Ppn2, a novel Zn²⁺-dependent polyphosphatase in the acidocalcisome-like yeast vacuole

Phosphorus is an essential macronutrient and a building block of nucleic acids, lipids and high-energy molecules. It is taken-up by cells in the form of inorganic phosphate (P_i), the scarcity of which is often is a limiting factor for growth of microorganisms and plants (Elser, 2012). Therefore, all organisms have developed elaborate systems of acquiring phosphate from the environment and complex regulatory mechanisms for adapting their metabolism in response to phosphate availability.

In all kingdoms of life, an integral part of phosphate metabolism is the synthesis of inorganic polyphosphate (polyP), a linear polymer ranging from several P_i residues to thousands of P_i residues in length (Kornberg et al., 1999). It is usually accumulated in acidic, Ca²⁺- and phosphate-rich membrane-enclosed compartments, called acidocalcisomes or acidocalcisome-related organelles (Docampo et al., 2005; Docampo and Huang, 2016). Substantial stores of polyP exist in prokaryotic and eukaryotic microorganisms. PolyP metabolism is tightly integrated into the overall energy metabolism of the cell and polyP is used by the cells as a P_i reserve and in phosphorylation and polyphosphorylation reactions (Azevedo et al., 2015; Livermore et al., 2016; Nocek et al., 2008; Rao et al., 2009). PolyP is important in detoxification of heavy metals and in counteracting alkaline and osmotic stress (Keasling, 1997; Pick and Weiss, 1991; Rohloff and Docampo, 2008). In addition, infectivity and persistence of prokaryotic and eukaryotic parasites depend on their polyP stores (Galizzi et al., 2013; Kim et al., 2002; Lander et al., 2013). The ability of polyP to stabilize proteins and act as a chaperone increases stress tolerance of cells (Azevedo et al., 2015; Dahl et al., 2015; Gray et al., 2014).

Mammalian cells contain much lower polyP levels compared to microorganisms (Kumble and Kornberg, 1995), but several important functions of mammalian polyP have been discovered. Upon activation, polyP is released from platelet-dense granules and modulates blood clotting and inflammation (Morrissey and Smith, 2015; Müller et al., 2009). PolyP can activate extracellular receptors (Dinarvand et al., 2014), modulate the excitability of neurons (Stotz et al., 2014) and protect cells from amyloid toxicity (Cremers et al., 2016). Although at times controversial (Faxalv et al., 2013), these findings have boosted interest in the polyP metabolism of eukaryotes. Further exploration of the role of acidocalcisomes and polyphosphates is hampered by our incomplete knowledge of the enzyme systems involved in polyP metabolism, even in the simple model organism Saccharomyces cerevisiae.

When phosphate is abundant in the medium, yeast accumulate large amounts of polyP, mainly inside the vacuole (Indge, 1968), an acidic, acidocalcisome-like organelle that degrades and recycles cellular compounds, accumulates and stores nutrients, and is tightly integrated into the signaling networks of the cell (Li and Kane, 2009). PolyP is used as a phosphate reserve that can sustain growth under phosphate scarcity or under acute high metabolic demand, as for example for dNTP synthesis during the S phase of the cell cycle (Bru et al., 2016). Synthesis and degradation of polyP are required to buffer cytosolic P_i concentration, and this is essential for cells in order to adapt to changing phosphate availability in the environment. That cells actively regulate cytosolic P_i is also underscored by the existence of a dedicated transcriptional program, the phosphate signal transduction (PHO) pathway (Thomas and O'Shea, 2005; Vardi et al., 2013), which responds to cytosolic P_i levels. Since polyP metabolism does not only respond to cytosolic P_i concentration but can even set it (Desfougères et al., 2016a), the exploration of polyP synthesis and turnover is crucial for understanding P_i metabolism and nutrient signaling in general. This can only be achieved once the major proteins involved in polyP metabolism have been identified.

In yeast, polyP is synthesized from nucleoside triphosphates by the vacuolar transporter chaperone (VTC) complex, which localizes to the vacuole membrane and the cell periphery, likely the endoplasmic reticulum (Hothorn et al., 2009; Uttenweiler et al., 2007). Concomitantly with synthesis, VTC transfers a growing polyP chain into the vacuole lumen (Gerasimaite et al., 2014). This process depends on the proton gradient across the vacuole membrane. Cells lacking VTC lose virtually all their polyP, suggesting that VTC is the major polyP polymerase in yeast (Desfougères et al., 2016a; Hothorn et al., 2009; Ogawa et al., 2000).

