The gluconate shunt is an alternative route for directing glucose into the pentose phosphate pathway in fission yeast

Glycolysis and the pentose phosphate pathway both play a central role in the degradation of glucose in all domains of life. Another metabolic route that can facilitate glucose breakdown is the gluconate shunt. In this shunt glucose dehydrogenase and gluconate kinase catalyze the two-step conversion of glucose into the pentose phosphate pathway intermediate 6-phosphogluconate. Despite the presence of these enzymes in many organisms, their only established role is in the production of 6-phosphogluconate for the Entner-Doudoroff pathway. In this report we performed metabolic profiling on a strain of Schizosaccharomyces pombe lacking the zinc-responsive transcriptional repressor Loz1 with the goal of identifying metabolic pathways that were altered by cellular zinc status. This profiling revealed that loz1Δ cells accumulate higher levels of gluconate. We show that the altered gluconate levels in loz1Δ cells result from increased expression of gcd1. By analyzing the activity of recombinant Gcd1 in vitro and by measuring gluconate levels in strains lacking enzymes of the gluconate shunt we demonstrate that Gcd1 encodes a novel NADP+-dependent glucose dehydrogenase that acts in a pathway with the Idn1 gluconate kinase. We also find that cells lacking gcd1 and zwf1, which encode the first enzyme in the pentose phosphate pathway, have a more severe growth phenotype than cells lacking zwf1. We propose that in S. pombe Gcd1 and Idn1 act together to shunt glucose into the pentose phosphate pathway, creating an alternative route for directing glucose into the pentose phosphate pathway that bypasses hexokinase and the rate-limiting enzyme glucose-6-phosphate dehydrogenase.

The Embden-Meyerhof-Parnas pathway (glycolysis) and the pentose phosphate pathway play a central role in the degradation of glucose in organisms from all domains of life. In glycolysis, glucose is broken down to pyruvate to provide energy in the form of ATP, metabolic intermediates, and reduced nicotinamide adenine dinucleotide. In the pentose phosphate pathway, glucose is broken down to provide reducing equivalents in the form of NADPH and pentose sugars that are biosynthetic precursors of nucleic acids and amino acids (1)(2)(3). As the processes of glycolysis and the pentose phosphate pathway run in parallel, all cells have mechanisms to tightly regulate the flow of glucose into each pathway. On entering cells, the majority of glucose is phosphorylated to glucose 6-phosphate (Glu-6-P) in the first step of glycolysis. The intermediate Glu-6-P can then be further broken down in the process of glycolysis, or shunted into the pentose phosphate pathway. In the first committed step of the pentose phosphate pathway, Glu-6-P dehydrogenase catalyzes the dehydrogenation of Glu-6-P. This irreversible, rate-limiting step of the pentose phosphate pathway is typically highly regulated within cells, and therefore holds a prominent position in determining the overall flow of glucose into the pentose phosphate pathway (1,2).
The gluconate shunt is a less studied metabolic route that can facilitate the breakdown of glucose. In the gluconate shunt, glucose is oxidized by glucose dehydrogenase to gluconate, which is then phosphorylated by gluconate kinase to produce 6-phosphogluconate (4,5). As 6-phosphogluconate is the second intermediate of the pentose phosphate pathway, the gluconate shunt potentially creates a route for directing glucose into the pentose phosphate pathway that bypasses the rate-limiting enzyme Glu-6-P dehydrogenase. However, the only established role for the gluconate shunt is found in plants, algae, cyanobacteria, and some bacteria, which all use the Entner-Doudoroff (ED) 2 pathway to degrade glucose or gluconate (4,6,7). In the ED pathway 6-phosphogluconate is dehydrated to generate 2-keto-3-deoxygluconate-6-phosphate, which is then cleaved to generate pyruvate and glyceraldehyde 3-phosphate. The gluconate shunt is therefore a metabolic route that can be used to direct glucose and gluconate to the ED pathway (6,8). Despite this established role for the gluconate pathway, glucose dehydrogenase and gluconate kinase activities have been detected in mammals, fission yeast, and flies, which all lack the key ED pathway enzymes (5, 9 -13).
In this report we used metabolomics profiling to identify metabolites that accumulate in a fission yeast strain lacking the transcriptional repressor Loz1. In Schizosaccharomyces pombe Loz1 plays a central role in zinc homeostasis by regulating the expression of genes required for zinc uptake and zinc storage (14,15). Loz1 also regulates the availability of zinc in cells by controlling the levels of the non-essential, abundant zinc-binding enzyme alcohol dehydrogenase 1 (Adh1). Specifically, under zinc-limiting conditions, inactivation of Loz1 3 results in the derepression of an adh1 antisense transcript, and the strong antisense transcription in turn inhibits the expression of adh1 (16,17). Although the regulation of adh1 gene expression has been well characterized at a transcriptional level, it is largely unclear if the lowered levels of Adh1 in zinc-deficient cells also affects cellular metabolism. As Loz1 regulates the expression of adh1, and potentially other abundant zinc-binding proteins, the goal of this study was to determine whether the changes in transcription that are mediated by Loz1 also affect cell metabolism. By using metabolomic analyses to screen for metabolites whose levels were altered in fission yeast mutants with impaired Loz1 function, we found that the metabolite that showed the highest fold-increase in loz1⌬ cells was D-gluconate.
Here we show that the higher levels of D-gluconate in loz1⌬ cells results from increased expression of gcd1, a gene encoding a novel NADP ϩ -dependent glucose dehydrogenase. We also find that the function of gcd1 overlaps with zwf1, the gene encoding Glu-6-P dehydrogenase. We propose that in fission yeast the gluconate shunt creates an alternative route for directing glucose into the pentose phosphate pathway that bypasses the rate-limiting enzyme Glu-6-P dehydrogenase.

