Sequences of the full-length ORFs were obtained from the M. truncatula genome v4.0 (Tang et al., 2014). Employing Gateway^TM technology (Invitrogen), PCR-amplified full-length ORFs were recombined into the donor vector pDONR221. Sequence-verified entry clones were recombined with the destination vector pK7WG2D for overexpression and pK7GWIWG2(II) for silencing in hairy roots (Karimi et al., 2007). All primers used for cloning are reported in Supplementary Table 1.

The bait vector was obtained by cloning the truncated version of MKB1 (MKB1ΔC: AA1-AA237) in the pGBT9 vector. As the prey cDNA library, we opted for a previously generated library from root nodules of M. truncatula A17 inoculated with a Sinorhizobium meliloti strain (Baudin et al., 2015). Screening of the library was performed by transformation of the PJ69-4A yeast strain with the bait by the PEG-LiAc method, subsequently super-transforming this bait strain with the Y2H cDNA library and plating on synthetic defined media devoid of Leu, Trp, and His. PCR was performed on prey plasmids of all transformants on the selective plates using vector-specific primers. After PCR purification (GeneJET PCR Purification; Thermo Scientific^TM), amplicons were subjected to Sanger sequencing for identification of potential interactors of the bait MKB1ΔC. A complete overview of identified candidate interactors of MKB1ΔC with their gene identity and annotation is provided in Supplementary Table 2.

Subsequent Y2H assays were performed essentially as described (Cuéllar Pérez et al., 2013). Bait and prey were fused to the GAL4 activation domain or GAL4 DNA-binding domain via cloning into the pGAL424gate/pDEST22 or pGBT9gate/pDEST32, respectively (Cuéllar Pérez et al., 2013). Yeast transformants were selected on synthetic defined (SD) medium lacking Leu and Trp (Clontech, Saint -Germain-en-Laye, France). For Y3H analysis, a construct with a third potential interaction partner, N-terminally fused to a nuclear localization signal was generated by cloning into the pMG426-NLS vector (Nagels Durand et al., 2012), which was subsequently co-transformed with the bait and prey constructs. For Y3H, transformants were selected on SD medium lacking Leu, Trp, and Ura (Clontech, Saint -Germain-en-Laye, France). For both Y2H and Y3H assays, three individual colonies were grown overnight in liquid cultures at 30°C, and 10- or 100-fold dilutions were dropped on control and selective media lacking His in addition to the plasmid auxotrophy markers (Clontech).

The E2 UBCs of Arabidopsis were collected from Kraft et al. (2005). From clade VI, E2 UBCs of H. sapiens and S. cerevisiae were also selected together with the M. truncatula E2 UBC Medtr3g062450. Protein sequences were aligned with ClustalW. The phylogenetic tree was generated in MEGA7 software (Kumar et al., 2016), by the neighbor-joining method (Saitou and Nei, 1987), and bootstrapping was done with 1,000 replicates. The evolutionary distances were computed using the JTT matrix-based method and are in the units of the number of amino acid substitutions per site (Jones et al., 1992). The analysis involved 41 amino acid sequences. All positions containing gaps and missing data were eliminated. There was a total of 112 positions in the final dataset. Evolutionary analyses were conducted in MEGA7 (Kumar et al., 2016).

Sterilization of M. truncatula seeds (ecotype Jemalong J5), transformation of seedlings by Agrobacterium rhizogenes (strain LBA 9402/12), and the subsequent generation of hairy roots were carried out as described previously (Pollier et al., 2011). Hairy roots were cultivated for 21 days in liquid medium prior to sampling for RNA, protein, and metabolite extraction.

One hundred milligrams of frozen roots of three independent transgenic lines were ground in a Retsch MM300 mixer, and total RNA was extracted using the Qiagen RNeasy kit (Qiagen). One microgram of RNA was used for cDNA synthesis using the iScript^TM cDNA Synthesis Kit (Bio-Rad). qRT-PCR was performed in the LightCycler 480 System (Roche) using the Fast Start SYBR Green I PCR mix (Roche) via the following program: pre-incubation (95°C, 10 s), 45 amplification cycles (incubation 95°C, 10 s; annealing 65°C, 15 s; elongation 72°C, 15 s). Relative expression levels using multiple reference genes were calculated using qBase (Hellemans et al., 2007). The M. truncatula 40S ribosomal protein S8 and translation elongation factor 1a were used as reference genes. Primer sequences are presented in Supplementary Table 1.

