Multiple C2 domains and Transmembrane region Proteins (MCTPs) tether membranes at plasmodesmata

In eukaryotes, membrane contact sites (MCS) allow direct communication between organelles. Plants have evolved unique MCS, the plasmodesmata intercellular pores, which combine endoplasmic reticulum (ER) - plasma membrane (PM) contacts with regulation of cell-to-cell signalling. The molecular mechanism and function of membrane tethering within plasmodesmata remains unknown. Here we show that the Multiple C2 domains and Transmembrane region Protein (MCTP) family, key regulators of cell-to-cell signalling in plants, act as ER - PM tethers specifically at plasmodesmata. We report that MCTPs are core plasmodesmata proteins that insert into the ER via their transmembrane region whilst their C2 domains dock to the PM through interaction with anionic phospholipids. A mctp3/4 loss-of-function mutant induces plant developmental defects while MCTP4 expression in a yeast Δtether mutant partially restores ER-PM tethering. Our data suggest that MCTPs are unique membrane tethers controlling both ER-PM contacts and cell-cell signalling.

Intercellular communication is essential for the establishment of multicellularity, and 71! evolution gave rise to distinct mechanisms to facilitate this process. Plants have developed 72! singular cell junctions -the plasmodesmata-which span the cell wall and interconnect nearly 73! every single cell, establishing direct membrane and cytoplasmic continuity throughout the 74! plant body (Tilsner et al, 2016). Plasmodesmata are indispensable for plant life. They control 75! the flux of non-cell-autonomous signals such as transcription factors, small RNAs, hormones 76! and metabolites during key growth and developmental events (Gallagher et al, 2014;Tilsner 77! et al, 2016;Vatén et al, 2011;Carlsbecker et al, 2010;Benitez-Alfonso et al, 2013;Wu et al, 78! 2016;Han et al, 2014;Daum et al, 2014;Nakajima et al, 2001;Xu et al, 2011;Ross-elliott et 79! al, 2017). Over the past few years, plasmodesmata have emerged as key components of plant 80! defence signalling Wang et al, 2013;Lim et al, 2016). Mis-regulation of 81! plasmodesmata function can lead to severe defects in organ growth and tissue patterning but 82! also generate inappropriate responses to biotic and abiotic stresses (Wu et al, 2016;Wong et al, 83! 2016;Han et al, 2014;Sager & Lee, 2014;Caillaud et al, 2014;Faulkner et al, 2013).

95!
Where it enters the pores, the ER becomes constricted to a 15 nm tube (the desmotubule) 96! leaving little room for lumenal trafficking. According to current models, transfer of molecules 97! occurs in the cytoplasmic sleeve between the ER and the PM. Constriction of this gap, by the 98! deposition of callose, is assumed to be the main regulator of the pore size exclusion limit 99! (Vatén et al, 2011;Zavaliev et al, 2011). Recent work however, suggests a more complex 100! picture where plasmodesmal ER-PM gap is not directly related to the pore permeability and 101! may play additional roles (Nicolas et al, 2017a(Nicolas et al, , 2017b. Newly formed plasmodesmata (type 102! ! 4! I) exhibit such close contact (~2-3nm) between the PM and the ER, that no electron-lucent 103! cytoplasmic sleeve is observed (Nicolas et al, 2017a). During subsequent cell growth and 104! differentiation the pore widens, separating the two membranes, which remain connected by 105! visible electron-dense spokes, leaving a cytosolic gap (type II). This transition has been 106! proposed to be controlled by protein-tethers acting at the ER-PM interface 107! Nicolas et al, 2017b). Counterintuitively, type I plasmodesmata with no apparent cytoplasmic 108! sleeve are open to macromolecular trafficking and recent data indicate that tight ER-PM 109! contacts may in fact favour transfer of molecules from cell-to-cell (Nicolas et al, 2017a).

110!
The close proximity of the PM and ER within the pores, and the presence of tethers qualifies 111! plasmodesmata as a specialised type of ER-to-PM membrane contact site (MCS) (Tilsner et 112! al, 2016;Bayer et al, 2017). MCS are structures found in all eukaryotic cells which function 113! in direct inter-organellar signalling by promoting fast, non-vesicular transfer of molecules and 114! allowing collaborative action between the two membranes (Burgoyne et al, 2015;Prinz, 2014;115! Phillips & Voeltz, 2016;Gallo et al, 2016;Eden et al, 2010Eden et al, , 2016Ho et al, 2016;Chang et 116! al, 2013;Kim et al, 2015;Petkovic et al, 2014;Zhang et al, 2005;Omnus et al, 2016). In 117! yeast and mammalian, MCS protein tethers are known to physically bridge the two 118! organelles, to control the intermembrane gap and participate in organelle cross-talk. Their 119! molecular identity/specificity dictate structural and functional singularity to different types of 120! MCS (Eisenberg-Bord et al, 2016;Henne et al, 2015). To date, the plasmodesmal membrane 121! tethers remain unidentified, but by analogy to other types of MCS it seems likely that they 122! play important roles in plasmodesmal structure and function, and given their unique position 123! within a cell-to-cell junction may link intra-and intercellular communication.