The situation with polyP degradation is more complex. Three proteins display polyphosphatase activities. Ppx1 is a highly processive exopolyphosphatase, degrading polyP by sequentially cleaving phosphate units from the end of a polyP chain (Wurst and Kornberg, 1994; Wurst et al., 1995). It is localized in the cytosol and therefore separated from the major polyP stores in the vacuole (Huh et al., 2003). Its deletion does not prevent degradation of polyP in response to low phosphate availability. It might, however, counteract the toxicity of cytosolic polyP (Gerasimaite et al., 2014). Ppn1 preferentially cleaves long polyP chains in the middle, but the end products of its reaction are triphosphate as well as monomeric P_i (Shi and Kornberg, 2005), suggesting that Ppn1 might also display exophosphatase activity. Ppn1 is localized inside the vacuole (Sethuraman et al., 2001). A third enzyme, Ddp1, has initially been described as diadenosine hexaphosphate and diphosphoinositol hydrolase. It has robust endopolyphosphatase activity in vitro (Cartwright and McLennan, 1999; Lonetti et al., 2011; Safrany et al., 1999). Similar to Ppx1, Ddp1 is localized in the cytosol and nucleus (Huh et al., 2003). Whereas the role of Ddp1 and its mammalian homologs in the metabolism of inositol pyrophosphates is well established, it is unclear whether its endopolyphosphatase activity is utilized in vivo.

A series of studies have detected polyphosphatase activities in different cellular compartments, such as the cell envelope, mitochondria, nucleus, cytosol and vacuoles (Lichko et al., 2006, 2003; Wurst et al., 1995). The enzymes responsible for these activities have been characterized by their cellular localization, molecular mass of the partially purified enzyme, the preferred metal ions and their differential sensitivity to inhibitors. However, the genes encoding these activities have never been identified, precluding studies of their in vivo roles. To our best knowledge, the only attempt to identify a gene by sequencing the N-terminus of a partially purified polyphosphatase led to the identification of the flavoprotein Yhb1, for which a role in polyP degradation has not been further explored (Lichko et al., 2000).

The difficulties in biochemically identifying missing polyphosphatase activities might be related to the fact that many phosphatases act on a wide spectrum of molecules in vitro – RNA, DNA, proteins and small molecules. This increases the chance that a promiscuous activity of unrelated enzymes towards polyP will be detected in extracts and mistaken for a genuine polyphosphatase activity (Mohamed and Hollfelder, 2013). In our efforts to identify missing S. cerevisiae polyphosphatases we chose a targeted approach based on data mining.

As the vacuole stores most of the polyP in yeast, we scanned databases for promising candidates – vacuolar proteins of unknown function that contain motifs suggesting that they might hydrolyze phosphodiester or phosphoric anhydride bonds. We turned our attention to YNL217w, an uncharacterized open reading frame (ORF) coding for a 327-amino-acid (aa) protein with a weak similarity to metallophosphatases. High-throughput screens had suggested that the product of this gene might localize to the vacuole and interact with the vacuolar polyphosphatase Ppn1 (Breitkreutz et al., 2010; Huh et al., 2003). We term this protein Ppn2.

Ppn2 is a member of the phosphoprotein phosphatase (PPP) family, which is defined by three highly conserved sequence motifs (-GDXHG-, -GDXVDRG- and -GNHE-) that contain the catalytic residues. BLASTP searches in the non-redundant protein database did not reveal similarities of Ppn2 outside the ∼100-amino-acid stretch containing these conserved motifs. The closest homolog of Ppn2 in the Protein Data Bank (PDB) is a symmetrically cleaving diadenosine tetraphosphate (Ap4A) hydrolase from Shigella flexneri 2a (Wang et al., 2006), which could be used to generate a homology model (Fig. 1A,B).

Ppn2 has a predicted transmembrane domain between residues 8 and 29, placing its putative catalytic domain at its C-terminus (Reynolds et al., 2008) (Fig. 1C). We investigated the localization of Ppn2 in the vacuole by genomically tagging it with yeGFP at the N- or C-terminus in the protease-deficient yeast strains BJ3505 and BY4742 pep4Δ. In order to test the topology of the protein, we subjected isolated vacuoles containing yeGFP–Ppn2 and Ppn2–yeGFP to limited proteolysis with proteinase K. yeGFP was readily cleaved from yeGFP–Ppn2, both in intact and detergent-solubilized vacuoles, indicating that the N-terminus faces the cytosol. In contrast, Ppn2–yeGFP could only be cleaved after solubilization of the vacuole membrane, indicating that the C-terminus of Ppn2 is localized in the lumen (Fig. 1D). Both constructs appear functional in vivo because they restore the typical polyP length distribution, which is lost in ppn2Δ cells (see below and Fig. 1E).

The fluorescence signals of the N- and C-terminally tagged proteins showed different patterns. While Ppn2–yeGFP stained exclusively the vacuole lumen, yeGFP–Ppn2 appeared on the vacuole membrane (Fig. 1F, left panel). Cleavage of yeGFP exposed to proteases in the vacuole lumen did not provide a satisfactory explanation for the lumenal localization, because the constructs had been expressed in protease-deficient strains. Furthermore, the majority of Ppn2–yeGFP appeared as an intact fusion protein in western blots of vacuole extracts prepared from these cells (Fig. 1D, see lane ‘-’ for the Ppn2–yeGFP sample). Alternatively, one of the tags might interfere with sorting of Ppn2. Numerous hydrolases reach the vacuole via the multivesicular body (MVB) pathway, where they are first sorted into luminal vesicles of late endosomes, which then fuse with vacuoles and deliver these vesicles for degradation in the vacuole lumen (Hurley and Emr, 2006). Similar to the process for other hydrolases that are sorted via the MVB pathway, such as Ppn1 (Reggiori and Pelham, 2001) and Cps1 (Odorizzi et al., 1998), Ppn2 exposes only an 8-amino-acid N-terminal peptide to the cytosol. We propose that the attachment of a bulky GFP tag might render this peptide inaccessible for recognition by the ESCRT machinery and disturb sorting. We tested whether Ppn2 uses the MVB pathway by expressing Ppn2–yeGFP in a vps4Δ strain, in which sorting into luminal MVB vesicles is blocked (Babst et al., 1997). In this mutant, proteins using the MVB pathway remain in an enlarged MVB and on the vacuole membrane. We found Ppn2–yeGFP on the vacuole membrane of vps4Δ cells and in a large dot next to the vacuole, which is typical for the enlarged MVBs of a vps4Δ mutant (Fig. 1F, right panel). This suggests that Ppn2 reaches the vacuole lumen via the MVB pathway.

