Brassinosteroid signals cooperate with katanin‐mediated microtubule severing to control stamen filament elongation

Abstract

Proper stamen filament elongation is essential for pollination and plant reproduction. Plant hormones are extensively involved in every stage of stamen development; however, the cellular mechanisms by which phytohormone signals couple with microtubule dynamics to control filament elongation remain unclear. Here, we screened a series of Arabidopsis thaliana mutants showing different microtubule defects and revealed that only those unable to sever microtubules, lue1 and ktn80.1234, displayed differential floral organ elongation with less elongated stamen filaments. Prompted by short stamen filaments and severe decrease in KTN1 and KTN80s expression in qui‐2 lacking five BZR1‐family transcription factors (BFTFs), we investigated the crosstalk between microtubule severing and brassinosteroid (BR) signaling. The BFTFs transcriptionally activate katanin‐encoding genes, and the microtubule‐severing frequency was severely reduced in qui‐2. Taken together, our findings reveal how BRs can regulate cytoskeletal dynamics to coordinate the proper development of reproductive organs.

BZR1‐family transcription factors induce cell elongation in stamen filaments in Arabidopsis thaliana via activation of katanin expression.

Introduction

Plant reproduction depends on the coordinated development of the pistils and stamens, with defects in either resulting in plant infertility (Regan & Moffatt, 1990; Sanders et al, 1999; Wellmer et al, 2014). Arabidopsis thaliana flowers contain two short outer stamens and four long inner stamens, the latter of which are the same length as the pistil to achieve successful self‐pollination. Late stamen development consists of filament elongation, pollen maturation, and anther dehiscence (Goldberg et al, 1993). In the late floral developmental stages, especially stages 12 and 13, the stamen filaments rapidly elongate and anthers dehisce to release pollen grains onto the stigma for pollination and subsequent fertilization (Smyth et al, 1990; Scott et al, 2004).

Coordinated stamen and pistil development is regulated by diverse external environmental factors and internal phytohormone cues (Richards et al, 2001; Tabata et al, 2010; Ye et al, 2010; Song et al, 2013; Qi et al, 2015; Saito et al, 2015). Brassinosteroids (BRs) play crucial roles in regulating multiple processes during plant development and stress responses (Mitchell et al, 1970; Nolan et al, 2020). Notably, recent studies revealed that key genes encoding putative BR‐degrading enzymes control heterostyly in primrose (Primula vulgaris) flowers (Huu et al, 2016; Li et al, 2016). Deficiencies in the biosynthesis or perception of the BRs have been shown to attenuate stamen filament elongation in A. thaliana, resulting in male infertility and abolishing seed production (Ye et al, 2010). Specifically, mutants associated with the BRs biosynthesis pathway, such as constitutive photomorphogenesis and dwarfism (cpd), br‐insensitive 2 (bin2), br‐insensitive 1–201 (bri1‐201), and dwarf4 (dwf4), produce shorter filaments than wild type and are unable to produce seeds (Szekeres et al, 1996; Bouquin et al, 2001; Li et al, 2001; Kim et al, 2005). The six core transcription factors in the BR signaling pathway, BES1, BZR1, BEH1, BEH2, BEH3, and BEH4, are functionally redundant. A quintuple mutant (qui‐2) displays male sterility because of defects in its stamen filament elongation and microsporocyte development (Chen et al, 2019b).

The plant cytoskeleton plays crucial functions in various cellular processes essential for cell morphogenesis and organogenesis (Chen et al, 2016; Ruan et al, 2018). In particular, cortical microtubules, which provide tracks for moving cellulose synthase complexes during cellulose biosynthesis, play a central role in plant cell morphogenesis (Lloyd & Chan, 2004; Paredez et al, 2006; Lindeboom et al, 2013). The loss of function of a series of genes encoding microtubule‐associated proteins, respectively, result in a dwarf phenotype or cause morphogenetic defects in specific cell types; for example, knockout mutants of either Kinesin4A/FRA1 or KTN1, the latter of which encodes a katanin p60 subunit, have a dwarf phenotype with fragile fibers (Bichet et al, 2001; Burk et al, 2001; Bouquin et al, 2003; Kong et al, 2015). CLASP is a microtubule‐related protein involved in cell division and cell expansion. The clasp‐1 was dwarfed, produced fewer cells in the root division zone, and had defects in cell expansion (Ambrose et al, 2007; Ruan et al, 2018). GCP4 is indispensable for the formation of the γ‐tubulin ring complex during microtubule nucleation in Arabidopsis. AmiR‐GCP4 plants exhibited dwarfism and produced abnormal microtubules in its guard cells, with a swelling of the pavement cells in the leaf epidermis (Kong et al, 2010).

Two previous studies have shed light on the molecular mechanisms that connect cellular signaling of plant growth to cytoskeletal organization: one uncovered that ROP6 Rho GTPase along with its effector RIC1 activate katanin‐mediated microtubule severing to promote the ordering of cortical microtubules in epidermal pavement cells (Lin et al, 2013); another revealed that blue light act upstream the microtubule‐severing protein katanin to control the reorientation of cortical arrays from transverse to longitudinal, thus inhibiting the rapid elongation of hypocotyl (Lindeboom et al, 2013). However, the mechanism by which microtubules are dynamically remodeled to regulate stamen filament elongation has not been reported. Moreover, it remains unknown if and how the phytohormones regulate microtubule remodeling to control stamen filament elongation.

Previously, we revealed that KTN80 has four paralogous genes that are functionally redundant, and the phenotype of the quadruple mutant ktn80.1234 is consistent with that of the KTN1 null allele, lue1. Importantly, we uncovered that KTN80 plays a precision guidance role to target the KTN1/KTN80 complex to microtubule‐severing site to achieve precise microtubule severing (Wang et al, 2017). In the present study, we discovered that filament elongation defects occur in lue1 and ktn80.1234 in A. thaliana, which are deficient in their microtubule‐severing ability. Intriguingly, we further revealed that the BZR1‐family transcription factors (BFTFs) transcriptionally activate the expression of KTN1 and KTN80s. Our findings provide evidence that the BR signaling pathway is associated with microtubule‐severing events to regulate proper filament elongation. We shed light on the cellular mechanism by which the coordination between BRs signaling and microtubule severing is precisely regulated to achieve successful reproduction.

Results

Filament elongation failure specifically occurs in mutants defective in microtubule severing

To explore the cellular biological mechanism underlying the microtubule dynamics involved in filament elongation, we carefully examined the stamen developmental phenotype of a collection of A. thaliana mutants which contain mutations in genes encoding microtubule‐associated proteins (MAPs) with distinct roles, such as microtubule nucleation, severing, and bundling. The representative genes included SPR3/GCP2 and GCP4, encoding the microtubule nucleators GCP2 and GCP4, respectively (Seltzer et al, 2007; Nakamura & Hashimoto, 2009; Kong et al, 2010); AUG6, encoding a component of the augmin complex (Miao et al, 2019); EB1a/1b/1c, encoding the microtubule plus‐end tracking proteins (Bisgrove et al, 2008); FRA1/Kinesin4A, encoding a kinesin motor involved in cell wall biosynthesis (Kong et al, 2015); CLASP, encoding a membrane‐tethering protein involving in microtubule plus‐end dynamics (Mimori‐Kiyosue et al, 2005; Ambrose et al, 2007; Bratman & Chang, 2008; Ruan et al, 2018); and KTN1 and KTN80s, encoding microtubule‐severing proteins (Bouquin et al, 2003; Wang et al, 2017). Remarkably, differential floral organ elongation was only found in lue1 and ktn80.1234 (Figs 1, and EV1A and B), which have less elongated filaments compared to the normally elongated pistil (Fig 1C–E). Among the other mutants, amiR‐GCP4, amiR‐AUG6, and clasp‐1 show severe developmental defects with a dwarf statue and small floral organs; however, the long stamens and the pistil were still identical in length (Figs 1A and B, and EV2A, B and G). fra1, spr2, and spr3 also have various developmental defects, but stamen filament elongation remains unaffected (Figs 1A and B, and EV2C, D and F). In addition, although the eb1 triple mutant exhibits spiral growth phenotype, the length of its stamens is similar to that of wild type (Fig EV2E). These results indicate that stamen filament elongation is specifically affected by the loss of katanin function.