124!
Here, we have reduced the complexity of the previously published Arabidopsis plasmodesmal 125! proteome (Fernandez-Calvino et al, 2011) through the combination of a refined purification 126! protocol (Faulkner & Bayer, 2017) and semi-quantitative proteomics, to identify ~120 127! proteins highly enriched in plasmodesmata, and identify tether candidates. Amongst the most 128! abundant plasmodesmal proteins, members of the Multiple C2 domains and Transmembrane 129! region Proteins (MCTPs) were enriched in post-cytokinetic plasmodesmata with tight ER-PM 130! gap compared to mature plasmodesmata with wider gap and sparse spokes, and exhibit the 131! domain architecture characteristic of membrane tethers, with multiple lipid-binding C2 132! domains in the N-terminal, and multiple transmembrane domains in the C-terminal region.

133!
Using live cell imaging, molecular dynamics, and yeast complementation, we show that 134! MCTP properties are consistent with a role in ER-PM membrane tethering at plasmodesmata.

Identification of plasmodesmal ER-PM tethering candidates 143!
To identify putative plasmodesmal MCS tethers, we decided to screen the plasmodesmata 144! proteome for ER-associated proteins (a general trait of ER-PM tethers (Henne et al, 2015;145! Eisenberg-Bord et al, 2016)) with structural features enabling bridging across two 146! membranes. Published plasmodesmata proteome reported the identification of more than 1400 147! proteins in Arabidopsis (Fernandez-Calvino et al, 2011), making the discrimination of true 148! plasmodesmata-associated from contaminant proteins a major challenge. To reduce the 149! proteome complexity and identify core plasmodesmata proteins, we used a refined 150! plasmodesmata purification technique (Faulkner & Bayer, 2017) (Nikolovski et al, 2012;Dunkley et al, 160! 2006) and used as a basis for selecting MCS-relevant candidates.

161!
Alongside, we also analysed changes in protein abundance during the ER-PM tethering 162! transition from very tight contacts in post-cytokinetic plasmodesmata (type I) to larger ER-163! PM gap and sparse tethers in mature plasmodesmata (type II) (Nicolas et al, 2017a). For this 164! we obtained a similar semi-quantitative proteome from four and seven-day old culture cells, 165! enabling a comparison of plasmodesmata composition during the tethering transition (Nicolas 166! et al, 2017a) (Supplementary Fig. 2).

173!
Amongst the 47 plasmodesmal proteins differentially enriched, all MCTPs were more 174! In summary we concluded that whatever the tissue and organ considered, AtMCTP4 is 276! strongly and consistently associated with plasmodesmata but also presents a steady 277! right panels and c), quantitatively confirming the loss of plasmodesmata association when the 293! C2 modules were deleted. We therefore concluded that, similar to the HsE-Syt and AtSYT 294! ER-PM tether families (Giordano et al, 2013;Levy et al, 2015;Pérez-Sancho et al, 2015b) (Zimmermann et al, 2018) for remote homology detection. The 309! ! 11! searches yielded a total of 1790 sequence matches, which contained almost all human and A. 310! thaliana C2 domains. We next clustered the obtained sequences based on their all-against-all 311! pairwise similarities in CLANS (Frickey & Lupas, 2018 (Giordano et al, 2013;Saheki et al, 2016;Pérez-Lara et al, 2016;Abdullah et 318! al, 2014;Marty et al, 2014). By comparison to the C2 domains of these membrane-trafficking 319! and -tethering proteins, the C2 domains of most other proteins do not make any connections 320! to the C2 domains of AtMCTPs at the P-value cut-off chosen for clustering (1e-10). Thus, 321! based on sequence similarity, the plant AtMCTP C2 domains are expected to bind 322! membranes.

323!
We next asked whether the C2 modules of MCTPs are sufficient for PM association in vivo.

