The Medicago truncatula DMI1 Protein Modulates Cytosolic Calcium Signaling

In addition to establishing symbiotic relationships with arbuscular mycorrhizal fungi, legumes also enter into a nitrogen-fixing symbiosis with rhizobial bacteria that results in the formation of root nodules. Several genes involved in the development of both arbuscular mycorrhiza and legume nodulation have been cloned in model legumes. Among them, Medicago truncatula DMI1 ( Doesn't Make Infections 1) is required for the generation of nucleus-associated calcium spikes in response to the rhizobial signaling molecule Nod factor. DMI1 encodes a membrane protein with striking similarities to the archaebacterial calcium-gated potassium channel MthK. The cytosolic C-terminus of DMI1 contains an RCK (regulator of the conductance of K + ) domain which in MthK acts as a calcium-regulated gating ring controlling the activity of the channel. Here we show that a dmi1 mutant lacking the entire C-terminus acts as a dominant-negative allele interfering with the formation of nitrogen-fixing nodules and abolishing the induction of calcium spikes by the G-protein agonist Mastoparan. Using both the full length DMI1 and this dominant-negative mutant protein we show that DMI1 increases the sensitivity of a sodium- and lithium-hypersensitive yeast mutant towards those ions and that the C-terminal domain plays a central role in regulating this response. We also show that DMI1 greatly reduces the release of calcium from internal stores in yeast, while the dominant negative allele appears to have the opposite effect. This work suggests that DMI1 is not directly responsible for Nod factor-induced calcium changes, but does have the capacity to regulate calcium channels in both yeast and plants. agonist Mastoparan. Interestingly, sensitivity of plant gene expression to Mastoparan is retained in dmi1 null mutants, putting the role of DMI1 in the Ca2+ oscillatory machinery into question. This study provides direct evidence that DMI1 is able to regulate the release of Ca 2+ from internal stores, both in the model legume Medicago truncatula and in Saccharomyces cerevisiae . We that the assays


