Cod1p/Spf1p is a P-type ATPase involved in ER function and Ca²⁺ homeostasis

The ER is the site of essential cellular processes that require stringent regulation of lumenal ion levels. The ER also serves as a source of releasable Ca²⁺ that can be mobilized for various cellular demands. Consequently, the ion levels of the ER are under continuous, dynamic control in order to serve these essential functions. The ion levels of many membrane compartments are maintained by the action of a superfamily of ATP-dependent ion pumps known as P-type ATPases (Axelsen and Palmgren, 1998; Scarborough, 1999). These proteins catalyze ATP-dependent active transport of ions across numerous cellular membranes. The P-type ATPases that have been characterized are highly specific, transporting only a single ion, or sometimes two related ones. However, there are a large number of P-type ATPases identified solely by sequence, having unknown function and specificity, that await assignment of biochemical and physiological functions. In mammals, one mechanism of ER ion regulation occurs through the action of the sarcoplasmic/endoplasmic reticulum calcium ATPase (SERCA)^* pumps, which are localized to the ER and belong to the type IIa, calcium-specific subfamily of P-type ATPases (Carafoli and Brini, 2000; Axelsen and Palmgren, 2001). Although clearly important in animals, SERCA pumps are absent in other organisms, such as fungi, that nevertheless control ER ion content (Catty et al., 1997; Axelsen and Palmgren, 1998). Thus, it is likely that a more widely conserved P-type ATPase is responsible for ER ion maintenance in these organisms, and perhaps play an important, yet until now unsuspected, role in mammals as well.

Because fungi do not posses SERCA ATPases, the mechanisms by which calcium is supplied to the yeast ER had remained enigmatic. The absence of SERCA pumps led several groups to suggest that the yeast ER is supplied with calcium solely by the action of the Golgi-localized pump Pmr1p (Duerr et al., 1998; Marchi et al., 1999; Strayle et al., 1999). This model is reasonable because pmr1Δ mutants have a number of phenotypes related to the function of the ER (Duerr et al., 1998; Strayle et al., 1999). Like SERCA, Pmr1p belongs to the type IIa family of P-type ATPases and some phenotypes of pmr1Δ mutants can be complemented by expression of SERCA1a or related ATPases (Liang et al., 1997; Duerr et al., 1998; Talla et al., 1998; Degand et al., 1999). Pmr1p has been shown to transport both calcium and manganese, but how the Golgi apparatus–localized Pmr1p supplies calcium to the ER is unclear (Mandal et al., 2000; Wei et al., 2000).

We and others have recently identified another P-type ATPase in yeast, Cod1p/Spf1p, that adds a new dimension to the regulation of ER ion supply (Suzuki and Shimma, 1999; Cronin et al., 2000). Cod1p belongs to the type V subfamily of P-type ATPases that are broadly conserved in eukaryotes but for which no ionic substrates have yet been identified (Axelsen and Palmgren, 1998). Close homologues of Cod1p are present in Drosophila melanogaster, Caenorhabditis elegans, Dictyostelium discoideum, Arabidopsis thaliana, and humans (Costanzo et al., 2000). Cod1p is the only member of the type V subfamily for which any phenotypic information is known. We identified COD1 in a genetic screen for mutants incapable of regulating the degradation of the ER membrane protein Hmg2p (Cronin et al., 2000). The involvement of Cod1p in the degradation of an ER protein as well as other ER phenotypes implies that Cod1p functions in the ER (Suzuki and Shimma, 1999; Cronin et al., 2000). Furthermore, the cod1 mutant phenotype suggests a role for Cod1p in calcium regulation because the mutant phenotype could be partially suppressed by exogenous calcium (Cronin et al., 2000). These observations led us to propose that Cod1p functions to supply Ca²⁺ to the ER (Cronin et al., 2000).

In this paper, we directly examined the role of Cod1p in ER function and calcium regulation in vivo and the biochemical requirements for Cod1p ATPase activity in vitro. We have now demonstrated that Cod1p localized to the ER by both immunofluorescence and subcellular fractionation, and have provided phenotypic evidence supporting a role for Cod1p in ER function and cellular calcium regulation. Further, we have distinguished phenotypically the roles of Cod1p and Pmr1p in ER function and quality control. Finally, we purified the Cod1 protein and directly demonstrated the ATPase activity of Cod1p in vitro. Taken together, our data indicated that Cod1p plays a significant role in ER function and cellular Ca²⁺ homeostasis that is not equivalent to, nor redundant with, the role of Pmr1p or other P-type ATPases. Moreover, the presence of Cod1p homologues in the genomes of all sequenced metazoans suggested that the function of Cod1p is likely to be more widely conserved in eukaryotes than that of more restricted, but better known, pumps such as the SERCA ATPases.

Many of the phenotypes of the cod1 mutants suggested that Cod1p functions in the ER (Suzuki and Shimma, 1999; Cronin et al., 2000). Accordingly, we determined the localization of Cod1p by both immunofluorescence microscopy and density gradient fractionation (Fig. 1). To facilitate localization, we constructed the 3MYC::COD1 allele of COD1 expressing a protein with three myc epitopes fused to the amino terminus of the protein. 3myc–Cod1p was expressed from the unmodified COD1 promoter in single copy, preventing aberrant localization due to overexpression. Strains expressing 3myc–Cod1p are phenotypically wild type in all ways tested (unpublished data). The pattern of staining for 3myc–Cod1p was typical for ER localization, with a distinct ring around the nucleus and additional staining near the cell periphery. The distribution of Cod1p, though slightly more extensive, overlapped considerably with the distribution of Kar2p, a protein residing in the lumen of the ER (Fig. 1 A) (Rose et al., 1989; Preuss et al., 1991). We also saw the same pattern of perinuclear and peripheral staining with Sec61p, a component of the ER translocation apparatus (Fig. 1 B) (Stirling et al., 1992).