Figure 1. Filament elongation is specifically reduced in microtubule‐severing defective mutants.

1. Phenotypes of wild type and nine microtubule‐associated gene mutants.
2. Comparison of the filament and stigma phenotypes of wild type and nine microtubule‐associated gene mutants. Scale bar, 0.5 mm.
3. Gynoecium lengths in lue1, ktn80.1234, and wild type in anthesis. N = 35 gynoeciums from nine wild‐type plants, n = 84 gynoeciums from 10 lue1 plants, n = 63 gynoeciums from 10 ktn80.1234 plants. Error bars indicate SD.
4. Stamen lengths in lue1, ktn80.1234, and wild type at anthesis. N = 70 stamens from nine wild‐type plants, n = 168 stamens from 10 lue1 plants, n = 126 stamens from 10 ktn80.1234 plants. Error bars indicate SD. Student's t test, **P < 0.01. Wild type versus lue1, **P < 0.0001; wild type versus ktn80.1234, **P < 0.0001.
5. Ratio of the stamen to stigma lengths in lue1, ktn80.1234, and wild type. N = 35 stamen‐stigma pairs from nine wild‐type plants, n = 84 stamen‐stigma pairs from 10 lue1 plants, n = 63 stamen‐stigma pairs from 10 ktn80.1234 plants. Error bars indicate SD. Student's t test, **P < 0.01. Wild type versus lue1, **P < 0.0001; wild type versus ktn80.1234, **P < 0.0001.

Figure EV1. Identification of flower morphology and pollen activity of microtubule‐severing defective mutants.

1. Comparison of flowers in wild type, ktn80.12 double mutant, ktn80.34 double mutant, ktn80.1234 quadruple mutant, and lue1. Scale bar, 0.5 mm.
2. Comparison of filaments and stigmas in wild type, ktn80.12 double mutant, ktn80.34 double mutant, ktn80.1234 quadruple mutant, and lue1. Scale bar, 0.5 mm.
3. Morphologies of pollen grains from wild type, ktn80.1234, and lue1. Scale bar, 20 μm.
4. Alexander's staining of anthers of wild type, ktn80.1234, and lue1. Scale bar, 200 μm.

Figure EV2. Statistics on the gynoecium length and stamen length from microtubule‐associated gene mutants.

1. Statistics on the gynoecium length and stamen length from amiR‐AUG6. N = 81 gynoeciums from nine wild‐type plants, n = 162 stamens from nine wild‐type plants; N = 32 gynoeciums from eight amiR‐AUG6 plants, n = 64 stamens from eight amiR‐AUG6 plants. Error bars indicate SD. Student's t test, **P < 0.01. Wild type versus amiR‐AUG6 in gynoecium length, **P = 0.004; wild type versus amiR‐AUG6 in stamen length, **P = 0.0023.
2. Statistics on the gynoecium length and stamen length from amiR‐GCP4. N = 81 gynoeciums from nine wild‐type plants, n = 162 stamens from nine wild‐type plants; N = 39 gynoeciums from 12 amiR‐GCP4 plants, n = 78 stamens from 12 amiR‐GCP4 plants. Error bars indicate SD. Student's t test, *P < 0.05. Wild type versus amiR‐GCP4 in gynoecium length, *P = 0.0141.
3. Statistics on the gynoecium length and stamen length from fra1. N = 81 gynoeciums from nine wild‐type plants, n = 162 stamens from nine wild‐type plants; N = 79 gynoeciums from 10 fra1 plants, n = 158 stamens from 10 fra1 plants. Error bars indicate SD.
4. Statistics on the gynoecium length and stamen length from spr2. N = 81 gynoeciums from nine wild‐type plants, n = 162 stamens from nine wild‐type plants; N = 46 gynoeciums from nine spr2 plants, n = 92 stamens from nine spr2 plants. Error bars indicate SD.
5. Statistics on the gynoecium length and stamen length from eb1 triple. N = 81 gynoeciums from nine wild‐type plants, n = 162 stamens from nine wild‐type plants; N = 54 gynoeciums from 12 eb1 triple plants, n = 108 stamens from 12 eb1 triple plants. Error bars indicate SD.
6. Statistics on the gynoecium length and stamen length from spr3. N = 81 gynoeciums from nine wild‐type plants, n = 162 stamens from nine wild‐type plants; N = 118 gynoeciums from 11 spr3 plants, n = 236 stamens from 11 spr3 plants. Error bars indicate SD.
7. Statistics on the gynoecium length and stamen length from clasp‐1. N = 81 gynoeciums from nine wild‐type plants, n = 162 stamens from nine wild‐type plants; N = 39 gynoeciums from six clasp‐1 plants, n = 78 stamens from six clasp‐1 plants. Error bars indicate SD. Student's t test, **P < 0.01. Wild type versus clasp‐1 in gynoecium length, **P < 0.0001; wild type versus clasp‐1 in stamen length, **P < 0.0001.

The lue1 and ktn80.1234 produced fewer seeds than wild type (Bouquin et al, 2003; Wang et al, 2017). To explore the cause of low seed setting in the microtubule‐severing defective mutants, we conducted a cryo‐scanning electron microscopy (Cryo‐SEM) experiment, which revealed that the morphology of mature pollen in lue1 and ktn80.1234 mutants does not show difference compare to the wild type (Fig EV1C), as was further confirmed by the Alexander dyeing experiment (Fig EV1D). To further verify pollen viability of the microtubule‐severing defective mutants, we performed manual pollination. First, wild type, lue1, and ktn80.1234 pollens were pollinated on wild type stigma, respectively, and F1 progenies did not show any significant difference in silique length, seed setting, and seed number among the three cross combinations (Appendix Fig S1A, B, G, and H). Similarly, when wild type, lue1 and ktn80.1234 pollens were pollinated on the lue1 and ktn80.1234 stigma, respectively, F1 progenies also did not see any significant difference in silique length, seed setting, and seed number whatever the pollen donor is wild type or mutants (Appendix Fig S1C–H). It is worth noting that even if pollen function does not seem affected in the lue1 and ktn80.1234 mutants, both silique size and seed number are reduced compared to the wild type (Appendix Fig S1G and H), indicating that katanin is also required for normal fertilization or fruit development. Taken together, these results demonstrated that the aberrant filament elongation specifically occurs in mutant which is defective in microtubule‐severing, and that these mutants produce fewer seeds due to the unsuccessful pollination caused by their short filaments.

Compromised cell elongation leads to short stamen filaments in mutants defective in microtubule‐severing

The capacity for maintaining normal cell division and differentiation is a critical determinant of organ morphological characteristics and performance in plants (Rasmussen et al, 2013; Abera et al, 2014). To ascertain the appearance of unique short filaments in the microtubule‐severing defective mutants, we further used Cryo‐SEM to investigate the epidermal cell morphology in these plants. In lue1 and ktn80.1234, an obvious reduction in filament cell length was found compared to wild type (Fig EV3A and B). In contrast, both microtubule‐severing mutants showed a remarkable increase in filament cell width relative to wild type (Fig EV3A and C). Based on the above situation, these findings indicate that the microtubule‐severing defective mutants have a higher cell width/length ratio than wild type (Fig EV3D), which is the explanation for this phenomenon that the growth anisotropy of filament cells is impaired in mutants defective in microtubule severing.

Figure EV3. Microtubule‐severing defective mutants lack filament cell elongation capability.