324!
Fluorescent protein fusions of the C2A-D or C2B-D modules without the TMR were 325! generated for NbMCTP7, AtMCTP3, AtMCTP4, AtMCTP6, AtMCTP9 as well as 326! AtMCTP1/FTIP and AtMCTP15/QKY and expressed in N. benthamiana. We observed a 327! wide range of sub-cellular localisations from cytosolic to PM-associated and in all cases 328! plasmodesmata association was lost .

329!
To further investigate the potential for MCTP C2 domains to interact with membranes, we 330! employed molecular dynamics modelling. We focussed on AtMCTP4, as a major 331! plasmodesmal constituent and whose loss-of-function in conjunction with AtMCTP3, induces 332! severe plant development defects (Liu et al, 2018) (Fig. 3 showed that all individual C2 domains of AtMCTP4 can interact with lipids and dock on the 341! membrane surface when a "PM-like" lipid composition was used (Fig. 5a). The PC-only 342! membrane showed only weak interactions, whilst the PC:PS membrane allowed only partial 343! ! 12! docking (Fig. 5a). Docking of AtMCTP4 C2 domains arose mainly through electrostatic 344! interactions between lipid polar heads and basic amino acid residues at the protein surface.

345!
We further tested two other MCTP members, namely AtMCTP15/QKY and NbMCTP7, 346! which possess four rather than three C2 domains. We found that similar to AtMCTP4, the 347! individual C2 domains of AtMCTP15/QKY and NbMCTP7 exhibited membrane interaction 348! in the presence of the negatively charged lipids ( Supplementary Fig. 8).

349!
Our molecular dynamics data thus suggests that membrane docking of the AtMCTP4 C2 350! domains depends on the electrostatic charge of the membrane and more specifically on the 351! presence of PI4P, a negatively-charged lipid which has been reported as controlling the 352! electrostatic field of the PM in plants (Simon et al, 2016).

353!
To confirm the importance of PI4P for MCTP membrane interactions and thus, potentially 354! subcellular localisation, we used a short-term treatment with phenylarsine oxide (PAO), an 355! inhibitor of PI4-kinases (Simon et al, 2016). We focused on Arabidopsis root tips where 356! effects of PAO have been thoroughly characterised (Simon et al, 2016). In control-treated  (Fig. 5b). This control not only demonstrates that the PAO treatment was successful, 368! but also highlights that the majority of PI4P was normally found at the PM, rather than the 369! to the PM, we used a yeast Δtether mutant line deleted in six ER-PM tethering proteins 380! resulting in the separation of the cortical ER (cER) from the PM (Manford et al, 2012) and 381! expressed untagged AtMCTP4. To monitor recovery in cortical ER, and hence, ER-PM 382! contacts, upon AtMCTP4 expression, we used Sec63-RFP (Metzger et al, 2008) as an ER 383! marker combined with confocal microscopy. In wild-type cells, the ER was organised into 384! nuclear (nER) and cER. The cER was visible as a thread of fluorescence along the cell 385! periphery, covering a large proportion of the cell circumference ( Fig. 6a white arrows). By 386! contrast and as previously reported (Manford et al, 2012), we observed a substantial reduction 387! of cER in the Δtether mutant, with large areas of the cell periphery showing virtually no 388! associated Sec63-RFP (Fig. 6a). When AtMCTP4 was expressed into the Δtether mutant line, 389! we observed partial recovery of cER, visible as small regions of Sec63-RFP closely apposed 390! to the cell cortex. We further quantified the extent of cER in the different lines by measuring 391! the ratio of the length of cER (Sec63-RFP) against the cell perimeter (through calcofluor wall 392! staining) and confirmed that ATMCTP4 expression induced an increase of cER from 7.3 % to 393! 23.1% when compared to the Δtether mutant (Fig. 6b). This partial complementation is 394! similar to results obtained with yeast deletion mutants containing only a single endogenous 395! ER-PM tether, IST2, or all three isoforms of the tricalbin (yeast homologs of HsE-Syts) 396! (Manford et al, 2012), supporting a role of AtMCTP4 as ER-PM tether. 397!

399!
In plants, communication between cells is facilitated and regulated by plasmodesmata, ~50 400! nm diameter pores that span the cell wall and provide cell-to-cell continuity of three different 401! organelles: the PM, cytoplasm, and ER. The intercellular continuity of the ER and the 402! resulting architecture of the pores make them unique amongst eukaryotic cellular junctions, 403! and qualify plasmodesmata as a specialised type of ER-PM MCS Tilsner 404! et al, 2016). Like other types of MCS, the membranes within plasmodesmata are physically 405! connected but so far the molecular components and function of the ER-PM tethering 406! machinery remain an enigma.

