Proteomic and Functional Consequences of Hexokinase Deficiency in Glucose-repressible Kluyveromyces lactis

The analysis of glucose signaling in the Crabtree-positive eukaryotic model organism Saccharomyces cerevisiae has disclosed a dual role of its hexokinase ScHxk2, which acts as a glycolytic enzyme and key signal transducer adapting central metabolism to glucose availability. In order to identify evolutionarily conserved characteristics of hexokinase structure and function, the cellular response of the Crabtree-negative yeast Kluyveromyces lactis to rag5 null mutation and concomitant deficiency of its unique hexokinase KlHxk1 was analyzed by means of difference gel electrophoresis. In total, 2,851 fluorescent spots containing different protein species were detected in the master gel representing all of the K. lactis proteins that were solubilized from glucose-grown KlHxk1 wild-type and mutant cells. Mass spectrometric peptide analysis identified 45 individual hexokinase-dependent proteins related to carbohydrate, short-chain fatty acid and tricarboxylic acid metabolism as well as to amino acid and protein turnover, but also to general stress response and chromatin remodeling, which occurred as a consequence of KlHxk1 deficiency at a minimum 3-fold enhanced or reduced level in the mutant proteome. In addition, three proteins exhibiting homology to 2-methylcitrate cycle enzymes of S. cerevisiae were detected at increased concentrations, suggesting a stimulation of pyruvate formation from amino acids and/or fatty acids. Experimental validation of the difference gel electrophoresis approach by post-lysis dimethyl labeling largely confirmed the abundance changes detected in the mutant proteome via the former method. Taking into consideration the high proportion of identified hexokinase-dependent proteins exhibiting increased proteomic levels, KlHxk1 is likely to have a repressive function in a multitude of metabolic pathways. The proteomic alterations detected in the mutant classify KlHxk1 as a multifunctional enzyme and support the view of evolutionary conservation of dual-role hexokinases even in organisms that are less specialized than S. cerevisiae in terms of glucose utilization.

Glucose is the preferred substrate for ATP regeneration, the formation of metabolic precursors, and the maintenance of a reductive potential in eukaryotes from yeast to humans (1,2). In accordance with its pivotal significance to central metabolism, the utilization of glucose is strictly controlled by the sugar itself (2). Several lines of experimental evidence classify hexokinases as dual-role enzymes that, in addition to their involvement in the uptake and phosphorylation of glucose, contribute to glucose sensing and signaling in various eukaryotic species (3)(4)(5). The exploration of glucose-dependent signal transduction in turn is of outstanding importance for the understanding of metabolic disorders such as diabetes (6) and the switch from respiration to aerobic glycolysis occurring in early tumor development (7). Substantial insights into the molecular mechanisms of glucose sensing and signaling and, in particular, the role of hexokinases were obtained from studies using the traditional eukaryotic model organism Saccharomyces cerevisiae. The phenomenon of glucose repression in this yeast is mediated through effects on transcription, RNA stability, and protein degradation (8). Most interestingly, glucose abundance is accompanied by the translocation of hexokinase isoenzyme ScHxk2 and transcription factor ScMig1 into the nucleus, where both proteins act as constituents of a heterooligomeric repressor complex preventing the expression of glucose-repressible genes, whereas glucose exhaustion results in the retranslocation of both proteins into the cytosol (3). Taking into consideration the existence of genes predicted to encode Mig1 homologs in a variety of yeasts (8), evolutionary conservation of transcriptional regulation including dual-role hexokinases and Mig1-type transcriptional repressors seems conceivable.
The use of S. cerevisiae as a model organism for higher eukaryotes is limited by its Crabtree-positive phenotype, which is reflected by a predominance of fermentation over respiration despite the presence of oxygen (9). In addition, the genetic redundancy resulting from a whole-genome duplication in the evolutionary history of the genus (10) complicates the functional analysis of its genes. With respect to glucose phosphorylation and signaling, the genome duplication event is associated with the expression of two hexokinases (ScHxk1 and ScHxk2), one glucokinase (ScGlk1), and one glucokinase paralog (ScEmi2). In comparison, the genome of the Crabtreenegative model organism Kluyveromyces lactis, in which no genome duplication has occurred (10), encodes a single hexokinase (KlHxk1) (11) and a single glucokinase (KlGlk1) (12). Prominent physiological features of K. lactis are its growth on lactose as the sole carbon source (13), the limited exploitation of its glucose uptake capacity during aerobic growth, and the low extent of aerobic ethanol accumulation (14). Reduced sensitivity to glucose repression is another functional peculiarity of K. lactis that is highly strain-dependent and seems to be mediated by multiple genes (15)(16)(17). Contrary to the situation in S. cerevisiae, where components of the respiratory system are subject to glucose repression (18), respiration in K. lactis remains essentially unaffected when glucose is abundantly available (19). Repression by glucose is more pronounced in strain JA6 than in other strains of the same genus and appears to be related to the existence of two tandemly arranged glucose transporter genes (KHT1 and KHT2) at the RAG1 locus (17), whereas the genes encoding the unique low-affinity glucose permease Rag1 (15-17, 20 -22) and the high-affinity glucose transporter KlHgt1 present in sequenced reference strain CBS2359/152 are both lacking in the JA6 genome (23). KHT1 and RAG1 encode proteins that differ only at their C termini and are similarly regulated at the transcriptional level (21). Deletion of KHT1 does not prevent fermentative growth when respiration is blocked by antimycin A (resistance against antimycin A on glucose ϭ Rag ϩ phenotype), as KHT2 expression apparently provides sufficient glucose uptake capacity. The influence of rag5 mutations on high-affinity and low-affinity glucose transport (11,20,24) and their coincidence with relief from glucose repression (16,25) identified RAG5 (KLLA0D11352g) and its gene product KlHxk1 as further components of the glucose signaling cascade in K. lactis.
