ROP INTERACTIVE PARTNER b interacts with RACB and supports fungal penetration into barley epidermal cells

RHO of Plants (ROP) G-proteins are key components of cell polarization processes in plant development. The barley (Hordeum vulgare) ROP protein RACB, is a susceptibility factor in the interaction of barley with the barley powdery mildew fungus Blumeria graminis f.sp. hordei (Bgh). RACB also drives polar cell development, and this function might be coopted during formation of fungal haustoria in barley epidermal cells. In order to understand RACB signaling during the interaction of barley with Bgh, we searched for potential downstream interactors of RACB. Here, we show that ROP INTERACTIVE PARTNER b (RIPb, synonym: INTERACTOR OF CONSTITUTIVE ACTIVE ROP b; ICRb) directly interacts with RACB in yeast and in planta. Over-expression of RIPb supports susceptibility of barley to Bgh. RIPb further interacts with itself at microtubules. However, the interaction with activated RACB takes place at the plasma membrane. Both, RIPb and RACB are recruited to the site of fungal attack around the neck of developing haustoria suggesting locally enhanced ROP activity. We further assigned different functions to different domains of the RIPb protein. The Nterminal coiled-coil CC1 domain is required for microtubule localization, while the C-terminal coiled-coil CC2 domain is sufficient to interact with RACB and to fulfill a function in susceptibility at the plasma membrane. Hence, RIPb appears to be localized at microtubules and is then recruited by activated RACB for a function at the plasma membrane during formation of the haustorial complex.


