Microtubule-associated IQD9 guides cellulose synthase velocity to shape seed mucilage

Arabidopsis seeds release large capsules of mucilaginous polysaccharides, which are shaped by an intricate network of cellulosic microfibrils. Cellulose synthase complexes is guided by the microtubule cytoskeleton, but it is unclear which proteins mediate this process in the seed coat epidermis (SCE). Using reverse genetics, we identified IQ67 DOMAIN 9 (IQD9) and KINESIN LIGHT CHAIN-RELATED 1 (KLCR1) as two highly expressed genes during seed development and comprehensively characterized their roles for cell wall polysaccharide biosynthesis and cortical microtubule (MT) organization. Mutations in IQD9 as well as in KLCR1 lead to compact mucilage capsules with aberrant cellulose distribution, which can be rescued by transgene complementation. Double mutant analyses revealed that their closest paralogs (IQD10 and KLCR2, respectively) are not required for mucilage biosynthesis. IQD9 physically interacts with KLCR1 and localizes to cortical MTs to maintain their organization in SCE cells. Similar to the previously identified TONNEAU1 (TON1) RECRUITING MOTIF 4 (TRM4) protein, IQD9 is required to maintain the velocity of cellulose synthases. Our results demonstrate that IQD9, KLCR1 and TRM4 are MT-associated proteins that are required for seed mucilage architecture. This study provides the first direct evidence that members of the IQD, KLCR and TRM families have overlapping roles in guiding the distribution of cell wall polysaccharides. Therefore, SCE cells provide an attractive system to further decipher the complex genetic regulation of polarized cellulose deposition.


Introduction 58
The seed coat epidermal (SCE) cells of some Angiosperms, including the model plant 59 Arabidopsis thaliana, synthesize large amounts of hydrophilic polysaccharides (North et al.,60 2014; Voiniciuc et al., 2015c;Šola et al., 2019). Although the mucilage capsules that rapidly 61 encapsule Arabidopsis seeds upon hydration are pectin-rich, they can be regarded as 62 specialized secondary cell walls because they also contain hemicelluloses that are typical of 63 woody tissues (Voiniciuc et al., 2015c). Substituted  was previously generated (Yang et al., 2019). Transient expression was performed in N. 161 benthamiana leaves as previously described (Grefen et al., 2010). In short, Agrobacterium 162 tumefaciens GV3101 cells containing the desired constructs were mixed with the P19 viral 163 suppressor (OD600 = 0.7 for each) and incubated for 4 h (at 18 °C, 200 rpm) before 164 infiltration into the lower side of leaves from 5-week-old plants. 165 166

Staining and quantification of mucilage area 167
Around 30 seeds were hydrated in water for 30 min and stained with 300 µl of 0.01% 168 (w/v) RR (Sigma-Aldrich; R2751) for 15 min at 125 rpm in 24-well plates. After rinsing with 169 water, the stained seeds were re-suspended in 300 µl of water and imaged with a Leica 170 M165FC stereomicroscope equipped with MC170 HD camera. The mucilage and seed 171 projected areas were quantified using an existing ImageJ pipeline (Voiniciuc et al., 2015b). To view surface morphology, around 30 seeds were mixed with 500 µL of 0.01% (w/v) 180 propidium iodide for 15 min. Seeds were rinsed twice with water and imaged using a Leica 181 LSM 900 with 10X/0.3 objective (excitation 488 nm, emission 600-650 nm). 182 183

Seed polysaccharide quantification 184
Non-adherent mucilage was extracted by gently mixing 5 mg seeds in water for 30 min 185 at 125 rpm and subsequently the adherent mucilage was isolated using a ball mill (Retsch; 186 MM400) for 30 min at 30 Hz, as previously described (Voiniciuc, 2016). The two mucilage 187 fractions, spiked with ribose and inositol respectively, were hydrolyzed and quantified via 188 high-performance anion exchange chromatography with pulsed amperometric detection 189 (HPAEC-PAD), as described   Crystalline cellulose was quantified using the Updegraff reagent (Updegraff, 1969) and 194 the anthrone colorimetric assay (Foster et al., 2010), as previously adapted for Arabidopsis 195 whole seeds (Voiniciuc et al., 2015b). 196 197