INTRODUCTION
Nitrogen and phosphorus are essential macronutrients that frequently limit plant growth. In order to meet their phosphorus requirements many plants undergo symbiotic interactions with arbuscular mycorrhizal fungi. In addition, leguminous plants also establish a mutualistic symbiotic interaction with rhizobial bacteria that results in the formation of root nodules wherein atmospheric nitrogen is fixed by the bacteria and transferred to the plant.
The rhizobium-legume symbiosis is initiated by the production of bacterial lipochitooligosaccharidic signals, termed Nod factors, that are produced in response to flavonoids exuded by legume roots. The perception of Nod factors initiates downstream responses, such as specific ion fluxes, root hair deformations and induction of early nodulation (ENOD) genes (Oldroyd and Downie, 2004). Similarly, diffusible Myc factors are able to elicit symbiotic responses in host plants but the structures of these signals are unknown. Within 1-2 minutes of Nod factor application, a transient increase in cytosolic free calcium ([Ca 2+ ] cyt ) is detectable in root hair cells (Felle et al., 2000). This Ca 2+ transient has been shown to depend on Ca 2+ influx through the plasma membrane and is localized to the root hair tip (Shaw and Long, 2003). Following this initial [Ca 2+ ] cyt transient, repetitive nucleus-associated Ca 2+ oscillations are observed in the cytosol (Erhardt et al., 1996;Walker et al., 2000). It has been shown in a number of plant and animal systems that Ca 2+ oscillations encode information that can be translated into downstream responses (Dolmetsch et al., 1998;Allen et al., 2001) and there is also evidence that Ca 2+ spiking is relevant for the activation of downstream responses in nodulation signaling (Miwa et al., 2006;Sun et al., 2007).
Since Ca 2+ spiking appears to be a crucial component of the nodulation process, we were interested in how the Ca 2+ signal is generated. [Ca 2+ ] cyt elevations are mediated by the activation of Ca 2+ -permeable channels in membranes delineating compartments of higher [Ca 2+ ]. In contrast to animals, the molecular identity of Ca 2+ -permeable channels is largely unknown in plants. The sole exception is the TPC1 gene, encoding a Ca 2+ -activated and Ca 2+ -permeable channel in the vacuolar membrane (Peiter et al., 2005b). Because Nod factor-induced Ca 2+ spiking is confined to the perinuclear area, TPC1 is not likely to be involved in this response.
Genes for Ca 2+ -permeable channels in the plant ER or nuclear envelope have not been identified, but biochemical and biophysical studies on fractionated membranes have revealed a number of Ca 2+ conductances: Similar to the situation in some animals cells, Ca 2+ -permeable channels in the ER can be activated by cyclic ADP ribose (cADPR) (Navazio et al., 2001) and inositol 1,4,5-trisphosphate (InsP 3 ) (Muir and Sanders, 1997). Also, the NADP metabolite nicotinic acid adenine dinucleotide phosphate (NAADP) has been demonstrated to release Ca 2+ from plant ER-derived vesicles (Navazio et al., 2000). In addition, a voltage-dependent, Ca 2+ -sensitive, and Ca 2+ -selective channel has been identified in ER-derived membranes from Bryonia tendrils (Klüsener et al., 1995). However, such in vitro studies are unable to determine whether these conductances are actually present in the nuclear envelope, which is part of the ER fraction.
Although the molecular identities of the components that make up the Ca 2+ oscillator in the Nod factor-stimulated root hair are not known, pharmacological studies have provided some information on the mechanism of oscillations. Two inhibitors of InsP 3 -activated channels, TMB-8 and 2-APB, inhibited expression of ENOD11 (Charron et al., 2004), and 2-APB also blocked Ca 2+ spiking (Engstrom et al., 2002). Recent reports that the G-protein agonist Mastoparan and its synthetic analog Mas7 activate Ca 2+ oscillations and nodulation gene expression in a manner analogous to Nod factor-induced responses support the notion that Ca 2+ spiking may involve phospholipid signaling (Charron et al., 2004;Sun et al., 2007). Evidence for the contribution of type II A ATPases in the Ca 2+ spiking response stems from experiments demonstrating a block of the response by the specific inhibitor cyclopiazonic acid (Engstrom et al., 2002). The most likely scenario is therefore a release of Ca 2+ from the nucleus-associated compartment by ligand-activated channels followed by Ca 2+ retrieval by type II A Ca 2+ pumps.
Forward genetic studies in model legumes, such as Medicago truncatula and Lotus japonicus have revealed genes that are required for both legume nodulation and Nod factor signaling. Among them, the DMI (Doesn't Make Infections) genes play a very early role in Nod factor signaling and are also necessary for the establishment of arbuscular mycorrhization, indicating a common signaling pathway. dmi1 mutants are affected in many responses to Nod factors and particularly in the Nod factorinduced Ca 2+ spiking response (Wais et al., 2000). Cloning of DMI1 revealed that it encodes a putative ion channel with strong homologies to the MthK Ca 2+ -gated K + channel (Ané et al., 2004). However, in contrast to the MthK channel, the putative filter region of the DMI1 protein does not contain the GYGD motif found in K + -7 selective channels, leaving open the possibility that DMI1 may constitute a ligandactivated Ca 2+ channel. In favor of this idea, DMI1 contains an RCK domain that may be involved in the binding of Ca 2+ and/or other unknown ligands (Jiang et al., 2002).
Subsequently, two proteins homologous to DMI1 were identified from L. japonicus: CASTOR and POLLUX (Imaizumi-Anraku et al., 2005). Unexpectedly, fusions of both proteins with GFP (Green Fluorescent Protein) revealed localization to plastids in onion cells and pea roots, which is difficult to reconcile with an involvement in nucleus-associated Ca 2+ spiking. However, it has recently been demonstrated for DMI1 that this protein is localized in the nuclear envelope (Riely et al., 2007). The facts that DMI1 is essential for Nod factor-induced Ca 2+ spiking and localizes to the nuclear envelope around which the major Ca 2+ changes occur made it a strong candidate for playing a direct role in the formation of Ca 2+ spikes. Analysis of dmi1 mutants, however, suggested that DMI1 is not required for Mastoparan-induced Ca 2+ spiking and ENOD11 expression (Charron et al., 2004;Sun et al., 2007). Because Mastoparan and Nod factor are likely to activate the same Ca 2+ oscillatory mechanism, DMI1 is unlikely to function as the Ca 2+ channel directly mediating Nod factor-induced Ca 2+ oscillations.
To elucidate the role of DMI1 in Nod factor signaling we extended the analysis of dmi1 mutant alleles. Here we show that dmi1-2, a nod-mutant that lacks the entire C-terminus (including the RCK domain) but maintains the full channel domain, appears to act as a dominant-negative allele. Our first indication of this was the finding that dmi1-2 inhibits Mas7-induced Ca 2+ oscillations. We have validated the dominant negative nature of dmi1-2 by showing that this allele interferes in the ability of wild-type DMI1 to activate appropriate nodulation. In order to understand better the function of DMI1 we used yeast as a heterologous expression system. When expressed in yeast, DMI1 and dmi1-2 respectively induce hypersensitivity and hypertolerance to high Li + and Na + concentrations. The notion that this phenotype is caused by altering intracellular Na + and Li + sensitivity rather than ion accumulation is supported by the facts that DMI1 is targeted to the yeast ER and that both the fulllength and the truncated protein alter a Li + -sensitive hexose-induced [Ca 2+ ] cyt transient originating from internal stores. The combination of mutant analysis and heterologous expression in this study provides further evidence that DMI1 is unlikely to function as an oscillatory Ca 2+ channel, and demonstrates that DMI1 has the capacity to regulate Ca 2+ release channels in both yeast and plants. We hypothesize that DMI1 may function in symbiosis signaling by regulating Ca 2+ channel activity following Nod factor and Myc factor perception.