We independently tested the localization of 3myc–Cod1p by fractionation of intracellular membranes on sucrose density gradients in the presence of EDTA or magnesium as described in the Materials and methods (Fig. 1, C and D). A characteristic of ER microsomes is that their buoyancy in a sucrose gradient depends on the presence or absence of magnesium in the gradient, allowing unambiguous assignment of a membrane protein to the ER by testing the behavior of the protein under both conditions (Roberg et al., 1997). The distribution of 3myc–Cod1p in all gradients closely overlapped the distribution of the ER-localized Sec61p. When fractions were prepared with 2 mM MgCl₂, the ER containing 3myc–Cod1p and Sec61p cofractionated with the plasma membrane marker Gas1p (Doering and Schekman, 1996) and separated from the Golgi apparatus–derived membranes labeled by Emp47 (Schröder et al., 1995). When fractions were prepared with EDTA, 3myc–Cod1p and Sec61p cofractionated with Emp47p and separated from Gas1p, thus confirming that the fractions containing 3myc–Cod1p were derived from the ER.

Calcium concentration within the cytosol and secretory pathway regulates the expression of a number of genes, and mutations that perturb cellular calcium induce compensatory changes in gene expression (Garrett-Engele et al., 1995; Matheos et al., 1997; Locke et al., 2000). For instance, deletion of PMR1 produces a five- to sevenfold increase in the transcription of PMC1, a vacuolar calcium ATPase (Marchi et al., 1999, Locke et al., 2000). We investigated the effect of cod1 deletion on the expression of FKS2, PMC1, and ENA1, genes known to be regulated in response to calcium concentration, using previously described β-galactosidase reporter plasmids (Fig. 2, A and B) (Matheos et al., 1997; Locke et al., 2000). The expression of each of the three calcium-activated reporters was three- to sevenfold higher in the cod1Δ mutant than in wild-type cells grown under normal laboratory conditions and in the absence of any external calcium stimulus. This effect of cod1Δ was similar to that reported earlier for pmr1Δ (Locke et al., 2000). Importantly, transient treatment with 200 mM CaCl₂ induced β-galactosidase activity to the same level both in wild-type and cod1Δ cells, indicating that deletion of COD1 did not interfere with the ability of the reporter constructs to respond to calcium stimulation (unpublished data). Likewise, treatment of cultures with the calcineurin inhibitor cyclosporin A diminished expression of the reporter genes, indicating that the increased expression in cod1Δ was largely calceneurin dependent (Fig. 2 B; unpublished data). A portion of the increase may be calcineurin independent, as the treatment did not reduce expression to the same level as in wild type. The increased β-galactosidase activity in the cod1Δ mutants was specific to the calcium-regulated reporters. Expression of the CYC1–β-galactosidase reporter, which is not responsive to calcium (Matheos et al., 1997), did not differ significantly between wild-type and cod1Δ strains (Fig. 2 A).

Hypersensitivity to an extracellular ion can indicate a defect in intracellular transport of that ion. For example, mutants defective in the calcium and manganese pump Pmr1p are hypersensitive to the presence of calcium and manganese in the growth medium (Lapinskas et al., 1995; Wei et al., 2000) (Fig. 3 A). Thus, we tested the ability of the cod1Δ null mutant to grow in high levels of extracellular calcium or other cations. We found that cod1Δ mutants had over twice the sensitivity to calcium as wild-type cells, sensitivity equivalent to that of pmr1Δ mutants. The combined deletion of both cod1 and pmr1 further increased sensitivity to extracellular calcium to four times that of wild-type cells. The defect was specific to calcium, as cod1Δ mutants were no more sensitive to other divalent ions than wild-type cells (unpublished data). For example, deletion of COD1 did not affect manganese sensitivity, which was greatly increased in pmr1Δ strains, and the hypersensitivity of a pmr1Δ null mutant to manganese was not further exacerbated by a cod1Δ allele (Fig. 3 A).

The increase in expression of calcium-regulated genes and hypersensitivity to calcium suggested that Cod1p might influence total cellular ion content under some conditions. For example, deletion of PMR1 results in increased cellular calcium and increased cellular manganese (Lapinskas et al., 1995) (Fig. 3 B). We examined the level of a number of ions in wild-type and cod1Δ cultures by inductively coupled plasma optical emission spectrometry (ICP-OES) as described in the Materials and methods (Fig. 3 B). Deletion of COD1 alone did not affect cellular calcium levels. However, deletion of both COD1 and PMR1 produced a synergistic increase in intracellular calcium compared with cod1Δ or pmr1Δ alone. This synergistic effect was specific to calcium: although pmr1Δ cells also displayed increased manganese levels, the combined deletion of cod1 and pmr1 did not further increase total cellular manganese (Fig. 3 B). Deletion of COD1 either alone or in the presence of pmr1Δ had no effect on the total cellular concentration of magnesium, iron, zinc, copper, or any other ion detected in yeast cultures, demonstrating that the effect of cod1Δ was specific.

The localization of Cod1p to the ER and the previously reported phenotypes of the cod1 mutants suggested that Cod1p plays a major role in the function of the ER. We tested the sensitivity of cod mutants to the reducing agent DTT and the glycosylation inhibitor tunicamycin, both of which (through distinct mechanisms) alter protein folding in the ER (Fig. 4 A). cod1-1 cells had twice the sensitivity to tunicamycin as wild-type cells, but maintained normal sensitivity to DTT. Treatment of wild-type cells with either agent provokes the accumulation of unfolded proteins in the ER, inducing the unfolded protein response (UPR) (Frand and Kaiser, 1998; Pollard et al., 1998). We measured the unfolded protein response with a UPR reporter plasmid (4xUPRE–green fluorescent protein [GFP]) consisting of four unfolded protein response elements driving expression of the GFP coding region (4xUPRE; Fig. 4 B; Pollard et al., 1998). Strikingly, the GFP reporter in the cod1Δ mutant was constitutively expressed at 10 times the level of expression in wild-type cells in the absence of any additional stimulus. Furthermore, in cod1Δ, maximum expression of the UPR reporter required only half the dose of tunicamycin (2 μg/ml) needed to stimulate maximum expression in wild-type cells. As expected from the growth sensitivity experiments, the cod1Δ mutant displayed wild-type sensitivity to DTT in the unfolded protein response. Maximal induction of the reporter protein required 2 mM DTT in both wild-type and cod1Δ.