Recombinant glutathione S-transferase (GST)–MKB1ΔC fusion proteins [truncated with RING mutation (MKB1ΔCmRING) or without mutation (MKB1ΔC)] were purified according to the manufacturer’s instructions with Glutathione Sepharose 4B resin columns (GEHealthcare) from transformed E. coli cells, pretreated for 2 h with isopropyl-b-D-1-thiogalactopyranoside (IPTG). A protein refolding step to assure the full ion Zn charge of the GST–MKB1 fusion proteins was included by incubation with refolding buffer (20 mM HEPES, pH 7.4, 0.02 mM ZnCl₂, 1.5 mM MgCl₂, 150 mM KCl, 0.2 mM EDTA, 20% glycerol, 0.05% Triton X-100) for 1 h at 4°C. Ubiquitination reactions were performed in a total volume of 30 ml using 15 ml of the refolded GST–MKB1 bound to glutathione resin. The reaction contained 300 ng of GST–MKB1 fusion protein as E3 ubiquitin ligase, 250 ng of the ubiquitin-activating enzyme (UBE1) from rabbit (BostonBiochem), 400 ng of human recombinant UBCH5A protein (BostonBiochem) or C-terminally 6xHis-tagged Medtr3g062450 from the pDEST17 vector, and 2 mg of HA-Ub from human (BostonBiochem) in ubiquitination buffer (50 mM HEPES, pH 7.4, 2 mM ATP, 5 mM MgCl₂, 2 mM DTT, 0.02 mM ZnCl₂). The ubiquitination reactions were incubated for 1 h at 30°C and stopped by adding Laemmli buffer (4% SDS, 20% glycerol, 10% 2-mercaptoethanol, 0.004% bromophenol blue, 0.125 M Tris-HCl, pH 6.8). Samples were resolved on 8% SDS–PAGE, followed by protein immunoblot analysis with anti-HA (Qiagen) and anti-GST (GE Healthcare) antibodies.

For co-localization analysis, the ORFs of MKB1, MASH, and Medtr3g062450, including and lacking their stop codon, were cloned in pDONR221 to obtain entry clones that were subsequently used to generate CaMV35S promoter-driven C-terminal and N-terminal GFP fusion constructs in pFAST-R05 and pFAST-R06 destination vectors, respectively, via single LR Gateway^TM (Invitrogen) reactions (Shimada et al., 2010).

For agro-infiltration, wild-type tobacco (Nicotiana benthamiana) plants were grown for 3–4 weeks. Tobacco infiltration of lower epidermal leaf cells with the Agrobacterium tumefaciens strains was performed as described in Boruc et al. (2010).

Image acquisition was obtained with a Zeiss 710 inverted confocal microscope, equipped with a 63× water-corrected objective (n.a. 1.2) using the following settings for EGFP and mCHERRY detection: EGFP excitation at 488 nm, emission filter 500–530 nm, mCHERRY excitation at 559 nm, and emission filter of 630–660 nm. Confocal images were acquired using the ZEN software package attached to the confocal system. Confocal images were processed with ImageJ¹.

Protein extraction from M. truncatula hairy roots and determination of HMGR protein levels by immunoblot analysis was carried out as described (Pollier et al., 2013).

M. truncatula hairy roots (five biological repeats of three independent transgenic lines per transgene construct) were grown for 21 days in liquid medium and upon harvest immediately frozen in liquid nitrogen. Processing and metabolite extraction from 400 mg of the hairy root tissue was performed as described (Pollier et al., 2011). LC-ESI-FT-ICR-MS analysis was carried out using an Acquity UPLC BEH C18 column (150 × 2.1 mm, 1.7 mm; Waters, Waltham, MA, United States) mounted on an LC system consisting of an Accela pump and autosampler (Thermo Electron Corporation, Waltham, MA, United States) coupled to an LTQ FT Ultra (Thermo Electron Corporation) via an electrospray ionization source operated in negative mode. A gradient was run using acidified (0.1% formic acid) water:acetonitrile (99:1, v/v; solvent A) and acetonitrile:water (99:1, v/v; solvent B): 0 min, 5% B; 30 min, 55% B; and 35 min, 100% B. The injection volume was 10 μl, the flow rate 300 ml/min, and the column temperature 40°C. Negative ionization was obtained with a capillary temperature of 150°C, sheath gas of 25 (arbitrary units), auxiliary gas of 3 (arbitrary units), and a spray voltage of 4.5 kV. Full MS spectra between m/z 120 and 1,400 were recorded. MSn spectra (MS2 and two dependent MS3 scan events, in which the two most abundant daughter ions were fragmented) were generated from the most abundant ion of each full MS scan. The collision energy was set at 35%. The resulting chromatograms were integrated and aligned using the Progenesis QI software (Waters). The PCA was performed with MetaboAnalyst 4.0 with Pareto-scaled mass spectrometry data and standard settings² (Chong et al., 2018).