dmi1-2 Suppresses Mas7-Induced Calcium Oscillations
The G-protein agonist Mastoparan, which originates from wasp venom, and its synthetic analog Mas7 have been shown to induce Ca 2+ oscillations and nodulation gene expression in M. truncatula root hair cells in a manner analogous to Nod factorinduced responses (Pingret et al., 1998;Sun et al., 2007). Mas7-induced Ca 2+ oscillations and gene induction do not require NFP, DMI1 or DMI2 that are necessary for Nod factor-induced Ca 2+ spiking (Charron et al., 2004;Sun et al., 2007). We spiking. This is consistent with what we have shown in earlier studies using the dmi1-1 allele (Sun et al., 2007), that contains a premature stop codon within the RCK domain ( Fig. 1A). In addition we saw Mas7-induced Ca 2+ oscillations in the dmi1-4 allele that contains a premature stop codon within the channel region ( Fig. 1A, B).
Surprisingly, in the dmi1-2 mutant Mas7-induced Ca 2+ spiking was not observed (Fig.   1B), suggesting that this allele was interfering with the Mas7 mode of action. dmi1-2 contains a premature stop codon soon after the last transmembrane domain of the channel region, thus the mutant protein should have a complete channel domain, but will lack the entire cytosolic C-terminus including the RCK domain. This mutation does not affect the transcript level (Supplemental Fig. S3). Since the RCK domain represents the regulatory component of the MthK channel, we hypothesize that the product of dmi1-2 may represent an unregulated channel. The fact that dmi1-2 could suppress Mas7 action indicated that this mutation may act as a dominant-negative allele.

