A formin-mediated cell wall- plasma membrane- cytoskeleton continuum is required for symbiotic infections in Medicago truncatula

Plant cell infections are tightly orchestrated by cell wall (CW) alterations, plasma membrane (PM) resident signalling processes and dynamic remodelling of the cytoskeleton. During root nodule symbiosis these processes result in morpho-dynamic responses including root hair swelling and curling, PM invagination and polar growth of a tubular infection structure, the infection thread (IT). However, the molecular details driving and guiding these PM remodelling events remain to be unravelled. Here, we studied a formin protein (SYFO1) in M. truncatula that is specifically induced during rhizobial infection. Phenotypical analysis of syfo1 mutants clearly indicates that the encoded protein is required for efficient rhizobial colonization of root hairs. SYFO1 itself creates a proteinaceous bridge between the CW and the polarized cytoskeleton. It binds to CW components via a proline-rich N-terminal segment, which is indispensable for its function. On the cytoplasmic side of the PM SYFO1 is associated with actin accumulations supporting the hypothesis that it contributes to cell polarization in vivo. This is further sustained by the fact that cell shape changes can be induced in a stimulus-dependent manner in root protoplasts expressing SYFO1. Taken together we provide evidence for the evolutionary re-wiring of a generic cytoskeleton modulator into a symbiosis-specific response.


INTRODUCTION 41
Legumes have the unique ability to symbiotically associate with rhizobia to maintain a 42 nitrogen-fixing mutualism. Intracellular colonization of Medicago truncatula roots by the 43 compatible rhizobium Sinorhizobium meliloti initiates from young, growing root hairs. In the 44 course of the interaction an organogenetic program is executed in the root cortex and the 45 pericycle that results in the development of nodules, in which symbiotic nitrogen fixation takes 46 place [1,2]. The first morphological step to establish this symbiosis comprises a rhizobial trap, 47 where a growing root hair engulfs the symbiont by physically curling around it [3,4]. 48 Phenomena such as root hair deformation and root hair branching, steps that precede bacterial 49 trapping in legumes [5,6], have been generally observed in plants in response to the 50 microtubule-stabilizing agent taxol [7], in mutants like the kinesin mrh2 [8] or upon over-51 expression of the formin protein AtFH8 [9] and the ROP GTPase RHO-OF PLANTS 2 [10]. 52 In contrast, incomplete curls were only observed at low frequency in Arabidopsis mutants 53 affected in the loci CEN1, CEN2, CEN3 [11] and in the scn1-1 mutant, where the corresponding 54 locus encodes the RhoGTPase GDP dissociation inhibitor SUPERCENTIPEDE1 [12]. This 55 indicates full root hair curling to represent a rather specific invention. The entrapment of the 56 symbiont completes with its full enclosure between root hair cell walls in a structure called the 57 'infection chamber' (IC) [3][4][5]13]. This is followed by exocytotic secretion of host-derived cell 58 wall loosening enzymes such as NODULE PECTATE LYASE [14] which, most likely, 59 subsequently allows the formation of a negatively curved plasma membrane (PM) structure 60 that further elongates into a tube-like channel, the 'infection thread' (IT) [3,4,15]. Parallel to 61 these morphological changes a set of transcription factors genetically re-programs host root 62 cells to allow transcellular IT progression and nodule organogenesis [16][17][18][19][20][21][22]. 63 Molecularly, initial root hair responses are triggered upon the recognition of bacterial 64 signalling molecules, called Nod Factors, by host LysM-type receptor-like kinases [23][24][25][26][27][28]. 65 8 rescued when generating independent transgenic roots expressing a ProSYFO1:  construct in these mutant backgrounds (Fig. 2D, F), demonstrating that the 162 insertion in the SYFO1 locus is causative for these phenotypes. 163 As it was previously demonstrated that the inoculation of Lotus root hairs with its symbiont M. 164 loti resulted in a strong polarization and bundling of actin filaments with a strong accumulation 165 of F-actin in the root hair tip [34,35], we tested whether this pattern is affected in syfo1 mutants. 166 In the absence of S. meliloti, longitudinal actin filaments were observed in young growing root 167 hairs within the infection zone of Medicago wild-type plants ( in almost 90% of all root systems of wild-type plants (Fig. S7E). However, this pattern was 171 strongly reduced in both syfo1 mutant alleles where only about 30% of all tested roots contained 172 root hairs responding with the above-mentioned pattern (Fig. S7E). 173 174 SYFO1 associates with polar actin assemblies under symbiotic conditions 175 Since our data indicated a symbiosis-specific role of SYFO1 in root hair polarization, we 176 investigated spatial and temporal dynamics of SYFO1 at subcellular resolution. In roots 177 expressing the ProSYFO1:SYFO1-GFP construct, we observed a weak homogenous signal of 178 the SYFO1 protein at the PM of root hairs in the absence of rhizobia (Fig. 3A). The underlying 179 low, basal expression was also detected by qRT-PCR (Fig. S3A) while it was most likely too 180 weak when using the nuclear-localized GFP reporter to test promoter activity (Fig. S1). 181 Interestingly, SYFO1 strongly accumulated in subapical and apical foci at root hair tips prior 182 to deformation at 2 dpi with S. meliloti (Fig. 3B, C), which strongly resembled actin patterns 183 observed upon Nod Factor application as reported earlier [34]. In root hairs that 184 morphologically responded by deformation (Fig. 3D) and curling (Fig. 3E), SYFO1 distributed 9 again along the PM with only mild accumulations at the apical region (Fig. 3D). SYFO1 also 186 resided along the infection thread membrane even though to a much weaker extent ( 3I-I''). To unambiguously verify that SYFO1 is the key driver of these protrusions we isolated 204 protoplasts from our syfo1-1 and syfo1L-1 mutants. No protrusions were found in the absence 205 of rhizobia in any of the used genotypes. Upon inoculation of protoplasts with S. meliloti, those 206 expressing wild-type SYFO1 (wild-type, syfo1L-1) or over-expressing SYFO1-GFP (OE) 207 developed protrusions while they were entirely absent on protoplasts isolated from the syfo1-1 208 mutant (Fig. 3J). This demonstrates that focal accumulation of SYFO1 can drive cell 209 polarisation and membrane deformations in a stimulus-dependent manner. 210

