Genetic and physical interactions between the organellar mechanosensitive ion channel homologs MSL1, MSL2, and MSL3 reveal a role for inter‐organellar communication in plant development

Abstract Plant development requires communication on many levels, including between cells and between organelles within a cell. For example, mitochondria and plastids have been proposed to be sensors of environmental stress and to coordinate their responses. Here we present evidence for communication between mitochondria and chloroplasts during leaf and root development, based on genetic and physical interactions between three Mechanosensitive channel of Small conductance‐Like (MSL) proteins from Arabidopsis thaliana. MSL proteins are Arabidopsis homologs of the bacterial Mechanosensitive channel of Small conductance (MscS), which relieves cellular osmotic pressure to protect against lysis during hypoosmotic shock. MSL1 localizes to the inner mitochondrial membrane, while MSL2 and MSL3 localize to the inner plastid membrane and are required to maintain plastid osmotic homeostasis during normal growth and development. In this study, we characterized the phenotypic effect of a genetic lesion in MSL1, both in wild type and in msl2 msl3 mutant backgrounds. msl1 single mutants appear wild type for all phenotypes examined. The characteristic leaf rumpling in msl2 msl3 double mutants was exacerbated in the msl1 msl2 msl3 triple mutant. However, the introduction of the msl1 lesion into the msl2 msl3 mutant background suppressed other msl2 msl3 mutant phenotypes, including ectopic callus formation, accumulation of superoxide and hydrogen peroxide in the shoot apical meristem, decreased root length, and reduced number of lateral roots. All these phenotypes could be recovered by molecular complementation with a transgene containing a wild type version of MSL1. In yeast‐based interaction studies, MSL1 interacted with itself, but not with MSL2 or MSL3. These results establish that the abnormalities observed in msl2 msl3 double mutants is partially dependent on the presence of functional MSL1 and suggest a possible role for communication between plastid and mitochondria in seedling development.


| INTRODUC TI ON
Plastids and mitochondria are found in almost every plant cell and are involved in all aspects of plant biology. In plants, as in animals, mitochondria are involved in multiple cellular processes, including cellular respiration and co-enzyme synthesis (Rébeillé, Alban, Bourguignon, Ravanel, & Douce, 2007;Schertl & Braun, 2014).
Mitochondrial activity is thought to protect against photoinhibition and oxidative damage to chloroplasts by dissipating excess redox equivalents from the chloroplasts under high light conditions (Yoshida, Terashima, & Noguchi, 2007). Conversely, mitochondrial respiration has long been understood to be modulated by light. For example, the alternative oxidase AOX1a (a component of the mitochondrial electron transport chain) is up-regulated by light (Yoshida & Noguchi, 2011;Yoshida, Watanabe, Kato, Sakamoto, & Noguchi, 2008).
The mechanism by which chloroplasts and mitochondria communicate is not fully understood. While there is evidence for the transfer of lipids via physical contact between chloroplasts and mitochondria (Jouhet et al., 2004), further validation is required (Delfosse et al., 2015). Communication may be mediated through the diffusion of factors through the cytosol, through direct contacts with other organelles (de Souza, Wang, & Dehesh, 2017), or via signals to the nuclear genome (retrograde signaling) that are then conveyed to the other organelle (Chan, Phua, Crisp, McQuinn, & Pogson, 2016;Kleine & Leister, 2016;de Souza et al., 2017;Woodson & Chory, 2012).
We have been studying the effect of organellar osmotic stress on plant development. We previously showed that two members of the MscS-Like (MSL) family of mechanosensitive ion channels, MSL2 and MSL3, serve to maintain osmotic homeostasis in plastids during normal growth and development (Haswell & Meyerowitz, 2006;Veley, Marshburn, Clure, & Haswell, 2012). MSL proteins are homologs of the mechanosensitive channel MscS (Mechanosensitive channel of small conductance), which serves as an "osmotic safety valve" to protect Escherichia coli against lysis during extreme hypoosmotic shock (Levina et al., 1999;Naismith & Booth, 2012). MSL2 and MSL3 are localized to the inner chloroplast membrane and msl2 msl3 double mutants produce a range of plastid defects, including enlarged and round epidermal cell plastids, defective chloroplast division and abnormal ultrastructure in the proplastids of the shoot apical meristem (Wilson, Jensen & Haswell, 2011;Wilson, Basu, Bhaskara, Verslues, & Haswell, 2014;Wilson, Mixdorf, Berg, & Haswell, 2016;Haswell & Meyerowitz, 2006). Furthermore, msl2 msl3 plants have multiple developmental defects, including dwarfing and leaf variegation. After culture on solid media, they form ectopic calluses at the meristem, a process that is dependent on superoxide accumulation in plastids . All of these developmental phenotypes can be interpreted as direct or indirect consequences of plastid osmotic dysregulation, as all are suppressed when plants or cells are supplied with osmotic support (Veley et al., 2014;Wilson et al., 2014Wilson et al., , 2016. MSL2 and MSL3 can partially rescue an MS channel mutant E. coli strain, suggesting that they form MS ion channels as shown for several other members of the family (Haswell & Meyerowitz, 2006;Lee et al., 2016;Maksaev & Haswell, 2012;Hamilton et al., 2015), but their electrophysiological characterization remains elusive.
MSL2 and MSL3 are two members of a 10-gene family in the genome of Arabidopsis thaliana (Haswell, 2007). Another member, MSL1, is also found in endosymbiotic organelles. Subcellular fractionation and GFP-fusion protein localization experiments demonstrate that MSL1 localizes to the inner mitochondrial membranes (Haswell & Meyerowitz, 2006;Lee et al., 2016). The mature form of MSL1 provides a mechanically activated ion channel activity in excised membrane patches (Lee et al., 2016). Plants harboring the null msl1-1 allele (hereafter referred to as msl1) are indistinguishable from the wild type under normal growth conditions. However, plant mitochondria isolated from msl1 mutants exhibit increased transmembrane potentials when the F 1 F 0 ATP synthase is inhibited.
Compared to the wild type, msl1 mutants also show a larger increase in mitochondrial glutathione oxidation in response to oligomycin,