The present paper is focused on the functional significance of the unique hexokinase KlHxk1 expressed in glucose-repressible strain JA6 of K. lactis (25). Structural investigations revealed prominent molecular similarities of KlHxk1 and hexokinase isoenzyme ScHxk2 of S. cerevisiae (26) suggesting similar functions. The two kinases exhibit 73% identity on the protein level and carry similar amino acids at 61 additional positions. KlHxk1 and ScHxk2 establish dynamic monomerhomodimer equilibria (25,27) that, according to the Ostwald dilution law, depend on the enzyme concentration in a range that is likely to meet physiological conditions. 1 Homodimer stability is drastically reduced as a consequence of phosphorylation of the conserved amino acid serine-15 present in the highly similar N-termini of both enzymes (28,29). In the case of KlHxk1, the crystal structure of a symmetrical ring-shaped homodimer of the phosphorylated enzyme displays two structurally equivalent phosphoserine-15 residues in the intersubunit interface at positions that are critical for dimer stability (29). The regulatory significance of serine-15 phosphorylation was made apparent by the finding that in S. cerevisiae this modification promotes the association of nuclear ScHxk2 with the ScXpo1 exportin and thereby facilitates translocation of the enzyme into the cytosol when glucose availability is low (30,31). Phosphoserine-15 ScHxk2 is dephosphorylated by the ScGlc7/ScReg1 phosphoprotein phosphatase complex (3,32) to allow its nuclear import and contribution to glucose repression when the sugar is abundantly available again. In contrast, the subcellular distribution of KlHxk1 and its dependence on glucose availability are still unknown. Serine-15, however, is part of a conserved N-terminal sequence motif (K7-M16) that in S. cerevisiae is essentially involved in the shuttle of ScHxk2 between cytosol and nucleus and, in particular, in glucose signaling at the level of the ScMig1 repressor complex (3,32). Recently, protein kinases ScTda1/ ScYmr291w (33) and ScSnf1 (3) were reported to be implicated in ScHxk2 phosphorylation at serine-15 in response to glucose limitation (30). In comparison, neither the condition triggering KlHxk1 phosphorylation at the equivalent site nor the corresponding protein kinase of K. lactis are known. The identification of a K. lactis gene encoding a hypothetical protein (KLLA0A09713p) of still unknown function that is homologous to ScTda1/ScYmr291w, however, provokes the hypothesis that KlHxk1 and the upstream KlHxk1-S15 kinase might fulfill key roles in glucose-dependent signal transduction in K. lactis.
In order to verify the expectation that KlHxk1 represents a dual-function enzyme that, in addition to its catalytic function in glycolysis, contributes to the transcriptional control of central metabolic pathways in K. lactis, the proteomic consequences of RAG5 disruption and concomitant KlHxk1 deficiency were studied by means of difference gel electropho-resis (DIGE) 2 (34). The DIGE approach compared RAG5 wildtype strain JA6 (subsequently referred to as "wild-type") as a reference to the congenic rag5 null mutant strain JA6⌬rag5 (subsequently referred to as "mutant"), which lacks both KlHxk1 protein and hexokinase catalytic activity. With consideration of the involvement of KlHxk1 in glucose repression (16), the proteomic response of K. lactis to the rag5 null mutation was analyzed in detail upon growth in high-glucose medium and was compared with the proteomic changes taking place in media containing galactose or glycerol as the sole carbon source. The overall experimental outcome not only confirms and extends current knowledge on KlHxk1 involvement in glucose sensing and signaling (16,20,24), but also indicates novel functions of the RAG5 gene and its gene product KlHxk1 in chromatin remodeling and general stress response in addition to roles in protein turnover and in tricarboxylic acid, short-chain fatty acid, and amino acid metabolism.
Cell Culture for DIGE Analysis-Strains were grown in yeast nitrogen base (YNB) medium (pH 5.5) consisting of 0.69% (w/v) YNB without amino acids supplemented with adenine, uracil, and amino acids according to Ref. 36 at 30°C and 200 rpm. Uracil was omitted when plasmid selection was required. Glucose, galactose, or glycerol was added at an initial concentration of 2% (w/v). Yeast cells were grown stepwise in YNB medium in precultures of 5 ml for 16 h (1), 25 ml for 8 h (2), and 100 ml for 16 h (3) in the presence of the indicated carbon source. The initial OD (600 nm/1 cm) of preculture 3 was 0.02, whereas cultivation of the main cell culture (500 ml) used for protein extraction started at an OD (600 nm/1 cm) of 0.3. To prepare the latter culture, cells from preculture 3 were washed twice using YNB medium without a carbon source and resuspended in YNB medium supplemented with the indicated carbon source. The cells were finally incubated until an OD (600 nm/1 cm) of 1 was reached, collected by centrifugation (3,500 ϫ g, 8 min, 4°C), washed twice with ice-cold double-distilled water, frozen, and stored at Ϫ80°C until further use. Because of impaired growth of strain JA6⌬rag5 on glucose, precultures 1-3 of the mutant strain were supplemented with 2% (w/v) galactose as the sole carbon source and precultures 2 and 3 were scaled up to give 50 ml and 200 ml, respectively, to allow sufficient biomass accumulation. Cells from the latter preculture 3 were washed with and resuspended in YNB medium containing 2% (w/v) glucose to give another preculture (4) (500 ml) with a starting OD (600 nm/1 cm) of 1. Preculture 4 was cultivated for 16 h under the above conditions. Glucose-grown mutant strain main cell cultures were finally prepared from preculture 4 using YNB medium supplemented with 2% (w/v) glucose and propagated from a starting OD (600 nm/1 cm) of 1 for 6 h. Cells were harvested, washed, and frozen as specified.
Protein Extraction-The frozen yeast cells were thawed and resuspended on ice in 3 ml of DIGE lysis buffer (30 mM Tris/HCl, 7 M urea, 2 M thiourea, 4% (w/v) CHAPS, 10 mM dithiothreitol (DTT), 1x protease inhibitor mix, pH 9.1) per gram of wet cell mass. Cells were disrupted by repeated use of a French ® Pressure Cell Press (SLM Aminco, Rochester, NY) equipped with a mini cell at 20,000 psi (138 MPa) followed by nine cycles of 10 s of sonication and 10 s of cooling on ice using a UP100H-type ultrasonic processor (Dr. Hielscher, Teltow, Germany) at 20% output corresponding to a 28-m amplitude. Samples were re-collected via centrifugation (5,000 ϫ g, 30 s, 4°C) after the third and sixth cycle. Insoluble material was removed after the last cycle of sonication and cooling by centrifugation (16,000 ϫ g, 15 min, 4°C). The resulting supernatants containing the solubilized proteins (subsequently referred to as "proteomes") were subjected to protein determination employing the RC DC™ Protein Assay (Bio-Rad, Hercules, CA), frozen, and stored at Ϫ80°C until further use.