The unfolded protein response is mediated by the transcription factor Hac1p (Cox and Walter, 1996; Nikawa et al., 1996). We tested the necessity of the UPR in the cod1 mutant by crossing a cod1-1 null mutant to a hac1Δ mutant (Fig. 4 C). Sporulation of the diploid and dissection of 22 of the resulting ascospores revealed the synthetic lethality of hac1Δ cod1-1 double mutants. We confirmed this result using cod1Δ::Kan^R and hac1Δ::URA3 alleles. In 19 tetrads, no colonies with the Ura⁺, Kan^R phenotype were observed. We obtained identical results using cod1Δ::HIS3 and hac1Δ::URA3 null alleles in the unrelated MYY290 genetic background (14 tetrads analyzed). Microcolonies, three to five cells in diameter, were seen by microscopy on the dissection plate, but these did not grow further even after prolonged incubation. The presence of the microcolonies indicated that the cod1Δ, hac1Δ spores were capable of germinating but incapable of sustained growth. In contrast, double mutants of hac1 and YOR291w, the closest yeast homologue to COD1, were viable (unpublished data).

pmr1 mutants have phenotypes indicative of a perturbed UPR, including hypersensitivity to DTT and tunicamycin, inositol auxotrophy, and increased expression of the ER chaperone Kar2p (Duerr et al., 1998). We examined the viability of a pmr1Δ hac1Δ double mutant in the same genetic background in which cod1Δ hac1Δ cells were inviable (Fig. 4 D). Surprisingly, the pmr1Δ hac1Δ double mutants were viable. The pmr1Δ hac1Δ double mutant retained phenotypes of both parent strains: hypersensitivity to EGTA (pmr1Δ) and inositol auxotrophy (hac1Δ). The cod1Δ mutant was prototrophic for inositol in all strains tested (unpublished data). Thus, although Cod1p and Pmr1p had synthetic effects on whole cell calcium that implied shared functions, they also clearly had distinct, nonoverlapping ER functions.

The constitutive activation of the UPR and the constitutive degradation of Hmg2p in the cod1Δ mutant raised the possibility that UPR played a role in regulating Hmg2p degradation (cod1Δ mutants cannot regulate the degradation of Hmg2p (Cronin et al., 2000). In wild-type cells, treatment with lovastatin (a 3-hydroxy-3-methylglutaryl coenzyme A reductase [HMGR] inhibitor) slows Hmg2p degradation and raises Hmg2p levels (Hampton and Rine, 1994). This regulation can be measured by flow cytometry using the Hmg2p–GFP reporter protein (Hampton et al., 1996; Gardner and Hampton, 1999). We tested the effect of UPR activation on the regulated degradation of the Hmg2p–GFP by treatment with the glycosylation inhibitor tunicamycin (Fig. 5 A). Hmg2p–GFP degradation in cells treated with 4 μg/ml tunicamycin (a dose that fully induces the UPR) still responded to treatment with lovastatin in the same manner as cells grown without tunicamycin. In fact, Hmg2p–GFP degradation was moderately slowed by tunicamycin treatment. Thus, the unregulated, constitutive degradation of Hmg2p present in cod1 mutants cannot be attributed to activation of the unfolded protein response or to defective glycosylation. We deleted HAC1 to determine if the unfolded protein response was required for regulation of Hmg2p degradation (Fig. 5 B). The regulated degradation of Hmg2p–GFP in hac1Δ cells treated with lovastatin was indistinguishable from that seen in wild-type cells under the same treatment despite reports that some degradation substrates require the UPR for ER-associated degradation.

Genes closely homologous to COD1 have been identified in several metazoans including humans. These putative orthologues are more similar to COD1 than COD1 is to its closest yeast homologue, YOR291w. For example, D. melanogaster expresses a gene GH06032, which we refer to as dCOD1, with 40% identity and 58% similarity in the protein sequence to Cod1p (Costanzo et al., 2000). We constructed a modified version of dCOD1 for expression in yeast, encoding dCod1p with a hemagglutinin (HA) epitope tag fused to the amino terminus. We expressed HA–dCod1p from several different promoters in the cod1Δ mutant to test its ability to complement the mutant phenotype. cod1Δ mutants cannot regulate the degradation of the HMGR isozyme Hmg2p (Cronin et al., 2000) (Fig. 6). This defect in the regulated degradation of Hmg2p causes cod1Δ mutants to be hypersensitive to the HMGR inhibitor lovastatin in the appropriate genetic background. Expression of dCOD1 from the strong TDH3 or PMA1 promoters rescued growth of cod1Δ mutants on lovastatin, whereas expression of dCOD1 from the weaker COD1 promoter was not sufficient to restore lovastatin resistance (Fig. 6). As expected from the reversal of drug sensitivity, dCOD1 also restored regulation of Hmg2p degradation, as measured by flow cytometry using the Hmg2p–GFP reporter protein (Fig. 6 B).

Complementation of cod1Δ phenotypes was specific for P-type ATPases closely related to Cod1p. We tested the ability of the type IIa P-type ATPases SERCA1a and ACA3 (from A. thaliana) to complement the cod1 mutant phenotype with respect to Hmg2p–GFP degradation. Both of these genes complement the pmr1 mutant phenotype, but neither could complement the cod1 mutant phenotype, nor could overexpression of PMR1 (Fig. 6 B; unpublished data). Similarly, overexpression of HA::COD1 or HA::dCOD1 could not complement the sensitivity of pmr1Δ mutants to EGTA, suggesting that the roles of Cod1p and Pmr1p in ER function do not overlap significantly (unpublished data). The lack of cross-complementation also raises the possibility that the similar effects of pmr1 and cod1 on total cellular calcium may arise through different mechanisms.