To identify novel interactors of MKB1, a yeast two-hybrid (Y2H) screen was performed using an available prey M. truncatula cDNA library (Baudin et al., 2015) and MKB1 devoid of its membrane spanning domain (MKB1ΔC; amino acid 1–237, to allow retrieving interactors of the catalytic domain in the Y2H system) as bait (Figure 1A). Identification of interacting preys was performed by Sanger sequencing of the respective cDNA inserts of yeast colonies that survived selection. For this study, only prey inserts identified in at least two independent transformants were considered for further in-depth analysis (Figure 1B and Supplementary Table 2). From these candidates, the full-length coding sequences were cloned de novo for binary interaction validation by Y2H again using MKB1ΔC as the bait. Interaction with ubiquitin, the E2 UBC and the HSP40 protein encoded by Medtr3g092130, Medtr3g062450, and Medtr3g100330, respectively, could be confirmed and were subjected to further analysis (Figure 1C).

Besides ubiquitin itself, an obvious potential additional member of the canonical MKB1 complex is the E2 UBC encoded by Medtr3g062450, which clusters with the clade VI E2 UBCs (Supplementary Figure 1). Group VI is the largest group of E2 UBCs comprising more promiscuous E2 UBCs that function in vitro with multiple E3s from different families to mediate K48-linked poly-ubiquitination, typically reported to target proteins to the proteasome for degradation, in a process called the ubiquitin–proteasome system (UPS) (Callis, 2014). Because any plant genome is predicted to encode tens of E2 UBCs, we wanted to evaluate whether MKB1ΔC uniquely interacts with E2 UBCs from clade VI. Therefore, a Y2H screen was set up using a publicly available library of Arabidopsis E2 UBCs, which contains 30 (out of 37) different E2 UBCs (Kraft et al., 2005; Nagels Durand et al., 2016). As expected, MKB1ΔC interacted with AtUBC8-11 and AtUBC28-30, which are both members of clade VI, but also with AtUBC15 and 18 that both belong to clade VII (Figure 2). Arabidopsis E2 UBCs of clade VII are related to the human UBC E2 W (Ube2W), which is reported to catalyze N-terminal ubiquitination of its target proteins (Callis, 2014). To our knowledge, to date, no specific activity or function has yet been assigned to either AtUBC15 or 18. Taken together, our interaction data show that MKB1ΔC can interact with a specific subset of E2 UBCs that are related to N-terminal ubiquitination and K48 poly-ubiquitination.

Key structural elements of RING E3 ligases, to which also MKB1 belongs, are the two loop-like regions that coordinate the Zn²⁺ ions, surrounding a shallow groove formed by the central α-helix. Together, these elements serve as an interface for interaction with the UBC domains of E2 UBCs (Deshaies and Joazeiro, 2009; Metzger et al., 2014). To assess whether Medtr3g062450 encodes a possible canonical E2 UBC for MKB1, an in vitro ubiquitination assay was performed as previously reported and in which it was shown that the human protein HsUBCH5A catalyzes auto-ubiquitination of MKB1ΔC in vitro (Pollier et al., 2013). We could demonstrate in vitro auto-ubiquitination activity of GST-tagged MKB1ΔC in the presence of recombinant His-tagged Medtr3g062450 protein and HA-tagged ubiquitin (Figure 3A). Furthermore, we evaluated whether the auto-ubiquitination activity of MKB1ΔC by the E2 UBC is dependent on the integrity of the E3 RING domain. Several functional studies of RING E3s typically employ mutations in the zinc-binding residues to inactivate the RING domain (Deshaies and Joazeiro, 2009). However, such mutations perturb the overall RING domain structure. Therefore, we targeted MKB1 residues necessary to sustain the UBC-RING contact sites instead, determined by a sequence alignment of the RING domain of MKB1 with the conserved RING-like U-box of human CHIP and the RING domains of human c-CBL and cIAP2 (Figure 3B). The predicted contact site Ile^AA39 was replaced with a charged residue, Arg^AA39, in MKB1ΔC (Figure 3B) resulting in a RING-dead MKB1ΔC version (MKB1ΔCmRING) that is different from the one previously generated in Pollier et al. (2013) by substituting the Cys³⁷ and Cys⁴⁰ residues by Ser residues. The absence of ubiquitination of MKB1ΔCmRING indicates that the auto-ubiquitination activity of MKB1ΔC by Medtr3g062450 is dependent on the integrity of its RING domain (Figure 3A). These results suggest that the Medtr3g062450 E2 UBC is a canonical E2 UBC for MKB1, catalyzing auto-ubiquitination of MKB1, and, consequently, possibly also ubiquitination of MKB1 targets.