K E Y W O R D S
Arabidopsis, chloroplast, mechanosensitive channel, mitochondria, reactive oxygen species high temperature, and cadmium treatments, as measured with a redox-sensitive fluorescent reporter (mito-roGFP2). These data show that MSL1 plays a role in maintaining mitochondrial redox homeostasis during abiotic stress, but how direct these effects are and the role (if any) played by membrane stretch or ion flux is not yet known.
The presence of MSL channels in both chloroplast and mitochondrial envelopes, combined with existing evidence for integration of organellar responses to environmental and metabolic signals, led us to propose that MSL1, MSL2, and MSL3 may interact to coordinate a cellular response to osmotic stresses. To begin to test this idea, we characterized the genetic and physical interactions between MSL1, MSL2, and MSL3 in Arabidopsis. Our results reveal an unexpected genetic relationship whereby loss of MSL1 enhances some but suppresses other phenotypes previously observed in the msl2 msl3 mutant. We also document new phenotypes in the msl2 msl3 root and show that these are also ameliorated in the msl1 msl2 msl3 triple mutant. Finally, we demonstrate that MSL1 and MSL2 are capable of interacting with themselves in the split-ubiquitin yeast two hybrid assay, and that MSL2 and MSL3 interact with each other, but not with MSL1. These results point to a complex interplay between osmotic stress signals from the chloroplast and the mitochondria that lead to developmental outcomes in both the shoot and the root.
Transmembrane domains and overall topology were predicted with Aramemnon (Schwacke & Flügge, 2018). Amino acid sequences were aligned using Clustal Omega 1.2.4 and default settings (Sievers & Higgins, 2018). Percent identity and similarity were calculated as number of identical or similar residues in the alignment divided by the total number of positions in the alignment, including gaps.

| Generation and validation of msl1 msl2 msl3 triple mutant and msl1 msl2 msl3 + MSL1 g complementation lines
The msl1 msl2 msl3 triple mutant was generated by crossing the msl1-1 mutant (first reported in Lee et al., 2016) to the msl2-3 msl3-1 double mutant (first reported in Wilson, et al., 2011). Triple mutant plants were identified in the F3 generation by PCR genotyping. A genomic copy of the MSL1 locus (including all sequence from 1,207 bp upstream of the ATG to 208 bp downstream of the TAG, including introns) was cloned into the pBGW backbone to make the molecular complementation construct MSL1g (Lee et al., 2016). To generate homozygous msl1 msl2 msl3 lines complemented with a genomic copy of MSL1 (msl1 msl2 msl3 + MSL1g), MSL1g was introduced into the msl1 msl2 msl3 background via Agrobacteriummediated floral dip (Clough & Bent, 1998). Siblings homozygous for the presence or absence of MSL1g were identified in the T3 as lines exhibiting 100% or 0% Basta-resistance, respectively. All lines were validated by PCR genotyping. The MSL1 locus and our approach to genotyping the genomic locus of MSL1 in the presence of MSL1g are shown in Supporting Information Figure S1.