dmi1-2 Interferes with DMI1 During Nodule Development
In order to assess further the effect of dmi1-2 we attempted complementation of this mutant with the wild type gene. Transformation of DMI1 under the control of its native promoter and inoculation with S. meliloti led to nodule formation in hairy roots of both the dmi1-3 and dmi1-2 mutants ( Table 1). The transformed roots of the dmi1-2 mutant produced about half as many nodules as those on the dmi1-3 mutant; this was due to a decrease (from 90% to 70%) in the numbers of transformed roots producing nodules and a decrease (of about 40%) in the numbers of nodules formed on those roots that did produce nodules. Furthermore, there was a mixed population of nodules on complemented dmi1-2 roots, with most nodules being white and appearing ineffective, although occasionally pink nodules did form ( Fig. 2A). In contrast complemented dmi1-3 roots predominantly had pink nodules. We separately analyzed acetylene reduction in pink and white nodules from the complemented dmi1-2 plants. Overall, acetylene reduction was greatly reduced in complemented dmi1-2 compared with complemented dmi1-3 roots (Fig. 2B) and the frequency of nodules capable of reducing acetylene was also much lower with the complemented dmi1-2 roots (Fig. 2B). Microscopy of white nodules on complemented dmi1-2 roots revealed that these nodules predominantly lacked bacteroids within the cells of the nitrogen-fixing zone of the nodule (Fig. 2C, 2E), but did show swollen infection threads containing many bacteria (asterisked in Fig. 2E). In contrast the few pink nodules on complemented dmi1-2 plants were able to reduce acetylene and showed many infected plant cells (arrowed in Fig. 2D) and normal bacteroid development Based on the fact that dmi1-2 interferes with Mas7-induced Ca 2+ spiking and cannot be fully complemented by DMI1 we conclude that dmi1-2 probably acts as a dominant-negative allele with the capacity to interfere with Ca 2+ release in plant cells.

DMI1 Exacerbates the Sensitivity of a Na + -and Li + -Sensitive Yeast Mutant
Because it was difficult to infer a function of DMI1 from in planta studies, we chose to express DMI1 in a heterologous system. Budding yeast (Saccharomyces cerevisiae) has been successfully used in the past to characterize plant ion channels and transporters. As DMI1 shares strong homology to cation channels, we expressed DMI1 in yeast mutants defective in cation homeostasis.
The cch1∆mid1∆ mutant is unable to grow on Ca 2+ -depleted media because it lacks a plasma membrane Ca 2+ channel (Fischer et al., 1997), whereas the trk1∆trk2∆tok1∆ mutant requires high K + concentrations in the medium due to the lack of high-affinity K + uptake systems (Bertl et al., 2003). Expression of DMI1 did not suppress the phenotypes of either strain (Supplemental Figure S1).
Unlike plant cells, wild type yeast tolerates very high levels of Na + in the medium. This is mainly due to the presence of four Na + /Li + pumps, ENA1 -ENA4, in the yeast plasma membrane, facilitating a highly efficient removal of Na + and Li + from the cytosol (Quintero et al., 1996). Accordingly, an ena1∆-ena4∆ deletion mutant, G19, is hypersensitive to Na + and Li + (Quintero et al., 1996). As Li + can affect the same molecular targets as Na + , sensitivity to both ions is commonly linked (Murguia et al., 1995). Figure 3A shows that expression of the full-length DMI1 cDNA in the G19 mutant resulted in a further increase of Na + and Li + sensitivity of this strain. The increased Na + sensitivity was not due to osmotic effects because the DMI1expressing G19 was not more sensitive to 400 mM K + or 800 mM mannitol (data not shown). DMI1 also increased the Li + sensitivity of the wild type strain W303 (data not shown).
Given our interpretation that the dmi1-2 allele acts in a dominant-negative manner in planta, we tested the effect of dmi1-2 cDNA expression on Li + and Na + sensitivity of yeast. As shown in Figure 3A, Li + sensitivity, and to a lesser extent Na + sensitivity, of the G19 strain were decreased in dmi1-2 transformants. This provides further evidence that the dmi1-2 protein is partially functional and suggests that dmi1-2 acts opposite to wild type DMI1.
The decreased Li + /Na + sensitivity of the dmi1-2-expressing yeast is difficult to reconcile with a location of this protein in the yeast plasma membrane. Moreover, we observed that Na + accumulation from medium containing 10-200 mM Na + was not affected by DMI1 expression (Fig. 3B), which suggests that DMI1 does not alter Li + /Na + conductance in the yeast plasma membrane. Thus, DMI1 may affect the internal sensitivity of yeast cells to Li + and Na + .