loz1⌬ cells have increased levels of gluconate
Our previous studies revealed that cells lacking the transcriptional repressor Loz1, or cells expressing the hypomorphic allele loz1-1, have impaired zinc homeostasis and reduced levels of the enzyme Adh1 (14). To determine whether impaired Loz1 activity affected cell metabolism, wild-type, loz1⌬, and loz1-1 cells were grown to exponential phase in the nutrientrich YES medium. Cells were then harvested and metabolites identified using both GC-MS and LC-MS. 314 unique metabolites were detected. 11 of these accumulated Ͼ2-fold in both loz1-1 and loz1⌬ cells relative to the wild-type control, including a variety of lipids and hydrolyzed phospholipids, the amino acid ergothioneine, the organic acids citrate, cis-aconitate, and gluconate, and the alcohol 2,3-butanediol (Table 1). Of these metabolites, gluconate, a naturally occurring derivative of glucose, showed the highest fold-increase in loz1⌬. As little is known about the biological function of gluconate in eukaryotes, we chose to further investigate how and why gluconate levels were regulated by Loz1.

The Idn1 gluconate kinase is required for the breakdown of gluconate
Studies of glucose metabolism in cell-free extracts of S. pombe have demonstrated the presence of two enzymes involved in gluconate metabolism, an NADP ϩ -dependent glucose dehydrogenase, which catalyzes the oxidation of D-glucose to D-gluconate, and gluconate kinase, which phosphorylates D-gluconate to produce 6-phosphogluconate ( Fig. 1) (5,18). A single hexose transporter named Ght3 has also been identified that facilitates uptake of gluconate when glucose is limiting (19). As the growth medium used for the metabolic profiling contained 3% glucose, which results in the strong repression of the Ght3 gluconate uptake system (supplemental Fig. S1), we predicted that the increased levels of gluconate in loz1⌬ cells were likely to be a result of altered expression of one or both of the enzymes involved in gluconate metabolism.
The simplest explanation for the increase in gluconate levels in the loz1 mutants was that the expression or activity of the gluconate kinase was reduced in these strains ( Fig. 2A). In S. pombe a single gene named idn1 encodes a protein with high sequence similarity to characterized gluconate kinases (supplemental Fig. S2). To test whether Idn1 was required for the breakdown of gluconate levels in vivo, wild-type and idn1⌬ cells were grown to exponential phase in YES medium and gluconate levels were measured by LC-MS/MS. As shown in Fig. 2B, idn1⌬ cells accumulated ϳ300-fold higher levels of gluconate relative to the wild-type control, consistent with Idn1 catalyzing the phosphorylation of gluconate to 6-phosphogluconate. As a complementary approach to examine Idn1 activity, we employed an enzymatic assay to measure gluconate levels in cell extracts. In these assays, intracellular gluconate levels are coupled to the consumption of 6-phosphogluconate (see "Experimental procedures"). When gluconate levels were measured by this method, idn1⌬ cells accumulated ϳ28-fold higher levels of gluconate relative to the wild-type control (Fig. 2C). This phenotype was also rescued by the introduction of a plasmid expressing an Idn1-GFP fusion protein from the constitutive pgk1 promoter (idn1⌬ Idn1-GFP). These results are consistent with the predicted role of Idn1 in phosphorylating gluconate in vivo.
As decreased expression of gluconate kinase would potentially lead to gluconate accumulating within cells, we next tested whether idn1 expression was altered in loz1⌬ cells. To determine whether the transcription of idn1 was controlled by Loz1, a construct was generated in which ϳ1200 bp of the idn1 promoter, extending from the open reading frame, was fused to the lacZ reporter gene. In wild-type and loz1⌬ cells expressing 3 The descriptive names of genes or proteins discussed in the text are as follows:  Zinc-dependent regulation of the Gcd1 glucose dehydrogenase this reporter, no significant changes in ␤-galactosidase activity were apparent following growth overnight in a zinc-limited minimal medium (ZL-EMM) supplemented with 0 -200 M zinc (Fig. 2D). The presence of a fully functional Loz1 also had little effect on idn1 mRNA levels, as similar levels of idn1 transcripts accumulated in wild-type, loz1-1, and loz1⌬ cells grown to exponential phase in the zinc-rich YES medium (Fig. 2E). This was in contrast to the Loz1-regulated zrt1 and adh4 mRNA controls, which accumulated to higher levels in loz1 mutants grown in the zinc-replete YES medium. Idn1 protein levels were also not regulated by Loz1 or zinc as similar levels of the functional Idn1-GFP protein were detected in idn1⌬ loz1⌬ and idn1⌬ cells (Fig. 2F), and in idn1⌬ cells grown over a range of zinc levels (Fig. 2G). Although the above experiments do not eliminate a model in which Loz1 influences the activity of Idn1, the results indicate that the ability of loz1⌬ and loz1-1 cells to accumulate gluconate is not a result of altered idn1 gene expression.