Difference Gel Electrophoresis-Equal amounts of protein extracted from 10 cultures of wild-type and mutant cells grown in media containing 2% (w/v) glucose, 2% (w/v) galactose, or 2% (w/v) glycerol (60 samples in total) were mixed to set up an internal standard for multiplex matching of DIGE images, spot normalization, and calculation of abundance changes. Internal standard, wild-type, and mutant proteomes were differentially labeled with N-hydroxysuccinimidyl ester derivatives of the cyanine dyes Cy2, Cy3, and Cy5 following the manufacturer's instructions for minimal labeling (GE Healthcare, Munich, Germany). For analytical gels, 50 g of protein were labeled with 200 pmol of fluorescent dye derivative in a total volume of 25 l using DIGE lysis buffer as the solvent. Cy2 was exclusively used for labeling the internal standard that was run on each gel in parallel with two wild-type and/or mutant proteomes. To avoid any labeling bias, Cy3 and Cy5 were randomly utilized to label the wild-type and mutant proteomes. The labeling reaction was stopped by the addition of 1 l of 10 mM L-lysine. Three differentially labeled proteomes were mixed, reduced by the addition of 3.75 l of 2 M DTT corresponding to a final concentration of 50 mM of the reductant, supplemented with 0.75 l of immobilized pH gradient (IPG) buffer pH 3-10 (GE Healthcare, Munich, Germany), and completed using sample rehydration solution (7 M urea, 2 M thiourea, 4% (w/v) CHAPS) to obtain a final volume of 150 l. Isoelectric focusing was performed using an Ettan IPGphor 3 unit (GE Healthcare, Munich, Germany). The samples were cuploaded at the anodic side of IPG strips (24 cm, pH 3-10, linear) previously treated with 450 l of strip rehydration buffer (0.5% (v/v) IPG buffer, pH 3-10, 7 M urea, 2 M thiourea, 4% (w/v) CHAPS, 1.2% (v/v) DeStreak reagent) and separated at 20°C. The following conditions were applied, which correspond to a total of 114,575 Vh: 150 V/3 h, linear gradient; 300 V/3 h, linear gradient; 1,000 V/3 h, linear gradient; 10,000 V/3 h, linear gradient; and 10,000 V/95,000 Vh, step. The strips were treated with equilibration buffer (50 mM Tris/HCl, 6 M urea, 20% (v/v) glycerol, 2% (w/v) SDS, 130 mM DTT, pH 8.8) for 20 min and then allowed to equilibrate for another 20 min with a modified buffer lacking DTT but containing 135 mM iodoacetamide for sulfhydryl group alkylation. Subsequent SDS-PAGE using 12.5% (w/v) gels (20.5 ϫ 25.5 cm) was carried out at 25°C and 5 mA per gel for the first hour and 20 W per gel until completion in an Ettan Daltsix system (GE Healthcare, Munich, Germany).
Image Acquisition and Data Analysis-Digital DIGE images of wildtype, mutant, and internal standard proteomes were generated by scanning the two-dimensional gels (30 in total) with a Typhoon 9410 2 The abbreviations used are: DIGE, difference gel electrophoresis; YNB, yeast nitrogen base; DTT, dithiothreitol; IPG, immobilized pH gradient; AAR, average abundance ratio; DML, post-lysis dimethyl labeling; KlHxk1, hexokinase isoenzyme 1 of Kluyveromyces lactis.
Variable Mode Imager (GE Healthcare, Munich, Germany) using excitation/emission wavelengths of 488 nm/520 nm for Cy2, 532 nm/ 580 nm for Cy3, and 633 nm/670 nm for Cy5. Images (90 in total) were matched and normalized and equivalent protein spots in different gels were identified by employing the fully automated computer-assisted alignment module (batch processor) of DeCyder 2-D Differential Analysis Software Version 7.0 (GE Healthcare, Munich, Germany). Matches were manually revised using the biological variation analysis module. Statistical data treatment employed the extended data analysis component of DeCyder 2-D Differential Analysis Software Version 7.0 (GE Healthcare, Munich, Germany). The significance of proteomic differences between wild-type and mutant strains grown on the same carbon source was analyzed via t test for independent samples. For type-I error adjustment, the method of Benjamini and Hochberg (37) integrated in the DeCyder software was applied using a false discovery rate of 5%, which corresponded to a critical p value of 1.7 ϫ 10 Ϫ3 , indicating rejection of the null hypotheses for spots with lower p values. In order to limit the extent of false negative decisions, normalized spots exhibiting at least 3-fold abundance changes (average abundance ratio (AAR) Ն 3 or Յ Ϫ3) were considered in the analysis of the proteomic consequences of KlHxk1 deficiency. In the context of the present paper, the AAR indicates the ratio of the average amount of a protein in the mutant proteome to that in the reference proteome for proteins occurring at an increased level or the negative inverse ratio for proteins occurring at a decreased level. Proteins fulfilling the above criteria are subsequently referred to as "hexokinase-dependent proteins" (for statistical details, see "Discussion").
Protein Identification-Gel discs containing hexokinase-dependent proteins were excised from preparative two-dimensional gels loaded with 150 to 1,000 g of protein and stained using colloidal Coomassie Brilliant Blue G-250 (38). Prior to excision, spot patterns were manually aligned with the fluorescent image of the master gel to allow for correct spot localization. In the case of low-abundance proteins, the alignment was facilitated by the addition of Cy2-labeled internal standard (50 g per gel) to the unlabeled proteome. The excised gel discs were washed twice with deionized water and once using 50% (v/v) acetonitrile in 25 mM ammonium bicarbonate solution, shrunk by dehydration in 100% (v/v) acetonitrile, and dried in a vacuum centrifuge (240 ϫ g, 30 min, 25°C). The dried material was incubated overnight with 20 l of mass-spectrometry-grade trypsin (Promega, Mannheim, Germany) solution (2.5 g/ml, 50 mM ammonium bicarbonate used as the solvent) at pH 7 and 37°C. For peptide extraction, the proteolyzed samples were mixed with 20 l of 0.5% (v/v) trifluoroacetic acid (TFA) in acetonitrile, sonicated for 5 min in an ultrasonic bath sonicator, and finally incubated for another 5 min at 20°C in a shaker at 800 rpm. After centrifugation, supernatant removal, and repetition of the extraction procedure using 20 l of 100% (v/v) acetonitrile per sample, the pooled supernatants were vacuum dried. The extracted peptides were dissolved in 5 l of 0.1% (v/v) TFA in water, sonicated for 5 min as above, and incubated at 20°C for another 5 min. Aliquots (0.5 to 1.0 l) of each sample were pipetted on an AnchorChip™ target plate (Bruker Daltonics, Leipzig, Germany), allowed to sit for 3 min at 20°C to 25°C, and subsequently mixed on the plate with the same volume of ␣-cyano-4-hydroxycinnamic acid. Peptide analysis was performed with an Ultraflex TM Automated Highperformance MALDI-TOF/TOF Mass Spectrometry System (Bruker Daltonics, Bremen, Germany) in reflectron mode. Peak lists were generated using Bruker Daltonics flexAnalysis Software Version 2.2. Peptide mass fingerprint analysis employed MASCOT Application Software Version 2.2 (Matrix Science, London, UK) using the following criteria for searching against the NCBInr 20121118 database (21,581,546 sequences, 7,400,919,093 residues): taxonomy, fungi; mass accuracy, 50 to 90 ppm; fixed modifications, carbamidomethy-lation of cysteine; variable modifications, methionine oxidation and deamidation of asparagine and/or glutamine; mass values, monoisotopic; maximum number of missed cleavage sites, one. Known peptide mass values of contaminating tryptic keratin fragments and autodigestion products of trypsin were excluded from the database search. Peptide mass fingerprint scores higher than 74 (p Ͻ 0.05) were considered to indicate significant protein identification. For insignificant peptide mass fingerprint scores, MALDI-TOF/TOF sequencing of selected tryptic peptides was performed applying a mass tolerance setting for fragment ions of Ϯ0.8 Da. Extended database searches included the UniProtKB/Swiss-Prot 2012_10 (538,259 sequences, 191,113,170 residues) and UniProtKB/TrEMBL 2012_10 (27,122,814 sequences, 8,765,290,755 residues) databases.