To directly study the biochemistry of Cod1p, we constructed a gene that expressed a modified version of the Cod1 protein with nine histidine residues and an HA epitope tag inserted just after the amino-terminal methionine. The plasmid complemented the cod1Δ mutant phenotype (Fig. 7). Cells expressing the histidine-tagged Cod1 protein from the native COD1 promoter at single copy or the much stronger TDH3 promoter from a 2μ plasmid were wild type for regulating Hmg2p–GFP degradation (Fig. 7; unpublished data). We also constructed a derivative of this plasmid expressing the mutant D487N-Cod1p. In this mutant protein, the aspartyl phosphorylation site strictly conserved in all P-type ATPases was replaced by an asparagine residue. The plasmid expressing the D487N-Cod1p was unable to complement the cod1Δ null phenotype as expected from previous studies (Fig. 7; Suzuki, 2001).

We purified wild-type histidine-tagged Cod1p and the nonfunctional D487N protein using nickel affinity chromatography as described in detail in the Materials and methods section (Fig. 8, A and B). The purification yielded a single major band corresponding to the tagged Cod1p. Minor products detected in the elutions were the same in the purifications of both wild-type and the D487N mutant proteins.

We tested the purified proteins for the ability to hydrolyze ATP using a sensitive radiochemical assay described previously (Bais, 1975; Ghosh et al., 1990; Mandal et al., 2000). The assay measures the free inorganic phosphate produced from [γ³²-P]ATP, taking advantage of the fact that activated charcoal will sequester unhydrolyzed ATP but not the liberated ³²Pi. Wild-type Cod1p hydrolyzed ATP in a time-dependent manner, with free phosphate increasing linearly with time (Fig. 8 C). In contrast, the mutant D487N protein produced no detectable ATPase activity (Fig. 8 C). The reaction catalyzed by the eluted Cod1p was dependent on the concentration of ATP and could be modeled using Michaelis-Menten kinetics (Fig. 8 D). In the presence of 5 mM magnesium, Cod1p exhibited a K_m of ∼15 μM for ATP hydrolysis and a V_max of ∼150 nmol Pi/min/mg. As with other P-type ATPases, the activity of Cod1p was inhibited by 500 μM vanadate (unpublished data).

It has previously been demonstrated that the ATPase activity of P-type pumps can be stimulated by the presence of its ionic substrates. Thus, stimulation of ATPase activity by an ion normally indicates that the ion is transported by the ATPase. For example, the ATPase activity of the Golgi apparatus–localized Pmr1p is stimulated with calcium and manganese, the substrates it transports (Mandal et al., 2000). Accordingly, we tested calcium to see if it would stimulate the ATPase activity of Cod1p (Fig. 9, A and B). Surprisingly, calcium failed to increase the ATPase activity of Cod1p. In fact, higher concentrations of calcium (>10 μM free calcium) inhibited the activity of Cod1p (Fig. 9, A and B). We next tested various biologically relevant cations with the aim of identifying which ion was the substrate of Cod1p. The range of ions tested included Ca²⁺, Sr²⁺, Ba²⁺, Mn²⁺, Fe²⁺, Fe³⁺, Co²⁺, Ni²⁺, Cu²⁺, Zn²⁺, and Cd²⁺, none of which stimulated activity above the level seen with magnesium alone (Fig. 9, A and B; unpublished data). Replacing the K⁺ in the assays with Li⁺, Na⁺, Rb⁺, or Cs⁺ did not affect the activity of the protein (unpublished data). To ensure that trace amounts of any possible contaminating cations were not stimulating the activity of Cod1p, we assayed the ATPase activity in the presence of increasing concentrations of the chelators TPEN, 1,10-phenanthroline, and EDTA (Fig. 9 C). TPEN and 1,10-phenanthroline have been used to remove metals such as zinc from metalloenzymes (Arslan et al., 1985; Dawson et al., 1986; Bays et al., 2001). None of the chelators diminished the activity of Cod1p; rather, the chelating agents all stimulated the activity of the enzyme, suggesting that heavy metals may inhibit the activity of the enzyme. The only ion for which Cod1p showed any dependence was Mg²⁺.

Mg²⁺ in the form of an MgATP complex is required for all characterized P-type ATPases. Studies of other P-type ATPases have shown a requirement for additional free Mg²⁺ in order to obtain maximal activity (Skou, 1974; Ahlers, 1981; Plesner and Plesner, 1981; Brooker and Slayman, 1983). In these studies, maximal stimulation of the ATPase was usually observed when the total concentration of magnesium was one to three times the total concentration of ATP. We investigated the dependence of Cod1p on Mg²⁺ by varying the concentration of added MgCl₂ while keeping a total ATP concentration of 50 μM (Fig. 9, D and E). Because of the large excess of Mg²⁺ relative to the concentration of ATP, the MgATP concentration was nearly constant above 1.25 mM. Cod1p activity increased up to 10 mM Mg²⁺, after which it began to decline slowly (Fig. 9, D and E; unpublished data). The increase in Cod1p activity between 1.25 and 10 mM indicated a dependence on free Mg²⁺, because in this range, the concentration of MgATP remained constant. However, the increase in total activity due to free magnesium was on the order of twofold at maximum.