Our attempts to purify the hydrophilic domain of recombinant Ppn2 failed, as the protein remained insoluble. Therefore, we decided to investigate its activity using protein immunoprecipitated from yeast vacuoles. We overexpressed Ppn2 with an N-terminal 3HA tag in yeast, using the strong GPD1 promoter (P_GPD1-3HA-Ppn2). The protein was immunoprecipitated from vacuole detergent lysates with anti-HA antibody (Fig. 2A). When washed beads were incubated with synthetic polyP-60 (i.e. a 60-residue polyP chain) in the presence of ZnCl₂, a robust degradation activity was detected (Fig. 2B). No degradation was observed when vacuoles with non-tagged Ppn2 were used for immunoprecipitation.

High-throughput screens had suggested an interaction between Ppn2 and another vacuolar polyphosphatase, Ppn1. During our initial experiments, we found that the alkaline phosphatase Pho8 also co-immunoprecipitated with Ppn2 with high yield. Almost equal proportions of the total amount were co-precipitated for both proteins, suggesting that they form a stable complex. This is of interest because a recent report showed that Pho8 influences polyP levels (Kizawa et al., 2016) and the homologous vacuolar alkaline phosphatase from Candida utilis has polyphosphatase activity (Fernandez et al., 1981). Therefore, we deleted PPN1 and/or PHO8 in P_GPD1-3HA-PPN2 cells and measured the activity of the immunoprecipitated 3HA–Ppn2. Immunoprecipitation efficiency of 3HA–Ppn2 was the same in all strains (Fig. 2A). Polyphosphatase activity was neither affected by ppn1Δ nor by a pho8Δ mutation, suggesting that Ppn2 carries a polyphosphatase activity that is independent of Ppn1 and Pho8 (Fig. 2B).

We identified conserved motifs of the PPP-family in Ppn2 and mutated Asp68 and Asp98 to Asn (Fig. 1A). These residues chelate the metal ion that is required for catalysis by lambda protein phosphatase (Zhuo et al., 1994). We expressed wild-type (wt) and mutant Ppn2 with an N-terminal 3HA tag from a centromeric plasmid. All proteins were expressed at similar levels, and could be immunoprecipitated with similar efficiency (Fig. 2C). As expected, 3HA–Ppn2^wt showed robust degradation activity, while 3HA–Ppn2^D68N and 3HA–Ppn2^D98N were nearly inactive. Therefore, the conserved residues in the putative active center of Ppn2 likely play a similar role to those in other members of the PPP-superfamily. This result further supports the conclusion that Ppn2 is directly responsible for the observed polyphosphatase activity.

In order to examine the metal requirement of 3HA–Ppn2, we immunoprecipitated it in the presence of EDTA and measured its activity after re-addition of different divalent ions at 1 mM concentration (Fig. 2D). 3HA–Ppn2^D68N remained inactive under all conditions. In contrast to other known yeast polyphosphatases, which can use Mg²⁺ for catalysis, 3HA–Ppn2^wt remained inactive in the presence of Mg²⁺. Only Zn²⁺ and Co²⁺ supported the catalytic activity of 3HA–Ppn2^wt. Zn²⁺ is more likely to be used in vivo because it is stored to high levels inside vacuoles (Simm et al., 2007). Co²⁺ has similar ionic radius and the same coordination geometry as Zn²⁺, therefore it can effectively substitute for Zn²⁺, and sometimes for other divalent metal ions, in the active sites of enzymes (Andreini et al., 2008). However, its concentration inside the cell is low and it is commonly accepted that only vitamin B₁₂-dependent enzymes obligatorily require cobalt ions, where Co³⁺ serves as a redox center. More abundant divalent metal ions are more likely to be employed as Lewis acids in other enzyme classes.

When we followed degradation of polyP-60 by 3HA–Ppn2^wt by separating the reaction products on a gel, we noticed that shorter polyP chains accumulated before polyP disappeared entirely (Fig. 3A). This pattern is consistent with Ppn2 acting as a non-processive exopolyphosphatase or an endopolyphosphatase that cleaves inside the polyP chain. In order to test this more clearly, we measured the activity of 3HA–Ppn2^wt by a Malachite Green assay that detects free P_i ions. Very low activity was detected, even with 8-fold higher protein concentration, compared to the preceding experiment where polyP-60 degradation had been followed by DAPI fluorescence (Fig. 3B). This residual activity disappeared completely when PHO8 and PPN1 were deleted, suggesting that Ppn2 by itself does not release P_i.