Salt stress treatments 198
Germination assays were performed in 24-well culture plates as described previously 199 (Yang et al., 2021). Around 35 seeds were hydrated in 500 µl of water or 150 mM CaCl2 200 solution per well. All the seeds were vernalized for 66 h (dark, 4°C), transferred to a chamber 201 with constant light (100-120 µmol m -2 s -1 ), 22°C and 60% humidity. The seeds were imaged 202 every 24 h with a Leica M165FC stereomicroscope and defined as germinated when the 203 radicle length was > 70 µm. 204 For the seedling salt stress assay, the seeds were placed on ½ Murashige and Skoog agar 205 plates, stratified for 66 h (dark, 4°C) and grown vertically in the climate-controlled chamber 206 described above. Five-day-old seedlings of similar size were transferred to fresh agar plates 207 with or without 100 mM NaCl and growth was imaged using a Nikon D5600 digital camera. The bottom 7 cm of mature stems were harvested and homogenized using a ball mill 214 (Retsch MM400) for 10 min at 30 Hz. The AIR was extracted by sequential washes with 1 215 mL of 70% (v/v) ethanol, 1 mL of 1:1 (v/v) chloroform:methanol and 1 mL of acetone. Stem 216 AIR was hydrolyzed and quantified via HPAEC-PAD exactly as described for seed mucilage, 217 using ribose as the internal standard. 218 219

Confocal microscopy and image analysis 220
For fluorescence co-localization assays in N. benthamiana, leaf discs at 3 days post-221 infiltration were imaged with a Zeiss LSM880 inverted confocal microscope using a 40X/1. All images were processed uniformly using ImageJ. The maximum projection of the Z-242 stack or time-lapse view was generated by frames using the Z Project tool and average 243 intensity. For the colocalization evaluation, the intensity plot analysis was done by "RGB 244 Profile Plot" plugin. The Pearson correlation coefficient of region of interest (ROI) from 245 single frame was quantified with "Coloc 2". Kymograph analysis of proteins and velocity 246 quantification of GFP-CESA3 were performed as previously described (Vellosillo et al., 247 2015). Briefly, the time-lapse stack was generated with "Walking Average" plugin. The GFP-248 CESA3 track was depicted using segmented line on time average image and transferred to 249 time-lapse stack. The kymographs were generated using the "MultipleKymograph" plugin, 250 and the slope of each line was used to calculate the particle velocity. showed high seed coat activity for pIQD9 (Fig. S1). IQD9 and IQD10 displayed partially 279 overlapping expression profiles but share only 47% amino acid identity (Fig. S1). Furthermore, 280 iqd9-1 iqd10-1 (i9 i10) double mutant seeds showed compact mucilage capsules resembling 281 the iqd9 single mutants (Fig. 1c). 282 Since IQD-KLCR interactions have been reported (Bürstenbinder et al., 2013), we also 283 assessed if KLCR1/CMU1 (At4g10840) and KLCR2/CMU2 (At3g27960) are involved in 284 mucilage biosynthesis. KLCR1 was highly expressed throughout the seed development, while 285 KLCR2 transcription peaks at the pre-globular and globular stages, before mucilage 286 biosynthesis (Fig. S2a). Two knockout klcr1 alleles, klcr1-1 and cmu1 (Fig. S2b, Table S1), 287 resembled the iqd9 compact mucilage defect (Fig. 1a), while klcr2 seeds displayed WT-like 288 mucilage. Both iqd9 as well as klcr1 mutants reduced mucilage capsule area by 30-40% 289 compared to WT (Fig. 1d), without altering seed size. The double mutant i9 i10 and klcr1-1 290 klcr2-2 (k1 k2) phenocopied the mucilage structure of iqd9-1 and klcr1-1 respectively, 291 indicating no functional redundancy between the related genes. Transgene complementation 292 of i9 i10 with IQD9 and of klcr1-1 with KLCR1 fully rescued the compact mucilage defects 293 (Fig. 1). Both iqd9 and klcr1 mutants resembled the mucilage phenotype of trm4, which has 294 SCE cells with disorganized MTs (Yang et al., 2019). Moreover, iqd9 klcr1 and iqd9 trm4 295 double mutants displayed compact RR-stained mucilage capsules equivalent to the single 296 mutants (Fig. S2c-f). 297 298 IQD9 and KLCR1 are specifically required for cellulose distribution in mucilage 299 Monosaccharide analysis of non-adherent and adherent mucilage fractions revealed that 300 the iqd9, klcr1 and trm4 mutants did not alter the content or adherence of matrix 301 polysaccharides to the seed surface (Fig. 1e). Surprisingly, despite no impact on RR staining, 302 the iqd10-1 mutation correlated with a small but statistically significant reduction in total 303 extractable monosaccharides (Fig. 1c-e). The iqd10-1 mutation decreased rhamnose (Rha) 304 and galacturonic acid (GalA) content in non-adherent mucilage, along with galactose (Gal) 305 and arabinose (Ara) in adherent mucilage (Fig. 2). Nevertheless, compared to the 306 galactoglucomannan-deficient muci10 mutant with compact mucilage, all the MT-related 307 mutant seeds released relatively normal levels of mucilage glycans (Fig. 2). 308