SPCC794.01c encodes an NADP ؉ -dependent glucose dehydrogenase
An alternative explanation for the increased gluconate in the loz1 mutants is that the unknown glucose dehydrogenase is expressed at higher levels in the loz1-1 and loz1⌬ cells. To gain insight into whether the first step of the gluconate shunt was regulated by Loz1, gluconate levels were compared in loz1⌬ idn1⌬ and idn1⌬ cells. As shown in Fig. 2, B and C, ϳ2-fold higher levels of gluconate accumulated in loz1⌬ idn1⌬ relative to idn1⌬. These results reveal that the increase in gluconate in loz1⌬ cells is independent of Idn1, and are consistent with gluconate synthesis being higher in loz1⌬.
As Loz1 represses its target gene expression when zinc is in excess, we predicted that the gene encoding the unknown glucose dehydrogenase might be expressed at higher levels in zincdeficient cells. When published microarray data were searched for putative NADP ϩ -dependent oxidoreductases that were regulated by zinc, we noted that SPCC794.01c, a gene encoding a putative NADP ϩ -dependent Glu-6-P dehydrogenase, was induced in response to zinc deficiency in multiple analyses (17,20). In S. pombe, two additional genes (zwf1 and SPAC3C7.13c) are also predicted to encode NADP ϩ -dependent Glc-6-P dehydrogenase, which catalyzes the first committed step of the pentose phosphate pathway (Fig. 1). Although all of the three putative Glc-6-P dehydrogenases from S. pombe have multiple conserved domains that are common to this family of proteins (supplemental Fig. S3), a notable exception was that the residues predicted to be involved in coordinating the phosphate moiety of Glc-6-P (21,22) were not conserved in SPCC794.01c (Fig. 3A). This lack of conservation suggested that SPCC794.01c might differ in its substrate specificity relative to other Glc-6-P dehydrogenases.
To examine the substrate specificity of SPCC794.01c, Histagged recombinant SPCC794.01c was purified from Escherichia coli (Fig. 3B). The activities of His-SPCC794.01c, the His tag alone (EV), were then assayed by measuring NADPH production in the presence of a given substrate (Fig. 3C). The activity of Glc-6-P dehydrogenase from Saccharomyces cerevisiae (Sc.Zwf1) was also measured as a control. As expected, incubation of the control enzyme (Sc.Zwf1) with each of the substrates led to an ϳ8.5-fold increase in NADPH levels in the presence of Glc-6-P relative to the His tag control (EV). This activity was specific to Glc-6-P as no increase in NADPH was observed in the presence of glucose or galactose substrates. In contrast, incubation of SPCC794.01c in the presence of glucose resulted in an ϳ12-fold increase in NADPH levels relative to the His tag control, and in the presence of Glc-6-P, a 3-fold increase in NADPH. No increase in NADPH was observed in the presence of a galactose substrate. Together these experiments reveal that SPCC794.01c has different substrate specificity than other Glc-6-P dehydrogenases and that it is able to use glucose as a substrate in vitro. As SPCC794.01c functions as a glucose dehydrogenase, it was named Gcd1 for Glucose dehydrogenase 1.
To test whether Gcd1 is necessary for the increased gluconate in loz1⌬ cells in vivo, wild-type, loz1⌬, and loz1⌬ gcd1⌬ cells were grown to exponential phase in YES medium and gluconate levels were measured by LC-MS/MS (Fig. 3D). As expected, higher levels of gluconate accumulated in loz1⌬ cells relative to the wild-type control. This increase was not apparent in loz1⌬ gcd1⌬ indicating that Gcd1 is required for the increased levels of gluconate in loz1⌬ cells.