Next, we blocked the end groups of polyP-60 and polyP-300 by phosphoamidate linkage to spermidine (Choi et al., 2010) in order to render it resistant to exopolyphosphatase activity. This modified polyP could not be digested by purified recombinant S. cerevisiae yeast exopolyphosphatase (sc)Ppx1, serving as proof of successful modification (Fig. 3C). The modified polyP was still degraded by immunoprecipitated 3HA–Ppn2^wt, but not by the catalytically inactive 3HA–Ppn2^D68N. Thus, Ppn2 carries endopolyphosphatase activity.

In order to determine whether, in addition to Ppn1 and Ppn2, vacuoles contain other major polyphosphatases, we sought to measure polyP hydrolysis activity in vacuoles from ppn1Δ ppn2Δ cells. However, our intense attempts to isolate vacuoles from this strain have failed. We had noticed previously that our vacuole isolation procedure yields vacuoles with low levels of polyP, even when the cells express hyperactive VTC and contain higher polyP levels than in wild type (Desfougères et al., 2016a). This is consistent with our observation that polyP is degraded during vacuole isolation (A.M., unpublished observation). We hypothesized that the failure to obtain vacuoles from ppn1Δ ppn2Δ might be caused by their inability to degrade polyP. In line with this assumption, the problem was overcome by deleting the catalytic subunit of the polyP polymerase (VTC4) in addition to the polyphosphatases. vtc4Δ ppn1Δ ppn2Δ cells gave good vacuole yields and allowed us to proceed with the experiments.

We compared the degradation of polyP-300 by detergent lysates of vacuoles from vtc4Δ, vtc4Δ ppn1Δ, vtc4Δ ppn2Δ and vtc4Δ ppn1Δ ppn2Δ strains (Fig. 4). In the presence of Mg²⁺, the major polyphosphatase activity in vacuole lysates stemmed from Ppn1, as no activity was detected in vacuole extracts of ppn1Δ vtc4Δ and ppn1Δ ppn2Δ vtc4Δ strains. In the presence of Zn²⁺ ions, ppn1Δ vtc4Δ and ppn2Δ vtc4Δ both showed slower degradation than in wild type. This activity was completely absent in the ppn1Δ ppn2Δ vtc4Δ strain. Our assays yielded no evidence for vacuole-associated polyphosphatase activities other than Ppn1 and Ppn2, but they do not exclude that such activities exist and become significant under different experimental conditions. These results do, however, suggest that both Ppn1 and Ppn2 make major contributions to vacuolar polyP degradation.

In order to test for a role of Ppn2 in polyP metabolism in vivo, we quantified polyP levels in ppn1Δ, ppn2Δ and ppn1Δ ppn2Δ cells in the lag phase, logarithmic and stationary phase. Deletion of vacuolar polyphosphatases did not lead to increased polyP accumulation in vivo (Fig. 5A). In addition, vacuolar structure, which depends on the accumulation of polyP (Desfougères et al., 2016a), was not altered in the deletion mutants (Fig. 5C). By contrast, the average chain length showed strong changes (Fig. 5B). In logarithmically growing BY4742 wild-type yeast, polyP chain length ranges from 3 to >300 P_i residues. ppn1Δ ppn2Δ cells contained only chains longer than 300 units in all growth phases. As previously reported (Sethuraman et al., 2001), ppn1Δ cells contained longer polyP chains in the stationary phase. During the lag phase and logarithmic phase, however, polyP chain length distribution in ppn1Δ cells was similar to that in wild type. The effect of Ppn2 was the strongest during early logarithmic phase, when ppn2Δ cells showed a clear shift towards longer chains. Taken together with previous observations (Desfougères et al., 2016b), which suggested that the amount of polyP may be sensed inside the vacuole, our results suggest that polyP accumulation is mostly controlled by regulating the rate of the synthesis rather than degradation. In line with this, deletion of additional genes, such as PHO8 and PPX1, also did not increase cellular polyP levels in logarithmically growing cells (Fig. S1). Our results show that Ppn1 and Ppn2 make distinct contributions to vacuolar polyphosphate turnover in different growth phases. Furthermore, they suggest that the VTC complex is a highly processive enzyme that synthesizes very long chain polyP, and that the wide variation in chain lengths observed in wild-type cells results from polyP degradation by Ppn1 and Ppn2.

We next examined the effects of vacuole polyphosphatase knockouts on cell growth. No differences were detected in synthetic complete (SC) medium, when the cultures were started from logarithmically growing precultures (Fig. 6A). However, in SC medium with low levels of phosphate (SC-P_i), ppn1Δ ppn2Δ cells reduced their growth rate earlier and reached a lower final density than did wild-type cells. The same phenotype was observed with vtc4Δ cells, which do not produce polyP. Loss of only one of the two vacuolar polyphosphatases did not produce a growth phenotype, suggesting that under these conditions Ppn1 and Ppn2 function redundantly.