In order to reliably identify low-abundance proteins and verify low-score hexokinase-dependent proteins that were not adequately characterized with the MALDI-TOF/TOF approach, the extracted tryptic peptides were purified on C18 StageTips (39) and separated on a Proxeon Easy nLC nano-HPLC system (Proxeon Biosystems, Odense, Denmark) fitted with an in-house packed 15-cm analytical reverse-phase column with a 75-m inner diameter containing Reprosil-AQ Pur 3-m C18 reverse-phase beads (Dr. Maisch GmbH, Ammerbuch-Entringen, Germany). Peptide separation was accomplished by applying a 65-min gradient from 5% to 60% (v/v) acetonitrile in 0.1% formic acid. The effluent was directly electrosprayed into a Velos-Orbitrap mass spectrometer (Thermo Fisher Scientific, Waltham, MA) equipped with an electrospray ion source (Proxeon Biosystems, Odense, Denmark). The recorded spectra were analyzed using MaxQuant Software Package Version 1.2.2.5 (40) by matching the data to the UniProt K. lactis database (version 2012) with a false discovery rate of 1% (peptides and proteins). The fixed and variable modifications were set to carbamidomethylation of cysteines and methionine oxidation, respectively. For further data analysis, the R statistical software package was used.
Experimental Determination of pI and M r Marker proteins for twodimensional electrophoresis (Sigma-Aldrich, Taufkirchen, Germany) were labeled with Cy3 and added to the Cy2-labeled internal standard. Their DIGE coordinates were used to determine experimental pI values (pI(exp)) by employing an algorithm integrated in DeCyder Software Version 7.0 (GE Healthcare, Munich, Germany). Experimental M r values (M r (exp)) are based on manually determined migration distances of the above marker proteins complemented by pre-stained marker proteins for one-dimensional electrophoresis (PageRuler™ Prestained Protein Ladder, Fermentas, St. Leon-Rot, Germany) and application of GraphPad Prism Software Version 4.03 (GraphPad Software, La Jolla, CA). M r (exp) values of individual K. lactis proteins were calculated from manually determined migration distances using a best-fit equation provided by the GraphPad Prism software.
DIGE Analysis of KlHxk1 Phosphorylation-Spike-ins of purified serine-15-phosphorylated and unphosphorylated KlHxk1 (29) were prepared by differentially labeling 5 g of each of the two proteins with Cy3 (green fluorescence) and Cy5 (red fluorescence), respectively, in a final volume of 25 l. DIGE analysis performed to identify the two KlHxk1 species in the two-dimensional images and twodimensional gels employed samples consisting of 0.2 g of each of the two fluorescently labeled enzymes, 50 g of Cy2-labeled internal standard, and 99.6 g of unlabeled internal standard to reach a final loading amount of 150 g of protein per gel. In order to improve spot separation and detection, twice the amount of each of the two labeled KlHxk1 forms, 50 g of Cy2-labeled internal standard, and 99.2 g of unlabeled internal standard were analyzed on pH 4 -7 IPG strips. Preparative two-dimensional gels covering the same pH range were loaded with 1 mg of unlabeled internal standard and subjected to blue silver staining (38). The proteins present in the discs cut out of these gels at the DIGE coordinates of the fluorescently labeled unphospho-rylated and serine-15-phosphorylated KlHxk1 were subjected to phosphopeptide analysis as described below.
Phosphopeptide Analysis-Purified unphosphorylated and serine-15 phosphorylated KlHxk1 (29) was subjected to SDS-PAGE in 10% (w/v) acrylamide gels followed by in-gel proteolysis according to a two-step protocol using endopeptidases LysC (Wako, Osaka, Japan) and trypsin (Promega, Mannheim, Germany) (41). The recovered peptides were desalted on C18 Empore Solid Phase Extraction Disks (3M, Neuss, Germany) (39), separated on a 15-cm analytical reversephase column with a 75-m inner diameter containing Reprosil-AQ Pur 3-m C18 reverse-phase beads (Dr. Maisch GmbH, Ammerbuch-Entringen, Germany) using a gradient of 5% to 60% acetonitrile in 0.1% formic acid, and ionized on a Proxeon ion source (Proxeon Biosystems, Odense, Denmark). Fragment spectra of the phosphopeptides were collected using a Velos-Orbitrap mass spectrometer (Thermo Fisher Scientific, Waltham, MA) in multistage activation mode. The spectra were analyzed with MaxQuant Software Package Version 1.2.2.5 (40) and used to develop a method for single reaction monitoring. In order to verify the single reaction monitoring method and determine the retention time during reverse-phase chromatography, phosphopeptide KGS(p)MADVPANIMEQIHGIETIFTVSSEK corresponding to KlHxk1 amino acids 13-40 was synthesized via Fmoc (N-(9-fluorenyl)methoxycarbonyl) solid phase synthesis on a ResPep SL Microscale Automated Synthesizer (Intavis, Cologne, Germany). The proteins present in the discs cut out of the preparative twodimensional gels at the DIGE coordinates of the fluorescently labeled unphosphorylated and serine-15-phosphorylated KlHxk1 were converted to peptides according to an in-gel digestion protocol using sequencing-grade trypsin (Promega, Mannheim, Germany) (42). Recovered peptides were analyzed by applying the chromatographic setup described above, but recorded on a Q-TRAP 5500 mass spectrometer (AB Sciex, Toronto, ON, Canada) in multiple reaction monitoring mode. The data were quantified using MultiQuant Software Package Version 1.2.0.6 (AB Sciex, Toronto, ON, Canada) and further analyzed by use of the R statistical analysis language.
Homology Analysis and Assignment of Protein Functions-Despite substantial genomic and physiological differences between S. cerevisiae and K. lactis, about 82% of the K. lactis metabolic genes are annotated as S. cerevisiae homologs (43). Therefore, S. cerevisiae was taken as a primary reference for the annotation of hexokinasedependent proteins of K. lactis using the NCBI HomoloGene database system (www.ncbi.nlm.nih.gov/homologene) when evidence for protein existence on the protein or transcript level according to Uni-ProtKB was missing. The annotation of K. lactis proteins exhibiting homology to two or more S. cerevisiae proteins always refers to the S. cerevisiae protein with the higher score as calculated with BLASTp. In three cases, homology analysis identified genes existing only in yeast species other than S. cerevisiae. The assignment of molecular functions and biological processes to K. lactis proteins or their yeast homologs employed the GO and KEGG databases, respectively. In two indicated cases, functional assignment was complemented by literature information.