Introduction
The interaction of plants with powdery mildew fungi is a model for the biology of cell-autonomous responses to fungal parasites (Dörmann et al., 2014).
The powdery mildew fungus Blumeria graminis f.sp. hordei (Bgh) is a biotrophic ascomycete specifically adapted to barley (Hordeum vulgare) and grows on the plant´s surface. In the beginning of its life cycle Bgh has to penetrate an epidermal cell in order to establish a haustorium for nutrient uptake (Hahn et al., 1997;Voegele et al., 2001) and to provide a surface for the translocation of virulence effector proteins into the host cell (Catanzariti et al., 2007). During all stages of fungal invasion, the epidermal host cell stays intact. Host cytosol and fungal haustorium are separated by the extrahaustorial matrix and the plant-derived extrahaustorial membrane (EHM).
Plant host cells polarize in very early phases of the interaction with fungi. A reorganization of the cytoskeleton was shown in different pathosystems, as well as the accumulation of peroxisomes, mitochondria, Golgi bodies and ER at the site of pathogen attack (Kobayashi et al., 1997;Takemoto et al., 2003;Koh et al., 2005;Takemoto et al., 2006;Fuchs et al., 2016). This is accompanied by relocation of the nucleus to the site of attack (Gross et al., 1993;Scheler et al., 2016). Polarization is considered important for effective defense, in particular for the focal formation of papilla or cell wall appositions, which requires localized deposition of callose, other cell wall glucans and phenolic compounds at the attempted penetration site (McLusky et al., 1999;Hückelhoven, 2007;Chowdhury et al., 2014).
However, it is reasonable to assume, that host cell polarization is also important for successful pathogen establishment, for instance for the generation of the EHM (Scheler et al., 2016;Kwaaitaal et al., 2017).
ROP GTPases (RHO of Plants, also called RAC for rat sarcoma-related C3 botulinum toxin substrate) are small monomeric G-proteins that form a plantspecific RHO subfamily. ROPs can cycle between an actively signaling GTPbound state and an inactive GDP-bound state and are crucial for polarity of diverse types of plant cells (Feiguelman et al., 2018). While activation is mediated by Guanosin Nucleotide Exchange Factors (GEF) enabling the exchange of GDP to GTP, inactivation is facilitated by GTPase Activating Proteins (GAP) which activate the intrinsic GTPase function of the Gprotein, leading to GTP hydrolysis. ROPs seem to fulfill different functions depending on particular downstream factors called ROP-effectors. For instance Arabidopsis thaliana ROP2 suppresses light induced stomata opening by interacting with ROP Interactive CRIB Motif Containing Protein7 (RIC7), which in turn interacts and inhibits the exocyst vesicle tethering complex subunit Exo70B1 (Hong et al., 2015). ROP2 is additionally involved in pavement cell lobe interdigitation by interacting with RIC4 for actin assembly in lobes and at the same time inhibiting RIC1 which organizes microtubules together with katanin KTN1 downstream of ROP6 (Fu et al., 2005;Lin et al., 2013). In these pathways, RIC proteins are considered scaffolds for connecting activated ROPs with downstream effector proteins in G-protein signaling.
Another class of downstream interactors are ROP Interactive Partners (RIPs) first called Interactors of Constitutive Active ROPs (ICRs) (Lavy et al., 2007;Li et al., 2008). ICRs/RIPs represent a second class of plantspecific proteins connecting ROP signaling to downstream effectors. So far, little is known about these proteins. Arabidopsis knockout plants of RIP1/ICR1 have defects in pavement cell development, root hair development as well as root meristem maintenance showing an involvement of RIP1/ICR1 in different developmental processes. RIP1/ICR1 seems to be able to interact with different ROP proteins and was found to interact downstream with SEC3a of the exocyst complex and thereby possibly controlling the localization of the auxin transporter PIN1 (Lavy et al., 2007;Hazak et al., 2010;Hazak et al., 2014). Additionally it was reported, that RIP1 acts in pollen tube formation where it interacts with ROP1 at the plasma membrane of the pollen tube tip (Li et al., 2008). RIP3 (also called ICR5 or MIDD1 for Microtubule Depletion Domain1) plays a key role in xylem cell development in Arabidopsis. During the formation of the secondary cell wall in progenitor cells, RIP3 interacts with ROP11 and the kinesin KIN13A, which leads to local microtubule depletion and the formation of secondary wall pits (Mucha et al., 2010;Oda et al., 2010;Fukuda, 2012, 2013).
The barley ROP protein RACB is involved in root hair outgrowth and controls asymmetric cell division of subsidiary cells in stomata development (Scheler et al., 2016). RACB and RACB-associated proteins influence arrays and stability of filamentous actin and the microtubule cytoskeleton (Opalski et al., 2005;Hoefle et al., 2011;Huesmann et al., 2012). Next to its function in polar cell development, RACB is also a susceptibility factor in the interaction with the powdery mildew fungus Bgh. Over-expression of constitutively activated RACB (CA RACB) enhances the penetration success of Bgh into barley epidermal cells, silencing of RACB leads to a decreased penetration rate (Schultheiss et al., 2002;Schultheiss et al., 2003;Hoefle et al., 2011).
RACB's function in susceptibility seems not to be dependent on defense suppression, but rather on the exploitation of developmental signaling mechanisms of the host (Scheler et al., 2016). A Bgh effector candidate, ROP-Interactive Peptide1 (ROPIP1), binds directly to activated RACB.
Expression of ROPIP1 in barley cells negatively influences microtubule stability and leads to an increased penetration rate of Bgh into barley epidermal cells (Nottensteiner et al., 2018). RACB further interacts with the class VI receptor-like cytoplasmic kinase ROP-Binding Kinase1 (RBK1).
Activated RACB supports in vitro kinase activity of RBK1, but RBK1 acts in resistance rather than susceptibility. This seems to be explained by the interaction of RBK1 with S-Phase Kinase1-Associated (SKP1)-Like Protein (SKP1-like), which is part of an E3-ubiquitin ligase complex and both RBK1 and SKP1-like can limit the abundance of the RACB protein (Huesmann et al., 2012;Reiner et al., 2015). Another interactor of RACB is the Microtubule-Associated ROP GTPase Activating Protein1 (MAGAP1), a CRIB-motif containing ROP-GAP. MAGAP1 and RACB recruit each other to the cell periphery and to the microtubule cytoskeleton, and MAGAP1 apparently counters the susceptibility effect of RACB, while silencing of MAGAP1 leads to increased susceptibility to Bgh (Hoefle et al., 2011).
In this study, we identified barley RIPb as another downstream interactor of RACB. We investigated the effect of RIPb on susceptibility by transient overexpression and RNAi knockdown of RIPb in single epidermal cells, and the interaction between RIPb and RACB by Yeast-Two-Hybrid assays and ratiometric bimolecular fluorescence complementation (BiFC). RIPb and RACB co-localize and presumably interact at the plasma membrane, at the microtubule cytoskeleton, and at the site of fungal invasion. To further investigate the structure-function relationship of RIPb, we tested a series of