The second candidate member of the MKB1 E3 ligase complex is the HSP40 encoded by Medtr3g100330, which we named MKB1-supporting heat-shock protein 40 (MASH). Notably, mining of the transcriptome data available on the Medicago truncatula Gene Expression Atlas (MtGEA) (He et al., 2009) indicated that MASH expression was highly correlated with that of MKB1 and its target HMGR1 (Figure 4A). For instance, a concerted upregulation of these three genes is observed in M. truncatula cell suspension cultures upon methyl JA (MeJA) treatment, in roots and shoots upon drought stress and in root hydroponic systems in high-salt conditions. Expression of Medtr3g062450 is not co-regulated with these three genes (Figure 4A), which may correspond to its plausible pleiotropic role as E2 UBC in other, MKB1-independent UPS processes. Based on its domain organization, MASH belongs to the subtype III of HSP40s that possess a canonical J-domain (Figure 4B) and generally act as obligate HSP70 co-chaperones that assist in diverse processes of cellular protein metabolism (Misselwitz et al., 1998; Laufen et al., 1999; Fan et al., 2003; Walsh et al., 2004; Craig et al., 2006; Rajan and D’Silva, 2009; Kampinga and Craig, 2010). The structure of the J-domain is conserved across all kingdoms and consists of four helices with a tightly packed helix II and III in antiparallel orientation. A flexible loop containing a highly conserved and functionally critical HPD signature motif, pivotal to trigger ATPase activity of HSP70s, connects both helices (Figure 4B; Laufen et al., 1999; Walsh et al., 2004). Hydrophobicity analysis of MASH revealed that it does not encompass a clear trans-membrane domain, indicating that it would not reside in the ER membrane as its potential ER membrane-anchored partner MKB1, but possibly is active in the cytoplasm to which also the catalytic part of MKB1 is exposed (Figure 4C). This was confirmed by co-localization studies in Agro-infiltrated N. benthamiana leaves, in which MASH predominantly showed a nucleocytosolic localization, whereas the E2 UBC Medtr3g062450 showed both nucleocytosolic and ER localization (Figure 4D). Co-expression of free MKB1 did not alter MASH localization either (Supplementary Figure 2). This result is not surprising given our actual difficulties in visualizing or detecting GFP-tagged MKB1 protein in Agro-infiltrated N. benthamiana leaves, either in the wild-type or ring-dead version. An MKB1-GFP signal was rarely visible, even in the co-localization assays; hence, we could not robustly determine its localization in this set-up. Accordingly, bimolecular fluorescence complementation (BiFC) experiments to detect in planta interaction between MKB1 and either MASH or E2 UBC Medtr3g062450 all consistently failed. Attempts to express and visualize GFP-tagged MKB1 protein in stably transformed M. truncatula hairy root lines were not successful either.

To determine the physiological role and relevance of MASH in the MKB1 E3 ligase complex, a functional analysis was carried out in planta. To this end, three independent stable MASH overexpression (MASH^OE), MASH knock-down (MASH^KD), MKB1 knock-down (MKB1^KD), and GUS overexpression (CTR) M. truncatula hairy root lines were generated (Figure 5A). MASH^KD roots displayed a strikingly similar phenotype to that of MKB1^KD hairy roots, as previously reported by Pollier et al. (2013; Figure 5B). At the morphological level, MASH^KD and MKD1^KD roots both showed dissociation of hairy roots into “caltrop”-like structures (Figure 5C). Comparable phenotypes were not observed in MASH^OE roots (Figures 5B,C), correlating with the previously described absence of a phenotype in MKB1^OE roots (Pollier et al., 2013).