Immunodetection of KlHxk1-The anti-KlHxk1 polyclonal antibody used in the present study was generated by immunizing rabbits with purified nonphosphorylated KlHxk1 (29) and subjecting the antiserum to immunoaffinity chromatography on a HiTrap NHS-activated HP column (GE Healthcare, Munich, Germany) reacted with purified KlGlk1 to eliminate cross-reacting immunoglobulin. Strains JA6, JA6⌬rag5, and JA6⌬rag5/pTSRAG5 were pre-grown in YNB medium supplemented with 2% (w/v) galactose for 15 h. The cells were washed twice using YNB medium without a carbon source, resuspended in 25 ml of the YNB-galactose medium to give an OD (600 nm/1 cm) of 0.3, propagated until an OD (600 nm/1 cm) of 1.0 was reached, and washed with ice-cold lysis buffer consisting of 50 mM imidazole, 50 mM NaCl, 5 mM -aminocaproic acid, pH 7.0, supplemented with 1 mM 4-(2-aminoethyl)-benzenesulfonyl fluoride, 1x proteinase inhibitor mixture (EDTA-free; Roche, Mannheim, Germany), and 1x phosphatase inhibitor cocktails I ϩ II (Sigma-Aldrich, Steinheim, Germany). Protein extracts were prepared by vigorously shaking the cells suspended in 0.1 ml of lysis buffer with 0.1 g of 0.4-mmdiameter glass beads (Carl Roth, Karlsruhe, Germany) in 1.5-ml plastic tubes five times for 1 min each time at 25 Hz in a Mixer Mill MM 200 (Retsch, Haan, Germany) with 1-min intervals for sample cooling on ice. The supernatants obtained via centrifugation (12,000 ϫ g, 10 min, 4°C) were subjected to SDS-PAGE in 10% (w/v) acrylamide gels, and the separated proteins were transferred onto a PVDF membrane (Millipore, Schwalbach, Germany). The membrane was probed with the affinity-purified anti-KlHxk1 polyclonal antibody and bound IgG detected with HRP-conjugated goat anti-rabbit IgG (Cell Signaling Technology, Danvers, MA) using Immobilon TM Western Chemiluminescent HRP Substrate (Millipore, Schwalbach, Germany).
Repression Assay-Strains JA6, JA6⌬rag5, and JA6⌬rag5/pTSRAG5 were pre-grown in YNB medium supplemented with 2% (w/v) galactose, shifted into YNB medium containing 2% (w/v) glucose, and cultivated for ϳ15 h. The cells were washed twice using YNB medium without a carbon source, resuspended in YNB medium containing 2% (w/v) of the indicated carbon source(s), and propagated from a starting OD (600 nm/1 cm) of 0.3 until an OD (600 nm/1 cm) of 1.0 was reached. The ␤-galactosidase activity of protein extracts was determined according to Ref. 22 using o-nitrophenyl-␤-D-galactopyranoside as a substrate. One milliunit of ␤-galactosidase activity catalyzes the formation of 1 nmol o-nitrophenolate per minute at 30°C.
Nephelometric Growth Analysis-The utilization of different carbon sources by strains JA6, JA6⌬rag5, and JA6⌬rag5/pTSRAG5 was monitored in liquid media at 30°C in a NEPHELOstar Galaxy laserbased microplate nephelometer (BMG LABTECH, Offenburg, Germany) equipped with 96-well Cellstar ® suspension culture plates (Greiner BioOne, Solingen, Germany) based on the detection of particulate matter via forward light scattering. Cells were grown to stationary phase in 5 ml of YNB medium containing 2% (w/v) galactose at 200 rpm and 30°C and were subsequently washed twice with and resuspended in YNB medium without a carbon source to adjust to an OD (600 nm/1 cm) of 2. This suspension was used to inoculate YNB medium containing 2% (w/v) glucose, 2% (w/v) fructose, 2% (w/v) galactose, or 2% glycerol in the wells of the Cellstar ® plates. Typically, 5 l of cell suspension and 200 l of medium were applied per well. Medium lacking any carbon source was used as a control. Evaporation and condensation were avoided by sealing the plates with gas-permeable plastic film (Breathe-Easy, Carl Roth, Karlsruhe, Germany). Growth data are based on three independent experiments, each of which consisted of assays performed in triplicate.

KlHxk1 Deficiency of Mutant Strain JA6⌬rag5 of K. lactis-
The absence of KlHxk1 protein and catalytic activity in strain JA6⌬rag5 was examined via different experimental approaches, all verifying the null mutation. Firstly, severely impaired growth of the mutant on glucose (Fig. 1A) and almost no growth on fructose (Fig. 1B) together with normal growth on galactose (Fig. 1C) and glycerol (Fig. 1D) indicated the absence of a functional hexokinase. The slow growth on glucose (Fig. 1A) is consistent with the expression of a recently identified glucokinase (KlGlk1) of unknown physiological function in strain JA6⌬rag5 (12). Secondly, the rate of phosphorylation of glucose and fructose determined according to Ref. 12 with extracts of glucose-and galactose-grown JA6⌬rag5 was less than 3% compared with JA6 control extracts (data not shown). Thirdly, Western blot analysis using polyclonal anti-KlHxk1 primary antibodies ( Fig. 2A)  Fourthly, glucose repression of ␤-galactosidase as observed in strain JA6 (Fig. 2B, dataset 1) was essentially relieved in strain JA6⌬rag5 during growth on glucose and abolished during growth on glucose ϩ galactose (Fig. 2B, dataset 2) and was fully reestablished upon transformation of the mutant with multicopy plasmid pTSRAG5 (Fig. 2B, dataset 3). Finally, the absence of KlHxk1 in the proteome of strain JA6⌬rag5 was verified by determining the AAR of the KlHxk1 protein ( Table I). The latter analysis had to take into account the existence in wild-type cells of two molecular species of KlHxk1 differing in the state of phosphorylation at serine-15 (29) and consequently exhibiting different isoelectric points and positions in the two-dimensional gel (Figs. 3 and 4). DIGE experiments employing spike-ins of purified differentially labeled unphosphorylated and serine-15-phosphorylated KlHxk1 (Fig. 3) identified spot 968 as the phosphoserine-15 enzyme. Mass spectrometric quantification applying a single reaction monitoring method specific for the 13 KGS(p)MADV-PANIMEQIHGIETIFTVSSEK 40 tryptic phosphopeptide verified the latter result by indicating a 20-fold higher signal intensity for spot protein 968 relative to spot protein 976 (supplemental Fig. S1 and supplemental Table S1). With these findings in mind, the AAR values in Table I essentially confirm the anticipated absence of KlHxk1 in the mutant with respect to spot protein 968, whereas the data calculated for spot protein 976 require special consideration (see "Discussion"). In accordance with above data, strain JA6⌬rag5 was considered appropriate for analyzing the proteomic consequences of KlHxk1 deficiency in glucose-repressible K. lactis.