RIPb truncations regarding their function in the interaction of barley with
Bgh and their role for protein-protein interaction.

Identification of ICR/RIP proteins in barley
Previous studies have shown that ICR/RIP proteins are a class of proteins with little sequence similarity (Li et al., 2008). All ICR/RIP proteins identified so far in Arabidopsis contain an N-terminal QEEL motif and a C-terminal QWRKAA motif. These motifs are present in respective N-and C-terminal coiled-coil domains. Based on this, we performed bioinformatic analyses and identified three high confidence genes coding for ICR/RIP proteins in barley (Supplemental Fig. S1). It appears that in several grasses the first glutamic acid in the QEEL motif is exchanged to aspartic acid (QDEL).
Because we did not observe a clear orthology to individually numbered Arabidopsis thaliana ICR/RIP proteins and phylogenetic analysis was ambiguous as well, we named the barley proteins RIPa/ICRa (HORVU3Hr1G087430), RIPb/ICRb (HORVU1Hr1G012460) and RIPc/ICRc (HORVU3Hr1G072880) (Supplemental Fig. S2). We also identified three ICR/RIP proteins in rice containing the QDEL motif as well as the QWRKAA motif (Os01g61760, Os05g03120 and OsJ_03509 (Yu et al., 2005)).
Alignments of the barley ICRs/RIPs with the ICR/RIP proteins from rice and the five ICR/RIP proteins previously identified in Arabidopsis (ICR1/RIP1 (At1g17140), ICR2/RIP2 (At2g37080), ICR3/RIP5 (At5g60210), ICR4/RIP4 (At1g78430) and ICR5/RIP3/MIDD1 (At3g53350)) show little overall amino acid sequence conservation between the grasses and Arabidopsis, except for the conserved QD/EEL motif at the more N-terminal part and the QWRKAA motif at the more C-terminal part of the protein. The latter was shown to be necessary for ROP interaction (Lavy et al., 2007). The alignment also shows conservation of several lysine residues at the very Cterminus, which were shown before to be important for membrane localization of other ICR/RIP proteins (Li et al., 2008) S1).
Phylogenetic analysis shows that HvICRa/HvRIPa and HvICRb/HvRIPb are more closely related to each other, than to HvICRc/ HvRIPc, which is located on an independent branch of the tree (Supplemental Fig. S2). Each one ICR/RIP from rice (Oryza sativa ssp. japonica) and Brachypodium distachyon appear to be orthologous to HvICRa/HvRIPa, HvICRb/HvRIPb and HvICRc/HvRIPc respectively. (Supplemental Fig. S2).

RIPb influences susceptibility of barley to Bgh
Semiquantitative reverse transcription PCR showed that all three barley