1. Morphologies of stamen filament cells from wild type, lue1, and ktn80.1234. Scale bar, 50 μm.
2. Stamen filament cell length from wild type, lue1, and ktn80.1234. N = 61 cells from five wild type plants, n = 65 cells from five lue1 plants, n = 67 cells from five ktn80.1234 plants. Error bars indicate SD. Student's t test, **P < 0.01. Wild type versus lue1, **P < 0.0001; wild type versus ktn80.1234, **P < 0.0001.
3. Stamen filament cell width from wild type, lue1, and ktn80.1234. N = 61 cells from five wild type plants, n = 65 cells from five lue1 plants, n = 67 cells from five ktn80.1234 plants. Error bars indicate SD. Student's t test, **P < 0.01. Wild type versus lue1, **P < 0.0001; wild type versus ktn80.1234, **P < 0.0001.
4. Ratio of cell width/length from wild type, lue1, and ktn80.1234. N = 61 cells from five wild type plants, n = 65 cells from five lue1 plants, n = 67 cells from five ktn80.1234 plants. Error bars indicate SD. Student's t test, **P < 0.01. Wild type versus lue1, **P < 0.0001; wild type versus ktn80.1234, **P < 0.0001.
5. Morphologies of the filament transections from wild type, lue1, and ktn80.1234. Scale bars, 40 μm.
6. Transection areas of wild type, lue1, and ktn80.1234 filaments. N = 34 filaments from five wild type plants, n = 33 filaments from 10 lue1 plants, n = 32 filaments from 10 ktn80.1234 plants. The box plots are shown with median (horizontal line), 25^th and 75^th percentiles (box edges) and interquartile range (whiskers). Student's t test, **P < 0.01. Wild type versus lue1, **P < 0.0001; wild type versus ktn80.1234, **P < 0.0001.
7. Number of outermost cells on wild type, lue1, and ktn80.1234 filaments. N = 34 filaments from five wild type plants, n = 33 filaments from 10 lue1 plants, n = 32 filaments from 10 ktn80.1234 plants. The box plots are shown with median (horizontal line), 25^th and 75^th percentiles (box edges) and interquartile range (whiskers).
8. Number of cells in the transects in wild type, lue1, and ktn80.1234 filaments. N = 34 filaments from five wild type plants, n = 33 filaments from 10 lue1 plants, n = 32 filaments from 10 ktn80.1234 plants. The box plots are shown with median (horizontal line), 25^th and 75^th percentiles (box edges) and interquartile range (whiskers).

To assess if a change in cell number might also contribution to stamen elongation defects, we used resin sections to observe transverse sections of the stamen filaments with the aim of exploring their morphology and quantifying the number of component cells. The cross‐sectional area of the filaments produced by the microtubule‐severing defective mutants was larger than wild type, with the highest increase being approximately 33% greater than the control (Fig EV3E and F). A further analysis of the filament cell number suggested that there were no significant differences in the numbers of outermost cells or the total cells between the mutants and wild type (Fig EV3G and H). These results implied that the shortening and thickening of the filaments are the result of the abnormal elongation of individual cells rather than the number of cells produced. This is consistent with the cell morphology of roots of mutants (Bichet et al, 2001), indicating that abnormal morphology of filament cells in lue1 and ktn80.1234 mutants is caused by defective cell elongation (growth anisotropy) rather than cell division.

Filament epidermal cells in ktn80.1234 and lue1 display a disordered microtubule network caused by abolished microtubule severing

To ascertain the cellular mechanisms regulating the formation of the short filaments, we examined the microtubule arrays and organizations in the filament cells of the microtubule‐severing defective mutants. We performed live‐cell imaging using transgenic line expressing mCherry‐TUB6 (encoding β‐tubulin 6), which labels the cortical microtubules. Wild type has well‐aligned microtubule arrays perpendicular to the direction of cell elongation (Fig 2A and C), while the microtubule‐severing defective mutants lue1 and ktn80.1234 produce a disordered microtubule network (Fig 2B, D, E and F). Parallel or oblique microtubule arrays were mainly observed in wild‐type filament cells, while mesh‐like microtubule arrays were mostly seen filament cells of microtubule‐severing defective mutants (Fig 2B). We next measured the density of microtubule arrays in wild type and microtubule‐severing defective mutants lue1 and ktn80.1234, and the results showed that the density obviously higher in microtubule‐severing defective mutants compared to that of wild type (Fig 2G). We further employed the order parameter S ₂ to quantify the degree of alignment of microtubule arrays. For a completely isotropic (disordered) system, S ₂ is close to 0, while for systems with completely parallel or anti‐parallel microtubules, S ₂ is close to 1 (Deinum et al, 2017). Compared with wild type, the S ₂ of microtubule arrays was lower in microtubule‐severing defective mutants lue1 and ktn80.1234 (Fig 2H). Notably, in wild‐type filaments, katanin complexes functions to precisely sever microtubules at crossovers and branching nucleation sites in most cases, while in ktn80.1234 filaments, the loss of KTN80s function results in disrupted KTN1 recruitment (Fig 3A and B). Thus, in ktn80.1234 filament cells, microtubule severing is diminished, which is consistent with the findings in the epidermal pavement cells (Wang et al, 2017). To further investigate this mechanism, time‐lapse imaging was employed. Over time, wild‐type filament cell microtubules become better aligned due to the effective microtubule severing performed by katanin complexes at crossovers and branching nucleation sites (Fig 3C and Movie EV1), while ktn80.1234 exhibited net‐like microtubule arrays that remain unchanged at the end of the time series (Fig 3D and Movie EV2). Collectively, these results suggest that the disordered microtubule networks exhibited by the mutants arise from abnormal microtubule severing, which affects filament morphogenesis and results in plant infertility.

Figure 2. The microtubule‐severing defective mutants display abnormal microtubule organization in the stamen filament cells.

• A
  Comparison of microtubule organization in the stamen filament cells of wild type, lue1, and ktn80.1234. Scale bar, 10 μm.
• B
  Microtubule morphology distribution in wild type and katanin mutants. Seven stamen filaments from different plants were analyzed in wild type, five stamen filaments from different plants were analyzed in lue1 and ktn80.1234, respectively. Error bars indicate SD. Statistically significant differences between groups were tested using one‐way ANOVA, **P < 0.01. Wild type versus lue1 in parallel microtubule arrays, **P < 0.0001; wild type versus ktn80.1234 in parallel microtubule arrays, **P < 0.0001; wild type versus lue1 in oblique microtubule arrays, **P < 0.0001; wild type versus ktn80.1234 in oblique microtubule arrays, **P < 0.0001; wild type versus lue1 in disordered microtubule arrays, **P < 0.0001; wild type versus ktn80.1234 in disordered microtubule arrays, **P < 0.0001.
• C
  Microtubule organization in wild‐type stamen filament cells. Scale bar, 10 μm.
• D–F
  Microtubule organization in the stamen filament cells of lue1 and ktn80.1234. Arrows indicate microtubule crossovers. Dotted lines indicate abnormal microtubules. Scale bar, 10 μm.
• G
  Microtubule density in wild type and katanin mutants. Six cells from different plants were analyzed in wild type, and five cells from different plants were analyzed in lue1 and ktn80.1234, respectively. Error bars indicate SD. Student's t test, **P < 0.01. Wild type versus lue1, **P = 0.0031; wild type versus ktn80.1234, **P = 0.0098.
• H
  S ₂ values in wild type and katanin mutants. Five cells from different plants were analyzed in wild type, and five cells from different plants were analyzed in lue1 and ktn80.1234, respectively. Error bars indicate SD. Student's t test, **P < 0.01. Wild type versus lue1, **P = 0.0033; wild type versus ktn80.1234, **P = 0.0006.

Figure 3. The ktn80.1234 quadruple mutant lacks proper microtubule severing and disrupts the recruitment of KTN1 in the stamen filament cells.

1. The recruitment of GFP‐labeled KTN1 (pseudo‐colored in red) to the microtubules (mCherry‐TUB6; pseudo‐colored in green) is completely disrupted in ktn80.1234 filament cells compared with wild type. Scale bar, 5 μm.
2. Severing frequency (severing events per 100 μm² per second) from all severing events. One hundred and sixty five severing events were detected in area of 1,087 μm² from three cells in wild type over 5 min. Fifteen severing events were detected in area of 1,187 μm² from three cells in ktn80.1234 over 5 min. Error bars indicate SD. Student's t test, **P < 0.01. Wild type versus ktn80.1234, **P < 0.0001.
3. Time‐lapse images showing that KTN1 is associated with microtubule severing at crossover sites in wild‐type filament cells. GFP‐labeled KTN1 fluorescent particles are pseudo‐colored in red, and microtubules (labeled with mCherry‐TUB6) are shown in green. Yellow arrows indicate KTN1 at microtubule crossover sites. Asterisks indicate the location at which a microtubule has been severed at a crossover site. Scale bar, 5 μm.
4. Time‐lapse images showing that KTN1 cannot be recruited to crossover sites, and no severing was observed in ktn80.1234 filament cells. Yellow open arrows indicate microtubule crossovers, which remain stable for a long time. Scale bar, 5 μm.