It has previously been reported that silencing of MKB1 results in an altered metabolism, manifested in an altered flux toward TS biosynthesis. Detailed metabolic profiling of MKB1^KD hairy roots showed a higher accumulation of mono-glycosylated saponins, including 3-O-Glc-medicagenic acid, and reduced levels of high-level glycosylated saponins such as soyasaponin I (Pollier et al., 2013). To verify whether the MKB1^KD phenotype of the MASH^KD roots was also reflected in its metabolite composition, metabolite profiling by liquid chromatography electrospray ionization Fourier transform ion cyclotron resonance mass spectrometry (LC-ESI-FT-ICR-MS) was carried out on CTR, MASH^KD, MASH^OE, and MKB1^KD lines. A principal component analysis (PCA) on the LC-ESI-FT-ICR-MS dataset was carried out and revealed grouping of the samples derived from MKB1^KD and MASH^KD roots. These samples were clearly separated from the samples derived from CTR and MASH^OE roots (Figure 6A), implying a similar trend in the metabolic profile of the MKB1^KD and MASH^KD roots, and no major differences in the metabolite composition of CTR and MASH^OE roots. Relative quantification of diagnostic mono-glycosylated TSs, such as 3-O-Glc-medicagenic acid, in the various hairy root samples showed that these metabolites were significantly more highly abundant in both MKB1^KD and MASH^KD roots (Figure 6B). Conversely, like in MKB1^KD roots, several high-level glycosylated TSs, such as soyasaponin I, were significantly less abundant in MASH^KD roots (Figure 6B). Although there were still significant differences in the levels of these TSs between MKB1^KD and MASH^KD roots, it could be concluded that the trends in the alterations at the metabolite level in MKB1^KD and MASH^KD roots were similar. No significant differences between CTR and MASH^OE roots were observed for these metabolites, except for soyasaponin I (Figure 6B).

Finally, MKB1^KD hairy roots have been shown to also exert a TS-specific negative feedback on the transcriptional level (Pollier et al., 2013). To evaluate whether MASH^KD roots showed a similar transcript profile, quantitative reverse transcription PCR (RT-qPCR) was performed on TS-specific biosynthesis genes in MASH^KD, MKB1^KD, MASH^OE, and CTR roots. Expression of the BAS, CYP93E2, CYP716A12, UGT73F3, and UGT73K1 genes, all encoding TS-specific enzymes, was strongly downregulated in MASH^KD roots, similar to MKB1^KD roots (Figure 7 and Supplementary Table 3). Conversely, we did not detect a general downregulation of sterol-specific biosynthesis genes in MASH^KD roots, in accordance with what was observed in MKB1^KD roots (Supplementary Figure 3). Finally, and importantly, MKB1 transcript levels were not downregulated in the MASH^KD roots (Figure 5A), supporting that the observed MASH^KD phenotypes can be attributed to MASH silencing and are not caused by a mere downregulation of MKB1. Hence, together, these data indicate that silencing of MASH does not affect the transcriptional regulation of triterpenes in general, but specifically affects TS biosynthesis, as is the case with MKB1 silencing (Pollier et al., 2013).

Given that all observable MKB1^KD phenotypes were mirrored in the MASH^KD roots, we hypothesized that loss of MASH function would also affect the targeted degradation of HMGR by MKB1 in M. truncatula. To investigate whether MASH indeed assists in the MKB1-mediated degradation of HMGR, we monitored both HMGR transcript and protein levels in planta in MASH^KD roots, as well as in MKB1^KD and CTR roots, following MeJA application. The rationale behind this experimental design is that we had previously shown that effects of MKB1 silencing on in planta HMGR protein levels could only be detected after MeJA application (Pollier et al., 2013). Indeed, as confirmed in the analysis conducted here, MeJA-induced HMGR transcript upregulation, which is observed in all lines (Supplementary Figure 4A), only resulted in detectably higher HMGR protein levels in MeJA-elicited MKB1^KD roots but not in control roots (Supplementary Figures 4B–E). Unexpectedly, however, HMGR protein levels did not significantly increase in MASH^KD roots following MeJA treatment, and showed a similar trend as the CTR lines (Supplementary Figure 4D), suggesting that degradation of HMGR may not be significantly perturbed by loss of MASH function.