RIPb interacts with RACB
In order to determine the subcellular localization of RIPb, we transiently Recorded images further confirmed that YFP-RIPb is also present in the cytosol, where it co-localized with additionally co-expressed soluble CFP, and at the cell periphery or plasma membrane (Fig. 2). Co-expression with constitutively activated RACB-G15V (CA RACB) resulted in depleted cytosolic localization of YFP-RIPb, when compared to soluble red fluorescing mCherry. At the same time, microtubule localization and cellperipheral localization was still clearly detectable. Cytoplasmic depletion of YFP-RIPb was not observed when we co-expressed dominant negative RACB-T20N (DN RACB) (Fig. 3A). This change in RIPb localization might be best explained if RACB recruits RIPb to the cell periphery/plasma membrane. To test this, we co-expressed YFP-RIPb with the plasma membrane marker pm-rk (Nelson et al., 2007;Weis et al., 2013), either alone or in presence of CA RACB (Supplemental Fig. S5). YFP-RIPb alone showed some overlapping signal with pm-rk, but the peak in the signal profile was slightly displaced due to additional cytosolic signal (Supplemental Fig. S5A).
However, in presence of CA RACB we recorded a shift of the YFP-RIPb peak towards the peak of the plasma membrane marker (Supplemental Fig. S5B).
This supports that RIPb gets recruited to the plasma membrane by activated RACB, but also shows that RIPb itself localized close to or attached to the plasma membrane. Ratiometric Bimolecular Fluorescence Complementation (BiFC) experiments further supported the interaction of RIPb with RACB.
YFP fluorescence was reconstituted when nYFP-RIPb and cYFP-CA RACB were co-expressed in leaf epidermal cells (Fig. 3B, C). By contrast, coexpression of nYFP-RIPb and cYFP-DN RACB did not result in clear BiFC and the strength in signals were on average less than 10% of the signals recorded for the interaction with CA RACB (Fig. 3B, C). We observed the complemented CA RACB-RIPb YFP complex signals either exclusively at the plasma membrane or at cortical microtubules and the plasma membrane but hardly in the cytosol, when compared to mCherry (Fig. 3B). We further confirmed a direct interaction between both wild type RACB (RACB WT) and CA RACB with RIPb ( Fig. 3D), respectively, in yeast. These experiments together suggest a direct interaction between RIPb and RACB in planta.  (Lavy et al., 2007;Mucha et al., 2010). RIPbCC2 was also able to interact with CA RACB but not with DN RACB in BiFC assays and this interaction took place at the plasma membrane (Supplemental Fig. S7). To confirm this, we again co-expressed a YFP fusion of RIPbCC2 with the plasma membrane marker pm-rk either alone or in presence of CA RACB (Supplemental Fig. S5).