Absence of polyphosphatase activity in vacuole extracts from double-knockout cells had suggested that Ppn1 and Ppn2 might be the most important vacuole polyphosphatases. In order to study their contribution in vivo, we assayed polyP mobilization after transferring the cells from P_i-rich medium to SC-P_i. Transfer to low-P_i medium induces rapid polyP degradation (Harold, 1966). Wild-type cultures fully degraded polyP within 2 h in SC-P_i medium. ppn1Δ cultures showed a considerable delay and degraded their polyP only after 4–6 h. ppn2Δ cultures showed similar degradation kinetics to that of wild-type cells. ppn1Δ ppn2Δ cultures maintained their polyP stores throughout the incubation (Fig. 6B). We analyzed the chain length distribution during P_i limitation by gel electrophoresis (Fig. 6C). P_i limitation in wild-type cells was accompanied by a shift in polyP towards shorter chain lengths, followed by its complete disappearance after 2 h of incubation. Whereas ppn2Δ had little effect, polyP degradation was delayed in ppn1Δ cells, with shorter polyP chains prevailing at the later time-points. PolyP in ppn1Δ ppn2Δ cells was, however, not converted into short chains, even after 6 h in low-P_i medium. Taken together, these observations show that Ppn1 and Ppn2 have a major role in physiologically triggered polyP mobilization that is relevant for growth.

S. cerevisiae has a complex system of polyP-degrading enzymes. Two polyphosphatases – Ppx1 and Ppn1 – were identified two decades ago, but it has soon been discovered that the deletion of both abolishes neither the in vivo mobilization polyP in response to P_i scarcity nor polyP degradation by cell extracts in vitro (Lichko et al., 2006, 2008), but only slow these processes down. The contribution of the third yeast enzyme with polyphosphatase activity, Ddp1, was addressed by studying strains overexpressing Ddp1, which did not provide a clear answer concerning its in vivo relevance for polyP degradation (Trilisenko et al., 2015). The situation for Ddp1 is complicated by its role in metabolism of inositol pyrophosphates, which are crucial stimulators of polyP synthesis (Cartwright and McLennan, 1999; Safrany et al., 1999; Wild et al., 2016). Furthermore, its cytosolic and nuclear localization make it an unlikely candidate for degradation of the major vacuolar polyP stores. An earlier study had found two polyphosphatase activities associated with the vacuole (Wurst et al., 1995). While one of them was likely Ppn1 (Shi and Kornberg, 2005), the other one has remained unidentified. In this work, we show that a previously uncharacterized ORF, YNL217w, encodes the novel vacuolar polyphosphatase Ppn2, which, together with Ppn1, plays a major role in yeast polyP metabolism.

Why do the cells use two vacuolar enzymes for polyP degradation? Ppn1 and Ppn2 both have endopolyphosphatase activities and are sorted to the vacuole via the MVB pathway, but their sequences are unrelated. Having two unrelated enzymes with a similar function provides a means to differentially regulate polyP degradation in response to different stimuli or metabolic demands. The change of Ppn1 activity during culture growth has been noted previously (Shi and Kornberg, 2005). In line with this study, we observed that Ppn1 is important for defining polyP chain length in the stationary phase, whereas Ppn2 is most important at the beginning of logarithmic phase.

Ppn1 and Ppn2 seem to also have different biochemical properties. While Ppn1 was initially described as an endopolyphosphatase, subsequent studies have shown that it could release P_i (Andreeva et al., 2015; Sethuraman et al., 2001; Shi and Kornberg, 2005). Our data suggests that Ppn2 is exclusively an endopolyphosphatase and is unable to release P_i. Although this result could be a consequence of the particular in vitro reaction conditions and should be interpreted with some caution, it might point to an important difference between Ppn1 and Ppn2. Ppn1 might be responsible for mobilization of polyP stores in response to increased demand for P_i, while Ppn2 might control polyP chain length.

An earlier study found that polyP of 3 and 4 P_i residues is generated from longer chains (Lusby and McLaughlin, 1980). We extend this observation and demonstrate in vivo that the VTC complex produces polyP chains longer than 300 P_i residues. The typical broad distribution of polyP chain lengths, as seen in wild-type cells, should be ascribed to vacuolar polyphosphatases acting on these very long chains. It remains to be determined what advantage is provided to the cell by the ability to control the polyP chain length. It has been proposed that the accumulation of very long chains, or the lack of short chains of polyP, is toxic for yeast (Sethuraman et al., 2001). We were unable to reproduce this finding in our ppn1Δ and ppn1Δ ppn2Δ cells, perhaps because we used a different strain background. However, important physicochemical properties of polyP are defined by its length that might be of biological importance. First, the end and middle groups of polyP have different pK_a values. While the middle groups are strongly acidic, the pK_a1 of the end groups is between 6.5 and 7.2 – close enough to the luminal pH of vacuoles to influence the buffering capacity of the vacuole lumen (Lee and Whitesides, 2010). Shifting the average chain length towards shorter chains can increase the concentration of the end groups and might play a role in controlling vacuolar and cytosolic pH. In line with this, degradation of longer polyP chains to triphosphate has been implicated in counteracting alkaline stress in the alga Dunaliella salina, and polyP degradation is observed during alkaline stress in Trypanosoma cruzi (Pick and Weiss, 1991; Ruiz et al., 2001). Second, at acidic pH (pH 5–6), longer polyP chains are better chelators of divalent metal ions than very short chains (polyP-3 or polyP-4). This can influence the ability of vacuoles to retain metal ions. Third, depending on ionic strength, pH and the presence of divalent metal ions, polyP chains readily form gels that can induce phase separations and provide separate reaction spaces (Cini and Ball, 2014). Finally, the ability of polyP to interact with proteins can depend on its chain length. This is illustrated by the observations that long polyP chains are more effective chaperones in bacteria (Gray et al., 2014), and that polyP of different chain length activates blood clotting by different mechanisms and displays different activities in pro-inflammatory signaling (Dinarvand et al., 2014; Smith et al., 2010).