Comparative DIGE Analysis of JA6 and JA6⌬rag5 Proteomes-The DIGE method allowed for the resolution of 2,851 fluorescent spots in the digital image of the master gel containing all of the K. lactis proteins present in the wild-type and mutant proteomes during growth in high-glucose medium (Fig. 4). In comparison, the total number of predicted proteincoding genes in the K. lactis genome is 5,108 (44). Due to Cells were grown in YNB medium supplemented with the indicated carbon sources. Growth was monitored at 30°C in a NEPHELOstar Galaxy laser-based microplate nephelometer. Cell density is given in relative nephelometric units (RNU). Data represent three independent experiments, each of which was performed in triplicate. For clarity, only one out of four data points is plotted. differences in relative protein abundance and variation in the brightness and brilliance for image generation, the master gel image (Fig. 4) selected from 30 DIGE images scanned at the Cy2-specific wavelength displayed a number of fluorescent spots that apparently deviated from the former spot number. Spot analysis and mass spectrometric peptide determination identified 59 fluorescent spots containing 45 individual proteins that occurred in 60 different molecular forms in the mutant proteome at a minimum 3-fold enhanced or reduced level (false discovery rate of 5%) as a consequence of KlHxk1 deficiency. The identity and function(s) of these hexokinasedependent proteins as determined by in-gel tryptic digestion, mass spectrometric peptide analysis, and database searches are summarized in supplemental Table S2, where proteins are   ordered according to their AAR value, and Tables II-IV contain  protein groups organized according to assigned functions. Functional Groups of Hexokinase-dependent Proteins of K. lactis-Hexokinase-dependent proteins that are involved in or related to glycolysis and/or gluconeogenesis, hexose metabolism, and glucosaccharide turnover are listed in Table II, and Fig. 5 gives an overview of reactions of central metabolic pathways that are catalyzed by them. Except for inositol 3-phosphate synthase (spot 927), external NADH-ubichinone oxidoreductase 1 (spot 937), and two molecular species of pyruvate decarboxylase (spots 828 and 858), these proteins were detected at significantly increased levels in the mutant proteome, suggesting their RAG5/KlHxk1-dependent repression in wild-type cells during growth on glucose. The detection of three molecular species of glyceraldehyde-3-phosphate dehydrogenase (spots 1717, 1733, and 1736) exhibiting experimentally determined M r and pI values that only slightly deviated from those of the authentic enzyme and of two additional molecular species exhibiting clearly deviating DIGE coordinates (spots 2502 and 2533) indicates covalent modifications of unknown nature and significance. Spot protein 470, identified as a ␤-glucosidase precursor that in its active state is involved in glucosaccharide hydrolysis, represents another high-abundance hexokinase-dependent protein that in wild-type cells apparently is subject to RAG5/KlHxk1-dependent glucose repression. Interestingly, three enzymes of the lactose-galactose regulon (45) were identified (␤-galactosidase, galactokinase, and galactose-1-phosphate uridylyltransferase) at increased concentrations with ␤-galactosidase (spots 226 -228) and galactokinase (spots 871 and 875) occurring in different molecular forms. In the case of ␤-galactosidase (spot 228), the extraordinarily high AAR value reflects drastic changes at the level of transcription, translation, and/or proteolysis without excluding other types of covalent modification. The detection of increased levels of the latter three enzymes is likely to reflect the adaptation of K. lactis metabolism to the utilization of alternative carbon sources when a drastically reduced glucose phosphorylation capacity is limiting the utilization of this sugar. Table III summarizes the proteomic response of strain JA6 to KlHxk1 deficiency with regard to ethanol, acetate, propionate, and tricarboxylic acid metabolism, and Fig. 5 gives an overview of affected reactions of central metabolic pathways. Again, the majority of listed proteins were detected at significantly increased levels in the mutant proteome, with acetyl-CoA synthetase 1 displaying the most pronounced abundance change (spot 589). Particularly striking was the assignment of three hexokinase-dependent proteins to the mitochondrial 2-methylcitrate cycle, which allows the conversion of propionate originating in the course of the degradation of odd-chain fatty acids and certain amino acids into pyruvate, a common precursor of biosynthetic and catabolic pathways (46). These proteins are homologs of S. cerevisiae mitochondrial citrate synthase Cit3 (spot 1213) with dual-  (2), and JA6⌬rag5/pTSRAG5 (3). Cells were grown in YNB medium supplemented with 2% (w/v) galactose, shifted into YNB medium containing 2% (w/v) glucose, washed with YNB medium without a carbon source, transferred into YNB medium containing 2% (w/v) of the indicated carbon source(s), and propagated until an OD (600 nm/ 1 cm) of 1 was reached. The activity of ␤-galactosidase in the protein extracts was determined using o-nitrophenyl-␤-D-galactopyranoside as a substrate. One milliunit (mU) of ␤-galactosidase activity catalyzes the formation of one nanomole of o-nitrophenolate per minute at 30°C. substrate specificity for citrate and 2-methylcitrate (47), mitochondrial 2-methylisocitrate lyase (spot 837), and 2-methylcitrate dehydratase (spot 965). Their detection at increased proteomic levels not only verifies the existence of the 2-methylcitrate cycle in K. lactis but also suggests its repression by glucose in a RAG5/KlHxk1-dependent way. The listing of mitochondrial aldehyde dehydrogenase and cytosolic alcohol dehydrogenases 1 and 2 (spots 992, 1411, and 1414) in Table III takes into account their potential roles in acetate and ethanol formation, respectively, without excluding contributions to gluconeogenesis (see Table II) via the glyoxylate cycle and/or to redox state maintenance and stress response (see Table IV). In the case of alcohol dehydrogenase isoenzyme 2, the inverse abundance changes determined for spot proteins 1411 and 1414 might reflect post-translational regulation of this enzyme.
Hexokinase-dependent proteins exhibiting actual or predicted functional relations to chromatin remodeling, amino acid and protein metabolism, redox state maintenance, and stress response are listed in Table IV. The only enzyme in Table IV with assigned functions in central metabolic path-ways (spot 1828) is also included in Fig. 5. Remarkably, the highest AAR determined in the present study (AAR 37.5) corresponds to the KLLA0C16225p protein (spot 2392), which is weakly similar to the S. cerevisiae Gre1 hydrophilin of unknown function (48). The findings that the ScGRE1 gene is induced under hypoxic conditions (49) and that the response to hypoxia in K. lactis is linked to glucose metabolism (50) might reflect a functional relation between the KLLA0C16225p protein and KlHxk1. In the case of the protein KLLA0D07414p (spot 615), the assignment of a molecular function is similarly unsatisfactory. This protein exhibits weak similarity to the Fmo1 thiol-specific monooxygenase of S. cerevisiae, which probably is required for correct folding of disulfide-bonded proteins; however, homologs of greater similarity are found in other yeasts. The identification of hexokinase-dependent proteins according to functions of their yeast homologs in translation, proteolysis, and amino acid metabolism (e.g. spot proteins 933, 1905, and 1100) suggests stimulation of protein turnover presumably as a result of enhanced oxidative protein damage and limited glycolytic energy supply in the mutant. The latter observation is complemented by the   Table S1). For further experimental details, see "Experimental Procedures." abundance change of the K. lactis KLLA0F24838p homolog of the ISWI chromatin-remodeling complex ATPase Isw2 of S. cerevisiae (spots 471 and 568), possibly indicating the association of nuclear events taking place in the mutant at the transcriptional level with the detected proteomic alterations. In comparison, the relative abundance of two K. lactis proteins exhibiting homology to the S. cerevisiae phenylpyruvate decarboxylase Aro10 (spots 743 and 749) was decreased in the mutant despite this enzyme's role in amino acid degradation.