RIPb truncations show distinct subcellular localization and function
Similar to full length RIPb, YFP-RIPbCC2 showed co-localization with pmrk but the peak was slightly shifted in the signal profile probably due to additional cytosolic signal (Supplemental Fig. S5C). Co-expression with CA RACB again shifted the signal towards the plasma membrane, resulting in an overlay of the two peaks in the signal profile (Supplemental Fig. S5D).
RIPb was also able to interact with itself in yeast. The Va-region may be important for this, since only full length RIPb and truncations containing this region were able to interact in yeast (Fig. 4C). In order to look for specific subcellular localizations in planta, we created YFP-tagged protein fusions of these truncations. Like YFP-RIPbCC2, YFP-RIPbVaCC2 localized strongly to the cell periphery, presumably the plasma membrane, with weak cytosolic background (Fig. 4D). YFP-RIPbCC1Va was located in the cytosol and at the microtubules. Additionally, co-expression experiments with YFP-RIPbCC1Va and CA RACB show that YFP-RIPbCC1Va was not positioned at the cell periphery despite the presence of CA RACB (Supplemental Fig.   S6), suggesting that the CC2 domain is necessary for the recruitment by CA RACB. However, YFP-RIPbCC1 and YFP-RIPbVa were exclusively detected in the cytosol (Fig. 4D). Hence, both the CC1 domain and the Va domain appear to be required but not sufficient for microtubule association. Double mutation of D85 and E86 of the QDEL motif did not lead to a loss of microtubule localization (Supplemental Fig. S6B). The QDEL motif itself might therefore not be necessary for microtubule localization. Since the Va domain was required for dimerization and microtubule association, RIPb might localize to the microtubules as a dimer or oligomer. This was further supported because BiFC-signals recorded after co-expression of nYFP-RIPb with cYFP-RIPb occur exclusively at the microtubules and show less cytosolic background (Fig. 5A), when compared to YFP-RIPb alone, which may be detected both in its monomeric and its dimeric/oligomeric form (Fig.   2). Quantification of reconstituted YFP fluorescence showed significantly stronger signal intensities for the nYFP-RIPb/cYFP-RIPb homodimer compared to the co-expression of MAGAP1-nYFP with cYFP-RIPb, which are not supposed to interact and used as a negative control. As a positive control for the negative control, we co-expressed MAGAP1-nYFP with cYFP-CA RACB, which again showed high YFP complementation signals (Fig. 5A,   B). Signal quantification showed high signal overlap between the complemented YFP fluorescence signal and microtubule marker RFP-MAGAP1-Cter over a linear region of interest (Fig. 5C, D). Since there appeared to be little cytosolic background in the nYFP-RIPb/cYFP-RIPb BiFC images, we measured the ratio between microtubule and cytosolic signal within each cell. We then compared signal ratios within YFP-RIPb expressing cells and those expressing the nYFP-RIPb/cYFP-RIPb BiFC pair.
As a positive control for microtubule localization, we also measured the signal ratio for the microtubule marker RFP-MAGAP1-Cter. The data show that the microtubule/cytosol ratio for nYFP-RIPb/cYFP-RIPb BiFC was far higher than that of YFP-RIPb, indicating a more exclusive microtubule localization of the di/oligomer (Fig. 5E). In fact, the signal ratio for nYFP-RIPb/cYFP-RIPb was similar to that of the microtubule marker RFP-MAGAP1-Cter.
Results from Lavy et al. (2007) and Mucha et al. (2010) suggest, that ICRs/RIPs lacking a functional QWRKAA motif lose the ability to interact with ROPs and that either CC1 or CC2 domains can bind to further downstream signaling components. This indicates that RIPb might be able to fulfill a ROP signaling function through one of these domains. To test the functionality of RIPb truncations, we tested their effect on penetration success of Bgh on barley. Interestingly over-expression of RIPbCC2 strongly increased susceptibility by about 75% (Fig. 1C). In contrast, overexpression of the CC2-domain of RIPa did not lead to a significant increase in susceptibility (Supplemental Fig. S4B). The effect of RIPbCC2 completely disappeared when we expressed RIPbVaCC2, containing additionally the Va-domain. The CC1-domain alone also increased susceptibility by about 35% and this effect was also reduced when we expressed the longer RIPbCC1Va truncation (Fig. 1C). In order to investigate the possible influence of protein levels on the influence on susceptibility, we measured the fluorescence intensity of YFP-tagged fusion proteins relative to an internal mCherry control in single transformed cells. We found that fluorescence intensities of full length RIPb and YFP-RIPbVaCC2 were lower than that of RIPbCC1 and RIPbCC2 (Supplemental Fig. S8). Hence, protein expression levels might have influenced the strength of induced susceptibility but CC1 and particularly CC2 alone were sufficient to support fungal penetration success. Oda et al. (2010) showed that in Arabidopsis RIP3/ICR5/MIDD1 is involved in local microtubule depolymerization during xylem cell development.
Microtubule depletion might also influence the outcome of the interaction of barley and Bgh. Indeed, we recently showed that barley RIPa influences microtubule organization when co-expressed with barley RAC1 and MAGAP1 (Hoefle et al., 2020). To see if this could be the case for RIPb in barley, we co-expressed RIPb and RIPbCC2, respectively, with the microtubule marker RFP-MAGAP1-Cter and evaluated microtubule organization as described by Nottensteiner et al. (2018). In the empty vector control, we found 68% of microtubules in a well-organized parallel state, while about 17% of cells show disordered but intact microtubules and the rest of the cells showed fragmented microtubules (Supplemental Fig. S9).
In cells expressing RIPb or RIPbCC2 we observed a similar pattern. The amount of fragmented microtubules was a little lower in cells expressing RIPb and a little higher in cells expressing RIPbCC2, but no statistically significant changes were observed.

