; Rast-Somssich et al., 2015). Seeds were stratified in water for 3 days at 4°C, sown on pots containing soil mix (1:1:1 perlite, vermiculite and peat) and grown in growth chambers under long-day conditions (16 h of light and 8 h of darkness). For vascular phenotype analysis and RT-qPCR experiments, plants were watered with 50 μM GA₃ (Sigma), 10 μM PAC (Duchefa) or mock solution once a week.

For in vitro growth, seeds were surface sterilized and sown on half-strength MS (Duchefa) plates with 1% (w/v) sucrose, 8 g/l agar (pH 5.7). Seeds were stratified for 3-5 days at 4°C, and grown in growth chambers under continuous light (50-60 μmol m⁻² s⁻¹) at 22°C.

A pENTR vector carrying the coding sequence (CDS) of KNAT1 was obtained from SALK Institute, and transferred via LR clonase II (Invitrogen) into the pGADT7 (Clontech) yeast two-hybrid vector to create a GAL4-activation domain fusion. GAI deletions and the truncated version of RGA without the DELLA domain (RG52) (Gallego-Bartolomé et al., 2012) were fused to the GAL4-binding domain of pGBKT7 (Clontech) yeast two-hybrid vector. Direct interaction assays in yeast were carried out following the Clontech small-scale yeast transformation procedure. Yeast strain Y187 was transformed with GAL4-activation domain constructs, whereas yeast strain Y2HGold was transformed with GAL4-binding domain constructs. Diploid cells with both plasmids were obtained by mating and selected in SD/-Leu/-Trp/-His and with 3-aminotriazol (3-AT) (Sigma) to test interactions.

A pENTR vector carrying M5GAI has been previously described (Gallego-Bartolomé et al., 2012). For the co-immunoprecipitation assay in Nicotiana benthamiana, M5GAI and KNAT1 CDS were transferred via LR clonase II (Invitrogen) into pEarleyGate-203 and -201 to create the myc-M5GAI and HA-KNAT1 fusions, respectively. Each construct was introduced into Agrobacterium tumefaciens C58 cells that were used to infiltrate N. benthamiana leaves. Discs from infiltrated leaves were collected after 3 days, and proteins were extracted in a buffer containing 50 mM Tris-HCl (pH 7.5), 1× protease inhibitor cocktail (Roche), 0.1% Nonidet P-40 and 10% (v/v) glycerol. Total proteins were then incubated with anti-myc paramagnetic beads (Miltenyi) for 2 h at 4°C under slight rotation. The remaining steps were conducted following manufacturer′s instructions. myc-GAI detection was performed by using a 1:1000 dilution of anti-myc antibody (clone 9E10; Roche); HA-KNAT1 was detected by using a 1:5000 dilution of anti-HA antibody (clone 3F10; Roche). RGA immunodetection was performed using a 1:1000 dilution of polyclonal anti-RGA antibodies (Agrisera) that specifically recognize this Arabidopsis DELLA protein (Crocco et al., 2015).

For RNA-sequencing (RNA-seq) analysis, 7-day-old wild-type (No-0) and 35S::KNAT1 seedlings growing in ½ MS plates under continuous light as described above were transferred to a liquid growing medium supplemented with 10 μM PAC for 18 h. Seedlings were then incubated with 10 µM PAC+100 μM GA₃ or maintained in 10 µM PAC for 5 h. Three biological replicates were collected for RNA-seq.

Total RNA was extracted with RNeasy Plant Mini Kit (Qiagen) according to manufacturer's instructions, treated with the DNase Kit (Ambion), and then frozen at −80°C until analyzed. The RNA concentration and integrity (RIN) were measured in a RNA nanochip (Bioanalyzer, Agilent Technologies 2100). The preparation of the libraries and the sequencing were carried out by the Genomic Service of the University of Valencia (Spain). RNA-seq libraries were generated using the TruSeq Stranded mRNA Sample Preparation Low Sample (LS) Protocol (Illumina) and sequenced on a NextSeq 500 sequencer (Illumina) with a depth of 10 M. To estimate expression levels, the RNA-seq reads were pre-processed to eliminate adapters by using the package Trim.fastaq and then mapped to the Arabidopsis reference genome using TopHat (Trapnell et al., 2012). Transcript counts were calculated with HTSeq-count software (Anders et al., 2015). Differentially expressed genes were determined with DESeq2 (Love et al., 2014) and edgeR packages (Robinson et al., 2010) using as criteria fold change ≥2 and P<0.01.