Qui‐2 has filament development defects and abnormal cell morphology

The quintuple qui‐2 mutant, which lacks five core transcription factors of the BR signaling pathway, also displays short stamen filaments (Fig 4A). To explore whether the BR signaling pathway is involved in regulating katanin‐mediated microtubule severing, we measured the lengths of the gynoecium and stamen in qui‐2. Importantly, we showed that unlike in the microtubule‐severing defective mutants, the filaments and the gynoecium of qui‐2 are both shorter than that in wild type (Fig 4B and C). The qui‐2 has a lower ratio of stamen to stigma length than wild type (Fig 4D). To investigate the cellular basis causing its filament developmental defects, we carried out Cryo‐SEM and semi‐thin stamen cross sections to observe the cellular morphology (Fig 4E and Appendix Fig S2A). Clear structural differences were identified in qui‐2 filament cells, which were much shorter and slightly wider than their counterparts in wild type (Fig 4F and G). This led to a huge increase in the cell width/length ratio and indicated that lengthwise growth was inhibited, causing cells to expand in a direction perpendicular to the filament itself (Fig 4H). The semi‐thin stamen cross sections revealed no differences in the transected area of the filaments or the number of cells produced by qui‐2 and wild type (Appendix Fig S2B). Taken together, these results indicate that the BFTFs are essential for normal cell morphology by sustaining the ability for filament cells to elongate, and that the loss of their function results in filament development defects, in accordance with the phenotype of the microtubule‐severing defective mutants.

Figure 4. The qui‐2 has short filaments and short filament cells.

1. Comparison of the filament and stigma phenotypes in wild type and qui‐2. Scale bars, 0.2 mm.
2. Gynoecium length in wild type and qui‐2 at anthesis. N = 81 gynoeciums from nine wild type plants, n = 42 gynoeciums from 15 qui‐2 plants. Error bars indicate SD. Student's t test, **P < 0.01. Wild type versus qui‐2, **P < 0.0001.
3. Stamen length in wild type and qui‐2 at anthesis. N = 162 stamens from nine wild‐type plants, n = 84 stamens from 15 qui‐2 plants. Error bars indicate SD. Student's t test, **P < 0.01. Wild type versus qui‐2, **P < 0.0001.
4. Ratio of the stamen to stigma length in wild type and qui‐2 at anthesis. N = 81 stamen‐stigma pairs from nine wild‐type plants, n = 42 stamen‐stigma pairs from 15 qui‐2 plants. Error bars indicate SD. Student's t test, **P < 0.01. Wild type versus qui‐2, **P < 0.0001.
5. Comparison of the morphologies of the stamen filament cells in wild type and qui‐2. Scale bars, 10 μm.
6. Stamen filament cell length in wild type and qui‐2. N = 61 cells from five wild type plants, n = 62 cells from 15 qui‐2 plants. Error bars indicate SD. Student's t test, **P < 0.01. Wild type versus qui‐2, **P < 0.0001.
7. Stamen filament cell width in wild type and qui‐2. N = 61 cells from five wild type plants, n = 62 cells from 15 qui‐2 plants. Error bars indicate SD. Student's t test, *P < 0.05. Wild type versus qui‐2, *P = 0.0156.
8. Ratio of filament cell width/length in wild type and qui‐2. N = 61 cells from five wild type plants, n = 62 cells from 15 qui‐2 plants. Error bars indicate SD. Student's t test, **P < 0.01. Wild type versus qui‐2, **P < 0.0001.

BFTFs transcriptionally activate the expression of KTN1 and  KTN80s

This similarity of qui‐2 phenotype to the short filaments of the ktn80.1234 and lue1 led us to hypothesize that the BR‐related transcription factors specifically regulate KTN80s and KTN1 expression to participate in filament development in A. thaliana. It was previously reported that the four KTN80s act redundantly to manipulate microtubule organization (Wang et al, 2017). To test whether the expression of these four KTN80s and KTN1 responded to BR hormonal signals, we treated the seedlings with BRs. The qRT‐PCR results showed that the expression levels of the four KTN80s and KTN1 were upregulated 1 h after the BRs treatment, with CPD and SAUR‐AC1 as the control of negative and positive, respectively (Sun et al, 2017; Appendix Fig S3A). Meanwhile, to further analyze whether and how BR signals control filament development, we detected KTN80s and KTN1 expression in the filaments of various mutants of the BR signal transduction pathway genes. Significantly, two BRs pathway mutants, bes1‐1 bzr1‐1 beh1‐1 beh3‐1 beh4‐1 (qui‐1) and qui‐2 (Chen et al, 2019b), which displayed a disrupted transcriptional regulation of the transduction of the BRs signaling, as well as the BR‐insensitive mutant bri1‐116 (Zhang et al, 2009), showed a decrease in KTN80s and KTN1 expression compared with wild type (Appendix Fig S3B and D). On the contrary, the gain‐of‐function point mutation of BZR1, bzr1‐1D, which caused increased BZR1 activity, showed a significant increase in the expression of these genes in the filament tissue (Appendix Fig S3C). Taken together, the results imply that the BR signals may make a substantial contribution to the regulation of KTN80s and KTN1 expression.

In order to clarify the roles of KTN80s and KTN1 during plant growth and development, we obtained the expression patterns of KTN80s and KTN1 from the Genevestigator (https://genevestigator.com/). Notably, the expression of KTN1, KTN80.3, and KTN80.4 is much higher in all tested tissues than that of KTN80.1 and KTN80.2 (Appendix Fig S4A). We further performed qRT‐PCR to finely quantify the expression levels of KTN1/KTN80 genes in different tissues including various floral tissues. The results showed that these five genes show differential expression pattern in different tissues, and KTN80.1, KTN80.3, and KTN80.4 show highest expression level in stamen filaments (Appendix Fig S4B–F). In order to further verify the roles of KTN80s and KTN1 in filament elongation, we investigated their expression profiles using pKTN80::GUS and pKTN1::GUS. Both genes were highly expressed in the filament tissues, with no significant difference between their expression levels (Appendix Fig S5A). Chromatin immunoprecipitation sequencing (ChIP‐seq) data previously revealed that BZR1 can bind the promoter regions of many microtubule‐associated genes, including the microtubule‐severing gene KTN80.4 and the microtubule‐stabilization gene CLASP (Sun et al, 2010), and BZR1/BES1 was also previously reported to negatively regulate CLASP expression to sustain cell proliferation (Ruan et al, 2018). This indicates that BZR1 could directly bind to the KTN80.4 promoter and regulate its expression (Appendix Fig S5B). We therefore selected KTN80.4 as a representative of katanin complexes for our study of their regulation by the BRI1–BR–BAK1 signaling pathway.

To further explore the relationship of BR signaling pathway and KTN80.4, a sequence analysis was performed on the ~3 kb KTN80.4 promoter, revealing several conserved BZR1/BES1 motifs (Fig 5A). We subsequently employed a PCR‐assisted binding‐site selection method to determine the consensus motifs preferentially bound by BZR1/BES1. Compared with wild type, the P7 fragment in the KTN80.4 promoter displayed the highest enrichment of bound BZR1‐YFP in the pBZR1::BZR1‐YFP line (Fig 5A and B). To further verify these results, we purified the BZR1 and BES1 proteins to perform electrophoretic mobility shift assays (EMSAs; Fig EV4A), which showed that BZR1 (Fig 5D) and BES1 (Fig EV4B) could bind to the P7 region, as evidenced by the shifted P7 bands. When we mutated the P7 region by replacing the E‐box motif (CAGTTG) with (AAAAAA), no band shift was detected (Figs 5D and EV4B). This was consistent with the previous finding that the cloning of ~1 kb of the KTN80.4 promoter region fully rescued the assembly of katanin complexes as well as the growth defects (Wang et al, 2017).

Figure 5. BZR1 binds to the promoter region of KTN80.4 and directly activates its activity.