The actual or predicted molecular functions and biological processes assigned to the 45 identified hexokinase-dependent proteins of K. lactis glucose-repressible strain JA6 are classified in Fig. 6. In line with the metabolic limitations resulting from the KlHxk1 deficiency, carbohydrate metabolism represents the most frequently met functional category, followed by amino acid metabolism. According to expectations, genetic information processing represents another large functional category. By contrast, the effect of KlHxk1 deficiency on energy metabolism appears to be significantly less pronounced, suggesting the existence of compensatory metabolic pathways in the mutant. DISCUSSION The present exploratory study addressed the significance of the single hexokinase KlHxk1 in general metabolism and, in particular, glucose sensing and signaling in the Crabtreenegative yeast K. lactis favoring respiration over fermentation (9). In comparison to the extensively studied respiro-fermentative model organism S. cerevisiae, glucose-dependent signal transduction is less understood and/or different in K. lactis. Glucose repression of invertase encoded by the KlINV1 gene, for example, does not require the transcriptional repressor KlMig1 in K. lactis (51), whereas ScMig1 is indispensable for invertase (SUC2) repression in S. cerevisiae. Genetic studies of hexokinase functions in K. lactis using rag5 mutants indicated the involvement of RAG5 and its gene product KlHxk1 in the transcriptional regulation of low-and high-affinity glucose transport (11,20,24) and in glucose repression of several enzymes (16,25), but also in glucose-induced transcription of the KlPDC gene encoding pyruvate decarboxylase (52). As cell morphology and functions are primarily determined by the proteome, the absence of complementary information on the proteome level initiated the systematic proteomic comparison of hexokinase wild-type strain JA6 and hexokinase null mutant strain JA6⌬rag5 in different metabolic situations as presented in this paper.
The application of the DIGE method to the analysis of the most prominent proteomic changes occurring in K. lactis during growth on glucose as a consequence of RAG5 disruption and concomitant KlHxk1 deficiency resulted in the identifica- FIG. 4. DIGE image of the K. lactis proteome (black and white image of the master gel) and annotation of hexokinase-dependent proteins. The identification of K. lactis proteins occurring during growth on glucose (2% initial concentration) at altered concentrations in the mutant proteome as a consequence of RAG5 disruption and concomitant hexokinase deficiency employed an AAR setting of Ն 3/Յ Ϫ3 and a false discovery rate of 5% according to Ref. 37. The boundaries surround fluorescent spots containing hexokinase-dependent proteins or indicate their position in the two-dimensional gel when the absolute amount of the respective protein was low. The dot within each boundary marks the center of the protein mass, which generally represents the optimal picking location for protein identification. The proteins are annotated in accordance with the spot numbers used in supplemental Table S2, in which submitted or recommended names, molecular functions, and biological processes are additionally indicated. The positions of serine-15-phosphorylated (spot 968) and unphosphorylated KlHxk1 (spot 976) are labeled in green and red, respectively. For further experimental details, see "Experimental Procedures." tion of 60 protein species present at a minimum 3-fold altered (mostly increased) concentration in the mutant proteome (supplemental Table S2). Benjamini-Hochberg correction for 2,851 detected protein spots resulted in a critical p value of 1.7 ϫ 10 Ϫ3 , which controls the probability of false positive identifications (false discovery rate ϭ 5%), and the predefin-  ition of at least 3-fold abundance changes according to Ref.
53 guarantees a statistical power of ϳ80%. The corresponding calculation took into account observed variability data expressed in terms of coefficients of variation, which turned out to be less than 0.52 for 95% of the proteins listed in supplemental Table S2. The identified protein species correspond to 45 individual K. lactis proteins from which they are derived by means of covalent modifications that remain to be identified. The molecular functions assigned to these hexokinase-dependent proteins and their relative abundance in the mutant proteome (supplemental Table S2) strongly support the hypothesis that KlHxk1 plays a key role in transforming the extracellular glucose signal into a specific proteomic pattern mainly by limiting the amount of proteins that are dispensable when glucose is abundantly available. This important conclusion is impressively illustrated by the greatly elevated proteomic levels of ␤-galactosidase and maltase (Table II), confirming previous results showing that these enzymes are subject to KlHxk1-dependent glucose repression (16,25). The above hypothesis is consistent with the missing detection of proteins fulfilling the AAR Ն 3/Յ Ϫ3 criterion when DIGE analysis was performed following cultivation on galactose or glycerol as the sole carbon source. The latter finding likely reflects an independence of galactose and glycerol utilization of hexokinase catalytic and regulatory functions. It should be noted, however, that in total four proteins from supplemental Table S2 were identified in cells grown on galactose and glycerol that fulfilled the 5% false discovery rate criterion but exhibited less than 3-fold abundance changes (spot numbers 937 (AAR ϭ Ϫ1.31) and 1916 (AAR ϭ 1.56) for galactose and spot numbers 825 (AAR ϭ 1.73) and 862 (AAR ϭ 1.31) for glycerol). The identification of three hexokinase-dependent proteins that are homologs of 2-methylcitrate cycle enzymes of S. cerevisiae (Table III) indicates on the protein level the existence of this pathway in K. lactis and suggests its repression in a RAG5/KlHxk1-dependent way. The failed detection in the present study of elevated levels of invertase and malate dehydrogenase in the mutant does not necessarily contradict the increased catalytic activities of these enzymes found in a rag5 mutant in a condition of glucose repression (16) because enzyme activity is not stringently correlated with enzyme concentration. In comparison, the decreased level of glucoseinducible pyruvate decarboxylase (supplemental Table S2, spot 828) confirms the observation of reduced transcription of the KlPDC1 gene in a rag5 mutant (52). The additional identification of hexokinase-dependent proteins predicted to be functionally related to chromatin remodeling, amino acid and protein metabolism, redox state maintenance, and stress response (Table IV) reinforces the idea that KlHxk1 exerts functions beyond hexose phosphorylation and glucose-dependent signal transduction. Finally, the proteomic data from the present study confirm the sensitivity to RAG5/ KlHxk1-dependent glucose repression of K. lactis genes encoding proteins/enzymes that are required for the utilization of alternative carbon sources and for gluconeogenesis when glucose is limiting (16,22,25,54), whereas genes encoding proteins involved in respiration remain essentially unaffected (19). The obvious consequence of KlHxk1 deficiency is the slow growth of the mutant on glucose (Fig. 1A), which likely reflects limited utilization of the sugar initiated by the KlGlk1 glucokinase (12). This impairment might cause metabolic conditions differing from those existing in wild-type cells that could affect transcription (55) and consequently the proteome of the mutant. The lacking dependence of glucose repression in S. cerevisiae on the glucose phosphorylating capacity of the cell (56), however, alleviates this concern, at least with respect to proteins involved in RAG5/KlHxk1-dependent glucose repression if evolution has preserved basic regulatory mechanisms in the two yeasts. The latter consideration might not apply to glucose induction signaling because in K. lactis the expression of the Rag1 and Kht1 glucose transporters is influenced by the glycolytic flux (21,57) and, in the case of Rag1, RAG5/KlHxk1 (20,57). Apart from the fact that only abundance changes meeting the AAR Ն 3/Յ Ϫ3 criterion were considered in the present study, the interpretation of missing Kht1 identification has to take into account that the level of an individual protein represents the complex result of transcription, translation, proteolysis, and/or additional covalent protein modification(s), in addition to limitations associated with incomplete protein solubilization as discussed below.