For RT-qPCR, RNA from 28- or 35-day-old plants grown under long-day conditions as described above was extracted and treated with a DNase Kit (Ambion) to eliminate genomic DNA. Poly(dT) cDNA was prepared from 1.5 μg of total RNA with PrimeScript 1st strand cDNA Synthesis Kit (Takara Bio) and analyzed on 7500 Fast Real-Time PCR System (Applied Biosystems) with SYBR Premix Ex Taq II (Tli RNaseH Plus) ROX plus (Takara Bio) according to the manufacturer's instructions. All individual reactions were carried out in triplicate. Expression levels were normalized to those of ACT8. Primer sequences are shown in Table S3.

The 6xHis-KNAT1 recombinant protein in the pHGWA vector was expressed in the BL21 Rosetta 2 (DE3) pLysS (Novagen) E. coli strain with auto-inducible medium (ZYM5052) for 24 h at 25°C. It was then purified by binding onto a HisTrap HP column (GE Healthcare Life Sciences) and eluted with imidazole. The 6xHis-MBP-RGA recombinant protein was co-expressed with the chaperone Tig in BL21 cells carrying the pTf16 plasmid (Takara) and induced with 0.1 mM IPTG for 16 h at 12°C, purified by binding onto a MBP-Trap HP column (GE Healthcare Life Sciences) and eluted with maltose. The elution buffer was replaced by EMSA buffer [15 mM HEPES-KOH (pH 7.5); 40 mM KCl; 0.1 mM dithiothreitol; 10% glycerol] by filtration through a Sephadex-G25 HiTrap column (GE Healthcare Life Sciences). Oligonucleotide probes were labeled by filling the ends with the Klenow enzyme (Fermentas) in the presence of ³²P-dCTP. The EMSA reaction was performed with 1 ng of ³²P-labeled probe, 2 μg of poly(dI-dC) and 100 ng of KNAT1 alone or combined with RGA or MBP (1:5 to 1:20 ratio as indicated), and incubated at room temperature for 20 min. The binding reactions were analyzed by electrophoresis on 6% native acrylamide gel in 0.5×TBE buffer. The gels were then dried and autoradiographed at −80°C overnight.

Vascular phenotype of hypocotyls was analyzed with phloroglucinol staining. Briefly, hypocotyls were fixed in FAE solution (5% formaldehyde, 10% acetic acid, 50% ethanol) by vacuum infiltration for 5 min. Samples were then dehydrated through ethanol solutions up to 70% ethanol, embedded in paraffin wax using a Leica TP1020 tissue processor, sectioned using a Microm microtome and mounted on slides. Slides were placed in histoclear for 10 min for paraffin removal and then incubated 2×5 min in absolute ethanol. Samples were stained with a saturated 150 mM solution of phloroglucinol (Sigma-Aldrich) for 2 min and then soaked in 50% (v/v) HCl. Photographs were taken immediately with a Leica DM5000B microscope and a Leica DFC550 digital camera. Quantification of secondary growth was carried out as previously described (Liebsch et al., 2014) through the xylem II/total xylem ratio using ImageJ software.

Arabidopsis lines expressing GFP-RGA and KNAT1-CFP were crossed and plants homozygous for both reporters were used for analysis. Plants were grown for 4 weeks. In order to promote DELLA accumulation, watering was supplied with 10 µM paclobutrazol (Duchefa) once the plants had developed the first pair of true leaves. Hypocotyls were hand cut with a razor blade and fixed with 4% paraformaldehyde, cleared with ClearSee solution (Kurihara et al., 2015) and stained with Direct Red 23 (Pontamine Fast Scarlet 4B, Sigma) as described by Ursache et al. (2018) with minor modifications. Cleared and stained hypocotyl sections were then placed into a drop of ClearSee on 0.3 mm cavity slides for imaging with a Zeiss LSM 780 confocal microscope. eCFP and GFP/Direct Red 23 were sequentially visualized after excitation with 405 and 488 nm laser lines, respectively. Emission filters were set to 466-481 nm for eCFP, 503-517 nm for GFP and 594-613 nm for Direct Red 23. Emission spectra for eCFP and GFP were verified within individual nuclei with the ‘lambda scan’ mode of the microscope.

We are grateful to Javier Forment (IBMCP, Valencia, Spain) for his help with the RNA-seq analysis, and to David Esteve-Bruna (IBMCP, Valencia, Spain) for insightful comments on the manuscript and technical help. We also thank Miltos Tsiantis (MPZI, Cologne, Germany) for seeds of the Arabidopsis line expressing a CFP-tagged version of KNAT1.