If the physiological function of Ppn1 were to release P_i, Ppn2 could accelerate this reaction by increasing the concentration of free polyP ends (substrates of exopolyphosphatase activity) and by decreasing the concentration of the competing long chains that are the preferred substrates for the endopolyphosphatase activity of Ppn1. Under the growth conditions that we have employed, we have not obtained compelling evidence for this mechanism, as ppn2Δ cells did not show a significant delay in polyP degradation in low-P_i medium. However, our in vitro observations still argue for such a role: polyphosphatase defects in the vacuole lysates of ppn1Δ and ppn2Δ strains (as shown in Fig. 4) could only be detected when polyP-300 was used as a substrate. With shorter polyP-60 the differences were less pronounced and poorly reproducible. It is possible that the polyP degradation assays that we performed in SC-P_i medium under standard laboratory conditions were not competent to reveal a similar effect in vivo.

The major polyP store of the yeast cell is inside the vacuole. Surprisingly, deletion of both the vacuole polyphosphatases Ppn1 and Ppn2 did not lead to increased polyP accumulation in steady state. This suggests that polyP levels of the cell are mainly controlled through synthesis and not through degradation. This interpretation is in line with our earlier findings showing that hyper-activation of the VTC complex by point mutations or overexpression of the regulatory subunit leads to higher polyP levels in vivo (Desfougères et al., 2016a) and that the polyP content of vacuoles may be sensed and communicated to the cytosol (Desfougères et al., 2016b). Studies on the mobilization of this store have so far been hampered by the fact that polyP turnover could be inhibited only partially because only a single vacuolar polyphosphatase was known. The identification of Ppn2 now allows experimenters to stabilize the vacuolar polyP pool, even under P_i starvation conditions, where new polyP cannot be made. This suggests that Ppn1 and Ppn2 are the major players in the polyP mobilization required for growth. The discovery of the second vacuolar polyphosphatase thus opens the way for addressing the regulation of the vacuolar phosphate pool and for us to explore its impact on phosphate and energy homeostasis of the cell.

Accumulation of large polyP amounts as a P_i reserve is a feature that is conserved in free-living microorganisms. In higher multicellular organisms, polyP assumes regulatory and signaling functions and may participate in bone mineralization (Müller et al., 2015). Discovering and studying functions of polyP that are unrelated to storage in yeast is complicated because of the extreme abundance of the vacuolar polyP pool (Gerasimaite et al., 2014; Indge, 1968; Urech et al., 1978), which can mask polyP dynamics in other cellular compartments. Selectively making this polyP pool inert by use of ppn1Δ ppn2Δ mutations could help experimenters to study polyP functions in other compartments, such as polyphosphorylation of nuclear proteins (Azevedo et al., 2015). Protein polyphosphorylation requires an active VTC complex. However, it is unknown whether the polyP that is used for polyphosphorylation of nuclear proteins originates from the vacuolar pool. VTC is also found in the cell periphery, likely the endoplasmic reticulum, which is continuous with the nuclear membrane. It is not inconceivable that VTC might hand over polyP chains directly to cytosolic or nuclear proteins that use them for protein polyphosphorylation. The possibility of immobilizing the vacuolar polyP pool by the elimination of Ppn1 and Ppn2 may allow us to address such questions. It will also be instrumental in searching for and identifying potential additional polyphosphatases (Lichko et al., 2003).

Creatine kinase, creatine phosphate and ATP were from Roche. PolyP-60 and polyP-300 were kind gifts from Toshikazu Shiba, Regenetiss Inc., Japan. Recombinant scPpx1 was purified as previously described (Gerasimaite et al., 2014). The concentration of polyP is reported as the concentration of P_i units. It was determined by degrading polyP with recombinant scPpx1 and measuring the released P_i with a Malachite Green assay (Hothorn et al., 2009).