| Superoxide and hydrogen peroxide detection
For superoxide detection, 21-day-old seedlings of the indicated genotypes were collected and treated side-by-side. First, they were vacuum-infiltrated for 4 min in 0.1% weight-to-volume nitro blue tetrazolium in 10 mM potassium phosphate buffer pH 7.8 with 10 mM NaN 3 , incubated for 1 hr in the dark, and then cleared with an ascending series of ethanol solutions (30%, 50%, 70%, 80% and 95%). This protocol was adapted from Hoffmann et al., 2005. Images of stained seedlings were captured with a dissecting microscope and camera. Hydrogen peroxide detection was performed as described in Wu et al., 2012 with the following modifications: seedlings of the indicated genotypes were collected and treated side-by-side. They were incubated for 3 hr in 0.1 mg ml −1 3,3-diaminobenzidine pH 3.8, and vacuum-infiltrated for 5 min. Tissue was incubated overnight in the dark and cleared with an ascending ethanol series (30%, 50%, 70%, 80% and 95%), then imaged as for superoxide staining above.

| Mating-based Split-Ubiquitin System
Physical interactions between MSL1, MSL2, and MSL3 were determined using the mating-based split-ubiquitin system described in Obrdlik et al. (2004). cDNAs encoding the mature version of each protein were cloned and recombined into the destination vector pEarleyGate103 (Earley et al., 2006) using LR Clonase II (Thermo Fisher Scientific). MSL sequences were PCRamplified from destination vectors using primers attB1-F (5′-ACA AGTTTGTACA A A A A AGCAGGCTCTCCA ACCACCATG-3′) and attB2-R (5′-TCCGCCACCACCAACCACTTTGTACAAGAAAG CTGGGTA-3′). PCR products were co-transformed with digested pMetYCgate (digested with PstI+HindIII) into yeast strain THY.AP4 (selected on Synthetic Complete media lacking leucine), and with digested pXNGate21-3HA (digested with EcoRI+SmaI) into yeast strain THY.AP5 (selected on Synthetic Complete media lacking tryptophan and uracil). pMetYCgate and pXNGate21-3HA were obtained from the Arabidopsis Biological Resource Center, https://abrc.osu.edu. Cells were mated for 2 days on Synthetic Complete media lacking Leu, Trp, and Ura for selection of diploids. Interactions between proteins were determined via growth after 3 days on Synthetic Minimal media lacking adenine, histidine, leucine, tryptophan, and uracil. This media was also supplemented with 150 μM Methionine.

Both crystallography and biochemical experiments establish that
EcMscS forms a homoheptameric mechanosensitive ion channel (Bass, Strop, Barclay, & Rees 2002;Miller et al., 2003). Each EcMscS which is proteolytically cleaved after organellar import. Similarly, it is likely that the chloroplast-targeting N-terminal peptides of MSL2 and MSL3 (indicated in green) are cleaved after directing preprotein to the chloroplast (Haswell & Meyerowitz, 2006).

| Loss of MSL1 exacerbates the leaf notching, rumpling, and variegation observed in msl2 msl3 double mutant plants
In order to address the possibility of coordination between plastids and mitochondria, we first investigated genetic interactions between MSL1, MSL2, and MSL3. To do so, we compared the whole seedling phenotypes of 24-day-old wild type plants, msl1 mutants, msl2 msl3 double mutants, msl1 msl2 msl3 triple mutants, and msl1 msl2 msl3 triple mutants complemented with a transgene containing a genomic copy of MSL1 (msl1 msl2 msl3 + MSL1g) ( Figure 2). As previously reported, msl2 msl3 plants exhibit leaf notching, rumpling and variegation (Wilson et al., 2011).
This effect was suppressed in msl1 msl2 msl3 + MSL1g seedlings, indicating that the increase in phenotypic severity in the msl1 msl2 msl3 triple mutant can be attributed to a defect at the MSL1 locus.