Gluconate accumulates in zinc-limited cells in a manner that is dependent upon Gcd1
As Gcd1 was necessary for the Loz1-dependent increase in gluconate we next tested whether gcd1 was a Loz1 target gene. To determine whether the expression of gcd1 was dependent upon Loz1, a reporter construct containing ϳ1450 bp of the gcd1 promoter fused to the lacZ gene was integrated into the genome of wild-type and loz1⌬ cells. As Loz1 represses gene expression when zinc is in excess, ␤-galactosidase activity was examined following growth in ZL-EMM or ZL-EMM supplemented with 1-200 M Zn 2ϩ . An ϳ10-fold increase in ␤-galactosidase activity was observed in zinc-limited wild-type cells, whereas high levels of ␤-galactosidase activity were observed under all conditions in loz1⌬ (Fig. 4A). Cellular gcd1 mRNA levels were also dependent upon zinc and Loz1 (Fig. 4, B and C).

Zinc-dependent regulation of the Gcd1 glucose dehydrogenase
To examine the effects of zinc on Gcd1 protein levels, a strain was generated that expressed the endogenous Gcd1 protein fused to 13 myc epitope tags. As shown in Fig. 4D, there was an ϳ2-3-fold increase in levels of Gcd1-Myc protein in cells grown overnight under zinc-limiting conditions relative to the levels of Gcd1-Myc that accumulated in zinc-replete cells. Thus, Gcd1 accumulates in both zinc-limited and zinc-replete cells, yet higher levels accumulate in zinc-starved cells consistent with the increased expression of gcd1 under this condition.
To determine whether increased expression of gcd1 in zinclimited cells also resulted in gluconate accumulation, gluconate levels were measured in strains grown in ZL-EMM with or without a 200 M zinc supplement. Zinc-limited wild-type cells accumulated ϳ3-fold higher levels of gluconate relative to zincreplete cells (Fig. 4E). This fold-change was reduced in loz1⌬ mutants, whereas deletion of gcd1 resulted in lower levels of gluconate accumulating under zinc-limiting conditions. Although these results are consistent with the Loz1-dependent derepression of gcd1 leading to increased gluconate accumulation in zinc-limited cells, the levels of gluconate that accumu-lated were small relative to the large increase in gluconate seen in idn1⌬ (Fig. 4F). The increases in gluconate in idn1⌬ cells were largely dependent upon Gcd1 as lower levels of gluconate were detected in gcd1⌬ idn1⌬ cells compared with idn1⌬. Taken together these results are consistent with Gcd1 and Idn1 acting in the same pathway. They also suggest that most of the gluconate generated within cells is rapidly phosphorylated by Idn1.

The function of Gcd1 overlaps with Zwf1
Why would the expression of gcd1 be increased in response to zinc limitation? As Gcd1 is a NADP ϩ -dependent enzyme and the phosphorylation of gluconate by Idn1 requires 1 ATP, there is no obvious energetic advantage of using the gluconate shunt to generate 6-phosphogluconate instead of hexokinase and Glc-6-P dehydrogenase (Fig. 1). However, as these reactions could run in parallel, a potential explanation for the increased expression of gcd1 is that it allows increased carbon flux into the pentose phosphate pathway to increase NADPH regeneration and/or the levels of needed biosynthetic interme-