The regulation of ions in the ER is of central importance to the cell, given the numerous functions of this organelle. However, the only known supplier of ions to the ER was the Golgi apparatus–localized Pmr1p, which has been shown to influence the concentration of calcium in the ER (Strayle et al., 1999). Our previous work and that of others raised the possibility that Cod1p might play an important role with regard to ER ion regulation (Suzuki and Shimma, 1999; Cronin et al., 2000). In this work, we show that Cod1p is localized to, and centrally involved in, the function of the ER and in calcium regulation, and demonstrate the biochemical activity of Cod1p. Our cross-complementation studies, along with the broad conservation of Cod1p among metazoans, indicate that Cod1p shares similar functions in a broad range of organisms. Thus, it appears that Cod1p and its orthologues play a unique role in the functioning of the ER, quite distinct from other known P-type ATPases.

Our studies clearly show that Cod1p is a functionally important ER protein. Unique among yeast P-type ATPases, Cod1p localized to the ER membrane, as measured by immunofluorescence and direct biochemical analysis. The Cod1p subcellular localization is consistent with the functions and phenotypes of this ion pump. Our study and other studies had previously demonstrated phenotypes in ER-associated processes such as N-linked glycosylation and regulation of HMGR stability (Suzuki and Shimma, 1999; Cronin et al., 2000). The experiments above showed that Cod1p was critical in maintaining the proper ER folding environment and was required for viability of mutants deficient in the UPR, presumably due to its role in maintaining ER ion balance, as discussed below. Furthermore, at least in our strains, Cod1p had a more important role in this regard than Pmr1p, because the same synthetic lethality with hac1Δ was not observed with a pmr1Δ null. Taken together, our localization and phenotypic studies showed Cod1p to be a critical and conserved participant in ER function, with numerous phenotypes that arise from its action on ER physiology.

The identity of Cod1p as a P-type ATPase and its importance in the normal functioning of the ER suggests that it functions to control the balance of one or more ions in the lumen of the ER. The idea that this ion might be calcium first arose from our previous observations of cod1 suppression by manipulating exogenous calcium (Cronin et al., 2000). In this work, we further examined the role of Cod1p in calcium maintenance in a variety of ways. Increased expression of calcineurin-related genes can indicate depletion of secretory pathway calcium store, and a cod1Δ null showed the expected increase in expression of the calcineurin-activated genes PMC1, ENA1, and FKS2 (Cunningham and Fink, 1996; Matheos et al., 1997). In this assay, it is interesting to note that the magnitude of the effect of COD1 inactivation was similar to that reported for pmr1Δ nulls (Locke et al., 2000). Similarly, the specific growth hypersensitivity of cod1Δ strains to calcium chloride is reminiscent of the growth sensitivities of mutants lacking calcium-specific P-type ATPases. For example, deletion of the vacuolar calcium ATPase PMC1 results in calcium hypersensitivity, and deletion of the Golgi apparatus ATPase PMR1 leads to calcium and manganese hypersensitivity (Cunningham and Fink, 1994a; Lapinskas et al., 1996; Wei et al., 2000). Finally, the synergistic effect of cod1Δ and pmr1Δ on total cellular calcium, but no other ions, also pointed to a role for COD1 in maintaining cellular calcium homeostasis.

Although the calcium-related phenotypes were suggestive of direct involvement in calcium transport, our cross-complementation experiments and biochemical data suggested that Cod1p may influence the ER in a different manner. The known Ca^2+_transporting ATPases group together by sequence homology (type II) and are generally capable of substituting for each other's function in vivo. This property has been used extensively to clone new Ca²⁺ pumps and aid in the characterization of Ca²⁺ specificity of pumps cloned by other means (Liang et al., 1997; Harper et al., 1998; Talla et al., 1998; Degand et al., 1999). For example, SERCA1a from rabbit, ECA1 from A. thaliana, and SMA1 from Schistosoma mansoni all restore some functions to yeast mutants lacking the genes encoding the endogenous Ca²⁺ pumps Pmr1p and Pmc1p (Liang et al., 1997; Talla et al., 1998; Degand et al., 1999). It should be noted that the ability of exogenous pumps to complement yeast mutants does not strictly depend on the localization of the exogenous pump. Thus, functions of the Golgi apparatus–localized Pmr1p and the vacuolar Pmc1p can be substituted by the exogenous SERCA1 pump localized to the ER in yeast (Degand et al., 1999). In contrast to the apparent facility with which these pumps replace each other, neither SERCA nor ACA2, nor overexpressed PMR1, could suppress phenotypic defects of the cod1Δ null mutant. Similarly, overexpression of Cod1p did not suppress the EGTA sensitivity of pmr1Δ. Thus, if Cod1p contributes to the ER by directly pumping calcium, it must be doing so in a manner distinct from the previously known type II calcium pumps.

To further examine whether or not Cod1p was directly involved in calcium transport, we purified the Cod1 protein to identify the biochemical requirements for Cod1p ATPase activity in vitro. P-type ATPases exhibit enhanced activity in the presence of their substrates, such that the ATPase activity of SERCA1 is stimulated by calcium, the sodium/potassium ATPase by sodium and potassium (Koenderink et al., 2000), Pmr1p by manganese and calcium (Mandal et al., 2000), and the Menkes protein by copper (Voskoboinik et al., 2001). Surprisingly, calcium was not required for the biochemical activity of Cod1p, and at higher concentrations (>100 μM), Ca²⁺ inhibited activity. In fact, in the apparent absence of any substrate, Cod1p hydrolyzed ATP at a rate similar to that of other P-type ATPases. For example, Pmr1 has a Vmax of near 200 nmol Pi/min/mg in the presence of its preferred substrates Ca²⁺ and Mn²⁺ (Mandal et al., 2000), whereas Cod1p exhibits a Vmax for ATP hydrolysis of ∼150 nmol Pi/min/mg in reactions containing solely the eluted product, magnesium, potassium, and micromolar quantities of sodium. The metal chelators EDTA, TPEN, and 1,10-phenanthroline each activated (or de-inhibited?) Cod1p activity when used at concentrations that should have greatly reduced the free concentration of any divalent ions in the assay with the exception of magnesium. Cod1p appears to be acting quite differently from known P-type pumps.