1. Schematic representation of the KTN80.4 promoter showing the positions of two motifs, BRRE and E‐box motifs, which have affinity for the BZR1 family transcription factor (BFTF). ATG denotes the translation start site. TAA denotes the translation stop site. Red arrow indicates E‐box motifs of P7 region, brown lines indicate motif sites, and gray rectangles indicate introns. Fourteen PCR fragments (P1–P14) were designed for the ChIP analysis.
2. ChIP analysis of BZR1 binding to the KTN80.4 promoter region in wild type and pBZR1::BZR1‐YFP plants. Flowers at stages 13 and 14 were harvested for ChIP analysis performed using the anti‐GFP antibody. The red rectangle indicates the P7 region showing the highest enrichment. ACT2 was used as an internal control in PCR quantification. The amounts of DNA fragments were normalized against the input data. Error bars denote the SD of three biological replicates. Student's t test, **P < 0.01, *P < 0.05. Wild type versus pBZR1::BZR1‐YFP in P5, **P < 0.0001; wild type versus pBZR1::BZR1‐YFP in P6, **P < 0.0001; wild type versus pBZR1::BZR1‐YFP in P7, **P < 0.0001; wild type versus pBZR1::BZR1‐YFP in P8, **P < 0.0001; wild type versus pBZR1::BZR1‐YFP in P9, *P = 0.0167; wild type versus pBZR1::BZR1‐YFP in P13, *P = 0.0405.
3. Activation of KTN80.4 by BZR1. The proKTN80.4::LUC and mproKTN80.4::LUC constructs were transformed into Nicotiana benthamiana leaves alongside either a pro35S::4xmyc or pro35S::BZR1‐4xmyc.
4. EMSA showing that BZR1 binds to the E‐box of the KTN80.4 promoter regions in vitro. Nonlabeled (free) native probe in 10‐ to 100‐fold molar excess relative to the biotin‐labeled native probe was used as a cold competitor. The mutant probe was used as a negative control.
5. Quantitative analysis of the luminescence intensity in (C). Similar results were observed in three biological replicates. Error bars indicate SD. Student's t test, **P < 0.01. pro35S::4xmyc in proKTN80.4::LUC versus pro35S::BZR1‐4xmyc in proKTN80.4::LUC, **P < 0.0001.

Source data are available online for this figure.

Figure EV4. BES1 binds the KTN80.4 promoter region and directly activates its activity.

1. Coomassie Blue‐stained purified MBP‐BZR1 and MBP‐BES1 proteins in the EMSA, related to Fig 5D. Left of lane 1: protein marker; lane 1: MBP‐BZR1 protein; lane 2: MBP‐BES1 protein. 10 μl loaded per lane. MBP‐BZR1 and MBP‐BES1 migrate at ~88 kDa.
2. EMSA showing that BES1 binds to the E‐box of the KTN80.4 promoter region in vitro. Nonlabeled (free) native probe in 10‐ to 100‐fold molar excess relative to the biotin‐labeled native probe was used as a cold competitor. The mutant probe was used as a negative control.
3. Activation of KTN80.4 by BES1. The proKTN80.4::LUC and mproKTN80.4::LUC constructs were transformed into Nicotiana benthamiana leaves alongside the pro35S::4xmyc or pro35S::BES1‐4xmyc.
4. Quantitative analysis of the luminescence intensity in (C). Similar results were observed in three biological replicates. Error bars indicate SD. Student's t test, **P < 0.01. pro35S::4xmyc in proKTN80.4::LUC versus pro35S::BES1‐4xmyc in proKTN80.4::LUC, **P < 0.0001.

Source data are available online for this figure.

To gain further insights into the fact that BZR1/BES1 binding to the KTN80.4 promoter region is essential for upregulation of KTN80.4 by the BRs signals in planta, luciferase enzyme activity assays were carried out in Nicotiana benthamiana leaves. The KTN80.4 promoter sequence was used to drive expression of the luciferase‐encoding gene, LUC, in the transgenic plants. The coexpression of pKTN80.4::LUC with p35S::BZR1‐4xMYC resulted in an enhanced luciferase enzyme activity compared with the p35S::4xMYC negative control (Fig 5C and E), and the BES1 results were consistent with the results (Fig EV4C and D). Furthermore, the ability of BZR1/BES1 to bind the promoter of pKTN80.4 (mpKTN80.4) was confirmed to be reduced by the mutation of the E‐box in P7 region (Figs 5D and EV4C). Taken together, these results indicate that BRs signals promote filament elongation by transcriptionally activating the expression of katanin‐encoding genes.

Qui‐2 has disordered microtubule severing and forms a complicated cortical microtubule network

Based on the similar cellular morphologies of the microtubule‐severing defective mutants and qui‐2 (Fig EV3A and E), along with our knowledge that the BFTFs directly activate KTN1 and KTN80s expression, we speculated that katanin‐mediated microtubule severing plays vital roles in the BRs‐associated filament development. We also hypothesized that qui‐2 would have abnormalities in the distribution of its microtubule‐severing events and the resulting microtubule arrays. To determine whether the self‐organization of the microtubule arrays is affected by katanin‐mediated microtubule severing, we performed live cell in qui‐2. We found that filament cells of wild‐type flowers displayed principally parallel microtubule arrays or aligned oblique microtubule arrays, while filament cells in qui‐2 exhibited mainly disordered microtubule arrays (Fig EV5A and B). To gain better insight into the microtubule organization in wild type and qui‐2, we measured the microtubule arrangements in each of the filament cells. A total of 80% of the cells in wild type displayed the parallel arrangement or oblique microtubule arrays, with the rest possessing mesh‐like microtubule arrays. By contrast, mesh‐like microtubule arrays were found in over 60% of qui‐2 filament cells (Fig EV5C). Compared to wild type, the density of microtubule arrays is significantly higher in qui‐2 (Fig EV5D), and the S ₂ value is significantly lower than in wild type (Fig EV5E), indicating a more disordered microtubule arrangement in qui‐2. Taken together, these results support our hypothesis that the BFTFs integrates with microtubule severing in microtubule assembly, thus regulating the elongation of stamen filament cells.

Figure EV5. The qui‐2 displays abnormal microtubule organization in its stamen filament cells.

• A, B
  Global diagram of the microtubule arrays of the filament cells in (A) qui‐2 and (B) wild type. The microtubule organization is magnified in the lower right corner. Scale bars, 15 μm.
• C
  Microtubule morphology distribution in wild type and qui‐2. Seven stamen filaments from wild type and seven stamen filaments from qui‐2 were analyzed. Error bars indicate SD. Statistically significant differences between groups were tested using one‐way ANOVA, **P < 0.01. Wild type versus qui‐2 in parallel microtubule arrays, **P < 0.0001; wild type versus qui‐2 in oblique microtubule arrays, **P < 0.0001; wild type versus qui‐2 in disordered microtubule, **P < 0.0001.
• D
  Microtubule density in wild type and qui‐2. Six cells were analyzed in wild type and qui‐2, respectively. Error bars indicate SD. Student's t test, **P < 0.01. Wild type versus qui‐2, **P < 0.0001.
• E
  S ₂ values in wild type and qui‐2. Five cells were analyzed in wild type and qui‐2, respectively. Error bars indicate SD. Student's t test, **P < 0.01. Wild type versus qui‐2, **P = 0.0003.