DMI1 Localizes to the Endoplasmic Reticulum in Yeast
To localize DMI1 in yeast we first constructed C-and N-terminal fusions of DMI1 with GFP. Yeast cells expressing DMI1::GFP or GFP::DMI1 showed a punctate pattern of GFP fluorescence in the cytoplasm (not shown). However, neither construct produced the Na + -or Li + -hypersensitive phenotypes observed in DMI1expressing G19 yeast, indicating the fusion proteins were either alternatively targeted or not functional (data not shown). Therefore we localized DMI1 by indirect immunofluorescence microscopy. Probing of fixed, permeabilized DMI1-expressing yeast cells with an antiserum raised against N-terminal peptides of DMI1 stained the yeast cells in an ER-like pattern ( Fig. 4 A,B). Fluorescence was absent in vector control cells (not shown). The precise localization of DMI1 in yeast was determined by immunogold electron microscopy of high pressure frozen / freeze substituted yeast cells, a method that maintains the integrity of subcellular compartments and sample antigenicity. As shown in Figure 4D, immunogold label was most prominent along the ER and to a lesser degree at the nuclear envelope, consistent with the immunofluorescence observations ( Fig. 4B) and reflecting the continuity of these two compartments. In control cells, weak background labeling was observed on cell walls and in vacuoles, but notably absent from the ER and nuclear envelope (Fig. 4C).

Transient in Yeast
We hypothesized that DMI1 might exacerbate Li + sensitivity by interfering in a Moreover, these Ca 2+ phenotypes are independent of Li + and Na + . Taken together, these results suggest that the altered Li + /Na + sensitivity associated with DMI1 and dmi1-2 is mediated by an effect on [Ca 2+ ] cyt homeostasis.

DISCUSSION
The aim of this study was to investigate further the role of DMI1 in Nod factor signaling by combining mutant analysis in planta with bioassay of ion homeostasis in the heterologous yeast system. DMI1 is required for the generation of Nod factor-

dmi1-2, a semi-dominant negative allele
Our Ca 2+ imaging experiments show that, unlike other dmi1 mutant alleles, the dmi1-2 allele fails to elicit the Mas7-induced Ca 2+ response, suggesting that dmi1-2 interferes with Ca 2+ channel activation. This mutant allele also interferes with the development of functional (nitrogen-fixing) nodules following complementation with the wild-type DMI1 gene. Taken together, these observations suggest a direct role of DMI1 in both Nod factor-induced Ca 2+ spiking and late nodule development. In The dmi1-2 gene product lacks the entire cytosolic C-terminus, including the RCK domain. In the bacterial potassium channel MthK, the RCK domain regulates channel activity in response to Ca 2+ or other ligands, such as NAD (Jiang et al., 2002). We hypothesize that the dmi1-2 protein forms a deregulated channel subunit that may affect the activity of other channels (see below). This is supported by our yeast experiments: Whereas DMI1 suppresses a hexose-induced Ca 2+ transient in yeast (and this is correlated with enhanced sensitivity to Na + and Li + ), the expression of dmi1-2 in yeast has the opposite effect, extending the hexose-induced Ca 2+ transient and increasing resistance to high concentrations of Na + and Li + . The fact that dmi1-2 had opposite effects to DMI1 in yeast is also a strong indication that the unusual effects of dmi1-2 in M. truncatula are due to this allele rather than occasional second-site mutations that might persist in the back-crossed dmi1-2 line. Although the initial description of dmi1 mutants describes dmi1-2 as a recessive allele (Catoira et al., 2000), this characterization was based on a presence/absence scoring of  The dominant negative nature of the dmi1-2 allele is likely due to the formation of multimers between DMI1 and dmi1-2. Cation channels homologous to DMI1 (e.g. MthK) consist of four pore-forming subunits (Jiang et al., 2002), and an identical structure is to be expected for DMI1. Heteromers consisting of DMI1 and dmi1-2 subunits are likely to be defectively regulated, because their gating ring is incomplete.
Dosage-dependent negative dominance of a cation channel mutation in plants has also been demonstrated for the tetrameric Shaker-like K + channel KAT1 (Kwak et al., 2001).