309
Since the matrix polysaccharide composition could not account for the compact 310 mucilage defects of iqd9 and klcr1, we then examined the structure of cellulose using S4B, a 311 specific fluorescent dye (Anderson et al., 2010). The iqd9 and klcr1 mutant seeds extruded 312 less cellulose upon hydration (Fig. 3a) and had ~30% shorter rays atop each columella 313 compared with WT (Fig. 3b). Moreover, the mutant seeds lacked the diffuse cellulose 314 staining that was observed between the WT rays. The S4B-stained seeds of i9 i10 and k1 k2 315 resembled the iqd9 and klcr1 single mutants, while iqd10 and klcr2 had WT-like seeds. The 316 cellulose defects of i9 i10 and klcr1 mutants were rescued by IQD9 and KLCR1 transgene 317 complementation, respectively, using their native promoters (Fig. 3a). While all the iqd9, 318 klcr1 and trm4 mutant combinations examined showed short S4B-stained cellulosic rays 319 compared to WT (Fig. 3 and Fig. S3), counterstaining of mucilage with calcofluor displayed 320 relatively normal content of other β-glucans (Fig. S3). Despite evident changes in the 321 architecture of cellulose extruded from hydrated seeds, these MT-related mutants did not alter 322 the crystalline cellulose content of whole seeds (Fig. 3c). Mutations in IQD9 or KLCR1 did 323 not alter seed shape (Fig. S4), nor the germination rate or sensitivity to salt (Fig. S5). and siliques (Fig. S1). Except for anthers, pIQD9 showed higher activity than pIQD10 in 332 reproductive organs and seeds (Fig. S2). IQD10 was expressed highest in Arabidopsis stems 333 and its ortholog in Populus deltoides (PdIQD10) affects the development of the woody stem 334 (Badmi et al., 2018). However, stem cross-sections of the iqd10 single mutants and the i9 i10 335 double mutant did not show the irregular xylem (irx) phenotype observed in secondary cell 336 wall mutants such as irx14 (Fig. S6a). Furthermore, iqd stems had normal monosaccharide 337 composition (Fig. S6b), while irx14 stems were xylan-deficient as previously described 338 To determine the subcellular localization of IQD9 proteins, we first co-expressed IQD9-343 GFP fusion proteins and the MT marker RFP-TUB6 in N. benthamiana leaf epidermal cells. 344 IQD9-GFP localized in striated arrays that overlapped with RFP-TUB6 at the cell cortex (Fig.  345 S7), and could be abolished by treating cells with MT-depolymerizing oryzalin. Next, we co-346 expressed IQD9-GFP with mCherry-tagged KLCR1 (mCherry-KLCR1) in N. benthamiana 347 and found that they were co-localized in arrays resembling MTs (Fig. 4a-c). IQD9 and 348 KLCR1 still co-localized when their striated patterns were disassembled by oryzalin 349 treatment (Fig. 4c-e). We validated that these two proteins physically interact using co- To investigate the distribution of IQD9-GFP in Arabidopsis, we examined its subcellular 361 localization under the control of its native promoter in the complemented iqd9 line, which 362 rescued the mucilage defects (Fig. 1). While undetectable in young seedlings, IQD9-GFP 363 fluorescence was evident during seed coat development (Fig. S8a), particularly at the peak 364 stage of mucilage biosynthesis. Z-stack maximum projections displayed IQD9-GFP proteins 365 in MT arrays, near the PM and inside the nucleus (Fig. 5a). IQD9-GFP displayed circular 366 arrays around the cytoplasmic column, resembling previously described CESA trajectories 367 during mucilage production (Griffiths et al., 2015). At SCE cell boundaries, IQD9-GFP 368 proteins co-localized with the PM stained by FM4-64. Time-lapse imaging revealed that 369 IQD9-GFP proteins were static (the vertical lines in kymograph; Fig. 5b), as previously noted 370 for KLCR/CMU proteins (Liu et al., 2016). Highly immobile KLCR1-GFP proteins, 371 expressed under its native promoter in the complemented klcr1 line, were also associated 372 with both MTs and PM throughout SCE development (Fig. S8b), but lacked the nuclear 373 localization observed for IQD9-GFP. In cross-sectional views of live SCE cells, both IQD9-374 GFP and KLCR1-GFP were localized primarily as striated arrays adjacent to the mucilage 375 pocket (Fig. 5c,d). 376