Anti-HA monoclonal mouse antibody HA.11 clone 16B12 (mouse ascites) (Covance, # MMS-101R) was used for immunoprecipitations. Anti-HA polyclonal antibody from rabbit (Abcam ab137838) was used at 1:2000 for western blotting. Anti-GFP (Torrey Pines Biolabs TP401), affinity purified from rabbit antisera, was used at 1:5000 for western blotting. Polyclonal antibodies to Pho8 and Vam3 were raised in rabbits using proteins expressed in E. coli as described (Müller et al., 2002; Pieren et al., 2010). The unpurified sera was used for western blotting at dilutions of 1:10,000. The secondary antibodies conjugated to IRDye 800 were from LiCor. Specificity of all antibodies had been verified by blotting of vacuoles from knockout strains or from strains expressing non-tagged proteins, respectively.

Strains are listed in Table S1. Genes were deleted by replacing a complete ORF with a marker cassette (Janke et al., 2004; see Table S2 for PCR primers). Integration of the cassette into the correct region was confirmed by colony PCR, with one of the primers complementary to the promoter of the target gene and another primer complementary to the marker gene on the cassette. pRSx0x series plasmids were used as templates for amplifying cassettes with auxotrophic markers. PEP4 and PHO8 genes were deleted using plasmids pTS15 and pGP10, respectively (a kind gift from Tom Stevens, Eugene, OR). BY4742 ppx1Δ ppn1Δ was constructed from BY4742 ppn1Δ::kanMX (Euroscarf, acc.no. Y14286). BY4742 vtc4Δ ppn1Δ was constructed from BY4742 vtc4Δ (Euroscarf, Acc.no. Y16780).

To construct plasmids pRS415-P_GPD-PPN2 and pRS415-P_GPD-3HA-PPN2, yeast genomic DNA was isolated from strains BY4742 P_GPD-PPN2 and BY4742 P_GPD-3HA-Ppn2 as described previously (Breitkreutz et al., 2010) and the target genes were amplified using primer pairs RG077+RG074 and RG076+RG074, respectively. The resulting fragments were cloned into the pRS415GPD vector using BamHI and HindIII, or SpeI and HindIII restriction sites. The point mutations D68N and D98N were introduced by the QuikChange mutagenesis procedure using primers RG078+RG079 and RG080 and RG081, respectively.

Cell growth was assessed by measuring the optical density at 600 nm (OD_600nm). To investigate polyP accumulation, growth in low-P_i medium and degradation of polyP under phosphate starvation, precultures were grown in SC medium at 30°C for 24 h. From these, fresh cultures were inoculated and grown overnight at 30°C to reach an OD_600nm of 1–2. Then, the cells were sedimented (2 min, 3000 g), washed once with low phosphate medium (SC-P_i) and incubated at 30°C for 6 h. At the indicated time-points, 2.5 OD_600nm units of cells were collected for polyP isolation.

To investigate growth in SC medium and the length of polyP chains at different growth phases, precultures were grown in SC for 48 h at 30°C. Then, the cells were diluted to an OD_600nm of ∼0.1 with fresh medium and further incubated at 30°C. The following samples were collected and processed immediately: lag phase (at OD_600nm of ∼0.1, 3 h incubation), early exponential phase (at OD_600nm of ∼0.4–0.6, 6–7 h incubation), stationary phase (at OD_600nm of 3–4, 24 h incubation). PolyP was isolated from 2–2.5 OD_600nm units of cells by phenol/chloroform and glass bead extraction as described previously (Lonetti et al., 2011) and dissolved in 100 µl TE buffer (10 mM Tris-HCl pH 7.4, 0.5 mM EDTA). Isolated polyP was quantified by enzymatic hydrolysis with an excess of recombinant scPpx1: ∼50 nmol of isolated polyP was incubated in 100 µl of 5 mM Tris·HCl pH 7.0, 1.5 mM MgCl₂ and 0.4 µg/ml scPpx1 at 37°C for 1.5–2 h. An aliquot of 10 µl was mixed with 90 µl water and phosphate was quantified by adding 150 µl Malachite Green reagent (86 µl 28 mM ammonium heptamolybdate in 12.5% H₂SO₄ and 64 µl 0.76 mM malachite green in 0.35% polyvinyl alcohol) and measuring absorbance at 595 nm.

For vacuole isolation, the cells were grown in 1 l YPD medium at 30°C overnight to an OD_{600 nm} of 0.8–1.9 and vacuoles were isolated by floatation in a stepwise Ficoll 400 gradient as described previously (Gerasimaite et al., 2014). Unless indicated otherwise, the fractionation solutions contained 0.5× protease inhibitor cocktail (PIC): 50 µM pefabloc SC, 50 ng/ml leupeptin, 25 µM O-phenanthroline and 250 ng/ml pepstatin A, and 1 mM PMSF. Vacuole concentration is expressed as concentration of total vacuolar protein as determined by a Bradford assay, using bovine serum albumin (BSA) as a standard.

The cells were grown in SC medium overnight to an OD_{600 nm} of 1–2. For staining the vacuole membrane, 0.2 OD_{600 nm} units of cells were resuspended in 1 ml SC medium containing 10 µM FM4-64 and incubated at 30°C for 1 h. The cells were washed once with a fresh medium and chased for an additional hour. Cells were observed using a LEICA DMI6000 B inverted microscope with a 100×1.4 NA lens, connected to a Hamamatsu digital CCD camera (ORCA-R2 C10600-10B) and an X-Cite^® series 120Q UV lamp. Appropriate filters were used to visualize GFP and FM4–64 fluorescence.