| msl1 msl2 msl3 triple mutants form shooty outgrowths in place of the ectopic calluses observed in msl2 msl3 double mutants
Since the msl1 lesion exacerbated leaf phenotypes in the msl2 msl3 background, we hypothesized that the same would be true for other msl2 msl3 phenotypes, including the production of meristematic callus previously observed in msl2 msl3 seedlings grown on solid media . Seedlings were grown vertically on solid media for 19-21 days at 21°C under a 16 hr light regime and the shoot apex examined (Figure 3). Under these conditions, msl1 seedlings were indistinguishable from the wild type and meristematic calluses were not observed in either background. Consistent with our earlier report, callus-like growth at the shoot apex was observed in ~70% of msl2 msl3 seedlings. Unexpectedly, no callus was formed F I G U R E 1 Predicted Topology of EcMscS, AtMSL1, and AtMSL2. Experimentally determined or predicted membrane topology of the indicated monomers. Each dot represents one amino acid. Amino acids corresponding to the conserved MscS Domain (as defined in Haswell, 2007)

| MSL1 is required for meristematic reactive oxygen species accumulation in the msl2 msl3 background
Double msl2 msl3 mutants accumulate the reactive oxygen species (ROS) superoxide (O 2 − ) and hydrogen peroxide (H 2 O 2 ) at the shoot apex at levels higher than the wild type .
To determine the role of MSL1 in the accumulation of ROS, seedlings were grown on solid media for 21 days and stained with 3,3′-diaminobenzidine (DAB, which indicates H 2 O 2 ) or nitrotetrazolium blue chloride (NBT, which indicates O 2 − ) ( Figure 4). As previously observed, levels of NBT and DAB were higher in msl2 msl3 mutants shoot apices than in the wild type. Single msl1 mutants were indistinguishable from the wild type. In the apices of msl1 msl2 msl3 triple mutants, DAB and NBT staining were greatly reduced compared to msl2 msl3 double mutants. In addition, strong meristematic DAB and NBT staining was recovered in msl1 msl2 msl3 + MSL1g plants, indicating that MSL1 is required for meristematic ROS accumulation in addition to callus formation in msl2 msl3 plants. These results are also consistent with our previous observation that callus formation in msl2 msl3 seedlings is dependent on O 2 − accumulation in the shoot apex , and we propose that the lack of callus formation in msl1 msl2 msl3 plants can be attributed to the absence of apical O 2 − accumulation when MSL1 is mutated. We note that NBT (but not DAB) staining in the cotyledons and leaves of the msl1 msl2 msl3 triple mutant was elevated compared to all other genotypes (Supporting Information Figure S2).

| msl2 msl3 mutants have shorter roots and few lateral roots per unit length, and MSL1 is partially required for these root defects
Only aerial phenotypes of the msl2 msl3 mutant have been documented. To begin to assess root phenotypes in this mutant, we grew seedlings vertically on solid media for 13 days. As shown in Figure 5, msl2 msl3 seedlings had primary roots averaging 1.4 cm in length, over four times shorter than Col-0 roots, which averaged 6.8 cm. phenotype, and appears to be involved in the reduction in lateral roots.

| MSL1, MSL2, and MSL3 interact in an organelle-specific manner
To begin to assess whether these genetic relationships might be mediated through direct protein-protein interactions, we used the mating-based split-ubiquitin system (mbSUS), a version of the classic yeast two hybrid modified for the analysis of membrane proteinprotein interactions (Grefen, et al., 2009;Obrdlik et al., 2004). In this assay, interactions between proteins are assessed by virtue of their ability to bring together two fragments of ubiquitin, Nub and Cub. When the two fragments are brought together, they catalyze the cleavage of an artificial transcription factor (LexA-VP16) that is translationally fused to Cub, thereby allowing activation of reporter genes. We tested mature (lacking transit peptides) versions of MSL1, MSL2, and MSL3 for interaction in this assay ( Figure 6).
Mating yeast strains expressing MSL1, 2, or 3-Cub-LexA to a strain expressing Nub WT , a version of Nub that does not require interaction However, MSL1 did not interact with MSL2 or with MSL3. Also, MSL3 did not interact with itself.