A possible explanation is that the in vitro ATPase activity demonstrated by Cod1p is simply uncoupled from pumping activity. This might occur if the factors coupling Cod1p to a specific ion or ions were lost during the purification of the enzyme. This model suggests that the inability of the Drosophila homologue of Cod1p to complement the codΔ mutant at low copy was due to poor interaction of dCod1p with the required factors in yeast. A requirement for specificity-determining factors also leaves open the possibility for a direct role for Cod1p in calcium transport. Such coupling factors would be quite unexpected. The ATPase activity of some pumps bas been uncoupled from transport by cis mutations, and several mammalian P-type ATPases have β subunits. However none of the β subunits appear to modify the substrate specificity of the enzyme. Additionally, although we have been able to measure calcium transport by Pmr1p, Pmc1p, rabbit SERCA, and other Ca²⁺-specific ATPases in yeast microsomes (Sorin et al., 1997; Marchi et al., 1999; Ton et al., 2002), we have been unable to detect calcium transport by Cod1p-containing microsomes (unpublished data).

Another possibility is that magnesium is a substrate for Cod1p. P-type ATPases in general need free magnesium in order to hydrolyze ATP. The concentration of free magnesium needed for activation depends on the concentration of ATP present, such that generally the ratio of magnesium to ATP needed falls within the range of 1–3 mol magnesium per mol ATP (Ahlers, 1981; Plesner and Plesner, 1981;Brooker and Slayman, 1983; Skou, 1974). In the case of Cod1p, we obtained the greatest activity at a relatively high concentration of 10 mM magnesium, a concentration 200 times the concentration of ATP in the assay mix. A role for Cod1p in Mg²⁺ is attractive given the biological importance of magnesium. At present, magnesium supply and regulation in yeast are poorly understood and information about the intracellular distribution of magnesium is scarce (Beeler et al., 1997). ATPases in the lumen of the ER, such as Kar2, and other enzymes within the secretory pathway would presumably need magnesium in the lumen. It would seem fitting that yeast cells would have a way to transport magnesium into the lumen of the ER. Although the high level of magnesium needed for the activity of Cod1p does not demonstrate that Cod1p pumps magnesium, it suggests magnesium may be a good starting point for further biochemical study.

A third possibility is that Cod1p transports nonmetallic substrate. Recently, several groups have gathered evidence that the type IV group of P-type ATPases are involved in the translocation of aminophospholipids necessary for the maintenance of membrane lipid asymmetry (Tang et al., 1996; Chen et al., 1999; Ding et al., 2000; Gomes et al., 2000), though this idea has been challenged (Siegmund et al., 1998; Marx et al., 1999). The type IV ATPases that have been studied, including the Golgi apparatus–localized yeast protein Drs2p, are apparently specific for phosphatidylserine. Yeast lacking DRS2 are defective in translocating a fluorescently labeled phosphatidylserine, a defect that can be complemented by expression of the A. thaliana ALA1, and the ATPase activity of bovine ATPase II is stimulated by phosphatidylserine (Tang et al., 1996; Ding et al., 2000; Gomes et al., 2000). Although the type V pumps, such as Cod1p, are most homologous to Ca²⁺-transporting ATPases, when the sequence comparisons are restricted to certain core domains, the type V pumps are more similar to the type IV ATPases than to other members of the P-type ATPase family (Axelsen and Palmgren, 1998, 2001). The similar features of core domain sequences between the type IV and V groups suggest that an investigation of nonmetallic substrates for Cod1p may be fruitful.

Understanding the role of Cod1p has significance for the understanding of the possible functions of the other type V ATPases. Genes predicted to encode type V ATPases are present in all of the eukaryotic genomes that have been sequenced to date, yet none of the other members of the type V family are known other than by sequence homology. A number of genes have been identified, including one in humans, which are predicted to encode proteins with a high degree of homology to Cod1p (>35% identity and 55% similarity; Costanzo et al., 2000). Our demonstration that one of these, the D. melanogaster dCod1p, rescues the cod1Δ mutant phenotype indicated experimentally that the function of these proteins is conserved and these proteins are orthologous to Cod1p. More distant homologues of COD1 in the type V subgroup are even lesser known and likely have other functions. YOR291w, the second yeast type V ATPase has less homology to COD1 than dCOD1 (Costanzo et al., 2000). Deletion of YOR291w produces none of the phenotypes displayed by cod1Δ mutants (this paper; Cronin et al., 2000). In fact, we have been unable to detect any phenotypes associated with this gene (unpublished data). Thus, the degree of sequence similarity (or disparity) to Cod1p appears to be a useful predictor of function.

In conclusion, we have greatly extended the phenotypic characterization of Cod1p, defining its role in calcium regulation and ER function, and we have provided the first biochemical exploration of Cod1p. This study provides the foundation for further biochemical studies of Cod1p that will likely illuminate the obscurity surrounding ion regulation and enhance our understanding of the widely conserved, yet little known, type V group of P-type ATPases.

Restriction enzymes, Vent DNA polymerase, and T4 DNA ligase were obtained from New England Biolabs, Inc. The 9E10 cell culture supernatant was produced in our lab from cells (American Type Culture Collection accession no. CRL 1729) grown in RPMI1640 culture medium (Life Technologies) with 10% FCS and supplements. HA.11 anti-HA antibody was obtained from BAbCo. CY5-conjugated anti–rabbit antibody was a gift from Dr. Kit Pogliano (UCSD). Affinity-purified HRP-conjugated goat anti–mouse antibodies were purchased from Sigma-Aldrich. ECL chemiluminescence immunodetection reagents were from Amersham Pharmacia Biotech. Escherichia coli total lipid extract was obtained from Avanti Polar Lipids, Inc. All other chemical reagents were obtained from Sigma-Aldrich or Fisher Scientific.

pRH1112 (CEN, LEU2, HA::COD1) was constructed from pRH810 by using synthetic overlap extension to add sequence coding for an HA epitope tag to the 5′ end of COD1, just after the start codon. pRH1312 was constructed from pRH1112 by replacing the native promoter of COD1 between SacI and SphI sites with a PCR-amplified TDH3 (glyceraldehyde 3-phosphate dehydrogenase) promoter.