Through the analysis of image sequences, we tracked sufficient katanin‐associated events in wild type and qui‐2 (Fig 6A–C, Movies EV1 and EV3). As expected, the katanin fluorescent particles localizing on microtubules were markedly decreased in qui‐2 (Fig 6D), which was in accordance with the expression levels changes of KTN1 and KTN80s compared to wild type (Appendix Fig S3B). Among these events, the fluorescent katanin particles at microtubule lateral lattice and nucleation sites were largely reduced. However, the fluorescent katanin particles at microtubule crossovers only show slight changes between wild type and qui‐2. This bias could be caused by the fact that there exist more microtubule crossovers in qui‐2 than in wild type. To evaluate the localization of katanin fluorescent particles on microtubule crossovers in consideration of different microtubule arrays between qui‐2 and wild type, we detected the proportion of katanin‐associated microtubule crossovers. In wild type, the majority of microtubule crossovers (71.2%) are associated with katanin fluorescent particles. On the other hand, only less than half of the microtubule crossovers (38.3%) are associated with katanin fluorescent particles in qui‐2 (Fig 6F). These observations further confirmed that the organization of katanin complexes on microtubule crossovers is suppressed. We next calculated the normalized severing frequency on microtubule crossovers in wild type and qui‐2. Predictably, the normalized severing frequency is decreased in qui‐2 (Fig 6E). Besides, we also noticed that the residency time of fluorescent katanin particles at microtubule crossovers and nucleation sites is reduced compared with wild type (Fig 6G). Many of katanin fluorescent particles disappear soon after targeting the microtubules and fail to trigger severing events (Fig 6C and Movie EV4). Collectively, our live‐cell observations revealed that suppressed microtubule severing activity in qui‐2 is caused by both decreased microtubule recruitment of katanin complexes and reduced severing frequency. Consequently, a mesh‐like cortical microtubules are formed in qui‐2 stamen filament cells, leading to defective elongation of stamen filaments and failed pollination.

Figure 6. The qui‐2 has reduced microtubule severing and katanin residency.

• A–C
  Time‐lapse images showing that KTN1 is associated with microtubule severing at crossover sites in (A) wild type and (B and C) qui‐2 filament cells. GFP‐labeled KTN1 fluorescent particles are pseudo‐colored in red, and microtubules (labeled with mCherry‐TUB6) are shown in green. Yellow arrows indicate KTN1 at the microtubule crossover sites. Asterisks indicate the location at which a microtubule has been severed at a crossover site in (A and B), or the location at which a microtubule has not been severed at a crossover site in (C). Scale bar, 5 μm in (A–C).
• D
  Katanin localization frequency on microtubules (events per 100 μm² per second). Purple indicates katanin at microtubule lateral lattice. Green indicates katanin on microtubule crossovers. Orange indicates katanin on microtubule nucleation sites. Four hundred and forty katanin localization events were detected in area of 2,466 μm² from five cells in wild type. Four hundred and eighteen katanin localization events were detected in area of 1,720.57 μm² from four cells in qui‐2. The data represent means SD. Student's t test, **P < 0.01. Wild type versus qui‐2 in katanin at microtubule lateral lattice, **P = 0.0018; wild type versus qui‐2 in katanin on microtubule crossovers, P = 0.5374; wild type versus qui‐2 in katanin on microtubule nucleation sites, **P = 0.0032.
• E
  Normalized severing frequency (severing events per 100 μm² per second per crossover) at crossovers were calculated in (D). Error bars indicate SD. Student's t test, **P < 0.01. Wild type versus qui‐2, **P = 0.0011.
• F
  Comparison of katanin recruitment on microtubule crossovers. Green parts indicate the microtubule crossovers without katanin fluorescent particles during observation. Orange parts indicate microtubule crossovers associated with katanin fluorescent particles during observation. A total of 208 microtubule crossovers from wild type (observed area, 2,377.45 μm²) and 107 microtubule crossovers from qui‐2 (observed area, 1,175.38 μm²) were taken into count, respectively.
• G
  Comparison of the residency times of katanin in different microtubule positions. A total of 415 katanin localization events from five cells of wild type and 309 katanin localization events from three cells of qui‐2 were analyzed. Error bars indicate SEM. Student's t test, **P < 0.01. Wild type versus qui‐2 in inactive katanin fluorescent particles at microtubule crossovers, **P < 0.0001; wild type versus qui‐2 in inactive katanin fluorescent particles at microtubule nucleation sites, **P < 0.0001.

Discussion

Our findings revealed an intriguing natural phenomenon that, among the various mutants of microtubule‐associated genes, the microtubule‐severing defective mutants, lue1 and ktn80.1234, display male sterility because of their short stamen filaments. Here, we combined genetic, histological, biochemical, and live‐cell imaging data to reveal the mechanism by which the BFTFs transcriptionally activate the expression of KTN1 and KTN80s, which in turn precisely regulates microtubule severing and remodels the proper organization of the microtubule array to promote stamen filament elongation (Fig 7).

Figure 7. A working model showing that BZR1 family members bind to the KTN1/KTN80s promoter region and activate its expression to regulate filament elongation in the BR signaling pathway.

Katanin‐mediated microtubule severing plays both a specific and an indispensable role in regulating stamen filament elongation

The microtubule cytoskeleton has been showed to play crucial roles in cell elongation during plant cell morphogenesis (Ehrhardt & Shaw, 2006). In response to developmental signals and environmental stimuli, microtubules undergo dynamic remodeling achieved by various activities, such as nucleation, severing, damage, and repairing (Lindeboom et al, 2013). Intriguingly, we uncovered that stamen filament failure specifically occurs in mutants with microtubule‐severing defects. Katanin‐mediated microtubule severing has been shown to play a central role in generating well‐aligned microtubule arrays to sustain rapid cell elongation in plants, such as rapidly elongating hypocotyls or root cells where cortical microtubules are organized in transverse arrays perpendicular to the growing direction (Webb et al, 2002; Chen et al, 2014; Peaucelle et al, 2015; Elliott & Shaw, 2018). During stamen development, stamen filaments keep much shorter than the pistil till stage 12. Stamen filament growth accelerates at stage 12 and reaches the same height with the pistil at stage 13 (Smyth et al, 1990). Thus, we speculate that proper microtubule‐severing activity is required for rapid elongation of stamen filaments at stages 12 and 13. Detailed expression profile analysis further revealed that KTN1 and KTN80s are differentially expressed between the developing stamens and pistils, with both being highly expressed at floral stages 12 and 13 when the stamen filaments rapidly elongate (Fig 1 and Appendix Fig S5A). Therefore, our findings further support the specificity and necessity of microtubule severing in the reorganization of microtubule arrays to sustain rapid elongation of plant cells.

The BR signaling pathway couples with katanin‐mediated microtubule remodeling to regulate stamen filament elongation

So far, the molecular mechanisms that connect cellular signaling of plant growth to cytoskeletal organization and function have not been well delineated. As mentioned above, two previous publications have shed light on these mechanisms, both involving signaling to katanin, by small G protein ROP6 and by perception of blue light, respectively (Lin et al, 2013; Lindeboom et al, 2013). Importantly, we here uncovered the BFTFs transcriptionally activate the expression of KTN1 and KTN80s to regulate microtubule‐severing events, thus controlling proper stamen filament elongation. Given that six BFTF‐encoding genes and four KTN80‐encoding genes play redundant roles in the BR signaling pathway and microtubule severing, respectively (Wang et al, 2017; Chen et al, 2019a, 2019b). Using KTN80.4 as a representative component, we performed multifaceted elaborate experiments and demonstrated that BZR1 and BES1 could transcriptionally activate the expression of KTN80.4, thus confirmed the interplay of BRs signaling with katanin‐mediated microtubule severing to regulate stamen filament elongation. Future researches should also explore whether different BFTFs regulate different KTN80s to regulate filament cell elongation.

As expected, we observed that less katanin particles localize at microtubule crossovers and microtubule severing frequency is reduced in qui‐2 cells, when compared to wild type. We also noticed that the duration time of katanin complexes on microtubule crossovers and nucleation sites is reduced when they fail to trigger microtubule severing (Fig 6). One possibility is that the formation of katanin complexes is compromised due to lower dosage of KTN1 and KTN80s, thus producing more incompetent katanin complexes that are less stable and show shorter residency time at microtubules in qui‐2. In addition, it should be noticed there were a set of genes encoding microtubule‐associated proteins, which were identified by ChIP‐seq of BZR1 (Appendix Fig S5B). Thus, another possibility could be that these factors play important roles either in facilitating the formation or in promoting microtubule recruitment of KTN1/KTN80 complexes. Yet it needs to further explore whether these factors are involved in filament elongation by regulating katanin‐dependent microtubule severing.