407!
Here, we provide evidence that members of the MCTP family, some of which have been 408! described as key regulators of intercellular trafficking and cell-to-cell signalling (Vaddepalli 409! et al, 2014;Liu et al, 2018Liu et al, , 2012, also act as ER-PM tethers inside the plasmodesmata pores. Whilst several MCTPs have previously been characterised as regulators of cell-to-cell 413! trafficking or signalling (Liu et al, 2012(Liu et al, , 2018Vaddepalli et al, 2014;Liu et al, 2017), only 414! some have also been localised to plasmodesmata, whilst other studies reported alternative 415! localisations which include PM, ER, Golgi, endosomes and cytoplasm (Trehin et al, 2013;416! Liu et al, 2017416! Liu et al, , 2018416! Liu et al, , 2012Kraner et al, 2017;Vaddepalli et al, 2014), perhaps depending on 417! the isoform, orientation of fluorescent protein fusions and expression context. Here, we have 418! identified several MCTPs (6-10 out of 16) as belonging to the most abundant proteins at 419! plasmodesmata through both in vivo and proteomic data. This includes AtMCTP3 and 420! AtMCTP4 recently identified as modulators of SHOOTMERISTEMLESS trafficking (Liu et 421! al, 2018), for which we find that a Atmctp3/Atmctp4 loss-of-function mutant displays a severe 422! developmental phenotype, including defects in the root QC, that agrees with the findings of 423! Liu et al, 2018. Whereas Liu et al. (2018 observed endosomal-localisation of AtMCTP3 and 424! AtMCTP4, our data suggest they are primarily plasmodesmata-associated. We therefore 425! propose that MCTPs are core plasmodesmata-constituents and that AtMCTP3 and AtMCTP4 426! may possibly regulate the transport of SHOOTMERISTEMLESS, at the pores. 427!

MCTPs as plasmodesmata-specific ER-PM tethers 429!
While ER-PM contacts within plasmodesmata have been observed for decades (Ding et al, 430! 1992;Tilsner et al, 2011;Tilney et al, 1991;Nicolas et al, 2017b), the molecular identity of 431! the tethers has remained elusive. Here we propose that MCTPs are prime plasmodesmal 432! membrane tethering candidates as they possess all required features: 1) strong association 433! with plasmodesmata; 2) structural similarity to known ER-PM tethers such as HsE-Syts and 434! AtSYTs (Levy et al, 2015;Pérez-Sancho et al, 2015b;Giordano et al, 2013) possessing an 435! ER-inserted TMR at one end and multiple lipid-binding C2 domains at the other for PM 436! docking; 3) the ability to partially restore ER-PM tethering in a yeast Δtether mutant.

437!
Similarly to other ER-PM tethers (Eisenberg-Bord et al, 2016;Wong et al, 2016;Henne et al, 438! 2015;Giordano et al, 2013), MCTP C2 domains dock to the PM through electrostatic 439! interaction with anionic lipids, especially PI4P and to a lesser extent PS. In contrast with 440! animal cells, PI4P is found predominantly at the PM in plant cells and defines its electrostatic 441! signature (Simon et al, 2016). Although plasmodesmata are MCS, they are also structurally 442! unique: both the ER and the PM display extreme, and opposing membrane curvature inside 443! the pores; the ER tubule is linked to the PM on all sides; and the membrane apposition is 444! unusually close (2-3 nm in type I post-cytokinetic pores (Nicolas et al, 2017a)). So while 445! ! 15! structurally related to known tethers, MCTPs are also expected to present singular properties.

446!
For instance, similar to the human MCTP2, MCTPs could favour ER membrane curvature 447! through their TMR (Joshi et al, 2017). Plasmodesmata also constitute a very confined 448! environment, which together with the strong negative curvature of the PM, may require the 449! properties of MCTP C2 domains to differ from that of HsE-Syts or AtSYTs. All of these 450! aspects will need to be investigated in the future. 451!

Inter-organellar signalling at the plasmodesmal MCS? 453!
In yeast and animals, MCS have been shown to be privileged sites for inter-organelle 454! signalling by promoting fast, non-vesicular transfer of molecules such as lipids (Gallo et al, 455! 2016;Saheki et al, 2016;Wong et al, 2016). Unlike the structurally analogous tethering 456! proteins AtSYTs and HsE-Syts, MCTPs do not harbour known lipid-binding domains that 457! would suggest that they participate directly in lipid transfer between membranes. However,

Combining organelle tethering and cell-to-cell signalling functions 471!
Several members of the MCTP family have previously been implicated in regulating either 472! macromolecular trafficking or intercellular signalling through plasmodesmata. 473! AtMCTP1/FTIP interacts with, and is required for phloem entry of the Flowering Locus T 474! (FT) protein, triggering transition to flowering at the shoot apical meristem (Liu et al, 2012).