Plasmid pOT2-GH06032 containing the dCOD1 cDNA was purchased from Research Genetics. pRH1388 (CEN, LEU2, HA::dCOD1) expressing HA–dCod1p from the COD1 promoter was constructed as follows. Primers oRH1328 and oRH1329 were used to amplify the proximal part of the dCOD1 ORF before the AgeI site, adding sequence coding for an HA tag just after the start codon. This PCR product was cut with AatII and AgeI and ligated with the BglII–AatII fragment of pRH1112 and the AgeI–BglII fragment of pOT2-GH06032. The HA–dCOD1 ORF was subsequently transferred to a number of plasmids including pRH1389 (2μ LEU2 P_TDH3-HA::dCOD1) pRH1431 (2μ URA3 P_PMA1-HA::dCOd1) using AatII and BglII.

pRH1432 expressed 9HIS–HA–Cod1p from the TDH3 promoter on a high copy plasmid. Synthetic overlap extension was used to add sequence coding for nine histidine residues to the 5′ end of P_TDH3-HA::COD1, just after the start codon and before the sequence coding for the HA epitope tag. Oligonucleotide primers oRH1079, oRH1082, oRH1354, and oRH1355 were used to amplify from the template plasmid pRH1331 (2μ, LEU2, P_TDH3-HA::COD1). The resulting PCR product was subcloned into pRH1332 (2μ, URA3), using BsiWI and EagI to yield pRH1432. Plasmid pRH1478 expressed the D487N mutant of 9HIS–HA–Cod1p from the TDH3 promoter. It was constructed by removing the BstEII–BstEII fragment coding for the D487N mutation from pCS210 and placing it into the same sites of pRH1432.

Yeast strains were grown in minimal media (yeast nitrogen base without amino acids; Difco) with glucose and the appropriate supplements as described previously, except that leucine supplementation was increased to 60 mg/liter. Experiments were performed in minimal media at 30°C unless otherwise noted. For protein purification, yeast strains were grown in minimal media with glucose supplemented with 1 g/liter adenine, 1 g/liter methionine, 1.5 g/liter leucine, 1.5 g/liter lysine, 1 g/liter histidine, and 20 g/liter casamino acids (Difco). Crosses and transformations were done using standard techniques.

The 3myc–Cod1p allele expressed in RHY1870 (Table I) was constructed using a PCR epitope-tagging method (Schneider et al., 1995). URA3 and the flanking myc epitope tags from pMPY–3xMYC were amplified with primers oRH1190 and oRH1191. The primers added sequences homologous to the 5′ end of COD1. Yeast were transformed with the PCR product and selected for acquisition of uracil prototrophy. Transformants were screened by the fluorescence plate assay for the Cod⁻ phenotype. Cod⁻, Ura⁺ candidates were grown in YPD to allow “pop-out” of URA3 mediated by the direct repeats of the myc epitope tag followed by counterselection for loss of URA3 on media containing 5-fluoroorotic acid. The resulting Ura⁻, Cod⁺ candidates were then tested for expression of the 3myc–Cod1p by western analysis.

The cod1Δ::KanMX allele in RHY1722 (HMG2, cod1Δ::KanMX) was constructed by PCR-mediated disruption of the cod1Δ::LEU2 allele from RHY1402, which was derived from a cross of RHY792 and RHY1202 (Cronin et al., 2000). The resulting strain (RHY1401) was crossed to RHY795 (HMG2, ura3-52). Sporulation and dissection yielded RHY1722.

RHY1708 (1MYC::HMG2) and RHY1709 (1MYC::HMG2, cod1Δ::LEU2) were used for β-galactosidase assays. They were constructed by treatment of RHY791 and RHY1202 (Cronin et al., 2000) with 5-fluoroorotic acid. The β-galactosidase reporter plasmids were transformed into the strains using standard techniques.

hac1Δ strains were constructed as follows. RHY1647 (hac1Δ::URA3, COD1) and RHY1664 (hac1Δ::URA3, COD1) were constructed by transforming RHY534 and RHY872, respectively, with the BamHI fragment of pHAKO (Cox and Walter, 1996). RHY534 is a diploid version of JRY1159 (Hampton and Rine, 1994). RHY1884 was constructed by transforming MYY290 (Smith and Yaffe, 1991) with the BamHI fragment of pHAKO (Cox and Walter, 1996). RHY2179 (pmr1Δ::HIS3) was constructed by transforming MYY291 (Smith and Yaffe, 1991) with the AatII–AatII fragment from pL127-6 (Cunningham and Fink, 1994b). RHY2277 was isolated from a cross of RHY2179 and RHY1884.

Susceptibility of mutants to DTT, tunicamycin, MnCl₂, and other divalent cations was tested in minimal media. Susceptibility to MnCl₂, CaCl₂, and EGTA was tested in liquid YPD buffered with succinate to pH 5.5. Low-density liquid cultures of wild-type or mutant cells were used to serially dilute the tested agent. The resulting cultures were incubated and measured for optical density at 600 nm.

Immunofluorescence was performed essentially as described in detail previously (Rossanese et al., 1999), except that the cells were adhered to a multiwell slide instead of a coverslip. Images were captured using a DeltaVision confocal microscope and processed with the accompanying software.

β-Galactosidase assays were performed as previously described (Guarente, 1983), except that freeze–thawing in liquid nitrogen was used to lyse cells.

Fractionation was performed as described by Roberg et al., 1997. Protein from each fraction was separated by SDS-PAGE, transferred to nitrocellulose, probed with the appropriate antibodies, and detected by chemiluminescence using the ECL system. NIH Image 1.6 was used to determine signal intensity.