Zinc-dependent regulation of the Gcd1 glucose dehydrogenase
diates. In support of this hypothesis, there are precedents for the regulation of NADPH levels by zinc in other yeast (23). loz1⌬ cells also accumulate ergothioneine (Table 1), which is potentially generated from precursors supplied by the pentose phosphate pathway (24). To test whether the enzymes of the gluconate shunt influenced total cellular NADPH regeneration, NADP ϩ and NADPH levels were measured in wild-type and gcd1⌬ cells following growth to exponential phase in YES medium. No significant changes in the NADP ϩ /NADPH ratio were observed in gcd1⌬ compared with wild-type cells (supplemental Fig. S4). We also observed no obvious growth defect of gcd1⌬ cells in zinc-limiting medium, zinc-replete medium, or in the presence of strong oxidants such as H 2 O 2 ( Fig. 5 and data not shown).
As we were unable to find any phenotype that was a result of loss of gcd1, we next tested whether there was redundancy between the gluconate shunt and the first steps of the pentose phosphate pathway. As two genes (zwf1 and SPAC3C7.13c) encode Glc-6-P dehydrogenase in S. pombe, we initially performed RNA blot analysis to examine their expression in exponentially growing cells. We also examined the expression of gnd1, which encodes the pentose phosphate enzyme phosphogluconate dehydrogenase, as a control. The RNA blot analysis revealed that zwf1 and gnd1 mRNAs were abundantly expressed in wild-type and loz1⌬ cells (Fig. 4C). In contrast, we were unable to detect SPAC3C7.13c using this method. Due to the low expression of SPAC3C7.13c, our further experiments focused on zwf1. Genome-wide deletion studies in S. pombe suggest that zwf1⌬ cells are inviable (25). Nevertheless, we were able to delete zwf1 in a diploid background and isolate viable, albeit slow growing, haploid zwf1⌬ cells using tetrad dissection analysis. We were also able to isolate viable haploid strains lacking zwf1 and gcd1 from crosses of zwf1⌬ to gcd1⌬. When the growth phenotypes of these mutants were compared, cells lacking gcd1 did not display any significant growth defect on YES medium, or in this medium supplemented with zinc or the zinc chelator EDTA (Fig. 5). In contrast, zwf1⌬ cells had a strong growth defect under all conditions. Importantly, cells lacking both zwf1 and gcd1 exhibited a more severe growth defect than cells lacking only zwf1. These results suggest redundancy between Zwf1 and Gcd1, and that Gcd1 functions overlap with Zwf1.

Discussion
Enzymes involved in gluconate metabolism are found in many organisms, however, few studies have examined the biological role of these enzymes in vivo. Here we demonstrate that gcd1 encodes a novel NADP ϩ -dependent glucose dehydrogenase that is required for gluconate synthesis, and that the Idn1 gluconate kinase is necessary for gluconate breakdown in vivo. As previous studies have shown that Idn1 specifically phosphorylates gluconate to produce 6-phosphogluconate (5, 18), and Residues that are predicted to coordinate Glc-6-P are highlighted by white text. The alignment was performed using Clustal-Omega (EMBL-EBI). B, SDS-PAGE analysis of purified recombinant His-tagged SpSPCC794.01c, or the His tag (EV). Gels were stained with Coomassie Brilliant Blue. A protein ladder with sizes in kDa is shown on the left (C). Activity of 1 g of purified His tag (EV), His-tagged SpSPCC794.01c, or 0.5 units of S. cerevisiae Zwf1 in the presence of NADP ϩ and the indicated substrate. Reactions were allowed to go to completion and enzyme activity was determined by measuring NADPH generation via absorbance at 340 nm (D). The indicated strains were grown to exponential phase in YES medium and gluconate levels were measured by LC-MS/MS. Data represents the average values from 5 independent cultures with error bars representing standard deviations.

Zinc-dependent regulation of the Gcd1 glucose dehydrogenase
Gcd1 functions overlap with Zwf1, we propose that in fission yeast the gluconate shunt provides an alternative route for directing glucose into the pentose phosphate pathway that bypasses hexokinase and Glc-6-P dehydrogenase.
Studies of the gluconate shunt enzymes have so far mostly been limited to organisms that are able to metabolize 6-phosphogluconate by the ED pathway (6 -8). Despite the established role for these enzymes, glucose dehydrogenase and gluconate kinase activities have been detected in mammals and S. pombe, which both lack the key enzymes of the ED pathway (5,13,26). An important factor that has previously limited studies of gluconate metabolism in these organisms in vivo is that the genetic identity of the glucose dehydrogenase(s) was unclear. The identification of Gcd1 has therefore provided a means of examining gluconate metabolism in organisms without a functional ED pathway.
Gcd1 differs from traditional Glc-6-P dehydrogenases in that it contains nonconsensus amino acids in its putative Glc-6-P binding pocket. In the well characterized Glc-6-P dehydrogenases from humans and Leuconostoc mesenteroides, lysine and tyrosine residues within the Glc-6-P binding pocket are critical for Glc-6-P binding and catalysis (21,22,27). In Gcd1 the equivalent tyrosine is replaced by a phenylalanine, whereas the lysine is replaced with a glutamate (Fig. 3B). As Glc-6-P is itself negatively charged, the incorporation of a negatively charged glutamate into the Glc-6-P binding pocket instead of a positively charged lysine could be an important feature of Gcd1 that results in it preferentially using glucose instead of Glc-6-P as a substrate. An enzyme with similarities to Gcd1 is found in flies. Drosophila contains two putative Glc-6-P dehydrogenases, Zw and CG7140. Zw contains consensus amino acids in the Glc-6-P binding pocket, whereas CG7140 contains amino acid substitutions that replace the conserved tyrosine and lysine residues (supplemental Fig. S3). Although it is currently unknown if CG7140 is able to use glucose as a substrate, the similarity of CG7140 with Gcd1 suggests that it also may have a  Plates were incubated at 31°C for 3-6 days before photography.