IQD9 maintains MT organization in SCE cells 378
Proper MT organization is essential for the establishment of mucilage architecture. The 379 MT marker RFP-TUB6, which formed circular arrays around the cytoplasmic column of SCE 380 cells at 7 DPA (Yang et al., 2019), was introduced into the iqd9-1 mutant by crossing. In 381 contrast to the WT background, circular MT arrays were undetectable in all iqd9 SCE cells 382 expressing RFP-TUB6 (Fig. 6). While MT organization was severely disrupted in the seed 383 coat, both WT and iqd9 displayed transversely oriented RFP-TUB6 arrays in hypocotyl 384 epidermal cells (Fig. S9), despite some variation in fluorescence intensity. Therefore, the 385 distribution of cortical MTs in the seed coat depends heavily on IQD9, while cytoskeleton 386 organization in other tissues likely requires additional IQDs. 387

388
The loss of IQD9 reduces CESA3 velocity 389 CESA3 is a key subunit of the CSC that polarly deposits cellulose in seed mucilage 390 pockets (Griffiths et al., 2015). Since cellulose distribution is disordered in iqd9 mucilage 391 (Fig. 3), we hypothesized that IQD9 influences CSC motility at the cell cortex. Consistent 392 with previous results (Griffiths et al., 2015), time-lapse images revealed GFP-CESA3 393 proteins moved in a unidirectional, clockwise manner around the cytoplasmic column of SCE 394 cells (Fig. 7). While this pattern was still present, the velocity of GFP-CESA3 particles 395 decreased from 135.9 ± 8.2 nm min -1 in WT cells to only 92.8 ± 17.3 nm min -1 in iqd9 (mean 396 ± SD; at least 170 measurements of 9 cells from 3 plants per genotype). Consistent with their 397 mucilage staining phenotypes (Fig. 1, Fig. S2 and Fig. S3), the movement of GFP-CESA3 398 was also reduced in trm4 cells, akin to iqd9 (Fig. 7). In these mutant seeds, CSC movement 399 appeared to be uncoupled from MTs, a behavior previously described for CESA proteins in 400 klcr (cmu) mutants (Liu et al., 2016), but could not be monitored in greater detail due to the 401 severe disruption of RFP-TUB6 localization in SCE cells ( Fig. 6; Yang et al., 2019). Taken 402 together, IQD9 is a novel protein required for seed mucilage biosynthesis by maintaining 403 cortical MT arrays and the speed of CESA movement, which influence cellulose distribution. In this study, we discovered that IQD9 and its interactor KLCR1 localize to cortical 427 arrays that resemble the circular paths of MTs (Fig. 6) (Fig. S1c), which is sufficient for MT localization in vivo (Sugiyama 434 et al., 2017). The highly immobile IQD9 proteins could function similarly to KLCR1/CMU1, 435 its binding partner (Fig. 4), to stabilize the cortical MT arrays of SCE cells and sustain CSC 436 speed during cellulose deposition. Consistent with this hypothesis, oryzalin treatment of SCE 437 cells severely disrupted the trajectory and velocity of GFP-CESA3 (Griffiths et al., 2015). 