A SYFO1-mediated cell wall-plasma membrane-cytoskeleton continuum is required for 211 symbiotic responses in root hairs. 212
As actin binding of formins is generally mediated by the FH2 domain that is also present in the 213 cytosolic domain of SYFO1 ( Fig. 1A'), we examined the extracellular region, which is less 214 prominently found among formin proteins. bleached region and a mobile fraction of about 24% for wild-type SYFO1 whereas the mobile 228 fraction for SYFO1 ΔPRR was significantly higher (57%) (Fig. 4C-E). This clearly indicates that 229 the PRR segment within the SYFO1 ECD anchors this formin to the cell wall. To address 230 whether the cell wall association is required for SYFO1 function we conducted genetic 231 complementation experiments, we generated transgenic roots expressing SYFO1 ΔPRR under the 232 control of the native SYFO1 promoter in our syfo1-1 and syfo1-2 alleles. In contrast to the full-233 length SYFO1 (Fig. 2), the deletion of the PRR fully abolished the ability to complement the 234 root hair deformation phenotype of syfo1 mutants (Fig. 4F cell wall association and consequently restricts the lateral mobility also in SYFO1 (Fig. 4E). 257 Its ability to initiate membrane protrusions in cell wall-depleted protoplasts further suggests 258 that SYFO1 is involved in actin nucleation and filament elongation (Fig. 3G-G'; Fig. S8). In 259 line with non-symbiotic formins, such as AtFH4 from Arabidopsis that re-localizes to infection 260 sites of the powdery mildew fungus Blumeria graminis [48], we hypothesize that SYFO1 261 evolved to specifically mediate targeted secretion of cell wall constituents and other cargo 262 material to sustain symbiotic root hair responses including root hair curling and later stages of 263 infection. This entirely depends on the ability of SYFO1 to associate with the cell wall (Fig. 4)  264 where it maintains a cell wall-plasma membrane-cytoskeleton continuum that cannot be 265 functionally complemented by other, non-symbiotic formins that remain being expressed in 266 syfo1 mutants. 267 268 13

Plant growths and phenotypical analysis 270
For phenotypical analysis Medicago truncatula wild-type R108, syfo1-1, syfo1-2, syfo1L-1 and 271 syfo1L-2 seeds were scarified and sterilized before being sown on 1% agar plates for 272 germination and kept in darkness at 4°C for 3 days for vernalization. Germination was allowed 273 for up to 24 hours at 24°C before transferring the seedlings to plates containing Fahraeus  Signal peptide and transmembrane domains were predicted from proteins using signal v5.0 316 (10.1038/s41587-019-0036-z) and TMHMM v2.0c (PMID: 9783223G) respectively using 317 default parameters. 318 To look for relaxation (K<1) or intensification (K>1) of selection acting on different lineages 319 of interest in Eudicots (Table S3), we used the RELAX program (10.1093/molbev/msu400). 320 This method calculates different synonymous and non-synonymous substitution rates (ω = 321 N/ S) using the phylogenetic tree topology for both foreground and background branches. 322 Because (i) the programs used does not accept gaps in codon sequences and (ii) there is a 323 negative correlation between the number of sequences and the number of ungapped positions, 324 we used different numbers of input sequences for RELAX analysis. Protein sequences from 325 SYFO1 and SYFO1L orthologs were aligned using MUSCLE v3.8.382. Short sequences were 326 excluded to maximize sequence number while limiting gapped positions compared to SYFO1 327 and SYFO1L sequences of Medicago using a custom R script. We opted for 151 CDS 328 sequences corresponding to 1008 positions (Table S3). 329 330

Construct design 331
The constructs used in this study were designed using Golden Gate cloning [53]. 2.5 kb 332 upstream of the SYFO1 start codon were chosen as putative promoter region. A Golden Gate 333 compatible full-length genomic DNA version (Medtr5g036540.1) was synthesized 334 (GENEWIZ, Germany) by removing the BpiI and BsaI restriction sites via silent mutations. 335 All cloning primers are listed in Table S4. To select transgenic roots a pUbi:NLS-mCherry or 336 pUbi:NLS-2xCerulean cassette was additionally inserted into the different T-DNAs containing 337 the transgenes of choice as previously described [50]. Level II and level III constructs were 338 assembled based on the principle described earlier [53]. An overview about all designed 339 constructs is provided in Table S5.  340   341   342 16

Confocal Laser-Scanning Microscopy and FRAP 343
For imaging the NLS-GFP reporter module, sectioned nodules, protein localisation and 344 plasmolysis we used a Leica TCS SP8 confocal microscope equipped with a 20x HCX PL APO 345 water immersion lenses (Leica Microsystems, Mannheim, Germany). GFP was excited with a 346 White Light Laser (WLL) at 488 nm and the emission was detected at 500-550 nm. mCherry 347 fluorescence was excited using a WLL at 561nm and emission was detected between 575-630 348 nm. Samples, co-expressing two fluorophores were imaged in sequential mode between frames. 349