| D ISCUSS I ON
It has been proposed that plastids and mitochondria interact through signaling or metabolic pathways to coordinate cellular responses (Bobik & Burch-Smith, 2015), prompting us to initiate an analysis of the genetic and physical interactions between three members of the MscS-Like (MSL) family of mechanosensitive ion channels. These three proteins are localized to the mitochondria (MSL1, Lee et al., 2016) or to the chloroplast (MSL2 and MSL3, Haswell & Meyerowitz, 2006). While msl1 mutant plants have no obvious developmental phenotype, msl2 msl3 mutants exhibit crumpled, variegated, and notched leaves and after growth on solid media they produce callus at the shoot apex. Double msl2 msl3 mutants also accumulate ROS at the shoot apex. Here we document two additional phenotypes in the msl2 msl3 mutant, including a shorter primary root and reduced number of lateral roots than the wild type. In addition, we found that introducing the msl1 allele into the msl2 msl3 background exacerbated leaf phenotypes but ameliorated callus production, ROS accumulation, and the root phenotypes.
There are multiple molecular explanations for genetic interactions between proteins localized to different compartments.
One possibility is that they are actually not in different compartments; that MSL1 could move to the chloroplast or MSL2 can move to the mitochondrion. Dual targeting to both the mitochondria and the chloroplast has been observed for many plant proteins but is difficult to predict Xu, et al., 2013). We have not observed dual localization in our experiments with MSL1-, MSL2-, or MSL3-GFP fusion proteins, but it remains a possibility that protein levels below the level of detection are dual localized. We considered the possibility of the formation of heteromeric channels, which might explain cross-organelle effects with very low levels of dual-targeted proteins. However, in our mbSUS experiments, we did not observe any interactions between MSL1 and MSL2 or MSL3, though we did see robust interaction between MSL2 and MSL2, and also strong interaction between MSL2 and MSL3 ( Figure 6). Whether MSL3 forms a homomeric channel or is only able to form a heteromeric channel with MSL2 remains to be determined. Taken together, these data suggest that the observed genetic interactions between MSL1, MSL2, and MSL3 are unlikely to be mediated by direct protein-protein interactions.
Instead, the MSL1/2/3 genetic interactions we observed may reflect an interaction between two organelle signaling pathways that clear. In the case of the root phenotypes, we note that msl1 msl2 msl3 seedlings were larger than msl2 msl3 seedlings ( Figure 5). It is possible that the higher root length and number of lateral roots in msl1 msl2 msl3 may be an indirect effect of larger seedling size. Our current working hypothesis is that msl2 msl3 mutant plastids produce or potentiate an osmotic stress signal that requires MSL1 function in the mitochondria for its production or action. When MSL1 is absent, the osmotic stress signal generated in the plastids is not propagated, resulting in exacerbated (leaf morphology) or attenuated (callus, ROS, root) phenotypes.
With respect to callus production, one mechanism by which mitochondria might affect plastid osmotic stress signaling is through the modulation of ROS levels. We previously showed that ROS accumulation in the shoot apex leads to and is required for apical callus formation in msl2 msl3 plant . Furthermore, msl1 mutants show increased mito-roGFP signal in response to multiple abiotic stresses, indicating that the mitochondrial glutathione pool is oxidized under these conditions (Lee et al., 2016). We proposed that MSL1 is required to prevent over-reduction of the respiratory chain and ROS production under conditions of high membrane potential. The data presented here suggest that a mitochondrial signal associated with MSL1 functions upstream of meristematic superoxide accumulation and the production of callus that is caused by osmotically stressed plastids in the msl2 msl3 mutant. One possibility is that osmotically stressed plastids in some way induce a ROS-related stress signal in mitochondria, which turn leads to the accumulation of ROS in meristematic cells and the production of callus. We propose that in the absence of MSL1, the signal from plastids is not efficiently received or propagated, perhaps because msl1 mutant mitochondria are unable to normalize their own ROS levels and therefore have an abnormal response to a subtle ROS signal from osmotically stressed plastids.
The leaf morphology and leaf superoxide accumulation phenotypes of the msl2 msl3 double mutant were exacerbated in the msl1 msl2 msl3 background, while the meristematic ROS accumulation, callus production and root phenotypes were suppressed. One explanation for these differences may be variation in the metabolic coupling of plastids and mitochondria. Eliminating MSL1, and thereby disrupting mitochondrial redox homeostasis, may have different effects in source tissues that are actively photosynthesizing and photorespiring (such as leaves) and sink tissues (such as roots and meristems) that are not.
To summarize, we show here that the loss of MSL1 can attenuate or exacerbate the developmental effects of plastid osmotic stress observed in the msl2 msl3 mutant. We hypothesize a signaling relationship between the two organelles that impacts a range of developmental processes, from cell identity at the shoot apex to the elaboration of lateral roots. Additional experiments are needed to determine how osmotically stressed plastids lead to these developmental phenotypes, and why many of them are modulated by the presence of mitochondrial MSL1. JSL, MEW RR, and ESH analyzed data.