Zinc-dependent regulation of the Gcd1 glucose dehydrogenase
broader substrate specificity. Other organisms may therefore express and utilize Gcd1-like NADP ϩ -dependent glucose dehydrogenases.
Although homologs of Gcd1 appear to be limited to fission yeast and flies, glucose dehydrogenase and gluconate kinase activities have been detected in mammals suggesting that they also may use a related pathway (9,12,13,26,28). It is therefore interesting to note that cells lacking gcd1 and idn1 accumulate ϳ15-fold higher levels of gluconate relative to the wild-type control (Fig. 4F). In addition, the molecular weight of the glucose dehydrogenase that was originally purified from S. pombe was 66.5 kDa (5) and the molecular mass of Gcd1 is 53.5 kDa. Together these observations suggest the presence of a Gcd1independent mechanism for generating gluconate in S. pombe. Thus, future studies characterizing this alternative pathway for gluconate production in fission yeast may provide further information regarding gluconate metabolism in mammals.
A clue to the function of the gluconate pathway in fission yeast was revealed by our observations that gcd1 is regulated at a transcriptional level by Loz1 and zinc, and that cells lacking zwf1 and gcd1 have a more severe growth defect than cells lacking only zwf1. Loz1 plays a central role in zinc homeostasis by regulating the expression of genes required for zinc uptake and zinc storage (14). The Loz1-dependent derepression of gcd1 in zinc-limited cells therefore suggests that increased flux of glucose through the gluconate pathway is important for zinc homeostasis or cell survival during longer periods of zinc limitation. Studies of zinc homeostasis in S. cerevisiae have found that MET30 expression is increased in zinc-limited cells (23). Met30 is a component of the SCF(Met30) ubiquitin ligase that targets the Met4 transcription factor for degradation. As Met4 typically activates the expression of genes required for the NADPH expending process of sulfate assimilation, it is thought that the zinc-dependent regulation of MET30 and Met4 conserves NADPH for defense against the increased levels of oxidative stress that arise from zinc starvation (23). Because flux of glucose through the gluconate shunt into the pentose phosphate pathway would also yield 2 NADPH molecules per glucose, increased expression of gcd1 could be an alternative mechanism to protect zinc-deficient cells from the oxidative stress associated with this growth condition. However, we did not see any major changes in NADPH regeneration in cells lacking gcd1 and we observed increased levels of NADP ϩ and NADPH in zwf1⌬ and zwf1⌬ gcd1⌬ cells (supplemental Fig. S4). One potential limitation with the above experiments is that other NADPH-producing pathways could be up-regulated and compensate for the loss of zwf1 and/or gcd1. In support of this hypothesis, in S. cerevisiae disruption of the ZWF1 gene results in a methionine auxotrophy, because high levels of NAPDH are required to sustain methionine biosynthesis (29,30). However in S. pombe, neither zwf1⌬ nor zwf1⌬ gcd1⌬ are methionine auxotrophs (data not shown). Thus, a better understanding of the net contributions of different routes of NADPH synthesis in S. pombe may shed additional light into the role of Zwf1 and Gcd1 in NADPH regeneration. Additional important experiments include the use of different labeled forms of glucose and gluconate to determine the extent to which glucose in cells is metabolized by the gluconate shunt and pentose phosphate pathway.
In addition to its role in NADPH regeneration, the pentose phosphate shunt is a major anabolic pathway that produces pentose sugars that are biosynthetic precursors of nucleic acids and amino acids. An alternative explanation for increased expression of gcd1 in zinc-limited cells is therefore that it generates needed biosynthetic precursors. In support of this hypothesis, higher levels of ergothioneine, a derivative of the amino acid histidine, were detected in loz1⌬ and loz1-1 cells (Table 1). Although it is not yet known if the Loz1-dependent increase in ergothioneine levels are dependent upon the gluconate shunt, ergothioneine has antioxidant functions (24). Increased ergothioneine production in zinc-deficient cells may therefore be a different metabolite route to protect cells from oxidative stress.
Another metabolite that accumulated to higher levels in loz1⌬ cells was 2,3-butandiol (Table 1). In S. cerevisiae, deletion of the alcohol dehydrogenase genes ADH1, ADH3, and ADH5 results in a large increase in the synthesize of 2,3-butandiol, potentially as a mechanism to reduce the buildup of the toxic metabolite acetaldehyde (31). Interesting future studies would therefore be to test whether the increase in 2,3-butandiol in loz1⌬ cells is a mechanism to help these cells contend with the low expression of adh1 in this strain. Another phenotype of cells expressing the loz1⌬ and loz1-1 alleles is that they hyperaccumulate zinc. Although most cells do not naturally hyperaccumulate zinc, an interesting exception is found in the prostate. Prostate epithelial cells are atypical in that they possess high levels of zinc in the cytosol and mitochondria to facilitate citrate secretion (32,33). The high levels of zinc in the mitochondria inhibit the activity of the Fe-S requiring enzyme aconitase that oxidizes citrate to isocitrate, which in turn allows citrate to build up for secretion. It is therefore noteworthy that citrate and cis-aconitate accumulated in loz1⌬ and loz1-1 cells. Thus, future studies to determine why the levels of metabolites such as citrate and ergothioneine are altered in loz1⌬ cells may further our knowledge of zinc homeostasis in yeast and other organisms.