In mammalian and yeast cells, the INSIG proteins bridge the ERAD E3 ubiquitin ligases GP78 and HRD1 to their targets, including the HMGR proteins (Burg and Espenshade, 2011; Wangeline et al., 2017; Johnson and DeBose-Boyd, 2018). This interaction is dependent on the membrane spanning SSD, which is absent in plant HMGR proteins. Nonetheless, we wanted to assess whether MASH can recruit M. truncatula HMGR1 to the MKB1 machinery and thus act as an INSIG-analog, but then as a cytosolic version that would connect the catalytic domains of HMGR1 and MKB1, which are both exposed to the cytosol. A similar function has been reported for cytosolic HSP40-type chaperones in the degradation of membrane-localized hepatic P450s (Kim et al., 2016). To this end, a Y2H assay was performed to explore the potential binary interaction between MASH and M. truncatula HMGR1 devoid of its membrane domain (HMGR1ΔN). However, no interaction was detected (Supplementary Figure 5A). Next, we hypothesized that MASH may only bind HMGR1 in the presence of MKB1. Therefore, a yeast three-hybrid (Y3H) assay was performed. However, also here, no interaction was observed between MKB1ΔC and HMGR1ΔN in the presence of MASH (Supplementary Figure 5B). In conclusion, our data do not support a possible role for MASH as a mediator of MKB1-HMGR interaction, at least not on its own.

To assess whether MASH-MKB1 interaction may be conserved in other plant species, we assessed the interaction between the Arabidopsis RMA-type E3 ubiquitin ligase homologs of MKB1, called RMA1 to RMA3, against MASH, its closest homolog in M. truncatula, as well as the closest Arabidopsis MASH homologs. Arabidopsis RMAs have been previously reported as ERAD-type E3 ubiquitin ligases with possible roles in growth and development (Matsuda et al., 2001; Lee et al., 2009; Son et al., 2009). Only the Arabidopsis RMA1 devoid of its membrane domain, RMA1ΔC, could indeed directly interact with the M. truncatula MASH (Figure 8). The closest homolog of MASH in M. truncatula, encoded by Medtr5g066100, did not interact with MKB1ΔC, nor with any of the Arabidopsis RMAs. Therefore, interaction between MASH and MKB1 seems specific, and no redundancy seems to exist for MASH functioning in M. truncatula, as also evidenced by the strong phenotype of the MASH^KD roots.

Next, direct binding between the truncated Arabidopsis RMAs, MKB1ΔC, and the three closest Arabidopsis MASH homologs, encoded by At2G01710, At5G09540, and At5G64360, was assessed. These three homologs were identified through the PLAZA tool (Van Bel et al., 2018) and also corresponded to the first three BlastP hits for MASH in the Arabidopsis genome. We could observe specific interaction between AT5G64360 and RMA1ΔC, RMA2ΔC and MKB1ΔC (Figure 8). Together, these data suggest that a putative role of MASH in the RMA-type E3 ubiquitin ligase machinery could be conserved in Arabidopsis.