The orientation of Ppn2 relative to the vacuole membrane was probed by limited proteolysis of Ppn2 N- and C-terminal fusions with yeGFP. Isolated vacuoles were resuspended in PS buffer (10 mM PIPES-KOH pH 6.8, 200 mM sorbitol) to 0.02 mg/ml and mixed with an equal volume of proteinase K (20 µg/ml in PS buffer). The samples were incubated for 5 min on ice and proteins were precipitated with 10% trichloroacetic acid (TCA). Untreated controls received either no proteinase K at 0 min or proteinase K was added after TCA. The samples were fractionated on SDS-PAGE gels and yeGFP was detected by western blotting.

All solutions contained 1×PIC and 1 mM PMSF. 200 µg vacuoles in 200–400 µl were mixed with 1 ml PS buffer, collected by centrifugation (7000 g, 10 min, 4°C) and gently resuspended in 360 µl solubilization buffer (20 mM PIPES-KOH pH 6.8, 150 mM KCl). 40 µl 5% Triton X-100 was added, the samples were shaken vigorously at 4°C for 20 min, centrifuged (15 min, 22,000 g at 4°C). 4 µl HA.11 clone 16B12 monoclonal antibody was added to the supernatant together with 30 µl protein-G–agarose beads (Roche) equilibrated in the solubilization buffer. The samples were incubated on a rotating wheel (8°C, 2 h), then the beads were washed once with solubilization buffer with Triton X-100 and three times with solubilization buffer, and were then resuspended in 180 µl of solubilization buffer. 1.5 µl of the resulting suspension was used for an 80 µl reaction (single time-point). This corresponds to 1.5 µg of starting vacuole material. The immunoprecipitation efficiency varied between 12 and 35% of the total input in different experiments.

Polyphosphatase activity of immunoprecipitated 3HA–Ppn2 was assayed in a buffer containing 20 mM PIPES-KOH pH 6.8, 150 mM KCl, 1 mM ZnCl₂ and 30 µM polyP-60. The reactions were started by adding 3HA–Ppn2 adsorbed onto beads (1.5 µl beads in 80 µl) and incubated at 27°C with vigorous shaking. At the indicated time points, 80 µl aliquots were withdrawn and the reaction was stopped by mixing with 160 µl of stop buffer (10 mM PIPES-KOH pH 6.8, 150 mM KCl, 15 µM DAPI, 12 mM EDTA). PolyP was detected by measuring the characteristic DAPI-polyP fluorescence (excitation at 415 nm, emission at 550 nm, 530 nm cut-off) in a black 96-well plate.

When P_i release was monitored, the reactions were performed in 150 µl volume of the same buffer in the presence of 15 µl beads. The reactions were incubated at 27°C for 30 min, then 100 µl aliquots were withdrawn and mixed with 10 µl 1 M H₂SO₄. In parallel, the same quantity of polyP-60 was hydrolyzed with 1 M HClO₄ (30 min, 99°C). Released P_i was detected by adding 150 µl Malachite Green reagent and measuring absorbance at 595 nm.

For measuring polyphosphatase activities of vacuole lysates, vacuoles were isolated without PIC. The reactions were started by adding vacuoles (final concentration 0.002 mg/ml) to a buffer containing 20 mM PIPES-KOH pH 6.8, 150 mM KCl, 1 mM ZnCl₂ or MgCl₂, 0.1% Triton X-100, 1×PIC and 1 mM PMSF. After different periods of incubation at 27°C, 80 µl aliquots were withdrawn, the reactions were stopped and the remaining polyP was detected as described above. Values for the 0 min time-point represent reactions that had been performed with heat-inactivated (10 min, 99°C) vacuoles.

Terminal phosphate groups of polyP chains were blocked by attaching spermidine via a phosphoamidate linkage as described (Choi et al., 2010). Briefly, 5 mM polyP-60 or polyP-300 were incubated in 100 µl of 100 mM MES-NaOH pH 6.5 in the presence of 300 mM N-(3-dimethylaminopropyl)-N′-ethylcarbodiimide hydrochloride (EDAC) and 70 mM spermidine-HCl at 37°C overnight. In the control reaction, EDAC was omitted. At the end of the reaction, LiCl was added to give a 1 M final concentration and unreacted spermidine-HCl and EDAC were removed by gel filtration on a sephadex G-25 column equilibrated with 1 M LiCl. PolyP was precipitated with 500 µl acetone, centrifuged (22,000 g, 30 min), washed once with acetone, air-dried and dissolved in water. The concentration was estimated by DAPI fluorescence using unmodified polyP-60 as a standard. 30 µM of chemically modified and control polyP was used in the reaction with immunoprecipitated 3HA–Ppn2.

We thank Toshikazu Shiba (Regenetiss Inc., Japan) for synthetic polyP-60 and polyP-300. We are grateful to Andrea Schmidt, Lydie Michaillat-Mayer and Yann Desfougères for yeast strains and discussions.