Strains, media, and plasmids
A full list of strains and genotypes can be found in supplemental Table S1. Strains were either grown to exponential phase at 31°C in YES (yeast extract ϩ supplements) medium, or in a derivative of Edinburgh minimal medium that lacks zinc (ZL-EMM) (17). Strain ABY1116 (gcd1-13MYC::kanMX6) was generated by homologous recombination using standard procedures. The Pet32a-Gcd1 fusion construct was generated by introducing gcd1 into the XhoI/SacI site of the vector Pet32a. The idn1-lacZ and gcd1-lacZ reporters were generated by amplifying ϳ1 kb of the respective promoter regions with PCR primers that contained BamHI/EagI restriction sites. Each PCR fragment was then cloned into similar sites of the vector JK-lacZ (17). All vectors derived from plasmid JK148 were linearized with NruI before integration into the genome.

Zinc-dependent regulation of the Gcd1 glucose dehydrogenase RNA blot and immunoblot analysis
Total RNA was extracted by using hot acidic phenol and RNA blots run using standard protocols. Probes for RNA blots were generated from purified PCR products with the MAXIscript T7 kit (Thermo Fisher Scientific) according to the manufacturer's instructions. Crude protein extracts for immunoblot analysis were obtained using a trichloroacetic acid precipitation (34). Crude protein extracts were separated on 10% (w/v) SDS-PAGE gels, and immunoblot analysis was performed using standard procedures. Immunoblots were incubated with the primary antibodies anti-c-Myc (Sigma C3956), anti-GFP (Sigma G1544), and anti-Act1 (Abcam ab3280), and secondary antibodies IR-Dye800CW-conjugated anti-mouse IgG (LICOR) and IRDye680-conjugated anti-rabbit IgG (LICOR). Signal intensities were analyzed using the Odyssey Infrared image system (LICOR).

Recombinant protein purification and glucose dehydrogenase assays
BL21(DE3) pLysS cells containing Pet32a-gcd1 or the empty Pet32a vector were pregrown for 8 h in lysogeny broth ϩ 15% glycerol at 31°C. gcd1 expression was induced by 0.5 mM 1 mM isopropyl ␤-D-thiogalactopyranoside and cells were grown for a further 12-15 h at 31°C before lysis by sonication. Cell lysates were collected by centrifugation, and His-tagged proteins were purified using nickel-nitrilotriacetic acid Super flow (Qiagen) columns pre-equilibrated with lysis buffer (phosphate-buffered saline (PBS) ϩ 0.1% Tween ϩ 10 mM imidazole). Following washing, proteins were eluted with containing PBS ϩ 0.05%, Tween 20, and 250 mM imidazole. Protein fractions were stored in 50% (v/v) glycerol at Ϫ80°C. The concentration and purity of proteins was determined using the Bradford assay and SDS-PAGE analysis, respectively. 1 g of Gcd1, 1 g of Pet32a HIS tag, or 0.5 units of ScZwf1 (Sigma G7877) were diluted to 50 l in PBS ϩ 0.1% Tween 20. The protein was added to a solution containing 1 M MgCl 2 , 2.5 mM NADP (Calbiochem 481972), and 2.5 mM with one of the following sugars, glucose (Fisher Scientific D18), galactose (Fisher Scientific BP656), or Glc-6-P (Acros Organics 446980010). Reactions were allowed to go to completion and NADPH levels were measured via absorbance at 340 nm using a Synergy H1 plate reader.