Analysis of GFP fluorescence in living cells by flow microfluorimetry was performed on a Becton Dickinson FACScalibur^® flow microfluorimeter using settings appropriate for GFP. Strains were typically grown into early log phase in minimal media. After addition of drugs, cultures were allowed to incubate 4 h before analysis.

The total amount of calcium and other ions in whole cells and subcellular fractions was determined by ICP-OES. To determine the total ion content of whole cells, cells were grown in YPD, pH 5.5, to OD₆₀₀ of 1.5 ml of the culture was collected by vacuum filtration onto 0.4-μm filters. The cells were then washed with 5 ml deionized water. The filters were transferred to microcentrifuge tubes and digested in 500 μl of 33% nitric acid overnight at 65°C. Samples were diluted with 3 ml of deionized water before analysis using a PerkinElmer Optima 3000XL spectrometer and software.

9HIS–HA–Cod1p was prepared by detergent extraction of the protein from ER membranes followed by nickel affinity chromatography. ER membranes were prepared by sucrose gradient centrifugation. Before harvesting, cells were treated with 10 mM sodium azide. The cells were washed in 10 mM sodium azide, 5 mM Tris, pH 7.5, and then resuspended in lysis buffer (4 M sorbitol, 1 mM EDTA, 0.1 M K_xH_xPO₄, pH 7.5) plus protease inhibitors (10 mM PMSF, 2 μg/ml chymostatin, 1 μg/ml leupeptin, 1 μg/ml pepstatin) so that the packed yeast cells made up one third of the volume. The cells lysed with four 30-s pulses with 1-min intervals in a Bead Beater^® (BioSpec Products, Inc.). The crude lysates was clarified by centrifugation at 350 g in a Sorvall SS-34 rotor. The clarified lysates were loaded onto a sucrose step gradient and centrifuged for 2 h at 174,000 g in a Beckman Coulter SW28 rotor. ER membranes containing 9HIS–HA–Cod1p were collected from the interface between 30% and 54% sucrose. The collected membranes were diluted with an equal volume of P&R buffer (0.5 M sucrose, 10 mM MES/KOH, pH 6, 150 mM potassium chloride) with protease inhibitors (1 mM PMSF, 2 μg/ml chymostatin, 1 μg/ml leupeptin, 1 μg/ml pepstatin). The membranes were pelleted by centrifugation for 1 h at 174,000 g. The pelleted membranes were resuspended in P&R buffer with protease inhibitors and stored at −90°C in 110-μl aliquots. Protein concentration in the membranes was measured using the BCA assay (Pierce Chemical Co.).

Detergent extracts of the membranes were prepared as follows. 10 mg of membrane was suspended in 10 ml of S-buffer (20 mM Hepes/Tris, pH 7.0, 20% glycerol, 0.5% E. coli lipid extract, 6 mM β-mercaptoethanol, 10 mM imidazole, protease inhibitors, and 1.5% n-octyl-β-d-glucopyranoside) and allowed to shake gently at 4°C for 2.5 h. The resulting suspension was centrifuged at 100,000 g for 1 h to pellet insoluble material.

The detergent-solubilized 9HIS–HA–Cod1p was then purified by nickel affinity chromatography. 250 μl of NTA-Ni agarose beads (QIAGEN) per milliliter of solubilized membranes was washed twice with deionized water and then twice with S-buffer. The equilibrated beads were added to the detergent extract and gently shaken at 4°C for 30 min. The extract with beads was then transferred to 2 ml Micro Bio-Spin columns (Bio-Rad Laboratories). Unbound material was allowed to flow through by gravity elution. Each column was washed twice with 1 ml of S-buffer plus 200 mM NaCl, and the column was washed a third time with S-buffer. 9HIS–HA–Cod1p was eluted after a 5-min incubation with 0.2 ml per column of S-buffer plus 300 mM imidazole. The eluted material was immediately frozen in a dry ice–ethanol bath and stored at −90°C in 60-μl aliquots.

ATPase activity was measured using the radiometric assay described previously (Bais, 1975; Mandal et al., 2000) with some modifications. Assays were done in a 50-μl volume and generally contained 50 mM Hepes/Tris, pH 7.0, 100 mM KCl, 5 mM MgCl₂, 1 mM EDTA. [γ³²P]ATP (30 Ci/mmol; Amersham Pharmacia Biotech) was diluted 10-fold with 1 mM NaATP to the final concentration desired. Variations in the assay conditions are noted in the figures. The concentrations of free ions and MgATP complexes were calculated using WINMAXC v2.05 (Bers et al., 1994) with supplementary constants from Data for Biochemical Research (Dawson et al., 1986). EDTA was excluded from experiments determining the effect of transition metals on Cod1p, because the high affinity of EDTA (and other chelators) for transition metals prevented accurate determination of the free ion within the desired range. Solutions were checked for contaminating ions by ICP-OES. Vanadate was prepared by the method of Gordon (1991).

The authors thank members of the Hampton and Rao labs for assistance with various parts of the project. The authors thank Dr. Chise Suzuki (National Food Research Institute, Tsukuba, Japan) for providing material prior to publication, Dr. Kyle Cunningham (Johns Hopkins University), Dr. Hans Rudolph (Universitat Stuttgart, Stuttgart, Germany), Dr. Peter Walter (UCSF), Dr. Jeff Harper (The Scripps Research Institute, La Jolla, CA), and Dr. Patrice Catty (Commissariat à l'Energie Atomique, Grenoble, France) for providing plasmids and strains, Dr. Jeff Harper, Karina Quon, and Kevin Walda (Scripps Institute for Oceanography, La Jolla, CA) for assistance with ICP-OES, Kit Pogliano and Marc Sharp for use of the Deltavision microscope, and Dr. Robert Rickert (UCSD) for use of the FACScalibur^® flow microfluorimeter and software.