We also expect that overexpression of KTN1 could genetically rescue the filament elongation defects of qui‐2. However, numerous studies have shown that proper dynamic remodeling of microtubules is required for plant cell development, and altered microtubule organization either by underexpression or by overexpression of a tubulin‐encoding gene or a MAP (microtubule‐associated protein)‐encoding gene usually causes abnormal plant cell development (Zhong et al, 2002; Ambrose et al, 2007; Burk et al, 2007; Kirik et al, 2007; Korolev et al, 2007; Zhou et al, 2007; Wan et al, 2014; Zhu et al, 2015). For example, either knock‐out or overexpression of FRA1/Kinesin 4A, encoding a kinesin motor, results in a dwarf phenotype in Arabidopsis (Zhong et al, 2002; Zhou et al, 2007). In particular, prior work has demonstrated that the KTN1 overexpression line shows growth retardation with defective filament elongation as shown in the KTN1 knockout line, fra2 (Burk et al, 2007), and that KTN80a overexpression in rice shows retarded root growth (Wan et al, 2014). We further generated the KTN1 overexpression line and confirmed the dwarf growth phenotype with filament elongation defects, as shown in the KTN1 knockout line, lue1 (Appendix Fig S6). Thus, we no longer expect genetic rescue of short stamen filament in qui‐2 by overexpressing KTN1. Alternatively, we are expecting more advanced methods to monitor spatiotemperal transcription activation of katanin genes and precise microtubule severing, which orchestrate the regulation on stamen filament elongation.

Stamen development is regulated by the crosstalk among the BR, auxin, jasmonic acid (JA), and gibberellin (GA) pathways in Arabidopsis thaliana

Many of the studies of filament elongation have focused on the development of stamens, and the molecular mechanisms by which phytohormone signaling pathways regulate filament development, involving both synergistic and antagonistic actions among hormone signals (Nagpal et al, 2005; Cheng et al, 2009; Song et al, 2013; Qi et al, 2015). Auxin has been shown to play a major role in tapetum development and filament elongation, with the tir1 afb1 afb2 afb3 quadruple auxin perception mutant and arf6‐2 arf8‐3, an auxin response factor mutant, both producing short filaments (Nagpal et al, 2005; Cecchetti et al, 2013). And MSG2, an auxin primary response gene IAA19, may be one of the master genes that regulate elongation of stamen filaments in the catch‐up phase of stamen growth (Tashiro et al, 2009) and a flower‐specific splice variant of the auxin response factor ARF8 regulates stamen elongation in Arabidopsis (Ghelli et al, 2018). It has also been shown that the mechanism by which auxin transport facilitates the regulation of filament elongation is distinct from the way in which it coordinates anther dehiscence and pollen maturation during stamen development (Cecchetti et al, 2004, 2008). In addition, JA is shown to play crucial roles in filament elongation; the allene oxide synthase mutant aos and the COR‐insensitive mutant coi1 display male sterility because of defects in their filament elongation (Park et al, 2002; Thines et al, 2007). Furthermore, the myc2 myc3 myc4 myc5 quadruple mutant and the myb21 myb24 double mutant, which lack components of the bHLH–MYB transcription complex downstream of JA signaling, exhibit obvious defects in stamen development (Qi et al, 2015). The role of GA in stamen filament development has also been studied. The stamen development of ga1‐3, a GA synthesis mutant, stagnates at the 10^th stage of flower development, and stamen filaments failed to elongate normally (Cheng et al, 2004). The GA synthesis defective mutants, ga20ox and ga3ox, also show defects in stamen filament elongation (Hu et al, 2008; Rieu et al, 2008; Plackett et al, 2012).

Importantly, accumulating evidence support the notion that multiple phytohormones act in concert to regulate stamen filament elongation. For example, auxin has been shown to regulate stamen development by influencing JA biosynthesis, which in turn induces the expression of transcription factors MYB21 and MYB24 to promote stamen filament elongation (Reeves et al, 2012). A recent report illustrated that the auxin response factors ARF6 and ARF8 activate the transcription of DWARF4 to facilitate BR biosynthesis, which modulates growth anisotropy and regulates leaf shape (Xiong et al, 2021). Thus, another possibility could be that core transcriptional factors of BR, JA, auxin, and GA pathways might form a complex to regulate stamen filament development in Arabidopsis, as BZR1, ARF6 and PIF4 have been shown to form a complex to regulate hypocotyl elongation in Arabidopsis (Oh et al, 2014). However, the mechanisms underlying their crosstalk and coordinated activities are yet to be fully elucidated.

Materials and Methods

Plant materials and growth conditions

The Columbia ecotype of Arabidopsis (Col‐0) was used in this study. ktn80.1234, ktn80.12, ktn80.34, lue1, qui‐1, qui‐2, bzr1‐1D, bri1‐116, pBZR1::BZR1‐YFP have been described (Zhang et al, 2009; Wang et al, 2017; Chen et al, 2019b). The pTUB6::mCherry‐TUB6‐pKTN1::VisGreen‐KTN1 marker line was described in our previous studies (Wang et al, 2017; Liu et al, 2019). Various cross combinations qui‐2 (bes1 ⁺) and the pTUB6::mCherry‐TUB6‐pKTN1::VisGreen‐KTN1 marker line (T6‐K1) were performed to produce materials to observe the dynamics of katanin complexes and microtubules. Their homozygote qui‐2 (T6‐K1) were verified by genotyping. After vernalizing at 4°C for 3 days, all the mentioned wild type and mutant seeds were planted and grown in greenhouse at 22°C under 16 h‐light/8 h‐dark photoperiod condition. Marker line plasmid (T6‐K1) were transferred into Agrobacterium tumefaciens GV3101 using the Agrobacterium tumefaciens‐mediated transformation procedures. For every transformation, phenotypes of transgenic plants were verified in at least five independent transgenic lines.

Induced gene expression method was performed according to a previous report with minor modifications (Saito et al, 2015). As follows, 20–30 wild‐type seeds were incubated in a 150‐ml Erlenmeyer flask with 40 ml liquid MS medium containing 1% sucrose under continuous light (20–36 μmol m⁻² s⁻¹) with shaking (130 rpm). Plants were treated with 10 nM epibrassinolide (eBL; Sigma) and harvested at the designated time points.

Nicotiana benthamiana plants used for transient expression assays were grown on soil in greenhouse at 22°C under 16 h‐light/8 h‐dark photoperiod condition at 22°C for 3–4 weeks.

RNA extraction and gene expression analyses

Total RNA was extracted from different tissues of A. thaliana using RNAprep Pure Plant Kit (TIANGEN) reagent and cDNA was prepared from 2 μg of total RNA with Superscript III reverse transcriptase (Invitrogen) and an oligo dT primer according to the manufacturer's instructions.

The qRT‐PCR assays were performed in a 20 μl reaction volume using SYBR Green real‐time PCR master mix (Toyobo), and expression levels of target genes were normalized to UBQ5 (At3G62250) using the $2^{-\Delta \Delta C_{\mathrm{T}}}$ method (Livak & Schmittgen, 2001; Wang et al, 2017). All experiments were carried out in triplicate on the BioRad CFX96 system. The primers used in qRT‐PCR are shown in Appendix Table S1.

Histochemical staining and Alexander staining

About 1 kb promoter sequence before the start codon (ATG) of the four KTN80s and KTN1 coding sequence was amplified from Arabidopsis (Col‐0) gDNA and cloned into the pCAMBIA1381 vector, which was confirmed by sequencing respectively. The pKTN80s::GUS and pKTN1::GUS construct was introduced into Arabidopsis (Col‐0) using the Agrobacterium tumefaciens. Filament tissues from the transformed plants were obtained in floral stage 12 and 13 and were stained for GUS activity for 7 h, as described previously (Jefferson, 1989). The anthers were stained with the Alexander solution for 40 min (Alexander, 1969). The stained tissues were viewed under a stereo microscope and photographed using a digital camera. For semi‐thin section, the filament tissues were fixed overnight in 4% (w/v) paraformaldehyde in phosphate‐buffered saline, pH 7.4, at 4°C, dehydrated in ethanol and embedded in Technovit 7100. The samples were cut into sections (thickness, 5 μm) with a Leica RM2265 microtome, stained with 0.05% toluidine blue (Sigma‐Aldrich).