RACB and RIPb co-localize at the site of fungal attack
Since RIPb and RACB can interact in planta and both proteins can influence barley susceptibility to Bgh, we wanted to know whether RIPb and RACB would localize to the sites of fungal penetration. Therefore, we transiently co-expressed YFP-RIPb and CFP-RACB in single epidermal cells and inoculated the leaves with conidia of Bgh. At 24 h after inoculation, we observed ring-like accumulations of both YFP-RIPb and CFP-RACB at the site of fungal penetration around the haustorial neck. Cytosolic mCherry appeared also in regions at the site of attack but was less spatially confined than YFP-RIPb and CFP-RACB (Fig. 6A). We observed even more pronounced fluorescence at infection sites, when YFP-tagged RIPb was coexpressed with CFP-CA RACB. In this context, we detected clear accumulation of RIPb and CA RACB at the site of fungal penetration, though independent of the outcome of the penetration attempt. If the penetration was successful, a clear ring-like localization pattern around the haustorial neck could be observed. However, if the fungal penetration was not successful we detected a more fringed accumulation of both proteins, possibly representing membrane domains around papilla protrusions (Fig.   6B). Since RIPbCC2 had a stronger influence on fungal penetration success than full length RIPb, we also imaged YFP-RIPbCC2 when co-expressed with CFP-CA RACB. Interestingly, there was a very strong localization of both proteins around the haustorial neck region in penetrated cells, but also in some instances at sites of repelled fungal attempts (Fig. 6C). The ringlike accumulation of RIPbCC2 around the haustorial neck was also visible later at 48 hours after the inoculation (Fig. 6D). There was also constantly local aggregation of cytoplasm at the sites of attack, but measurements of the ring-like YFP-RIPbCC2 fluorescence showed signal intensities were clearly more confined to the cell periphery compared to cytosolic mCherry fluorescence (Fig. 6E).