One of the principal aims of the Y2H screen that we launched, was to identify the “adaptor” protein that would connect MKB1 to its target(s), in particular HMGR. This was encouraged by previous findings, in particular the observation that MKB1 can target yeast HMG2P and thereby complement a S. cerevisiae yeast strain devoid of the HRD1 E3 ubiquitin ligase (Pollier et al., 2013). Although M. truncatula uses a different family of ERAD E3 ubiquitin ligases from those directing HMGR for destruction in yeast (HRD1), they thus appeared to be compatible, suggesting that both ERAD E3 ubiquitin ligases rely on a common adaptor. Our Y2H screen revealed the HSP40 chaperone MASH as a direct MKB1 interactor, which at first sight represented an excellent candidate for a possible mediator of MKB1-HMGR interaction. Indeed, several precedents for such a role of HSP40 proteins exist in the field. In yeast, the E3 ubiquitin ligases HRD1 and DOA10 make use of an ER-resident HSP70-binding protein 3 (BiP3) to survey client ERAD substrates other than HMGR (Ruggiano et al., 2014). The DOA10 complex is also known to target ERAD substrates with lesions in the cytosolic domain, and is surveyed by the cytosolic HSP70, SSA1P, and the HSP40s, YDJ1P, and HLJ1P (Ruggiano et al., 2014). Interestingly, YDJ1P appears to be the closest homolog of MASH in S. cerevisiae. Likewise, in mammalian cells, an adaptor function has been suggested for cytosolic HSP40-type chaperones, again in association with HSP70 proteins, to mediate the interaction between the U-box type E3 ubiquitin ligase CHIP and its UPS target, the membrane-localized hepatic P450 protein CYP3A4 (Kim et al., 2016) or between human RMA1 and its UPS target cystic fibrosis transmembrane conductance regulator (CFTR) (Grove et al., 2011). However, our data do not support an adaptor role for the HSP40 chaperone MASH in M. truncatula, at least not by itself. No interaction with HMGR could be observed, nor did loss of MASH function significantly affect in planta HMGR levels. It is possible that the HMGR-MKB1 adaptor protein is also a membrane protein, like MKB1 and HMGR themselves, and the INSIGs. Hence, it would not have been possible to isolate it through a classical Y2H screen. Screens through other methods, such as affinity purification coupled to mass spectrometry, as we have recently established in M. truncatula hairy roots (Goossens et al., 2016a) and which can be adapted to isolate membrane protein complexes (Bassard et al., 2012), may offer a potent alternative, as well as the recently developed proximity labeling method with TurboID (Arora et al., 2019), which is particularly useful to detect integral membrane protein–protein interactions. Both methods may also allow revealing alternative MKB1 substrates and/or co-chaperones such as HSP70 proteins, which may reveal a multi-protein adaptor complex to bridge MKB1 and its targets.

It appears that with MASH and the clade IV E2 UBCs, M. truncatula MKB1 has recruited cytosolic ERAD machinery components to facilitate the degradation of ER-localized targets. Possibly this may apply to plant RMA-type ERAD E3 ubiquitin ligases in general, as evidenced by the conserved interaction between the Arabidopsis MKB1-MASH homologs. Because the Arabidopsis MKB1-homolog RMA2 can also interact with the Arabidopsis clade VI E2 UBC29 (Arabidopsis Interactome and Mapping Consortium, 2011) besides the MASH homolog AT5G64360 and because Arabidopsis RMA1 has been reported to accept ubiquitin from mammalian clade VI E2 UBCs for in vitro auto-ubiquitination (Matsuda et al., 2001), we postulate that the putative role of MASH in the RMA-type E3 ubiquitin ligase machinery could be conserved in Arabidopsis and possibly other (dicot) plant species as well. As an alternative to a direct role in surveying the ERAD of substrates, the interaction with MASH may aid in preserving the stability of the ERAD E3 ubiquitin ligase itself. Such a possible stabilizing role of MASH was suggested by some preliminary data. For instance, in some of our transient expression assays in Agro-infiltrated N. benthamiana leaves, co-expression of MASH with tagged MKB1 appeared to stabilize the MKB1 protein and increase its accumulation levels. However, given the variable and low amounts of detectable tagged MKB1 protein, robust visualization by confocal imaging, or quantification of the MKB1 accumulation levels by immunoblot analysis resulted impossible; hence, strong postulation on a possible role of MASH as an MKB1-stabilizing chaperone needs more experimental support. Unfortunately, also all of our efforts to visualize or assess MKB1 protein levels in M. truncatula were unsuccessful, so to date, we did not manage to further probe this postulation. Nonetheless, precedents for the necessity for such a role exist in the field. In yeast, the membrane protein HRD3 is always present in a stoichiometric complex with HRD1 and is essential for the execution of HRD-dependent protein degradation because loss of HRD3 causes unrestricted self-degradation of HRD1 (Vashistha et al., 2016). An analogous system seems to exist for the multi-protein Skp/Cullin/F-box (SCF)-containing E3 ubiquitin ligase complexes. Indeed, the HSP-complex HSP90-HSP40/SGT1b was recently shown to stabilize the F-box component TRANSPORT INHIBITOR RESPONSE1 of the auxin receptor SCF complex in Arabidopsis in response to low and high temperatures to maintain proper plant growth and development (Wang et al., 2016). Research on MASH-MKB1 homologs in other plants, in parallel with further characterization of the MASH-MKB1 machinery in M. truncatula will allow to elucidate the functioning of this vital machinery for plant protein quality control.