The slow growth of strain JA6⌬rag5R on glucose (Fig. 1A) deserves special consideration because the correlation of instantaneous growth rate and gene expression detected in S. cerevisiae (58) might exist in K. lactis as well and interfere with the identification of hexokinase-dependent proteins. The present study did not address this phenomenon, but comparison of the AAR values in supplemental Table S2 and the growth rate response data of S. cerevisiae (58) identified two classes of hexokinase-dependent proteins of K. lactis: proteins whose relative abundance in the mutant qualitatively matches the growth-rate-dependent change of expression of their homologs in S. cerevisiae, and proteins that do not match this correlation (besides proteins that are not listed in Ref. 58). This situation, together with the finding that only one ribosomal protein was detected at an altered level in the mutant proteome (spot 1674; Table IV and supplemental Table S2), which contradicts the strong growth-rate dependence of ribosomal protein expression in S. cerevisiae (58), chal-lenges a general influence of growth rate on gene expression in K. lactis. The above considerations illustrate the complexity of gene deletion studies and their limitations, which are essentially attributable to the interdependence of enzyme activity, growth rate, and gene expression.
In order to recognize proteomic differences that might have been caused by the different precultivation conditions applied to generate a sufficient biomass of the hexokinase mutant rather than by the null mutation, the proteomes of both strains were compared after precultivation on galactose and final cultivation on glucose by post-lysis dimethyl labeling (DML) and electrospray ionization MS/MS. The DML data in supplemental Table S3 (columns K and L) indicate that the proteomic consequences of KlHxk1 deficiency are largely similar with respect to the two precultivation conditions. In detail, for 27 protein species the AAR values deviated from their respective mean by not more than 20%, and the AAR values of 8 additional protein species indicated the same abundance change tendency when 3-fold or higher abundance changes were considered as in the DIGE approach. For an additional 22 protein species, concordant abundance changes of less than 3-fold were determined. The DML data support the view that final cultivation on glucose as applied in the DIGE approach is appropriate to largely abolish proteomic differences that might have been caused by the different precultivation conditions.
Difference gel electrophoresis not only separates proteins and thereby reduces sample complexity prior to proteolytic digestion and subsequent mass spectrometry, but also allows prior-to-proteolysis abundance determinations, in particular for multiple forms of proteins that may arise through covalent modifications such as proteolysis or phosphorylation or combinations thereof. This singularity represents a key advantage FIG. 6. Assignment of hexokinase-dependent proteins of RAG5 wild-type strain JA6 of K. lactis to functional categories according to KEGG pathway and Gene Ontology (GO) biological process information. The categorized proteins were detected by DIGE at altered concentrations in the proteome of glucose-grown rag5 null mutant strain JA6⌬rag5 as a consequence of RAG5 disruption and concomitant KlHxk1 deficiency (for a full list of hexokinase-dependent proteins, see supplemental Table S2). Categorization of hexokinase-dependent proteins with no explicit KEGG or GO annotation is based on the annotation of the closest yeast homolog. Proteins assigned more than one function are accordingly considered in more than one functional category. The sum of footprints of the categorized proteins is normalized to equal 100%. of the DIGE method that led to its application in the present study. By contrast, methodological constraints complicate the evaluation of proteomic data obtained via the same method. First of all, the detection of a total of 2,851 fluorescent spots containing different protein species versus 5,108 predicted protein-coding genes in the K. lactis genome (44) certainly is related to the transcriptional state of the mutant, but it likely is additionally affected by incomplete protein extraction despite the application of harsh chemical and physical conditions. Limited protein solubilization represents a general methodological problem that may be overcome by combining total cell digestion, peptide separation, and quantitative mass spectrometry; however, even solubilized proteins undergoing high abundance changes may escape from detection and mass spectrometric identification when their absolute amount is limiting. This constraint is likely to apply to regulators of gene expression, and indeed, no transcription factor was identified as a hexokinase-dependent protein. In addition, the DIGE method is not appropriate for determining the individual contributions of two or more proteins present in a particular protein spot to the overall abundance of the spot protein, because spot analysis is based on total fluorescence intensities (see supplemental Table S2, protein spots 471, 568, 1411, and 1905). In view of the large number of unequivocally identified hexokinase-dependent proteins (Tables II-IV), however, this uncertainty does not compromise the basic conclusions of this study. Image analysis represents another challenge, as unspecific background fluorescence and protein interference may affect the determination of relative abundance changes, especially when low-abundance proteins are analyzed. The latter limitation is likely the cause of the unexpectedly low absolute AAR values of unphosphorylated KlHxk1 (spot 976; Table I) accounting for as little as ϳ0.1% of the K. lactis total protein subjected to DIGE analysis ( Figs. 3 and 4). Finally, the assignment of molecular functions and biological processes to the identified hexokinase-dependent proteins represents a serious challenge that is complicated by the fact that evidence for protein existence according to the UniProtKB database still is at the "inferred from homology" and "predicted" levels for the vast majority of corresponding K. lactis gene loci.
The DIGE approach was validated by DML and electrospray ionization MS/MS (supplemental Table S3). Relative protein abundance data determined via DML were compared with the AAR values resulting from DIGE analysis (supplemental Table  S3, column J versus column K). In detail, for 16 protein species the AAR values determined via the two proteomic methods deviated from their respective means by not more than 20%, and AAR values indicated the same abundance change tendency for an additional 22 protein species when 3-fold or higher abundance changes were considered as in the DIGE approach. For an additional 18 protein species, concordant abundance changes of less than 3-fold were determined. When evaluating these data one has to consider that DML detects peptides from all the forms of a given protein that result from co-and/or post-translational modifications, whereas DIGE analysis typically refers to single protein species that are separated prior to mass spectrometric analysis of their peptides. The above findings, together with the broad variety of molecular functions and biological processes assigned to the large number of prominent hexokinase-dependent proteins, support the view that KlHxk1 plays a dual role as a catalyst and key regulator and suggest the existence of novel hexokinase signaling networks in K. lactis that remain to be explored.