Discussion
ICR/RIP proteins are considered scaffold proteins in ROP signaling. Next to RICs, ICRs/RIPs might be key factors in branching of ROP signaling in plants. It appears that so far most described downstream functions of ROPs are mediated through either RIC or ICR/RIP proteins. ICRs/RIPs contain a characteristic QWRKAA motif in the CC2 domain, which was previously described as the motif responsible for ROP interaction (Lavy et al., 2007).
Our results support this, since only full length RIPb and truncations containing this motif interacted with RACB and were subcellularly recruited by CA RACB (Fig. 3, Fig. 4, Supplemental Fig. S5 Fig. S1). The function of this motif, however, remains more elusive. Although the CC1 domain is important for microtubule localization of RIPb (Fig. 4), amino acid exchanges in the QDEL motif did not result in loss of microtubule association (Supplemental Fig. S6). For this study, we focused on a possible RACB signaling mechanism via ICR/RIP proteins during the interaction of barley and Bgh. Barley RIPb interacts with CA and wild type RACB in yeast, supporting that it is a potential downstream interactor of RACB. Over-expression of RIPb but not of RIPa or RIPc increased penetration rate of Bgh into transformed epidermal barley cells (Fig. 1, Supplemental Fig. S4A). RIPb silencing had no significant effect on the interaction between epidermal cells and Bgh (Fig.   1B). This might be due to residual transcript or protein amounts of RIPb after transient knockdown or due to convergence in RACB downstream signaling which could compensate for the lack of RIPb during the interaction. For instance RIC171 might act as an alternative downstream interactor of RACB (Schultheiss et al., 2008), and it is possible that even more interactors of RACB are involved, because ROP proteins are considered signaling hubs (Nibau et al., 2006). Hence silencing of only one signaling branch might not have a significant effect on the interaction, whereas over-expression could support a certain RACB downstream branch and therefore has an effect.
Additionally, RACB is not the only barley ROP that can support fungal penetration success (Schultheiss et al., 2003), and hence even RACBindependent ROP signaling could compensate for RACB-RIPb functions in RIPb-silenced cells.
RIPb shows diverse subcellular localizations. Next to cytosolic localization, we observed localization at the plasma membrane and at the microtubule cytoskeleton (Fig. 2). The N-terminal CC1 domain is necessary but not sufficient for microtubule localization, since the RIPbVaCC2 truncation lacking the CC1-domain did not localize to microtubules. The CC1 domain alone did also not show microtubule localization. The central Va domain alone was insufficient for microtubule association but it appeared to be required for both microtubule association and for RIPb-RIPb interaction (Fig.   4). BiFC experiments further suggested that the RIPb-RIPb interaction takes mainly place at microtubules (Fig. 5). Interestingly, truncated versions of RIPb, which contain the Va domain, did not induce susceptibility when overexpressed, whereas RIPbCC1 and particularly RIPbCC2 induced susceptibility, similar to or much stronger than the full length protein. We therefore speculate that dimerization or oligomerization of RIPb at  Fig. S4B), and therefore this effect appears specific for RIPb.
RIPbCC2 was able to interact with RACB in yeast and in planta (Fig. 4,   Supplemental Fig. S7). Furthermore, RIPb did not localize to the cell periphery anymore without the CC2 domain (RIPbCC1Va) even in presence of CA RACB (Supplemental Fig. S6). This together suggests, that the CC2 domain of RIPb is responsible both for ROP interaction and for a function, which my take place at the plasma membrane.
The N-terminal CC1 domain of RIPb is required for microtubule association but might interact with signaling components as well. This could explain the susceptibility effect of over-expression of RIPbCC1, although the CC1 domain itself does not interact with RACB (Fig. 1C, Fig. 4C). Interestingly, the CC1 domain of Arabidopsis AtRIP3/ICR5/MIDD1 is required for interaction with KINESIN13A (Mucha et al., 2010). It could hence be that RIPb fulfills a dual function via different domains of the protein.
BiFC experiments showed interaction between RACB and RIPb at the microtubules and at the plasma membrane. Since RACB alone does not localize to microtubules (Schultheiss et al., 2003) it seems that RIPb is able to recruit RACB to microtubules when over-expressed. The interaction between the susceptibility-inducing CC2 domain and RACB on the other hand takes place at the plasma membrane (Supplemental Fig. S5, S6).
These results suggest that RACB also recruits RIPb to the plasma membrane during susceptibility signaling and that recruitment of RACB to microtubules perhaps limits this effect. We speculate that in this experimental setup, recruitment of RACB to microtubules brings RACB into proximity of microtubule-located MAGAP1, which presumably inactivates RACB (Hoefle et al., 2011). This might explain why full length RIPb has a less strong effect on susceptibility when compared to RIPbCC2, which cannot recruit RACB to the microtubules. We found that protein levels of YFP-RIPbCC2 are higher than the levels of full length YFP-RIPb when transiently expressed in epidermal cells (Supplemental Fig. 6). Since both constructs are driven by a CaMV35S promotor, a different posttranscriptional regulation or protein turnover might be the most plausible explanation for this. The difference in protein levels can influence the effect that both proteins have, which would also confirm our notion, that RIPbCC2 might be a less regulated functional version of RIPb. However, since RIPbVaCC2 showed similar protein levels to RIPb, but had no influence on the outcome of the interaction between barley and Bgh it is unlikely that protein levels alone explain different efficacies of over-expression constructs.
We observed co-localization of RIPb and RACB and of RIPbCC2 and RACB at the site of fungal attack. In interactions where the fungus was able to penetrate the host cell, a ring of RIPb and RACB or CA RACB around the haustorial neck at the plasma membrane, was observed. However, we also observed signals at repelled penetration attempts around the formed papilla, indicating that accumulation of these two proteins alone is not sufficient to render all cells susceptible. RACB possesses a C-terminal CSIL motif, which is predicted to mediate protein prenylation at the cysteine residue, and is necessary for plasma membrane association and function in susceptibility (Schultheiss et al., 2003). Additionally, RACB has a polybasic stretch close to the C-terminus (Schultheiss et al., 2003) shown for other ROPs to be although at least for RIPb we found no strong evidence that the microtubule cytoskeleton might be affected (Supplemental Fig S9). Together, our data support a hypothesis according to which RIPb is localized at microtubules from which it is recruited to RACB signaling hotspots at the plasma membrane by activated RACB. There it might interact with further proteins of the RACB signaling pathway but also with RACB-independent factors to facilitate fungal entry into barley epidermal cells. The fact that the putative fungal effector ROPIP1 binds RACB and destabilizes barley microtubules (Nottensteiner et al., 2018) adds another level of complexity, on which ROPIP1 may foster release of RIPb from microtubules for its function in susceptibility.