Tobacco leaf transient expression system for KTN80.4 promoter activity

The coding sequence of BZR1/BES1 was cloned to the Gateway donor vector pEntry‐topo‐SD and then recombined into the pGWB17 vector under the control of the Cauliflower mosaic virus 35S promoter by LR reactions (Invitrogen). Meanwhile the KTN80.4 promoter region was cloned into binary vector pGWB435 by LR reactions (Invitrogen) to generate the reporter construct pKTN80.4::LUC. Site‐Directed Mutations of pKTN80.4 were carried out using Mut Express MultiS Fast Mutagenesis Kit V2 (Vazyme). The transient expression assay was performed in N. benthamiana leaves according to previous described methods (Shang et al, 2010). The vector was transferred into Agrobacterium tumefacien GV3101 by electric shock. Agrobacterium tumefaciens GV3101 containing effector (p35S::4xmyc or p35S::BZR1/BES1‐4xmyc) or reporter (pKTN80.4::LUC or mpKTN80.4::LUC) were collected with 3,220 g, washed with twice injection buffer, and suspended in injection buffer (10 mM MES, pH 5.7, 10 mM MgCl₂ and 50 mM acetosyringone) to an OD₆₀₀ = 0.8, and Agrobacterium tumefaciens GV3101 containing effector and Agrobacterium tumefaciens GV3101 containing reporter were mixed with 1:1. After 2 h of co‐culture, injection was given, and then dark treatment was performed for 36 h. The detached leaves were sprayed with 1 mM luciferin (Promega), and the LUC signal was captured with a low‐light cooled charge‐coupled device camera (Night owl LB985, Berthold Technologies, Germany), and relative LUC activity was measured.

Chromatin immunoprecipitation and real‐time PCR

Chromatin immunoprecipitation was performed according to a previous report with minor modifications (He et al, 2012). Filament tissues form floral stage 13 and 14 containing pBZR1::BZR1‐YFP and wild type were harvested and fixed with 1% formaldehyde under vacuum for 30 min and terminated with 2.5 M glycine. Chromatin was sonicated with a Bioruptor 30 times in lysis buffer (50 mM HEPES, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% SDS, 1% Triton X‐100, 0.1% sodium deoxycholate, and 1 mM PMSF with 1 × Roche protease inhibitor cocktail) and then incubated at 4°C for 4 h with GFP antibody (Abcam). The sonicated chromatin regarded as an input. Immunoprecipitated complexes were collected using protein A/G beads (Millipore) and washed with LiCl washing buffer (0.25 M LiCl, 0.5% Nonidet P‐40, 0.5% sodium deoxycholate, 1 mM EDTA, and 10 mM Tris–HCl, pH 8.0). Reverse protein DNA cross‐linking was performed by incubating the immunoprecipitated complexes at 95°C for 10 min. Quantitative real‐time PCR analysis was performed using specific primers corresponding to different promoter regions. The primers are showed in Appendix Table S1.

Protein expression and electrophoretic mobility shift assay (EMSA)

To construct plasmids for the expression of ATBZR1 and ATBES1 protein in Escherichia coli, the full‐length coding sequences of ATBZR1 and ATBES1 were amplified and cloned into pET21a‐MBP vector, and the resulting constructs were transformed into E. coli strain BL21 (DE3). The recombinant ATBZR1 and ATBES1 protein was purified using Streptactin Beads 4FF following the manufacturer's procedures (Smart‐lifesciences, Changzhou, China). Oligonucleotide probes containing a E‐box motif (CAGTTG) and a mutant E‐box motif (AAAAAA) were synthesized and labeled with biotin at the 3′ end (Invitrogen). For competition experiments, nonlabeled probe was added to the reactions. EMSA was performed using a LightShift Chemiluminescent EMSA kit (Thermo Scientific). The binding reaction was performed in a 20 μl reaction mixture containing the ATBZR1 and ATBES1 protein, 20 nmol of synthetic biotin‐labeled DNA probe, and 50 ng poly(dI‐dC). The reaction mixtures were incubated at room temperature for 30 min and then the samples were separated on a 6% native polyacrylamide gel in 0.5 × TBE buffer (45 mM Tris‐borate, 1 mM EDTA, pH 8.0). Electrophoresis was performed at 15 V/cm in 0.5 × TBE buffer. For the competition assay, appropriate amounts of unlabeled DNA fragments were used as competitors and were added to the reaction prior to the addition of the proteins. The labeled probes were transferred to positively charged nylon membrane and detected using the chemiluminescent nucleic acid detection module provided with the kit. Probe sequences are shown in Appendix Table S1.

Tissue preparation and scanning electron microscopy

Filament tissues form floral stage 13 were harvested and were stabilized on the specimen chamber with conducting resin. They were frozen in the liquid nitrogen via frozen transmission system (Quorum PP3010T). The samples were broken and sublimated in −70°C for 10 min. After sputtering with platinum (10 mA, 60 s), the imaging was performed in a high‐resolution SEM 7401F (JEOL, Japan). The prepared samples could be directly inserted into the specimen chamber without being removed from the high vacuum.

Measurement of gynoecium and filament length

To measure carefully the filament and gynoecium length, flowers at floral stage 13 for each genotype were harvested and placed on 1/2 MS medium, and the gynoecium and one of the two longer filaments for each flower were measured using a type microscope (Leica M205FA) and ImageJ.

Microscopy and image analysis

Filament form floral stages 12 and 13 were cut and put in an imaging glass slide for microscopy as described (Wang et al, 2017). Live cell imaging was captured under a spinning disk confocal microscope equipped with the Yokogawa® Nipkow CSU‐X1 spinning disk scanner, Hamamatsu EMCCD 9100‐13, Nikon TiE inverted microscope with the Perfect Focus System. For the Perkin Elmer system, GFP and mCherry were motivated by the 488 nm and 561 nm lasers. A 100X plan Apo oil immersion TIRF objective (NA = 1.49) was used for image acquisition, which were a maximum Z projection of image series acquired in an interval of 0.3 μm.

Images were processed and analyzed using Volocity (Perkinelmer) and Image J. As follows, movies were acquired every 5 s during the course of 300 s, and the time‐lapsed images were used for evaluating the relationship between katanin complexes and microtubules severing events, and for calculating of residency time of katanin complexes fluorescent particles. Then, reduction of the images noise was carried out using the plugins “enhance contrast” and “subtract background” of ImageJ.

To quantify the degree of alignment of microtubule arrays, the order parameter S ₂ , was employed. It takes values between 0, for a perfectly isotropic microtubule array and 1, for a system in which all microtubules are aligned. The S ₂ values were calculated as described in a previous study (Deinum et al, 2017). In brief, we measured the length and the orientation of each microtubule in individual filament cell from acquired confocal images. At least five represented cells from each mutant and wild type were taken into count, respectively.

The frequency of microtubule severing was calculated as described in our previous study (Wang et al, 2017). In brief, we measured the area of an entire filament cell in the acquired time‐lapsed series, and all the severing events were tracked during the course of the image stack. The frequency of microtubule severing was then calculated as counts of severing events occurs per second per 100 square microns. Since the microtubule severing events are frequently occurs at microtubule crossovers, we also employed crossover numbers for normalization of the severing frequency.

Statistical analysis

Statistical analysis was performed using GraphPad Prism 7.0. Data were compared either by Student's two tailed t‐test or one‐way ANOVA followed by post hoc Tukey's test. For box plots, the boxes show the first and third quartiles, split by median. Details of statistical analysis, number of quantified entities (n), and measures of dispersion can be found in the figure legends.

Author contributions

Jie Wang: Resources; formal analysis; investigation; writing – original draft; writing – review and editing. Guangda Wang: Data curation; formal analysis; writing – review and editing. Weiwei Liu: Formal analysis; investigation. Huanhuan Yang: Investigation. Chaofeng Wang: Investigation. Weiyue Chen: Resources; investigation. Xiaxia Zhang: Investigation. Juan Tian: Investigation. Yanjun Yu: Investigation. Jia Li: Resources; investigation. Yongbiao Xue: Resources; supervision; investigation. Zhaosheng Kong: Supervision; funding acquisition; writing – original draft; project administration; writing – review and editing.

Disclosure and competing interests statement

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Movie EV1

Movie EV2

Movie EV3

Movie EV4

Source Data for Expanded View

PDF+

Source Data for Figure 5

The EMBO Journal (2023) 42: e111883

Data availability

All data are available from the corresponding authors upon request.