Metabolomics and gluconate measurements
For metabolomic profiling, 50 ml of cells were grown to an A 600 of 1.0 in YES medium. Cells were then washed in ddH 2 O and pellets were frozen at Ϫ80°C. Further sample preparation, extraction of metabolites, ultra high performance LC-MS and GC-MS, and metabolite detection were performed by Metabolon. For further LC-MS/MS analyses to measure gluconate levels cells were grown as described above. Cells were then washed in ddH 2 O and cell pellets were frozen at Ϫ80°C. Frozen cell pellets were plunged into a boiling water bath. After the addition of 1 ml of 65°C 50 M [U-13 C]fumarate, samples were boiled for a further 10 min, vortexed for 5 min in the presence of zirconia beads, boiled for a further 5 min, and then vortexed for a final 5 min. Cell extracts were centrifuged for 10 min, 4°C, 14,000 ϫ g, and the supernatant filtered through 0.2-m cellu-lose acetate filters. Filtered supernatants were frozen at Ϫ80°C and lyophilized until further analysis. Dry pellets were resuspended in 1 ml of ddH 2 O and filtered through a 3-kDa Amicon Ultra 0.5-ml filter. LC was performed with a ultra high-pressure liquid chromatography 1290 from Agilent Technologies, using an IonPac AS11 column (250 ϫ 2 mm, Dionex) and a guard column AG11 (50 ϫ 2 mm, Dionex) at a flow rate of 0.35 ml/min. A gradient of KOH was generated as previously described (35,36). The MS/MS analysis was performed with a QTRAP 5500, from AB Sciex (Framingham, MA) in negative polarity using multiple reaction monitoring mode. The gluconate was monitored using the parent/daughter transition of 195.2/74.9 and the following parameters: Ϫ50 V declustering potential, Ϫ26 V collision energy, Ϫ45 V cell exit potential, and Ϫ10V entrance potential. Under our chromatographic conditions (anion exchange chromatography with basic eluent), the only form of gluconate (gluconic acid, gluconate, and gluconolactone) detected was gluconate. The level of gluconate was quantified using external standard curves of a commercially available gluconate standard, and by using [U-13 C 4 ]fumarate as an internal standard to account for any loss of sample during extraction, filtration etc. For gluconate enzymatic assays, cell pellets were resuspended in 1 ml of H 2 O. Cells suspensions were boiled for 5 min and lysed by bead beating for 5 min. Gluconate levels were then measured using the D-gluconate/Dglucono-␦-lactone assay kit (Megazyme) per the manufacturer's instructions. Briefly in these assays, residual 6-phosphogluconate was removed from boiled cell extracts by incubation with ATP, NADP ϩ , and 6-phosphogluconate dehydrogenase. After the removal of 6-phosphogluconate, the basal level of NADPH was determined by measuring fluorescence excitation: 340 nm and emission 440 nm. IdnK was then added to generate 6-phosphogluconate from gluconate. As 6-phosphogluconate is a substrate for 6-phosphogluconate dehydrogenase further increases in NADPH levels are proportional to the levels of gluconate within cells. Final gluconate levels were determined by comparing the increase in NADPH levels to that obtained with known concentrations of gluconate. All samples were normalized to protein concentration.

␤-Galactosidase assays
␤-Galactosidase assays were performed as described previously, and activity units were calculated as follows: (⌬A 420 ϫ 1000)/(min ϫ ml of culture ϫ culture absorbance at 600 nm). Errors bars represent mean Ϯ S.D. from three independent experiments.

NADP ؉ /NADPH measurements
This procedure was carried out with a modified version of the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide and phenazine ethosulfate cycling assay described by (23,37). 5-ml cultures were grown to an A 600 ϳ1.0 in YES medium. 2 ϫ 1.5 ml of cells were then washed with ddH 2 O, and were resuspended in 250 l of 0.1 M HCl or 0.1 M NaOH to measure NADP ϩ and NADPH levels, respectively. Cells were broken open and the endogenous proteins were denatured by freezing at Ϫ80°C for 30 min, boiling for 3 min, and vortexing with glass beads for 5 min. Cell debris was removed by centrifugation and Zinc-dependent regulation of the Gcd1 glucose dehydrogenase cycling reactions set up in 96-well plates. For each reaction 20 l of supernatant was added to 100 l of 0.1 M Tris, pH 7.4. Samples were mixed by low speed vortexing and the reaction initiated by the addition of 100 l of detection buffer (0.1 M Tris, pH 7.4, 0.1% (v/v) Tween 20, 10 mM MgCl 2 , 10 mM Glc-6-P (Acros Organics 446980010), 1 mM 3-(4,5-dimethylthiazol-2yl)-2,5-diphenyltetrazolium bromide (Sigma M5655), 0.2 mM phenazine ethosulfate (Sigma P4544), and 5 units of ScZwf1 (Sigma G7877)). Reactions were kept in the dark and color development was monitored spectrophotometrically at 595 nm. Reactions were stopped by the addition of 0.45 M NaCl and the concentrations of NADP ϩ and NADPH were determined based of a standard curve of known concentrations of NADPH. Final concentrations were normalized to the initial cell density of the culture.