Conclusions
Over the last years, the impact of susceptibility factors for plant -pathogen interactions has become more and more obvious. Barley susceptibility factor RACB might be a key player in cellular polarization during fungal invasion.
Here we identified RIPb as a potential downstream interactor of activated RACB in susceptibility. RACB and RIPb might be involved in fine-tuning of cell polarization in advantage of the fungus. It will be important to identify further interactors of RIPb and of its strongly susceptibility-supporting CC2 domain. This may establish a deep understanding of the components and mechanisms of subcellular reorganizations in the cell cortex, which support the biotrophic parasite Bgh in accommodation of its haustorium in an intact epidermal cell.

Biological Material
Barley (Hordeum vulgare) cultivar Golden Promise was used in all experiments. Plants were grown under long day conditions with 16h of light and 8h in the dark with a relative humidity of 65% and light intensity of 150 µM s -1 m -2 at a temperature of 18°C.
Powdery mildew fungus Blumeria graminis f.sp. hordei race A6 was cultivated on wild type Golden Promise plants under the conditions described above and inoculated by blowing spores into a plastic tent that was positioned over healthy plants or transformed leaf segments.

Cloning procedures
HvRIPb (HORVU1Hr1G012460) was amplified from cDNA using primers

Determination of Susceptibility
Transiently transformed barley leaves were inoculated with Bgh 24 h after bombardment for over-expression constructs and 48 h after bombardment for gene silencing constructs. 24 h after inoculation penetration rate into the transformed cells was determined by fluorescence microscopy as described before (Hückelhoven et al., 2003).

Protein localization and Protein -Protein Interaction in planta
Localization of HvRIPb and co-localization of HvRIPb and HvRACB were determined by transiently transforming barley epidermal cells with plasmids encoding fluorophore fusion proteins. Imaging was done with a Leica TCS SP5 microscope equipped with hybrid HyD detectors. CFP was excitated at 458nm and detected between 465nm and 500nm. YFP was excitated at 514nm and detected between 525nm and 570nm. Excitation of mCherry and RFP was done at 561nm and detection between 570nm and 610nm.

For ratiometric quantification of BiFC experiments Mean Fluorescence
Intensity (MFI) was measured over a region of interest at the cell periphery.
Background signal was subtracted and ratio between YFP and mCherry signal was calculated. At least 25 cells were analyzed per construct for each experiment. Images were taken 24 hours to 48 hours after transformation by particle bombardment.
To evaluate the microtubule to cytosol signal ratio, ten cells per construct were measured. MFI in each cell was measured either on cytosolic strands or along microtubules on three different regions of interest each, in single imaging plains. Average MFI was calculated for cytosol and microtubules, respectively. Afterwards ratio between average microtubule signal and average cytosolic signal was calculated.

Yeast Two-Hybrid assays
For targeted yeast two-hybrid assays, HvRIPb and its truncations were introduced into pGADT7. Introduction of HvRACB into pGBKT7 was