DMI1 effects on Na + /Li + sensitivity and Ca 2+ signaling in yeast
Expression of DMI1 and dmi1-2 in yeast resulted in increased and decreased Na + /Li + sensitivity, respectively. The Na + accumulation data and the localization of DMI1 to yeast ER and nuclear membranes argue in favor of a model wherein DMI1 alters the sensitivity to intracellular Na + /Li + in yeast, rather than altering Na + /Li + accumulation. Li + effects have been extensively studied in mammals, and to some extent also in yeast. In mammals, Li + inhibits glycogen synthase kinase-3 (GSK-3) and inositol phosphatases. Inhibition of the latter causes a disruption of the inositol cycle and thereby decreased levels of cellular inositol trisphosphate (InsP 3 ) and disturbances in [Ca 2+ ] cyt signaling (Williams and Harwood, 2000). Inositol phosphatase homologs in plants and yeast have also been demonstrated to be sensitive to Li + , and to a lesser extent also to Na + (Murguia et al., 1995;Quintero et al., 1996;Lopez et al., 1999). The yeast genes IMP1 and IMP2 encode Li + -and Na + - in InsP 3 phosphorylation (Tisi et al., 2002;Tisi et al., 2004). In agreement with a critical role of the inositol cycle in hexose sensing, Li + has been shown to affect the hexose-induced [Ca 2+ ] cyt transient (Csutora et al., 2005). However, this may also be attributed to the inhibition of phosphoglucomutase (PGM), a further Li + -sensitive yeast enzyme which is also required for the hexose-induced [Ca 2+ ] cyt response (Fu et al., 2000;Masuda et al., 2001).
Irrespective of the primary Li + target, it is most likely that Li + sensitivity in yeast is at least partially due to disturbances in [Ca 2+ ] cyt signaling and that DMI1 and dmi1-2 also interfere in this signaling network. The results of our luminometric assays clearly support such a notion. We demonstrate that DMI1 inhibits and dmi1-2 prolongs the activity of the full-length or the truncated DMI1 proteins. In support of this idea, a mammalian homolog of DMI1, BKCa, has been demonstrated to be still conductive after removal of its entire C terminus (Piskorowski and Aldrich, 2002).

Conclusions
Nod factors induce rapid changes in both Ca 2+ concentration and gene expression. Mutations and inhibitors that abolish Nod factor-induced Ca 2+ spiking block gene induction, indicating a specific role for Ca 2+ spiking in signal transduction.
The duration and number of Ca 2+ spikes is critical for Nod factor-induced ENOD11 expression in root hair cells (Miwa et al., 2006). A number of genes have been defined that are necessary for Nod factor-induced Ca 2+ spiking, including the putative cation channel DMI1. This protein localizes to the nuclear envelope of M. truncatula root hair cells and this localization correlates with the nuclear association of Ca 2+ spiking. We have shown previously that DMI1 is not necessary for Mas7-induced Ca 2+ oscillations, which appear to mimic Nod factor-induced Ca 2+ spiking (Sun et al., 2007). Here we demonstrate that dmi1-2, a dominant-negative allele, can interfere with Mas7-induced Ca 2+ spiking. This suggests that DMI1 is not the channel responsible for the oscillatory Ca 2+ response, but instead that DMI1 may act at a level close to that of Mas7 in the Nod factor signaling cascade. Our work in yeast indicates that DMI1 interferes with the release of Ca 2+ from ER-stores, which normally supply hexose-induced Ca 2+ transients. It is interesting to note that in both Nod factorinduced Ca 2+ spiking and hexose-induced [Ca 2+ ] cyt transients the Ca 2+ signal is dependent on phospholipid signaling. We propose that DMI1 is not the Ca 2+ channel responsible for Ca 2+ oscillations, but rather that DMI1 regulates these channel(s), perhaps by mobilizing an as yet undefined cation and thereby altering the membrane potential. The localization of DMI1 to the ER membrane in yeast and nuclear membrane in plants complicates a direct electrophysiological analysis of DMI1.
Nevertheless, analysis of mutant alleles, such as those employed here, offer insight into the mode of action of DMI1. Similar structure-function analyses should provide novel insights into the activation of Ca 2+ oscillations in plant systems.