438 The reduced velocity of CESA3-containing CSCs in iqd9 and trm4 SCE cells (Fig. 7) shows 439 that multiple classes of proteins are required to shape the circular MT arrays and cellulose 440 distribution. and trm4 seeds released matrix polysaccharides with a composition that was similar to WT 466 (Fig. 2). Even though the iqd10-1 mutation partially reduced certain monosaccharides (Fig.  467 2), the mucilage released from these seeds showed a WT-like appearance (Figs. 1 and 3). Even though IQD9 and IQD10 promoters were active in both vegetative and 496 reproductive organs (Fig. S1), IQD9 was indispensable only for the organization of seed 497 mucilage polysaccharides (Figs. 1-3). While the expression of IQD9-GFP under its native 498 promoter was detected only in the general seed coat (Fig. 5 and Fig. S4), KLCR1-GFP was 499 expressed more ubiquitously in complemented lines. The transcription of IQD9, IQD10 and 500 IQD13 was previously associated with secondary cell wall biosynthesis (Mutwil et al., 2008). 501 Even though we detected transcriptional activity in the vasculature (Fig. S1), the absence of 502 IQD9 and/or IQD10 did not cause irx phenotypes found in stems with defective cellulose-503 hemicellulose networks ( The localization of soluble IQD9 and KLCR1/CMU1 (Fig. 5, Liu et al., 2016) proteins 516 near or at the PM suggests that they could interact with membrane-bound CSC components, 517 which travel in a spiral pattern during mucilage synthesis (Griffiths et al., 2015). TRM4 was 518 previously shown to maintain MT organization and directly bind CESA3 (Yang et al., 2019) 519 to enhance its mobility (Fig. 7). Although the mechanism that connects IQD9 and KLCR1 to 520 TRM4 requires further investigation, we provide the first evidence that members of these 521 three MT-associated families cooperate to direct cellulose deposition. Additional CSC-related 522 genes are expressed during mucilage production (Griffiths & North, 2017), but their putative 523 roles in mucilage biosynthesis remain to be investigated. Exploring the interactome of IQD9, 524 KLCR1, and TRM4 could reveal novel targets to fine-tune the biosynthesis of cellulose, the 525 most abundant renewable material on our planet. In addition to plant studies, the growing 526 arsenal of proteins found to influence cellulose biosynthesis could be rapidly expressed and 527 engineered in surrogate hosts . Yeast species such as Pichia pastoris have 528 already been used to express a Populus CESA capable of producing cellulose microfibrils in 529 vitro (Purushotham et al., 2016) and to identify essential protein co-factors for CESA-like 530 enzymes that catalyze hemicellulose elongation . Therefore, synthetic 531 biology advances combined with attractive plant models, such as the Arabidopsis SCE cells, 532 provide exciting avenues to refine the fibers that shape plants and many industrial products.