481!
Whilst the mechanisms by which these MCTP proteins regulate intercellular 482! transport/signalling have not been elucidated, MCTP physical interaction with mobile factors 483! or receptor is critical for proper function (Vaddepalli et al, 2014;Liu et al, 2017Liu et al, , 2018Liu et al, , 2012.

484!
In AtMCTP1/FTIP, the interaction is mediated by the C2 domain closest to the TMR (Liu et 485! al, 2017). For the C2 domains of HsE-Syts, conditional membrane docking is critical for their 486! function and depends on intramolecular interactions, cytosolic Ca 2+ and the presence of

666!
In Arabidopsis cultured cells, transition from type I to type II plasmodesmata is associated 667! with a change in ER-PM contact site architecture, from very tight contact (~3 nm) with no 668! visible cytoplasmic sleeve (type I) to larger ER-PM distance (10 nm to more) with an electron 669! lucent cytosolic sleeve and sparse spoke-like elements (type II) (Nicolas et al, 2017a). We 670! analysed the plasmodesmata proteome from four days old cultured cells where type I 671! plasmodesmata represent 70% of the total plasmodesmata population and at seven days where 672! this proportion is reversed and type II become predominant (Nicolas et al, 2017a (Kraner et al, 2017) water at a final OD 600 of 0.3 for individual constructs, 0.2 each for the combination of two.

950!
The ectopic silencing suppressor 19k was co-infiltrated at an OD 600 of 0.15. Agroinfiltrated N. 951! benthamiana leaves were imaged 3-4 days post infiltration at room temperature. ~ 2 by 2 cm 952! leaf pieces were removed from plants and mounted with the lower epidermis facing up onto 953! glass microscope slides.

954!
Transgenic Arabidopsis plants were grown as described above. For primary roots, lateral roots 955! and hypocotyl imaging, six to seven days old seedlings or leaves of 5-8 leaf stage rosette 956! plants were mounted onto microscope slides. For shoot apical meristem imaging, the plants 957! were first dissected under a binocular then transferred to solid MS media and immediately 958! observed using a water-immersion long-distance working 40X water immersion objective.

968!
For quantification of NbMCTP7 co-localisation with VAP27.1, SYT1 and PDCB1, co-969! expression of the different constructs was done in N. benthamiana. An object based method 970! was used for colocalization quantification (Bolte & Cordelières, 2006) . Images from different 971! conditions are all acquired with same parameters (zoom, gain, laser intensity etc.) and 972! ! 49! channels are acquired sequentially. These images are processed and filtered using ImageJ 973! software (https://imagej.nih.gov/ij/) in order to bring out the foci of the pictures. These foci 974! were then automatically segmented by thresholding and the segmented points from the two 975! channels were assessed for colocalization using the ImageJ plugin Just Another 976! Colocalization Plugin (JACoP) (Bolte & Cordelières, 2006). This whole process was 977! automatized using a macro (available upon demand).

980!
Brightness and contrast were adjusted on ImageJ software (https://imagej.nih.gov/ij/). or Arabidopsis) were acquired by sequential scanning of mCHERRY-PDCB, PDLP1-mRFP 989! or aniline blue (plasmodesmata markers) in channel 1 and GFP/YFP-tagged MCTPs in 990! channel 2 (for confocal setting see above). About thirty images of leaf epidermis cells were 991! acquired with a minimum of three biological replicates. Individual images were then 992! processed using ImageJ by defining five regions of interest (ROI) at plasmodesmata (using 993! plasmodesmata marker to define the ROI in channel1) and five ROIs outside plasmodesmata.

994!
The ROI size and imaging condition were kept the same. The GFP/YFP-tagged MCTP mean 995! intensity (channel 2) was measured for each ROI then averaged for single image. The 996! plasmodesmata index corresponds to intensity ratio between fluorescence intensity of MCTPs 997! at plasmodesmata versus outside the pores. For the plasmodesmata-index of RFP-HDEL, 998! PDLP1-RFP and mCHERRY-PDCB1 we used aniline to indicate pitfields. R software was 999! used for making the box plots and statistics.