Plant Material and Growth Methods
M. truncatula lines used were Jemalong A17 and homozygous lines of the dmi1-2 and dmi1-3 mutants that had been backcrossed once. Seeds were scarified in concentrated sulfuric acid for eight minutes, surface-sterilized in 12% sodium hypochlorite, imbibed in sterile water and plated on 1% deionized-water agar plates.
Seeds were subsequently stratified for 24 or 48 hours at 4 °C and germinated by incubating at room temperature overnight. All nodulation assays were performed using a suspension of approximately 10 7 cfu ml -1 of S. meliloti strain ABS7M (pXLGD4) as described previously. For Ca 2+ imaging seedlings were grown overnight on Buffered Nodulation Agar medium at pH 6.5 (Erhardt et al., 1996)  carries uidA under the 35S promoter and this was used to identify transformed roots as described by (Jefferson, 1987). Acetylene reduction measurements were carried out on 21 day old nodules as described by Somasegaran and Hoben (1994). Results were expressed as nanomoles ethylene produced per hour per nodule. Nodules were cut in half longitudinally and fixed, sectioned and stained as described previously in Lodwig et al. (2005).

DMI1 expression analysis
Root samples were collected from 7-day old M. truncatula seedlings grown on Fahraeus medium. Total RNA was extracted using QIAGEN RNeasy ® Plant Mini kit.
DNA contamination was removed using Ambion's DNA-free ® kit and total RNA was quantified using RiboGreen ® RNA quantification kit (Molecular Probes/Invitrogen). A total of 0.5 µg of RNA was used to synthesize cDNA, using oligo dT primers with SuperscriptIII ® cDNA synthesis kit of Invitrogen.

Calcium Imaging in Medicago Root Hairs
For the Ca 2+ analysis we microinjected root hair cells with Ca 2+ -responsive dyes as described by Erhardt et al. (1996), with slight modifications as described by
This pYES2:DMI1 construct was mutagenized using the QuickChangeII Site-Directed Mutagenesis Kit (Stratagene, La Jolla, CA) kit to mimic the dmi1-2 mutation. Yeast was transformed with plasmids using a simplified protocol (Elble, 1992).
Transformants were selected and maintained on synthetic complete (SC) agar plates containing 20 g L -1 glucose and lacking the appropriate selective markers (Sherman, 2002).
Growth assays on plates and in liquid were performed as described (Peiter et al., 2005a). For drop assays, yeast was grown overnight in selective SC medium containing 20 g L -1 galactose to a density of c. 10 7 cells mL -1 , centrifuged, resuspended in sterile water to 1 · 10 7 cells ml -1 , and diluted 10-, 100-and 1000-fold.

Na + Accumulation by Yeast
Yeast was grown shaking (180 rpm) at 30°C in SC-Ura (galactose) medium to early stationary phase (c. 5 · 10 7 cells ml -1 ). NaCl was added to 50 ml aliquots as indicated in Results. After 3 hours, cultures were centrifuged (5 min, 3200 g, 4 °C) and washed twice with 25 ml 10 mM CaCl 2 solution at 4 °C. Cell densities were determined using a haemacytometer (Improved Neubauer). Pellets were dried at 80 °C for 3 d, weighed and digested in HNO 3 using a microwave accelerated reaction system (Mars 5, CEM). Na + concentration in the digests was determined by atomic absorption spectroscopy (Spectr-AA20, Varian, Palo Alto, CA).

Production of Polyclonal Antisera
Soluble, tetrameric multiple antigenic peptides corresponding to the DMI1 amino acid sequence 65-FLGIGSTSRKRRQPPPPPSKPPVNLIPPHPR-95 coupled to a trilysine core (Tam, 1988) were synthesized at the Molecular Genetics Instrument Facility at the University of Georgia (Athens, GA). Polyclonal, anti-DMI1 antisera were generated in rabbit against this peptide by Chemicon International (Ramona, CA).

Indirect Immunofluorescence Microscopy of Yeast
Yeast strains were grown overnight in SC-Ura (galactose) medium to a density of c. 10 7 cells mL -1 and fixed by adding 1/10 volume of formaldehyde (37 %). h. Controls omitted either the primary antibodies or used the pre-immune serum.

[Ca 2+ ] cyt Determination in Yeast by Aequorin Luminescence
Yeast strains were transformed with a pEVP11AEQ plasmid carrying the APOAEQUORIN gene (Batiza et al., 1996) and maintained on selective SC plates.